Structural Mechanics Module

User’s Guide
Structural Mechanics Module User’s Guide
© 1998–2022 COMSOL
Protected by patents listed on www.comsol.com/patents, or see Help>About COMSOL Multiphysics on the
File menu in the COMSOL Desktop for less detailed lists of U.S. Patents that may apply. Patents pending.
This Documentation and the Programs described herein are furnished under the COMSOL Software License
Agreement (www.comsol.com/sla) and may be used or copied only under the terms of the license
agreement.
COMSOL, the COMSOL logo, COMSOL Multiphysics, COMSOL Desktop, COMSOL Compiler,
COMSOL Server, and LiveLink are either registered trademarks or trademarks of COMSOL AB. All other
trademarks are the property of their respective owners, and COMSOL AB and its subsidiaries and products
are not affiliated with, endorsed by, sponsored by, or supported by those trademark owners. For a list of such
trademark owners, see www.comsol.com/trademarks.
Version: COMSOL 6.1

Contact Information
Visit the Contact COMSOL page at www.comsol.com/contact to submit general inquiries
or search for an address and phone number. You can also visit the Worldwide Sales Offices
page at www.comsol.com/contact/offices for address and contact information.

If you need to contact Support, an online request form is located on the COMSOL Access
page at www.comsol.com/support/case. Other useful links include:

• Support Center: www.comsol.com/support

• Product Download: www.comsol.com/product-download
• Product Updates: www.comsol.com/support/updates
• COMSOL Blog: www.comsol.com/blogs
• Discussion Forum: www.comsol.com/forum
• Events: www.comsol.com/events
• COMSOL Video Gallery: www.comsol.com/videos
• Support Knowledge Base: www.comsol.com/support/knowledgebase

Part number: CM021101

C o n t e n t s

Chapter 1: Introduction

About the Structural Mechanics Module 30

A Quick Overview of the Structural Mechanics Module . . . . . . . . 30
What Problems Can It Solve? . . . . . . . . . . . . . . . . . . 31
The Structural Mechanics Module Physics Interface Guide . . . . . . . 34
Common Physics Interface and Feature Settings and Nodes . . . . . . 43
Where Do I Access the Documentation and Application Libraries? . . . . 44

Overview of the User’s Guide 48

Chapter 2: Structural Mechanics Modeling

Study Types 55
Introduction . . . . . . . . . . . . . . . . . . . . . . . . 55
Stationary Analysis . . . . . . . . . . . . . . . . . . . . . . 55
Eigenfrequency Analysis . . . . . . . . . . . . . . . . . . . . 57
Mode Analysis . . . . . . . . . . . . . . . . . . . . . . . . 62
Time-Domain Analysis . . . . . . . . . . . . . . . . . . . . . 63
Frequency-Domain Analysis . . . . . . . . . . . . . . . . . . . 64
Mode Superposition. . . . . . . . . . . . . . . . . . . . . . 67
Harmonic Perturbation . . . . . . . . . . . . . . . . . . . . 69
Modal Reduced-Order Models. . . . . . . . . . . . . . . . . . 72
Linearized Buckling Analysis. . . . . . . . . . . . . . . . . . . 77
Bolt Pretension Study . . . . . . . . . . . . . . . . . . . . . 78
Random Vibration (PSD) Study . . . . . . . . . . . . . . . . . 78
Response Spectrum Analysis Study . . . . . . . . . . . . . . . . 79

Selecting the Physics Interface 81

Solid Mechanics . . . . . . . . . . . . . . . . . . . . . . . 81
Elastic Waves, Time Explicit. . . . . . . . . . . . . . . . . . . 84
Shell and Plate . . . . . . . . . . . . . . . . . . . . . . . . 85

CONTENTS |3
 Membrane . . . . . . . . . . . . . . . . . . . . . . . . . 86
Beam . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
Pipe Mechanics . . . . . . . . . . . . . . . . . . . . . . . 87
Truss . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
Wire . . . . . . . . . . . . . . . . . . . . . . . . . . . 88

Selecting Discretization 89
Shape Function Order . . . . . . . . . . . . . . . . . . . . . 89
Lagrange and Serendipity Shape Functions . . . . . . . . . . . . . 90
Choosing Shape Functions in Multiphysics Models . . . . . . . . . . 90
Implicit Shape Function Orders . . . . . . . . . . . . . . . . . 91

Coupling Different Element Types 93

Introduction to the Element Types . . . . . . . . . . . . . . . . 93
Coupling Techniques . . . . . . . . . . . . . . . . . . . . . 93

Applying Loads 102

Units, Orientation, and Visualization . . . . . . . . . . . . . . 102
Load Cases . . . . . . . . . . . . . . . . . . . . . . . . 104
Singular Loads . . . . . . . . . . . . . . . . . . . . . . . 106
Moments in the Solid Mechanics Interface . . . . . . . . . . . . 108
Pressure . . . . . . . . . . . . . . . . . . . . . . . . . 108
Acceleration Loads . . . . . . . . . . . . . . . . . . . . . 108
Temperature Loads — Thermal Expansion . . . . . . . . . . . . 109
Hygroscopic Swelling . . . . . . . . . . . . . . . . . . . . 109
Total Loads . . . . . . . . . . . . . . . . . . . . . . . . 110
Loads with a Spatial Variation . . . . . . . . . . . . . . . . . 110
Variables for Loads . . . . . . . . . . . . . . . . . . . . . 111

Defining Constraints 113

Rigid Body Motion . . . . . . . . . . . . . . . . . . . . . 113
Orientation . . . . . . . . . . . . . . . . . . . . . . . . 114
Prescribed Displacements, Velocities, and Accelerations . . . . . . . 115
Symmetry Constraints . . . . . . . . . . . . . . . . . . . . 116
Elemental and Nodal Constraints. . . . . . . . . . . . . . . . 119
Suppressing Constraints on Lower Dimensions . . . . . . . . . . 120
Kinematic Constraints . . . . . . . . . . . . . . . . . . . . 122
Rotational Joints . . . . . . . . . . . . . . . . . . . . . . 122

4 | CONTENTS
Attachments . . . . . . . . . . . . . . . . . . . . . . . 122

Calculating Reaction Forces 125

Using Predefined Variables to Evaluate Reaction Forces . . . . . . . 125
Using Weak Constraints to Evaluate Reaction Forces . . . . . . . . 126
Using Surface Traction to Evaluate Reaction Forces. . . . . . . . . 127
Evaluating Surface Traction Forces on Interior Boundaries . . . . . . 128

Introduction to Material Models 129

Material Models for Structural Mechanics . . . . . . . . . . . . 129
Combination of Material Models . . . . . . . . . . . . . . . . 131
Entering Material Data . . . . . . . . . . . . . . . . . . . . 131
Introduction to Linear Elastic Materials . . . . . . . . . . . . . 132
Introduction to Viscoelastic Materials . . . . . . . . . . . . . . 132
Mixed Formulation . . . . . . . . . . . . . . . . . . . . . 133
Modeling Damage . . . . . . . . . . . . . . . . . . . . . 135
About the Material Databases for the Structural Mechanics Module . . 136
Using External Materials . . . . . . . . . . . . . . . . . . . 137
References . . . . . . . . . . . . . . . . . . . . . . . . 150

Modeling Piezoelectric Problems 151

About Piezoelectric Materials . . . . . . . . . . . . . . . . . 151
Piezoelectric Material Orientation . . . . . . . . . . . . . . . 152
Piezoelectric Losses. . . . . . . . . . . . . . . . . . . . . 160
References for Piezoelectric Damping . . . . . . . . . . . . . . 167
Piezoelectric Coupling . . . . . . . . . . . . . . . . . . . . 168
Create the Piezoelectric Effect Interface and Define Domains . . . . . 169
Complete Settings of Piezoelectric Materials . . . . . . . . . . . 170
Add Damping and Loss . . . . . . . . . . . . . . . . . . . 171
Define Material Properties . . . . . . . . . . . . . . . . . . 171
Coupling Piezoelectricity with Acoustics . . . . . . . . . . . . . 173

Modeling Electrostrictive and Ferroelectroelastic Materials 174

Electrostriction Coupling . . . . . . . . . . . . . . . . . . . 174
Ferroelectricity . . . . . . . . . . . . . . . . . . . . . . 174
Multiphysics Interfaces . . . . . . . . . . . . . . . . . . . . 174

CONTENTS |5
 Modeling Magnetostrictive Materials 176
Magnetostriction Coupling . . . . . . . . . . . . . . . . . . 176
Linear vs. Nonlinear Magnetostriction . . . . . . . . . . . . . . 176
Create the Magnetostriction Interface and Define Domains. . . . . . 177

Mechanical Damping and Losses 181

Damping Sources. . . . . . . . . . . . . . . . . . . . . . 181
About Damping . . . . . . . . . . . . . . . . . . . . . . 182
Rayleigh Damping . . . . . . . . . . . . . . . . . . . . . 184
Loss Factor Damping . . . . . . . . . . . . . . . . . . . . 186
Viscous Damping . . . . . . . . . . . . . . . . . . . . . . 187
Equivalence Between Loss Factor, Rayleigh, and Viscous Damping . . . 187
Explicit Complex-Valued Damping . . . . . . . . . . . . . . . 188
Viscoelastic Damping . . . . . . . . . . . . . . . . . . . . 188
Piezoelectric Damping . . . . . . . . . . . . . . . . . . . . 188
Adding Damping in the Modal Solver . . . . . . . . . . . . . . 189

Modeling Geometric Nonlinearity 190

Geometric Nonlinearity, Frames, and the ALE Method . . . . . . . 190
Geometric Nonlinearity for the Solid Mechanics Interface . . . . . . 198
Geometric Nonlinearity for the Shell, Plate, Membrane, Beam and
Truss Interfaces . . . . . . . . . . . . . . . . . . . . . 199
Solving Geometrically Nonlinear Problems . . . . . . . . . . . . 199
Prestressed Structures. . . . . . . . . . . . . . . . . . . . 200
Geometric Nonlinearity for the Piezoelectric Material. . . . . . . . 203
References . . . . . . . . . . . . . . . . . . . . . . . . 205

Contact Modeling 206

Overview. . . . . . . . . . . . . . . . . . . . . . . . . 206
Setting Up a Contact Problem . . . . . . . . . . . . . . . . . 207
Contact Pairs . . . . . . . . . . . . . . . . . . . . . . . 216
Meshing for Contact Analysis . . . . . . . . . . . . . . . . . 218
Settings for Contact Nodes . . . . . . . . . . . . . . . . . . 219
Quasistatic Contact Analysis . . . . . . . . . . . . . . . . . 229
Dynamic Contact Analysis . . . . . . . . . . . . . . . . . . 230
Multiphysics Contact Analysis . . . . . . . . . . . . . . . . . 231
Special Types of Contact Problems . . . . . . . . . . . . . . . 233
Fallback Conditions to Contact Regions . . . . . . . . . . . . . 234

6 | CONTENTS
Solver Settings for Contact Analysis. . . . . . . . . . . . . . . 235
Monitoring the Solution . . . . . . . . . . . . . . . . . . . 239
Dependent Variables in Contact Analysis. . . . . . . . . . . . . 240
Important Contact Variables . . . . . . . . . . . . . . . . . 243
References for Contact Modeling . . . . . . . . . . . . . . . 245

Activating and Deactivating Material 246

Springs and Dampers 249

Defining Multiphysics Models 251

Thermal-Structure Interaction . . . . . . . . . . . . . . . . . 251
Acoustic-Structure Interaction. . . . . . . . . . . . . . . . . 252
Thermal-Electric-Structural Interaction . . . . . . . . . . . . . 252

Thermally Coupled Problems 254

Temperature-Dependent Material Data . . . . . . . . . . . . . 254
Thermal Expansion . . . . . . . . . . . . . . . . . . . . . 256
Constraints and Thermal Expansion. . . . . . . . . . . . . . . 262
Thermoelastic Damping . . . . . . . . . . . . . . . . . . . 263

Fluid-Structure Interaction 265

Deforming Fluid Domains . . . . . . . . . . . . . . . . . . 265
Union or Assembly . . . . . . . . . . . . . . . . . . . . . 266
Contact and FSI . . . . . . . . . . . . . . . . . . . . . . 267
One-Way FSI . . . . . . . . . . . . . . . . . . . . . . . 268

Component Mode Synthesis 270

Introduction . . . . . . . . . . . . . . . . . . . . . . . 270
COMSOL Implementation . . . . . . . . . . . . . . . . . . 270
Applications. . . . . . . . . . . . . . . . . . . . . . . . 271
Working With CMS Models . . . . . . . . . . . . . . . . . 272

Computing Mass Properties 284

Effective Properties of Periodic Structures 287

RVE Modeling Using the Cell Periodicity Node . . . . . . . . . . 287

CONTENTS |7
 Modeling Pretensioned Bolts 290

Simplified Modeling of Bolt Threads 299

Modeling Embedded Structures and Reinforcements 301

Modeling Thin Layers 306

The Thin Layer Node . . . . . . . . . . . . . . . . . . . . 306
Interior and Exterior Boundaries . . . . . . . . . . . . . . . . 308
Material Models . . . . . . . . . . . . . . . . . . . . . . 309
Loads and Boundary Conditions . . . . . . . . . . . . . . . . 309
Result Evaluation and Visualization . . . . . . . . . . . . . . . 310

Modeling Cracks 312

The Crack Node . . . . . . . . . . . . . . . . . . . . . . 312
J-integrals. . . . . . . . . . . . . . . . . . . . . . . . . 313
Stress Intensity Factors . . . . . . . . . . . . . . . . . . . 316
Contact Between Crack Faces . . . . . . . . . . . . . . . . . 317

Buckling Analysis 319

Linearized Buckling Analysis. . . . . . . . . . . . . . . . . . 319
Nonlinear Buckling Analysis . . . . . . . . . . . . . . . . . . 323
Modeling Imperfection in a Buckling Analysis . . . . . . . . . . . 323

Performing a Random Vibration Analysis 328

Performing a Response Spectrum Analysis 336

Stress Linearization 344

Solver Settings for Structural Mechanics 350

Symmetric Matrices . . . . . . . . . . . . . . . . . . . . . 350
Selecting Iterative Solvers . . . . . . . . . . . . . . . . . . 351
Solver Settings for Viscoelasticity, Creep, and Viscoplasticity . . . . . 352
Specifying Tolerances and Scaling for the Dependent Variables . . . . 353
Solver Settings . . . . . . . . . . . . . . . . . . . . . . . 353

8 | CONTENTS
 Using Reduced Integration 355

Result Presentation 359

Local Orientations . . . . . . . . . . . . . . . . . . . . . 359
How an Expression Is Evaluated . . . . . . . . . . . . . . . . 362
Quality Settings . . . . . . . . . . . . . . . . . . . . . . 364
Gauss Point Evaluation. . . . . . . . . . . . . . . . . . . . 365
Postprocessing of Eigenmodes . . . . . . . . . . . . . . . . . 366
Plotting Applied Loads . . . . . . . . . . . . . . . . . . . . 368

Part Libraries 373

Chapter 3: Structural Mechanics Theory

Solid Mechanics Theory 379

Introduction . . . . . . . . . . . . . . . . . . . . . . . 379

Frames and Coordinate Systems 381

Material and Spatial Coordinates . . . . . . . . . . . . . . . . 381
Coordinate Systems. . . . . . . . . . . . . . . . . . . . . 382
Boundary System . . . . . . . . . . . . . . . . . . . . . . 383

Analysis of Deformation 384

Lagrangian Formulation . . . . . . . . . . . . . . . . . . . 384
Deformation Measures . . . . . . . . . . . . . . . . . . . 387
Invariants of Strain . . . . . . . . . . . . . . . . . . . . . 391
Inelastic Strain Contributions . . . . . . . . . . . . . . . . . 394
Deformed Geometry . . . . . . . . . . . . . . . . . . . . 398
Axial Symmetry . . . . . . . . . . . . . . . . . . . . . . 400
Plane Strain . . . . . . . . . . . . . . . . . . . . . . . . 403
Generalized Plane Strain . . . . . . . . . . . . . . . . . . . 404

Stresses 409
Defining Stress. . . . . . . . . . . . . . . . . . . . . . . 409
Invariants of the Stress Tensor . . . . . . . . . . . . . . . . 410
Plane Stress . . . . . . . . . . . . . . . . . . . . . . . . 413

CONTENTS |9
 Initial Stresses and Strains . . . . . . . . . . . . . . . . . . 414
External Stress. . . . . . . . . . . . . . . . . . . . . . . 415
External Strain . . . . . . . . . . . . . . . . . . . . . . . 417
Axial Symmetry and Stresses . . . . . . . . . . . . . . . . . 418

Equation of Motion 419

Material Models 422

Linear Elastic Material . . . . . . . . . . . . . . . . . . . . 425
Mixed Formulation . . . . . . . . . . . . . . . . . . . . . 437
Nonlinear Elastic Materials . . . . . . . . . . . . . . . . . . 439
Linear Viscoelasticity . . . . . . . . . . . . . . . . . . . . 451
Large Strain Viscoelasticity . . . . . . . . . . . . . . . . . . 480
Hyperelastic Materials . . . . . . . . . . . . . . . . . . . . 483
Distributed Fiber Models . . . . . . . . . . . . . . . . . . . 510
Elastoplastic Materials . . . . . . . . . . . . . . . . . . . . 514
Failure Criteria for Concrete, Rocks, and Other Brittle Materials . . . 552
Elastoplastic Soil Models . . . . . . . . . . . . . . . . . . . 558
Creep and Viscoplasticity. . . . . . . . . . . . . . . . . . . 574
Inelastic Strain Rate . . . . . . . . . . . . . . . . . . . . . 593
Shape Memory Alloy . . . . . . . . . . . . . . . . . . . . 595
Piezoelectricity . . . . . . . . . . . . . . . . . . . . . . 600
Magnetomechanics . . . . . . . . . . . . . . . . . . . . . 610
Magnetostriction and Piezomagnetism . . . . . . . . . . . . . . 614
Electrostriction . . . . . . . . . . . . . . . . . . . . . . 627
Ferroelectroelasticity . . . . . . . . . . . . . . . . . . . . 628
Rigid Material . . . . . . . . . . . . . . . . . . . . . . . 632
Damage Models . . . . . . . . . . . . . . . . . . . . . . 642
Safety Factor Evaluation . . . . . . . . . . . . . . . . . . . 652

Formulation of the Equilibrium Equations 673

Equation Implementation . . . . . . . . . . . . . . . . . . . 673

Study Types 675

Frequency-Domain Studies . . . . . . . . . . . . . . . . . . 675
Eigenfrequency Studies . . . . . . . . . . . . . . . . . . . 679
Random Vibration Theory . . . . . . . . . . . . . . . . . . 682
Response Spectrum Analysis Theory . . . . . . . . . . . . . . 689

10 | C O N T E N T S
Linear Buckling . . . . . . . . . . . . . . . . . . . . . . 707

Damping 709
Rayleigh Damping . . . . . . . . . . . . . . . . . . . . . 709
Loss Factor Damping . . . . . . . . . . . . . . . . . . . . 710
Viscous Damping . . . . . . . . . . . . . . . . . . . . . . 711
Maximum Loss Factor . . . . . . . . . . . . . . . . . . . . 712
Wave Attenuation . . . . . . . . . . . . . . . . . . . . . 713

Loads and Constraints 717

Distributed Loads . . . . . . . . . . . . . . . . . . . . . 717
Gravity and Base Excitation . . . . . . . . . . . . . . . . . . 718
Frame Acceleration Forces . . . . . . . . . . . . . . . . . . 719
Spring Foundation and Thin Elastic Layer . . . . . . . . . . . . . 721
Spring-Damper Theory . . . . . . . . . . . . . . . . . . . 725
Added Mass . . . . . . . . . . . . . . . . . . . . . . . . 730
Rigid Motion Suppression . . . . . . . . . . . . . . . . . . 731
Limited Displacement . . . . . . . . . . . . . . . . . . . . 732
Low-Reflecting Boundary Condition . . . . . . . . . . . . . . 733
Cyclic Symmetry and Floquet Periodic Conditions . . . . . . . . . 734
Symmetry Condition with Translation . . . . . . . . . . . . . . 735
Roller Boundary Condition . . . . . . . . . . . . . . . . . . 736
Thermal Expansion of Constraints . . . . . . . . . . . . . . . 738
Fluid-Structure Interaction . . . . . . . . . . . . . . . . . . 740
Periodic Cell Theory . . . . . . . . . . . . . . . . . . . . 741
Rigid Connector . . . . . . . . . . . . . . . . . . . . . . 744
Continuity Condition . . . . . . . . . . . . . . . . . . . . 747

Contact Analysis Theory 751

Contact Search and Kinematics . . . . . . . . . . . . . . . . 751
Definition of Contact Tractions . . . . . . . . . . . . . . . . 755
Normal Contact . . . . . . . . . . . . . . . . . . . . . . 755
Tangential Contact with Friction . . . . . . . . . . . . . . . . 762
Unbiased Formulation and Self-Contact . . . . . . . . . . . . . 767
Adhesion . . . . . . . . . . . . . . . . . . . . . . . . . 768
Decohesion . . . . . . . . . . . . . . . . . . . . . . . . 770
Wear . . . . . . . . . . . . . . . . . . . . . . . . . . 781

CONTENTS | 11
 Bolt Modeling Theory 783
Tightening Torque . . . . . . . . . . . . . . . . . . . . . 783

Stress Linearization 786

J-Integral Theory 789

Embedded Elements 793

Theory for Thin Layers 797

Deformation Gradient . . . . . . . . . . . . . . . . . . . . 797
Solid Approximation . . . . . . . . . . . . . . . . . . . . 798
Membrane Approximation . . . . . . . . . . . . . . . . . . 799
Spring Approximation . . . . . . . . . . . . . . . . . . . . 800
Material Models in Thin Layers . . . . . . . . . . . . . . . . 800
Spring Material. . . . . . . . . . . . . . . . . . . . . . . 801

Reduced Integration and Hourglass Stabilization 804

Average Rotation 806

Energy Quantities 808

Elastic Energy . . . . . . . . . . . . . . . . . . . . . . . 808
Kinetic Energy . . . . . . . . . . . . . . . . . . . . . . . 810
Dissipated Energy . . . . . . . . . . . . . . . . . . . . . 810
Mechanical Energy Flux . . . . . . . . . . . . . . . . . . . 811
Energy Variables . . . . . . . . . . . . . . . . . . . . . . 812

References 815

Chapter 4: Solid Mechanics

The Solid Mechanics Interface 828

Domain, Boundary, Edge, Point, and Pair Nodes for Solid Mechanics . . 838
Initial Values . . . . . . . . . . . . . . . . . . . . . . . 843
Change Thickness . . . . . . . . . . . . . . . . . . . . . 844

12 | C O N T E N T S
Change Cross Section . . . . . . . . . . . . . . . . . . . . 844
Linear Elastic Material . . . . . . . . . . . . . . . . . . . . 845
Nonlinear Elastic Material . . . . . . . . . . . . . . . . . . 853
Elastoplastic Soil Material. . . . . . . . . . . . . . . . . . . 859
Hyperelastic Material . . . . . . . . . . . . . . . . . . . . 866
Shape Memory Alloy . . . . . . . . . . . . . . . . . . . . 878
Phase Transformation Direction . . . . . . . . . . . . . . . . 883
Piezoelectric Material . . . . . . . . . . . . . . . . . . . . 883
Piezomagnetic Material . . . . . . . . . . . . . . . . . . . 887
Magnetostrictive Material. . . . . . . . . . . . . . . . . . . 891
Viscoelasticity . . . . . . . . . . . . . . . . . . . . . . . 891
Mullins Effect . . . . . . . . . . . . . . . . . . . . . . . 898
Plasticity . . . . . . . . . . . . . . . . . . . . . . . . . 899
Set Variables . . . . . . . . . . . . . . . . . . . . . . . 907
Creep . . . . . . . . . . . . . . . . . . . . . . . . . . 909
Additional Creep . . . . . . . . . . . . . . . . . . . . . . 914
Viscoplasticity . . . . . . . . . . . . . . . . . . . . . . . 915
Porous Plasticity . . . . . . . . . . . . . . . . . . . . . . 923
Soil Plasticity . . . . . . . . . . . . . . . . . . . . . . . 929
Concrete . . . . . . . . . . . . . . . . . . . . . . . . . 933
Rocks . . . . . . . . . . . . . . . . . . . . . . . . . . 935
Fiber . . . . . . . . . . . . . . . . . . . . . . . . . . 937
Thermal Expansion (for Fibers) . . . . . . . . . . . . . . . . 939
Thermal Expansion (for Materials) . . . . . . . . . . . . . . . 940
Hygroscopic Swelling . . . . . . . . . . . . . . . . . . . . 943
Intercalation Strain . . . . . . . . . . . . . . . . . . . . . 945
Initial Stress and Strain. . . . . . . . . . . . . . . . . . . . 947
External Stress. . . . . . . . . . . . . . . . . . . . . . . 948
External Strain . . . . . . . . . . . . . . . . . . . . . . . 952
Inelastic Strain Rate . . . . . . . . . . . . . . . . . . . . . 955
Damage . . . . . . . . . . . . . . . . . . . . . . . . . 958
Activation . . . . . . . . . . . . . . . . . . . . . . . . 963
Safety . . . . . . . . . . . . . . . . . . . . . . . . . . 964
Damping . . . . . . . . . . . . . . . . . . . . . . . . . 970
Mechanical Damping . . . . . . . . . . . . . . . . . . . . 976
Coupling Loss . . . . . . . . . . . . . . . . . . . . . . . 978
Dielectric Loss. . . . . . . . . . . . . . . . . . . . . . . 979
Conduction Loss (Time-Harmonic) . . . . . . . . . . . . . . . 981

CONTENTS | 13
 External Stress-Strain Relation. . . . . . . . . . . . . . . . . 982
Rigid Material . . . . . . . . . . . . . . . . . . . . . . . 984
Initial Values (Rigid Material) . . . . . . . . . . . . . . . . . 988
Fixed Constraint (Rigid Material) . . . . . . . . . . . . . . . . 989
Prescribed Displacement/Rotation . . . . . . . . . . . . . . . 990
Applied Force (Rigid Material) . . . . . . . . . . . . . . . . . 993
Location Nodes . . . . . . . . . . . . . . . . . . . . . . 995
Applied Moment (Rigid Material) . . . . . . . . . . . . . . . . 995
Mass and Moment of Inertia (Rigid Material) . . . . . . . . . . . 997
Center of Mass Nodes. . . . . . . . . . . . . . . . . . . . 999
Spring Foundation (Rigid Material) . . . . . . . . . . . . . . . 999
Free. . . . . . . . . . . . . . . . . . . . . . . . . . . 1002
Prescribed Displacement . . . . . . . . . . . . . . . . . . . 1003
Prescribed Velocity . . . . . . . . . . . . . . . . . . . . . 1006
Prescribed Acceleration . . . . . . . . . . . . . . . . . . . 1009
Fixed Constraint . . . . . . . . . . . . . . . . . . . . . . 1011
Thermal Expansion (for Constraints) . . . . . . . . . . . . . . 1013
Roller . . . . . . . . . . . . . . . . . . . . . . . . . . 1016
Symmetry . . . . . . . . . . . . . . . . . . . . . . . . 1018
Symmetry Plane . . . . . . . . . . . . . . . . . . . . . . 1021
Antisymmetry . . . . . . . . . . . . . . . . . . . . . . . 1023
Rigid Motion Suppression . . . . . . . . . . . . . . . . . . 1024
Body Load . . . . . . . . . . . . . . . . . . . . . . . . 1026
Gravity . . . . . . . . . . . . . . . . . . . . . . . . . 1028
Base Excitation . . . . . . . . . . . . . . . . . . . . . . 1030
Rotating Frame . . . . . . . . . . . . . . . . . . . . . . 1031
Linearly Accelerated Frame . . . . . . . . . . . . . . . . . . 1034
Boundary Load . . . . . . . . . . . . . . . . . . . . . . 1036
Edge Load . . . . . . . . . . . . . . . . . . . . . . . . 1040
Point Load . . . . . . . . . . . . . . . . . . . . . . . . 1041
Point Load, Free . . . . . . . . . . . . . . . . . . . . . . 1045
Ring Load . . . . . . . . . . . . . . . . . . . . . . . . 1047
Ring Load, Free . . . . . . . . . . . . . . . . . . . . . . 1048
Point Load (on Axis) . . . . . . . . . . . . . . . . . . . . 1050
Thin Layer . . . . . . . . . . . . . . . . . . . . . . . . 1052
Prescribed Displacement (Thin Layer) . . . . . . . . . . . . . . 1053
Fixed Constraint (Thin Layer) . . . . . . . . . . . . . . . . . 1055
Roller (Thin Layer) . . . . . . . . . . . . . . . . . . . . . 1056

14 | C O N T E N T S
Face Load (Thin Layer) . . . . . . . . . . . . . . . . . . . 1057
Boundary Load (Thin Layer) . . . . . . . . . . . . . . . . . 1059
Spring Material. . . . . . . . . . . . . . . . . . . . . . . 1061
Spring Foundation . . . . . . . . . . . . . . . . . . . . . 1062
Thin Elastic Layer. . . . . . . . . . . . . . . . . . . . . . 1068
Predeformation . . . . . . . . . . . . . . . . . . . . . . 1072
Spring-Damper . . . . . . . . . . . . . . . . . . . . . . 1073
Source Point (for Spring-Damper) . . . . . . . . . . . . . . . 1079
Destination Point (for Spring-Damper) . . . . . . . . . . . . . 1079
Source Point Nodes. . . . . . . . . . . . . . . . . . . . . 1079
Destination Point Nodes . . . . . . . . . . . . . . . . . . . 1080
Added Mass . . . . . . . . . . . . . . . . . . . . . . . . 1080
Periodic Condition . . . . . . . . . . . . . . . . . . . . . 1083
Adiabatic Heating. . . . . . . . . . . . . . . . . . . . . . 1086
Cell Periodicity . . . . . . . . . . . . . . . . . . . . . . 1088
Boundary Pair . . . . . . . . . . . . . . . . . . . . . . . 1092
Low-Reflecting Boundary . . . . . . . . . . . . . . . . . . . 1093
Rigid Connector . . . . . . . . . . . . . . . . . . . . . . 1093
Center of Rotation Nodes . . . . . . . . . . . . . . . . . . 1104
Thermal Expansion (Rigid Connector) . . . . . . . . . . . . . . 1104
Applied Force (Rigid Connector) . . . . . . . . . . . . . . . . 1106
Applied Moment (Rigid Connector) . . . . . . . . . . . . . . . 1108
Mass and Moment of Inertia (Rigid Connector) . . . . . . . . . . 1110
Spring Foundation (Rigid Connector) . . . . . . . . . . . . . . 1111
Attachment . . . . . . . . . . . . . . . . . . . . . . . . 1113
Thermal Expansion (Attachment). . . . . . . . . . . . . . . . 1115
Reduced Flexible Components . . . . . . . . . . . . . . . . 1117
Component Definition. . . . . . . . . . . . . . . . . . . . 1119
Fixed Joint . . . . . . . . . . . . . . . . . . . . . . . . 1120
Source Filter . . . . . . . . . . . . . . . . . . . . . . . 1121
Destination Filter. . . . . . . . . . . . . . . . . . . . . . 1121
Contact . . . . . . . . . . . . . . . . . . . . . . . . . 1122
Friction . . . . . . . . . . . . . . . . . . . . . . . . . 1134
Slip Velocity. . . . . . . . . . . . . . . . . . . . . . . . 1138
Adhesion . . . . . . . . . . . . . . . . . . . . . . . . . 1141
Decohesion . . . . . . . . . . . . . . . . . . . . . . . . 1143
Wear . . . . . . . . . . . . . . . . . . . . . . . . . . 1146
Continuity . . . . . . . . . . . . . . . . . . . . . . . . 1149

CONTENTS | 15
 Bolt Pretension . . . . . . . . . . . . . . . . . . . . . . 1150
Bolt Selection . . . . . . . . . . . . . . . . . . . . . . . 1153
Bolt Thread Contact . . . . . . . . . . . . . . . . . . . . 1156
Crack . . . . . . . . . . . . . . . . . . . . . . . . . . 1158
Crack Closure . . . . . . . . . . . . . . . . . . . . . . . 1160
Face Load (Crack) . . . . . . . . . . . . . . . . . . . . . 1162
J-Integral . . . . . . . . . . . . . . . . . . . . . . . . . 1163
Reverse Crack Front . . . . . . . . . . . . . . . . . . . . 1165
Port . . . . . . . . . . . . . . . . . . . . . . . . . . . 1166
Elastic Predeformation. . . . . . . . . . . . . . . . . . . . 1170
Phase . . . . . . . . . . . . . . . . . . . . . . . . . . 1172
Harmonic Perturbation . . . . . . . . . . . . . . . . . . . 1174
Stress Linearization . . . . . . . . . . . . . . . . . . . . . 1175
Section Forces . . . . . . . . . . . . . . . . . . . . . . . 1178
Test Material . . . . . . . . . . . . . . . . . . . . . . . 1181
Local System Results . . . . . . . . . . . . . . . . . . . . 1183
Average Rotation. . . . . . . . . . . . . . . . . . . . . . 1185
Wave Speeds . . . . . . . . . . . . . . . . . . . . . . . 1187
Important Variables in the Solid Mechanics Interface . . . . . . . . 1188

Chapter 5: Shell and Plate

Modeling with Plates and Shells 1194

Introduction . . . . . . . . . . . . . . . . . . . . . . . 1194
Dependent Variables . . . . . . . . . . . . . . . . . . . . 1195
Layered and Nonlayered Shells . . . . . . . . . . . . . . . . 1197
Working with Section Stiffness . . . . . . . . . . . . . . . . 1199
Loads . . . . . . . . . . . . . . . . . . . . . . . . . . 1199

Theory for the Shell and Plate Interfaces 1201

General Theory for the Shell and Plate Interfaces . . . . . . . . . 1201
Theory for FSDT Laminated Shell . . . . . . . . . . . . . . . 1224
Theory for Hyperelastic Shells . . . . . . . . . . . . . . . . . 1239
Results Evaluation in a Laminate . . . . . . . . . . . . . . . . 1239
Theory for Section Stiffness . . . . . . . . . . . . . . . . . . 1240
References for the Shell Interface . . . . . . . . . . . . . . . 1241

16 | C O N T E N T S
The Shell and Plate Interfaces 1242
Domain, Boundary, Edge, Point, and Pair Nodes for the Shell and
Plate Interfaces . . . . . . . . . . . . . . . . . . . . . 1247
Initial Values . . . . . . . . . . . . . . . . . . . . . . . 1249
Thickness and Offset . . . . . . . . . . . . . . . . . . . . 1250
Thickness Change . . . . . . . . . . . . . . . . . . . . . 1252
Linear Elastic Material . . . . . . . . . . . . . . . . . . . . 1253
Layered Linear Elastic Material. . . . . . . . . . . . . . . . . 1255
Shell Local System . . . . . . . . . . . . . . . . . . . . . 1259
Layered Hyperelastic Material . . . . . . . . . . . . . . . . . 1260
Viscoelasticity . . . . . . . . . . . . . . . . . . . . . . . 1262
Plasticity . . . . . . . . . . . . . . . . . . . . . . . . . 1263
Set Variables . . . . . . . . . . . . . . . . . . . . . . . 1265
Creep . . . . . . . . . . . . . . . . . . . . . . . . . . 1265
Additional Creep . . . . . . . . . . . . . . . . . . . . . . 1267
Viscoplasticity . . . . . . . . . . . . . . . . . . . . . . . 1268
Thermal Expansion (for Materials) . . . . . . . . . . . . . . . 1269
Hygroscopic Swelling . . . . . . . . . . . . . . . . . . . . 1273
Initial Stress and Strain. . . . . . . . . . . . . . . . . . . . 1276
External Stress. . . . . . . . . . . . . . . . . . . . . . . 1279
External Strain . . . . . . . . . . . . . . . . . . . . . . . 1281
Inelastic Strain Rate . . . . . . . . . . . . . . . . . . . . . 1283
Damage . . . . . . . . . . . . . . . . . . . . . . . . . 1284
Mullins Effect . . . . . . . . . . . . . . . . . . . . . . . 1285
Safety . . . . . . . . . . . . . . . . . . . . . . . . . . 1286
Damping . . . . . . . . . . . . . . . . . . . . . . . . . 1287
Section Stiffness . . . . . . . . . . . . . . . . . . . . . . 1289
Fixed Constraint . . . . . . . . . . . . . . . . . . . . . . 1291
Prescribed Displacement/Rotation . . . . . . . . . . . . . . . 1293
Thermal Expansion (for Constraints) . . . . . . . . . . . . . . 1296
Prescribed Velocity . . . . . . . . . . . . . . . . . . . . . 1298
Prescribed Acceleration . . . . . . . . . . . . . . . . . . . 1300
Pinned . . . . . . . . . . . . . . . . . . . . . . . . . . 1303
No Rotation . . . . . . . . . . . . . . . . . . . . . . . 1304
Simply Supported. . . . . . . . . . . . . . . . . . . . . . 1306
Symmetry . . . . . . . . . . . . . . . . . . . . . . . . 1307
Symmetry Plane . . . . . . . . . . . . . . . . . . . . . . 1310
Antisymmetry . . . . . . . . . . . . . . . . . . . . . . . 1312

CONTENTS | 17
 Body Load . . . . . . . . . . . . . . . . . . . . . . . . 1313
Face Load . . . . . . . . . . . . . . . . . . . . . . . . 1315
Edge Load . . . . . . . . . . . . . . . . . . . . . . . . 1317
Point Load . . . . . . . . . . . . . . . . . . . . . . . . 1320
Point Load, Free . . . . . . . . . . . . . . . . . . . . . . 1322
Ring Load . . . . . . . . . . . . . . . . . . . . . . . . 1325
Ring Load, Free . . . . . . . . . . . . . . . . . . . . . . 1327
Spring Foundation . . . . . . . . . . . . . . . . . . . . . 1329
Predeformation . . . . . . . . . . . . . . . . . . . . . . 1333
Added Mass . . . . . . . . . . . . . . . . . . . . . . . . 1333
Point Mass . . . . . . . . . . . . . . . . . . . . . . . . 1335
Point Mass Damping . . . . . . . . . . . . . . . . . . . . 1336
Periodic Condition . . . . . . . . . . . . . . . . . . . . . 1337
Layered Adiabatic Heating . . . . . . . . . . . . . . . . . . 1341
Boundary to Boundary. . . . . . . . . . . . . . . . . . . . 1344
Edge to Boundary . . . . . . . . . . . . . . . . . . . . . 1345
Edge to Edge . . . . . . . . . . . . . . . . . . . . . . . 1348
Rigid Connector . . . . . . . . . . . . . . . . . . . . . . 1351
Thermal Expansion (Rigid Connector) . . . . . . . . . . . . . . 1359
Attachment . . . . . . . . . . . . . . . . . . . . . . . . 1360
Thermal Expansion (Attachment). . . . . . . . . . . . . . . . 1361
Phase . . . . . . . . . . . . . . . . . . . . . . . . . . 1362
Harmonic Perturbation . . . . . . . . . . . . . . . . . . . 1363

Chapter 6: Layered Shell

Theory for the Layered Shell Interface 1366

About Layered Shells . . . . . . . . . . . . . . . . . . . . 1366
Theory Background for the Layered Shell Interface . . . . . . . . . 1367

The Layered Shell Interface 1381

The Shell Properties and Interface Selection Sections . . . . . . . . 1385
Domain, Boundary, Edge, Point, and Pair Nodes for the Layered
Shell Interface. . . . . . . . . . . . . . . . . . . . . . 1386
Initial Values . . . . . . . . . . . . . . . . . . . . . . . 1390
Linear Elastic Material . . . . . . . . . . . . . . . . . . . . 1391

18 | C O N T E N T S
Hyperelastic Material . . . . . . . . . . . . . . . . . . . . 1393
Piezoelectric Material . . . . . . . . . . . . . . . . . . . . 1395
Viscoelasticity . . . . . . . . . . . . . . . . . . . . . . . 1399
Mullins Effect . . . . . . . . . . . . . . . . . . . . . . . 1400
Plasticity . . . . . . . . . . . . . . . . . . . . . . . . . 1401
Set Variables . . . . . . . . . . . . . . . . . . . . . . . 1403
Creep . . . . . . . . . . . . . . . . . . . . . . . . . . 1404
Additional Creep . . . . . . . . . . . . . . . . . . . . . . 1405
Viscoplasticity . . . . . . . . . . . . . . . . . . . . . . . 1407
Thermal Expansion (for Materials) . . . . . . . . . . . . . . . 1408
Hygroscopic Swelling . . . . . . . . . . . . . . . . . . . . 1409
Initial Stress and Strain. . . . . . . . . . . . . . . . . . . . 1411
External Stress. . . . . . . . . . . . . . . . . . . . . . . 1412
External Strain . . . . . . . . . . . . . . . . . . . . . . . 1414
Inelastic Strain Rate . . . . . . . . . . . . . . . . . . . . . 1415
Damage . . . . . . . . . . . . . . . . . . . . . . . . . 1416
Activation . . . . . . . . . . . . . . . . . . . . . . . . 1417
Delamination . . . . . . . . . . . . . . . . . . . . . . . 1419
Safety . . . . . . . . . . . . . . . . . . . . . . . . . . 1422
Damping . . . . . . . . . . . . . . . . . . . . . . . . . 1423
Mechanical Damping . . . . . . . . . . . . . . . . . . . . 1425
Coupling Loss . . . . . . . . . . . . . . . . . . . . . . . 1426
Dielectric Loss. . . . . . . . . . . . . . . . . . . . . . . 1427
Rigid Material . . . . . . . . . . . . . . . . . . . . . . . 1428
Free. . . . . . . . . . . . . . . . . . . . . . . . . . . 1429
Prescribed Displacement . . . . . . . . . . . . . . . . . . . 1430
Prescribed Displacement, Interface . . . . . . . . . . . . . . . 1431
Prescribed Velocity . . . . . . . . . . . . . . . . . . . . . 1432
Prescribed Velocity, Interface . . . . . . . . . . . . . . . . . 1433
Prescribed Acceleration . . . . . . . . . . . . . . . . . . . 1434
Prescribed Acceleration, Interface . . . . . . . . . . . . . . . 1436
Fixed Constraint . . . . . . . . . . . . . . . . . . . . . . 1437
Fixed Constraint, Interface . . . . . . . . . . . . . . . . . . 1438
Thermal Expansion (for Constraints) . . . . . . . . . . . . . . 1439
Roller . . . . . . . . . . . . . . . . . . . . . . . . . . 1440
Roller, Interface . . . . . . . . . . . . . . . . . . . . . . 1441
Symmetry . . . . . . . . . . . . . . . . . . . . . . . . 1442
Antisymmetry . . . . . . . . . . . . . . . . . . . . . . . 1442

CONTENTS | 19
 Rigid Motion Suppression . . . . . . . . . . . . . . . . . . 1443
Body Load . . . . . . . . . . . . . . . . . . . . . . . . 1444
Face Load . . . . . . . . . . . . . . . . . . . . . . . . 1446
Rotating Frame . . . . . . . . . . . . . . . . . . . . . . 1447
Linearly Accelerated Frame . . . . . . . . . . . . . . . . . . 1448
Boundary Load . . . . . . . . . . . . . . . . . . . . . . 1449
Edge Load . . . . . . . . . . . . . . . . . . . . . . . . 1451
Line Load. . . . . . . . . . . . . . . . . . . . . . . . . 1452
Point Load . . . . . . . . . . . . . . . . . . . . . . . . 1453
Phase . . . . . . . . . . . . . . . . . . . . . . . . . . 1454
Spring Foundation . . . . . . . . . . . . . . . . . . . . . 1455
Spring Foundation, Interface . . . . . . . . . . . . . . . . . 1456
Thin Elastic Layer. . . . . . . . . . . . . . . . . . . . . . 1458
Thin Elastic Layer, Interface . . . . . . . . . . . . . . . . . . 1459
Predeformation . . . . . . . . . . . . . . . . . . . . . . 1460
Added Mass . . . . . . . . . . . . . . . . . . . . . . . . 1461
Added Mass, Interface . . . . . . . . . . . . . . . . . . . . 1462
Adiabatic Heating. . . . . . . . . . . . . . . . . . . . . . 1463
Rigid Connector . . . . . . . . . . . . . . . . . . . . . . 1466
Rigid Connector, Interface . . . . . . . . . . . . . . . . . . 1467
Thermal Expansion (Rigid Connector) . . . . . . . . . . . . . . 1468
Attachment . . . . . . . . . . . . . . . . . . . . . . . . 1469
Thermal Expansion (Attachment). . . . . . . . . . . . . . . . 1470
Continuity . . . . . . . . . . . . . . . . . . . . . . . . 1471

Chapter 7: Membrane

Theory for the Membrane Interface 1476

About Membranes . . . . . . . . . . . . . . . . . . . . . 1476
Theory Background for the Membrane Interface . . . . . . . . . . 1478

The Membrane Interface 1486

Boundary, Edge, Point, and Pair Nodes for the Membrane Interface. . . 1488
Linear Elastic Material . . . . . . . . . . . . . . . . . . . . 1490
Layered Linear Elastic Material. . . . . . . . . . . . . . . . . 1494
Viscoelasticity . . . . . . . . . . . . . . . . . . . . . . . 1498

20 | C O N T E N T S
 Plasticity . . . . . . . . . . . . . . . . . . . . . . . . . 1500
Set Variables . . . . . . . . . . . . . . . . . . . . . . . 1501
Creep . . . . . . . . . . . . . . . . . . . . . . . . . . 1502
Additional Creep . . . . . . . . . . . . . . . . . . . . . . 1503
Viscoplasticity . . . . . . . . . . . . . . . . . . . . . . . 1505
Wrinkling . . . . . . . . . . . . . . . . . . . . . . . . 1506
Thermal Expansion (for Materials) . . . . . . . . . . . . . . . 1507
Hygroscopic Swelling . . . . . . . . . . . . . . . . . . . . 1509
Initial Stress and Strain. . . . . . . . . . . . . . . . . . . . 1511
External Stress. . . . . . . . . . . . . . . . . . . . . . . 1513
Inelastic Strain Rate . . . . . . . . . . . . . . . . . . . . . 1515
Safety . . . . . . . . . . . . . . . . . . . . . . . . . . 1516
Damping . . . . . . . . . . . . . . . . . . . . . . . . . 1517
Symmetry . . . . . . . . . . . . . . . . . . . . . . . . 1519
Antisymmetry . . . . . . . . . . . . . . . . . . . . . . . 1521
Layered Adiabatic Heating . . . . . . . . . . . . . . . . . . 1522

Chapter 8: Beam

Modeling with Beams 1528

About Beams . . . . . . . . . . . . . . . . . . . . . . . 1528
In-Plane Beams . . . . . . . . . . . . . . . . . . . . . . 1528
3D Beam . . . . . . . . . . . . . . . . . . . . . . . . . 1529
Orientation of the Cross Section. . . . . . . . . . . . . . . . 1530
Geometric Variables . . . . . . . . . . . . . . . . . . . . 1531
Result Evaluation . . . . . . . . . . . . . . . . . . . . . . 1532
Geometric Nonlinearity . . . . . . . . . . . . . . . . . . . 1534
Modeling Stiffeners . . . . . . . . . . . . . . . . . . . . . 1535
Nonrigid Joints . . . . . . . . . . . . . . . . . . . . . . 1536

Theory for the Beam Interface 1540

Shape Functions . . . . . . . . . . . . . . . . . . . . . . 1540
Geometric Nonlinearity . . . . . . . . . . . . . . . . . . . 1542
Strain-Displacement/Rotation Relation. . . . . . . . . . . . . . 1545
Stress-Strain Relation . . . . . . . . . . . . . . . . . . . . 1545
Thermal Strain . . . . . . . . . . . . . . . . . . . . . . . 1546

CONTENTS | 21
 Hygroscopic Swelling . . . . . . . . . . . . . . . . . . . . 1546
Initial Load and Strain . . . . . . . . . . . . . . . . . . . . 1547
Implementation . . . . . . . . . . . . . . . . . . . . . . 1548
Theory for Section Stiffness . . . . . . . . . . . . . . . . . . 1550
Stress Evaluation . . . . . . . . . . . . . . . . . . . . . . 1552
Common Cross Sections . . . . . . . . . . . . . . . . . . . 1554

The Beam Interface 1571

Boundary, Edge, Point, and Pair Nodes for the Beam Interface . . . . 1574
Initial Values . . . . . . . . . . . . . . . . . . . . . . . 1576
Cross-Section Data . . . . . . . . . . . . . . . . . . . . . 1577
Section Orientation . . . . . . . . . . . . . . . . . . . . . 1581
Linear Elastic Material . . . . . . . . . . . . . . . . . . . . 1584
Thermal Expansion (for Materials) . . . . . . . . . . . . . . . 1586
Hygroscopic Swelling . . . . . . . . . . . . . . . . . . . . 1587
Initial Stress and Strain. . . . . . . . . . . . . . . . . . . . 1588
External Stress. . . . . . . . . . . . . . . . . . . . . . . 1590
Section Stiffness . . . . . . . . . . . . . . . . . . . . . . 1591
Prescribed Displacement/Rotation . . . . . . . . . . . . . . . 1593
Prescribed Velocity . . . . . . . . . . . . . . . . . . . . . 1597
Prescribed Acceleration . . . . . . . . . . . . . . . . . . . 1599
Pinned . . . . . . . . . . . . . . . . . . . . . . . . . . 1601
Thermal Expansion (for Constraints) . . . . . . . . . . . . . . 1602
No Rotation . . . . . . . . . . . . . . . . . . . . . . . 1603
Symmetry . . . . . . . . . . . . . . . . . . . . . . . . 1604
Antisymmetry . . . . . . . . . . . . . . . . . . . . . . . 1606
Edge Load . . . . . . . . . . . . . . . . . . . . . . . . 1607
Point Load . . . . . . . . . . . . . . . . . . . . . . . . 1610
Point Load, Free . . . . . . . . . . . . . . . . . . . . . . 1612
Point Mass . . . . . . . . . . . . . . . . . . . . . . . . 1613
Point Mass Damping . . . . . . . . . . . . . . . . . . . . 1614
Rigid Connector . . . . . . . . . . . . . . . . . . . . . . 1615
Attachment . . . . . . . . . . . . . . . . . . . . . . . . 1622
Beam End Release . . . . . . . . . . . . . . . . . . . . . 1622
Edge Group . . . . . . . . . . . . . . . . . . . . . . . . 1624
Phase . . . . . . . . . . . . . . . . . . . . . . . . . . 1624
Harmonic Perturbation . . . . . . . . . . . . . . . . . . . 1625

22 | C O N T E N T S
Chapter 9: Beam Cross Section

Using the Beam Cross Section Interface 1628

About Beams and Cross-Section Data . . . . . . . . . . . . . . 1628
Using the Beam Cross Section Interface . . . . . . . . . . . . . 1629

Theory for the Beam Cross Section Interface 1635

Cross-Section Properties. . . . . . . . . . . . . . . . . . . 1635
Computation of Stresses . . . . . . . . . . . . . . . . . . . 1649

The Beam Cross Section Interface 1652

Homogeneous Cross Section . . . . . . . . . . . . . . . . . 1653
Hole . . . . . . . . . . . . . . . . . . . . . . . . . . 1655
Safety . . . . . . . . . . . . . . . . . . . . . . . . . . 1657

Chapter 10: Truss

Modeling with Truss Elements 1660

Theory for the Truss Interface 1662

Theory Background for the Truss Interface . . . . . . . . . . . . 1662

The Truss Interface 1671

Boundary, Edge, Point, and Pair Nodes for the Truss Interface . . . . 1672
Cross-Section Data . . . . . . . . . . . . . . . . . . . . . 1674
Straight Edge Constraint . . . . . . . . . . . . . . . . . . . 1677
Linear Elastic Material . . . . . . . . . . . . . . . . . . . . 1678
Set Variables . . . . . . . . . . . . . . . . . . . . . . . 1679
Thermal Expansion (for Linear Elastic Material) . . . . . . . . . . 1680
Thermal Expansion (for Shape Memory Alloy) . . . . . . . . . . . 1682
Hygroscopic Swelling . . . . . . . . . . . . . . . . . . . . 1683
Initial Stress and Strain. . . . . . . . . . . . . . . . . . . . 1684
External Stress. . . . . . . . . . . . . . . . . . . . . . . 1684
Spring-Damper Material . . . . . . . . . . . . . . . . . . . 1686
Prescribed Displacement . . . . . . . . . . . . . . . . . . . 1687

CONTENTS | 23
 Pinned . . . . . . . . . . . . . . . . . . . . . . . . . . 1691
Thermal Expansion (for Constraints) . . . . . . . . . . . . . . 1691
Symmetry . . . . . . . . . . . . . . . . . . . . . . . . 1693
Antisymmetry . . . . . . . . . . . . . . . . . . . . . . . 1695
Edge Load . . . . . . . . . . . . . . . . . . . . . . . . 1696
Point Mass . . . . . . . . . . . . . . . . . . . . . . . . 1698
Point Mass Damping . . . . . . . . . . . . . . . . . . . . 1699

Chapter 11: Wire

Modeling Wires and Cables 1702

Theory for the Wire Interface 1703

Theory Background for the Wire Interface . . . . . . . . . . . . 1703

The Wire Interface 1706

Boundary, Edge, Point, and Pair Nodes for the Wire Interface. . . . . 1707
Elastic Wire . . . . . . . . . . . . . . . . . . . . . . . . 1709
Thermal Expansion (for Elastic Wire) . . . . . . . . . . . . . . 1710
Hygroscopic Swelling . . . . . . . . . . . . . . . . . . . . 1711
Initial Stress and Strain. . . . . . . . . . . . . . . . . . . . 1712
Pinned . . . . . . . . . . . . . . . . . . . . . . . . . . 1713
Thermal Expansion (for Constraints) . . . . . . . . . . . . . . 1713
Symmetry . . . . . . . . . . . . . . . . . . . . . . . . 1715
Antisymmetry . . . . . . . . . . . . . . . . . . . . . . . 1717
Edge Load . . . . . . . . . . . . . . . . . . . . . . . . 1718

Chapter 12: Pipe Mechanics

Theory for the Pipe Mechanics Interface 1722

About Pipes . . . . . . . . . . . . . . . . . . . . . . . . 1722
Effects of Internal Pressure . . . . . . . . . . . . . . . . . . 1723
Fluid Flow Forces. . . . . . . . . . . . . . . . . . . . . . 1727
Effects of Thermal Expansion . . . . . . . . . . . . . . . . . 1729
Stress Evaluation . . . . . . . . . . . . . . . . . . . . . . 1731

24 | C O N T E N T S
Connection Between Pipes and Structures . . . . . . . . . . . . 1736
Cross Sections. . . . . . . . . . . . . . . . . . . . . . . 1738

The Pipe Mechanics Interface 1743

Boundary, Edge, Point, and Pair Nodes for the Pipe Mechanics
Interface. . . . . . . . . . . . . . . . . . . . . . . . 1746
Pipe Cross Section . . . . . . . . . . . . . . . . . . . . . 1748
Section Orientation . . . . . . . . . . . . . . . . . . . . . 1752
Bend . . . . . . . . . . . . . . . . . . . . . . . . . . 1754
Fluid and Pipe Materials . . . . . . . . . . . . . . . . . . . 1756
Rigid Material . . . . . . . . . . . . . . . . . . . . . . . 1758
Thermal Expansion (for Materials) . . . . . . . . . . . . . . . 1760
Gravity . . . . . . . . . . . . . . . . . . . . . . . . . 1762
Base Excitation . . . . . . . . . . . . . . . . . . . . . . 1763
Linearly Accelerated Frame . . . . . . . . . . . . . . . . . . 1764
Fluid Load . . . . . . . . . . . . . . . . . . . . . . . . 1766

CONTENTS | 25
 Chapter 13: Multiphysics Interfaces and Couplings

The Thermal Stress, Solid Interface 1772

The Thermal Stress, Shell Interface 1774

The Thermal Stress, Layered Shell Interface 1776

The Thermal Stress, Membrane Interface 1778

The Joule Heating and Thermal Expansion Interface 1780

Overview of Fluid-Structure Interaction Interfaces 1783

The Fluid-Solid Interaction Interface 1785

The Fluid-Solid Interaction, Fixed Geometry Interface 1788

The Fluid-Shell Interaction Interface 1790

The Fluid-Shell Interaction, Fixed Geometry Interface 1792

The Fluid-Membrane Interaction Interface 1794

The Fluid-Membrane Interaction, Fixed Geometry Interface 1796

The Fluid-Pipe Interaction, Fixed Geometry Interface 1798

The Solid Thin-Film Damping Interface 1799

The Shell Thin-Film Damping Interface 1801

The Fluid-Solid Interaction, Viscoelastic Flow Interface 1803

The Fluid-Solid Interaction, Viscoelastic Flow, Fixed Geometry

26 | C O N T E N T S
 Interface 1805

The Fluid-Solid Interaction, Conjugate Heat Transfer Interface

1807

The Fluid-Solid Interaction, Two-Phase Flow, Phase Field

Interface 1810

The Fluid-Solid Interaction, Two-Phase Flow, Phase Field,

Fixed Geometry Interface 1813

Overview of Electromagnetics-Structure Interaction Interfaces

1816

The Piezoelectricity Interface, Solid 1817

Coupling Piezoelectricity with Acoustics . . . . . . . . . . . . . 1820

The Piezoelectricity, Layered Shell Interface 1821

The Electrostriction Interface 1824

The Ferroelectroelasticity Interface 1825

The Magnetomechanics Interface 1826

The Magnetomechanics, No Currents Interface 1828

The Piezomagnetism Interface 1830

The Nonlinear Magnetostriction Interface 1833

Hygroscopic Swelling Coupling 1836

Multiphysics Couplings 1837

Thermal Expansion . . . . . . . . . . . . . . . . . . . . . 1837
Layered Thermal Expansion . . . . . . . . . . . . . . . . . . 1840
Hygroscopic Swelling . . . . . . . . . . . . . . . . . . . . 1843
Piezoelectric Effect . . . . . . . . . . . . . . . . . . . . . 1845

CONTENTS | 27
 Layered Piezoelectric Effect . . . . . . . . . . . . . . . . . . 1847
Electrostriction . . . . . . . . . . . . . . . . . . . . . . 1848
Magnetomechanical Forces . . . . . . . . . . . . . . . . . . 1852
Magnetic Forces . . . . . . . . . . . . . . . . . . . . . . 1854
Piezomagnetic Effect . . . . . . . . . . . . . . . . . . . . 1856
Nonlinear Magnetostriction . . . . . . . . . . . . . . . . . . 1858
Magnetostriction . . . . . . . . . . . . . . . . . . . . . . 1861
Lorentz Coupling 1861
Fluid-Structure Interaction 1863
Fluid-Structure Interaction, Pair . . . . . . . . . . . . . . . . 1865
Fluid-Structure Interaction, Fixed Geometry . . . . . . . . . . . 1868
Fluid-Pipe Interaction . . . . . . . . . . . . . . . . . . . . 1868
Structure Thin-Film Flow Interaction 1870
Shell Thin-Film Flow Interaction . . . . . . . . . . . . . . . . 1871

Couplings Between Structural Mechanics Interfaces 1874

Solid-Thin Structure Connection . . . . . . . . . . . . . . . . 1874
Solid-Beam Connection . . . . . . . . . . . . . . . . . . . 1877
Shell-Beam Connection . . . . . . . . . . . . . . . . . . . 1882
Layered Shell-Structure Transition . . . . . . . . . . . . . . . 1884
Layered Shell-Structure Cladding . . . . . . . . . . . . . . . . 1885
Layered Shell-Shell Connection . . . . . . . . . . . . . . . . 1886
Structure-Pipe Connection . . . . . . . . . . . . . . . . . . 1887
Embedded Reinforcement . . . . . . . . . . . . . . . . . . 1888
Beam Cross Section-Beam Coupling . . . . . . . . . . . . . . 1891
Beam-Beam Cross Section Coupling . . . . . . . . . . . . . . 1892

Chapter 14: Glossary

Glossary of Terms 1896

Index 1903

28 | C O N T E N T S
 1

Introduction

This guide describes the Structural Mechanics Module, an optional add-on

package that extends the COMSOL Multiphysics modeling environment with
customized physics interfaces that solve problems in the fields of structural and
solid mechanics, including special physics interface for modeling of shells,
membranes, beams, plates, trusses, wires, and pipes.

This chapter introduces you to the capabilities of this module and includes a
summary of the physics interfaces as well as information about where you can find
additional documentation and model examples. The last section is a brief overview
with links to each chapter in this guide.

• About the Structural Mechanics Module

• Overview of the User’s Guide

29
 About the Structural Mechanics
Module
In this section:

• A Quick Overview of the Structural Mechanics Module

• What Problems Can It Solve?
• The Structural Mechanics Module Physics Interface Guide
• Common Physics Interface and Feature Settings and Nodes
• Where Do I Access the Documentation and Application Libraries?

The Physics Interfaces and Building a COMSOL Multiphysics Model in

the COMSOL Multiphysics Reference Manual

A Quick Overview of the Structural Mechanics Module

The Structural Mechanics Module solves problems in the fields of structural and solid
mechanics, adding special physics interfaces for modeling shells and beams, for
example.

The physics interfaces in this module are fully multiphysics enabled, making it possible
to couple them to any other physics interfaces in COMSOL Multiphysics or the other
modules. Available physics interfaces include:

• Solid mechanics for 1D and 2D plane stress, plane strain, and generalized plane
strain, 2D axial symmetry, and 3D solids
• Beams in 2D and 3D, Euler and Timoshenko theory
• Pipes in 2D and 3D
• Truss elements in 2D and 3D
• Wires in 2D and 3D
• Shells and plates, Mindlin theory, 3D and 2D axial symmetry
• Membranes, 3D and 2D axial symmetry
• With the Composite Materials Module, the Layered Shell interface is also available.

30 | CHAPTER 1: INTRODUCTION
The module’s study capabilities include static, eigenfrequency, time dependent
(transient), frequency response, buckling, response spectrum, random vibration, and
parametric studies.

There are several material models:

• Linear Elastic Materials can be isotropic, orthotropic, or fully anisotropic, and you
can use local coordinate systems to specify material properties.
• Linear Viscoelastic Materials
• Piezoelectric Materials
• Magnetostrictive Materials are available when used together with the AC/DC
module.
• Material models for hyperelasticity, metal plasticity, porous plasticity, creep,
viscoplasticity, nonlinear elasticity, soil plasticity, concrete, rocks, and clay are
available with the optional Nonlinear Structural Materials Module and
Geomechanics Module.

Large deformations, as well as contact and friction, can also be modeled.

Coupling structural analysis with thermal analysis is one example of multiphysics easily
implemented with the module, which provides predefined multiphysics couplings for
thermal stress and other types of multiphysics. Piezoelectric materials, coupling the
electric field and strain in both directions are fully supported inside the module
through special multiphysics couplings solving for both the electric potential and the
displacements. Structural mechanics couplings are common in simulations done with
COMSOL Multiphysics and occur in interaction with, for example, fluid flow (fluid–
structure interaction, FSI), chemical reactions, acoustics, electric fields, magnetic
fields, and optical wave propagation.

What Problems Can It Solve?

The Structural Mechanics Module contains a set of physics interfaces adapted to a
broad category of structural-mechanics analyses. The module serves as an excellent
tool for the professional engineer, researcher, and teacher. In education, the benefit of
the short learning curve is especially useful because educators need not spend excessive
time learning the software and can instead focus on the physics and the modeling
process.

A short summary is given below.

ABOUT THE STRUCTURAL MECHANICS MODULE | 31

 STATIC ANALYSIS
In a static analysis, the load and constraints are fixed in time.

EIGENFREQUENCY ANALYSIS
An eigenfrequency analysis finds the damped or undamped eigenfrequencies and
mode shapes of a structure, sometimes referred to as the free vibration of a structure.
Prestress effects and damping can also be taken into account.

TRANSIENT ANALYSIS
A transient analysis finds the transient response for a time-dependent model, taking
into account mass and mass moment of inertia. The transient analysis can be either
direct or using a modal solution.

FREQUENCY RESPONSE ANALYSIS

A frequency response analysis finds the steady-state response to harmonic loads. The
frequency-response analysis can be either direct or using a modal solution. Effects of
prestress can be included.

LINEAR BUCKLING STUDY

A linear buckling analysis uses the stiffness coming from stresses and material to
define an eigenvalue problem where the eigenvalue is a load factor that, when
multiplied with the actual load, gives the critical load in a linear context.

PARAMETRIC ANALYSIS
A parametric analysis finds the solution dependence due to the variation of a specific
parameter, which could be, for instance, a material property or the position of a load.

RESPONSE SPECTRUM ANALYSIS

Response spectrum analysis provides a method to estimate peak values of, for example,
displacements and stresses when a structure is subjected to a short, non-deterministic
event like an earthquake or a shock.

RANDOM VIBRATION ANALYSIS

For steady-state random dynamic loading, like wind or waves, it is possible to perform
random vibration analysis where the input is given in terms of a power spectrum
density (PSD).

32 | CHAPTER 1: INTRODUCTION
THERMAL STRESS
In a transient thermal stress study, the program neglects mass effects, assuming that
the time scale in the structural mechanics problem is much smaller than the time scale
in the thermal problem.

LARGE DEFORMATIONS
You can also enable geometric nonlinearity for all structural mechanics interfaces. The
engineering strain is then replaced with the Green–Lagrange strain and the stress with
the second Piola–Kirchhoff stress. To solve the problem, the program uses a total
Lagrangian formulation.

ELASTOPLASTIC MATERIALS
An elastoplastic analysis involves a nonlinear material with or without hardening.
Several isotropic and kinematic hardening models are available.

The material models allow large strains.

The elastoplastic material models are available in the Solid Mechanics, Shell, Layered
Shell, Membrane, and Truss interfaces.

CREEP AND VISCOPLASTIC MATERIALS

A number of different material models for creep and viscoplasticity are available. In
these materials, the rate of strain depends on the stress.

HYPERELASTIC MATERIALS
In hyperelastic materials, the stresses are computed from a strain energy density
function. They are often used to model rubber and biological tissue, but are also used
in acoustic elasticity. Many different models are available.

The hyperelastic materials are available in the Solid Mechanics, Shell, Layered Shell,
and Membrane interfaces.

NONLINEAR ELASTIC MATERIALS

The nonlinear elastic materials are intended for materials that exhibit a nonlinear
behavior already at small strains. Some brittle material as well as soils show this
behavior.

ABOUT THE STRUCTURAL MECHANICS MODULE | 33

 VISCOELASTIC MATERIALS
Viscoelastic materials have a time-dependent response, even if the loading is constant.
Viscoelasticity is available in the Solid Mechanics, Shell, Layered Shell, and Membrane
interfaces.

• Hyperelastic, elastoplastic, creep, viscoplastic, and nonlinear elastic

material models are available with the Nonlinear Structural Materials
Module.
• Additional functionality and material models for geomechanics and soil
mechanics — nonlinear elasticity, soil plasticity, concrete, rock, and clay
material models — is available with the Geomechanics Module.

CONTACT MODELING
You can model contact between parts of a structure. The Solid Mechanics, Shell,
Layered Shell, and Membrane interfaces support contact with or without friction.
Three contact algorithms are available: penalty, augmented Lagrangian, and Nitsche
methods. The contact models can be augmented with adhesion and decohesion.

SPECIAL MODELING TECHNIQUES

In the Structural Mechanics Module, you will find support for many important special
modeling techniques. Some examples are:

• Bolt pretension
• Bolt thread modeling
• Stress linearization
• Rigid connectors
• Transitions between solid, shell, and beam elements

The Structural Mechanics Module Physics Interface Guide

The Poroelasticity interface requires and is coupled with the Structural

Mechanics Module, and is discussed in the Subsurface Flow Module
User’s Guide.

34 | CHAPTER 1: INTRODUCTION
 At any time, a new model can be created or physics interfaces added. Right-click the
Root (top) node and select Add Component or right-click a Component node and select
Add Physics.

Depending on the physics interface, specify parameters defining a problem on points,

edges (3D), boundaries, and domains. It is possible to specify loads and constraints on
all available geometry levels, but material properties can only be specified for the
domains, except for shells, membranes, beams, and trusses, where they are defined on
the boundary or edge level.

In the COMSOL Multiphysics Reference Manual:

• Studies and Solvers

• The Physics Interfaces
• For a list of all the core physics interfaces included with a COMSOL
Multiphysics license, see Physics Interface Guide.

PHYSICS INTERFACE ICON TAG SPACE AVAILABLE STUDY TYPE

DIMENSION

Acoustics

Elastic Waves

Elastic Waves, Time elte 3D, 2D, 2D time dependent

Explicit axisymmetric
Piezoelectric Waves, — 3D, 2D, 2D time dependent
Time Explicit axisymmetric
Fluid Flow

Fluid-Structure Interaction

Fluid-Solid Interaction2 — 3D, 2D, 2D stationary; time dependent

axisymmetric
Fluid-Shell Interaction2 — 3D, 2D stationary; time dependent
axisymmetric

ABOUT THE STRUCTURAL MECHANICS MODULE | 35

 PHYSICS INTERFACE ICON TAG SPACE AVAILABLE STUDY TYPE
DIMENSION

Fluid-Shell Interaction, — 3D, 2D stationary; stationary, one

Fixed Geometry2 axisymmetric way; time dependent; time
dependent, one way
Fluid-Membrane — 3D, 2D stationary; time dependent
Interaction2 axisymmetric
Fluid-Membrane — 3D, 2D stationary; stationary, one
Interaction, Fixed axisymmetric way; time dependent; time
Geometry2 dependent, one way
Fluid-Pipe Interaction, — 3D, 2D stationary; time dependent
Fixed Geometry7
Conjugate Heat Transfer, — 3D, 2D, 2D stationary; time dependent;
Fluid-Solid Interaction2 axisymmetric
Fluid-Solid Interaction, — 3D, 2D, 2D time dependent; time
Viscoelastic Flow2,10 axisymmetric dependent with phase
initialization
Fluid-Solid Interaction, — 3D, 2D, 2D time dependent; time
Two-Phase Flow, Phase axisymmetric dependent with phase
Field2, 6 initialization

36 | CHAPTER 1: INTRODUCTION
PHYSICS INTERFACE ICON TAG SPACE AVAILABLE STUDY TYPE
DIMENSION

Structural Mechanics

Solid Mechanics1 solid 3D, 2D, 2D stationary; eigenfrequency;

axisymmetric, eigenfrequency,
1D, 1D prestressed; mode analysis;
axisymmetric time dependent; time
dependent, modal; time
dependent, modal
reduced-order model;
frequency domain;
frequency domain, modal;
frequency domain,
prestressed; frequency
domain, prestressed,
modal; frequency domain,
modal reduced-order
model; frequency domain,
AWE reduced-order
model; response spectrum;
random vibration (PSD);
linear buckling; bolt
pretension
Thermal Stress, Solid2 — 3D, 2D, 2D stationary; time dependent
axisymmetric
Thermal Stress, Shell2, 4 — 3D, 2D stationary; time dependent
axisymmetric
Thermal Stress, — 3D, 2D stationary; time dependent
Membrane2, 4 axisymmetric
Joule Heating and Thermal — 3D, 2D, 2D stationary; time dependent
Expansion2 axisymmetric

ABOUT THE STRUCTURAL MECHANICS MODULE | 37

 PHYSICS INTERFACE ICON TAG SPACE AVAILABLE STUDY TYPE
DIMENSION

Shell shell 3D stationary; eigenfrequency;

eigenfrequency,
prestressed; time
dependent; time
dependent, modal; time
dependent, modal
reduced-order model;
frequency domain;
frequency domain, modal;
frequency domain,
prestressed; frequency
domain, prestressed,
modal; frequency domain,
modal reduced-order
model; response spectrum;
random vibration (PSD);
linear buckling
Plate plate 2D stationary; eigenfrequency;
eigenfrequency,
prestressed; time
dependent; time
dependent, modal; time
dependent, modal
reduced-order model;
frequency domain;
frequency domain, modal;
frequency domain,
prestressed; frequency
domain, prestressed,
modal; frequency domain,
modal reduced-order
model; response spectrum;
random vibration (PSD);
linear buckling

38 | CHAPTER 1: INTRODUCTION
PHYSICS INTERFACE ICON TAG SPACE AVAILABLE STUDY TYPE
DIMENSION

Beam beam 3D, 2D stationary; eigenfrequency;

time dependent, modal;
time dependent, modal
reduced-order model;
frequency domain;
frequency domain, modal;
frequency domain, modal
reduced-order model; time
dependent; response
spectrum; random
vibration (PSD); linear
buckling
Beam Cross Section bcs 3D, 2D stationary

Truss truss 3D, 2D stationary; eigenfrequency;

eigenfrequency,
prestressed; time
dependent; time
dependent, modal; time
dependent, modal
reduced-order model;
frequency domain;
frequency domain, modal;
frequency domain,
prestressed; frequency
domain, prestressed,
modal; frequency domain,
modal reduced-order
model; response spectrum;
random vibration (PSD);
linear buckling

ABOUT THE STRUCTURAL MECHANICS MODULE | 39

 PHYSICS INTERFACE ICON TAG SPACE AVAILABLE STUDY TYPE
DIMENSION

Wire wire 3D, 2D stationary; eigenfrequency;

eigenfrequency,
prestressed; time
dependent; time
dependent, modal; time
dependent, modal
reduced-order model;
frequency domain;
frequency domain, modal;
frequency domain,
prestressed; frequency
domain, prestressed,
modal; frequency domain,
modal reduced-order
model; response spectrum;
random vibration (PSD);
linear buckling
Membrane mbrn 3D, 2D stationary; eigenfrequency;
axisymmetric eigenfrequency,
prestressed; time
dependent; time
dependent, modal; time
dependent, modal
reduced-order model;
frequency domain;
frequency domain, modal;
frequency domain,
prestressed; frequency
domain, prestressed,
modal; frequency domain,
modal reduced-order
model; response spectrum;
random vibration (PSD)

40 | CHAPTER 1: INTRODUCTION
PHYSICS INTERFACE ICON TAG SPACE AVAILABLE STUDY TYPE
DIMENSION

Pipe Mechanics pipem 3D, 2D stationary; eigenfrequency;

time dependent, modal;
time dependent, modal
reduced-order model;
frequency domain;
frequency domain, modal;
frequency domain, modal
reduced-order model; time
dependent; response
spectrum; random
vibration (PSD)
Solid Thin-Film Damping — 3D, 2D, 2D stationary; eigenfrequency;
axisymmetric time dependent; time
dependent, modal; time
dependent, modal
reduced-order model;
frequency domain;
frequency domain, modal;
frequency domain, modal
reduced-order model; time
dependent; response
spectrum; random
vibration (PSD)
Shell Thin-Film Damping — 3D, 2D stationary; eigenfrequency;
axisymmetric time dependent; time
dependent, modal; time
dependent, modal
reduced-order model;
frequency domain;
frequency domain, modal;
frequency domain, modal
reduced-order model; time
dependent; response
spectrum; random
vibration (PSD)

ABOUT THE STRUCTURAL MECHANICS MODULE | 41

 PHYSICS INTERFACE ICON TAG SPACE AVAILABLE STUDY TYPE
DIMENSION

Piezoelectricity2 — 3D, 2D, 2D stationary; eigenfrequency;

axisymmetric eigenfrequency,
prestressed; time
dependent; time
dependent, modal;
frequency domain;
frequency domain, modal;
frequency domain,
prestressed; frequency
domain, prestressed,
modal; small-signal analysis,
frequency domain; linear
buckling
Piezoelectricity, Layered — 3D stationary; eigenfrequency;
Shell2 5 time dependent; frequency
domain
Magnetostriction2,3 — 3D, 2D, 2D stationary; eigenfrequency;
axisymmetric time dependent; frequency
domain; small-signal
analysis, frequency domain;
eigenfrequency,
prestressed; frequency
domain, prestressed
Magnetomechanics2,3 — 3D, 2D, 2D stationary; eigenfrequency;
axisymmetric time dependent; frequency
domain; small-signal
analysis, frequency domain;
eigenfrequency,
prestressed; frequency
domain, prestressed
Magnetomechanics, No — 3D, 2D, 2D stationary; eigenfrequency;
Currents2,3 axisymmetric time dependent; frequency
domain; small-signal
analysis, frequency domain;
eigenfrequency,
prestressed; frequency
domain, prestressed
Ferroelectroelasticity2,9 — 3D, 2D, 2D stationary; time dependent;
axisymmetric frequency domain

42 | CHAPTER 1: INTRODUCTION
PHYSICS INTERFACE ICON TAG SPACE AVAILABLE STUDY TYPE
DIMENSION

Electrostriction2,9 — 3D, 2D, 2D stationary; time dependent;

axisymmetric frequency domain
1
This physics interface is included with the core COMSOL Multiphysics software but has
added functionality for this module.
2 This physics interface is a predefined multiphysics coupling that automatically adds all the
physics interfaces and coupling features required.
3
Requires the addition of the AC/DC Module.

4 Requires the addition of the Heat Transfer Module.

5
Requires the addition of the Composite Materials Module.

6 Requires the addition of the CFD Module, or the Polymer Flow, or the Microfluidics
Module.
7
Requires the addition of the Pipe Flow Module.

8
Requires the addition of the Porous Media Flow Module.

9
Requires the addition of the AC/DC Module or the MEMS Module.

10
Requires the addition of the Polymer Flow Module.

Common Physics Interface and Feature Settings and Nodes

There are several common settings and sections available for the physics interfaces and
feature nodes. Some of these sections also have similar settings or are implemented in
the same way no matter the physics interface or feature being used. There are also some
physics feature nodes that display in COMSOL Multiphysics.

ABOUT THE STRUCTURAL MECHANICS MODULE | 43

 In each module’s documentation, only unique or extra information is included;
standard information and procedures are centralized in the COMSOL Multiphysics
Reference Manual.

In the COMSOL Multiphysics Reference Manual, see Table 2-4 for

links to common sections and Table 2-5 to common feature nodes.
You can also search for information: press F1 to open the Help
window or Ctrl+F1 to open the Documentation window.

Where Do I Access the Documentation and Application Libraries?

A number of online resources have more information about COMSOL, including
licensing and technical information. The electronic documentation, topic-based (or
context-based) help, and the Application Libraries are all accessed through the
COMSOL Desktop.

If you are reading the documentation as a PDF file on your computer,

the blue links do not work to open an application or content
referenced in a different guide. However, if you are using the Help
system in COMSOL Multiphysics, these links work to open other
modules, application examples, and documentation sets.

THE DOCUMENTATION AND ONLINE HELP

The COMSOL Multiphysics Reference Manual describes the core physics interfaces
and functionality included with the COMSOL Multiphysics license. This book also has
instructions on how to use COMSOL Multiphysics and how to access the electronic
Documentation and Help content.

Opening Topic-Based Help

The Help window is useful as it is connected to the features in the COMSOL Desktop.
To learn more about a node in the Model Builder, or a window on the Desktop, click
to highlight a node or window, then press F1 to open the Help window, which then

44 | CHAPTER 1: INTRODUCTION
displays information about that feature (or click a node in the Model Builder followed
by the Help button ( ). This is called topic-based (or context) help.

To open the Help window:

• In the Model Builder, Application Builder, or Physics Builder, click a node

or window and then press F1.
• On any toolbar (for example, Home, Definitions, or Geometry), hover the
mouse over a button (for example, Add Physics or Build All) and then
press F1.
• From the File menu, click Help ( ).
• In the upper-right corner of the COMSOL Desktop, click the Help ( )
button.

To open the Help window:

• In the Model Builder or Physics Builder, click a node or window and then
press F1.
• On the main toolbar, click the Help ( ) button.
• From the main menu, select Help>Help.

Opening the Documentation Window

To open the Documentation window:

• Press Ctrl+F1.
• From the File menu, select Help>Documentation ( ).

To open the Documentation window:

• Press Ctrl+F1.
• On the main toolbar, click the Documentation ( ) button.
• From the main menu, select Help>Documentation.

ABOUT THE STRUCTURAL MECHANICS MODULE | 45

 THE APPLICATION LIBRARIES WINDOW
Each model or application includes documentation with the theoretical background
and step-by-step instructions to create a model or application. The models and
applications are available in COMSOL Multiphysics as MPH-files that you can open
for further investigation. You can use the step-by-step instructions and the actual
models as templates for your own modeling. In most models, SI units are used to
describe the relevant properties, parameters, and dimensions, but other unit systems
are available.

Once the Application Libraries window is opened, you can search by name or browse
under a module folder name. Click to view a summary of the model or application and
its properties, including options to open it or its associated PDF document.

The Application Libraries Window in the COMSOL Multiphysics

Reference Manual.

Opening the Application Libraries Window

To open the Application Libraries window ( ):

From the File menu, select Application Libraries.

To include the latest versions of model examples, from the File>Help

menu, select ( ) Update COMSOL Application Library.

Select Application Libraries from the main File or Windows menus.

To include the latest versions of model examples, from the Help menu,
select ( ) Update COMSOL Application Library.

CONTACTING COMSOL BY EMAIL

For general product information, contact COMSOL at [email protected].

COMSOL ACCESS AND TECHNICAL SUPPORT

To receive technical support from COMSOL for the COMSOL products, please
contact your local COMSOL representative or send your questions to
[email protected]. An automatic notification and a case number will be sent to you

46 | CHAPTER 1: INTRODUCTION
by email. You can also access technical support, software updates, license information,
and other resources by registering for a COMSOL Access account.

COMSOL ONLINE RESOURCES

COMSOL website www.comsol.com

Contact COMSOL www.comsol.com/contact
COMSOL Access www.comsol.com/access
Support Center www.comsol.com/support
Product Download www.comsol.com/product-download
Product Updates www.comsol.com/support/updates
COMSOL Blog www.comsol.com/blogs
Discussion Forum www.comsol.com/forum
Events www.comsol.com/events
COMSOL Application Gallery www.comsol.com/models
COMSOL Video Gallery www.comsol.com/video
Support Knowledge Base www.comsol.com/support/knowledgebase

ABOUT THE STRUCTURAL MECHANICS MODULE | 47

 Overview of the User’s Guide
The Structural Mechanics Module User’s Guide gets you started with modeling using
COMSOL Multiphysics. The information in this guide is specific to this module.
Instructions how to use COMSOL in general are included with the COMSOL
Multiphysics Reference Manual.

As detailed in the section Where Do I Access the Documentation and

Application Libraries?, this information can also be searched from the Help
menu in COMSOL Multiphysics.

TABLE OF CONTENTS, GLOSSARY, AND INDEX

To help you navigate through this guide, see the Contents, Glossary, and Index.

MODELING WITH THE STRUCTURAL MECHANICS MODULE

The Structural Mechanics Modeling chapter gives you an insight on how to approach
the modeling of various structural mechanics problems.

STRUCTURAL MECHANICS THEORY

The Structural Mechanics Theory chapter introduces the general theory on which the
physics interfaces in the Structural Mechanics Module are based.

THE SOLID MECHANICS INTERFACE

The Solid Mechanics chapter describes The Solid Mechanics Interface, which is used
to model 3D solids, plane strain and plane stress 2D and 1D models, and 2D
axisymmetric models.

THE SHELL AND PLATE INTERFACES

The Shell and Plate chapter describes The Shell and Plate Interfaces, which are used to
model thin 3D structures (shell) and out-of-plane loaded plates (plate). The
underlying theory is described in Theory for the Shell and Plate Interfaces.

THE BEAM INTERFACE

The Beam chapter describes The Beam Interface, which contains Euler (Euler–
Bernoulli) and Timoshenko beams for modeling slender 3D and 2D structures.
Typical examples are frameworks and latticeworks. The underlying theory for the
physics interface is described in Theory for the Beam Interface.

48 | CHAPTER 1: INTRODUCTION
THE BEAM CROSS SECTION INTERFACE
The Beam Cross Section chapter describes The Beam Cross Section Interface, which
is used for computing cross section properties for beams. It can also be used for a
detailed evaluation of stresses in a beam when the section forces to which it is subjected
are known. The first section discusses Using the Beam Cross Section Interface, and the
underlying theory is described in Theory for the Beam Cross Section Interface.

THE TRUSS INTERFACE

The Truss chapter describes The Truss Interface, which models slender 3D and 2D
structures with components capable to withstand axial forces only. Typical applications
are latticeworks, but it can also be used for modeling cables. In the section Modeling
with Truss Elements, you will find a discussion about how to set up models using this
interface. The underlying theory for the physics interface is described in Theory for the
Truss Interface.

THE MEMBRANE INTERFACE

The Membrane chapter describes The Membrane Interface, which can be used for
prestressed membranes, cladding on solids, and balloons, for example. The underlying
theory for the physics interface is also included in Theory for the Membrane Interface.

THE MULTIPHYSICS INTERFACES

The Multiphysics Interfaces and Couplings chapter describes these physics interfaces
found under the Structural Mechanics branch when adding a physics interface:

• The Thermal Stress, Solid Interface combines a Solid Mechanics interface with a
Heat Transfer interface. The coupling appears on the domain level, where the
temperature from the Heat Transfer interface acts as a thermal load for the Solid
Mechanics interface, causing thermal expansion.
• The Thermal Stress, Shell Interface combines a Shell interface with a Heat Transfer
in Shells interface. The coupling appears on the boundary level, where the
temperature from the Heat Transfer in Shells interface acts as a thermal load for the
Shell interface, causing thermal expansion.
• The Thermal Stress, Layered Shell Interface combines a Layered Shell interface with
a Heat Transfer in Shells interface. The coupling appears on the boundary level,
where the temperature from the Heat Transfer in Shells interface acts as a thermal
load for the Layered Shell interface, causing thermal expansion.
• The Thermal Stress, Membrane Interface combines a Membrane interface with a
Heat Transfer in Shells interface. The coupling appears on the boundary level,

OVERVIEW OF THE USER’S GUIDE | 49

 where the temperature from the Heat Transfer in Shells interface acts as a thermal
load for the Membrane interface, causing thermal expansion.
• The Joule Heating and Thermal Expansion Interface combines solid mechanics
using a thermal linear elastic material with an electromagnetic Joule heating model.
This is a multiphysics combination of solid mechanics, electric currents, and heat
transfer for modeling of, for example, thermoelectromechanical (TEM)
applications.
• The Piezoelectricity Interface, Solid combines a Solid Mechanics interface with an
Electrostatics interface. Piezoelectric materials in 3D, 2D plane strain and plane
stress, and 2D axial symmetry can be modeled.
• The Piezoelectricity, Layered Shell Interface combines a Layered Shell interface with
an Electric Currents in Layered Shells interface. This makes it possible to model
piezoelectric effects in thin layered structures.
• The Electrostriction Interface combines a Solid Mechanics with an Electrostatics
interface. Using this interface, you can solve problems where strains are caused by
electrostrictive effects.
• The Ferroelectroelasticity Interface combines a Solid Mechanics with an
Electrostatics interface. Using this interface, you can solve problems involving
ferroelectric materials, for example within nonlinear piezoelectricity.
• Hygroscopic Swelling Coupling combines a Solid Mechanics with a Magnetic Fields
interface. Using this interface, you can solve problems in the magnetostrictive field
with linear as well as nonlinear material models.
• The Fluid-Solid Interaction Interface combines fluid flow with the Solid Mechanics
interface to capture the interaction between the fluid and the solid in a situation
where the fluid domain has significant deformation. The solid material exists on
domains which are adjacent to the fluid.
• The Fluid-Shell Interaction Interface combines fluid flow with the Shell interface to
capture the interaction between the fluid and the solid in a situation where the fluid
domain has significant deformation. The shell is modeled on the boundary of the
fluid.
• The Fluid-Membrane Interaction Interface combines fluid flow with the Membrane
interface to capture the interaction between the fluid and the membrane in a
situation where the fluid domain has significant deformation. The membrane is
modeled on the boundary of the fluid.
• The Fluid-Solid Interaction, Fixed Geometry Interface combines fluid flow with the
Solid Mechanics interface to capture the interaction between the fluid and the solid

50 | CHAPTER 1: INTRODUCTION
 in a situation where the fluid domain can be considered to be nondeforming. The
solid material exists on domains which are adjacent to the fluid.
• The Fluid-Shell Interaction, Fixed Geometry Interface combines fluid flow with the
Shell interface to capture the interaction between the fluid and the solid in a
situation where the fluid domain can be considered to be nondeforming. The shell
is modeled on the boundary of the fluid
• The Fluid-Membrane Interaction, Fixed Geometry Interface combines fluid flow
with the Membrane interface to capture the interaction between the fluid and the
membrane in a situation where the fluid domain can be considered to be
nondeforming. The membrane is modeled on the boundary of the fluid.
• The Fluid-Pipe Interaction, Fixed Geometry Interface combines flow computed
using the Pipe Flow interface with structural analysis in the Pipe Mechanics
interface. Different types of fluid loads are transferred to the structural analysis.
• The Fluid-Solid Interaction, Conjugate Heat Transfer Interface combines fluid flow
with the Solid Mechanics interface and the Heat Transfer in Solids and Fluids
interface. It combines fluid-structure interaction modeling with a nonisothermal
flow. Heat transfer is considered both in the fluid and in the solid in order to capture
thermal expansion effects.
• The Fluid-Solid Interaction, Two-Phase Flow, Phase Field Interface combines
two-phase fluid flow with the Solid Mechanics interface to capture the interaction
between the fluid and the solid in a situation where the fluid domain has significant
deformation. The solid material exists on domains which are adjacent to the fluid.
• The Fluid-Solid Interaction, Two-Phase Flow, Phase Field, Fixed Geometry
Interface combines two-phase fluid flow with the Solid Mechanics interface to
capture the interaction between the fluid and the solid in a situation where the fluid
domain can be considered to be nondeforming. The solid material exists on
domains which are adjacent to the fluid.

OVERVIEW OF THE USER’S GUIDE | 51

52 | CHAPTER 1: INTRODUCTION
 2

Structural Mechanics Modeling

The goal of this chapter is to give you an insight on how to approach the modeling
of various structural mechanics problems.

Some physics interfaces and features discussed in this chapter are only available with
certain products. For a detailed overview of the functionality available in each
product, visit https://www.comsol.com/products/specifications/

53
 In this chapter:

• Study Types • Component Mode Synthesis

• Selecting the Physics Interface • Computing Mass Properties
• Selecting Discretization • Effective Properties of Periodic
• Coupling Different Element Types Structures

• Applying Loads • Modeling Pretensioned Bolts

• Defining Constraints • Simplified Modeling of Bolt

Threads
• Calculating Reaction Forces
• Modeling Embedded Structures
• Introduction to Material Models
and Reinforcements
• Modeling Piezoelectric Problems
• Modeling Thin Layers
• Modeling Electrostrictive and
• Modeling Cracks
Ferroelectroelastic Materials
• Buckling Analysis
• Modeling Magnetostrictive
Materials • Performing a Random Vibration
Analysis
• Mechanical Damping and Losses
• Performing a Response Spectrum
• Modeling Geometric Nonlinearity
Analysis
• Contact Modeling
• Stress Linearization
• Activating and Deactivating
• Solver Settings for Structural
Material
Mechanics
• Springs and Dampers
• Using Reduced Integration
• Defining Multiphysics Models
• Result Presentation
• Thermally Coupled Problems
• Part Libraries
• Fluid-Structure Interaction

54 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

Study Types
Introduction
In this section, you will find information about when and how to apply the study types
which are available for structural mechanics problems:

• Stationary Analysis
• Eigenfrequency Analysis
• Mode Analysis
• Time-Domain Analysis
• Frequency-Domain Analysis
• Mode Superposition
• Harmonic Perturbation
• Modal Reduced-Order Models
• Linearized Buckling Analysis
• Bolt Pretension Study
• Random Vibration (PSD) Study
• Response Spectrum Analysis Study

For general information about study types and solvers, see Studies and
Solvers in the COMSOL Multiphysics Reference Manual

Stationary Analysis
You can consider a structural mechanics problem as stationary if the following two
criteria are fulfilled:

• The loads vary so slowly that inertial forces are negligible. Problems of this type are
referred to as quasi static.
• There are no explicit time dependencies in the material model. Viscoelasticity and
creep have such time dependencies.

To perform this type of analysis, you use a Stationary study step.

STUDY TYPES | 55
 In many cases, there is a variation in the load, even though the solution for each value
of the load can be considered as stationary. There are three conceptually different
cases:

• The load values are independent; it is just a number of different load cases you want
to compute. The load case handling functionality described in Load Cases is well
suited for this purpose.
• You want to study a nonlinear problem where the solution is path dependent, or
where the load must be increased in small increments in order to obtain a converged
solution. In this case you should use the parametric continuation solver. Create a
parameter under Global Definitions>Parameters, which you use to control the
variation of the load. Then select Auxiliary sweep under Study Extensions in the
settings for the Stationary solver. In the table for the auxiliary sweep parameters, add
the load controlling parameter, and define its range of variation.
• In a multiphysics problem, another physical quantity might be truly time dependent
but on a time scale that is “slow” from the structural mechanics point of view. This
is usually the case with, for example, problems coupled to heat transfer or diffusion.
If the problem also is such that the structural deformations do not affect the other
physics, it will be unnecessarily expensive to solve also the structural problem in the
time domain, irrespective of whether it is linear or nonlinear. In this situation, you
should first solve the other physics in a time-dependent study and then the structural
mechanics problem in a subsequent stationary study step using the time t as the
parameter in the auxiliary sweep.

For an example of how to set up such a study sequence, see

Fluid-Structure Interaction in a Network of Blood Vessels: Application
Library path Structural_Mechanics_Module/Fluid-Structure_Interaction/
blood_vessel.

CONSTRAINTS
A stationary problem is solvable only if the structure is sufficiently constrained. There
must not be any possible rigid body modes. Thus, no stress-free deformation states are
allowed.

For a more detailed discussion about sufficient constraints, see Rigid

Body Motion.

56 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

Eigenfrequency Analysis
An Eigenfrequency study solves for the eigenfrequencies (natural frequencies) and the
shapes of the corresponding eigenmodes.

When performing an eigenfrequency analysis, you can specify whether to look at the
mathematically more fundamental eigenvalue, , or the eigenfrequency, f, which is
more commonly used in a structural mechanics context. The relation between the two
is


f = – ---------
2i

where i is the imaginary unit.

The undamped eigenvalue problem is commonly written as

2
 K –  M u = 0

where K is the stiffness matrix, M is the mass matrix, u is the eigenmode displacement
vector, and   f is the angular frequency. If damping is present, the eigenvalue
equation is expanded to

2
 K + iC –  M u = 0 (2-1)

where C is the viscous damping matrix, and K can be complex-valued.

Because only the shape and not the size of the modes (eigenvectors) have physical
significance, the computed modes can be scaled arbitrarily. You can select the method
for scaling in the Eigenvalue Solver node of the solver sequence. If Scaling of eigenvectors
is set to Mass matrix, the eigenmodes u are orthogonalized with respect to the mass
matrix M so that

T
u i Mu i = 1 (2-2)

This is a common choice for the scaling of eigenvectors within the structural mechanics
field. The choice of eigenvector scaling does not affect for example the results of a
subsequent mode superposition analysis, but it will affect the interpretation of an
exported modal representation of the system.

STUDY TYPES | 57
 MODAL PARTICIPATION FACTORS
Modal (or mass) participation factors are useful tools when working with the modal
representation of a structure. Through them, you can get the following information:

• The fraction of the total mass of a structure that a certain number of modes
represent is a result. This can be important when judging if a set of modes forms a
good enough base for a mode superposition.
• The main direction of vibration for a certain mode can be seen from the relation
between the participation factors.
• When you have a large set of modes, an examination of the participation factors can
give information about the dominant modes.
To compute modal participation factors, a Participation Factors node must be present
under Definitions in the current component. When you add an Eigenfrequency study
from the Add Study window, such a node is automatically created.

You can also add it manually under Definitions>Physics Utilities. If you do that after an
eigenfrequency study has been run, you need to do an Update Solution in order to get
access to the variables containing the participation factors.

The modal participation factors are available as global variables, and these can for
example be displayed in a table using a Global Evaluation node under Derived Values in
the Results branch. The participation factor results are available as predefined variables
in the Definitions submenu for the component. In Table 2-1, the variables created from
a Participation Factors node is listed (assuming the default tag mpf1).
TABLE 2-1: PARTICIPATION FACTOR VARIABLES

VARIABLE DESCRIPTION

mpf1.mass Total mass

mpf1.CMJ Center of mass, J coordinate
mpf1.mEffLJ Effective mass, translation along J axis
mpf1.mEffRJ Effective mass, rotation around J axis
mpf.pfLJ Participation factor, translation along J axis
mpf.pfRJ Participation factor, rotation around J axis
mpf.pfLnormJ Normalized participation factor, translation along J axis
mpf.pfRnormJ Normalized participation factor, rotation around J axis

The normalized participation factors are those that would be obtained if mass matrix
scaled eigenmodes would have been used.

58 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

If a Participation Factors node is present in the model when an eigenfrequency study is
run, an evaluation group named Participation Factors is automatically generated. It
contains a table with the translational and rotational modal participation factors for all
computed eigenfrequencies.

If you would compute all eigenmodes of a structure, and sum all modal
masses, they will usually not exactly match the total mass of the structure.
The reason is that any mass which is associated with constrained degrees
of freedom is lost. This effect is discretization dependent. The mass lost
is a fraction of the mass of the elements having constrained nodes.

For an example showing how to compute modal participation factors and

modal mass, see In-Plane Framework with Discrete Mass and Mass
Moment of Inertia: Application Library path
Structural_Mechanics_Module/Verification_Examples/
inplane_framework_freq.

In the COMSOL Multiphysics Reference Manual:

• Eigenvalue Solver
• Studies and Solvers
• Postprocessing of Eigenmodes
• Derived Values, Evaluation Groups, and Tables

In the theory chapter of the Structural Mechanics User’s Guide:

• Modal Participation Factors

It is possible to compute eigenfrequencies for structures which are not fully

constrained; this is sometimes referred to as free-free modes. For each possible rigid
body mode, there is one eigenvalue which in theory is zero. The number of possible
rigid body modes for different geometrical dimensions is shown in the table below.
TABLE 2-2: NUMBER OF POSSIBLE RIGID BODY MODES

DIMENSION NUMBER OF RIGID BODY MODES

3D 6 (3 translations + 3 rotations)
2D axisymmetric 1 (Z-direction translation)

STUDY TYPES | 59
 TABLE 2-2: NUMBER OF POSSIBLE RIGID BODY MODES

DIMENSION NUMBER OF RIGID BODY MODES

2D (solid, beam, truss) 3 (2 translations + 1 rotation)

2D (plate) 3 (1 translation + 2 rotations)
The computed rigid body modes will in general not be recognizable as having pure
translation or rotation. Rather, they will contain linear combinations of all the
fundamental rigid body motions.

In a piezoelectric model, one more zero eigenfrequency could appear if you have not
set a reference value for the electric potential.

In practice, the natural frequencies of the rigid body modes are not computed as
exactly zero, but can appear as small numbers which may even be negative or complex.
If rigid body modes are present in the model, then it is important to use a nonzero
value in the Search for eigenfrequencies around text field in the settings for the
Eigenfrequency study step. The value should reflect the order of magnitude of the first
important nonzero eigenfrequency.

For an example showing an eigenfrequency computation in a model

having a rigid body mode, see Eigenfrequency Analysis of a Free
Cylinder: Application Library path Structural_Mechanics_Module/
Verification_Examples/free_cylinder.

DAMPING
If any type of damping is included in the model, an eigenfrequency solution
automatically returns the damped eigenvalues. The eigenfrequencies and, in general,
also the mode shapes are complex in this case. A complex-valued eigenfrequency can
be interpreted so that the real part represents the actual frequency, and the imaginary
part represents the damping. The damping ratio of the corresponding eigenmode can
be defined as

imag   i  imag   i 
 i = -------------------------  -------------------------
i real   i 

where the approximate expression is valid with high accuracy (within 2%) as long as the
damping is less than 0.2.

60 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

In a complex mode shape, there are phase shifts between different parts of the
structure, so that not all points reach the maximum at the same time under free
vibration.

Some damping types will still give real-valued eigenmodes, this is the case for Rayleigh
damping and loss factor damping.

If an eigenfrequency study is performed as a first step of a mode

superposition analysis, then all features that supply damping should be
disabled in this step. This can be done in the Physics and Variables Selection
section in the settings for the Eigenfrequency study step.

The damping is taken into account in the mode superposition studies.

See also the discussion under Mode Superposition.

PRESTRESSED ANALYSIS
In a loaded structure, the natural frequencies may be shifted due to stress stiffening.

To do a prestressed analysis, Include geometric nonlinearity must be selected in the

Eigenfrequency study step. This is automatic when you add the Eigenfrequency,
Prestressed study type.

The prestress loading can include a contact analysis, in which case the subsequent
eigenfrequency analysis provide as linearization around the current contact state.

Prestressed Structures

FREQUENCY DEPENDENT MATERIAL PROPERTIES

If the material data (stiffness or damping) is frequency dependent, the eigenvalue
problem will become nonlinear. This can, for example, occur for some viscoelastic
materials. In this case, the eigenvalue equation Equation 2-1 becomes

2
 K    + iC    –  M u = 0 (2-3)

The eigenvalue solver as such assumes that the matrices involved are constant, so they
must be evaluated at a certain frequency, the linearization point.

2
 K   L  + iC   L  –  M u = 0 (2-4)

STUDY TYPES | 61
 In order to get a correct solution to Equation 2-3, the linearization point L must be
close to the actual eigenvalue . This is in general possible only for one single
eigenfrequency at a time. You must solve this problem either by manual iteration, or
by using some type of scripting, for example through a model method.

Eigenmodes of a Viscoelastic Structural Damper: Application Library

path Structural_Mechanics_Module/Dynamics_and_Vibration/
viscoelastic_damper_eigenmodes.

The built-in viscoelastic materials use a special formulation that do not

generate a nonlinear eigenvalue problem, as long fractional derivatives are
not used.

Mode Analysis
The Mode Analysis study type ( ) is available with the Solid Mechanics interface in
2D plane strain.

Elastic waves can propagate over large distances in structures like rails and pipes, with
a generic name referred to as waveguides. After some distance of propagation in a
waveguide of uniform cross section, such guided waves can be described as a sum of
just a few discrete propagating modes, each with its own shape and phase speed. The
equation governing these modes can be obtained as a spatial Fourier transform of the
linearized time-harmonic equation in the waveguide axial z direction or by inserting
the assumption that the mode is harmonic in space,

u = ue – ikz z

and eliminating all out-of-plane z dependence.

Similar to the full time-harmonic equation, the transformed equation can be solved at
a given frequency with a nonzero excitation for most axial wave numbers kz. But at
certain discrete values the equation breaks down. These values are the propagation
constants or wave numbers of the propagating or evanescent waveguide modes. The

62 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

eigenvalue solver can solve for these propagation constants together with the
corresponding mode shapes.

The propagating wave number is a function of the frequency. The relation

between the two is commonly referred to as a dispersion curve.

The most common use for mode analysis is to define sources for a subsequent
time-harmonic simulation. If there is a component with one or more waveguide
connections, its behavior can be described by simulating its response to the discrete set
of propagating modes on the waveguide opening cross sections.

• Out-of-Plane Waves in the Structural Mechanics Theory Chapter

• Studies and Solvers and Mode Analysis in the COMSOL Multiphysics
Reference Manual

Time-Domain Analysis
There are two classes of problems where a stationary solution cannot be used:

• When the inertial forces no longer are negligible, the full problem as given by
Newton’s first law must be solved.
• When there are time dependencies in the material model, as for creep or
viscoelasticity.

The most general way of handling time-dependent problems is to use a Time Dependent
study. In this type of analysis, you can incorporate any type of nonlinearity, and there
are no limitations on the time dependence of the loads.

A time-domain solution can be preceded by a stationary study, if, for example, prestress
effects are needed.

For a linear problem including inertia, using the mode superposition method is often
much more efficient than using the standard direct method.

SOLVER SELECTION
The two classes of dynamic problems presented above have quite different properties.
The inertial forces in the full structural dynamics problem contain second-order time
derivatives of the displacements, whereas creep and viscoelasticity only have first-order

STUDY TYPES | 63
 derivatives. The physical and numerical properties of these equations differ
significantly.

There are two general solvers for time-dependent problems in COMSOL

Multiphysics.

• The Generalized alpha method, which is recommended for structural dynamics

problems. This is the default solver if Structural Transient Behavior is set to Include
inertial terms in the physics interface settings.
• The BDF method, which is recommended for first-order problems. This is the
default solver if Structural Transient Behavior is set to Quasistatic in the physics
interface settings.

In the COMSOL Multiphysics Reference Manual:

• Time-Dependent Solver
• Studies and Solvers

Frequency-Domain Analysis
In a frequency-domain analysis, you study the response to a harmonic steady state
excitation for certain frequencies. Such a steady state can prevail once all transient
effects have been damped out.

The response must be linear, so that the single frequency harmonic excitation gives a
pure harmonic response with the same frequency. The model may, however, contain
nonlinearities. The harmonic response is computed around a certain linearization
point. In such a case, the frequency-domain analysis can be considered as a very small
perturbation around that linearization point.

Harmonic Perturbation

64 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

All loads and responses in a frequency-domain analysis are in general complex-valued
quantities. If all loads do not have the same phase, you can describe the phase of a
certain load in two ways:

• Add a Phase subnode to the load, in which you directly give the phase angle.
• Enter the load as a complex value, for example as
100[N]*(1+0.3*i)/sqrt(1+0.3^2).

Most results from a frequency domain analysis are complex-valued. In many results
evaluation nodes, the real value of any result quantity will be shown. Assuming that
you want to display for example the displacement in the x direction, u, you have the
following options:

• Plot u or real(u). This gives the displacement at current (default zero) phase angle.
• Plot imag(u). This gives the displacement at a phase angle shifted 90 degrees from
the current value.
• Plot abs(u). This gives the amplitude of the displacement.
• Plot arg(u). This gives the phase angle of the displacement.

The reference phase, with respect to which the results above are reported can be
entered in the settings for the dataset.

STUDY TYPES | 65
 Result quantities that are nonlinear in terms of the displacements, such as principal
stresses, should be interpreted with great care in frequency domain. They will in
general not be harmonic, so the information about amplitude and phase is not reliable.

Some extra variables for postprocessing are created in a frequency-domain

analysis. As an example, in a Solid Mechanics interface with the name
solid, the following variables are defined:

• solid.disp — norm of displacement (at current phase angle)

• solid.vel — norm of velocity (at current phase angle)
• solid.acc — norm of acceleration (at current phase angle)
• solid.disp_rms — RMS displacement over a cycle
• solid.vel_rms — RMS velocity over a cycle
• solid.acc_rms — RMS acceleration over a cycle
• solid.uAmpX — amplitude of displacement in the X direction
• solid.uAmp_tX — amplitude of velocity in the X direction
• solid.uAmp_ttX — amplitude of acceleration in the X direction
• solid.uPhaseX — phase of X displacement, in radians
• solid.uPhase_tX — phase of X velocity, in radians
• solid.uPhase_ttX — phase of X displacement, in radians
• solid.mises — von Mises equivalent stress at current phase angle.
• solid.mises_peak — maximum von Mises equivalent stress over a
cycle.

The components in other coordinate directions are obtained by replacing

X by another coordinate name.

PRESTRESSED ANALYSIS
The shift in the natural frequencies in a prestressed structure may have a significant
effect on the frequency response. This is particularly important when the frequencies
of the load are close to any of the natural frequencies of the structure.

To do a prestressed analysis, Include geometric nonlinearity must be selected in the

Frequency Domain study step. This is automatic when you add the Frequency Domain,
Prestressed study type.

66 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

The prestress loading can include a contact analysis, in which case the subsequent
frequency domain analysis provides a linearization around the current contact state.

• Prestressed Structures
• Harmonic Perturbation

OBTAINING A TIME HISTORY

Sometimes you want to study the time history over a period for the results of a
frequency domain analysis. You can do that by adding a Frequency to Time FFT study
step. The frequency response results are then viewed as terms in a Fourier series, which
can be transformed to time domain. It is possible combine the results for several
frequencies into a single time history, under the assumption that they are all multiples
of the same fundamental frequency.

For examples showing how to obtain a time history from frequency

domain results, see

• Viscoelastic Structural Damper: Application Library path

Structural_Mechanics_Module/Dynamics_and_Vibration/
viscoelastic_damper_frequency.
• Vibration Analysis of a Deep Beam: Application Library path
Structural_Mechanics_Module/Verification_Examples/
vibrating_deep_beam.

Mode Superposition
Analyzing forced dynamic response for large models can be very time-consuming. You
can often improve the performance dramatically by using the mode superposition
technique. The following requirements must be met for a modal solution to be
possible:

• The analysis is linear. It is possible, however, that the structure has been subjected
to a preceding nonlinear history. The modal response can then be a linear
perturbation around that state.
• There are no nonzero prescribed displacements.
• The important frequency content of the load is limited to a range that is small when
compared to all the eigenfrequencies of the model, so that its response can be

STUDY TYPES | 67
 approximated with a small number of eigenmodes. In practice, this excludes wave
and shock type problems.
• If the modal solution is performed in the time domain, all loads must have the same
dependency on the time. This requirement can be relaxed if you use a reduced-order
model to represent the system, rather than using one of the mode superposition
studies.

When using the Structural Mechanics Module, there are three predefined study types
for mode superposition:

• Time Dependent, Modal

• Frequency Domain, Modal
• Frequency Domain, Prestressed, Modal

The two first of these study types consist of two study steps: One step for computing
the eigenfrequencies and one step for the modal response. The last one has three study
steps. Before the eigenfrequency step, you solve a static load case in order to get a
prestress state used in the eigenfrequency computation.

In practice, you have often computed the eigenfrequencies already, and then want to
use them in a mode superposition. In this case, start by adding an empty study, and
then add a Time Dependent, Modal or Frequency Domain, Modal study step to it. After
having added the study step this way, you must point the modal solver to the solution
containing the eigenfrequencies and eigenmodes. You do this by first selecting Show
default solver at the study level, and then selecting the eigenfrequency solution to be
used in the Eigenpairs section of the generated modal solver.

In a mode superposition, the deformation of the structure is represented by a linear

combination of its eigenmodes. The amplitudes of these modes are the degrees of
freedom of the reduced problem. You must select which eigenmodes to include in the
analysis. This choice is usually based on a comparison between the eigenfrequencies of
the structure and the frequency content of the load. As a rule of thumb, select
eigenmodes up to approximately twice the highest frequency of the excitation.

In the mode superposition formulation in COMSOL Multiphysics, the full model is

projected onto the subspace spanned by the eigenmodes. A problem having the
number of degrees equal to the number of included modes is then solved. Note that
this differs from many implementations of mode superposition, where it often is
assumed that the modal equations are totally decoupled.

68 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

An implication of this, is that it is not necessary to assume a certain structure of the
damping matrix. Any type of damping which is allowable in the corresponding analysis
of the full system can also be used in the modal based analysis.

For many common cases, the mode superposition analysis is not sensitive to whether
the eigenmodes were computed using damping or not. The reason is that the
eigenmodes of problems with Rayleigh damping and loss factor damping can be shown
to be identical to those of the undamped problem, so that the projection to the
subspace spanned by the eigenmodes is the same in both cases. For more general
damping, it is however recommended that you suppress all contributions to the
damping during the eigenfrequency step, and thus base the mode superposition on the
solution to the undamped eigenfrequency problem.

FREQUENCY DOMAIN ANALYSIS

All loads are assumed to have a harmonic variation. This is a perturbation type analysis,
so only loads having the Harmonic perturbation property selected are then included in
the analysis.

TIME-DEPENDENT ANALYSIS
Only the factor of the load which is independent of time should be specified in the load
features. The dependency on time is specified as Load factor under the Advanced section
of the modal solver. This factor is then applied to all loads.

• Modal Solver and Studies and Solvers in the COMSOL Multiphysics

Reference Manual
• Mechanical Damping and Losses

For an example showing how to perform mode superposition in time and

frequency domain, see Various Analyses of an Elbow Bracket:
Application Library path Structural_Mechanics_Module/Tutorials/
elbow_bracket.

Harmonic Perturbation
Analyses in the frequency domain assume that the problem your study is linear, at least
with respect to the response to the harmonic excitation. There may be other
nonlinearities, such that the structure has responded nonlinearly to a previous loading.

STUDY TYPES | 69
 This loading could, for example, have caused a large rotations or prestress of a rubber
membrane.

The concept of harmonic perturbation is in COMSOL Multiphysics used for

distinguishing the linear harmonic analysis from a possible prestress analysis. The most
important implication is that if a load has the Harmonic Perturbation selection, it is
applied only in a study that is of the perturbation type. A load without this selection is,
on the other hand, ignored in such a study. In this way two sets of loads can be
distinguished from each other. Technically speaking, the effect of marking a load as
Harmonic Perturbation is that the linper() operator is applied to the value of the load.

The default settings for the different structural mechanics study types in the frequency
domain are summarized in Table 2-3.
TABLE 2-3: DEFAULT PERTURBATION SETTINGS FOR STRUCTURAL MECHANICS STUDY TYPES

STUDY TYPE STUDY STEP PERTURBATION

Frequency Domain Frequency Domain No

Frequency Domain, Prestressed Stationary No
Frequency Domain, Perturbation Yes
Frequency Domain, Modal Eigenfrequency N/A
Frequency Domain, Modal Yes
Frequency Domain, Prestressed, Stationary N/A
Modal Eigenfrequency N/A
Frequency Domain, Modal Yes

Note the following:

• With the default settings you cannot use the same set of loads for a Frequency Domain
and a Frequency Domain, Modal study because only the latter responds to
perturbation loads.
• You can change the behavior of a Frequency Domain study to be of the perturbation
type by modifying the solver sequence. In the General section of the settings for the
Stationary Solver, change Linearity to Linear perturbation.
• A solver that does not have Linearity set to either Linear perturbation or Linear may
respond to nonlinear effects. There are multiphysics problems where this is wanted
because there may be a nonlinearity in another physics, even though the harmonic
solution within structural mechanics is linear. But if there are nonlinearities within

70 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

 the structural mechanics parts of the model, you must be careful with the solver
settings.

In a Frequency-Domain, Perturbation study step, that is when Linearity is set

to Linear perturbation, geometric nonlinearity will be incorporated in the
sense that there is a split between the material and spatial frames. This
makes it possible to take for example stiffness from follower loads into
account, and to use a contact solution as linearization point.

This frame split was introduced in version 5.3. As an effect, models

created in an earlier version, in which some expressions have a frame
dependency, may produce results that differ from before. Examples of
such cases are:

• Pressure loads
• Loads defined in coordinate systems with deformation dependent axis
orientation
• User-defined expressions containing spatial (“lowercase”) coordinates

In the COMSOL Multiphysics Reference Manual:

• Frequency Domain Perturbation Study Step

• Harmonic Perturbation — Exclusive and Contributing Nodes
• Built-In Operators (linper() operator)

For most load types, the use of Harmonic Perturbation is straightforward, but some
cases need a more detailed discussion:

• A Rigid Connector can be assigned a Harmonic Perturbation subnode in which you can
prescribe harmonic perturbation values to constrained degrees of freedom. If you
have added Applied Force or Applied Moment nodes under a Rigid Connector, you can
independently assign Harmonic Perturbation to these nodes, so that the loads are
considered as being of the perturbation type.

STUDY TYPES | 71
 • Even though initial stresses and strains are not usually considered as loads, you can
assign Harmonic Perturbation also to the Initial Stress and Strain nodes.
• Rather assigning Harmonic Perturbation to a load, you can write the load value
enclosed in the linper() operator. This is particularly useful when the feature that
provide the loading does not have the Harmonic Perturbation option.

For an example showing how to use harmonic perturbation, see Bracket

— Frequency-Response Analysis: Application Library path
Structural_Mechanics_Module/Tutorials/bracket_frequency.

Modal Reduced-Order Models

Reduced-order modeling seeks to reduce the number of degrees of freedom in a
physical model whilst still retaining the essential physics. For a lightly damped resonant
system driven at one of its resonant frequencies, it is reasonable to consider only the
contributions to the system of a small number (m) of modes within the signal
bandwidth. In some cases, a single mode is sufficient. A system with n degrees of
freedom has mass, stiffness, and damping matrices of size n-by-n. A reduced-order
representation of the system considering m modes has size m-by-m. The reduction in
complexity of the system, and the computational speed up is therefore significant when
m « n. This section describes the theory of the reduced-order system and gives
guidelines on how to obtain reduced-order models from a COMSOL model.

This can be employed in two different ways: Either you can use the built-in modal
solvers for the time or frequency domain, or you can export the small equivalent system
and analyze it outside COMSOL Multiphysics, for example, as a component in a larger
system simulation.

THE MODAL COORDINATE SYSTEM

Consider a mechanical system, with n degrees of freedom, described by an equation of
the form

ꞏꞏ ꞏ
Mu + Du + Ku = F (2-5)

where u is the displacement vector (size: n-by-1), K is the stiffness matrix (size:
n-by-n), D is the damping matrix (size: n-by-n), and M is the mass matrix (size:
n-by-n). In the frequency domain the problem takes the form

72 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

 2
–  M u 0 + iDu 0 + Ku 0 = F

where u = u0eit.

Initially consider the system in the absence of damping and forces. The undamped
system has n eigenvalues i, which satisfy the equation

ˆ =  2 Mu
Ku ˆ (2-6)
i i i

These eigenvectors can be shown to be orthogonal with respect to both M and K:

ˆ T Mu
u ˆ = 0 i  j , i  j (2-7)
j i

ˆ T Ku
u ˆ = 0 i  j , i  j (2-8)
j i

Next the following n-by-n matrix is constructed, with columns taken from the n
eigenvectors:

ˆ ,u
U = u ˆ u
ˆ
1 2 n

Then consider the following matrix:

u ˆ u
ˆ T Mu ˆT ˆ
1 1 1 Mu 2
ˆ T Mu
u ˆ u ˆT ˆ
2 1 2 Mu 2
T
U MU =
ˆ T Mu
u ˆ ˆT ˆ
n–1 n – 1 u n – 1 Mu n
ˆ T Mu
u ˆ ˆ T Mu
u ˆ
n n–1 n n

From Equation 2-7 it is clear that this is a diagonal matrix. Similarly, from
Equation 2-8 it is clear that UTKU is also diagonal.

From the properties of the eigenvectors it is possible to expand any function in terms
of the eigenvectors. Thus, the displacement u can be written as:

n
ˆ
u =  ai ui
i=1

This equation can also be expressed in the form:

STUDY TYPES | 73
 u = Ua (2-9)

where a is a column vector containing the coefficients ai as rows. In general a is

time dependent.

Now consider the original equation: Equation 2-5. First substitute for u using
Equation 2-9. Then transform the equation to the modal coordinate system by
premultiplying by UT. This gives:

T ꞏꞏ T ꞏ T T
U MUa + U DUa + U KUa = U F (2-10)

It has already been established that the matrices UTMU and UTKU are diagonal and
frequently a damping model is chosen that results in a diagonal damping matrix. For
example, in Rayleigh damping D  M + K, where  and  are constants. For a
general damping, the transformed damping matrix is however not diagonal. As an
alternative, a damping ratio, i, can be assigned to each mode.

EIGENVALUE SCALING
The precise form of Equation 2-10 is determined by the normalization adopted for the
eigenfunctions. In structural applications the eigenfunctions are often normalized such
that UTMU  I. This is referred to as mass matrix scaling in the eigenvalue solver. In
this case Equation 2-6 gives

ˆ T Ku
u ˆ =  2u
ˆ T Mu
ˆ = 2
i i i i i i

so that

T 2
U KU = diag   i 

where diag(i2) is the diagonal matrix with diagonal elements i2. Similarly, if
damping ratios for each mode are defined, the damping matrix can be expressed in the
form

T
U DU = diag  2 i  i 

Thus, if mass matrix scaling is used Equation 2-10 takes the form

ꞏꞏ ꞏ 2 T
a + diag  2 i  i a + diag   i a = U F (2-11)

It is also possible to scale the eigenvectors so that the point of maximum displacement
has given displacement. This is referred to as max scaling in the eigenvalue solver. For

74 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

an individual mode this scaling has a simple physical interpretation — the
corresponding component of a, ai, is the amplitude of the i:th mode, measured at the
point of maximum displacement, when the mode is driven by the force F. In this case
Equation 2-10 takes the form

ꞏꞏ ꞏ T
diag  m eff , i a + diag  c eff , i a + diag  k eff , i a = U F (2-12)

Here meff,i is the effective mass of the i:th mode, ceff,i = 2meff,iii is the effective
damping parameter for the mode, and keff,i is the effective spring constant. Each
element of the vector UTF gives the force component that acts on each of the
respective modes.

When using max scaling, it is the largest value of a degree of freedom

which is scaled to 1. The total displacement in that node will thus be
between 1 and 3 times the given scaling factor.

If degrees of freedom other than displacements are active in the

eigenfrequency problem, the maximum value may occur in another type
of degree of freedom such as electric potential or pressure. Consequently,
the peak displacement in that mode can then be less than the scaling
factor.

REDUCED-ORDER MODELS
The preceding discussion did not consider how to reduce the number of degrees of
freedom in the system. For systems in which the vector UTF has only a few significant
components (for example, components i = 1, …, m where m « n) the following
approximation can be made:

m
ˆ
u  ai ui
i=1

The expression for u in matrix becomes:

u = U'a'

where U' is now an m-by-n and a' is a vector with m components. The equation
system in modal coordinates now takes the form

T ꞏꞏ T ꞏ T T
U' MU'a' + U' DU'a' + U' KU'a' = U' F (2-13)

STUDY TYPES | 75
 The matrices U'TMU', U'TDU', and U'TKU' now have dimensions m-by-m.
Similarly, the vector U'TF has m components. This results in a significant reduction in
the system complexity.

REDUCED-ORDER MODELS WITH PHYSICAL DAMPING

If physically relevant damping is present in the system, the above theory must be
modified as the damping matrix is no longer diagonal in the modal coordinate system.
COMSOL can still handle this case as the modal solver does not assume that any of the
matrices are diagonal. In this case the eigenvalues become complex and the
eigenvectors split into right and left eigenvectors. The right eigenvectors Ur are
solutions of the equation:

2 ˆ + i Du
–i M u ˆ + Kuˆ
r ,i i r ,i r ,i = F

As in the previous section, for a reduced set of modes, it is assumed that:

u = U' r a'

where U'r is the n-by-m matrix containing the right eigenvectors chosen for the modal
analysis. Once again a' is a vector with m components. The system in modal
coordinates takes the form

T ꞏꞏ T ꞏ T T
U' l MU' r a' + U' l DU' r a' + U' l KU' r a' = U' l F

where U'l is the n-by-m matrix containing the left eigenvectors chosen for the modal
analysis.

The matrices U'lTMU'r, U'lTDU'r, and U'lTKU'r are no longer necessarily diagonal.
The modal solver accepts any linearly independent set of vectors to project the solution
vector and equations onto and constructs the reduced-order system accordingly.

ACCESSING THE REDUCED-ORDER MODEL MATRICES

The Model Reduction and Modal Reduced-Order Model study steps have the property that
they can assemble the modal matrices and make them available for output. In the Model
Reduction node, the Store reduced matrices check box must be selected.

After the model has solved, right-click the Results>Derived Values node and select
System Matrices. In the output section choose the Matrix to display in the list. The mass
matrix corresponds to the matrix U'lTMU'r the stiffness matrix corresponds to
U'lTKU'r, and the damping matrix corresponds to U'lTDU'r. The vector U'lTF is
available as the load vector. These matrices are given in a format that respects the

76 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

normalization of the preceding Eigenvalue Solver. To change this, select the Eigenvalue
Solver node, and change the Scaling of Eigenvectors setting under the Output section.
Use the Max setting if an equivalent Mass-Spring-Damper system is required, in which
case the modal amplitude corresponds to the maximum displacement of the mode.

Linearized Buckling Analysis

A linearized buckling analysis can be used to estimate the critical load at which a
structure becomes unstable. This is a predefined study type that consists of two study
steps: An initial step in which a unit load is applied to the structure, and a second step
in which an eigenvalue problem is solved for the critical buckling load.

COMSOL reports a critical load factor, ,which is the multiplier to the initial load at
which the structure becomes unstable. The corresponding eigenmode is the shape of
the structure in its buckled state.

The level of the initial load used is immaterial since a linear problem is solved. If the
initial load actually was larger than the buckling load, then the critical value of  is
smaller than 1. It is also possible that the computed value of  is negative. This signifies
that a reversed load will give the critical case.

When performing a linearized buckling study, it is possible to discriminate between live

and dead loads, where the former are the ones with respect to which the critical load
factor is computed, and the latter are assumed to be constant. In this case, two
different basic load cases need to be solved before the eigenvalue solution.

The buckling computed buckling modes can be used to provide an initial imperfection
for a subsequent nonlinear buckling analysis,

• For more details about how to model buckling, see Buckling Analysis.
• The numerical formulation is described in the section Linear Buckling
in the theory chapter.
• Settings for the solvers are described in Studies and Solvers, Linear
Buckling, and Buckling Imperfection in the COMSOL Multiphysics
Reference Manual.

STUDY TYPES | 77
 • Bracket — Linear Buckling Analysis: Application Library path
Structural_Mechanics_Module/Tutorials/bracket_linear_buckling
• Buckling Analysis of a Truss Tower: Application Library path
Structural_Mechanics_Module/Buckling_and_Wrinkling/
truss_tower_buckling

Bolt Pretension Study

The Bolt Pretension study step is a special case of a Stationary study step, where the
special degrees of freedoms used for modeling prestressed bolts are solved for. In all
other study types, these degrees of freedom are inactive. Typically, you include a Bolt
Pretension study step as the first step in a study in order to simulate the state after the
assembly of a bolted joint. You can then add any other types of study steps for
computing the effects of the service loads.

In a Bolt Pretension study step, it is possible to tension the bolts sequentially, or to

change the bolt prestress more than once, if an auxiliary sweep over some history
parameter is used.

Modeling Pretensioned Bolts

• Modeling of Pretensioned Bolts: Application Library path

Structural_Mechanics_Module/Tutorials/bolt_pretension_tutorial
• Prestressed Bolts in a Tube Connection: Application Library path
Structural_Mechanics_Module/Contact_and_Friction/tube_connection

Random Vibration (PSD) Study

A random vibration study is used to analyze the response of a structure subjected to
loads that randomly vary in time, but their statistical properties do not change with
time.

The input to a random vibration analysis is given in terms of power spectral densities
(PSD) and, in the case of several loading sources, the load cross-correlations.

78 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

The results of this type of analysis can only be interpreted in a statistical sense. Usually,
the root mean square (RMS) of the result quantities is studied.

The analysis is based on a mode superposition and the reduced-order model (ROM)
functionality. Except from the computation of eigenfrequencies and corresponding
eigenmodes, and the creation of the reduced model, the core of the computation is
performed during result evaluation.

The Random Vibration (PSD) study is mainly an entry point when adding studies. When
you select it, you actually get two studies and a number of nodes under Global
Definitions added to the model.

• See Performing a Random Vibration Analysis for a detailed description

of how to study random vibration.
• The theory is described in Random Vibration Theory.

• Bracket — Random Vibration Analysis: Application Library path

Structural_Mechanics_Module/Tutorials/bracket_random_vibration
• Random Vibration Test of a Motherboard: Application Library path
Structural_Mechanics_Module/Dynamics_and_Vibration/
motherboard_random_vibration
• Random Vibration Analysis of a Deep Beam: Application Library
path Structural_Mechanics_Module/Verification_Examples/
random_vibration_deep_beam

Response Spectrum Analysis Study

Response spectrum analysis is used for computing an approximation of the structural
response to transient, nondeterministic events, such as earthquakes or shocks.

The Response Spectrum Analysis study is mainly an entry point when adding studies.
What you actually get when you add such study is an Eigenfrequency study step together
with a Response Spectrum node under Definitions.

The actual response is computed on demand during result evaluation, using the
computed eigenfrequencies and modes. The settings for the evaluation are done in the
Response Spectrum 2D and Response Spectrum 3D datasets.

STUDY TYPES | 79
 If your response spectrum evaluation requires inclusion of missing mass correction,
you need also to compute a set of stationary load cases. To set up that analysis, use the
Create missing mass correction study button ( ) on the header of the Response
Spectrum section the in the Response Spectrum node settings.

• See Performing a Response Spectrum Analysis for a detailed

description of how to work with response spectrum evaluations.
• The theory is described in Response Spectrum Analysis Theory.
• The settings for the special datasets are described in Response
Spectrum 2D and Response Spectrum 3D in the COMSOL
Multiphysics Reference Manual.
• The Response Spectrum node is described in the COMSOL
Multiphysics Reference Manual.

• Earthquake Analysis of a Building: Application Library path

Structural_Mechanics_Module/Dynamics_and_Vibration/
building_response_spectrum
• Shock Response of a Motherboard: Application Library path
Structural_Mechanics_Module/Dynamics_and_Vibration/
motherboard_shock_response

80 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

Selecting the Physics Interface
The structural mechanics products contain a number of physics interfaces for a wide
range of applications. This section contains some guidelines for how to select an
appropriate physics interface for your analysis.

The basic physics interfaces for structural mechanics are:

• Solid Mechanics — General modeling of solid objects with extension in all

directions.
• Elastic Waves, Time Explicit — For efficient analysis of linear elastic wave
propagation in solid domain.
• Shell and Plate — For objects which are thin in one direction, but have significant
bending stiffness
• Membrane — For objects which are thin in one direction, and have negligible
bending stiffness
• Beam — For objects where two directions have significantly smaller dimensions
than the third; significant bending stiffness
• Pipe Mechanics — For analysis of pipes with internal pressure. This interface is
similar to the Beam interface, but specialized for analysis of pipes.
• Truss — For objects where two directions have significantly smaller dimensions than
the third; only axial forces can be transmitted
• Wire — Similar to Truss, but only tensile forces can be transmitted

For a detailed overview of the functionality available in each product, visit

https://www.comsol.com/products/specifications/

Solid Mechanics
The Solid Mechanics interface offers the most general modeling of structural
mechanics problems and is based on general principles of continuum mechanics. It is
the interface which contains the largest number of material models, and the most
advanced boundary conditions.

SELECTING THE PHYSICS INTERFACE | 81

 The drawback with using solid elements is that the models can become
computationally expensive, especially in 3D. For structures which are thin or slender,
you should consider using one of the specialized physics interfaces.

3D SOLID GEOMETRY
The degrees of freedom (dependent variables) in 3D are the global displacements u, v,
and w in the global x, y, and z directions, respectively.

Figure 2-1: Loads and constraints applied to a 3D solid using the Solid Mechanics
interface.

2D GEOMETRY

Plane Stress
The plane stress variant of the 2D physics interface is useful for analyzing thin in-plane
loaded plates. For a state of plane stress, the out-of-plane components of the stress
tensor are zero.

Figure 2-2: Plane stress is used to model plates where the loads are only in the plane; it does
not include any out-of-plane stress components.

The 2D physics interface for plane stress allows loads in the x and y directions, and
assumes that these are constant throughout the material’s thickness, which can vary
with x and y. The plane stress condition prevails in a thin (compared to the in-plane

82 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

dimensions) flat plate in the xy-plane loaded only in its own plane and without any
z direction restraint.

Plane Strain
The plane strain variant of the 2D physics interface that assumes that all out-of-plane
strain components of the total strain tensor z, yz, and xz are zero.

Figure 2-3: A geometry suitable for plane strain analysis.

Loads in the x and y directions are allowed. The loads are assumed to be constant
throughout the thickness of the material, but the thickness can vary with x and y.
Formally, the plane strain condition requires that the ends of the object are constrained
in the z direction, but it is often also used for central parts of an object that is long in
the z direction (compared to the in-plane dimensions). One example is a long tunnel
along the z-axis where it is sufficient to study a unit-depth slice in the xy-plane.

Generalized Plane Strain

Generalized plane strain is similar to plane strain in the sense that transverse stresses
can develop in the 2D cross section of a long object. The requirement that the
out-of-plane strain is zero, is however relaxed. Instead, an assumption about zero
resulting force in the transverse direction is used. Optionally, assumptions about zero
bending moments over the cross section can be added. Generalized plane strain
conditions prevail in the in the inner parts of a long object with free ends. For many
cases, generalized plane strain conditions is the 2D approximation that is closest to a
full 3D solution.

For an in-depth discussion of different aspects of 2D solid mechanics, see

https://www.comsol.com/blogs/
what-is-the-difference-between-plane-stress-and-plane-strain

SELECTING THE PHYSICS INTERFACE | 83

 AXISYMMETRIC GEOMETRY
The axisymmetric variant of the Solid Mechanics interface uses cylindrical coordinates
r,  (phi), and z. All properties are independent of the azimuthal angle  .

In the default version of the interface, displacements occur only in the r-z plane, and
there are two degrees of freedom, u and w. By selecting the Include circumferential
displacement option, you can model also torsion around the axis of rotational
symmetry. The azimuthal rotation degree of freedom v is then included. In addition,
many features, such as load features, allow values to be entered in the  direction.

The 2D axisymmetric geometry is viewed as the intersection between the original

axially symmetric 3D solid and the half plane   0, r  0. Therefore, the geometry is
drawn only in the half plane r  0, and it recovers the original 3D solid by rotating the
2D geometry about the z-axis.

Figure 2-4: Rotating a 2D geometry to recover a 3D solid.

• Axisymmetric Twist and Bending: Application Library path

Structural_Mechanics_Module/Verification_Examples/
axisymmetric_twist_and_bending

Elastic Waves, Time Explicit

The Elastic Waves, Time Explicit is used to compute the velocity and strain field in
solids with propagating linear elastic waves. In the Add Physics dialog, it is found in
the Acoustics branch, under Elastic Waves.

In general, the interface is suited for modeling the propagation of waves over large
distances relative to the wavelength, for example, ultrasound propagation for
nondestructive testing (NDT), or seismic waves. The interface exists in 2D
(generalized plane strain) and 3D.

84 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

The interface is based on the discontinuous Galerkin (dG or dG-FEM) method and
uses a time explicit solver. The method is very memory efficient and can solve problems
with many million degrees of freedom (DOFs). The method is also well suited for
distributed computing on clusters.

In the Acoustics Module User’s Guide:

• The Elastic Waves, Time Explicit Interface

• Theory for the Elastic Waves, Time Explicit Interface

Shell and Plate

The Shell interface is useful when the object is thin in one direction. Structures built
from welded or bolted flat plates are archetypal shell structures and so are pressure
vessels. The Shell interface is available in 3D and 2D axisymmetry.

The Plate interface is a specialization of the Shell interface, used for 2D modeling in
the XY-plane. A plate model has its main action in bending out of the plane, but can
optionally also treat in-plane forces. If the loads act only in the plane, using Solid
Mechanics with the Plane Stress option is a better choice.

Shells are modeled on boundaries, and the transverse direction is represented only by
the mathematical model. The degrees of freedom consist of displacements and
rotations at the modeled boundary. This results in an assumption where the in-plane
strains vary linearly through the thickness, and the stress in the thickness direction is
zero. The thickness of a shell does not have to be constant, although this is by far the
most common case.

When the Composite Materials Module is available, it is possible to model also

multilayered shells.

For nonlinear material models, a layered approach with a single layer is used. There is
a virtual mesh in the thickness direction, in order to accommodate the potential
variation of the material properties in the thickness direction.

Rather than computing the shell stiffness from material properties and thickness, you
can also directly enter that stiffness properties in tension, bending, and shear.

The Shell and Plate interfaces can be used both for “thin” and “thick” shells. Shear
deformations are taken into account; this is usually called Mindlin theory. The material
model is linear elastic.

SELECTING THE PHYSICS INTERFACE | 85

 When modeling with shells, it is important to keep track of “top” and “bottom” side
when applying loads and interpreting the results.

The in-plane stiffness of an elastic shell is proportional to the thickness h, while the
bending stiffness is proportional to h3. The difference in stiffness along different
directions can thus become very large. When an object is very thin, a shell model may
be numerically ill-posed due to the negligible bending stiffness. It is then better to use
the Membrane interface.

Membrane
The Membrane interface can be used for very thin objects, like cloth, where only
in-plane forces are important. Membranes can be considered as plane stress elements
but with an arbitrary, possibly curved configuration in space. The Membrane interface
is available in 3D and 2D axisymmetry.

In most applications, a membrane must be pretensioned in order to have a stable

configuration, so it will almost invariably be used in a geometrically nonlinear analysis.
The main exception is when it is used as a “cladding” on top of a solid, since it will
then be stabilized by the solid.

You can study configurations when there is local wrinkling in a membrane by adding
a special nonlinear material model.

In the Membrane interface a large number of different material models can be used.
When the Composite Materials Module is available, it is possible to model also
multilayered membranes.

Beam
A beam is an abstract model where only the extension in the axial direction is modeled
explicitly on an edge. The cross section is usually specified in terms of geometrical
properties such as area and moments of inertia. Several predefined cross-section types
are also available. The Beam interface is available in 3D and 2D.

The exact stress distribution in the beam is not explicitly modeled. It is actually not
even fully determined by the cross-sectional properties. Instead, six (in 3D) resultant
section forces are used: axial force, shear forces in two perpendicular directions, two
bending moments, and one twisting moment.

The Beam interface assumes linear elasticity.

86 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

Two formulations are available in the Beam interface:

• The classical Euler-Bernoulli beam theory, which is applicable for slender beams.
• Timoshenko theory, where also shear deformations are considered. This theory
makes it possible to use the Beam interface to model rather thick beams.

Pipe Mechanics
Pipes are similar to beams, and many properties of the Pipe Mechanics interface are
shared with the Beam interface. The most distinguishing feature is that the internal
pressure usually causes a significant part of the stresses in a pipe. Also, temperature
gradients usually occur through the pipe wall, rather than across the entire section.

The loads from internal pressure and drag forces can be taken directly from results in
the Pipe Flow interface. Similarly, the temperature in the pipe walls can be taken from
the Heat Transfer in Pipes interface.

In the Pipe Flow Module User’s Guide:

• The Pipe Flow Interface

• The Heat Transfer in Pipes Interface

Truss
The Truss interface has two main purposes:

• Modeling of trusses, consisting of straight bars carrying only axial forces

• As reinforcements, used in conjunction with other physics interfaces

The Truss interface is available in 3D and 2D.

For a truss model, only one geometrical property is needed, the cross-section area. The
material model can be linear elastic, elastoplastic, or a shape memory alloy. There is also
a special material model for creating spring/damper data.

The truss element has no stiffness in the directions perpendicular to its extension. For
trusses, this is usually not a problem since they are designed such that each member is
stabilized by its neighbors.

SELECTING THE PHYSICS INTERFACE | 87

 Wire
The Wire interface intended for modeling of cables and wires. The structures are often
prestressed, but also cables sagging under, for example, self-weight can be modeled.

The main difference between the Truss interface and the Wire interface is that the
wires cannot sustain any compressive forces. In reality the wire will wrinkle.
Numerically, this is handled by using a very low stiffness in compression.

88 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

Selecting Discretization
Shape Function Order
In structural mechanics analysis, the focus is often on the stresses and strains rather
than on the displacements. Since the strains are derivatives of the displacement field,
the accuracy of the strains will be one order less than the accuracy of the displacements.
For this reason, second-order shape functions are used as default in most of the
structural mechanics interfaces. Often this gives the best tradeoff between model size
and accuracy.

It is well known that using first-order shape functions in solid mechanics will give an
overly stiff solution, unless a very fine mesh is used. This is especially noticeable for
triangular and tetrahedral elements. This can, to some extent, be counteracted by using
reduced integration, see Using Reduced Integration.

If the purpose of the analysis is only to compute stiffness, rather than stresses, the use
of linear shape functions can still be justified. This is the default choice in the
Multibody Dynamics interface, available with the Multibody Dynamics Module.

If the solution contains discontinuities, for example when some type of front is moving
through the material, first-order elements and a fine mesh is often a good choice, since
the advantage of the higher-order elements lies in their ability to represent smooth
gradients.

TRUSS ELEMENTS
In the Truss interface the default is to use first-order shape functions, since the
elements are mainly used in a context where the axial force in each element is constant.

When truss elements share an edge with other structural elements, you should choose
the same discretization in both interfaces, usually quadratic.

BEAM ELEMENTS
The beam elements have only one set of shape functions, which cannot be changed.
The axial displacement and the twist are represented by first-order shape functions,
while the bending is represented by cubic Hermitian shape functions. This element can
then represent a constant axial force, a constant twisting torque, a linear bending
moment, and a constant shear force. This is the exact solution for a beam having no
distributed loads.

SELECTING DISCRETIZATION | 89
 A consequence of this formulation is that it may not possible to obtain a perfectly
conforming approximation if a beam shares an edge with elements from another
physics interface.

Lagrange and Serendipity Shape Functions

In the Solid Mechanics and Membrane interfaces, you can choose between two
families of shape functions: Lagrange and serendipity. The default is to use serendipity
shape functions.

The serendipity elements have the advantage of generating significantly fewer degrees
of freedom for structured meshes. The accuracy is in most cases almost as good as for
the Lagrange elements. The Lagrange elements are however less sensitive to strong
mesh distortions.

The serendipity shape functions differ from the Lagrange shape functions only for the
following element shapes:

• 2D: Quadrilateral elements of discretization order higher than 1

• 3D: Hexahedral, prism, and pyramid elements of discretization order higher than 1

In the COMSOL Multiphysics Reference Manual:

• The Lagrange Element (shlag)

• The Nodal Serendipity Element (shnserp)

When coupling two structural mechanics physics interfaces, the same type of shape
functions should be used in both interfaces to ensure conformity in displacement shape
functions. Since there is no difference between the two families of shape functions in
1D, this is not an issue when connecting edges.

Choosing Shape Functions in Multiphysics Models

In problems where several physics fields participate, the accuracy can sometimes be
improved by considering how the different fields interact. In structural mechanics, it
is common that other physics fields directly affect the inelastic strains. This is the case
in, for example, thermal expansion and hygroscopic swelling.

90 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

In thermal expansion, the elastic strain used in most constitutive relations is the
difference between the total strain, which is computed from derivatives of the
displacement field, and the thermal strain:

 el =  tot –  th =  tot –   T – T ref 

Since the thermal strain is directly proportional to the temperature, a consistent

approximation would be to use one step lower discretization order for the temperature
than for the displacements. When using built-in couplings, such as Thermal Expansion,
such a modification is not necessary. Any mismatch is automatically taken care of by
re-interpolating the thermal strains to an order that matches the general strain field.

Another type of coupling appears on the boundary between two domains having
different physics, as in fluid-structure interaction and acoustic-structure interaction.
When, for example, Thermoviscous Acoustics is coupled to Solid Mechanics, then the
time derivative of the displacement in the solid is set equal to the velocity in the
acoustic medium on the shared boundary. In this case, it makes sense to have the same
shape function order for these two fields.

Implicit Shape Function Orders

Some solid mechanics formulations contain other degrees of freedom in addition to
the displacements. The shape functions are then selected internally based on your
choice of displacement discretization.

VISCOELASTIC STRESS AND STRAIN

When using a viscoelastic material model, auxiliary degrees of freedom are added either
for the viscoelastic strains or the viscoelastic stresses, depending on whether a linear or
a nonlinear formulation is used. These degrees of freedom are local to the element, and
you can select either the discontinuous Lagrange or Gauss point data type.

The discontinuous Lagrange shape functions will have an order that is one below what
is used for the displacement shape functions.

If Gauss point data is used, the same integration points as used for the numerical
integration of the stiffness matrix are used. This order depends on the selected
displacement discretization order and whether reduced integration is used or not.

INELASTIC STRAINS
For material models like plasticity and creep, the inelastic strains are formally degrees
of freedom. They will be allocated at the same integration points as used for the

SELECTING DISCRETIZATION | 91
 numerical integration of the stiffness matrix. This order depends on the selected
displacement discretization order and whether reduced integration is used or not.

92 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

Coupling Different Element Types
In this section:

• Introduction to the Element Types

• Coupling Techniques

Introduction to the Element Types

In some engineering structures, the optimal idealization is a mixture of different
element types. Some examples are:

• Structures that are thin in large regions, but more three-dimensional at certain
locations. A mixture of solids and shells can then significantly reduce the model size.
• Plates or shells having beams as stiffeners.
• Truss elements acting as reinforcement bars in a concrete structure.
• A thin layer of one material on top of another material. In this case, an idealization
with shells or membranes covering the boundary of a solid can be useful.

When several physics interfaces are added in COMSOL Multiphysics, the default is
always that each physics interface has individual and unique degrees of freedom. In
structural mechanics, the first physics interface has the displacement variables (u, v, w),
then the second physics interface has (u2, v2, w2), and so on. This means that the
physics interfaces initially are independent, even when defined on the same geometrical
part. To get the intended interaction requires that a coupling is established between
the physics interfaces.

Various methods to couple different element types are discussed in this section.

Coupling Techniques
The following basic techniques to connect physics interfaces with displacement
degrees of freedom are discussed in this section:

• Renaming Degrees of Freedom

• Using Customized Coupling Features
• Using General Coupling Operators

COUPLING DIFFERENT ELEMENT TYPES | 93

 RENAMING DEGREES OF FREEDOM
The easiest coupling method is to rename the displacement degrees of freedom so that
these are the same for all physics interfaces. This is sufficient, for example, when using
membranes as cladding on a solid boundary or truss elements as reinforcement bars in
a solid.

In the Beam, Pipe Mechanics, Shell, and Plate interfaces, the deformation is described
also by rotational degrees of freedom. In the general case, these degrees of freedom
interact with the translational degrees of freedom in a connection.

In some special cases — for example, when a thin shell acts as cladding on a solid — it
is sufficient to make the degree of freedom names for the displacements common; the
rotational degrees of freedom are not important. If, however, a shell edge is connected
to a solid, it acts as a “hinge”, which in most cases is not the intended behavior. You
then need to use the more sophisticated techniques described next.

The default shape functions in the Solid Mechanics interface are of the
serendipity type, whereas Lagrange shape functions are used in the Shell
interface. If you are placing a shell element on the boundary of a solid
element, you must select Lagrange shape functions also in the Solid
Mechanics interface so that the two physics interfaces share the same node
points.

The shape functions used in the Beam and Pipe Mechanics interfaces have
special properties, and a beam cannot have the same degrees of freedom
as another physics interface if the same edge is shared.

Also, the representation of rotations differs between the Shell and Plate
interfaces (displacement of normal) and the Beam and Pipe Mechanics
interfaces (rotation angle). It is therefore not possible to use common
degree of freedom names for the rotational degrees of freedom.

USING CUSTOMIZED COUPLING FEATURES

There are a number of built-in couplings, by which you can add connections that are
nontrivial to set up manually:

• Shell Edge to Solid Boundary (3D)

• Shell Boundary to Solid Boundary (3D)
• Membrane Boundary to Solid Boundary (3D)

94 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

• Beam Point to Solid Boundary (2D)
• Beam Point to Solid (3D)
• Beam Edge to Solid Boundary (2D)
• Beam Edge to Solid Boundary (3D)
• Beam Edge to Shell Edge (3D)
• Beam Point to Shell Boundary (3D)
• Beam Point to Shell Edge (3D)
• Pipe Point to Solid Boundary (3D)
• Pipe Point to Shell Edge (3D)
• Embedded Reinforcement

Shell Edge to Solid Boundary (3D)

A shell can be coupled to a solid by adding a Solid-Thin Structure Connection
multiphysics coupling. In the settings, set Connection type to Solid boundaries to shell
edges. This situation typically occurs when you want to make a transition from a thin
region to one that is thicker. Usually, shell assumptions should be valid on both sides
of the transition. The solid geometry is expected to have the same thickness as the
thickness given in the Shell interface.

You can choose between two different formulations, by setting Method to either Rigid
or Flexible. The flexible version is significantly more accurate locally at the connected
solid boundary, but it comes with a cost in terms of some extra degrees of freedom.
Also, this method requires a large enough number of degrees of freedom in the
thickness direction of the solid. For second-order elements, typically three elements are
required.

Shell Boundary to Solid Boundary (3D)

A shell can also be coupled to a solid by adding a Solid-Thin Structure Connection
multiphysics coupling with Connection type set to Shared boundaries or Parallel
boundaries. This connection is used to add a shell on top of a solid as a ‘cladding’. It is
possible to include an offset distance. The boundaries may be coincident or parallel.

Membrane Boundary to Solid Boundary (3D)

A membrane can be coupled to a solid by adding a Solid-Thin Structure Connection
multiphysics coupling with Connection type set to Shared boundaries. When the thin
structure is a membrane, this is the only available connection type. It is used to add a
membrane on top of a solid as a ‘cladding’.

COUPLING DIFFERENT ELEMENT TYPES | 95

 Beam Point to Solid Boundary (2D)
A beam in 2D can be coupled to a solid by adding a Solid-Beam Connection
multiphysics coupling. In the settings, set Connection type to Solid boundaries to beam
points. This connection is intended for modeling a transition from a beam to a solid,
where beam assumptions are valid on both sides of the connection.

You can choose between two different formulations, by setting Method to either Rigid
or Flexible. The flexible version is significantly more accurate locally at the connected
solid boundary, but it comes with a cost in terms of some extra degrees of freedom.
Also, this method requires a large enough number of degrees of freedom in the
thickness direction of the solid. For second-order elements, typically three elements are
required.

Beam Point to Solid (3D)

A beam in 3D can be coupled to a solid by adding a Solid-Beam Connection
multiphysics coupling. In the settings, set Connection type to either Solid boundaries to
beam points, general. or Solid boundaries to beam points, transition. These two
couplings are fundamentally different.

The Solid boundaries to beam points, general connection is used for modeling a beam
with one end “welded” to the face of the solid. You can specify the size of the area on
the solid boundary that is connected to the endpoint of the beam in several different
ways.

The Solid boundaries to beam points, transition coupling is intended for modeling a
transition from a beam to a solid where beam assumptions are valid on both sides of
the connection. Thus, the geometry of the solid at the transition should match the
cross-section data given to the beam.

This connection type can include warping of the solid cross section. In order to
compute the warping properties, an extra PDE is solved over the cross-section
boundaries. To improve the performance, you should preferably solve for these
variables once in a separate stationary study or study step. In that study step, clear all
physics interfaces except the Solid-Beam Connection multiphysics coupling in the Physics
and Variables Selection section.

You can manually control whether to include warping or not. If not included, the setup
of the solver sequence is simplified, but there will be significant stress disturbances
close to the connection boundaries if the cross section is susceptible to warping.

There are four warping variables: one named Warping function and three named
Warping constant. In the successive study steps, you need to manually suppress

96 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

them. You can do so under the Dependent Variables node, where you first set Defined
by study step to User Defined. Then, for each of these four variables, clear the Solve for
this field check box.

Beam Edge to Solid Boundary (2D)

A beam in 2D can also be coupled to a solid by adding a Solid-Beam Connection
multiphysics coupling with Connection type set to Shared boundaries or Parallel
boundaries. This connection is used for adding a beam on top of a solid as a “cladding”.
It is possible to include an offset distance. The boundaries may be coincident or
parallel.

Beam Edge to Solid Boundary (3D)

A beam in 3D can also be coupled to a solid by adding a Solid-Beam Connection
multiphysics coupling with Connection type set to Solid boundaries to beam edges. This
connection is used for adding a beam that is “welded” along the surface of the solid.

Beam Edge to Shell Edge (3D)

A beam can be coupled to a shell by adding a Shell-Beam Connection multiphysics
coupling with Connection type set to either Shared boundaries or Parallel boundaries.
This connection is used for adding beams as stiffeners to shells. The edges may be
coincident or parallel. It is possible to prescribe that the beam has an offset from the
shell when a coincident edge is used.

Beam Point to Shell Boundary (3D)

A beam can be coupled to a shell by adding a Shell-Beam Connection multiphysics
coupling with Connection type set to Shell boundaries to beam points. This connection
is used for modeling a beam with one end “welded” to the face of the shell. You can
specify the size of the area on the shell boundary that is connected to the end point of
the beam in several different ways.

Beam Point to Shell Edge (3D)

A beam can be coupled to a shell by adding a Shell-Beam Connection multiphysics
coupling with Connection type set to Shell edges to beam points. This connection is
used for modeling a beam with one end “welded” to the edge of the shell. You can
specify the part of the shell edge that is connected to the end point of the beam in
several different ways.

Pipe Point to Solid Boundary (3D)

A pipe in 3D can be coupled to a solid by adding a Structure-Pipe Connection
multiphysics coupling. The coupling is intended for modeling a transition from a pipe
to a solid where beam assumptions are valid on both sides of the connection. Thus, the

COUPLING DIFFERENT ELEMENT TYPES | 97

 geometry of the solid at the transition should match the cross-section data given to the
pipe. The connection assumes that the pipe cross section is circular; if another cross
section is used, it is converted to an equivalent circular cross section. This means that
warping is not considered.

The connection can be considered an extension of the Solid boundaries to beam points,
transition coupling in Solid-Beam Connection to also account for radial deformation
of the pipe caused by the fluid pressure and the temperature difference over the cross
section.

Pipe Point to Shell Edge (3D)

A pipe in 3D can be coupled to a shell by adding a Structure-Pipe Connection
multiphysics coupling. The coupling is intended for modeling a transition from a pipe
to a shell where beam assumptions are valid on both sides of the connection. Thus, the
geometry of the shell at the transition should match the cross-section data given to the
pipe. The connection assumes that the pipe cross section is circular; if another cross
section is used, it is converted to an equivalent circular cross section. This means that
warping is not considered.

The connection can be considered an extension of the Shell edges to beam points
coupling in Shell-Beam Connection to also account for radial deformation of the pipe
caused by the fluid pressure and the temperature difference over the cross section.

Embedded Reinforcement
Lower dimension structural elements can be connected to a solid domain by adding
an Embedded Reinforcement multiphysics coupling. This connection supports
coupling truss, beam, and membrane elements to a Solid Mechanics interface. The
connection can either be rigid, or made by attaching springs between points on the
embedded structure and points in the solid. A more detailed discussion about this type
of modeling is given in Modeling Embedded Structures and Reinforcements.

The underlying theory and more details about the built-in couplings can
be found in

• Connection Between Shells and Solids

• Connection Between Shells and Beams
• Embedded Elements

98 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

 • Examples of all types of couplings between shells and beams are shown
in Connecting Shells and Beams: Application Library path
Structural_Mechanics_Module/Beams_and_Shells/shell_beam_connection
• An example of couplings between shells and solids is shown in
Connecting Shells and Solids: Application Library path
Structural_Mechanics_Module/Beams_and_Shells/shell_solid_connection
• An example of couplings between beams and solids is shown in
Connecting Beams and Solids: Application Library path
Structural_Mechanics_Module/Beams_and_Shells/beam_solid_connection

USING GENERAL COUPLING OPERATORS

The most general method of connecting parts modeled with different physics
interfaces is by using a General Extrusion operator. In this case the parts need not even
be in contact, so the connection is an abstraction.

An example could be a shell stiffened by beams. In practice, you would probably use
the built-in coupling described in Beam Edge to Shell Edge (3D) for this case, but the
example shows the principles.

In structure like this, the beam is usually placed at one side of the shell, so that the
centerline of the beam and the midsurface of the shell do not coincide. This difference

COUPLING DIFFERENT ELEMENT TYPES | 99

 must be taken into account, so the edges representing the beam are geometrically
disconnected from the midsurface of the shell.

Beam centerline

Mathematically, the connection between the beam and the shell can be expressed as


 u beam = u shell +    X beam – X shell 

  beam =  shell


or equivalently as


 u beam = u shell +   X beam – X shell   n a

  beam =  shell


Here,  is the rotation vector, which contains the rotational degrees of freedom in the
Beam interface. The rotation vector is also available as a variable in the Shell interface,
where it is derived from the rotational degrees of freedom a. The shell normal is
denoted by n. Small rotations are assumed.

To create the coupling:

1 Add a General Extrusion node under Definitions>Nonlocal Couplings. Select the line on
the shell midsurface as source. Enter data in the Destination Map.

100 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

2 Add a Prescribed Displacement/Rotation node in the Beam interface and select the
corresponding edge.
3 Enter data for the prescribed displacements and rotations, for example
genext1(u)+genext1(shell.thy)*zdist, where zdist is some expression
defining the distance from the beam axis to the shell midsurface.

Because a shell does not have a valid rotation degree of freedom around
its normal, the rotation of the beam should not be connected in that
direction.

In the COMSOL Multiphysics Reference Manual:

• Nonlocal Couplings and Coupling Operators and General Extrusion

• About Nonlocal Couplings

COUPLING DIFFERENT ELEMENT TYPES | 101

 Applying Loads
An important aspect of structural analysis is the formulation of the forces applied to
the modeled structure. You can use custom expressions, predefined or user-defined
coordinate systems, and even variables from other modeling physics interfaces.

Loads can be applied in the structural mechanics interfaces on the body, face, edge, or
point levels. You can also apply loads to special features like Rigid Material or Rigid
Connector. There is also an option to apply point loads to given coordinates, which do
not have to coincide with a geometrical point or a mesh node.

In this section:

• Units, Orientation, and • Acceleration Loads

Visualization • Temperature Loads — Thermal
• Load Cases Expansion
• Singular Loads • Hygroscopic Swelling
• Moments in the Solid Mechanics • Total Loads
Interface • Loads with a Spatial Variation
• Pressure • Variables for Loads

Units, Orientation, and Visualization

USING UNITS
Enter loads in any unit, independently of the base SI unit system in the model, because
COMSOL automatically converts any unit to the base SI unit system. To use the
feature for automatic unit conversion, enter the unit in square brackets, for example,
100[lbf/in^2].

The exception is random vibration analysis. In that case, no automatic unit conversions
are available, so you must enter loads in the base units of the model.

PREDEFINED AND CUSTOM COORDINATE SYSTEMS

In this module, different predefined coordinate systems are available when materials or
boundary conditions are specified. There is always the global coordinate system.
Depending on the dimensionality of the part being worked with, there can also be

102 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

predefined coordinate systems such as the local tangent and normal coordinate system
for boundaries.

Custom coordinate systems are also available and are useful, for example, to specify a
load in any direction without splitting it into components. From the Definitions
toolbar, select a Coordinate System ( ) from the menu.

Some coordinate systems can have solution dependent axis directions. If

you use such a system for defining a load, the directions of the load follow
the moving coordinate axis directions if the Include geometric nonlinearity
check box is selected under the Study settings section of the current study
step.

VISUALIZATION
If you have switched on the physics symbols (see Displaying Physics Symbols in the
Graphics Window — An Example in the COMSOL Multiphysics Reference Manual),
then an applied load is indicated by a symbol together with a coordinate system
indicator displaying the definition directions for the load. The actual direction or
magnitude of the load you enter is not, however, reflected by the symbol. As a load in
COMSOL Multiphysics can be a function of parameters, variables, the solution, or
results from other physics interfaces, it is not possible to display it with only the
information available in the individual load feature.

Once you have turned on the physics symbols for a certain physics interface, you can
fine-tune the display. Every feature which has associated physics symbols will now have
a check box Show physics symbols, by which you can control the display of the symbols
for that specific feature.

Loads are among the results for which predefined plots are generated, so you will
always have access to a visual feedback of the loads after the solution. How to work
with the default load plots is described in detail in the Plotting Applied Loads section.

Sometimes, especially if you have entered complicated load expressions in a large

model, it is important to inspect the load distribution before you run the analysis. You
will then need to generate a dataset and the predefined plots. The fastest way to do that

APPLYING LOADS | 103

 is to click Get Initial Value on the Study toolbar. Note that if you change the loads, you
will have to delete the dataset and generate it again.

In the COMSOL Multiphysics Reference Manual:

• Physics Symbols
• Using Units
• Coordinate Systems
• Plotting Applied Loads

For an example showing how to examine the load distribution, see

Bracket — Static Analysis: Application Library path
Structural_Mechanics_Module/Tutorials/bracket_static. This is also the first
example used in the Introduction to the Structural Mechanics Module.

Load Cases
For a Stationary or Frequency Domain study, you can define load cases and constraint
cases. Any load or constraint can be assigned to a load or constraint group, and then
be used conditionally.

For most load types, the load case acts as a simple multiplier, but some cases need a
more detailed discussion:

• A Prescribed Displacement or Prescribed Displacement/Rotation node can be assigned

both a constraint group and a load group. You can use the constraint group to
switch on and off the whole constraint. The load group acts as a multiplier to any
nonzero prescribed values of displacement and rotations, and will not have any
effect unless the constraint is active.
• When a load case multiplier is used for Thermal Expansion, the multiplier is applied
not to the actual temperature, but to the difference between the temperature and
the strain free reference temperature. The temperature difference, and thus the
thermal strain, is proportional to the load case multiplier.
• Since Thermal Expansion nodes are exclusive (only the last one given gives a
contribution for a certain domain), you cannot switch between different Thermal
Expansion nodes only by assigning them to different load cases.
• Hygroscopic Swelling behaves analogously to Thermal Expansion.

104 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

• For External Stress, the given stress values are multiplied by the load case multiplier.
• A Spring Foundation or Thin Elastic Layer node can be assigned a constraint group,
which you can use to switch the whole boundary condition on and off. If there is
also a Predeformation subnode, then you can assign a load group to that subnode.
The prescribed predeformation is then multiplied by the load case multiplier.
Predeformation nodes are exclusive, you cannot switch between them by assigning
them to different load cases.
• A Rigid Connector can be assigned both a constraint group and a load group. You can
use the constraint group to switch off the prescribed displacements and rotations.
The load group acts as a multiplier to nonzero prescribed values of displacement and
rotations.
• If you have added Applied Force or Applied Moment nodes under a Rigid Connector,
you can assign individual load groups to these nodes.
• If you have added Applied Force or Applied Moment nodes under a Rigid Material, you
can assign individual load groups to these nodes.

Any expression that acts purely as a load, that is, contributes only to the
right-hand side of the system of equations, can be part of the load case
handling. This is true even if, for example, the corresponding feature does
not have a setting for load cases, or if it is a contribution you have created
using equation based modeling.

To do this, you can modify any such expression expr to

if(group.<lgName>,group.<lgName>*expr). Here, <lgName> is the
parameter name you have chosen in the settings for the load group.

In most cases, the expression group.<lgName>*expr is sufficient. The

only reason to use the longer version with the if() statement if you want
to avoid that expr is evaluated for load cases in which it is not used.

For example, if you have a boundary load which partially is always active,
and partially is conditional, you can write 20[MPa]+group.lg1*10[MPa]
in the input field for the pressure.

In particular, this approach is useful for features that override each other,
like Thermal Expansion, since you can then accommodate several load
cases in a single node in the Model Builder.

APPLYING LOADS | 105

 In the COMSOL Multiphysics Reference Manual:

• About Load Cases

• Defining Load Groups and Constraint Groups
• Load Group
• Constraint Group

For an example about how to set up expressions for controlling position

and distribution of loads using load cases, see Tapered Cantilever with
Two Load Cases: Application Library path COMSOL_Multiphysics/
Structural_Mechanics/tapered_cantilever.

Singular Loads
In reality, loads always act on a finite area or over a volume. However, in a model loads
are sometimes defined on points or edges, which leads to a singularity. The reason for
this is that points and lines have no area, so the stress becomes infinite. Because of the
stress singularity, there are high stress values in the area surrounding the applied load.
The size of this area and the magnitude of the stress depend on both the mesh and the
material properties. The stress distribution at locations far from these singularities is
unaffected according a to a well-known principle in solid mechanics, the St. Venant’s
principle. It states that for an elastic body, statically equivalent systems of forces
produce the same stresses in the body, except in the immediate region where the loads
are applied.

Figure 2-5 shows a plate with a hole in plane stress loaded with a distributed load and
a point load of the same magnitude. The mesh consists of triangular elements with
quadratic shape functions. The high stress around the point load is dissipated within
the length of a few elements for both mesh cases. The stresses in the middle of the plate
and around the hole agree for the distributed load and the point load. The problem is
that due to the high stress around the singular load it is easy to overlook the high stress
region around the hole. When the point load is applied, the range must be manually
set for the stress plot to get the same visual feedback of the high stress region around
the hole in the two cases. This is because the default plot settings automatically set the
range based on the extreme values of the expression that is plotted.

Despite these findings it is good modeling practice to avoid singular loads because it is
difficult to estimate the size of the singular region. In the Structural Mechanics

106 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

Module, it is possible to define loads on all boundary types. However, avoid singular
loads altogether with elastoplastic or creep materials.

The Plasticity and Creep nodes are available as a subnode to Linear Elastic
Material nodes with the Nonlinear Structural Materials Module or the
Geomechanics Module.

normal mesh size

finer mesh size

Figure 2-5: A plate with a hole subject to a distributed load (left) and a point load (right).

For more details about singularities, see also

https://www.comsol.com/blogs/
singularities-in-finite-element-models-dealing-with-red-spots/

https://www.comsol.com/blogs/
applying-and-interpreting-saint-venants-principle/

APPLYING LOADS | 107

 Moments in the Solid Mechanics Interface
The Solid Mechanics interface, as opposed to the Beam, Pipe Mechanics, Plate, and
Shell interfaces, does not have rotational degrees of freedom. This makes the direct
specification of moment loads somewhat more complicated. To specify moments, you
can attach a Rigid Connector to the loaded area. The rigid connector has rotational
degrees of freedom, and it is possible to apply moments directly. For load application,
the flexible formulation of the rigid connector is particularly useful, since it avoids
artificial stiffening of the boundary where the load is applied.

Pressure
A pressure is a load acting toward the normal of a face of the structure. If there are
large deformations in the model and the Include geometric nonlinearity check box is
selected under the Study settings section of the current study step, the pressure acts as
a follower load. The pressure is then defined with respect to the geometry and, as the
geometry deforms locally, the orientation of the load changes. The size of the loaded
area can also change as an effect of straining.

Acceleration Loads
Within the structural mechanics interfaces, you will find four different types of loads
to describe acceleration loads:

• Gravity
• Base Excitation
• Rotating Frame
• Linearly Accelerated Frame

The two first nodes have are of a global type. They will be applied to all features in the
physics interface, and cannot have a spatial variation.

The Rotating Frame and Linearly Accelerated Frame nodes have a domain selection. They
can be applied selectively, and also depend on the coordinates. When applied to a set
of selected domains, such loads are applied also to lower dimensional objects, for
example, a point mass or an added mass on an edge.

All acceleration loads share the property that they are not applied to mass contributions
that belong to global features such as rigid connectors. There, you must add loads
explicitly.

108 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

When a structure is subjected to an acceleration which is applied to all its support
points, then it often more convenient to replace the support acceleration by a an
acceleration force, using a Base Excitation node. In particular, this is necessary when
using modal based dynamic analyses, in which nonzero prescribed displacements
cannot be used. Changing to a frame acceleration load does not affect the distribution
of forces and stresses. It will however imply that displacements, velocities, and
accelerations are measured relative to the supports, and not relative to a room fixed
coordinate system. The Base Excitation node does however also define absolute
acceleration variables, which are what would be measured by an accelerometer.

Temperature Loads — Thermal Expansion

When performing thermal expansion analysis, temperature loads are specified by
entering a temperature and a reference temperature in a thermal expansion subnode
which is available from the context menu (right-click the parent node, a Linear Elastic
Material node, for example) or from the Physics toolbar, Attributes menu. Enter a
constant temperature or an analytic expression that can depend on the coordinates or
dependent variables. For beams, plates, and shells it is also possible to specify bending
temperature loads. More details are available in the descriptions for each physics
interface.

When a separate physics interface is used to model heat transfer in the material, the
entry for the temperature is the dependent variable for the temperature from that
physics interface, typically T. In most cases, possible temperature variables from other
physics interfaces can be directly selected from a list.

• For more information about how to couple heat transfer analysis with
structural mechanics analysis, see Thermal-Structure Interaction. This
module also includes The Thermal Stress, Solid Interface.
• For a detailed discussion about thermal effects in structural mechanics
models, see Thermally Coupled Problems.

Hygroscopic Swelling
Some materials have the capability to absorb significant amounts of moisture through
diffusion processes. Changes in the moisture content may then cause volume changes.

To include the effects of hygroscopic swelling, the Hygroscopic Swelling subnode is

available from the context menu (right-click the parent node, Linear Elastic Material

APPLYING LOADS | 109

 node, for example) or from the Physics toolbar, Attributes menu. Enter a constant
concentration or an analytic expression that can depend on the coordinates or
dependent variables. For beams, plates, and shells it is also possible to specify bending
swelling loads caused by concentration gradient in the transverse direction. More
details are available in the descriptions for each physics interface.

When a separate physics interface is used to model the moisture diffusion in the
material, the entry for the concentration is the dependent variable for the
concentration from that physics interface, typically c. In most cases, possible
concentration variables from other physics interfaces can be directly selected from a
list.

The diffusion of the moisture into the material also adds to the mass density. You can
choose to automatically include this effect in a dynamic analysis, and also in mass
proportional loads, such as gravity and rotating frame loads.

Total Loads
You can specify a load either as a distributed load per unit length, area, or volume, or
as a total force to be distributed on a boundary. In the case of a total load, the applied
distributed load is the given load divided by the area (or length, or volume) on which
its acts. Thus, entering a total load is usually only meaningful when its orientation is
given by a Cartesian coordinate system.

Loads with a Spatial Variation

Since you can write any type of expression in an input text field, it is easy to describe
loads having a spatial variation.

HYDROSTATIC LOAD
Hydrostatic loading is a common special case of spatial variation. In this case, there is
often a fluid surface, above which there is no load. Such a load you can describe with
an expression like if(Z<ZSurf,rhoFluid*g_const*(ZSurf-Z),0). Here, ZSurf
and rhoFluid are assumed to be parameters containing the Z-coordinate of the fluid
surface and the mass density of the fluid respectively.

LOADS WITH A FAST SPATIAL VARIATION

If a load has a spatial variation which is fast relative to the element size, you may need
to increase the accuracy of the numerical integration used to compute the load
contribution. As a default, a load which varies no faster than the polynomial order of

110 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

the displacement shape functions can be integrated exactly. To change the integration
order, enable Equation View. In the Equation View node under the current load node
in the Model Builder tree, you can then increase the integration order.

The local stress state within the loaded element is still limited by what can be described
by the shape functions, but the total load applied on the structure will be more
accurate it you increase the integration order.

TRAVELING LOADS
Loads that are moving along the structure with time can be modeled using an
expression X-v*t, where v is the velocity of the load. The mesh independent point
load of the type Point Load, Free is particularly well suited for this type of modeling. If
a distributed load is modeled using this approach, it is often necessary to increase the
integration order as discussed in the previous section, since the load patch will typically
cover partial element faces.

Variables for Loads

Each node in which a load is given, such as Boundary Load or Point Load, creates a
number of variables which you have access to, for example during postprocessing.
These variables have standardized names. The names are constructed using the
following pattern: <phys>.<loadTag>_<loadType>_<geom><dir>, for example
solid.bndl1.F_Ax.

TABLE 2-4: THE LOAD VARIABLE NAMING SCHEME

DESCRIPTION EXAMPLES

<phys> Physics interface tag solid, truss2

<loadTag> Tag of the load feature bndl1, el2
<loadType> Force or moment F (force)
M (moment)
<geom> The type of object to which P (point)
the load is applied L (line)
A (area)
V (volume)
<dir> Orientation x, y, z, r (vector
components)
_Mag (magnitude of
the vector)

APPLYING LOADS | 111

 For loads which are do not have a geometrical selection, such as loads on Rigid
Connector and Rigid Material, the _<geom> part of the name is omitted. As an example,
a load created through an Applied Force node under a Rigid Connector, is contained in
a variable such as solid.rig2.rf4.Fy.

112 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

Defining Constraints
Defining the proper constraints for structural mechanics models is just as important as
defining the loads as together they make up the model boundary conditions. There are
many useful predefined physics features to define the constraints or to create
user-defined expressions that define constraints.

In this section:

• Rigid Body Motion

• Orientation
• Prescribed Displacements, Velocities, and Accelerations
• Symmetry Constraints
• Elemental and Nodal Constraints
• Suppressing Constraints on Lower Dimensions
• Kinematic Constraints
• Rotational Joints
• Attachments

Rigid Body Motion

In most cases, a structure must have a set of constraints which is sufficient to suppress
any rigid body motions. A stationary problem is solvable only if the structure is
sufficiently constrained. There must not be any possible rigid body modes. Thus, no
stress-free deformation states are allowed. In a dynamic analysis, rigid body motions
are admissible. The inertial forces will then balance the external forces.

The number of possible rigid body modes for different geometrical dimensions is
shown in the table below.
TABLE 2-5: NUMBER OF POSSIBLE RIGID BODY MODES

DIMENSION NUMBER OF RIGID BODY MODES

3D 6 (3 translations + 3 rotations)
2D axisymmetric 1 (Z-direction translation)
2D (solid, beam, truss) 3 (2 translations + 1 rotation)
2D (plate) 3 (1 translation + 2 rotations)

DEFINING CONSTRAINTS | 113

 If the model is under constrained, you may encounter the following problems:

• The solver reports that the stiffness matrix is singular.

• The solver reports that the stiffness matrix is ill-conditioned. Theoretically, the
matrix is singular for a structure with rigid body modes, but because of the roundoff
errors during the solution this is not exactly determined.
• A nonlinear analysis fails to converge.
• An iterative linear equation solver fails to converge.
• You get a solution with an extremely large displacement, orders of magnitude larger
than what is expected.

For a single body, it is seldom difficult to see whether it is fully constrained or not, but
for a more complex assembly, including several physics interfaces, or advanced
couplings and boundary conditions, it may not be trivial. If you suspect that rigid body
modes are a problem in your model, you can run an eigenfrequency analysis, and check
for modes with zero eigenfrequency as described in Eigenfrequency Analysis.

If there are no constraints which are dictated by the physical boundary conditions of
the structure, you can use the Rigid Motion Suppression feature to automatically
remove the rigid body motions. When you do this, the assumption is that the external
loads are in equilibrium. If not, reaction forces and stress concentrations will appear at
seemingly arbitrary points where the automatic constraints were placed.

As an alternative to applying constraints, you can also add elastic supports through a
Spring Foundation node to suppress rigid body motion.

Rigid Motion Suppression in the Structural Mechanics Theory chapter.

Orientation
You can specify constraints in global as well as in any previously defined local
coordinate system.

Coordinate Systems in the COMSOL Multiphysics Reference Manual

114 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

Prescribed Displacements, Velocities, and Accelerations
The most fundamental constraint is the prescribed displacement, where the individual
components of displacement or rotation can be prescribed to zero or nonzero values
for points, edges, boundaries, or domains.

For dynamic analysis, you can also directly prescribe the velocity or acceleration. The
conditions for prescribing displacements, velocities, or accelerations are mutually
exclusive for the same geometrical object since they prescribe the same degree of
freedom.

FREQUENCY DOMAIN
In frequency domain, a prescribed velocity vp or prescribed acceleration ap can be
directly interpreted as a prescribed displacement up:

vp
u p = ------
i

–ap
u p = --------
2
-


where  is the angular frequency.

A prescribed velocity with zero phase is assumed to have its peak at the
reference phase. As an effect, the corresponding peak displacement is
shifted by 90°. Similarly, a positive prescribed acceleration with zero phase
corresponds to a negative value of the displacement.

These definitions are particularly important if you mix prescribed velocity

or acceleration conditions with other boundary conditions.

TIME DOMAIN
In the case of a time-dependent analysis, the prescribed displacement is obtained as


u p  t  = u 0  t 0  + v p    d
t0

or

DEFINING CONSTRAINTS | 115

 t 
 
u p  t  = u 0  t 0  + v 0  t 0  + a p    d d
 

 
t0  t0 

where u0 and v0 are is given by the initial conditions. It is not possible to set explicit
initial conditions, but if initial values are taken from a previous study, they will be
respected. In order to compute the integrals, up is introduced as a separate degree of
freedom which is solved for by adding an extra ODE.

As prescribing the velocity or acceleration in time domain comes with an extra cost,
you should always consider using a prescribed displacement instead. As long as the
time history of the velocity or acceleration is a known a priori and does not depend on
the solution itself, this is possible.

• When the velocity or acceleration has a simple time dependence, you can integrate
it analytically one or two times to obtain the displacement, and directly prescribe the
displacement instead.
• When you have complicated known velocity or acceleration histories, for example
from measurements, you can use the integrate() operator. In this case, you enter
the prescribed displacement as integrate(my_data(tau),tau,0,t). Here
my_data is the measured data as function of time, and tau is a dummy integration
variable

When a local coordinate system is used for prescribing a prescribed velocity or

acceleration, the axis directions must be fixed in space. As an example, you cannot use
a Boundary System rotating with the deformation.

STATIONARY ANALYSIS
In a stationary analysis, the prescribed velocity and acceleration nodes can have two
different behaviors. As a default, they are ignored, but you can also select that the
degrees of freedom having a prescribed velocity or acceleration in a dynamic analysis
should be constrained to zero in a static analysis.

Symmetry Constraints
In many cases symmetry of the geometry and loads can be used to your advantage in
modeling. Symmetries can often greatly reduce the size of a model and hence reduce
the memory requirements and solution time. When a structure exhibits axial
symmetry, use the axisymmetric physics interfaces. A solid that is generated by rotating
a planar shape about an axis is said to have axial symmetry. In order to make use of

116 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

the axisymmetric physics interfaces, all loads and constraints must also be the same
around the circumference.

For other types of symmetry, use the predefined symmetry and antisymmetry
constraints. This means that no expressions need to be entered — instead just add the
type of constraint to apply to the model.

Physics Interface Axial Symmetry Node in the COMSOL Multiphysics

Reference Manual

If the geometry exhibits two symmetry planes (Figure 2-6), model a quarter of the
geometry by using the Symmetry node for the two selected surfaces.

Symmetry planes Apply symmetry constraints

Figure 2-6: If the geometry exhibits two symmetry planes, model a quarter of the geometry
by using the Symmetry feature for the two selected surfaces.

Both geometric symmetry and loads are important when selecting the
correct constraints for a model.

In an eigenfrequency or buckling analysis, the eigenmodes might be

nonsymmetric even if the structure is symmetric.

Figure 2-7 shows symmetric and antisymmetric loading of a symmetric geometry.

When modeling half of the geometry, the correct constraint for the face at the middle

DEFINING CONSTRAINTS | 117

 of the object would be Antisymmetry in the case of antisymmetric loading and
Symmetry in the case of symmetric loading of the object.

Symmetry plane Antisymmetry plane

Figure 2-7: Symmetry plane (left) and antisymmetry plane (right).

SYMMETRY IN 2D AXISYMMETRY
In an axisymmetric model, the only possible symmetry is when the symmetry plane is
normal to the Z-axis. For models in 2d axisymmetry, the Symmetry Plane node is used
for prescribing this type of symmetry.

Antisymmetry cannot exist in this case.

TRANSLATION OF THE SYMMETRY PLANE

In some situations, you may want to use a symmetry condition, in which the symmetry
boundary actually can move along its normal. This may for example the case when you
use symmetry conditions to terminate your modeled region even though the situation
is not truly symmetric. The best approximation may then be that the boundary remains
planar, but that there is no resultant reaction force from the boundary condition.

You can modify the symmetry condition, so that it can translate in various ways by
using the controls in the Normal Direction Condition section of the settings for the
Symmetry constraint. You can model the following cases:

• Reaction force free translation.

• Prescribed total force acting on the constrained part.
• The displacement in the normal direction is prescribed.

118 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

Note that allowing translation in the symmetry constraint is only meaningful if the
geometry selection corresponds to a single symmetry plane.

For an example showing how to force a boundary to remain plane, but

still allow it to translate in its normal direction using this special version
of Symmetry, see Thermo-Mechanical Analysis of a Surface-Mounted
Resistor: Application Library path Structural_Mechanics_Module/
Thermal-Structure_Interaction/surface_resistor.

Symmetry Condition with Translation in the Structural Mechanics

Theory chapter.

Elemental and Nodal Constraints

For most constraints, you can select between using elemental and nodal constraints.
To do this, select Advanced Physics Options, so that the Constraint Settings section is
displayed.

When using nodal constraints, one constraint is generated for each node within the
selection a certain constraint feature. With elemental constraints, the number of
constraints added at a node equals the number of elements connected to that node.
This means that if some values used in the constraints differ between the elements,
then different constraints will be generated by the elemental method, whereas with the
nodal method an average is computed at the node before adding the constraint.

When several constraints are present at a node, the internal constraint elimination
algorithm is responsible for reducing them to a minimum unique set. Using elemental
constraints will clearly put an extra burden on this algorithm, so whenever possible you
should use nodal constraints.

The two different options exist, since under some circumstances the actual constraints
can differ between the two methods. Consider for example a symmetry constraint,
where the displacement in the direction normal to the boundary is constrained by the
equation

un = 0

where n is the unit normal vector.

DEFINING CONSTRAINTS | 119

 If there are several intersecting symmetry planes, like in Figure 2-6, using nodal
constraints could cause a problem:

• If both boundaries are selected in the same Symmetry node, then only a single
constraint is applied for each node along the common edge, while you actually want
constraints along the normals of both planes. The normal used would be pointing
somewhere between the two planes, since a nodal constraint uses averaging of the
values from the adjacent elements.
• If two Symmetry nodes are used, so that the selection in any one of them only
contains boundaries without a normal direction discontinuity, the intended
constraints are added. On the common edge, there will be two contributions, one
from each Symmetry node, and each using the normal direction of its boundary. If
you want to use nodal constraints, you must set up your model in this way if the
constraints are orientation dependent.

Elemental constraints, on the other hand, can cause problems if the constraints added
by adjacent elements are not exactly the same. This could for example happen if the
normal orientation differs between neighboring elements. In such a case, a boundary
could behave as if it were fixed when a Symmetry, Antisymmetry, or Roller constraint is
applied. Such a situation could occur when the component consists of an imported
mesh, so that no underlying geometry exists.

The default type of the constraint, nodal or elemental, differs between different
constraint features. A nodal formulation is the default whenever it is considered safe,
like for a Fixed Constraint. Whenever the constraint can have a dependency on the
surface orientation, the default value is elemental.

See also Constraint Settings in the COMSOL Multiphysics Reference

Manual.

Suppressing Constraints on Lower Dimensions

Sometimes, boundary conditions on two adjacent objects can come into conflict on a
shared object.

For most constraints in the structural mechanics interfaces, you have the possibility to
select that certain objects of lower dimensions should be excluded from the main
selection. To do this, you must first select Advanced Physics Options. In the settings for
a constraint, like for example Prescribed Displacement, new sections named Excluded

120 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

Surfaces, Excluded Edges, and Excluded Points will then appear. In these sections, you
can select geometrical objects which should be excluded from the main selection when
the constraint is applied.

In the structural mechanics interfaces, there are many types of complex constraints,
and sometimes you may get conflicts or duplicates which makes the model either
overconstrained, or problematic for the automatic constraint elimination algorithm. If
you are aware of such situations, it is good practice to remove one of the potentially
conflicting constraints. One example of such a situation is when you have a Solid-Shell
Connection meeting a symmetry plane, as shown in Figure 2-8.

Figure 2-8: Example of potentially conflicting constraints

Here you would add a Symmetry condition on a boundary in the Solid Mechanics
interface, as well as a Symmetry condition on an edge in the Shell interface. But at the
same time, the displacements on whole boundary where the solid meets the shell are
controlled by shell degrees of freedom as an effect of the Solid-Shell Connection. As a
result, on the edge marked with Conflict in the sketch, the displacements will be
controlled both by the symmetry condition is Solid Mechanics, and implicitly through
the coupling, by the symmetry condition in the Shell interface. Particularly if the
geometry is curved, there is a risk that these constraints are not identical from a
numerical point of view. In this case, excluding the conflicting edge from the selected
boundary in the Solid Mechanics interface will make the behavior unique and fully
predictable.

Another example where constraints will come in conflict is if you want to constrain the
displacement on parts of the geometry using weak constraints, while keeping the
default pointwise constraints on other parts. If the same mesh node has both types of

DEFINING CONSTRAINTS | 121

 constraints, the solution will fail, so you must exclude any common geometrical objects
from the selection in one of the constraints.

See also Excluded Surfaces, Excluded Edges, and Excluded Points in the
COMSOL Multiphysics Reference Manual.

Kinematic Constraints
Kinematic constraints are equations that control the motion of solids, faces, edges, or
points. Add a Prescribed Displacement constraint to enter expressions for constraints.
You can define the equations using predefined coordinate systems as well as custom
coordinate systems. Special constraints, for instance to keep an edge of body straight
or to make a boundary rotate, require such constraint equations.

In the 3D and 2D Solid Mechanics interfaces and in the Shell and Beam
interface there is a special constraint called a Rigid Connector. A rigid
connector is applied to one or more boundaries, edges, or points and force
them to behave as connected to a common rigid body. The rigid
connector can be given prescribed displacements and rotations and thus
simplifies the realization of some constraints.

Rotational Joints
Joints between elements in The Truss Interface are automatically rotational joints
because the truss elements have no rotational degrees of freedom. For beams, however,
the rotational degrees of freedom are by default coupled between elements. To create
a rotational joint between two beam elements, add one additional Beam interface to a
geometry. Make sure that it is only active for the edge that includes the point where
the joint is positioned and that no other physics interface is active here. Couple the
translational degrees of freedom and leave the rotational degrees of freedom
uncoupled at the joint.

Attachments
An Attachment is a set of boundaries, edges, or points on a flexible or rigid component
used to connect it to another flexible or rigid component through a joint or spring. An

122 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

attachment can be defined on the boundaries of a solid element, on the edges of a shell
element, and on the points of a beam element.

When an attachment is connected to a flexible body, you can use two different
formulations: rigid or flexible.

In the rigid attachment formulation, all selected boundaries or edges behave as if they
were connected to a common rigid body. This may cause an unrealistic stiffening and
local stress concentrations.

In the flexible version, the boundaries are allowed to deform, and the rigid body
constraints are enforced only in an average sense.

The attachment formulation is similar to that of a rigid connector. In the rigid case,
the only degrees of freedom needed to represent this assembly are the ones needed to
describe the movement of a rigid body. In 2D this is just two in-plane translations, and
the rotation around the out-of-plane axis.

In 3D the situation is more complex. Six degrees of freedom are necessary, usually
selected as three translations and three parameters for the rotation. For finite rotations
any choice of three rotation parameters is singular at some specific set of angles. For
this reason, a four-parameter quaternion representation is used.

Some extra degrees of freedom are added for each attachment where the flexible
formulation is used.

When an attachment is defined on a rigid component, it does not create any degrees
of freedom of its own and directly picks the degrees of freedom of the rigid
component.

The formulation of the attachment feature is same as the rigid connector

formulation, which is discussed in Rigid Connector in the Structural
Mechanics Module User’s Guide.

Some useful information about the attachment feature:

• The attachment can be defined either on a flexible or on a rigid component. It is not

possible to select the boundaries from both types of components in a single
attachment. The reason is the different formulation of attachment for flexible or
rigid components.

DEFINING CONSTRAINTS | 123

 • The attachment’s center of rotation is the centroid of its selected boundaries, edges,
or points. In a joint, it is possible to select the attachment center of rotation as the
center of the joint.
• If the attachment is connected to a flexible domain, the forces and moments on an
attachment with a boundary selection are computed by summing the reaction forces
on the selected boundaries. There is an option to use these forces and moments to
evaluate the joint forces.
• For an attachment in the Shell interface, the selection consists of edges. The forces
and moments on an attachment are then computed by summing the reaction forces
and reaction moments on the selected edges. There is an option to use these forces
and moments to evaluate the joint forces.
• For an attachment in the Beam interface, the selection consists of points. The forces
and moments on an attachment are then computed by summing the reaction forces
and reaction moments on the selected points. There is an option to use these forces
and moments to evaluate the joint forces.
• When an attachment is defined in the Shell or Beam interface, the normal force
cannot be defined through Attachment reaction forces in the Friction subnode of a
joint.
• In a joint you select attachments to establish its connection with flexible, or
optionally, rigid bodies. This is accomplished by setting up a relation between the
source attachment and destination attachment degrees of freedom. Thus, the
motion of an attachment is governed by the joints where this attachment is used.
• An attachment which is not referenced by any joint acts as an unconstrained rigid
connector.

• The Multibody Dynamics Interface

• The Solid Mechanics Interface, The Shell and Plate Interfaces, and The
Beam Interface in the Structural Mechanics Module User’s Guide.

124 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

Calculating Reaction Forces
There are different ways to evaluate reaction forces and these are discussed in this
section.

• Using Predefined Variables to Evaluate Reaction Forces

• Using Weak Constraints to Evaluate Reaction Forces
• Using Surface Traction to Evaluate Reaction Forces
• Evaluating Surface Traction Forces on Interior Boundaries

The following sections describe the merits and costs of these methods.

Using Predefined Variables to Evaluate Reaction Forces

The results analysis capabilities include easy access to the reaction forces and moments.
They are available as predefined variables. The reaction force variables are available
only at the nodes, and not as a continuous field, so they are not suitable for graphic
presentation.

To compute the sum of the reaction forces over a region, use

Volume Integration, Surface Integration, or Line Integration under Results>
Derived Values or under an Evaluation Group node. The integration
method discovers that the reaction forces are discrete values and applies a
summation instead of an integration.

If you create an integration operator under Component>Definitions>

Nonlocal Couplings>Integration to sum reaction forces, you must explicitly
set Method to Summation over nodes.

Reaction forces are computed as the sum of the nodal values over the selected volume,
face, or edge. Reaction moments are calculated as the sum of the moment from the
reaction forces with respect to a reference point, and any explicit reaction moments (if

CALCULATING REACTION FORCES | 125

 there are rotational degrees of freedom). During postprocessing, you can modify the
coordinates of the reference point in the Parameters section of a result feature.

• Reaction forces are not available for eigenfrequency analysis or when

weak constraints are used.
• Reaction force variables are computed where there are constraints, that
is Dirichlet conditions like Fixed Constraint, or Prescribed Displacement.
Reaction force variables are found in the Reactions folder in the result
menus, and have names like solid.RFx and solid.RMz for an
x-directed force and a moment around the z-axis, respectively.
• From the physical point of view, there can be other sources of reaction
forces, such as the elastic and viscous forces from a Spring Foundation.
Such forces are not included in the reaction force variables, but are
placed in separate variables. You will find them in the Spring and
damping forces folder in the result menus.

• Total reaction force variables are available. In these variables, reaction

forces, spring forces, and damping forces have been integrated over the
whole physics interface. These variables are found in the Reactions
folder in the result menus, and have names like solid.RFtotalx and
solid.RMtotalz for an x-directed force and a moment around the
z-axis, respectively. These variables should thus, for a stationary
analysis, be equal to the total applied load.
• If reaction forces are summed independently for two adjacent
boundaries, the total sum is not the same as if the reaction forces were
summed for both boundaries in one operation. The values of the nodes
at the common edge always contain contributions from the elements
at both sides of the edge.

Derived Values, Evaluation Groups, and Tables in the COMSOL

Multiphysics Reference Manual

Using Weak Constraints to Evaluate Reaction Forces

Select the Use weak constraints check box to get accurate distributed reactions. Extra
variables that correspond to the reaction traction distribution are automatically added
to the solution components.

126 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

With weak constraints activated, COMSOL Multiphysics adds the reaction forces to
the solution components. The variables are denoted X_lm, where X is the name of the
constrained degree of freedom (as, for example, u_lm and v_lm). The extension lm
stands for Lagrange multipliers. It is only possible to evaluate reaction forces on
constrained boundaries in the directions of the constraints.

To compute the total reaction force on a boundary, integrate one of the

variables X_lm using Volume Integration, Surface Integration, or Line
Integration under Derived Values or under an Evaluation Group node.

If the constraint is defined in a local coordinate system, the degrees of

freedom for the weak constraint variables are defined along the directions
of that system.

Since the reaction force variables are added to the solution components, the number
of DOFs for the model increases slightly, depending on the mesh size for the
boundaries in question. Boundaries that are adjacent to each other must have the same
constraint settings. The reason for this is that adjacent boundaries share a common
node.

Using weak constraints affects the structure of the equation system to be solved, and
is not suitable for all types of equation solvers.

In the COMSOL Multiphysics Reference Manual:

• Derived Values, Evaluation Groups, and Tables

• Symmetric and Nonsymmetric Constraints

Using Surface Traction to Evaluate Reaction Forces

As an alternative method, you can obtain values of the reaction forces on constrained
boundaries by using boundary integration of the relevant components of the surface
traction vector.

For 2D and 1D axisymmetric components, multiply the surface traction

by the cross-section thickness, and for 1D components multiply the
surface traction by the cross-section area before integrating to calculate
the total reaction force.

CALCULATING REACTION FORCES | 127

 Two different types of surface traction results can be computed in COMSOL
Multiphysics:

The first type, contained in the variables interface.Tax, is computed from the
stresses. It is always available. Since the surface traction vector is based on computed
stress results, this method is less accurate for computing reactions than the other
methods.

The second type, contained in the variables interface.Tracx, is computed using a

method similar to the weak constraints, but without introducing the Lagrange
multipliers as extra degrees of freedom. The accuracy is high, but there is an extra
computational cost. These traction variables are computed only if the Compute
boundary fluxes check box in the Discretization section is selected for the Solid
Mechanics interface.

In case of geometric nonlinearity, the two types of traction variables are

interpreted differently. The interface.Tax variables are based on
Cauchy stress, and contains a force per current area. If you integrate them
you must use the spatial frame. The interface.Tracx variables are based
on First Piola–Kirchhoff stresses and contains a force per undeformed
area. An integration must then be done on the material frame.

Evaluating Surface Traction Forces on Interior Boundaries

As opposed to the other methods for reaction force computation, the boundary flux
based tractions are computed not only on external boundaries, but also on interior
boundaries. On interior boundaries, there are then two traction fields: One acting
from each of the domains sharing the boundary. These internal traction fields are
contained in the variables interface.iTracux and interface.iTracdx. The letters
u and d in the variable names indicate the upside and downside of the boundary,
respectively. If you need the value of the total force acting on an internal section
through your model, these variables can be integrated. The interface.iTracux and
interface.iTracdx variables are only available if the Compute boundary fluxes check
box is selected in the Discretization section of the physics interface, and there are
interior boundaries in your model.

Computing Accurate Fluxes in the COMSOL Multiphysics Reference

Manual

128 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

Introduction to Material Models
In this section:

• Material Models for Structural Mechanics

• Entering Material Data
• Introduction to Linear Elastic Materials
• Introduction to Viscoelastic Materials
• Mixed Formulation
• Modeling Damage
• About the Material Databases for the Structural Mechanics Module
• Using External Materials

Material Models for Structural Mechanics

The Structural Mechanics Module without any add-on modules provides the Linear
Elastic material with Viscoelasticity and Thermal Expansion modeling capabilities. It
also provides access to piezoelectric and magnetostrictive materials.

If you have the optional products Nonlinear Structural Materials Module or

Geomechanics Module, many other classes of nonlinear materials are also available.
These models can be modified and extended, and custom material models can be
defined.

You can also add a material model which you have coded yourself and made available
as a binary library file using an External Stress-Strain Relation.

• Modeling Piezoelectric Problems

• Modeling Magnetostrictive Materials
In the COMSOL Multiphysics Reference Manual:

• Working with External Materials

• External Material

INTRODUCTION TO MATERIAL MODELS | 129

 In Table 2-6 and Table 2-7 you can find an overview of the families of materials, and
their applicability in the various structural mechanics interfaces.
TABLE 2-6: MATERIAL MODELS IN THE PHYSICS INTERFACES ON DOMAINS AND BOUNDARIES

MATERIAL SOLID ELASTIC SHELL/ LAYERED MEMBRANE

MODEL MECHANICS WAVES, PLATE SHELL
TIME
EXPLICIT

Linear Elastic Material X — X X X

Layered Linear Elastic — — X — X
Material
Section Stiffness — — X — —
Elastic Waves, Time — X — — —
Explicit Model
Nonlinear Elastic X — — — X
Material
Hyperelastic Material X — — X X
Layered Hyperelastic — — X — —
Material
Shape Memory Alloy X — — — —
Piezoelectric Material X — — X —
Piezomagnetic X — — — —
Material
Viscoelasticity X — X X X
Plasticity X — X X X
Soil Plasticity X — — — —
Creep X — X X X
Elastoplastic Soil X — — — —
Material
Viscoplasticity X — X X X
Porous Plasticity X — — — —
Concrete X — — — —
Rocks X — — — —
Damage X — — — —
External Stress-Strain X — — — X
Relation
Rigid Material X — X X —

130 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

TABLE 2-7: MATERIAL MODELS IN THE PHYSICS INTERFACES ON EDGES

MATERIAL BEAM PIPE TRUSS

MODEL MECHANICS

Linear Elastic Material X — X

Fluid and Pipe — X —
Materials
Section Stiffness X — —
Plasticity — — X
Rigid Material X X —
Spring-Damper — — X
Material

Many of the material models can be augmented by effects like thermal expansion,
hygroscopic swelling, initial stresses and strains, external stress, and activation.

Combination of Material Models

It is possible to combine many of the effects in an additive manner. The models based
on elasticity all have the same structure where

1 An elastic strain is computed by removing all inelastic strains (for example, plastic or
thermal strains) from the total strain.
2 An “elastic stress” is computed from the elastic strains.
3 Any additional stresses (for example viscous stresses, or initial stresses) are added to
form the total stress.

This concept will give you a great freedom in combining effects. Some such useful
combinations are

• Plasticity and thermal expansion

• Plasticity and creep
• Thermal expansion and hygroscopic swelling
• Viscoelasticity and creep

Entering Material Data

For most material data, you have the option to choose between From material and
User defined. The preferred way of supplying the material data, is through the Materials

INTRODUCTION TO MATERIAL MODELS | 131

 node. If you are using data from the Material Library, this is the only option, but also
when supplying your own data this will improve clarity of model.

Constitutive matrices, such as the elasticity matrix for an anisotropic material, are in
many cases per definition symmetric. Only the upper diagonal of the matrix given as
input is used for forming the matrix used, so you need not enter the lower diagonal
part.

Introduction to Linear Elastic Materials

Linear elasticity forms the basis for the majority of structural mechanics simulations.

For isotropic linear elasticity, two parameters are enough to describe the material
behavior. The number of parameters increases to (at most) 21 for the fully anisotropic
case in 3D. When setting up a model, make sure that the material parameters are
defined in agreement with the type of relationship used. If necessary, transform the
material data before entering it in the physics interface. For example, for orthotropic
materials calculate the Poisson’s ratio xy by

Ex
 xy =  yx ------
Ey

In the theory section Linear Elastic Material

Introduction to Viscoelastic Materials

The generalized Maxwell, standard linear solid (SLS) and Kelvin-Voigt viscoelastic
materials are available. All the models are linear, and the corresponding materials can
be described as consisting of one or more branches with a spring and a dashpot acting
in parallel to a linear elastic material. For each viscoelastic branch, the shear modulus
and the relaxation time (or viscosity) are entered.

• Linear Viscoelasticity
• Viscoelasticity

132 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

Mixed Formulation
Nearly and fully incompressible materials can cause numerical problems if only
displacements are used in the interpolating functions. Small errors in the evaluation of
volumetric strain, due to the finite resolution of the discrete model, are exaggerated by
the high bulk modulus. This leads to an unstable representation of stresses, and in
general to underestimation of the deformation, because spurious volumetric stresses
might balance also applied shear and bending loads.

In several material models you will find settings named Use mixed formulation or
Compressibility, by which you can introduce a mixed formulation.

Use a mixed formulation when the material data is such that the deformation is close
to being incompressible. For an isotropic elastic material, this happens when Poisson’s
ratio approaches 0.5.

There are different approaches to assess which combinations of displacement shape

function types and auxiliary variable shape function types yield numerically sound and
effective elements. In general, the shape function order of the auxiliary variable should
be lower than for the displacement field, to avoid locking. However, this is not a
sufficient requirement. The inf-sup condition (Ref. 1) can be used to analytically or
numerically identify sound mixed element formulations. In general, the outcome of
such a test will depend not only on the shape function type combination, but also on
the element type. For instance, a sound and effective combination of shape function
types for a tetrahedral element is not necessarily suitable for a hexahedral element.
COMSOL Multiphysics provides four types of shape functions for the auxiliary
pressure or auxiliary volumetric strain variable. The different shape function types for
the auxiliary variable are described below. Depending on the particular context, one of
these types is implicitly selected using Automatic.

Note that in the case of a mixed mesh, you may have to sub-divide the
domain by element type, to be able to control the shape functions for the
auxiliary variable.

DISCONTINUOUS LAGRANGE
If this shape function type is selected, the auxiliary variable shape function is
discontinuous across element boundaries, and it is one order lower than the shape
function order of the displacements.

INTRODUCTION TO MATERIAL MODELS | 133

 CONTINUOUS
If this shape function type is selected, the auxiliary variable shape function is
continuous across element boundaries. It will be of the same type (Lagrange or
Serendipity) as that of the displacements, but one order lower. A special case is for a
linear displacement field, for which a discontinuous, constant, shape function is used.

LINEAR
If this shape function type is selected, the auxiliary variable is regarded as a linearly
interpolated field in the element. In the case of an auxiliary pressure, the interpolation
in a 3D isoparametric element is

p = p0 + 1  p1 – p0  + 2  p2 – p0  + 3  p3 – p0 

where p0, p1, p2, and p3 are auxiliary pressure coefficients, and 1, 2, and 3 are
isoparametric coordinates. Note that the field is linear in the local element coordinates,
and that it is not continuous across element boundaries.

The linear shape function type is not available for layered material
features.

CONSTANT
This shape function type represents a constant auxiliary variable in the element, p = p0.

The mixed formulation is useful not only for linear elastic materials but
also for nonlinear elastic materials, elastoplastic materials, hyperelastic
materials, and viscoelastic materials.

Note that some iterative solvers do not work well together with mixed
formulation because the stiffness matrix becomes indefinite.

For more details, see Mixed Formulation in the theory chapter.

134 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

Modeling Damage
Modeling problems with strain localization might turn into unstable behavior and
convergence difficulties. The following techniques can be used to help in such
situations:

• Use a displacement-controlled loading scheme, since damage is often associated

with a reduction in load carrying capacity, see the Cracking of a Notched Beam
example in the Geomechanics Module or in the Nonlinear Structural Mechanics
Module Application Libraries.
• Better convergence is often obtained when the Nonlinear method is set to Constant
(Newton) with a Damping factor equal to 1, and a Jacobian update on every iteration.
• Reduce the smallest and/or largest allowable step size in an Auxiliary sweep, or
restrict the time steps in a time-dependent study.
• Introducing a soft spot where a crack is expected can make the localization of strains
more stable, see the Brittle Damage in Uniaxial Tension example in the
Geomechanics Module or in the Nonlinear Structural Mechanics Module
Application Libraries. Alternatively, a random spatial distribution of the material
parameters could be employed to obtain a more stable solution.

When using the crack band method or no regularization at all, the following steps are
recommended:

• The size of the biggest mesh element h should not exceed 2EGf/ts2, where E is
the Young’s modulus, Gf is the fracture energy per unit area, and ts is the tensile
strength. Larger values of h will cause a snap-back of the stress-strain curve at the
material point level.
• Use linear shape functions for the displacement field. When using higher-order
shape functions, strains may localize in either rows of Gauss points, or entire
elements, depending on the stress state and numerical rounding errors.
• If cracks are located on a symmetry plane, the model parameters should be modified
so that the amount of dissipated energy is reduced by one half in the elements
adjacent to the symmetry plane. This can be achieved, for example, by reducing the
fracture energy in that row of elements.
When using the implicit gradient method, the element size should be sufficiently small
to resolve damaged zones. The same applies to the Phase field damage model, where it
is recommended that size of the mesh elements in the expected crack path follows
h < lint/2 for a linear displacement field, otherwise h < lint.

INTRODUCTION TO MATERIAL MODELS | 135

 While it is possible to solve brittle fracture problems with the Phase field damage model
by applying a fully coupled strategy, this can often exhibit poor or slow convergence.
An alternative and often more stable approach is to use a segregated solution strategy,
by splitting the evolution of the crack phase field and the displacement field in two
groups. This type of algorithmic operator split can conceptually be summarized as
follows for step n+1:

1 Initialization. The crack phase field, displacement field and state variables are
known at step n.
2 Update state variables. Update internal state variables used by the phase field
model with values from step n.
3 Solve for the Crack Phase Field. Compute the crack phase field variable in a Newton
step, with the displacement field frozen at step n.
4 Solve for the Displacement field. Compute displacement field variables in a Newton
step with the updated crack phase field.
This leads to a single-pass algorithm that is accurate only for sufficiently small
parameter or time steps.

The default solver will suggest the above single-pass algorithm for the
Phase field damage model when it is feasible to perform the operator split.
Cases where this is not possible include when some multiphysics
couplings are present in the model and when a segregated contact
algorithm has to be used.

An improvement to the method is to add a multi-pass correction by iterating over steps

3 and 4 in each increment; either until convergence is achieved or for a predefined
number of iterations. This type of strategy is demonstrated in the Brittle Fracture of
a Holed Plate example in the Geomechanics Module or in the Nonlinear Structural
Mechanics Module Application Libraries, where a multi-pass algorithm with a
maximum of 3 outer iterations is used. Note that although the solution is accepted
without requiring convergence of the outer problem, each sub group locally fulfills the
defined convergence criterion. Hence the displacement field can be considered as a
converged solution given the current crack phase field.

About the Material Databases for the Structural Mechanics Module

The Structural Mechanics Module includes these material databases: Liquids and
Gases, with temperature-dependent fluid dynamic and thermal properties, MEMS, an

136 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

extended solid materials library with metals, semiconductors, insulators, and polymers
common in MEMS devices, and a Piezoelectric database with over 20 common
piezoelectric materials. The materials include temperature-dependent fluid dynamic
and thermal properties.

In the COMSOL Multiphysics Reference Manual:

• MEMS Material Library

• Piezoelectric Materials Library
• Liquids and Gases Materials Library
• Materials

For an example of the MEMS materials database and Piezoelectric

materials database, see Piezoelectric Shear-Actuated Beam: Application
Library path Structural_Mechanics_Module/Piezoelectric_Effects/
shear_bender.

Using External Materials

If you need to use a constitutive model which is not available among the built-in
material models, it is possible to program it yourself. Such a material function, termed
an external material, is coded in C. If you already have an existing code in another
language like Fortran or C++, it is however possible to write a wrapper function to it.

Before moving to implementing your own material model, there are however two
other options to consider:

• Many material models like hyperelasticity, creep and plasticity have User defined as
one of the options in addition to the standard models. Any material model which
you can describe using built-in variables is most conveniently described here.
• A material model which can partially described in terms of a PDE can often be
implemented using one of the mathematical interfaces. Stresses or strains computed
in that interface are then injected in an existing material model using the External
Stress and External Strain subnodes.

There are two basic types of external material functions: those which completely
replace other material definitions in a domain, and those that just compute an inelastic
strain contribution to be used as part of an existing material model. The former is

INTRODUCTION TO MATERIAL MODELS | 137

 referenced from an External Stress-Strain Relation node, whereas the latter is
referenced from an External Strain subnode.

During the solution, an external material routine is always called for each Gauss point
during evaluation of stiffness matrices and computation of residuals. During result
presentation, the external material can be called from any location in the geometry, as
requested by for example graphs and point evaluations.

Almost invariably, you need to store state variables in the external material, such as for
example plastic strains. The state variables are stored at the Gauss points. If an external
material is called at another location, the state variables will be interpolated to that
location. This means that the state of the material may not be exactly consistent there,
which can lead to some artifacts during result presentation. You can avoid this problem
by using the gpeval operator.

In the COMSOL Multiphysics Reference Manual:

• Working with External Materials

• External Material

LIBRARY OF UTILITY FUNCTIONS

In order to simplify the task of writing the code for an external material, a library of
utility routines is provided. It provides a toolkit for operations common in solid
mechanics such as various tensor operations or computing principal stresses and
equivalent stresses.

LIST OF UTILITY FUNCTIONS

csext_add: Function that adds two matrices

/*
* Function: csext_add
* -------------------
* Description:
* Adds two (3 x 3) matrices 'A' and 'B' and stores the
* result 'A + B' in 'C'.
*
* Arguments:
* double A[3][3]
* double B[3][3]
* double C[3][3] (output)
*
* Return value:

138 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

 * void
*/
void csext_add(double A[3][3],double B[3][3], double C[3][3]);

csext_det: Function that computes the matrix determinant

/*
* Function: csext_det
* -------------------
* Description:
* Computes the determinant of a (3 x 3) matrix 'A' and
* returns the value.
*
* Arguments:
* double A[3][3]
*
* Return value:
* double
*
*/
double csext_det(double A[]3[3]);

csext_dev: Function that computes the deviator of a matrix

/*
* Function: csext_dev
* -------------------
* Description:
* Computes the deviator of a (3 x 3) matrix 'A'
* and stores the value in 'dev'.
*
* Arguments:
* double A[3][3]
* double dev[3][3] (output)
*
* Return value:
* void
*
*/
void csext_dev(double A[3][3], double dev[3][3]);

csext_dot: Function that computes the inner product of two matrices

/*
* Function: csext_dot
* -------------------
* Description:
* Computes the inner product of (3 x 3) matrices 'A' and 'B',

INTRODUCTION TO MATERIAL MODELS | 139

 * and returns 'A : B'.
*
* Arguments:
* double A[3][3]
* double B[3][3]
*
* Return value:
* double
*
*/
double csext_dot(double A[3][3],double B[3][3]);

csext_eig: Function that computes the spectral decomposition of a symmetric matrix

/*
* Function: csext_eig
* -------------------
* Description:
* Computes the eigenvalues and eigenvectors of a
* symmetric (3 x 3) matrix 'A'.
* The eigenvalues are stored in 'vals', sorted with the
* largest value in vals[0].
* The eigenvectors are stored column-wise in 'vecs' with
* the same ordering as the eigenvalues.
* Normal execution returns 0. nonzero means that the
* computation failed.
*
* Arguments:
* double A[3][3]
* double vals[3] (output)
* double vecs[3][3] (output)
*
* Return value:
* int
*
*/
int csext_eig(double A[3][3], double vals[3], double vecs[3][3]);

csext_gl: Function that computes the Green-Lagrange strain tensor

/*
* Function: csext_gl
* ------------------
* Description:
* Computes the Green-Lagrange strain tensor 'egl' based on
* the Right Cauchy-Green deformation tensor 'rcg'.
*
* Arguments:
* double rcg[3][3]

140 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

 * double egl[3][3] (output)
*
* Return value:
* void
*
*/
void csext_gl(double rcg[3][3],double egl[3][3]);

csext_inv: Function that computes the matrix inverse

/*
* Function: csext_inv
* -------------------
* Description:
* Computes the inverse of a (3 x 3) matrix 'A'.
* The inverse, if it exists, is stored in 'inv'.
* 0 is returned if successful, -1 if 'A' is numerically
* singular. The matrix 'A' is considered singular if
* abs(det(A))<tol.
*
* Arguments:
* double A[3][3]
* double tol
* double inv[3][3] (output)
*
* Return value:
* int
*
*/
int csext_inv(double A[3][3], double tol, double inv[3][3]);

csext_lcg: Function that computes the Left Cauchy–Green deformation tensor

/*
* Function: csext_lcg
* -------------------
* Description:
* Computes the Left Cauchy-Green deformation 'lcg' tensor
* based on the deformation gradient 'defgrad'.
*
*
* Arguments:
* double defgrad[3][3]
* double lcg[3][3] (output)
*
* Return value:
* void
*
*/

INTRODUCTION TO MATERIAL MODELS | 141

 void csext_lcg(double defgrad[3][3], double lcg[3][3]);

csext_linsolv: Function to solve a linear system of equations

/*
* Function: csext_linsolv
* -------------------
* Description:
* Solves a linear system of equations, 'Ax = b', for 'x',
* where 'A' is (n x n), and 'b' is (n x 1).
* The maximum allowed size of the system is n = 6.
*
* The solution vector 'x' is stored in 'b' as output.
* The elements of matrix 'A' are not preserved.
*
* If successful, 0 is returned.
* If the solution cannot be determined, -1 is returned.
* If n < 1 or n > 6, -2 is returned.
*
* Arguments:
* double *A
* double *b (input/output)
*
* Return value:
* int
*
*/
CSEXTUTILS_SYMBOLS int csext_linsolv(int n, double *A, double *b);

csext_lpolar: Function that computes the Left polar decomposition of a matrix

/*
* Function: csext_lpolar
* -------------------
* Description:
* Computes the Left polar decomposition F = VR,
* such that the deformation gradient 'defgrad' is
* multiplicatively decomposed into a rotation, 'R',
* and a stretch tensor, 'V'.
* If the polar decomposition fails, -1 is returned.
* If successful, 0 is returned.
*
* Arguments:
* double defgrad[3][3]
* double V[3][3] (output)
* double R[3][3] (output)
*
* Return value:
* int

142 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

 *
*/
int csext_lpolar(double defgrad[3][3], double V[3][3],
double R[3][3]);

csext_mises: Function that computes the von Mises equivalent stress

/*
* Function: csext_mises
* ---------------------
* Description:
* Computes and returns the von Mises equivalent stress based
* on a stress tensor 'sig'.
*
* Arguments:
* double sig[3][3]
*
* Return value:
* double
*
*/
double csext_mises(double sig[3][3]);

csext_mul: Function that multiplies two matrices

/*
* Function: csext_mul
* -------------------
* Description:
* Multiplies two (3 x 3) matrices 'A' and 'B'.
* The result 'AB' is stored in 'C'.
*
* Arguments:
* double A[3][3]
* double B[3][3]
* double C[3][3] (output)
*
* Return value:
* void
*
*/
void csext_mul(double A[3][3], double B[3][3], double C[3][3]);

csext_rcg: Function that computes the Right Cauchy–Green deformation tensor

/*
* Function: csext_rcg
* -------------------

INTRODUCTION TO MATERIAL MODELS | 143

 * Description:
* Computes the Right Cauchy-Green deformation tensor 'rcg'
* based on the deformation gradient 'defgrad'.
*
* Arguments:
* double defgrad[3][3]
* double rcg[3][3] (output)
*
* Return value:
* void
*
*/
void csext_rcg(double defgrad[3][3], double rcg[3][3]);

csext_rpolar: Function that computes the Right polar decomposition of a matrix

/*
* Function: csext_rpolar
* -------------------
* Description:
* Computes the Right polar decomposition F = RU,
* such that the deformation gradient 'defgrad' is
* multiplicatively decomposed into a rotation, 'R',
* and a stretch tensor, 'U'.
* If the polar decomposition fails, -1 is returned.
* If successful, 0 is returned.
*
* Arguments:
* double defgrad[3][3]
* double R[3][3] (output)
* double U[3][3] (output)
*
* Return value:
* int
*
*/
int csext_rpolar(double defgrad[3][3], double R[3][3],
double U[3][3]);

csext_spect: Function that computes a matrix based on its spectral decomposition

/*
* Function: csext_spect
* -------------------
* Description:
* Computes a symmetric (3 x 3) matrix 'A' based on
* its spectral decomposition A=Q*diag*Q^T.
* The matrix 'diag' is diagonal and stores the
* eigenvalues of 'A'. The matrix 'Q' stores the

144 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

 * eigenvectors (column-wise) of 'A', with the ordering
* corresponding to the eigenvalues in 'diag'.
* The vector 'd' stores the diagonal elements of 'diag'.
*
* Arguments:
* double Q[3][3]
* double d[3]
* double A[3][3] (output)
*
* Return value:
* void
*
*/
void csext_spect(double Q[3][3], double d[3], double A[3][3]);

csext_trace: Function that computes the matrix trace

/*
* Function: csext_trace
* ---------------------
* Description:
* Computes and returns the trace of a (3 x 3) matrix 'A'.
*
* Arguments:
* double A[3][3]
*
* Return value:
* double
*
*/
double csext_trace(double A[3][3]);

csext_transp: Function that computes the matrix transpose

/*
* Function: csext_transp
* ----------------------
* Description:
* Computes the transpose of a (3 x 3) matrix 'A'.
* The result is stored in 'transp'.
*
* Arguments:
* double A[3][3]
* double transp[3][3] (output)
*
* Return value:
* void
*
*/

INTRODUCTION TO MATERIAL MODELS | 145

 void csext_transp(double A[3][3], double transp[3][3]);

csext_add_voigt: Function that adds two matrices

/*
* Function: csext_add_voigt
* ---------------------------
* Description:
* Adds two matrices 'A' and 'B' stored on Voigt form.
* The result is stored in 'C'.
*
* Arguments:
* double A[6]
* double B[6]
* double C[6] (output)
*
* Return value:
* void
*/
void csext_add_voigt(double A[6], double B[6], double C[6]);

csext_dev_voigt: Function that computes the deviator of a symmetric matrix

/*
* Function: csext_dev_voigt
* ---------------------------
* Description:
* Computes the deviator of a (symmetric) matrix 'A' stored on
Voigt form.
* The result is stored in 'dev'.
*
* Arguments:
* double A[6]
* double dev[6] (output)
*
* Return value:
* void
*/
void csext_dev_voigt(double A[6], double dev[6]);

csext_dot_voigt: Function that computes the inner product of symmetric matrices

/*
* Function: csext_dot_voigt
* ---------------------------
* Description:
* Computes and returns the dot product (inner product) of two
* symmetric (3 x 3) matrices 'A' and 'B' stored on Voigt form.

146 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

 *
* Arguments:
* double A[6]
* double B[6]
*
* Return value:
* double
*/
double csext_dot_voigt(double A[6], double B[6]);

csext_mises_voigt: Function that computes the von Mises equivalent stress

/*
* Function: csext_mises_voigt
* ---------------------
* Description:
* Computes and returns the von Mises equivalent stress based
* on a stress tensor 'sig' on Voigt form.
*
* Arguments:
* double sig[6]
*
* Return value:
* double
*
*/
double csext_mises_voigt(double sig[6]);

csext_utils_trace_voigt: Function that computes the matrix trace

/*
* Function: csext_trace_voigt
* ---------------------------
* Description:
* Computes and returns the trace of a symmetric (3 x 3) matrix
* 'A' stored on Voigt form.
*
* Arguments:
* double A[6]
*
* Return value:
* double
*/
double csext_trace_voigt(double A[6]);

csext_from_voigt: Function to change from Voigt notation

/*

INTRODUCTION TO MATERIAL MODELS | 147

 * Function: csext_from_voigt
* -------------------
* Description:
* Converts a symmetric (3 x 3) matrix 'A' stored
* on Voigt form to matrix form. The result is stored in 'B'.
* If 'def' = 1, the values of last three elements of 'A' are
* un-altered when passed into 'B'.
* If 'def' = 2, the values of last three elements of 'A' are
* halved when passed into 'B'.
* Returns -1 if the value of 'def' is invalid, 0 otherwise.
*
* Arguments:
* double A[6]
* int def
* double B[3][3] (output)
*
* Return value:
* int
*
*/
int csext_from_voigt(double A[6], int def, double B[3][3]);

csext_to_voigt: Function to change to Voigt notation

/*
* Function: csext_to_voigt
* -------------------
* Description:
* Converts a symmetric (3 x 3) matrix 'A' to Voigt form.
* The result is stored in 'B'.
* If 'def' = 1, the values of the off-diagonal components
* of 'A' are un-altered when passed into 'B'.
* If 'def' = 2, the values of the off-diagonal components
* of 'A' are doubled when passed into 'B'.
* Returns -1 if the value of 'def' is invalid, 0 otherwise.
*
* Arguments:
* double A[3][3]
* int def
* double B[6] (output)
*
* Return value:
* int
*
*/
int csext_to_voigt(double A[3][3], int def, double B[6]);

148 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

csext_jac_conv: Function to convert the Jacobian
/*
* Function: csext_jac_conv
* ------------------------
* Description:
* Converts a Jacobian from 'dSde' to 'dSdF' using the deformation
* gradient 'defgrad', where:
* - 'S' is the 2:nd Piola-Kirchhoff stress tensor,
* - 'e' is the Green-Lagrange strain tensor,
* - 'F' is the deformation gradient.
* If 'def' = 1, dSde is defined using tensor shears.
* If 'def' = 2, dSde is defined using engineering shears.
* Returns -1 if the value of 'def' is invalid.
*
* Arguments:
* double dSde[6][6]
* int def
* double defgrad[3][3]
* double dSdF[6][9] (output)
*
* Return value:
* int
*
*/
int csext_jac_conv(double dSde[6][6], int def,
double defgrad[3][3], double dSdF[6][9]);

csext_unit: Function to define a unit matrix

/*
* Function: csext_unit
* -------------------
* Description:
* Initializes an (n x n) matrix 'A' to the identity matrix.
* Returns -1 if n < 1,
* 0 otherwise.
*
* Arguments:
* int n
* double *A (output)
*
* Return value:
* int
*
*/
int csext_unit(int n, double *A);

INTRODUCTION TO MATERIAL MODELS | 149

 csext_zero: Function to initialize a matrix
/*
* Function: csext_zero
* -------------------
* Description:
* Initializes an (m x n) matrix 'A' to zero.
* Returns -1 if m < 1 or n < 1,
* 0 otherwise.
*
* Arguments:
* int m
* int n
* double *A (output)
*
* Return value:
* void
*
*/
int csext_zero(int m, int n, double *A);

References
1. D. Chapelle and K.J. Bathe, The Inf-Sup test, Comp. Struct., Vol. 47, No. 4/5, pp.
537–545, 1993.

150 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

M o de li ng P i e z oel ect ri c Prob l ems
In this section:

• About Piezoelectric Materials

• Piezoelectric Coupling
• Create the Piezoelectric Effect Interface and Define Domains
• Complete Settings of Piezoelectric Materials
• Add Damping and Loss
• Define Material Properties

• Multiphysics Modeling Workflow in the COMSOL Multiphysics

Reference Manual.
• Piezoelectric Coupling
• Piezoelectricity in the theory section

About Piezoelectric Materials

Piezoelectric materials become electrically polarized when strained. From a
microscopic perspective, the displacement of atoms within the unit cell (when the solid
is deformed) results into electric dipoles within the medium. In certain crystal
structures, this combines to give an average macroscopic dipole moment or electric
polarization. This effect, known as the direct piezoelectric effect, is always
accompanied by the converse piezoelectric effect, in which the solid becomes strained
when placed in an electric field.

Within a piezoelectric, there is a coupling between the strain and the electric field,
which is determined by the constitutive relation:

T
S = sET + d E
(2-14)
D = dT +  T E

Here, the naming convention used in piezoelectricity theory is assumed: S is the strain,
T is the stress, E is the electric field, and D is the electric displacement field. The
material parameters sE, d, and T, correspond to the material compliance, the coupling
properties and the permittivity. These quantities are tensors of rank 4, 3, and 2,

MODELING PIEZOELECTRIC PROBLEMS | 151

 respectively. The tensors, however, are highly symmetric for physical reasons, and they
can be represented as matrices within an abbreviated subscript notation, which is
usually more convenient. In the Piezoelectricity interface, the Voigt notation is used,
which is a standard in the literature for piezoelectricity but which differs from the
defaults in the Solid Mechanics interface.

The Piezoelectric Material uses the structural mechanics nomenclature.

The strain is named  (instead of S) and the stresses are denoted by either
 or S (instead of T). This makes the names consistent with those used in
the other structural mechanics interfaces.

Equation 2-14 will, using the notation from structural mechanics, then read

T
 = sE  + d E
(2-15)
D = d +  0  rT E

Equation 2-14 (or Equation 2-15) is known as the strain-charge form of the
constitutive relations. The equation can be re-arranged into the stress-charge form,
which relates the material stresses to the electric field:

T
 = cE  – e E
(2-16)
D = e +  0  rS E

The material properties, cE, e, and S are related to sE, d, and T. It is possible to use
either form of the constitutive relations. In addition to Equation 2-14 or
Equation 2-16, the equations of solid mechanics and electrostatics must also be solved
within the material.

• Piezoelectric Coupling
• Modeling Piezoelectric Problems
• Piezoelectricity in the theory section

Piezoelectric Material Orientation

The orientation of a piezoelectric crystal cut is frequently defined by the system
introduced by the IRE standard of 1949 (Ref. 8). This standard has undergone a
number of subsequent revisions, with the final revision being the IEEE standard of

152 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

1987 (Ref. 9). Unfortunately, the 1987 standard contained a number of serious errors
and the IEEE subsequently withdrew it. COMSOL therefore adopts the preceding
1978 standard (Ref. 10), which is similar to the 1987 standard, for material property
definitions. Most of the material properties in the material library are based on the
values given in the book by Auld (Ref. 11), which uses the 1978 IEEE conventions.
This is consistent with general practice except in the specific case of quartz, where it is
more common to use the 1949 IRE standard to define the material properties.
COMSOL therefore provides additional sets of material properties consistent with this
standard for the case of quartz. Note that the material properties for quartz are based
on Ref. 12, which uses the 1949 IRE standard (the properties are appropriately
modified according to the different standards).

The stiffness, compliance, coupling, and dielectric material property matrices are
defined with the crystal axes aligned with the local coordinate axes. Note that the signs
of several matrix components differ between the 1949 and the 1978 standards (see
Table 2-8). In the absence of a user-defined coordinate system, the local system
corresponds to the global X, Y, and Z coordinate axes. When an alternative coordinate
system is selected this system defines the orientation of the crystal axes. This is the
mechanism used in COMSOL to define a particular crystal cut, and typically it is
necessary to calculate the appropriate Euler angles for the cut (given the thickness
orientation for the wafer). All piezoelectric material properties are defined using the
Voigt form of the abbreviated subscript notation, which is universally employed in the
literature (this differs from the standard notation used for the Solid Mechanics
interface material properties). The material properties are defined in the material
frame, so that if the solid rotates during deformation the material properties rotate
with the solid. See Modeling Geometric Nonlinearity.

Crystal cuts are usually defined by a mechanism introduced by the IEEE/IRE

standards. Both standards use a notation that defines the orientation of a virtual slice
(the plate) through the crystal. The crystal axes are denoted X, Y, and Z and the plate,
which is usually rectangular, is defined as having sides l, w, and t (length, width, and
thickness). Initially the plate is aligned with respect to the crystal axes and then up to
three rotations are defined, using a right-handed convention about axes embedded
along the l, w, and t sides of the plate. Taking AT cut quartz as an example, the IEEE
1978 standard defines the cut as: (YXl) 35.25°. The first two letters in the bracketed
expression always refer to the initial orientation of the thickness and the length of the
plate. Subsequent bracketed letters then define up to three rotational axes, which move
with the plate as it is rotated. Angles of rotation about these axes are specified after the
bracketed expression in the order of the letters, using a right-handed convention. For

MODELING PIEZOELECTRIC PROBLEMS | 153

 AT cut quartz only one rotation, about the l axis, is required. This is illustrated in
Figure 2-10. Note that within the 1949 IRE Standard AT cut quartz is denoted as:
(YXl) 35.25°, since the X-axis is rotated by 180° in this convention and positive
angles therefore correspond to the opposite direction of rotation (see Figure 2-9).
Table 2-9 summarizes the differences between the standards for different crystal cuts.

When defining the material properties of Quartz, the orientation of the

X, Y, and Z axes with respect to the crystal differs between the 1987 IEEE
standard and the 1949 IRE standard. Figure 2-9 shows the alignment of
the axis for the case of right-handed quartz. A consequence of this is that
both the material property matrices and the crystal cuts differ between the
two standards. Table 2-8 summarizes the signs for the important matrix
elements under the two conventions. Table 2-9 shows the different
definitions of the crystal cuts under the two conventions.

TABLE 2-8: SIGNS FOR THE MATERIAL PROPERTIES OF QUARTZ, WITHIN THE TWO STANDARDS COMMONLY
EMPLOYED.

IRE 1949 STANDARD IEEE 1978 STANDARD

MATERIAL RIGHT HANDED LEFT HANDED RIGHT HANDED LEFT HANDED

PROPERTY QUARTZ QUARTZ QUARTZ QUARTZ

s14 + + - -
c14 - - + +
d11 - + + -
d14 - + - +
e11 - + + -
e14 + - + -

TABLE 2-9: CRYSTAL CUT DEFINITIONS FOR QUARTZ CUTS WITHIN THE TWO STANDARDS COMMONLY
EMPLOYED AND THE CORRESPONDING EULER ANGLES FOR DIFFERENT ORIENTATIONS OF THE CRYSTAL
THICKNESS.

STANDARD REPRESENTATION AT CUT BT CUT

IRE 1949 Standard (YXl) +35.25 (YXl) 49

Y-thickness Euler (ZXZ: 035.250) (ZXZ: 0490)
angles
Z-thickness Euler (ZXZ: 0°,125.250) (ZXZ: 0 410)
angles
IEEE 1978 Standard (YXl) 35.25 (YXl) +49

154 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

TABLE 2-9: CRYSTAL CUT DEFINITIONS FOR QUARTZ CUTS WITHIN THE TWO STANDARDS COMMONLY
EMPLOYED AND THE CORRESPONDING EULER ANGLES FOR DIFFERENT ORIENTATIONS OF THE CRYSTAL
THICKNESS.

STANDARD REPRESENTATION AT CUT BT CUT

Y-thickness Euler (ZXZ: 0 35.250) (ZXZ: 0 490)

angles
Z-thickness Euler (ZXZ: 0 54.750) (ZXZ: 0 1390)
angles
When defining the material orientation, it is necessary to consider the orientation of
the plate with respect to the global coordinate system in addition to the orientation of
the plate with respect to the crystallographic axes. Consider the example of AT cut
quartz in Figure 2-10. The definition of the appropriate local coordinate system
depends on the desired final orientation of the plate in the global coordinate system.
One way to set up the plate is to orientate its normal parallel to the Y axis in the global
coordinate system. Figure 2-11 shows how to define the local coordinate system in this
case. Figure 2-12 shows how to define the local system such that the plate has its
normal parallel to the global Z axis. In both cases it is critical to keep track of the
orientation of the local system with respect to the global system, which is defined
depending on the desired orientation of the plate in the model.

There are also a number of methods to define the local coordinate system with respect
to the global system. Usually, it is most convenient to define the local coordinates with
a Rotated System node, which defines three Euler angles according to the ZXZ
convention (rotation about Z, then X, then Z again). Note that these Euler angles
define the local (crystal) axes with respect to the global axes — this is distinct from the
approach of defining the cut (global) axes with respect to the crystal (local) axes.

MODELING PIEZOELECTRIC PROBLEMS | 155

 Figure 2-9: Crystallographic axes defined for right-handed quartz in COMSOL and the
1978 IEEE standard (color). The 1949 standard axes are shown for comparison (gray).

Figure 2-9 is reproduced with permission from: IEEE Std 176-1987 -

IEEE Standard on Piezoelectricity, reprinted with permission from
IEEE, 3 Park Avenue, New York, NY 10016-5997 USA, copyright 1987,
by IEEE. This figure must not be reprinted or further distributed without
prior written permission from the IEEE.

156 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

Figure 2-10: Definition of the AT cut of quartz within the IEEE 1978 standard. The AT
cut is defined as: (YXl) 35.25°. The first two bracketed letters specify the initial
orientation of the plate, with the thickness direction, t, along the crystal Y axis and the
length direction, l, along the X axis. Then up to three rotations about axes that move with
the plate are specified by the corresponding bracketed letters and the subsequent angles. In
this case only one rotation is required about the l axis, of 35.25° (in a right-handed
sense).

MODELING PIEZOELECTRIC PROBLEMS | 157

 Figure 2-11: Defining an AT cut crystal plate within COMSOL, with normal in the
global Y direction. Within the 1978 IEEE standard the AT cut is defined as (YXl)
-35.25°. Start with the plate normal or thickness in the Ycr direction (a) and rotate the
plate 35.25° about the l axis (b). The global coordinate system rotates with the plate.
Finally rotate the entire system so that the global coordinate system is orientated as it
appears in COMSOL (c). The local coordinate system should be defined with the Euler
angles (ZXZ - 0, 35.25°, 0). (d) shows a coordinate system for this system in COMSOL.

158 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

Figure 2-12: Defining an AT cut crystal plate within COMSOL, with normal in the
global Z direction. Within the 1978 IEEE standard the AT cut is defined as (YXl)
35.25°. Begin with the plate normal in the Zcr direction, so the crystal and global systems
are coincident. Rotate the plate so that its thickness points in the Ycr direction (the starting
point for the IEEE definition), the global system rotates with the plate (b). Rotate the plate
35.25° about the l axis (d). Finally rotate the entire system so that the global coordinate
system is orientated as it appears in COMSOL (d). The local coordinate system should be
defined with the Euler angles (ZXZ: 0, -54.75°, 0). (e) shows a coordinate system for this
system in COMSOL.

MODELING PIEZOELECTRIC PROBLEMS | 159

 Piezoelectric Losses
Losses in piezoelectric materials can be generated both mechanically and electrically.

In the frequency domain, these can be represented by introducing complex material

properties in the elasticity and permittivity matrices. Taking the mechanical case as an
example, this introduces a phase lag between the stress and the strain, which
corresponds to a Hysteretic Loss. These losses can be added to the Piezoelectric
Material by three subnodes: Mechanical Damping, Coupling Loss, and Dielectric Loss.
The losses typically defined as loss factors (see below).

The hysteretic electrical losses are usually used to represent high frequency electrical
losses that occur as a result of friction impeding the rotation of the microscopic dipoles
that produce the material permittivity.

Low frequency losses, corresponding to a finite material conductivity, can be added to

the model through an Electrical Conductivity (Time Harmonic) subnode. This feature
operates only in the frequency domain.

In the time domain, the losses can be added by:

• Using the Rayleigh Damping option in the Mechanical Damping and Coupling Loss
subnodes to the Piezoelectric Material,.

Rayleigh Damping

• Using either the Dispersion, or Complex permittivity, or Maximum Loss Tangent option
in the Dielectric Loss subnode to the Piezoelectric Material.

These types of damping are also available in the frequency domain.

Rayleigh Damping

HYSTERETIC LOSS
In the frequency domain, the dissipative behavior of the material can be modeled using
complex-valued material properties, irrespective of the loss mechanism. Such hysteretic
losses can be applied to model both electrical and mechanical losses. For more
information about hysteretic losses, see Ref. 1 to Ref. 4.

160 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

For the case of piezoelectric materials, this means that the constitutive equations are
written as follows.

For the stress-charge formulation

T = c˜ E S – e˜ E
T

D = e˜ T +  ̃ E
0 rS

where T is the stress tensor, and S is the strain tensor.

For the strain-charge formulation,

S = s˜ E T + d˜ E
T

D = d˜ S +  ̃ E
0 rT

All constitutive matrices in the above equation can be complex-valued matrices, where
the imaginary parts define the dissipative function of the material.

Both the real and complex parts of the material data must be defined so as to respect
the symmetry properties of the material being modeled and with restrictions imposed
by the laws of physics.

A key requirement is that the dissipation density is positive; that is, there
is no power gain from the passive material. This requirement sets rules for
the relative magnitudes for all material parameters. This is important
when defining the coupling losses.

In COMSOL, you can enter the complex-valued data directly or by means of loss
˜
factors. When loss factors are used, the complex data X is represented as pairs of a
real-valued parameter

˜
X = real  X 

and a loss factor

˜ ˜
 X = imag  X   real  X 

the ratio of the imaginary and real part, and the complex data is then:

˜
X = X  1  j X 

MODELING PIEZOELECTRIC PROBLEMS | 161

 where the sign depends on the material property used. The loss factors are specific to
the material property, and thus they are named according to the property they refer to,
for example, cE. For a structural material without coupling, simply use s, the
structural loss factor.

The loss factors are defined so that a positive loss factor value usually corresponds to a
positive loss. The complex-valued data is then based on sign rules.

By default, there is no damping until at least one of the damping and losses related
subnodes is added.

For the Piezoelectric Material node, the following equations apply via the
corresponding three subnodes:

Mechanical Damping
m n m n m n
c˜ E = c E  1 + j cE 
m n m n m n
s˜ E = s E  1 – j sE 

where m and n refer to components of each matrix.

Coupling Loss
m n m n m n
e˜ =e  1 + j e 
m n m n m n
d˜ =d  1 + j d 

Dielectric Loss
m n m n m n
̃ rS =  rS  1 – j S 
m n m n m n
̃ rT =  rT  1 – j T 

Note that the multiplication is applied component-wise.

The loss factors can also be entered as scalar isotropic factors independently of the
material and the other coefficients.

A good check on the chosen values is to compute a number of eigenfrequencies,

possibly using some different sets of boundary conditions. All computed
eigenfrequencies must have a positive imaginary part in order to represent a damped
motion.

162 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

In practice, it is often difficult to find complex-valued data for each of the matrix
elements in the literature. Measuring the losses independently is a challenging task.

Two important particular cases are available in COMSOL, for which the
dielectric loss data can be entered based on experimentally measurable
quantities:

• Complex permittivity
• Maximum Loss Tangent

DIELECTRIC DISPERSION
The Dielectric Loss subnode can be set to use the Dispersion option. In such case, the
following equations need to be solved in the time domain:

D
 + J p = 0 (2-17)
t 

J p E
Jp + d =  0  rS (2-18)
t t

where you can specify two material parameters: the relaxation time d and the relative
permittivity increment rS. The latter can be either a matrix or a scalar quantity. This
model is a one-term version of the more general Debye dispersion model, Ref. 13.

The constitutive relation is assumed as

D = eS +  0   E (2-19)

where S is the strain tensor, and   is the relative permittivity in the high frequency
limit (that is, for excitations with a characteristic time much shorter than the relaxation
time d).

The parent Piezoelectric Material node has an input for the relative
permittivity, rS, which is used in stationary study. You can chose how this
input will be interpreted in the dispersion computations. The options are:

• High frequency limit. In such case,   =  rS will be used.

• Low frequency limit. In such case,   =  rS –  rS will be used. Note
that for consistency,   must be always positive valued.

MODELING PIEZOELECTRIC PROBLEMS | 163

 With the absence of free electric charges, Equation 2-17 and Equation 2-18 can be
combined and integrated in time to yield the following equation:

D
   D +  d +  0  rS E = 0
t

This is the equation form used in COMSOL Multiphysics for time-dependent analysis.

For the eigenfrequency and frequency domain analyses, the corresponding equation is:

    1 + j d  jD +  0  rS jE  = 0

In most cases, i can be factored out, so that the following equation is solved:

   D + j d  D +  0  rS E  = 0

This equation, together with the constitutive relation Equation 2-19 gives

   eS +  0  ' – j'' E  = 0 (2-20)

where

 rS
' =   + --------------------------2-
1 +  d  

and

 d  0 
rS
'' = ---------------------------2-
1 +  d  

Equation 2-20 shows how the dispersion parameters contribute to the polarization
and losses. Thus, the effective relative permittivity decreases with the excitation
frequency from the low frequency limit   +  rS down to the high frequency limit
  . The damping effect vanishes for both large and small frequencies, and it reaches
the maximum for   1d.

If your license includes either the AC/DC Module or MEMS Module,

more options for modeling dielectric dispersion can be found in
Dispersion section in the AC/DC Module User’s Guide.

The following two sections present two cases, for which the dielectric dispersion data
can be related to other experimentally measurable quantities. Both cases, can be used

164 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

in Eigenfrequency, Frequency Domain, and Time Dependent study. The software will
apply the dispersion model equations, for which the effective relaxation time and
relative permittivity increment are computed automatically based on the node input
parameters.

Complex permittivity
In this case, the complex relative permittivity is known at a certain reference frequency

'  f ref  – j''  f ref 

The relaxation time and relative permittivity contribution cane be computed as

–1 –1
 d =  ref ''  ' –   

2 –1
 rS = ' –   +  ''   ' –   

where  ref = 2f ref .

If the relative permittivity rS (input on the Piezoelectric Material parent node) is
selected to represent the low frequency limit, one has

2 –1
  = ' –  ''    rS – ' 

If rS is selected to represent the high frequency limit, one can simply use   =  rS
instead.

Maximum Loss Tangent

The loss tangent is defined as a function of the frequency

  f  = tan  = ''  '

so that the complex relative permittivity can be written as

 r  f  = '  I – j 

For many materials, the loss tangent reaches a maximum at certain frequency fref
within the frequency range of interested

 max = max  f  =   f ref 

The relaxation time cane be computed as

–1 2 12
 d =  ref   max +   max + I  

MODELING PIEZOELECTRIC PROBLEMS | 165

 where  ref = 2f ref .

If the relative permittivity rS (input on the Piezoelectric Material parent node) is
selected to represent the low frequency limit, the relative permittivity contribution is
computes as

 rS =  rS –  
–2
where   =   d  ref   rS .

If the relative permittivity rS is selected to represent the high frequency limit, it is
computed as

2
 rS =    d  ref  – I  

where   =  rS .

ELECTRICAL CONDUCTIVITY (TIME HARMONIC)

For frequency domain and eigenfrequency analyses, the effect of electrical conductivity
of the piezoelectric material (see Ref. 2, Ref. 5, and Ref. 6) can be included. Thus, in
addition to the displacement current, the conduction electric current term is used

   jD + J e  = 0

Je = e E

where e is the material electrical conductivity, and E is the electric field. The above
form of the equation is used for the eigenfrequency analysis in COMSOL
Multiphysics.

When the conduction loss is applied, the default eigenvalue analysis will
in most cases return a number of pure imaginary eigenfrequencies. To
avoid this, you can configure the solver to search for eigenvalues with real
part larger than zero.

Do not use any dielectric loss factor damping together with the
conduction loss in an eigenfrequency analysis.

166 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

 In the COMSOL Multiphysics Reference Manual:

• Selecting a Stationary, Time-Dependent, or Eigenvalue Solver

• Eigenvalue Solver

For the frequency domain analysis, the angular frequency is just a parameter, and the
equation can be transformed into

–1
   D – j  e E  = 0

which allows you to use both a dielectric loss factor and electrical conductivity in a
frequency response study. In such case, ensure that the loss factor refers to the
alternating current loss tangent, which dominates at high frequencies, where the effect
of ohmic conductivity vanishes (Ref. 7).

Conduction loss can be combined with Dielectric Dispersion for both eigenfrequency
and frequency domain analyses. The following equation forms are used, respectively,
in the frequency domain:

–1
   D + j d  D +  0  rS E +   d – j J e  = 0

and in eigenfrequency analyses:

    1 + j d  jD +  0  rS jE +  1 + j d  J e  = 0

References for Piezoelectric Damping

1. R. Holland and E.P. EerNisse, Design of Resonant Piezoelectric Devices, Research
Monograph No. 56, The M.I.T. Press, 1969.

2. T. Ikeda, Fundamentals of Piezoelectricity, Oxford University Press, 1990.

3. A.V. Mezheritsky, “Elastic, Dielectric, and Piezoelectric Losses in Piezoceramics:

How it Works all Together,” IEEE Transactions on Ultrasonics, Ferroelectrics, and
Frequency Control, vol. 51, no. 6, 2004.

4. K. Uchino and S. Hirose, “Loss Mechanisms in Piezoelectrics: How to Measure

Different Losses Separately,” IEEE Transactions on Ultrasonics, Ferroelectrics, and
Frequency Control, vol. 48, no. 1, pp. 307–321, 2001.

MODELING PIEZOELECTRIC PROBLEMS | 167

 5. P.C.Y. Lee, N.H. Liu, and A. Ballato, “Thickness Vibrations of a Piezoelectric Plate
With Dissipation,” IEEE Transactions on Ultrasonics, Ferroelectrics, and Frequency
Control, vol. 51, no. 1, 2004.

6. P.C.Y. Lee and N.H. Liu, “Plane Harmonic Waves in an Infinite Piezoelectric Plate
With Dissipation,” Frequency Control Symposium and PDA Exhibition, pp. 162–
169, IEEE International, 2002.

7. C.A. Balanis, “Electrical Properties of Matter,” Advanced Engineering

Electromagnetics, John Wiley & Sons, chapter 2, 1989.

8. “Standards on Piezoelectric Crystals, 1949”, Proceedings of the I. R. E.,vol. 37,

no.12, pp. 1378–1395, 1949.

9. IEEE Standard on Piezoelectricity, ANSI/IEEE Standard 176-1987, 1987.

10. IEEE Standard on Piezoelectricity, ANSI/IEEE Standard 176-1978, 1978.

11. B.A. Auld, Acoustic Fields and Waves in Solids, Krieger Publishing, 1990.

12. R. Bechmann “Elastic and Piezoelectric Constants of Alpha-Quartz,” Physical

Review B, vol. 110 no. 5, pp. 1060–1061, 1958.

13. N.S. Stoykov, T.A. Kuiken, M.M. Lowery, and A. Taflove, “Finite-element
time-domain algorithms for modeling linear Debye and Lorentz dielectric dispersions
at low frequencies,” IEEE Transactions on Biomedical Engineering, vol. 50, no. 9,
pp. 1100–1107, 2003.

Piezoelectric Coupling
The piezoelectric effect is an interaction between the mechanical and electrical physics,
where a stress applied on a piezoelectric material generates a voltage (direct effect) or
a voltage applied on it generates the deformation of the material (inverse effect). In
COMSOL Multiphysics, the Piezoelectricity interface is constituted of one Solid
Mechanics and one Electrostatics interface, which are coupled together by a Piezoelectric
Effect multiphysics feature. Hence a piezoelectric problem contains solid and
electrostatic domains, with at least one domain shared by the two physics interfaces
and with the piezoelectric coupling defined on it.

168 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

Create the Piezoelectric Effect Interface and Define Domains
A piezoelectric problem can be set up in different ways:

• By selecting Piezoelectricity from the Model Wizard,

• By choosing Piezoelectricity from the Add Physics menu when working in an existing
model, or
• By adding the corresponding features to create the coupling manually.

In the first two cases, by default all the domains in the model are assumed to be
piezoelectric materials.

When setting up the problem manually (that is, by adding single physics interfaces one
at a time) both Solid Mechanics and Electrostatics interfaces should be added. Then, you
have to specify which domains are in each physics, and which domains are to be
modeled as piezoelectric materials.

1 On the Solid Mechanics interface Settings window, locate the Domain Selection
section. Select the domains which undergo structural deformation, including the
piezoelectric material domains.
2 Go to the Solid Mechanics>Piezoelectric Material node (if it is not yet available, add
it). Select the domains where the piezoelectric effect applies. Domains which are not
piezoelectric can be modeled using other available material models.
3 Go to the Electrostatics node and under Domain Selection select the domains the
electrostatics equations must be solved. These domains include all the piezoelectric
domains, and any other insulating domains.

Since metals usually have a conductivity several orders of magnitude

higher than other materials, it is normal not to solve the electrostatics
equations in their corresponding domains, but instead the surfaces of the
material are represented as isopotentials with an appropriate boundary
condition (usually a terminal or floating potential).

4 Go to the Electrostatics>Charge Conservation, Piezoelectric node (if it is not yet

available, add it). Select the domains where the piezoelectric effect has to be
modeled.
5 A Multiphysics>Piezoelectric Effect node is already present if the coupling was added
using either the Model Wizard or Add Physics window. If the model is set up manually

MODELING PIEZOELECTRIC PROBLEMS | 169

 (that is, single physics interfaces are added), right-click the Multiphysics Couplings
node to add a Piezoelectric Effect coupling.
6 Confirm that all the domains where the piezoelectric material is present are selected.

Only domains that have both Charge Conservation, Piezoelectric selected in

the Electrostatics interface and Piezoelectric Material selected in the Solid
Mechanics interface are selected. The selection of this feature cannot be
edited. If several Solid Mechanics or Electrostatics interfaces are present,
select the correct ones.

7 Confirm that remaining domains are well assigned:

- Solid and electrostatic domains: In these domains, the electrostatics and
structural problems are solved independently without any piezoelectric coupling.
This is the case for insulators.
- Solid-only domains: In these domains, only mechanical phenomena are modeled,
and the electrostatics phenomena are neglected. Metals are typically modeled in
this way, because their conductivity is so high that their surfaces can be treated as
isopotential surfaces within the electrostatics problem. In some cases, insulators
are modeled using these settings, for example, when there is no potential applied
across the domain, and correspondingly solving the electrostatics equations
would produce a constant potential and waste computing resources.
- Electrostatics-only domains: Those domains are nonsolid. This is typically the
case of air, in which the electrostatics is solved but not mechanics.
- Nonsolid and nonelectrostatic domains: In those domains, another physics is
solved. A typical example is an acoustic domain in the case of piezoacoustic
modeling.

Complete Settings of Piezoelectric Materials

Go to the Solid Mechanics>Piezoelectric Material node. On the Settings window
complete these settings:

• Coordinate System Selection section: The material is poled in the x3 direction of the
coordinate system (x1, x2, x3) specified in this section. By default, it is set to the
global coordinate system. If the piezoelectric material is poled along another
direction, you need to define a coordinate system so that its third direction is aligned

170 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

 with the poling direction. Then, assign it as the coordinate system which orients the
material in the Coordinate System Selection section.

If a given piezoelectric material is present with several orientations (such

as stacked piezoelectric disks) you need to define several Piezoelectric
Material nodes and to assign a different coordinate system for each of
them.

• Piezoelectric Material Properties: Select whether the constitutive relation of

piezoelectric material is in Stress-charge or in Strain-charge form. This choice
defines the type of material properties that will be used.
• Geometric Nonlinearity: select the check box to force strains to be linear.
• Energy Dissipation: Select the check box to enable the calculation of the dissipated
energy.

Add Damping and Loss

In the physics toolbar you can add attributes to the Piezoelectric Material node,
especially the following damping and loss contributions:

• Mechanical Damping: Specify the domains of application, then choose if you want to
define a loss factor for cE, a loss factor for sE (in Strain-charge form), an isotropic
loss factor, or a Rayleigh damping.
• Dielectric Loss: Specify the domains of application, then choose if you want to define
a loss factor for rS, a loss factor for rT (in Strain-charge form), or dispersion.
• Coupling Loss: Specify the domains of application, then choose if you want to define
a loss factor for e, a loss factor for d (in Strain-charge form), or Rayleigh damping.
• Conduction Loss (Time-Harmonic): Specify the domains of application, then choose
how you want to define the Electrical conductivity.

Define Material Properties

Define material properties for the piezoelectric materials. The material library contains
several common piezoelectric materials under the piezoelectric section. If you want to
define your own piezoelectric material, you need to specify its properties by hand. The
required properties depend on whether the constitutive relations are in Stress-charge
or Strain-charge form, and which damping and loss attributes are created. Defining all

MODELING PIEZOELECTRIC PROBLEMS | 171

 piezoelectric settings before materials preselects the required properties and makes the
completion easier:

MECHANICAL PROPERTIES
3
• Density rho (SI unit: kg/m )
• Elasticity matrix cE (SI unit: Pa) in Stress-charge form.
• Compliance Matrix sE (SI unit: 1/Pa) in Strain-charge form.

ELECTROSTATIC PROPERTIES
• Relative Permittivity rS (dimensionless) in Stress-charge form.
• Relative Permittivity rT (dimensionless) in Strain-charge form.

COUPLING PROPERTIES
• Coupling matrix eES (SI unit: C/m2) in Stress-Charge form.
• Coupling matrix dET (SI unit: C/N) in Strain-charge form.

DAMPING AND LOSS PROPERTIES

• Either loss factor cE for elasticity matrix cE, or the loss factor sE for compliance
matrix sE: required when Mechanical Damping is present. Both are dimensionless,
the latter is valid only in Strain-Charge form.
• Either loss factor  S for electrical permittivity matrix rS, or loss factor  T for
electrical permittivity matrix rT: required when Dielectric Loss is present. TBoth
are dimensionless, the latter is valid only in Strain-Charge form.
• Either loss factor e for coupling matrix e, or loss factor d for coupling matrix d:
required when Coupling Loss is present. Both are dimensionless, the latter is valid
only in Strain-Charge form.
• Electrical conductivity e (SI unit: S/m): required when Conduction Loss is
present.

Working with Materials and Piezoelectric Materials Library in the

COMSOL Multiphysics Reference Manual.

172 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

Coupling Piezoelectricity with Acoustics
Using piezoelectric materials for an acoustic application is common, such as in sonars,
microphones, sensors, and so forth. This is why coupling piezoelectric devices with
acoustic domains is of particular interest for these applications.

Compared to a single piezoelectric model, you need to add a pressure acoustics

interface, for example, Pressure Acoustics, Frequency Domain or Pressure Acoustics,
Transient (depending on which study type you want to use) plus an Acoustic-Structure
Boundary coupling under the Multiphysics Couplings node. You can also directly create
the nodes that are needed for coupling by adding an Acoustic-Piezoelectric Interaction
interface from the Model Wizard or Add Physics windows. If solid and acoustic domains
are correctly defined, then the right coupling boundaries are automatically selected.
Then specify domains of application for each physics.

• Select solid domains and Piezoelectric Material domains in Solid Mechanics.

• Select electrostatic domains and Charge Conservation, Piezoelectric domains in
Electrostatics.
• Select acoustic domains in a Pressure Acoustics node.
• Under the Multiphysics branch, confirm that selections for Piezoelectric Effect and
Acoustic-Structure Boundary are the right ones. If several Pressure Acoustics, Solid
Mechanics, or Electrostatics interfaces are present, select the right ones that need to
be coupled in the multiphysics interfaces.
• Continue the modeling process by entering the settings for each physics interface
and feature and define materials.

MODELING PIEZOELECTRIC PROBLEMS | 173

 Modeling Electrostrictive and
Ferroelectroelastic Materials
Electrostriction Coupling
Electrostriction is an interaction between the mechanical and electric physics, where an
electric field applied on electrostrictive material generates the deformation of the
material (direct effect), and a stress applied on it changes the material polarization
(inverse effect). In contrast to linear piezoelectricity, the electrostrictive strain induced
in the polarized material is proportional to the square of the polarization. Thus, the
strain will remain the same if the direction of the applied electric field is changed to the
opposite.

Ferroelectricity
The ferroelectroelasticity and ferroelectricity phenomena are related to phase
transitions in materials. In its ferroelectric phase, the material exhibits spontaneous
polarization, so that it is constituted of domains with nonzero polarization even at zero
applied field. This is similar to permanent magnetism in ferromagnetics, which explains
the name used for such materials. Electrostriction in ferroelectroelastic materials can
be related to the domain rotation. Thus, the applied electric field can both rearrange
the domains resulting into the net polarization and rotate the domains mechanically.
Thus, the material extends in the direction of the electric field and contracts in the
direction perpendicular to the field. The domain rotation can be affected by an applied
mechanical stress, which also results into the effective polarization. At very large
electric fields, the electrostrictive effect saturates, as all ferroelectric domains in the
material are aligned along the direction of the applied field. Domain wall interactions
can also lead to a significant hysteresis in the polarization and strain.

Many piezoelectric materials exhibit such nonlinear ferroelectroelastic behavior at large

applied electric fields.

Multiphysics Interfaces
In COMSOL Multiphysics, the electrostrictive effect can be modeled using two
multiphysics interfaces: Electrostriction and Ferroelectroelasticity. Both interfaces are
constituted of one Solid Mechanics and one Electrostatics interface, which are coupled

174 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

together via an Electrostriction multiphysics feature. However, in case of
Ferroelectroelasticity, the corresponding Electrostatics interface will contain a dedicated
feature Charge Conservation, Ferroelectric.

The Electrostriction interface electrostatically can be used for modeling electrostatically

actuated structures in regimes when the electric polarization can be assumed to vary
linear with the applied electric field.

The Ferroelectroelasticity can be used for modeling nonlinear electromechanical

interaction in ferroelectric and piezoelectric materials at high applied electric fields.
Electric polarization in such materials depends nonlinearly on the applied electric field
including possible hysteresis and saturation effects.

Ferroelectricity in the AC/DC Module User’s Guide.

Electrostriction and Ferroelectroelasticity in the Structural Mechanics

Module User’s Guide.

MODELING ELECTROSTRICTIVE AND FERROELECTROELASTIC MATERIALS | 175

 Modeling Magnetostrictive Materials
Magnetostriction Coupling
Magnetostriction is an interaction between the mechanical and magnetic physics,
where a magnetic field applied on magnetostrictive material generates the deformation
of the material (direct effect or Joule effect), while a mechanical stress applied on it
changes the material magnetization (inverse effect or Villari effect).

The magnetostrictive strain has a nonlinear dependence on the magnetic field and the
mechanical stress in the material. However, the effect can be modeled using linear
coupled constitutive equations if the response of the material consists of small
deviations around an operating point (bias point). This type of coupling is reffed to as
Piezomagnetic Effect.

In COMSOL Multiphysics, there are two multiphysics interfaces for modeling either
linear or nonlinear magnetostriction, respectively:

• The Piezomagnetism Interface

• The Nonlinear Magnetostriction Interface

Both multiphysics interfaces, the Magnetostriction interfaces are constituted of one

Solid Mechanics and one Magnetic Fields interface, which are coupled together via either
Piezomagnetic Effect or Nonlinear Magnetostriction multiphysics coupling feature.
Hence a magnetostriction problem contains solid and magnetic domains, with at least
one domain shared by the two physics interfaces and with the magnetostriction
coupling defined on it.

Linear vs. Nonlinear Magnetostriction

For piezomagnetic model, it is possible to express the relation between the stress S,
strain , magnetic field H, and magnetic flux density B in either a stress-magnetization
form or strain-magnetization form. In COMSOL Multiphysics, both constitutive
forms can be used; simply select one, and the software will make all necessary
transformations if needed. You find all the necessary material data inputs within the
Piezomagnetic Material selected in the Solid Mechanics interface.

The nonlinear model of magnetostrictive strain can be used for the whole range from
full demagnetization to saturation magnetization. In case of nonlinear
magnetostriction, the magnetization model can be selected. The following options are

176 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

available: Langevin function, Hyperbolic tangent, User defined, and Linear. The last
option will make it possible to find an explicit expression for the magnetization. Note
however that such model does not have a proper saturation behavior, and thus it
should be used only in the operating range far from saturation. For all other choices,
the magnetization vector components will be treated as extra dependent variables.

For more details, see the corresponding theory section in Magnetostriction and
Piezomagnetism.

Create the Magnetostriction Interface and Define Domains

A magnetostriction problem can be set up in different ways:

• By selecting either Piezomagnetism or Nonlinear Magnetostriction from the Model

Wizard,
• By choosing either Piezomagnetism or Nonlinear Magnetostriction from the Add
Physics menu when working in an existing model, or
• By adding the corresponding coupling features to create the coupling manually.

In the first two cases, by default all the domains in the model are assumed to be
magnetostrictive materials.

PIEZOMAGNETISM
When setting up the problem manually (that is, by adding single physics interfaces one
at a time) both Solid Mechanics and Magnetic Fields interfaces should be added. Then,
you have to specify which domains are in each physics, and which domains are to be
modeled as piezomagnetic materials.

1 On the Solid Mechanics interface Settings window, locate the Domain Selection
section. Select the domains which undergo structural deformation, including the
piezomagnetic material domains.
2 Go to the Solid Mechanics>Piezomagnetic Material node. If it is not yet available, add
such node. Select the domains where the piezomagnetic effect applies. You find all
the necessary material data inputs within the Piezomagnetic Material. This includes
the elasticity, magnetic permeability and coupling matrices. Non piezomagnetic
domains can be modeled using any other available structural material model.
3 Go to the Magnetic Fields node and under Domain Selection select the domains the
magnetics equations must be solved. These domains include all the piezomagnetic
domains, and any other magnetic domains.

MODELING MAGNETOSTRICTIVE MATERIALS | 177

 4 Go to the Magnetic Fields>Ampere’s Law, Piezomagnetic node. If it is not yet available,
add such node. Select the domains where the piezomagnetic effect needs to be
modeled.
5 A Multiphysics>Piezomagnetic Effect node is already present if the coupling was
added using either the Model Wizard or Add Physics window. If the model is set up
manually (that is, single physics interfaces are added), right-click the Multiphysics
Couplings node to add a Piezomagnetic Effect coupling. If several Solid Mechanics or
Magnetic Fields interfaces are present, select the correct ones. By default, the model
will be solved as fully coupled. Alternatively, on the Piezomagnetic Effect coupling
node, you can select to include only either Joule effect or Villari effect. Confirm that
all the domains where the piezomagnetic material is present are selected.

Only domains that have both Ampere’s Law, Piezomagnetic selected in the
Magnetic Fields interface and Piezomagnetic Material selected in the Solid
Mechanics interface are applicable in the selection for Piezomagnetic Effect
coupling. The selection of this multiphysics coupling node cannot be
edited directly.

6 Confirm that remaining domains are well assigned:

- Solid and magnetic domains: In these domains, the magnetics and structural
problems are solved independently without any piezomagnetic coupling. This
represents deformable magnetic materials for which such coupling is negligible.
Exclude such domains from the selections for all Ampere’s Law, Piezomagnetic and
Piezomagnetic Material nodes under the corresponding interfaces.
- Solid-only domains: In these domains, only mechanical phenomena are modeled,
and the magnetics phenomena are neglected. This represents deformable solid
nonmagnetic materials. Exclude such domains from the selection in the Magnetic
Fields interface.
- Magnetics-only domains: Those domains are nonsolid. This is typically the case
of air, in which the magnetics is solved but not mechanics. Exclude such domains
from the selection in the Solid Mechanics interface.
- Nonsolid and nonmagnetic domains: In those domains another physics is solved.
Typical examples are acoustic domains and fluid flow domains, which might be

178 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

 present in certain multiphysics application. Exclude such domains from the
selections on both Magnetic Fields and Solid Mechanics interface nodes.

• Piezomagnetic Material
• Ampère’s Law, Piezomagnetic

NONLINEAR MAGNETOSTRICTION
When setting up the problem manually (that is, by adding single physics interfaces one
at a time) both Solid Mechanics and Magnetic Fields interfaces should be added. Then,
you have to specify which domains are in each physics, and which domains are to be
modeled as magnetostrictive materials.

Instead of Magnetic Fields interface, you can also use Rotating Machinery,
Magnetic interface.

1 On the Solid Mechanics interface Settings window, locate the Domain Selection
section. Select the domains which undergo structural deformation, including the
magnetostrictive material domains. The following structural material nodes are
supported to represent the magnetostrictive domains: Linear Elastic Material and
Hyperelastic Material (available with the Nonlinear Structural Materials Module).
2 Go to the Magnetic Fields node and under Domain Selection select the domains the
magnetics equations must be solved. These domains include all the magnetostrictive
domains, and any other magnetic domains.
3 Go to the Magnetic Fields>Ampere’s Law, Nonlinear Magnetostrictive node. If it is not
yet available, add such node. Select the domains where the magnetostriction effect
needs to be modeled. You can specify the magnetization model and enter the related
material data on the node.
4 A Multiphysics>Nonlinear Magnetostriction node is already present if the coupling was
added using either the Model Wizard or Add Physics window. If the model is set up
manually (that is, single physics interfaces are added), right-click the Multiphysics
Couplings node to add a Nonlinear Magnetostriction coupling. If several Solid
Mechanics or Magnetic Fields interfaces are present, select the correct ones. On the
coupling node, you can also select the magnetostriction model depending on the
material symmetry and enter the corresponding coupling data. By default, the
model will be solved as fully coupled. Alternatively, on the Nonlinear Magnetostriction

MODELING MAGNETOSTRICTIVE MATERIALS | 179

 coupling node, you can select to include only either Joule effect or Villari effect.
Select only the domains where the magnetostrictive effect needs to be modeled.

On the Nonlinear Magnetostriction coupling node, the applicable domains

in the selection are only domains that have both Ampere’s Law, Nonlinear
Magnetostrictive selected in the Magnetic Fields interface and Linear Elastic
Model (or Hyperelastic Model) selected in the Solid Mechanics interface.

5 Confirm that remaining domains are well assigned:

- Solid and magnetic domains: In these domains, the magnetics and structural
problems are solved independently without any magnetostriction coupling. This
represents deformable magnetic materials for which such coupling is negligible.
Exclude such domains from the selection in the Nonlinear Magnetostriction
coupling node.
- Solid-only domains: In these domains, only mechanical phenomena are modeled,
and the magnetics phenomena are neglected. This represents deformable solid
nonmagnetic materials. Exclude such domains from the selection in the Magnetic
Fields interface.
- Magnetics-only domains: Those domains are nonsolid. This is typically the case
of air, in which the magnetics is solved but not mechanics. Exclude such domains
from the selection in the Solid Mechanics interface.
- Nonsolid and nonmagnetic domains: In those domains another physics is solved.
Typical examples are acoustic domains and fluid flow domains, which might be
present in certain multiphysics applications. Exclude such domains from the
selections in both Magnetic Fields and Solid Mechanics interfaces.

• Ampère’s Law, Nonlinear Magnetostrictive

180 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

Mechanical Damping and Losses
Damping and losses are important factors for determining the response in
time-dependent and frequency domain studies, and sometimes also when computing
eigenfrequencies. This section describes how to model damping and loss using
different damping models. In this section:

In this section:

• About Damping
• Rayleigh Damping
• Loss Factor Damping
• Viscoelastic Damping
• Explicit Complex-Valued Damping
• Equivalence Between Loss Factor, Rayleigh, and Viscous Damping
• Piezoelectric Damping
• Adding Damping in the Modal Solver

Damping Sources
There are many sources of damping in a system. Some of them are:

• Dissipation in the material. This dissipation can be for example be proportional to

the amplitude of the strain rate (viscous damping) or to the amplitude of the strain
itself (hysteretic damping).
• Thermoelastic damping, which is a thermodynamic effect which is related not only
to the state in a point, but also to the gradients of the temperature field.
• Damping caused by the surrounding medium, often air or water.
• Friction between joined parts.
• Components intended to supply damping like a dashpot in a car suspension.

It is often difficult to separate and quantify these effects, so damping modeling is one
of the biggest challenges in structural dynamics.

MECHANICAL DAMPING AND LOSSES | 181

 About Damping
Phenomenological damping models are typically invoked to model the intrinsic
frictional damping present in most materials (material damping). These models are
easiest to understand in the context of a system with a single degree of freedom. The
following equation of motion describes the dynamics of such a system with viscous
damping:

2
d u du
m ---------2- + c ------- + ku = f  t  (2-21)
dt dt

In this equation u is the displacement of the degree of freedom, m is its mass, c is the
damping parameter, and k is the stiffness of the system. The time (t) dependent forcing
term is f(t). This equation is often written in the form:

2
d u- du 2 f t
--------- + 2 0 ------- +  0 u = --------- (2-22)
dt
2 dt m

where   c2m0 and 02  km. In this case  is the damping ratio (  1 for
critical damping) and 0 is the undamped resonant frequency of the system. In the
literature it is more common to give values of  than c. The damping ratio  can also
be readily related to many of the various measures of damping employed in different
disciplines. These are summarized in Table 2-10.
TABLE 2-10: RELATIONSHIPS BETWEEN MEASURES OF DAMPING

DAMPING DEFINITION RELATION TO

PARAMETER DAMPING RATIO

Damping ratio  = c  c critical –

Logarithmic u  t0   d  2
decrement  d = ln  ----------------------
u  t0 +  
 « 1
where t0 is a reference time and  is the period
of vibration for a decaying, unforced degree of
freedom.

182 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

TABLE 2-10: RELATIONSHIPS BETWEEN MEASURES OF DAMPING

DAMPING DEFINITION RELATION TO

PARAMETER DAMPING RATIO

Quality factor Q =    Q  1   2 

where  is the bandwidth of the amplitude  « 1

resonance measured at 1  2 of its peak.
Loss factor 1 Qh At the resonant
 = ------  --------
2  W h frequency:
  2
where Qh is the energy lost per cycle and Wh is
the maximum potential energy stored in the  « 1
cycle. The variables Qh and Wh are available as
solid.Qh and solid.Wh.

In the frequency domain, the time dependence of the force and the displacement can
be represented by introducing a complex force term and assuming a similar time
dependence for the displacement. The equations

jt jt
f  t  = Re  Fe  and u  t  = Re  Ue 

are written where  is the angular frequency and the amplitude terms U and F can in
general be complex (the arguments provide information on the relative phase of
signals). Usually the real part is taken as implicit and is subsequently dropped.
Equation 2-21 takes the following form in the frequency domain:

2
–  mU + jcU + kU = F (2-23)

where the time dependence has canceled out on both sides. Alternatively, this equation
can be written as:

2 2 F
–  U + 2j 0 U +  0 U = ----- (2-24)
m

There are three basic damping models available in the structural mechanics interfaces
for explicit modeling of material damping — Rayleigh damping, viscous damping,
and loss factor models based on introducing complex quantities into the equation
system. There are also other phenomena which contribute to the damping. Some
material models, such as viscoelasticity and plasticity are inherently dissipative. It is also
possible to model damping in spring conditions.

MECHANICAL DAMPING AND LOSSES | 183

 Rayleigh Damping
A common method of modeling damping is Rayleigh damping, where two damping
coefficients are specified. This type of damping is not directly related to any physical
process, but must be seen as a way to take the total damping of a structure into
account.

Rayleigh damping introduces damping in a form based on Equation 2-21. This means
that the method can be applied generally in either the time or frequency domain. The
parameter c in Equation 2-21 is defined as a fraction of the mass and the stiffness using
two parameters, dM and dK, such that

c =  dM m +  dK k (2-25)

Substituting this relationship into Equation 2-21 and rearranging into the form of
Equation 2-22 gives:

2
d u- 2 du 2 ft
---------
2
+   dM +  dK  0  ------- +  0 u = ---------
dt dt m

When there are many degrees of freedom m, k, and c become matrices and the
technique can be generalized.

Rayleigh damping can therefore be identified as equivalent to the damping ratio at

resonance of:

1  dM
 = ---  ----------- +  dK  0 (2-26)
2  0 

Note that Equation 2-26 holds separately for each vibrational mode in the system at
its resonant frequency. In the frequency domain it is possible to use frequency
dependent values of dM and dK. For example, setting dM  0 and dK  2/0
produces an equivalent viscous damping model at the resonant frequency 0.

While Rayleigh damping is numerically convenient, the model does not agree with
experimental results for the frequency dependence of material damping over an
extended range of frequencies. This is because the material damping forces behave
more like frictional forces (which are frequency independent) than viscous damping
forces (which increase linearly with frequency as implied by Equation 2-23). In the
frequency domain it is possible to introduce loss factor damping, which has the desired
property of frequency independence.

184 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

A complication with the Rayleigh damping model is to obtain good values for the
damping parameters dM and dK. A more physical damping measure is the damping
ratio, the ratio between actual and critical damping, often expressed as a percentage of
the critical damping. Commonly used values of the damping ratio can be found in the
literature.

Using Equation 2-26, this relationship at two frequencies, f1 and f2, with different
damping ratio, 1 and 2, results in an equation system that can be solved for dM and
dK:

1-
----------- f
4f 1 1  dM 1
=
1-  dK 2
----------- f
4f 2 2

The damping ratios are then

1 f2 – 2 f1
 dM = 4f 1 f 2 ---------------------------
2 2
f2 – f1
2 f2 – 1 f1
 dK = ---------------------------
2 2
  f2 – f1 

Using the same damping ratio, 1 = 2 = 0, does not result in a constant damping
factor inside the interval f1  f  f2. It can be shown that the damping factor is lower
inside the interval, as Figure 2-13 shows.
Damping ratio

Rayleigh damping
(f)

0
Specified damping

f1 f2 f

Figure 2-13: An example of Rayleigh damping.

MECHANICAL DAMPING AND LOSSES | 185

 Since the coefficients dM and dK should not be negative, the damping ratios are
constrained by the respective frequencies as

f1 2 f2
----  -----  ----
f2 1 f1

For many applications it is sufficient to leave dM as zero and to define damping only
using the dK coefficient. Then, according to Equation 2-26, a damping which
increases linearly with frequency is obtained. If the damping ratio f0 or loss factor
f0 is known at a given frequency f0, the appropriate value for dK is:

 dK =    f 0  =    2f 0 

In order to visualize the damping ratio as a function of frequency, click

Damping Ratio Preview ( ).

Loss Factor Damping

Loss factor damping (sometimes referred to as material damping, structural damping,
or hysteretic damping) can be applied in the frequency domain.

The loss factor is a measure of the inherent damping in a material when it is

dynamically loaded. It is typically defined as the ratio of energy dissipated in unit
volume per radian of oscillation to the maximum strain energy per unit volume.

In COMSOL Multiphysics the loss information appears as a multiplier to the elastic

constitutive matrix Dc

c
D =  1 + j s D

For a nonlinear elastic material, this applies to the tangential stiffness.

The use of loss factor damping traditionally refers to a scalar-valued loss factor s. But
there is no reason that s must be scalar. Because the loss factor is a value deduced from
true complex-valued material data, it can be represented by a matrix of the same
dimensions as the anisotropic stiffness matrix. Especially for orthotropic materials,
there should be a set of loss factors of all normal and shear elasticity modulus
components, and COMSOL allows all these options, so a more general expression is.

c
D mn =  1 + j s mn D mn

186 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

For hyperelastic materials the loss information appears as a stress contribution to the
second Piola–Kirchhoff stress, S:

W s
S q = j s ----------
E

Loss factor damping is available for frequency response analysis and damped
eigenfrequency analysis in all interfaces.

Viscous Damping
You can add an explicit viscous damping to several material models. Viscous damping
can be used both in time domain and frequency domain. In the viscous damping
model, and extra contribution, proportional to the strain rate, is added to the stress
tensor, as described in Viscous Damping.

You can specify the viscous damping for volumetric strains and shear strains
independently.

Equivalence Between Loss Factor, Rayleigh, and Viscous Damping

In frequency domain, it is possible to use Rayleigh damping in order to specify an
equivalent viscous damping. Set the stiffness damping parameter dK, to the loss
factor, , divided by the excitation frequency:

 
 dK = --------- = ----
2f 

The mass damping factor, dM, should be set to zero.

If, on the other hand, you would want to use a viscous damping, corresponding to a
certain Rayleigh stiffness damping, the conversion to bulk and shear viscosity can be
made using the expressions

 b = K dK
 v = G dK

where K and G are the bulk and shear moduli, respectively. Equivalently, you can
transform between loss factor damping and viscous damping,

MECHANICAL DAMPING AND LOSSES | 187

 K
 b = --------

G
 v = --------


Explicit Complex-Valued Damping

In frequency domain, it is possible to define damping by modeling the dissipative
behavior of the material using complex-valued material properties. In COMSOL
Multiphysics, you can enter the complex-valued data directly, using i or sqrt(-1) for
the imaginary unit.

However, entering complex-valued data directly into the material parameters is

incompatible with the built-in damping and viscoelasticity features.

Viscoelastic Damping
In some cases, damping is included implicitly in the material model. This is the case for
a viscoelastic material, where damping operates on the shear components of stress and
strain.

When viscoelasticity is modeled in the frequency domain, it will act as a loss factor
damping. The complex modulus G*() is the frequency-domain representation of the
stress relaxation function of viscoelastic material. It is defined as

G = G + jG =  1 + j s G

where G' is the storage modulus, G'' is the loss modulus, and their ratio sG''G' is
the loss factor. The term G' defines the amount of stored energy for the applied strain,
whereas G'' defines the amount of energy dissipated as heat; G', G'', and s can all be
frequency dependent.

Piezoelectric Damping
Piezoelectric losses are more complex and include coupling and electrical losses in
addition to the material terms. For damping in piezoelectric materials, see Piezoelectric
Losses.

For piezoelectric materials, dK is only used as a multiplier of the structural

contribution to the stiffness matrix when building-up the damping matrix as given by

188 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

Equation 2-25. In the frequency domain studies, you can use the coupling and
dielectric loss factors equal to dK to effectively achieve the Rayleigh damping
involving the whole stiffness matrix.

Adding Damping in the Modal Solver

In COMSOL it is possible to solve a problem for a set of modes in the absence of
damping, and then to use those solutions as a modal basis to solve a problem in the
time (using a time domain modal study) or frequency domain (using a Frequency
Domain, Modal study). In both of these cases it is possible to manually assign a
damping ratio to the computed modes in the time or frequency domain study. To do
that, right click on the study and choose Show Default Solver, then expand the solver
sequence until the Modal Solver node is visible. In the Settings window for that node,
add damping ratios for each of the modes.

For more details, see the section Modal Solver in the COMSOL
Multiphysics Reference Manual.

MECHANICAL DAMPING AND LOSSES | 189

 Modeling Geometric Nonlinearity
This section discusses how to model problems where displacements or strains are of a
size where the deformation of the structure has to be taken into account when
formulating the equations. Examples of the type of problems where this feature is
useful include:

• Thin structures, where the deflection is of the same order of magnitude as the
thickness.
• Where the structure exhibits large rotations. A rigid body rotation of only a few
degrees causes significant strains and stresses in a material where a linear strain
representation is used.
• Where the strains are larger than a few percent.
• Contact problems.
• Where a prestress must be taken into account for computing the dynamic response
of a structure.
• Buckling problems.
• Where a deformed mesh is used.
• Fluid-structure interaction problems.

• Contact Modeling
• Fluid-Structure Interaction
• The more formal theory is described in Analysis of Deformation

Geometric Nonlinearity, Frames, and the ALE Method

Consider the bending of a beam in the general case of a large deformation (see
Figure 2-14). In this case the deformation of the beam introduces an effect known as
geometric nonlinearity into the equations of solid mechanics.

Figure 2-14 shows that as the beam deforms, the shape, orientation, and position of
the element at its tip changes significantly. Each edge of the infinitesimal cube
undergoes both a change in length and a rotation that depends on position.
Additionally, the three edges of the cube are no longer orthogonal in the deformed

190 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

configuration (although typically for practical strains the effect of the
nonorthogonality can be neglected in comparison to the rotation).

From a simulation perspective it is possible to solve the equations of solid mechanics

on either a fixed domain (this is often called a Total Lagrangian formulation), or on
a domain that changes continuously with the deformation. The latter approach is often
called an Updated Lagrangian formulation. These two approaches also stand in
contrast to the Eulerian formulation which is often used for fluid mechanics. In an
Eulerian formulation the flow through a domain fixed in space is considered, while in
the Lagrangian formulation a fixed volume of material is considered.

In COMSOL Multiphysics, the concepts of a material frame and a

spatial frame are used. Equations formulated in the material frame will
give a Total Lagrangian formulation, while equations formulated in the
spatial frame will give an Eulerian formulation.

See Frames and Coordinate Systems in the theory chapter for more
details.

Solid mechanics in COMSOL Multiphysics is formulated on the material frame. This

is achieved by defining a displacement field for every point in the solid, usually with
components u, v, and w. At a given coordinate (X, Y, Z) in the reference configuration
(on the left of Figure 2-14) the value of u describes the displacement of the point
relative to its original position. The displacement is considered as a function of the
material coordinates (X, Y, Z), but it is not explicitly a function of the spatial
coordinates (x, y, z). The spatial coordinates give the current location in space of a
point in the deformed solid. As a consequence, it is only possible to compute
derivatives with respect to the material coordinates.

Taking derivatives of the displacement with respect to X, Y, and Z enables the

definition of a strain tensor. There are possible representations of the deformation. Any
reasonable representation must however be able to represent a rigid rotation of an
unstrained body without producing any strain. The engineering strain fails here, thus
it cannot be used for general geometrically nonlinear cases. One common choice for
representing large strains is the Green–Lagrange strain. It contains derivatives of the
displacements with respect to the original configuration. The values therefore
represent strains in material directions. This allows a physical interpretation, but it
must be realized that even for a uniaxial case, the Green–Lagrange strain is strongly
nonlinear with respect to the displacement. If an object is stretched to twice its original

MODELING GEOMETRIC NONLINEARITY | 191

 length, the Green–Lagrange strain is 1.5 in the stretching direction. If the object is
compressed to half its length, the strain would read -0.375.

An even more fundamental quantity is the deformation gradient, which contains the
derivatives of the deformed coordinates with respect to the original coordinates:

x
F = ------- = u + I
X

The deformation gradient contains all information about the local deformation in the
solid, and can be used to form many other strain quantities. As an example, the Green–
Lagrange strain is

1 T
 = ---  F F – I 
2

An element at a point (X, Y, Z) specified in the material frame moves with a single
piece of material throughout a solid mechanics simulation. It is often convenient to
define material properties in the material frame as these properties move and rotate
naturally together with the volume element at the point at which they are defined as
the simulation progresses. In Figure 2-14 this point is illustrated by the small cube
highlighted at the end of the beam, which is stretched, translated, and rotated as the
beam deforms. The three mutually perpendicular faces of the cube in the Lagrange

192 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

frame are no longer perpendicular in the deformed (spatial) frame. The deformed
frame coordinates in this frame are denoted (x, y, z) in COMSOL.

Figure 2-14: An example of the deformation of a beam showing the undeformed state
(left) and the deformed state (right) of the beam itself with an element near its tip
highlighted (top), of the element (center), and of lines parallel to the x-axis in the
undeformed state (bottom).

It is important to note that, as the solid deforms, the Lagrangian frame

becomes a nonorthogonal curvilinear coordinate system (see the lower
part of Figure 2-14 to see the deformation of the X-axis). Particular care
is therefore required when defining physics in this coordinate system.

FRAME CONTROL CONFLICTS

Under geometric nonlinearity, a Solid Mechanics interface (or a similar structural
mechanics interface such as Multibody Dynamics) will assume control over the spatial
frame in the domains where it is active. However, the definition of the spatial frame
must be unique, and there may be conflicts. Some examples are:

• A Solid Mechanics interface and a moving mesh feature (for example, Deforming
Domain or Rotating Domain) have a common selection. In this case, the selection in

MODELING GEOMETRIC NONLINEARITY | 193

 the Solid Mechanics interface will be marked as “not applicable”. Thus, it is not
possible to solve a structural mechanics problem on such a domain.
• There are more than one frame-controlling physics interfaces active on the same
domain.

When two physics interfaces are competing for frame control, you will get the error
message “Multiple moving frame specifications on the same selection” when trying to
run the study. To identify the problem, go to the settings for the study step, and select
Modify model configuration for study step. There, you will get an overview of the frame
control in the model. Note that it is quite possible that several features control the
spatial frame, as long as it is on different geometric selections.

When you select a physics interface in the tree view, you can click the Control Frame
Deformation button ( ) to toggle whether that interface should control the spatial
frame or not.

194 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

STRESS MEASURES
For example, in the spatial frame it is easy to define forces per unit area (known as
tractions) that act within the solid and to define a stress tensor that represents all of
these forces that act on a volume element. Such forces could be physically measured,
for example, using an implanted piezoresistor. The stress tensor in the spatial frame is
called the Cauchy or true stress tensor (in COMSOL Multiphysics this is referred to as
the spatial stress tensor). To construct the stress tensor in the Lagrangian frame a
tensor transformation must be performed on the Cauchy stress. This produces the
second Piola–Kirchhoff (or material) stress, which can be used with the material strain
to solve the solid mechanics problem in the (fixed) material frame. This is how the
Solid Mechanics interface works when geometric nonlinearities are enabled.

For the Cauchy stress tensor, both the force components and the normal to the area
on which the force is acting have fixed directions in space. This means that if a stressed
body is subjected to a pure rotation, the actual values of the stress components will
change. What was originally a uniaxial stress state might be transformed into a full
tensor with both normal and shear stress components. In many cases, this is neither
what you want to use nor what you would expect.

Consider for example an orthotropic material with fibers having a certain orientation.
It is much more plausible that you want to see the stress in the fiber direction, even if
the component is rotated. The Second Piola–Kirchhoff stress has this property as it is
defined along the material directions. In the figure below, an originally straight
cantilever beam has been subjected to bending by a pure moment at the tip. The
xx-component of the Cauchy stress and Second Piola–Kirchhoff stress are shown.
Since the stress is physically directed along the beam, the xx-component of the Cauchy
stress (which is related to the global x direction) decreases with the deflection. The
Second Piola–Kirchhoff stress, however, has the same through-thickness distribution
all along the beam, even in the deformed configuration.

MODELING GEOMETRIC NONLINEARITY | 195

 Figure 2-15: xx-components of the Cauchy stress tensor (top) and Second-Piola–Kirchhoff
stress tensor (below) for an initially straight beam

Another stress measure available in COMSOL Multiphysics is the First Piola–Kirchhoff

stress. It is a multiaxial generalization of the nominal (or engineering) stress. The stress
is defined as the force in the current configuration acting on the original area. The First
Piola–Kirchhoff stress is an unsymmetric tensor, and is for that reason less attractive to
work with. Sometimes you may also encounter the Kirchhoff stress, although it is not
used in COMSOL Multiphysics. The Kirchhoff stress is just the Cauchy stress scaled
by the volume change. It has little physical significance, but can be convenient in some
mathematical and numerical operations.

Unfortunately, even without a rotation, the actual values of all these stress
representations are not the same. All of them scale differently with respect to local
volume changes and stretches. This is illustrated in the graph below. The
xx-component of four different stress measures are plotted at the fixed end of the beam
from the example above. At this point, the beam axis coincides with the x-axis, so the
directions of all stress tensor components coincide. In the center of the beam, where
strains, and thereby volume changes are small, all values approach each other. For a
case with large rotation but small strains, the different stress representations can be
seen as pure rotations of the same stress tensor.

196 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

Figure 2-16: Stress distribution across the beam at the constrained end.

If you want to compute the resulting force or moment on a certain boundary based on
the stresses, there are in practice only two possible choices: Either integrate the Cauchy
stress over the deformed boundary, or integrate the First Piola–Kirchhoff stress over
the same boundary in the undeformed configuration. In COMSOL Multiphysics this
corresponds to selecting either Spatial frame or Material frame in the settings for the
integration operator.

ALE METHOD
In the case of solid mechanics, the material and spatial frames are associated directly
with the Lagrangian and Eulerian frames, respectively. In a more general case (for
example, when tracking the deformation of a fluid, such as a volume of air surrounding
a moving structure) tying the Lagrangian frame to the material frame becomes less
desirable. Fluid must be able to flow both into and out of the computational domain,
without taking the mesh with it. The arbitrary Lagrangian-Eulerian (ALE) method
allows the material frame to be defined with a more general mapping to the spatial or
Eulerian frame. In COMSOL Multiphysics, a separate equation is solved to produce
this mapping — defined by the mesh smoothing method (Laplacian, Winslow,
hyperelastic, or Yeoh) with boundary conditions that determine the limits of

MODELING GEOMETRIC NONLINEARITY | 197

 deformation (these are usually determined by the physics of the system, whilst the
domain level equations are typically being defined for numerical convenience). The
ALE method offers significant advantages since the physical equations describing the
system can be solved in a moving domain.

Deformed Geometry and Moving Mesh in the COMSOL Multiphysics

Reference Manual

Geometric Nonlinearity for the Solid Mechanics Interface

For the Solid Mechanics interface, you enable a geometrically nonlinear analysis for a
certain study step by selecting the Include geometric nonlinearity check box in the Study
Settings section of a study step.

If any active feature in the model requires the analysis to be geometrically nonlinear,
the Include geometric nonlinearity check box is selected automatically, and it cannot be
cleared. The following physics features force this behavior:

• Contact, because the deformation between the contacting boundaries must be part
of the solution.
• Moving mesh (when at least one deforming domain is active).
• Large strain plasticity.
• Hyperelastic materials, which are always formulated for large strains.

Usually you would also want to use geometric nonlinearity when a Moving Mesh
interface is present, but this setting is not forced by the program.

When you select a geometrically nonlinear study step, the behavior of several features
differs from that in a geometrically linear case:

• There is an important difference between using uppercase (X, Y, Z, R) and

lowercase (x, y, z, r) coordinates in expressions. The lowercase coordinates
represent the deformed position, and this introduces a dependency on the solution.
• Many features, such as coupling operators, can be specified as operating either in the
material (X, Y, Z) or the spatial (x, y, z) frame. This setting does not make a
difference unless a geometrically nonlinear analysis is performed. In most cases you
would want to do the operation in the material frame.

198 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

• The strain representation is changed from using engineering strains to Green–
Lagrange strains, unless the Geometrically linear formulation check box is selected in
the Geometric Nonlinearity section of a certain material.
• Where Green–Lagrange strains are used, Second Piola–Kirchhoff stresses are also
used. This means that material data must be given in terms of these quantities. This
distinction is important only when the strains actually are large.
• Pressure loads are interpreted as follower loads, so that the direction of the load as
well as the loaded area are deformation dependent.
• Rigid connectors take finite rotations into account.

Studies and Solvers in the COMSOL Multiphysics Reference Manual

Geometric Nonlinearity for the Shell, Plate, Membrane, Beam and

Truss Interfaces
For the Shell, Plate, Membrane, Beam, and Truss interfaces, a geometrically nonlinear
analysis is enabled in the same way described above. For the Membrane interface,
geometric nonlinearity must almost always be used, since it is nonlinear effects which
supply the stiffness in the transverse direction.

The geometric nonlinearity in the Beam interface is limited to large rotations and
displacements, but small strains.

The effect of using geometric nonlinearity in these interfaces is limited to the stress and
strain representation as the other effects described in Geometric Nonlinearity for the
Solid Mechanics Interface are not relevant.

Solving Geometrically Nonlinear Problems

Depending on the geometrically nonlinear effects that appear in your model, you may
have to use different solution strategies. Some problems in this class are strongly
nonlinear, while others show only a weak deviation from linearity. Some guidelines are:

• If the problem has a path dependent solution, then it must be solved in an

incremental way in order to give a correct solution. Problems including for example
plasticity or friction belong to this class. If you do the analysis in time domain, then
the solution is inherently incremental. If the analysis is stationary, invoke the

MODELING GEOMETRIC NONLINEARITY | 199

 parametric continuation solver by adding an Auxiliary Sweep, and ramp up some
loading parameter. In either case, make sure that the step size is not too large.
• Problem that have a unique solution, like an elastic model subjected to large
rotations or strains can be solved in a single static load step without loss of accuracy.
It is however possible that such an approach will not converge, in which case a
parametric continuation solver must be used.
• In problems involving large rotations, the default settings of the nonlinear solver
will sometimes give a too conservative solution strategy. You can often decrease the
solution time significantly by modifying the settings under Method and Termination
in the settings for the Fully Coupled or Segregated Step node in the solver sequence.
Set Nonlinear method to Constant (Newton) and use a rather high Damping factor. In
most cases the value 1 will work.
• If you model a situation which to a large extent is a rigid rotation, it is often
necessary to use tighter tolerances than usual in order to avoid spurious stresses.
Since the strains are computed from the differences of the displacements in an
element, even a small relative error in the displacements (which are large) can cause
significant strains. This will be visible in a case where the actual stresses are small.

Prestressed Structures
You can analyze eigenfrequency, frequency domain, or time-dependent problems
where the dynamic properties of the structure are affected by a preload, such as a
tensioned string.

Usually, a study of a prestressed problem includes using two study steps. The first step
is a Stationary step in which the static preload is applied. The preload step can be
computed with or without taking geometric nonlinearity into account. In the second
study step, where you compute the eigenfrequency or the frequency response, it is
necessary to take geometric nonlinearity into account. Even if the displacements and
strains are small, this is what gives the prestress contribution to the equations.

There are three predefined study sequences for prestressed dynamic analysis:

• Eigenfrequency, Prestressed
• Frequency Domain, Prestressed
• Frequency Domain, Prestressed, Modal

The prestressed study types assume that the loading causes small perturbations around
the prestressed state.

200 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

In a general nonlinear analysis, like a Stationary or Time Dependent study step, the full
combined effect of the prestress and other loads will automatically be included.

The same principles apply also to a linear buckling analysis, except that both study steps
should be geometrically linear. The nonlinear contribution is included in the
formulation of the buckling eigenvalue itself.

FOLLOWER LOADS
Loads that change orientation with deformation, such as a pressure, actually contribute
not only to the load but also to the stiffness. This is a physical effect and not just a
numerical artifact. Whether such loads are included or not in an Eigenfrequency study
step will affect the computed eigenfrequencies. If you for some reason do not want this
effect, you must suppress the load in the Physics and Variables section of the
Eigenfrequency node.

If you use a local coordinate system for describing a load, you must, in
case of geometric nonlinearity, pay attention to whether that coordinate
system has constant axis orientations or not. As an example, the default
boundary system has Frame set to Deformed Configuration, so that a load
represented in that system will behave as a follower load. Change to
Reference Configuration if the load should act in fixed directions.

In more general terms, any feature in which there is a dependency on the

choice of frame (material or spatial) can potentially affect the outcome of
a prestressed analysis.

CONTACT AND PRESTRESSED ANALYSIS

If a contact is included in the prestress load case, you can perform a subsequent
eigenfrequency or frequency-domain analysis in which the linearization will be made
around the computed contact state. The connection between the two boundaries is
elastic, using the penalty factor as stiffness, both for the penalty method and the
augmented Lagrangian method. If the augmented Lagrangian method is used for the
contact modeling, then the penalty factor must be constant and not dependent on the
iteration number.

MODELING GEOMETRIC NONLINEARITY | 201

 The contact state is considered as fixed, so there is no harmonic perturbation in contact
quantities, such as the contact pressure.

The gap distance in a contact analysis is computed from the coordinates

of the contacting boundaries in the spatial frame. The spatial frame is
defined by the displacements in the preload step and is not affected by the
displacements in a subsequent perturbation step.

INELASTIC STRESSES AND STRAINS

When inelastic stresses and strains are part of the problem description, you need to
make some extra considerations. Such contributions are formally part of the
constitutive model but generate load vector contributions.

There are three Preset study types which can be used to set up these two
study steps: Eigenfrequency, Prestressed; Frequency Domain, Prestressed;
and Linear Buckling.

If you want to explicitly prescribe the stress field for a prestressed analysis
rather than solving for it, you should not use the two study step
procedure. In such a case, prescribe the stress field using an Initial Stress
and Strain, External Stress, or External Strain node. Then add a separate
Eigenfrequency or Frequency Domain study and select Include Geometric
Nonlinearity in the settings for the study step.

Eigenfrequency, Prestressed, Frequency Domain, Prestressed, and Linear

Buckling in the COMSOL Multiphysics Reference Manual

• For an example of a general prestressed eigenfrequency analysis see

Bracket — Eigenfrequency Analysis: Application Library path
Structural_Mechanics_Module/Tutorials/bracket_eigenfrequency.
• For an example of an analysis where the stress state is explicitly
prescribed, see Vibrating Membrane: Application Library path
Structural_Mechanics_Module/Verification_Examples/
vibrating_membrane.

202 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

Geometric Nonlinearity for the Piezoelectric Material

PIEZOELECTRIC MATERIALS WITH LARGE DEFORMATIONS

The linear piezoelectric equations as presented in About Piezoelectric Materials with
engineering strains are valid if the model undergoes only relatively small deformations.
As soon as the model contains larger displacements or rotations, these equations
produce spurious strains that result in an incorrect solution. To overcome this
problem, so-called large deformation piezoelectric equations are required.

The Piezoelectric Material implements the large deformation piezoelectric equations

according to Yang (Ref. 1). Key items of this formulation are:

• The strains are calculated as the Green–Lagrange strains, ij:

1  u i u j u k u k
 ij = ---  -------- + -------- +
2  X j X i  --------
- ---------
X i X j 
(2-27)
k

Green–Lagrange strains are defined with reference to an undeformed geometry.

Hence, they represent a Lagrangian description. In a small-strain, large rotational
analysis, the Green–Lagrange strain corresponds to the engineering strain in
directions that follow the deformed body.
• Electrical field variables are calculated in the material directions, and the electric
displacement relation is replaced by an expression that produce electric polarization
in the material orientation of the solid.
• In the variational formulation, the electrical energy is split into two parts: The
polarization energy within the solid and the electric energy of free space occupied
by the deformed solid.

The first two items above result in another set of constitutive equations for large
deformation piezoelectricity:

T
S = cE  – e Em
P m = e +  0   rS – I E m

where S is the second Piola–Kirchhoff stress;  is the Green–Lagrange strain, Em and

Pm are the electric field and electric polarization in the material orientation; I is the
identity matrix; and cE, e, and rS are the piezoelectric material constants. The
expression within parentheses equals the dielectric susceptibility of the solid:

 =  rS – I

MODELING GEOMETRIC NONLINEARITY | 203

 The electric displacement field in the material orientation results from the following
relation

–1
D m = P m +  0 JC E m

where C is the right Cauchy–Green tensor

T
C = F F

Fields in the global orientation result from the following transformation rules:

–T
E = F Em
–1
P = J FP m
(2-28)
–1
D = J FD m
–1
v = V J

where F is the deformation gradient; J is the determinant of F; and v and V are the
volume charge density in spatial and material coordinates, respectively. The
deformation gradient is defined as the gradient of the present position of a material
point x  X + u:

x
F = -------
X

Finally, one can rewrite the constitutive equations as

T
S = cE  – e Em
–1
D m = e +  0  rS E m +  0  JC – I E m

DECOUPLED MATERIALS WITH LARGE DEFORMATIONS

The large deformation formulation described in the previous section applies directly to
materials not being piezoelectric if the coupling term is set to zero: e  0. In that case,
the structural part corresponds to the large deformation formulation described for the
solid mechanics interfaces.

The electrical part separates into two different cases: For solid domains, the electric
energy consists of polarization energy within the solid and the electric energy of free
space occupied by the deformed solid — the same as for the piezoelectric materials.
For nonsolid domains this separation does not occur, and the electric displacement in

204 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

these domains directly results from the electric field — the electric displacement
relation:

D = 0 r E

The Electrostatics Interface in the COMSOL Multiphysics Reference

Manual

LARGE DEFORMATION AND DEFORMED MESH

The Piezoelectricity Interface, Solid can be coupled with the Moving Mesh (ALE)
interface in a way so that the electrical degrees of freedom are solved in an ALE frame.
This feature is intended to be used in applications where a model contains nonsolid
domains, such as modeling of electrostatically actuated structures. This functionality is
not required for modeling of piezoelectric or other solid materials.

The use of ALE has impacts on the formulation of the electrical large deformation
equations:

• The first impact is that with ALE, the gradient of the electric potential directly
results in the electric field in the global orientation, and the material electric field
results after transformation.
• The most visible impact is on the boundary conditions. With ALE, any surface
charge density or electric displacement is defined per the present deformed
boundary area, whereas for the case without ALE, they are defined per the
undeformed reference area.

Deformed Geometry and Moving Mesh in the COMSOL Multiphysics

Reference Manual

References
1. J. Yang, An Introduction to the Theory of Piezoelectricity, Springer Science and
Business Media, N.Y., 2005.

2. A.F. Bower, Applied Mechanics of Solids, CRC Press, Boca Raton, FL (http://
www.solidmechanics.org), 2010.

MODELING GEOMETRIC NONLINEARITY | 205

 Contact Modeling
Overview
Situations where objects come into contact with each other occur frequently in
mechanical simulations. Setting up and solving such contact problems can be a
challenging task, and this section contains information about important aspects of
creating models involving contact.

In a contact analysis, you solve for the contact state and the contact forces. The most
fundamental quantity in a contact analysis is the distance between the objects which
may come into contact, the gap. If the gap is positive then there is no interaction. The
task of the contact algorithm is to ensure that the gap never becomes negative; that is,
to avoid overclosure. In order to accomplish this task, contact forces must be
introduced.

For normal contact, the state only consists of being in contact or not, and the force
variable is the contact pressure in the normal direction. With friction included, there
are two possible states for the relative tangential displacement when in contact:
sticking or sliding. The tangential force vector is added as force variable.

The contact analysis functionality in COMSOL Multiphysics also includes the

possibility to prescribe adhesion and decohesion between the contacting objects, and
to model removal of material by wear when the objects are sliding relative to each
other.

In this section:

• Setting Up a Contact Problem • Special Types of Contact Problems

• Contact Pairs • Fallback Conditions to Contact
• Meshing for Contact Analysis Regions

• Settings for Contact Nodes • Solver Settings for Contact Analysis

• Quasistatic Contact Analysis • Monitoring the Solution

• Dynamic Contact Analysis • Dependent Variables in Contact

Analysis
• Multiphysics Contact Analysis
• Important Contact Variables

206 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

Setting Up a Contact Problem
Mechanical contact can be modeled between boundaries in the Solid Mechanics,
Multibody Dynamics, Shell, Layered Shell, and Membrane interfaces. You can model
contact not only within a single physics interface, but also between two physics
interfaces, or even between a physics interface and any boundary having a mesh.

To model a mechanical contact problem, you must do the following fundamental

steps:

• In the finalization step of the geometry sequence, you should normally have Action
set to Form an assembly. If Form a union is used, then the contacting boundaries must
be geometrically separated in the initial configuration.
• Add Contact Pair nodes under Definitions. A contact pair consists of two sets of
boundaries, which are called the source and destination boundaries. Contact pairs
can also be added automatically, based on boundary adjacency when the Form an
assembly action is used. The geometric gap distance is a variable set up by the
contact pair, which also define operators for mapping variables or expressions
between the selected boundaries.
• Use the default Contact node or add new Contact nodes in the physics interface. In
the Contact node, you select the contact pairs to be used, and provide the settings
for the physical and numerical properties of the contact model.
• If relevant, add Friction, Slip Velocity, Adhesion, Decohesion, or Wear subnodes to
Contact.

CONTACT MODELING | 207

 • Select a suitable study type. You can analyze contact using the following types of
study steps:
- Stationary; commonly using an auxiliary sweep.
- Bolt Pretension
- Time Dependent
- Any type of prestressed analysis. In this case, the full contact problem is solved in
the preload step, and in the subsequent perturbation step the contact state is
considered as fixed.

• Documentation of the Contact, Friction, Slip Velocity, Adhesion,

Decohesion, and Wear nodes.
• Contact Analysis Theory.
• Identity and Contact Pairs in the COMSOL Multiphysics Reference
Manual.

In a multiphysics analysis, a contact problem can also incorporate for example changes
in the heat flux or electric current through the contacting boundaries. You will then
also need to add corresponding features in the other participating interfaces, like a
Thermal Contact node in the Heat Transfer in Solids interface. The contact state and
contact pressure used by other physics interfaces are always supplied by the structural
mechanics interface.

Because of the multiphysics capabilities, the setup of a contact problem is

split into two parts. The definition of the contact pair is made under
Definitions, and can be shared between several physics interfaces. This part
of the contact problem defines the geometric properties of the contact,
such as search and mapping operations between the selected boundaries.
The physics related definitions of the contact properties are then made in
the respective physics interfaces.

The fact that you add a Contact node to your model will automatically make all study
steps geometrically nonlinear. For the default Contact node, this requirement can be
removed by selecting Disconnect pair.

208 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

INCLUDING FRICTION
In real life, there is always some friction between contacting objects, but this is
sometimes ignored when setting up the mechanical contact problem. There are several
reasons to do this simplification:

• In many cases, only the normal forces are significant for the general force
distribution in the structure, while the frictional forces only cause minor local
effects.
• The values of the friction coefficients are difficult to obtain, and unless the structure
is assembled under well controlled conditions, the magnitude of the friction can
vary a lot. If there is a large degree in uncertainty in the input data, it can be argued
that the inclusion of friction does not add much to the quality of the results.
• Adding friction to a contact problem will often increase the computation time
significantly, or even cause convergence problems.

There are a number of situations when friction modeling cannot be avoided. Some of
them are:

• When a significant portion of the load is carried by shear tractions acting on the
contact boundaries.
• When a tangential force is necessary to maintain stability and to avoid rigid body
motions. In many cases, it is however possible to replace the friction by a suitable
boundary condition instead, as long as there are no resultant forces being resisted
by such a constraint.
• When modeling wear.
• When the frictional dissipation is an important component of a dynamics problem,
or when it is needed as a heat source in a thermal analysis.

In some cases, such as when a brake pad slides along a brake disc, the size and
orientation of the slip velocity is known. You can then employ a simplified form of
friction modeling by assuming the tangential contact to always be in a slip state, which
simplifies the computation of the friction forces. This is done using the Slip Velocity
node. This is particularly useful for wear modeling.

INCLUDING ADHESION AND DECOHESION

You can also specify that the contacting boundaries stick to each other, so that they will
not separate or slide. The onset of adhesion, when the boundaries become
permanently attached to each other, can be based on several criteria:

• When a certain contact pressure is exceeded.

CONTACT MODELING | 209

 • When the gap distance between the contact boundaries is smaller than a certain
value.
• When a user specified logical expression is fulfilled. This can for example be used if
an adhesive cures when a certain temperature is exceeded.
• From the start of the analysis. This case is particularly useful if you are interested in
modeling the subsequent tearing of a thin glue layer by decohesion.

If adhesion is active between the contact boundaries, it is possible to break the bond
by adding a decohesion model. You can choose between several different decohesion
models.

Adhesion and friction can be combined, but during the time that two boundaries are
bonded to each other through adhesion, any settings for friction are ignored.

INCLUDING WEAR
It is possible to model adhesive or abrasive wear of the material when the contacting
boundaries are sliding along each other. The removal of material from the domain
adjacent to the contact boundary can be modeled using three, fundamentally different,
approaches.

For the Solid Mechanics and the Multibody Dynamics interfaces, the most general
approach relies on the deformed geometry concept, where material is actually removed
by using an adaptive mesh technique. In the second, simplified, approach wear is
incorporated as an offset in the contact condition. This approach is computationally
less expensive, and is suitable as long as only small amounts of material are removed,
and wear does not change the orientation of the normal to the boundary significantly.
In the Shell and Membrane interfaces the structural domain is represented by only a
meshed surface. Therefore, a different, more suitable approach is used, in which the
thickness variable and the midsurface offset to the meshed boundary are modified.

Computing the amount of wear involves solving a rate equation, hence, it is only
possible to compute wear in time-dependent studies. The wear rate is typically a
function of the contact pressure and the relative slip velocity between the contacting
boundaries.

The Wear subnode is not available in the Layered Shell interface.

210 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

SELECTING THE CONTACT METHOD
In COMSOL Multiphysics, there are three classes of methods available for solving
contact problems: the penalty method, the augmented Lagrangian method, and the
Nitsche method. For all methods, the contact pair is asymmetric, that is, the
destination contact boundary is constrained not to penetrate the source boundary, but
not vice versa. However, it is possible to set up a symmetric formulation for the contact
problem by selecting the same boundaries as source and destination.

Penalty Method
The default penalty method is rather simple and robust method to introduce the
contact conditions. Roughly speaking, it is based on inserting a stiff distributed spring,
active only in compression, between the contacting boundaries. In addition to the
robustness, it has the advantage that no special solver is required, which makes it easier
to set up multiphysics problems and time-dependent studies. However, the penalty
method only enforces the contact conditions approximately. The contact forces
computed are thus less accurate than when using, for example, the augmented
Lagrangian method, and there is always some overclosure between the contacting
surfaces. For stick friction, there is also an ‘elastic’ deformation due to the penalization
of the constraint.

When using the penalty method, there is always a tradeoff between accuracy and
stability. While a large penalty factor will reduce nonphysical overclosures, the problem
may become ill-conditioned and unstable if it is too large. It might therefore be
beneficial to accept some penetration between the contacting objects. Note, however,
that if the penalty factor is too small, the contact condition may be violated.

Augmented Lagrangian Method

The augmented Lagrangian method provides better accuracy, but at a higher
computational cost, and is often less stable from a convergence point of view. All
contact conditions are enforced in a weak or integral sense, and thus evaluated in the
integration points on the destination boundary. For normal contact, it is thus possible
for a node to have a small penetration into the source boundary, even for a well
converged solution. Both, the contact pressure and the friction forces are added as
extra degrees of freedom to enforce the contact constraints.

When using the augmented Lagrangian method, it is possible to choose between a

segregated and a coupled solution method. The segregated solution method sets up
special type of segregated solver sequence, where the extra degrees of freedom related
to contact are updated in a separate segregated solver step. As its name implies, there
is no such need when using the coupled solution method, and you can more freely set

CONTACT MODELING | 211

 up the solver sequence. This less restrictive solver requirement can be particularly
useful for multiphysics problems where the penalty method does not provide sufficient
accuracy.

Nitsche Method
The Nitsche method for contact is an extension of the general method suggested by
J. Nitsche in 1971 to weakly impose Dirichlet conditions. Conceptually, it can be
viewed as an enhancement of the penalty method where the surface traction of the
adjacent domains is used to improve the accuracy of the contact condition. Also, as for
the penalty method, it has the advantage that no special solver is required, which makes
it easier to set up multiphysics problems and time-dependent studies.

The Nitsche method is intended as an alternative to the augmented Lagrangian

method for problems where the penalty method lacks sufficient accuracy. Relevant
examples where this might be the case includes:

• Large deformation problems with hyperelasticity

• Self-contact
• Problems where the structural stiffness of the global system is of importance

For the first two problems in particular, the accuracy and stability can be further
improved by increasing the quadrature order for the contact equations, see also
Quadrature Settings. Hence, the Nitsche method uses a higher quadrature order by
default than the other contact methods.

COMSOL Multiphysics supports three different formulations of the Nitsche method:

• Symmetric
• Skew-symmetric
• Nonsymmetric

The default nonsymmetric formulation is recommended for the majority of cases and
provides a good tradeoff between performance and stability. In cases where the default
formulation fails, the skew-symmetric formulation can be an alternative, but it includes
additional equations that are expensive to evaluate that can increase the solution time.

The Nitsche method is available in the Solid Mechanics and Multibody

Dynamics interfaces.

212 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

Dynamic Methods
Both the penalty and the augmented Lagrangian methods are additionally available in
a specialized formulation intended for dynamic contact problems. Both of these are
based on a viscous formulation that for the normal contact constrains the rate of the
gap to be zero, rather than the gap. Since the formulations are viscous, the duration of
the contact should be small, otherwise the overclosure will eventually grow and violate
the nonpenetration condition. This is especially the case for the dynamic penalty
method, where the gap rate is only approximately zero.

CONTACT DETECTION
The contact search is made only towards one side of the source and destination
boundaries, determined by the positive direction of the contact normal of the selected
boundary. For contact to be detected, this means that the source and destination
contact normals must point towards each other. Generally, the contact normal
coincides with the geometry normal that is determined by the direction of the
boundary. However, there are situations where this is not the case, or where the
Contact node can control the sign of the contact normal. Some common situations are
summarized in the following:

• If the contact boundary is on the exterior of a domain, then the geometry normal
always point outwards from the domain. This is the most common case, and then
no other considerations are needed.
• In some cases, like fluid-structure interaction, there may be a domain with a moving
mesh between the two contacting objects. The source and destination boundaries
will then, in the geometrical sense, be interior boundaries. In this case, the physics
interface (Solid Mechanics or Multibody Dynamics) defines a normal vector which
is pointing outward from the boundaries that are external to the physics interface.
Thus, this case is also handled automatically.
• When a boundary without an adjacent domain is selected, you need to be careful so
that the normal is pointing in the intended direction.
• When using the Shell, Layered Shell or Membrane interfaces, contact can potentially
occur on both sides of the boundary. In a single Contact node, you can only model
contact on one side. In the Contact Surface section, you can select whether the
contact should occur on the top side or bottom side. The ‘top’ and ‘bottom’ sides
are defined by the orientation of the physics interface normal, which may differ from
the geometry normal. In these interfaces, the orientation of the physics normal is
controlled by the Boundary System that is attached to each boundary through the

CONTACT MODELING | 213

 material models. The normal direction can be reversed using the settings in the
Boundary System node.

The actual normal vector used in the contact search algorithm can be
visualized by plotting <pair_tag>.n<coord_label>. For example, the
variable p1.nx gives the x-component of the spatial normal used by
contact pair p1. Plotting the contact normal vector can be useful to verify
that the source and destination normals point towards each other.

Visualizing the Orientation of Boundaries

In 2D, a boundary is represented by a line. If you follow the line from its start point
to its endpoint, the positive normal points to the left. The line orientations can be
visualized by selecting the Show edge direction arrows check box in the appropriate View
node under Definitions.

In 3D, the rules for the orientation of a boundary are more complicated. In general,
you have to visualize the normals, in order to check its orientation. This can be done
in different ways:

• For a mesh without physics, you will need to use the general result presentation
tools. Follow these steps:
- Run Get Initial Values for an arbitrary study in order to create data for result
presentation.
- Add a 3D Plot Group under Results.
- Add an Arrow Surface plot to the new plot group.
- In the Replace Expression ( ) dialog, select the geometry normal.

214 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

• When working with the Shell or Membrane interfaces, select Enable physics symbols
in the settings for the interface. You will then see the physics normals plotted if you
select a material model like Linear Elastic Material in the Model Builder.

CONTACT BETWEEN PHYSICS INTERFACES

Contact can seamlessly be modeled between Solid Mechanics, Shell and Membrane
interfaces. Equation are only added to Contact nodes that have an applicable
destination boundary, that is, a destination boundary that intersects with the selection
of the physics interface. Properties such as contact surface offset can, however, be
considered as long as either a source or a destination boundary is applicable in the
Contact node. Shell and Membrane offsets are automatically considered on both source
and destination boundaries.

SOURCE SELECTION OUTSIDE PHYSICS INTERFACES

In most cases, the source and destination boundaries belong to the same physics
interface. The only strict rule, however, is that the destination side belongs to the
physics interface in which the Contact node resides. The source side only needs to have
a mesh and can optionally have one or more physics interfaces attached to it.

If the source boundary is not part of the a physics interface with a Contact node, the
gap is computed using only the current location of its mesh, ignoring any physical
properties that may exist there. In this case, the Contact node has no control of the
contact normal used by the source boundary. Care must therefore be taken to ensure

CONTACT MODELING | 215

 that the contact normal of the source and destination points toward each other for the
contact detection to work.

There are two main scenarios where you may want to use a source selection that is not
in the current physics interface:

Fixed Rigid Wall

If one side of a contact pair can be considered as rigid and fixed in space, then it is
sufficient to add a meshed boundary at that location without any physics.

Moving Rigid Wall

This case is similar to the previous. In order to prescribe the path of the rigid wall, add
Moving Mesh with a Prescribed Deformation node under Definitions.

Contact Pairs

SELECTING SOURCE AND DESTINATION

To decide which boundaries to assign as source and destination in a contact pair,
consider the following guidelines:

• If there is a significant difference in the stiffness in the normal direction between the
source and destination boundaries, select the stiffer side as source. This is especially
important if the difference in stiffness is quite large (for example, more than a factor
of ten).
• When the contacting parts have approximately the same stiffness, consider the
geometry of the boundaries instead. Make a concave boundary the source and a
convex boundary the destination rather than the opposite.
• If one of the boundaries belongs to a part that is rigid it should be selected as the
source boundary. Rigid boundaries can be created in several ways, for example
- The underlying domain has the Rigid Material material model.
- The boundary or its underlying domain is constrained by, for example, Fixed
Constraint or Prescribed Displacement.
- The source boundary is meshed, but has no physics attached.
- The source boundary position is controlled by Prescribed Deformation under
Moving Mesh.
• If only one side of the contact pair is within a physics interface that has a Contact
node, that side must be the destination side.

216 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

For efficiency, include only those boundaries that can potentially come in contact in
the destination selection. All equations are formed on the destination boundary, and
includes performing the contact search which can be expensive.

If it is difficult to follow the above guidelines, the same boundaries can be selected as
both source and destination. Doing so results in an unbiased (or symmetric)
formulation that is less sensitive to, for example, difference in stiffness or mesh density
between the contacting boundaries. However, such a formulation involves evaluating
equation and contact mapping at additional points and can thus be more expensive to
use.

STRATEGIES FOR SELECTING CONTACT PAIRS

A Contact node can reference any number of contact pairs.

A contact pair is just a definition, and does not have to be referenced by any Contact
node.

Neither the source, nor the destination, in a contact pair has to be geometrically
contiguous. In practice, this means that you often only need a few contact pairs in a
model. The number of pairs actually needed will be determined by how many different
settings that are required in the Contact Pair and Contact nodes.

If you have many contact pairs in your model, it is a good idea to manually set the Label
of each pair in order to simplify the identification during subsequent selections in the
Contact nodes.

AUTOMATIC GENERATION OF CONTACT PAIRS

Contact pairs can be automatically generated during the finalization of the geometry
sequence. When Action is set to Form an assembly, you can select Create pairs, and use
Contact pair as Pair type. Boundaries which are in geometrical contact with each other
will then be placed in contact pairs. All contact pairs created are automatically added
to the default Contact node. If you do not want to use all pairs, either delete the pairs
from the model or disconnect the default Contact node in the relevant physics interface
by selecting Disconnect pair. For the latter alternative, add new Contact nodes in the

CONTACT MODELING | 217

 physics interface, and select which of available pairs to actually use for the contact
analysis.

The automatic pair generation will not know which side to use as source
or destination. Based on the suggestions in Selecting Source and
Destination above, you may need to switch selections using the Swap
Source and Destination ( ) button in the Source Boundaries section of the
contact pair settings.

SMALL SLIDING CONTACT

In some situations, the relative sliding between the contacting boundaries is small. This
is, for example, often the case for shrink fit simulations, when mounting a component
using prestressed bolts, or for partial decohesion of two components. The sliding
distance can be considered as small if it is significantly less than the length of a mesh
element edge.

In such cases, it is possible to simplify the problem by selecting the Mapping Method to
be Initial Configuration in the Contact Pair node. With this setting, a material point on
the destination boundary will see the same material point on the source boundary
throughout the entire simulation; that is, the mapping is constant. This setting will
make the contact analysis run faster and convergence to be more stable.

The analysis is geometrically nonlinear also when using this option, and the contact
region can still have arbitrarily large displacements and rotations.

Friction can be modeled. Even though there is no sliding in a geometrical sense, the
difference in tangential displacement is computed.

You cannot mix contact pairs with the two different types of Mapping
Method within the same Contact node.

Meshing for Contact Analysis

Once the source and destination boundaries are chosen, you should mesh the
destination finer than the source. Do not make the destination mesh just barely finer
than the source because this can cause nonphysical oscillations in the contact pressure.
Make the element size on the destination at least two times finer than on the source.
The reason is that the algorithm is asymmetric; the points on the destination side

218 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

connects to the source side, and not vice versa. With a coarse mesh on the destination
side, a large portion of an element (or even a whole element) on the source side could
be without connection to the destination.

It is always important that the geometry is well resolved, so that a curved contact
boundary actually is seen as curved rather than “faceted”. The density of the mesh
often needs to be finer than what would be needed to resolve stresses on a similar
boundary without the contact conditions. If the normal to the contact boundary
changes much from one element to the next, there is a risk that the contact analysis
does not converge.

If the source boundary is rigid, there are no requirements on the mesh size of the
destination boundary. In this case, you may use a significantly finer mesh on the source
boundary than on the destination boundary. This is sometimes necessary in order to
resolve the geometry well. On the other hand, if you have a flat rigid boundary, you
only need to mesh it with a single element.

Settings for Contact Nodes

THE CONTACT NODE

Penalty Factor
An important parameter in the settings for the Contact node is the penalty factor.
When running into convergence problems, check the penalty factor settings and
consider changing the current value. It is used by all contact methods, but its
interpretation differs:

• In the penalty method, the penalty factor has a straightforward interpretation as the
stiffness of a distributed spring connecting the two contacting boundaries. A higher
value will decrease the unphysical penetration, but can be detrimental to the
numerical conditioning of the stiffness matrix. A too small value can, however, result
in violation of the contact condition.
• In the augmented Lagrangian method, the penalty factor is a numerical parameter,
which affects the convergence properties of the algorithm. Its interpretation is
different depending on the chosen solution method. For a segregated method, the
penalty factor mainly affects the rate of convergence of the outer iteration loop; a
higher value typically leads to faster convergence, but can decrease the stability of
the algorithm for certain problems. For the coupled method, the penalty factor
mainly affects the structure of the underlying equations, and the solution is typically

CONTACT MODELING | 219

 less sensitive to the value of the penalty factor. However, the convergence properties
can be improved by tuning the penalty factor.
• In the Nitsche method, the penalty factor is a stabilization parameter which can
affect the robustness of the solution, but to a lesser extent the accuracy. The suitable
value of the penalty factor depends on the chose formulation, where the
skew-symmetric and nonsymmetric formulation are typically less sensitive than the
symmetric formulation.

In the augmented Lagrangian method and the Nitsche method, the value
of the penalty factor does not affect the accuracy of the final solution, like
it does in the penalty method.

High values of the penalty factor can lead to an ill-conditioned stiffness matrix and
convergence problems in the Newton iterations. This is often identified by that the
damping factor reported by the solver becomes less than 1 for many Newton iterations,
or by that the structure “jumps” into an unphysical state. Large errors returned from
the linear solver are also an indication that the stiffness matrix is ill-conditioned. If this
occurs, decrease the penalty factor.

The default value for the penalty factor is based on a characteristic stiffness. The
default is an “equivalent” Young’s modulus (Eequ) of the material on the destination
side. For linear elastic isotropic materials, Eequ is the actual Young’s modulus. For
other types of materials, Eequ is defined by an estimate of a similar stiffness at zero
strain. For materials that are user defined or in other ways nonstandard (for example,
anisotropic with large differences in stiffness in different directions), Eequ might need
to be replaced with another estimate.

For nonlinear materials in general, and for elastoplastic materials in

particular, there can be a significant change in stiffness during the solution
process. Choose the source and destination boundaries accordingly. You
may even have to adjust the penalty factor as the solution progresses when
such materials form a contact boundary.

When using the augmented Lagrangian formulation with a segregated solution

method, having a well-tuned penalty factor is important for the performance of the
outer contact iterations. The default value is selected as a compromise between speed
and stability, but with more weight on stability. The strategy is for each new step
(parametric step or time step) to start with a softened penalty factor, which is then

220 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

increased over the first four iterations. The purpose is to stabilize the problem in case
there are large overclosures during the first iterations. This is called relaxation.

In a situation where the contact is well established, using relaxation will however cost
extra iterations, and it may even lead to a loss of convergence.

For this method, the penalty factor can be tuned in several ways. You have three basic
choices, ranging from simple to advanced:

• With a Preset penalty factor, you can choose having it tuned for Stability, Speed, or
Bending. With Stability, relaxation is used in every step. With Speed, a constant
penalty factor is used, and the value that is used is equal to the final value obtained
when using Stability. For Bending, a constant and low penalty factor is used in all
iterations. The value corresponds to the initial value when using Stability. As the
name implies, this option is intended for bending dominated problems where the
structural stiffness is much lower than the bulk stiffness of the material.
• With Manual tuning, you can adjust the magnitude of the penalty factor, and also the
relaxation strategy.
• With User defined, you can enter any expression for the penalty factor.

For details about these settings, see the documentation of the settings for
the Contact node.

Some hints for selecting the penalty factor for the segregated augmented Lagrangian
method:

• Use relaxation only when large changes in the contact state is expected.
• If an analysis takes a long time, check the convergence graphs. If the contact
variables show a steady, but slow, convergence it may help to increase the penalty
factor. Increase by a factor of 2 – 5.
• If a model fails to converge, and the contacting parts appear to have moved far from
each other, it is probable that the penalty factor is too high. You can then either
decrease the total stiffness or add more relaxation.

Trigger Cutback
If, during the iterations, a contact problem comes into a state where it is far from the
converged solution, it is unlikely that the solution will ever converge. In such a case,
much computing time can be spent before the maximum number of iterations is
reached, and the solver makes an attempt with a smaller time or parameter step. To

CONTACT MODELING | 221

 speed up this process, you can select the Trigger cutback check box when using the
augmented Lagrangian method. You can then enter a logical expression that will force
the solver to immediately abandon the iterations and try a smaller step when fulfilled.
Such an expression can, for example, be a maximum displacement (like solid.disp >
5[mm]), based on what is physically realizable for the structure. The expression is
evaluated in all points on the boundary, but you can also use a global value, like an
average or a maximum.

Contact Surface Offset and Adjustment

It is possible to assign an offset to both the source and destination boundaries. When
an offset is given, the boundary used in the computations is not the geometrical
boundary, but a virtual boundary displaced by the offset value. You can use this option
for several purposes:

• When analyzing problems with for example shrink fit, nominal dimensions can be
used for the geometry, and the overclosure included in the gap by using a positive
offset.
• When there is a small clearance between two boundaries, a negative offset can be
used instead of changing the geometry.
• If geometries are such that a large overclosure exists in the initial configuration, the
contact algorithms may not converge. You can then add a negative offset, which is
slowly removed by letting it depend on time or on the parameter in the parametric
continuation solver.
• A positive offset can be used to avoid a complete collapse of a mesh that exists
between the source and destination boundaries. This is discussed in more detail in
Multiphysics Contact Analysis.

When the source and destination boundaries are curved, the discretization introduced
by the meshing may lead to small variations in the computed distance between the
source and destination boundaries, even though the geometrical shapes of the two
objects are ideal. When modeling for example a shrinkage fit, this effect can cause
significant fluctuations in the contact pressure. If you select Force zero initial gap, the
computed distance from destination to source will be adjusted by the initial gap
distance detected by the contact search. Positive gap distances smaller than the
tolerance gap are adjusted to be zero. By default, gap is set to Inf, which means that
all gaps and overclosures detected are adjusted to be zero. This adjustment can be
combined with an offset. The offset is applied to the adjusted gap value.

222 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

It is only the gap computation that is affected by this setting, the mesh as such is not
adjusted. This type of adjustment is most useful when the sliding is small, so that the
gap distance is always computed between the same points on source and destination.

You can only apply an offset to the source boundaries if they belong to the same physics
interface as the destination boundaries.

Initial Value
In the augmented Lagrangian method, where the contact pressure is a dependent
variable, it can be given an initial value. In force-controlled contact problems where no
other stiffness than the contact prohibits the deformation, the initial contact pressure
is crucial for convergence. If it is too low, the parts might pass through each other in
the first iteration. If it is too high, they will never come into contact.

Discretization
When using the augmented Lagrangian method it is possible to change the type and
order of the shape functions used for the contact pressure and friction force degrees of
freedom. The default is linear shape functions, which ensures that there are no
over-constraints in the contact interface. It is allowed to use a shape function order that
is equal to or lower than the order of the displacement field. Increasing the order of
the contact variables from the default can increase the accuracy of how well the contact
conditions are enforced, but can impair convergence and increase the computational
cost.

For a discretization other than Linear, the lumped solver is no longer optimal for the
contact pressure update when using a segregated solution method. In such cases, a
standard segregated step should be used. The default solver generation takes this into
account, but if you later modify the discretization, you should update the solver
sequence.

Quadrature Settings
The weak equations set up by the Contact node and its subnodes typically involve
discontinuous functions. These originate from the contact mapping, where the source
and destination meshes, in most cases, are nonconforming. The default quadrature
used in the numerical integration of these integrals is equal to the order used by the
displacement field. For a quadratic displacement field, this means integration order
equal to four.

In most situations, the default quadrature provides sufficient accuracy of the numerical
integration. However, it is sometimes necessary to increase the integration order for
the contact weak equations. This can improve the stability of the contact model since

CONTACT MODELING | 223

 it reduces numerical errors that might occur during assembly of the system matrices.
Note that using a too high integration order can significantly increase the cost of
assembling the matrices.

Jacobian Contribution
In the Advanced section of the Contact node, there is an option to specify the type of
Jacobian contribution from the contact equations. The default Automatic option will
choose a suitable setting depending on the mapping method used by the contact pair.
However, if controlled manually the Nonsymmetric option is the preferable choice
especially when the source boundaries undergoes large deformations, since it is more
robust. The Symmetric option can be attractive for large models since it preserves the
symmetry of the global stiffness matrix, as long as no other features cause it to be
nonsymmetric. This can decrease the solution time and memory requirements when
solving the model.

THE FRICTION NODE

When adding a Friction node, you can specify a constitutive model (friction model) for
the behavior of the tangential contact. This model includes conditions for switching
between sticking and sliding, as well as computation of the current friction forces.

Friction Parameters
The friction model specifies the threshold for the friction force in the contact pair. If
the computed (trial) friction force is above this threshold value, the contact is in a stick
state; if the (trial) friction force is above the threshold, the contact is in a slip (or
sliding) state.

Two predefined friction models are available based on the classical Coulomb law,
where the friction force is proportional to the contact pressure through the friction
coefficient. Both Coulomb laws are additionally generalized by allowing specification
of minimum (cohesion) sliding resistance and a cap that sets the maximum tangential
traction.

It is also possible to define the threshold for the friction force as an arbitrary expression
that may depend on any quantity in the model, for example temperature or position.
The only limitation is that the expression may not implicitly depend on the current
value of the friction force.

Friction Force Penalty Factor Control

This section provides similar settings as described in Penalty Factor of the Contact
node, but the penalty factor is here used to regularize the stick constraint. However,

224 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

the same considerations for how to set an appropriate value apply. For convenience, it
is also possible to utilize the penalty factor set in the parent contact node.

This section is not available when the Nitsche method is used. The same penalty factor
as set in the parent Contact node will be used for friction as well.

THE SLIP VELOCITY NODE

The Slip Velocity node facilitates a simplified form of slip friction modeling, which can
be used in the case that the direction and speed of the sliding is known. The same
friction models as for the Friction node are available. However, it is here assumed that
the tangential contact is in a sliding state, and that the slip velocity is known
beforehand. The latter is supplied to the feature as a user input in the local coordinate
system.

Knowing the slip velocity greatly simplifies the computation of friction forces. There is
no need to determine the transition between stick and slip contact, which can be
difficult.

THE ADHESION NODE

When using the penalty method, you can specify that the boundaries in the contact pair
should stick to each other after coming into contact by adding an Adhesion node.

The adhesive layer is conceptually without thickness, but by specifying an offset in the
Contact node, you can to some extent include the dimensions of the adhesive layer.

The adhesive layer always has a finite stiffness. For a glue layer, this represents the true
stiffness. For a more conceptual joining of two boundaries, this stiffness should be
considered in the same way as the penalty stiffness in the contact formulation. The
stiffness can differ between tension and compression: In compression the stiffness is
always taken as the penalty stiffness, whereas you can change the tensile stiffness.

THE DECOHESION NODE

When adhesion is active, it is possible to break the bond between the source and
destination boundaries by adding a Decohesion subnode to Contact. To model
decohesion, it is required that an Adhesion node is present and active in the same parent
Contact node.

Decohesion defines a cohesive zone model (CZM) based on interface damage

mechanics on the adhesive layer. Damage is assumed to be a scalar variable that initiates
as 0 and grows to 1 during decohesion, and in principle degrades the stiffness of the
adhesive layer. Since damage is a scalar, both the normal and tangential stiffness

CONTACT MODELING | 225

 components degrade simultaneously, irrespective of whether the actual loading
direction. However, the penalty stiffness of the contact condition is not affected by
damage.

Two alternative CZM are available. The Displacement-based damage models defines
damage growth as a function of a mixed mode displacement quantity. It comes with
several traction separation laws that associate the onset of damage with the peak
strength of the interface. For some of them, it is possible to choose between different
mixed mode failure criteria. The Energy-based damage models define damage growth
as a function of the stored undamaged elastic energy density of the interface. It also
comes with several different traction separation laws. However, these are more general
and define the onset of damage at an arbitrary elastic energy density. In principle, you
can define the model so that damage initiates immediately during loading of the
adhesive layer, that is for zero energy density. The strength of the interface is then
determined by the critical energy release rate and the shape of the damage evolution
function. In this way, the energy-based damage models can be viewed as a
regularization of linear elastic fracture mechanics.

Decohesion is an inherently unstable process. The elastic energy in the strained

adhesive layer is released during decohesion. Numerically, the decreasing branch of the
traction-separation curve manifests itself as a local negative stiffness. Such problems are
only possible to solve as long as the surrounding material can absorb the released
energy. The numerical stability is, furthermore, closely coupled to the physical stability
of the structure. The following points can help to set up a model with decohesion and
to overcome problems with convergence.

• If the structure is in a load-controlled situation, there is no possibility to continue

the analysis when the peak external load has been exceeded. Physically, this
corresponds to a sudden collapse of the component, or in the case of decohesion, a
rapid breakage of the adhesive layer.
• If the structure is loaded by prescribed displacements, it is usually possible to
continue the analysis further. It may, however, happen that the stored elastic energy
in the structure is large enough to force a complete breakdown of the adhesive layer
once a certain external displacement is reached. If possible, use prescribed
displacements in a decohesion analysis, and evaluate the applied load from the
reaction forces.
• Sometimes it is not possible to use prescribed displacements, for example if the load
is distributed. You can then add a Global Equation to control the loading rate by

226 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

 some other quantity that increases monotonically. This is the same technique as the
one used for post-buckling problems.
• To improve the robustness of the solver, it is sometimes beneficial to modify the
settings in the Method and Termination section of the Fully Coupled or Segregated
nodes in the solver sequence. For example, allow a larger number of iterations or try
a different nonlinear method. Often, the Constant (Newton) method can improve the
convergence of models with decohesion.
• The robustness of the solver can also be improved by modifying the parameter or
time stepping algorithm. For a stationary study, you can tune the step size in the
Parametric node, and for a time-dependent study, you can modify the time stepping
of the Time-Dependent Solver. A good idea is often to reduce the maximum allowed
step size of the solver and to allow for smaller step size than the default. Note that
if the maximum step size allowed is too large, the solver might bypass the
decohesion process altogether; in other words, even though a converged solution is
obtained, it might be invalid.
• The solution of the unstable failure due to decohesion is, to some degree, always
mesh dependent; see, for example, Ref. 1. It is therefore good practice to make sure
that the mesh of the interface and in its vicinity is fine enough to allow the energy
released during decohesion to properly redistribute in the structure. This can help
avoid solution jumps where several mesh elements are completely damaged in a
single step. Such solution jumps can be difficult for the solver to get pass, and even
if it does, the solution after the jump might be invalid.
• The true unstable failure of decohesion is a dynamic event. This can be analyzed
using a time-dependent solver, but the computational cost can be high. The inertial
forces will then balance the released energy, which is transformed to kinetic energy.
For time-dependent studies, it is possible regularize the CZM with a viscous delay by
selecting Delayed damage in the Regularization list. This option adds a delay to the
release of energy, which is controlled by the Characteristic time . Using this option can
help to suppress the instability of the solution when the step size or mesh size is too
large. If the viscous damage is used to stabilize a rate-independent decohesion

CONTACT MODELING | 227

 problem, the value of  must be chosen with care. As a rule of thumb,  should at least
be one or two orders of magnitude smaller than the expected time step.

For an example showing a decohesion analysis, including how to use a

global equation to control an unstable problem, see Mixed-Mode
Debonding of a Laminated Composite: Application Library path
Structural_Mechanics_Module/Contact_and_Friction/
cohesive_zone_debonding.

THE WEAR NODE

By adding a Wear subnode to a Contact node, it is possible to model adhesive or
abrasive wear of the material when the contacting boundaries are sliding along each
other. Since wear involves solving evolution equations, the Wear node only adds a
contribution for time-dependent studies. Moreover, wear is typically a slow process
where dynamic effects are of small significance. You should, therefore, usually set
Structural transient behavior to Quasistatic in the Structural Transient Behavior section
of the physics interface settings.

The most general technique to model the removal of material during the wear process
relies on the deformed geometry concept. When selecting the Deformed geometry
formulation, the wear feature adds a (hidden) Deforming Domain feature that controls
the material frame through an adaptive mesh smoothing. The wear, as computed in the
Wear node is fed as a (hidden) Prescribed Normal Mesh Displacement boundary
condition to the deforming domain, and thus describes the actual removal of material
from the geometry. When using this formulation, you must be aware that the adaptive
mesh means that state variables stored in Gauss points, for example plastic strains or
creep strains, will not represent the same material points all the time. Whether or not
this effect is acceptable must be judged on a case-by-case basis. Unless the amount of
material that is removed is large, or gradients are strong, this is mainly an issue close
to the boundary where material is removed by the wear process.

Alternatively, wear can be incorporated as an offset in the contact condition. This

formulation is computationally less expensive, and is suitable as long as only small
amounts of material are removed, and the wear does not change the orientation of the
normal to the boundary significantly.

The slip velocity used for the wear computation can be obtained from either a Friction
node or a Slip Velocity node, so one of these two nodes should be present and active
under the same Contact parent node. For the Generalized Archard wear model, this is a

228 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

requirement. In most cases, the orientation of the slip velocity is known a priori in a
wear analysis, in which case Slip Velocity provides the more efficient solution.

In general, modeling wear on the destination side is slightly more accurate, since it is
there that the contact pressure and slip velocity are originally computed. When
modeling wear on the source side, these quantities are mapped from the destination
boundary. Multiple Wear nodes under the same Contact node contributes with each
other, which means that it is possible to model wear on both source and destination
simultaneously. However, adding multiple wear contributions to either source or
destination may give unphysical results. On the source side it is possible to also use the
Rigid Material material model; this is not permitted on the destination due to general
restrictions of the Contact node.

• Documentation of the Contact, Friction, Slip Velocity, Adhesion,

Decohesion, and Wear nodes.
• Contact Analysis Theory in the Structural Mechanics Theory chapter.

Quasistatic Contact Analysis

When including contact in a stationary or quasistatic simulation, make sure that the
bodies are sufficiently constrained. If the bodies are not in contact in the initial
configuration, and there are no constraints on them, there will be possible rigid body
displacements. This will cause the solver to fail and must be avoided.

For a more detailed discussion about sufficient constraints, see

Constraints under Stationary Analysis.

Sometimes, as when simulating mounting processes, the structure is not fully

constrained until the contact is fully established. There are some strategies for how you
can deal with this problem.

• Create the geometry or set initial values for the displacement variables so that there
is a small penetration in the initial configuration.
• Use boundary conditions giving a prescribed displacement rather than a prescribed
force. When possible, this is usually the best way to stabilize problem. Note that you
can always obtain the force actually applied from the reaction forces.

CONTACT MODELING | 229

 • Add a temporary weak spring during the beginning of the simulation. Assuming
that a parameter p, ranging from 0 to 1, is used for applying the external load, you
can introduce a stabilizing spring with stiffness kx in the x direction as
–  p  10 
k x = k  1 – p 2

and similarly in any other direction that needs to be restrained. It is not important
whether the spring is applied to domains, boundaries, or edges, but it should not
create significant local forces. The value for the stiffness k should be chosen so that
the force generated by the spring balances the external load at a sufficiently small
displacement. A too weak spring will give a too large initial overclosure of the
contact boundaries. A too stiff spring might influence the solution too much.

Dynamic Contact Analysis

The contact formulations in COMSOL Multiphysics can be used in transient as well as
quasistatic analyses. In truly dynamic problems, where inertial effects are significant, a
contact formulation must conserve fundamental quantities, such as linear and angular
momentum and energy, across the contact pair. An important class of dynamic
problems, where these quantities must be conserved, is impact analysis.

To model dynamic contact events, two specialized contact methods are provided, the
Penalty, dynamic and the Augmented Lagrangian, dynamic methods. Both are based on
a viscous formulation that constrains the gap rate to be zero, ensuring that the normal
contact is dissipative and does not introduce any spurious energy contribution to the
system. Since the methods are dissipative, they are mainly intended for short duration
events, such as soft impact between two bodies. For prolonged interaction between
two bodies, energy dissipation can become significant, and overclosures can become
large, since the gap rate is only approximately zero. Both the dissipation and the
accuracy are controlled by a penalty factor that for these two methods conceptually
represents a dashpot, rather than a spring. It therefore has a characteristic time user
input that sets its magnitude. As a rule-of-thumb, it should be of the order of the
contact event duration, but the best choice must be decided on a case-by-case basis.

The Penalty, dynamic method also provides the possibility to combine the stiffness and
viscous based penalization of the normal contact. For impact analysis, it is often best
to use only the viscous formulation by setting the stiffness Penalty factor control to
Viscous only.

When modeling dynamic contact, the main interest is often the kinematics between the
contacting bodies. If you rely on the (default) adaptive time-stepping algorithm, the

230 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

solver typically also tries to resolve the wave propagation in the domains adjacent to
the contact pairs. This can cause unnecessarily small time steps and increase the
computation time. To avoid this, you can modify the solver to use a manual
time-stepping algorithm in the settings for the Time-Dependent Solver in the solver
sequence. Make sure to use time steps that are small enough to capture the contact
event, otherwise spurious energy contributions can result, and cause the problem to
“blow up”.

The time-dependent solvers in COMSOL Multiphysics introduce numerical damping

to stabilize the time stepping. This kind of stabilization is often necessary. However,
excessive numerical damping runs the risk of removing vital information from the
simulation. For this reason, the BDF solvers should be avoided for dynamic contact
analyses. The default solver suggests the generalized- solver when inertial terms are
included in the structural mechanics problem.

Regardless of the method that is used, and how the solver is set up, it is good practice
to do an a posteriori check of conservation of momentum and energy, to ensure that
the solution is acceptable.

For examples of dynamic contact problems, see

• Transient Rolling Contact: Application Library path

Structural_Mechanics_Module/Contact_and_Friction/
transient_rolling_contact.
• Impact Between Two Soft Rings: Application Library path
Structural_Mechanics_Module/Verification_Examples/ring_impact.

In both examples, an energy balance check is included. The latter model

also verifies the conservation of momentum.

Multiphysics Contact Analysis

Two different classes of multiphysics contact problems will be described in this section.

FLUXES THROUGH CONTACT SURFACES

In some contact problems, there is some kind of flux from one domain to another
through the contact zone. This can for example be a heat flux, an electric current, or
moisture transport. The common property here is that the other physics fields than the
displacements are present in domains where the solid mechanics problem is solved.

CONTACT MODELING | 231

 Typically, there will be a more or less perfect insulation as long as there is no contact,
but as soon as contact is established there will be a flux through the contact area.

This class of problems often exhibit a high degree of nonlinearity, which may lead to
convergence problems in the nonlinear solver. As an example, consider heat transfer
through the contact area, where initially only a small spot is in contact. The solution
for the temperature field is then extremely sensitive to the size of the contact area. If,
at the same time, the solid deforms due to thermal expansion, there may be large
changes in the contact area between iterations,

It is important to resolve the size of the contact area accurately, that is, to
use a very fine mesh in the contact area when modeling fully coupled
multiphysics problems.

If the contact area is larger, a fine mesh is not required because then the temperature
solution is not that sensitive to the size. If possible, start with an initial configuration
where the contact area is not very small.

You can use the contact variables (gap and contact pressure) in expressions for
quantities in other physics interfaces. As an example, a thermal resistance in the contact
region can depend on the contact pressure.

In many cases, the penalty method is preferred in multiphysics contact problems

because of its better stability and less restrictive requirements on solver selections. If
the contact conditions depend strongly on the contact pressure, use the augmented
Lagrangian method because if its higher accuracy.

FIELDS EXIST IN THE GAP

In this class of problems, a field exists between the domains controlled by solid
mechanics.

This is the case in, for example, fluid-structure interaction (FSI) problems. Here, the
equations in the fluid are solved on a domain with a moving mesh, so that the shape
of the fluid domain is controlled by the displacements of the solid. Another case of the
same type is when there is an electric field in an air gap.

If contact is established, the mesh in the original gap between the source and
destination boundaries will collapse. This must be avoided. The remedy is to add an
offset in the contact settings to either the source boundary, the destination boundary,
or both. If you do this, contact forces will be transmitted without the geometrical gap
being fully closed.

232 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

If, for example, this technique is used when modeling a valve, there will still be some
small flux even though the valve is closed, since there is a geometrical gap with the
width of the artificial offset. By choosing a suitably small offset, you can however make
that flux negligible.

Special Types of Contact Problems

INTERFERENCE FIT
Interference fits can be analyzed using contact modeling. This is necessary if you are
interested in checking the connection with respect to, for example, slipping or local
stresses.

There are two possible approaches for modeling interference fits, both equally valid:

• The actual geometry of the two parts before assembly is modeled. In this case, there
will be some overlap between the two domains.
• A nominal geometry in which the contacting boundaries have the same location is
created. In this case, the overlap is described as part of the contact modeling.

An imported CAD geometry can use either of these approaches, depending of the
strategy used during the geometry creation. Often, the geometrical parts are modeled
as nominal, and instead equipped by tolerance information that describe the amount
of interference.

True Geometry
With a true geometry, you can often immediately solve the contact problem.
Sometimes convergence problems may, however, appear, in particular if the material
model is nonlinear. The cause is often that the initial overlap is too far from the final
solution.

To deal with such a problem, add an offset in the in the settings for the Contact node.
The offset should be defined by a parameter, so that the boundaries of the two domains
are barely in contact in the initial state. Now, the offset can be reduced to zero
step-by-step, using an auxiliary sweep in the solver.

CONTACT MODELING | 233

 Nominal Geometry
When working with a nominal geometry, you always need to add an offset in the
Contact node. The offset equals the size of the interference. If needed for convergence
reasons, ramp up the offset using an auxiliary sweep in the solver.

Interference Fit Connection in a Mountain Bike Fork: Application

Library path Structural_Mechanics_Module/Contact_and_Friction/
mountain_bike_fork

SELF-CONTACT
To model self-contact, include the same boundaries in both the source and destination
selections of the Contact Pair definition. This will cause the boundaries to act as both
source and destination in the contact search and mapping. For mechanical contact, this
results in a unbiased (or symmetric) contact formulation, as the contact conditions are
formulated on both sides of the contact pair. Note that a source is not allowed to
partially intersect the destination when used for mechanical contact.

This technique to model self-contact means that some of the considerations discussed
in this chapter regarding contact modeling do not apply. For example, instead of the
recommendations in Meshing for Contact Analysis, it is recommended to use a
uniform mesh element size along the contacting boundaries. Self-contact is a case
where it might be necessary to increase the quadrature order used in the weak
equations, see Quadrature Settings.

For examples of self-contact problems, see

• Self-Contact of a Loaded Spring: Application Library path

Structural_Mechanics_Module/Contact_and_Friction/
loaded_spring_contact.
• Contact Analysis of a Rubber Boot Seal: Application Library path
Nonlinear_Structural_Materials_Module/Hyperelasticity/rubber_boot_seal.

Fallback Conditions to Contact Regions

In the Solid Mechanics, Shell, and Membrane interfaces, certain boundary conditions
may act as fallback conditions to a contact region, or any pair region, if their selection
overlaps with the source and destination boundaries. The Layered Shell interface does

234 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

not support any fallback conditions. When a boundary condition is considered as a
fallback, the condition is only active for regions that are not in contact.

In the Solid Mechanics interface, the following boundary conditions are by default
recognized as fallback conditions if their selection intersects any source and destination
boundary:

• Boundary Load
• Spring Foundation
• Added Mass
• Free

In the Shell and Membrane interfaces, the following boundary conditions can be
recognized as fallback conditions:

• Face Load
• Spring Foundation
• Added Mass
To enable the fallback condition in these nodes, change the Allowed region to Fallback
and nonpair regions in the Applicable Pair Regions section of settings window of the
node. To display the Applicable Pair Regions section, click the Show More Options
button ( ) and select Advanced Physics Options in the Show More Options dialog box.

The most common case when a fallback condition is used is when there is a pressure
load acting on the part of the boundary that is currently not in contact. In this case
you would add a Boundary Load or Face Load with a selection that intersects with the
source and destination boundaries.

Solver Settings for Contact Analysis

This section provides useful tips for tuning solver settings for contact analysis in
general, and tips specific to different methods.

GENERAL SETTINGS
The following solver settings can help to successfully perform contact simulations in
general:

• In a contact analysis, you almost invariably use an incremental approach. It is

possible to solve a problem without friction in a single stationary load step, but such
an approach will often fail to converge. In a stationary analysis, you should instead

CONTACT MODELING | 235

 use the parametric continuation solver, and gradually increase the load or
displacement. Enable it by selecting Auxiliary sweep under Study Extensions in the
settings for the Stationary solver.
• Use a direct solver, rather than an iterative solver, as linear equation solver if the
problem size allows it. Direct solvers are less sensitive and can provide better
convergence.
• As a default, the double dogleg nonlinear solver is selected when a stationary study
is generated and Contact nodes are present in the model. For the majority of contact
problems this solver has more stable convergence properties than the Newton
solver, which is the default solver for most other problems. Using otherwise similar
settings, the double dogleg solver tends to be somewhat slower than the Newton
solver on problems where both solvers converge. It is, however, often possible to
take larger parameter steps when using the double dogleg solver. For some
problems, the Newton solver can still be the better choice, so if you experience
problems using the default settings, try to switch solvers.
• For contact problems, it is often necessary to let the parametric solver use a defensive
strategy when going from one parameter step to the next. This can be controlled by
setting the value of Predictor in the Parametric. By default, the parametric solver will
do so by setting the predictor to Constant when contact is present. However, it can
sometimes be more efficient to use a more aggressive strategy by setting it to Linear.
• Always solve contact problems that involve friction or decohesion incrementally,
using a parametric or time-dependent solver. The evolution of the friction forces is
history dependent. For contact problems without friction, an incremental strategy
is not necessary but often a good choice.
• If the model includes friction, try solving the problem without friction first if
possible. When the study runs absent friction, enable friction again.
• The convergence of many contact problems can be improved by modifying the
parameter or time step algorithm. For a stationary study, you can tune the step size
in the Parametric node, and for a time-dependent study, you can modify the time
stepping in the Time-Dependent Solver.
• For models that include decohesion, see the suggestions under The Decohesion
Node.

THE PENALTY METHOD

Since the penalty method only adds a weak contribution to the physics, there are no
special solver requirements, apart from the suggestions that apply to contact analysis
in general. If the simulation shows poor convergence, or even diverges, it is for the

236 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

penalty method often related to the choice of the penalty factor. A too high penalty
factor may cause the ill-conditioning, which can manifest itself either in poor
convergence of the Newton solver or large errors in the linear solver. If this is
encountered, consider using a smaller penalty factor. However, a too small penalty
factor will deteriorate the accuracy of the simulation, and contact might even be lost if
the overclosure becomes too large.

THE SEGREGATED AUGMENTED LAGRANGIAN METHOD

The segregated solution method for the augmented Lagrangian formulation
introduces extra contact degrees of freedom for the contact pressure and the friction
forces. It also relies on a special solver sequence to obtain the correct solution. Some
general considerations when using this method are:

• The convergence check relies on the scaling of the degrees of freedom, but since
contact pressures and friction forces often are zero over parts of the simulation, you
should not rely on automatic scaling. When the solver sequence is first created, both
contact pressure and friction forces are given a manual scaling which is relevant for
typical metal-to-metal contact. You should in most cases change this to values
appropriate for your application. The variable scaling is accessed under Dependent
Variables in the solver sequence. Set the scale for each variable to a value that is
representative for the expected result. Too large values may give a too early
convergence, while too small values may lead to an excessive number of iterations.
• The default solver sequence generates one lumped step in the segregated solver for
each Contact node. This split of variables into different lumped steps does not
influence the solution as such; you can equally well group the contact variables in a
single lumped step. Each lumped step will however generate an individual curve in
the convergence plot, making it easier to pinpoint the source of possible
convergence problems. You can also increase the granularity even more by changing
Solver log to Detailed in the Advanced node in the solver sequence. This will give a
separate convergence curve for each dependent variable.

The default solver generates a solver sequence that is stable and gives and accurate
solution for the majority of contact problems. However, if convergence is difficult
other settings can be tested:

• The Segregated solver is generated with a termination technique set to Tolerance.

This setting ensures the convergence of the contact degrees of freedom, that is, that
their value only changes within the specified tolerance during the latest segregated
iteration. If, however, convergence of the contact DOFs is of less importance, you
can accept the solution after n segregated iterations instead. To do this, set the

CONTACT MODELING | 237

 termination technique to Iterations or tolerance and specify the maximum number
of segregated iterations. The solver will then continue to the next step if the
tolerance criteria is fulfilled, or if the maximum number of iterations is reached.
Note that the solution of the segregated groups is still converged in each segregated
iteration.
• For efficiency purposes, the nonlinear solver in the Segregated Step that includes the
displacement field is by default set to accept the solution after seven iterations,
regardless of convergence or not. If you notice in the solver log that the solution is
far from convergence after these seven iterations it can be necessary to change this
setting. Updating the contact DOFs with a nonconverged solution can cause the
overall problem to diverge. By changing the termination technique to Tolerance, the
segregated solver will instead do a cutback if such a situation is encountered.

THE COUPLED AUGMENTED LAGRANGIAN METHOD

The coupled solution method for the augmented Lagrangian formulation introduces
special contact degrees of freedom for the contact pressure and the friction forces.
Otherwise it places no special restriction on the solver sequence. Some general
considerations when using this method are:

• The convergence check relies on the scaling of the degrees of freedom, but since
contact pressures and friction forces often are zero over parts of the simulation, you
should not rely on automatic scaling. When the solver sequence is first created, both
contact pressure and friction forces are given a manual scaling typical for
metal-to-metal contact. You should in most cases change this to values appropriate
for your application. The variable scaling is accessed under Dependent Variables in the
solver sequence.
• Since the solution to the augmented Lagrangian can be non smooth, the default
double dogleg nonlinear solver in stationary studies is sometimes too conservative.
The convergence can in such cases often be improved by using a Newton solver, for
example, the Constant (Newton) with a full Jacobian update.
• Although the penalty factor does not affect the accuracy of the solution, it can have
significant influence on the convergence properties of the model.

THE NITSCHE METHOD

Since the Nitsche method only adds a weak contribution to the physics, there are no
special solver requirements, apart from the suggestions that apply to contact analysis
in general. If the simulation shows poor convergence, or even diverges, it is often
related to the choice of the penalty factor. A too low value may cause the solution to

238 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

be unstable, while as for the penalty method, a too high value may cause
ill-conditioning.

DYNAMIC CONTACT
The dynamic contact methods in general inherit the properties and suggested solver
settings from the corresponding standard method. With regards to solver settings, the
main difference for the dynamic methods is how to set up the time-dependent solver,
where it is recommended to use manual time stepping, see Dynamic Contact Analysis.

Monitoring the Solution

It is often useful to monitor the solution during a contact analysis. This can be done
in different ways.

Using the Results while solving functionality in the study step is a good practice. You
can either use a stress plot, or a plot of the contact pressure. In most cases, the scale of
a deformed plot should be set to 1 when monitoring contact problems. If you select
Results while solving in the Segregated node, the plot is updated after each iteration,
thus allowing you to monitor the convergence in detail.

For each contact pair, two global variables that can be used in probe plots are available.
These are the maximum contact pressure (<phys>.Tnmax_<pair>) and the minimum
gap distance (<phys>.gapmin_<pair>).

Looking at the convergence plot will give valuable information about the convergence
properties. There will, as a default, be one graph per Contact node in the Model Tree,
which will help you pinpoint the source of a convergence problem. You can also
increase the granularity even more by changing Solver log to Detailed in the Advanced
node in the solver sequence. This will give a separate convergence curve for each
dependent variable.

You can also select to include information about the contact state in the solver log. To
do that, select the Add contact status to solver log check box in the Advanced section of
the settings for the Contact node. For each contact pair, messages like

69 points of 120 are now in contact.

33 points started to stick. 72 points are now sticking.
12 points started to slide. 47 points are now sliding.

will be generated for each time or parameter step. Only changes are reported.

CONTACT MODELING | 239

 Dependent Variables in Contact Analysis
The Contact and its subnodes will generate a number of degrees of freedom depending
on the settings and study type. You will see these degrees of freedom appear under
Dependent Variables in the solver sequence. There are two types of extra variables
created:

• Variables changed until convergence is reached during the iterations. These variables
appear in the Lumped Step nodes in the Segregated solver or in the Fully Coupled
node.
• Variables used to store the state, once the iterations have converged for a certain
time or parameter step, so called internal degrees of freedom. Such state variables
are not immediately visible in the solver sequence, but it you select the Displacement
Field node under Dependent Variables in the solver sequence, you will see them listed
as ‘Internal variables’.

If you change settings in the Contact or Friction nodes after the solver
sequence has been generated, dependent variables may be added or
removed. The second case is never a problem, but when new dependent
variables are created, they are not automatically added to the groups in the
segregated solver. You may then encounter the error message
“Segregated solver steps do not involve all components.” You
can then either regenerate the solver sequence, or manually insert the
missing variables into the Lumped Step node.

In Table 2-11 the dependent variables that can be created by the Contact and its
subnodes are summarized. To shorten the variable names, the full scope has been
removed. As an example, the contact pressure variable for pair p1 in component comp1,
generated in the Solid Mechanics interface solid, will have the full name similar to
comp1.solid.Tn_p1. In the table, it is shown as Tn.

TABLE 2-11: DEPENDENT VARIABLES OCCURRING IN CONTACT ANALYSIS

VARIABLE NAME DESCRIPTION EXPLANATION CREATED SOLVER NODE

Tn Contact pressure The contact Contact: Lumped

pressure in the Augmented Step
normal direction Lagrangian used
Tt Friction force The friction force Friction: Lumped
vector Augmented Step
Lagrangian used

240 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

TABLE 2-11: DEPENDENT VARIABLES OCCURRING IN CONTACT ANALYSIS

VARIABLE NAME DESCRIPTION EXPLANATION CREATED SOLVER NODE

gap_old Previous gap Physical gap Contact: Penalty, Internal

distance variable from the dynamic or state
last converged Augmented variable
solution, Lagrangian,
dynamic used
cm_old Previous mapped The location on Friction node Internal
source the source where present state
coordinates this point was variable
located at last
converged
solution
sliptot_old Previous Total slip in this Friction or Slip Internal
accumulated slip point at last velocity: Store state
converged accumulated slip variable
solution selected
contact_old Contact variable Nonzero if the Friction node Internal
in previous step point was in present state
contact at last variable
converged
solution within
the friction
detection
tolerance
isContact_old Contact status Nonzero if the Contact: Add Internal
variable in point was in contact status to state
previous step contact at last solver log variable
converged selected
solution
isSliding_old Sliding friction Nonzero if the Contact: Add Internal
status variable point was in a contact status to state
previous step sliding state at last solver log variable
converged selected. Friction
solution node present.
isSticking_old Sticking friction Nonzero if the Contact: Add Internal
status variable point was in a contact status to state
previous step sticking state at solver log variable
last converged selected. Friction
solution node present.

CONTACT MODELING | 241

 TABLE 2-11: DEPENDENT VARIABLES OCCURRING IN CONTACT ANALYSIS

VARIABLE NAME DESCRIPTION EXPLANATION CREATED SOLVER NODE

Wfric Frictional Energy dissipated Friction present Segregated

dissipation density by friction and Compute step
frictional
dissipation
selected.
Wfric_old Previous frictional Energy dissipated Slip velocity Internal
dissipation density by friction at the present and state
last converged Compute variable
solution frictional
dissipation
selected.
activation_old Adhesion Nonzero if the Adhesion node Internal
condition at adhesion present state
previous step condition was variable
fulfilled at the last
converged
solution
cma_old Previous adhesive Position on the Adhesion node Internal
mapped source source where this present state
coordinates point was located variable
when adhesion
was detected.
um_max_old Maximum Maximum norm Decohesion node Internal
displacement of displacement present and state
jump at previous jump vector Displacement- variable
step reached until based damage
previous step. selected
Ydm_max_old Maximum damage Maximum value of Decohesion node Internal
energy at the stored present and state
previous step undamaged elastic Energy-based variable
energy at damage selected
previous step

242 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

TABLE 2-11: DEPENDENT VARIABLES OCCURRING IN CONTACT ANALYSIS

VARIABLE NAME DESCRIPTION EXPLANATION CREATED SOLVER NODE

vdmg_old Viscous damage at Value of the Decohesion node Internal

previous step viscous damage present and state
variable at the Delayed damage variable
previous step selected
Wdmg_old These variables Decohesion node Internal
Wsd_old are related to the present and state
computation of Compute damage variables
dmg_old
the energy dissipation energy
dissipated by selected.
damage

Important Contact Variables

In this section you will find a summary of variables created in the contact node that
can be useful in postprocessing. In Table 2-12 important field variables are listed. For
the Shell, Layered Shell and Membrane interfaces, these are appended with the suffix
‘_top’ or ‘_bot’, depending on which side of the boundary that is in contact. All
variables are also defined per contact node, for example <phys>.<contact_tag>.Tn
is the contact pressure from the contact node with the tag <contact_tag>. These
variables can be useful if the model includes multiple contact nodes. Table 2-13
presents global variables related to contact that can be useful for postprocessing.
TABLE 2-12: IMPORTANT CONTACT VARIABLES

VARIABLE DESCRIPTION DEFINED IN COMMENTS

<phys>.Tn Contact pressure Contact Vector in global

coordinates also
available
<phys>.gap Gap distance including Contact
offsets
<phys>.Ttnorm Friction force norm Friction or Slip
Velocity
<phys>.Tt Friction force Friction or Slip Vector
Velocity
<phys>.qfric Friction dissipation Friction or Slip
rate Velocity
<phys>.fs Adhesive stress Adhesion Vector
<phys>.us Displacement jump Adhesion Vector
<phys>.adhesion Adhesion condition Adhesion

CONTACT MODELING | 243

 TABLE 2-12: IMPORTANT CONTACT VARIABLES

VARIABLE DESCRIPTION DEFINED IN COMMENTS

<phys>.bdmg Damage Decohesion

<phys>.um_max Maximum Decohesion If displacement-
displacement jump based damage
<phys>.Ydm_max Maximum damage Decohesion If energy-based
energy damage
<phys>.Wcnt Contact energy Contact
density
<phys>.Wfric Frictional dissipation Friction or Slip
density Velocity
<phys>.Wadhe Adhesive elastic Adhesion
energy
<phys>.Wdbmg Damage dissipation Decohesion
energy density

TABLE 2-13: GLOBAL CONTACT VARIABLES

VARIABLE DESCRIPTION DEFINED IN COMMENTS

<phys>.gapmin_ Minimum gap Contact

<pair_tag> distance, contact pair
<pair_tag>
<phys>.Tnmax_ Maximum contact Contact
<pair_tag> pressure, contact pair
<pair_tag>
<phys>.T_tot Total contact force Contact Vector
<phys>.Tn_tot Total contact force, Contact Vector
pressure contribution
<phys>.Tt_tot Total contact force, Friction or Slip Vector
friction contribution Velocity
<phys>.Wcnt_tot Total contact energy Contact Integration over
all contact
boundaries
<phys> Total frictional Friction or Slip Integration over
.Wfric_tot dissipation Velocity all contact
boundaries

244 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

TABLE 2-13: GLOBAL CONTACT VARIABLES

VARIABLE DESCRIPTION DEFINED IN COMMENTS

<phys> Total adhesive elastic Adhesion Integration over

.Wadhe_tot energy all contact
boundaries
<phys>.Wdmg_tot Total damage Decohesion Integration over
dissipation energy all contact
boundaries

References for Contact Modeling

1. J.L. Chaboche, F. Feyel, and Y. Monerie, “Interface debonding models: a viscous
regularization with limited rate dependency,” International Journal of Solids and
Structures, vol. 38, pp. 3127–3160, 2001.

CONTACT MODELING | 245

 Activating and Deactivating Material
In some applications, you may want to activate or deactivate material during an
analysis. An example is when you want to model the addition of material during
processes such as additive manufacturing or welding.

The information in this section is applicable if your license includes the

Structural Mechanics Module or the MEMS Module.

The Activation subnode can be used for this purpose. You enter an activation
expression to determine whether material is active or not. When this Boolean
expression is satisfied, the material is activated. Rather than truly adding or removing
material, the Activation subnode alters the stiffness and density of the material to
emulate this.

It is typically required that material is activated in a state of zero stress. Therefore,

Activation activates material in a stress-free state by removing all elastic strains present
at the point of activation.

ACTIVATION
Add an Activation subnode when you want to activate or deactivate one or several
domains selected in a Linear Elastic Material node. The Activation expression field is used
to define when material should be activated, and the Activation scale factor is used to
reduce the elastic stiffness and density of the material which is not active.

In Figure 2-17 a case is shown where the material in domains 1, 2, and 3 is to be

activated when an auxiliary sweep parameter para exceeds the value 1.5. The
activation scale factor has a default value of 10-5.

The activation condition can be any type of expression or function. A common case is
that it is a function of the temperature. The activation expression is evaluated in each
Gauss point. This means that an element can be partially activated. If you want to force
whole elements to be activated, you can for example put the activation expression
inside the centroid() operator.

246 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

Figure 2-17: Settings for the Activation subnode.

VARIABLES
Two useful variables are created when you add the Activation subnode. The variable
isactive is set to ‘1’ when the activation condition is satisfied, and it is ‘0’ otherwise.
The variable wasactive is used to record if the material has been active at any previous
step in the analysis. This variable can be used to “lock” the state of activation, once it
has been reached. Suppose that you want the material in the previous example to
remain active even if para later becomes less than 1.5. The activation expression for an
interface with the name solid could then be expressed as:

(para > 1.5) || solid.wasactive

Similarly, the variable wasinactive is used to record if the material has been inactive
at any previous step in the analysis.

ACTIVATING AND DEACTIVATING MATERIAL | 247

 Domains that are not selected in any Activation subnode always have the isactive
variable set to ‘1’.
TABLE 2-14: VARIABLES DEFINED BY THE ACTIVATION SUBNODE

VARIABLE DESCRIPTION

<physics>.isactive Current state of the material (1 or 0)

<physics>.wasactive Variable set to 1 if isactive has been 1 previously
<physics>.wasinactive Variable set to 1 if isactive has been 0 previously

RESULTS
If you have performed an analysis in which only part of the material is active, it is useful
to apply a Filter and only display the regions that actually are active; see Figure 2-18.
When an Activation subnode is added in a Solid Mechanics or Membrane interface,
such a Filter node is automatically added to the default stress plots.

Figure 2-18: Filter settings to only display active material.

248 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

S pr i ng s a nd Damp ers
The Spring Foundation and Thin Elastic Layer physics nodes supply elastic and
damping boundary conditions for domains, boundaries, edges, and points.

The features are completely analogous, with the difference that a Spring Foundation
connects the structural part on which it is acting to a fixed “ground,” while the Thin
Elastic Layer acts between two parts, either on an interior boundary or on a pair.

A Spring Foundation is most commonly used for simulating boundary conditions with
a certain flexibility, such as the soil surrounding a construction. Another important use
is for stabilizing parts that would otherwise have a rigid-body singularity. This is a
common problem in contact modeling before an assembly has actually settled. In this
case a Spring Foundation acting on the entire domain is useful because it avoids the
introduction of local forces.

A Thin Elastic Layer used as a pair condition can simulate thin layers with material
properties that differ significantly from the surrounding domains. Common
applications are gaskets and adhesives.

When a Thin Elastic Layer is applied on an interior boundary, it usually models a local
flexibility, such as a fracture zone in a geological model.

The following types of data are defined by these nodes:

• Spring Data
• Loss Factor Damping
• Viscous Damping

SPRING DATA
The elastic properties can be defined either by a spring stiffness or by a force as function
of displacement. The force as a function of displacement can be more convenient for
nonlinear springs. Each spring node has three displacement variables defined, which
can be used to describe the deformation dependency. These variables are named
<interface_name>.uspring1_<tag>, <interface_name>.uspring2_<tag>, and
<interface_name>.uspring3_<tag> for the three directions given by the local
coordinate system. In the variable names, <tag> represents the tag of the feature
defining the variable. The tag could, for example, be spf1 or tel1 for a Spring
Foundation or a Thin Elastic Layer, respectively. These variables measure the relative
extension of the spring after subtraction of any predeformation.

SPRINGS AND DAMPERS | 249

 In addition to explicitly supplying a spring stiffness, you can choose to enter elastic
material data and the layer thickness. The spring stiffness is then computed internally,
based on an assumption of plane strain conditions.

LOSS FACTOR DAMPING

The loss factor damping adds a loss factor to the spring data above, so that the total
force exerted by the spring with loss is

f sl =  1 + i f s

where fs is the elastic spring force, and  is the loss factor.

Loss factor damping is only applicable for eigenfrequency and frequency-domain

analysis. In time-dependent analysis the loss factor is ignored.

VISCOUS DAMPING
It is also possible to add viscous damping to the Spring Foundation and Thin Elastic
Layer features. The viscous damping adds a force proportional to the velocity (or in
the case of Thin Elastic Layer: the relative velocity between the two boundaries). The
viscosity constant of the feature can be made dependent on the velocity by using the
variables named <interface_name>.vdamper1_<tag>, <interface_name>
.vdamper2_<tag>, and <interface_name>.vdamper3_<tag>, which contain the
velocities in the three local directions.

Bracket — Spring Foundation Analysis: Application Library path

Structural_Mechanics_Module/Tutorials/bracket_spring

Spring Foundation and Thin Elastic Layer

250 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

Defining Multiphysics Models
The following modeling tips are about how to define multiphysics models. A good
place to start reading is in Building a COMSOL Multiphysics Model in the COMSOL
Multiphysics Reference Manual.

In this section:

• Thermal-Structure Interaction
• Acoustic-Structure Interaction
• Thermal-Electric-Structural Interaction

Thermal-Structure Interaction
The Thermal Stress, Solid Interface included with this module has a predefined
one-way approach for thermal-structure interaction (thermal stress), which combines
a Solid Mechanics interface with a Heat Transfer interface from the Heat Transfer
Module or COMSOL Multiphysics.

There are also similar multiphysics interfaces available for thin structures, as described
in The Thermal Stress, Shell Interface, The Thermal Stress, Membrane Interface, and
The Thermal Stress, Layered Shell Interface. The latter requires the Composite
Materials Module.

By default, COMSOL Multiphysics takes advantage of the one-way dependence and

solves the problem sequentially using the segregated solver. The solution for the
temperature is separated from the stress-strain analysis, which then uses the computed
temperature field from the heat transfer equation.

Using a single iteration in the segregated solver does not produce a

correct result if there are thermal properties that depend on the
displacements. Examples are when a heat source causes mechanical losses
(damping) in the material or when thermal contact is present.

DEFINING MULTIPHYSICS MODELS | 251

 Acoustic-Structure Interaction
When the Structural Mechanics Module is used together with an acoustics physics
interface from the Acoustics Module, it is possible to model a wide range of
acoustic-structure interaction problems.

There are several physics interfaces available that are documented and
described in the Acoustic-Structure Interaction Interfaces chapter in the
Acoustic Module User’s Guide

• The Acoustic-Solid Interaction, Frequency Domain Interface

• The Acoustic-Solid Interaction, Transient Interface
• The Acoustic-Piezoelectric Interaction, Frequency Domain Interface
• The Acoustic-Piezoelectric Interaction, Transient Interface
• The Solid Mechanics (Elastic Waves) Interface
• The Elastic Waves, Time Explicit Interface
• The Poroelastic Waves Interface
• The Acoustic-Solid-Poroelastic Waves Interaction Interface
• The Acoustic-Poroelastic Waves Interaction Interface
• The Acoustic-Shell Interaction, Frequency Domain Interface
• The Acoustic-Shell Interaction, Transient Interface
• The Thermoviscous Acoustic-Solid Interaction, Frequency Domain
Interface
• The Thermoviscous Acoustic-Shell Interaction, Frequency Domain
Interface

Thermal-Electric-Structural Interaction
The Joule Heating and Thermal Expansion Interface enables
thermal-electric-structural interaction. This is a combination of three physics
interfaces: Solid Mechanics, Heat Transfer in Solids, and Electric Currents.

The thermal-electric coupling is bidirectional, with Joule heating and

temperature-dependent electrical properties, while the temperature coupling to the
Solid Mechanics interface is unidirectional.

252 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

By default, COMSOL Multiphysics takes advantage of the one-way dependence and
solves the problem sequentially using the segregated solver. Temperature and electric
potential are solved using a coupled approach and then the stress-strain analysis uses
the computed temperature field from the heat transfer equation.

Using a single iteration does not produce a correct result if there are
thermal properties or electrical that depend on the displacements, making
the thermal-structure part into a bidirectional coupling.

DEFINING MULTIPHYSICS MODELS | 253

 T he r m a lly C o u p l ed Prob l em s
A wide class of structural mechanics problems are related to effects of variations in
temperature. In this section various such effects are discussed.

Temperatures can either be computed using another physics interface, usually Heat
Transfer in Solids, or directly be prescribed in the input for the various physics nodes.

For information about the predefined coupling between the Solid

Mechanics and Heat Transfer in Solids interfaces, see The Thermal Stress,
Solid Interface.

In this section:

• Temperature-Dependent Material Data

• Thermal Expansion
• Constraints and Thermal Expansion
• Thermoelastic Damping

Temperature-Dependent Material Data

Many material properties, such as Young’s modulus, coefficient of thermal expansion,
and yield stress, can have a significant dependence on temperature. In many cases,
materials supplied in the material libraries and databases have such dependencies
incorporated.

Materials in the COMSOL Multiphysics Reference Manual

If a material property under the Materials branch has a temperature dependence, you
have to input the temperature to be used in the Model Input section in the settings
window for the node in the physics interface that references the property. It is possible
that not all aspects of a material are defined in the same node in the Model Builder tree.
For example, if a problem is run with thermal expansion and plasticity, then:

• Young’s modulus, Poisson’s ratio, and mass density are given in the Linear Elastic
Material node.

254 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

• Yield stress and hardening function are given in the Plasticity node.
• Coefficient of thermal expansion is given in the Thermal Expansion (for Materials)
subnode.

The Plasticity node is available as a subnode to the Linear Elastic Material

node when you have either the Nonlinear Structural Materials Module or
the Geomechanics Module.

For each of these nodes, there is a Model Input section in the Settings window. Some
of these sections may be empty if none of the properties given in that node have a
temperature dependence. In general, you have to supply the temperature in all the
Model Inputs sections.

As a default, the value of the temperature T is obtained from a Common model input.
You can also select User defined to enter a value or expression for the reference
temperature locally. This can be done either by explicitly giving a temperature or by
selecting a temperature variable from another physics interface.

USING COMMON MODEL INPUT

When the option Common model input is selected in a physics node, you can see or
modify the value actually used by clicking the Go To Source button ( ). Doing that
will move you the node in the Model Builder that is defining the value of the property.
That location can be:

• The Default Model Inputs node under Global Definitions

• A Model Input node under Definitions->Shared Properties in the current component

If you want to create a model input value which is local to your current selection, click
the Create Model Input button . This will create a new Model Input node under
Definitions->Shared Properties in the current component, having the same selection as
in the current node.

See also Default Model Inputs and Model Input in the COMSOL
Multiphysics Reference Manual.

MASS DENSITY AND VOLUME REFERENCE TEMPERATURE

All structural mechanics interfaces are formulated on the material frame. This means
that the equations of motion are written for a certain volume in its initial configuration.

THERMALLY COUPLED PROBLEMS | 255

 The preservation of mass requires that the mass density is constant. In a structural
mechanics problem this means that the mass density must not change. If you are using
a material in which the density has a temperature dependence, you must specify a
specific temperature at which the value is evaluated. This is the volume reference
temperature. Conceptually, you can consider this as the temperature at which the
domain has the size in which it is drawn. In practice, the choice of reference
temperature is seldom an issue, unless your application requires extreme precision. The
density of a solid material has a rather slow variation with temperature, so in most cases
it is sufficient to use room temperature as reference.

If any material in the model has a temperature-dependent mass density, the Volume
reference temperature list will appear in the Model Input section of the material settings.
As a default, the value of Tref is obtained from a Common model input. You can also
select User defined to enter a value or expression for the reference temperature locally.

All effects of volume change with temperature are incorporated through the thermal
expansion effects.

See also

• Using Common Model Input.

• Default Model Inputs and Model Input in the COMSOL Multiphysics
Reference Manual.

Thermal Expansion
As the temperature changes, most materials react by a change of volume. For a
constrained structure, the stresses that evolve even with moderate temperature changes
can be considerable. The volume change can be represented a thermal strain th,
which produces stress-free deformations. For a linear elastic material, the constitutive
law is

 = C    –  th 

256 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

In the computations, the thermal expansion appears as a load, even though it formally
is a part of the constitutive relations.

You can include thermal expansion in a model either by adding a Thermal

Expansion (for Materials) subnode to the chosen material, or by using the
Thermal Expansion.

TEMPERATURE DEPENDENCE OF THE THERMAL EXPANSION

COEFFICIENT
When performing an analysis over a larger range of temperatures, you often need to
consider the temperature dependence in the values of the thermal expansion
coefficient itself.

As long as you are using materials from the COMSOL Material Library, everything is
handled internally. When you want to enter data from your own measurements or from
the literature, you do, however, need to be aware of some details in the definitions
used.

Tangent or Secant Data

Thermal expansion coefficients can appear in two forms: tangent and secant.

The tangent form states that the increment in length is

dL
-------- =  t  T dT (2-29)
L

where t is the tangential thermal expansion coefficient. This form, which is the
thermodynamic definition, is conceptually simple, because t is uniquely defined at
each temperature. It is, however, less convenient to use in practice because an
integration is required for determining the actual change in length for a finite
temperature difference.

The secant formulation, which is the default in COMSOL Multiphysics, is often used
in engineering:

L
-------- =   T T
L0

In the secant formulation, the actual values of  will however depend on the choice of
reference temperature, Tref, at which the material has the reference length L0:

THERMALLY COUPLED PROBLEMS | 257

 L
-------- =   T,T ref   T – T ref 
L0

Converting from Tangent Form to Secant Form

Equation 2-29 can be integrated, giving

ln  ------ =
L
L0   t   d (2-30)
T ref

Define

I  T,T ref  =   t   d

T ref

Thus,

L L I  T,T ref 
-------- = ------ –1 = e –1
L0 L0

giving the secant thermal expansion coefficient as

I  T,T ref 
e –1
  T,T ref  = ---------------------------------
 T – T ref 

For most materials and temperature ranges I  T,T ref  « 1 , which makes it possible to
approximate with the simpler expression

I  T,T ref 
  T,T ref  = ------------------------- (2-31)
 T – T ref 

If you have access to tangent data, you can choose between two different methods for
using them in COMSOL Multiphysics:

• In most of the physics interfaces, you can enter tangent data directly by selecting
Tangent coefficient of thermal expansion in the settings for Thermal Expansion. When
using this option, a numerical integration of Equation 2-30 will be performed each

258 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

 time the thermal strain is used. This will have a negative impact on the performance,
when compared to using a secant coefficient of thermal expansion.
• Precompute the expression in Equation 2-29 externally for the intended range of
temperatures. This can for example be done in a spreadsheet program. Enter the
computed result as a function, which is then used as any other secant temperature
dependent thermal expansion coefficient.

When using Tangent coefficient of thermal expansion, the integrate

operator is used. It is called with the two integration limits being the
reference temperature <phys>.Tref and the current temperature <phys>
.T, where <phys> is the tag of the physics interface. If you define the
expression for the coefficient of thermal expansion yourself, you must
ensure that it depends on a ‘free’ variable, and not use the same
temperature variable as you use to prescribe the current temperature
<phys>.T.

Thermal Expansion Coefficient Dependence on Reference Temperature

Letm(T) be the temperature-dependent function that represents the measured values
of the secant thermal expansion coefficient. The change in length of a sample at a given
temperature T with respect to the sample’s original length at a temperature Tm is called
dilation.

Note that by definition, the dilation at T = Tm is zero, so Tm denotes the strain-free

state of the material as far as the measured values of m(T) is concerned. Denote the
length of the sample at a temperature T as L(T) and the strain-free length as
L0 = L(Tm). The dilation can be then expressed as LTL(Tm. Using the definition
of the secant coefficient of thermal expansion, L(T) can be written as:

L  T  =  1 +  m  T   T – T m  L  T m  (2-32)

When using the measured data, it is possible that the strain-free state occurs at a
temperature Tref which differs from Tm. The dilation at any temperature T would then
be defined as LTL(Tref, where L(Tref) can be written as.

L  T ref  =  1 +  m  T ref   T ref – T m  L  T m  (2-33)

As a result of this shift in the strain-free temperature, it is necessary to redefine the

thermal expansion coefficient so that L(T) and L(Tref) can be related using
Equation 2-32 but with Tm replaced by Tref.

THERMALLY COUPLED PROBLEMS | 259

 L  T  =  1 +  r  T   T –  T ref   L   T ref   (2-34)

Here r(T) is the redefined thermal expansion coefficient, based on Tref. It can be
derived from the relations above. Using Equation 2-32 and Equation 2-34 there are
two ways of writing the current length L(T), so that

 1 +  r  T   T –  T ref   L   T ref   =  1 +  m  T   T – T m  L  T m  (2-35)

Equation 2-33 makes it is possible to eliminate L(Tref) and L(Tm) from

Equation 2-35, and after some algebra r(T) can then be written as

 m  T  –  m  T ref 
 m  T  +  T ref – T m  -----------------------------------------------
T – T ref
 r  T  = ------------------------------------------------------------------------------------------------------ (2-36)
1 +  m  T ref   T ref – T m 

Representation in COMSOL Multiphysics

Most materials listed in the material libraries and databases available with COMSOL
Multiphysics and its add-on products contain a function for the measured
temperature-dependent thermal expansion coefficient curve. You can find this from
the Materials branch, as shown in Figure 2-19. The Piecewise function named
alpha_solid_1 is the measured thermal expansion coefficient m(T).

Using Functions in Materials in the COMSOL Multiphysics Reference

Manual

The Material Contents section in Figure 2-19 shows the material property alpha,
which is the redefined thermal expansion coefficient r(T). The complete expression
for alpha is as follows:

(alpha_solid_1(T[1/K])[1/K]+(Tempref-293[K])*
if(abs(T-Tempref)>1e-3,(alpha_solid_1(T[1/K])[1/K]
-alpha_solid_1(Tempref[1/K])[1/K])/(T-Tempref),
d(alpha_solid_1(T[1/K]),T)[1/K]))/
(1+alpha_solid_1(Tempref[1/K])[1/K]*(Tempref-293[K]))

This is essentially Equation 2-36, but with a small modification to avoid problems if
T=Tref.

260 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

 m(T)

r(T)

Figure 2-19: An example in COMSOL Multiphysics showing the Materials branch and
where to find the temperature-dependent thermal expansion coefficient.

In the definition of alpha (to be more specific: <material_tag>.def.alpha) in

COMSOL Multiphysics, Tm is set as 293 K and Tref is obtained from the variable
Tempref, which typically fetches its value from the physics interface.

Using Your Own Material Data

If you use our own material data in COMSOL Multiphysics (via an interpolation or
any other function), you can still copy-paste the built-in expression for alpha into your
New Material. You just need to:

• Replace the function name alpha_solid_1(T[1/K])[1/K], with the function

name that you have assigned to the temperature-dependent measured thermal
expansion coefficient and use the correct temperature units.

Take care when describing the units. Temperature unit conversions can be
the cause of subtle errors because of the shift in zero-point value. Use
kelvin (K) as the temperature unit to the largest possible extent. As an
alternative, you can use the other absolute temperature scale, Rankine
(R). Avoid using Fahrenheit and Celsius unless you are completely
familiar with how the temperature unit conversion works.

THERMALLY COUPLED PROBLEMS | 261

 • Ensure that the value of Tm is changed from 293 K to the actual value of
temperature that was used as the strain-free temperature to compute m(T) in the
function that you created.

User-Defined Materials and Libraries in the COMSOL Multiphysics

Reference Manual

Constraints and Thermal Expansion

When a structure is undergoes thermal expansion, the fact that the motion is restricted
by constraints will in general cause stresses. There are three types of such effects:

• The global expansion is restricted by constraints at different locations, so that

internal forces are introduced throughout the structure.
• At a boundary that is constrained, local stresses can appear if the boundary is not
free to expand in the tangential direction.
• Internally, the same type of local constraint effects will be caused by rigid objects,
such as Rigid Material.

In many cases, not only the structure which actually is modeled deforms due to the
changes in temperature, but also the surroundings (which are approximated by
constraints) will deform. You can take this effect into account by adding a Thermal
Expansion subnode to the constraints. The constraints will then provide an extra
displacement based on a given temperature field. For thermal strains, which have a
simple variation in space (for example, linear temperature variations), it is possible to
completely offset the constraint stresses using this method. For more general cases, the
stresses caused by the constraint can be significantly reduced.

The thermal expansions of the constraints are independent of that of the material in
the adjacent domain, so that the surrounding structure can be made from another
material, or have a different temperature distribution.

262 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

You can, however, also inherit temperature and thermal expansion coefficient from the
domain being constrained. This is useful for the common case that the temperature
and materials are the same over the modeled structure body and its surroundings.

The spatial variation of the temperature and coefficient of thermal

expansion must be explicit functions of the material frame coordinates. It
is not possible to use a computed temperature distribution for the thermal
expansion of the constraints.

• Thermal Expansion (for Constraints)

• Thermal Expansion of Constraints in the theory section.

For an example showing how to relive the stress at constraints in a heated

structure, see Thermal Expansion in a MEMS Device: Application
Library path Structural_Mechanics_Module/Thermal-Structure_Interaction/
thermal_expansion.

Thermoelastic Damping
In most engineering problems, the coupling between temperatures and structural
problems can be considered as unidirectional. Only the thermal expansion is
considered.

The opposite effect, where changes in stress cause heat generation, may be important
in small structures vibrating at high frequencies. The Thermoelasticity interface,
available with the MEMS Module, is designed for analyzing such problems.

It is also possible to take this effect into account by adding the Thermoelastic Damping
node to the Heat Transfer in Solids interface. When you add a Thermal Expansion node
to a material in the Solid Mechanics interface, the heat source term is computed and
made available to the Heat Transfer in Solids interface.

When you add a Thermal Expansion node under the Multiphysics Couplings branch, it is
possible to select whether the thermoelastic damping effect should be taken into
account or not. The heat source contribution is then included automatically without
adding any data in the heat transfer interface.

THERMALLY COUPLED PROBLEMS | 263

 In either case, the heat source term is only present when Structural Transient Behavior
is set to Include inertial terms.

See also

• Entropy and Thermoelasticity

• Thermal Expansion (for Materials)
• The Fluid-Solid Interaction Interface

264 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

Fluid-Structure Interaction
Fluid-structure interaction (FSI) involves several phenomena where a solid structure
and a fluid interact with each other. The interaction has three possible components,
which can be more or less important:

• The pressure and viscous forces in the fluid provides a load on the boundary of the
solid. Usually, the pressure is dominant.
• The deformation of the structure changes the geometry of the fluid domain.
• The fluid sees the structure as a moving wall, which imposes a velocity at the
interface.

You can model FSI with four different structural mechanics interfaces: Solid
Mechanics, Multibody Dynamics, Shell, and Membrane. The fluid-flow can be
modeled with any domain-level physics interface from the Single-Phase Flow group and
the Two-Phase Flow groups under Multiphase Flow.

Deforming Fluid Domains

When a fluid-structure interaction multiphysics interface is added using the Model
Wizard, a Deforming Domain node can be added automatically in the Model Builder under
the Definitions node. This is the case for the following interfaces:

• Fluid-Solid Interaction
• Fluid-Shell Interaction
• Fluid-Membrane Interaction
• Fluid-Multibody Interaction
• Fluid-Multibody Interaction, Assembly

The Deforming Domain node is, however, not added for multiphysics interfaces denoted
‘Fixed Geometry’, which are intended for cases where the deformation of the fluid
domains is small everywhere.

A deforming domain represents domains and boundaries where the mesh can deform.
By default, the Deforming Domain node has an empty selection. You can then select any
fluid domain. However, this is only needed if the geometry of such a domain
experience significant changes due to the deformation or rotation of the adjacent solid
domains. Otherwise, the moving mesh computations could introduce unnecessary

FLUID-STRUCTURE INTERACTION | 265

 overhead and nonlinearity in the model (see Deformed Mesh Fundamentals).
Domains selected in the Solid Mechanics interface are not applicable. The
deformations of the solid parts are handled in the formulation of the Solid Mechanics
interface.

If no domains are selected in the Deforming Domain node, a fixed

geometry case is solved.

Under the Deforming Domain, you can also choose the Mesh smoothing type, by default
set to Hyperelastic. More information of the smoothing type can be found in
Deforming Domain chapter in the COMSOL Multiphysics Reference Manual

By default, the mesh is free at all external boundaries of the geometry, and it follows
the solid boundaries at the solid-fluid interfaces. You can also add other types of
boundary conditions for the mesh motion; for details, see Deformed Geometry and
Moving Mesh in the COMSOL Multiphysics Reference Manual.

Union or Assembly
In most cases you model FSI problems so that the geometry sequence is set up to form
a union, and the same multiphysics coupling, Fluid-Structure Interaction, is used
irrespective of the type of structural mechanics interface. This coupling will
automatically find all boundaries that are shared between the structure and the fluid.

There are, however, cases where the assembly mode must be chosen, particularly when
having mechanisms, as is common in the Multibody Dynamics interface. In that case,
the interface between the solid and the fluid is no longer formed by a common
boundary. Rather, it consists of two boundaries, located at the same place in space.
These boundaries will in general slide with respect to each other. To model this, you
use the Fluid-Structure Interaction, Pair multiphysics coupling. You must create
appropriate pairs containing the boundaries from both types of physics under
Definitions, and manually select them in the Fluid-Structure Interaction, Pair node.

UNION — FLUID-STRUCTURE INTERACTION

When using Fluid-Structure Interaction, the spatial frame also deforms with a mesh
deformation that is equal to the displacements u_solid of the solid within the solid
domains. The mesh is free to move inside the fluid domains, and it adjusts to the
motion of the solid walls. This geometric change of the fluid domain is automatically
accounted for in COMSOL Multiphysics by the ALE method.

266 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

ASSEMBLY — FLUID-STRUCTURE INTERACTION, PAIR
The Fluid-Structure Interaction, Pair coupling does not automatically transfer the mesh
deformation at the interface from structural displacements.

For the moving mesh, you must specify the deformation of the mesh manually. Add a
Prescribed Mesh Displacement node under the Moving Mesh node, in which you give the
structural displacement as the mesh displacement expression. The variables for the
displacement in the structure is provided by the multiphysics coupling.

The variable names to use have the form <tag>.u_solid, <tag>

.v_solid, and <tag>.w_solid, where <tag> is the tag of the
multiphysics coupling, for example fsip1.

Contact and FSI

Sometimes the structural deformations are so large that objects may come into contact
with each other with the fluid being squeezed in between. Modeling contact together
with FSI requires some special considerations. The mesh in the fluid domain may
deform, but the topology remains the same — a fluid domain cannot be split into two.
If you are to model a valve or a similar structure, then the two solid parts cannot come
exactly into contact.

By adding an offset in the settings for the Contact node, you can force the two sides of
the solid to experience contact at some distance before they meet in the geometrical
sense. This approach only will, however, leave a thin channel through which the fluid
can pass. The reduction in flow may be sufficient, but you can block it even further by
increasing the viscosity in the channel when the gap is closed. To do that, you can, for
example, compute the minimum gap anywhere in the contact pair, and then make the
viscosity a function of it. Another option is to compute the wall distance in the fluid
from both sides of the contact pair and use that information to modify the viscosity.
Do not increase the viscosity more than a couple of orders of magnitude, to avoid
numerical problems.

In configurations where you more or less completely cut off the whole flow, you must
pay particular attention to your boundary conditions. A prescribed flux will cause an
extreme pressure build-up upstream of the valve and thus unrealistically large forces on
the structure.

FLUID-STRUCTURE INTERACTION | 267

 One-Way FSI
For some fixed geometry FSI problems, you can consider the coupling as being
unidirectional. One physics interface affects the other, but it is not reciprocal. Typical
examples are when the fluid slightly deforms the structure, or when small structural
vibrations modify the fluid flow. In this case, it is not necessary to compute the solution
with both physics interface solved together. It is more efficient to first solve for the
governing physics interface only, and then the other one with results from the first as
input. To solve such a problem sequentially, you need to create a study configuration
manually, with one study step for each physics interface. The governing physics
depends on the coupling type. For the case of fluid loading on a structure, the
governing physics is the fluid one. For the case of velocity transmission to a fluid, the
governing physics is the solid one.

Deformed geometry FSI, or fixed geometry fully coupled FSI, should not
be solved using a one-way approach.

Below are the steps to follow to compute a one-way FSI problem sequentially:

1 In the study step settings windows, under the Physics and Variables Selection section,
clear the physics that is solved in the second step, so that only the governing physics
is selected.
2 Add a second study step to the study and in the settings windows make sure you
have the governing physics cleared.
If the selected study steps are of stationary type, you can generate the default solver
configuration, edit it if necessary, and compute the solution. The mapping of the
solution from the first to the second study step is done automatically.
In case of a transient problem, continue with the steps below:
3 In the second study step settings window, expand the section Physics and Variables
Selection. Under Initial values of variables solved for make sure the settings are
defined as in the table below:

Settings User controlled

Method Solution
Study <previous study step>
Selection First

268 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

4 Under Values of variables not solved for define the settings as in the table below:

Settings User controlled

Method Solution
Study <previous study step>
Selection All

In the case of a fluid loading to structure coupling type, the structural mechanics
problem can be treated as quasistatic. This can be handled by running the structural
analysis as a parametric sweep over a number of static load cases, where time is used as
the parameter.

For an example of one-way FSI using a quasistatic structural analysis, see

Fluid-Structure Interaction in a Network of Blood Vessels: Application
Library path Structural_Mechanics_Module/Fluid-Structure_Interaction/
blood_vessel.

FLUID-STRUCTURE INTERACTION | 269

 Component Mode Synthesis
Introduction
Component mode synthesis (CMS) is a family of procedures in which one or more
parts of a finite element model are represented by a small reduced-order model
(ROM). There are several possible algorithms through which such a reduction can be
made. In COMSOL Multiphysics, the popular Craig–Bampton method is used for
model reduction with a dual assembly technique to ensure compatibility between
components.

The reduced component is sometimes called a dynamic substructure. Essentially, it

contains small mass, damping, and stiffness matrices, tuned so that it can represent the
static stiffness, as well as a set of important eigenmodes and the inertia of the original
component. An important assumption is that the reduced structure is linear.

The reduced component accounts for the constraints applied on the component. It can
also contain, for example, information about loads applied to the component prior to
model reduction.

Since the reduced components only have a small number of degrees of freedom (often
of the order of 10-100), they are computationally more efficient than the original full
FEM components. To create a reduced component, it is necessary to both perform an
eigenfrequency analysis and to solve for a number of static load cases. These studies
are, however, computed at the component level, and are thus usually computationally
much cheaper than analyzing the full model.

For general information on reduced-order modeling and model

reduction, see also

• Reduced-Order Modeling

• Model Reduction

in the COMSOL Multiphysics Reference Manual.

COMSOL Implementation
The Component Mode Synthesis (CMS) technique is currently implemented in the
Solid Mechanics, Shell and Multibody Dynamics interfaces. You use the Reduced

270 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

Flexible Components node in either of these interfaces to control the generation and use
of such reduced components.

Reduced components are connected to each other and to nonreduced parts of the
model through Attachment features. It means that all physics features that can access
an attachment can be used to connect the reduced components. The most common
option is to use joints, but there are other alternatives, such as Spring-Damper, or
bearing foundations in the rotordynamics interfaces.

The global model can consist of either a set of reduced components, or a mix of
ordinary FE discretized domains and reduced components. The reduced components
can be used as linear parts in an otherwise nonlinear model.

A requirement to be able to create a valid reduced component is that its underlying

physics is linear with regards to material models, the strain formulation, and other
possible contributions such as springs and loads

COMSOL Multiphysics always forces a geometrically linear formulation

on selections that intersect that of a Reduced Flexible Components node.

When using a reduced component in the global model, the connection to

nonreduced parts must be such that the deformation of the reduced
component is limited to small rotations and small strains.

You can compute all type of results and visualize reduced components just as if they
are ordinary domains in your model. One important property of the reduced
component is that only the results for the reduced set of degrees of freedom are stored.
It means that for time-dependent studies with many time steps, the file sizes can be
reduced by orders of magnitude.

Applications
Some examples of applications of the CMS reduction technique include:

• In multibody dynamics models, you usually have a number of physical components,

coupled by joints. The ultimate reduction is then to use rigid domains. When it is
necessary to take the flexibility of a component into account, using a reduced
flexible component is attractive.

COMPONENT MODE SYNTHESIS | 271

 • In rotordynamics, it is common that the foundation supporting the rotor must be
part of the analysis. Such a foundation will almost invariably have linear properties.
It is then well suited to be considered as a reduced flexible component. This can
greatly reduce the number of the degrees of freedom in the rotordynamic analysis,
which is often computationally heavy.
• In many situations, only a smaller part of a model is nonlinear. It is then possible to
reduce the linear parts once and for all, and then iterate on a much smaller nonlinear
system of equations. For this to be feasible, it must however be possible to dissect
the structure in a way that can be represented by attachments.

Working With CMS Models

CREATING REDUCED COMPONENTS

A reduced component can be created in an applicable physics interface by adding a
Reduced Flexible Components node, in which you select all domains that are to be
reduced, as shown in Figure 2-20. Subsequently, one Component Definition subnode
should be added for each set of domains for which a reduced-order model (ROM) is
to be generated. By default, this step is automated and the selected domains are
grouped into disconnected components as detected from the geometry. With this
setting, a number of Component Definition subnodes are created and their selections are
set automatically. However, by setting Component definition to User defined in the
Reduced Flexible Components node, you can take manual control of the geometric
definition of the components by adding, removing, or modifying Component Definition
nodes. In each such node, select a number of domains that defines a component. This
may be necessary if domains are not physically adjacent, but connected by other means,
for example, by springs.

By default, disconnected geometries connect by the Continuity and Thin Elastic Layer
pair features are merged when automatically generating Component Definition
subnodes. This also applies to Boundary to Boundary, Edge to Boundary, and Edge to
Edge connection features in the Shell interface. The automatic handling of such
connections can be disabled by clearing Include connections and pairs in component
definition in the Reduced Flexible Components node.

272 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

Figure 2-20: A Reduced Flexible Components node with two automatically generated
Component Definition subnodes.

• If the components are geometrically disjoint, then Form Union should

be used in the geometry sequence. If not, Form Assembly is probably a
better choice, but then you may have to either create unions between
domains that are part of the same components, or edit the settings in
the Component Definition nodes manually.
• In the majority of cases, it is sufficient with one Reduced Flexible
Components node per physics interface. If you, however, do not want to
train all ROMs in one sweep, having a single Reduced Flexible
Components node per actual component can be useful. Another case is
if you decide to reduce one additional component after already having
reduced one or several other components. Then, adding an additional
Reduced Flexible Components node can be an alternative.

COMPONENT MODE SYNTHESIS | 273

 The Reduced Flexible Components node is only applicable on domains
where the material behavior is determined by a Linear Elastic Material node
or a Section Stiffness node in the Shell interface. You can, however, have
several such nodes with different properties and settings within one
reduced component.

A disconnected reduced component, as defined by a Component Definition subnode,

can only be connected to other parts of the geometry by attachment features. In fact,
to be able to create the ROM, each component must be connected to at least one
Attachment node. By using attachments, a reduced component can be connected to
any number of other parts, reduced or full, of the model assembly.

See also Attachments.

When working with CMS, the full static and dynamic behavior of a component is
represented by a number constraint modes and constrained eigenmodes that are
computed in the training phase. These are independent of each other and are used to
construct one ROM for each component. Each Attachment node connected to a
component will add a number of static load cases that describe the constraint modes:
six in 3D, and three in 2D. The number of eigenmodes to be used is controlled
manually. You can either define it for all components in the Reduced Flexible
Components node, or individually in each Component Definition subnode. To get an
accurate representation of the reduced component, always make sure to use a sufficient
number of eigenmodes to describe the dynamics of each component. During the
eigenfrequency training step, all attachments are treated as fixed constraints, hence, the
eigenmodes are always constrained.

Note that, from a computational point of view, each eigenmode is represented as a

degree of freedom when using the reduced component in a global analysis. For
computational efficiency, you should avoid using unnecessarily many eigenmodes.

To create the ROMs, a special study sequence needs to be set up for each Reduced
Flexible Components node. It should sweep over all Component Definition subnodes, and
for each component, compute the static load cases and the requested eigenmodes in
training study steps as outlined above. The results from these study steps are then used
in a Model reduction step to generate a ROM. By using the Configure CMS Study ( )
button in the Reduced Flexible Components node, the set-up of this special study

274 | CHAPTER 2: STRUCTURAL MECHANICS MODELING

sequence is automated. In the model tree, a generated ROM is represented as a node
with the label Reduced Component under Global Definitions as seen in Figure 2-21.

Figure 2-21: Nodes in the model tree that are added automatically when working with
CMS.

• When a Reduced Flexible Components node is added, a set of parameters

are also automatically added under the Global Definitions branch. These
are placed in the CMS Parameters node seen in Figure 2-21, which is
created by the first Reduced Flexible Components node in the model.
Additionally, each Reduced Flexible Components node creates an explicit
selection node with the label CMS Component (<tag>) in its model
component. The parameters and selection are used in the
corresponding CMS study and should therefore not be deleted or
modified. As a safeguard, they are regenerated if missing when the
Configure CMS Study ( ) action is executed.
• If you change the number of Component Definition nodes or add new
features to the physics interface after setting up the CMS study and
want to run it again, the safest option is to reconfigure the study by
using the Configure CMS Study ( ) action. Note that this will delete
any ROMs that have previously been generated. To save these in the
model, y