Composites modelling:
Introduction to Finite Elements
A.K. Pickett, 2013-2014
Institut fr Flugzeugbau, University of Stuttgart
IFB
Finite Elements: Overview of the method
and element types
Contents:
Popular Finite Element (FE) methods and commercial codes.
History of the FE method.
General overview of the FE method.
Analysis steps (Pre-processing (meshing) Solution (analysis) Post-
processing (results)).
Brief introduction to the main element types:
- 1D (line), 2D (planar) and 3D (solid).
- 2D types plain stress, plain strain, axi-symmetric.
- Plates and shells.
Distinction between linear and non-linear analyses.
Example applications involving different element types.
1
Composites modelling:
Introduction to Finite Elements
A.K. Pickett, 2013-2014
Institut fr Flugzeugbau, University of Stuttgart
IFB
Implicit method
Explicit method
k m f(t),x, x& , &x&
{P} = [K ]{ }
Load
Stiffness
Displacements
m&x& + cx& + kx = f(t)
m&x&n + kxn = fn (t)
advantages:
Static problems
Best for mildly non-linear
problems
(1) &x&n = m1(fn kxn )
(2) x& n+1/2 = x& n1/2 + tn&x&n
(3)
xn+1 = xn + tn+1/2x& n+1/2
advantages:
CPU efficient and robust
Very large model sizes possible
Highly non-linear materials
Large deformations
CPU efficient for contact problems
But there are plenty of other specialised methods
Finite difference, Boundary element, Finite strip.....
Meshless methods (SPH, FPM,....)
Eulearian,
Specialised methods such as control volume for liquid infusion
Etc, etc ......
IFB
Composites modelling:
Introduction to Finite Elements
A.K. Pickett, 2013-2014
Institut fr Flugzeugbau, University of Stuttgart
Typical Commercial Codes: Names, importance and applications
Implicit FE
Main codes
1. NASTRAN (large linear
structures)
2. ABAQUS (Non linear)
3. PAM-Implicit
4. MARC (Non linear)
5. ANSYS (general purpose)
6. IDEAS (meshing + FE)....
Literature
Industry
Car companies
Applications
Explicit FE
1.
2.
3.
4.
5.
6.
90%
1.
2.
3.
4.
5.
6.
DYNA3D
PAMCRASH
PAMSTAMP
PAMFORM
RADIOSS
ABAQUS explicit
1. CFD codes
2. Vehicle dynamics
3. Welding/Casting
Vibro-acoustics
4. Composites infusion
(PAM-RTM)
5. ..
10%
90%
10%
50%
50%
Stress Analysis
Stiffness
Eigenvalue (vibrations)
Flow problems; eg
Heat, Magnetism.
Fatigue
.
Others
1.
2.
3.
4.
5.
(due to importance of crash/safety/etc..)
Crash
Safety
Stamping
Biomechanics
..
IFB
A.K. Pickett, 2013-2014
Institut fr Flugzeugbau, University of Stuttgart
Composites modelling:
Introduction to Finite Elements
The inventors and example areas of application
Introduced by:
Turner, Clough, Martin and Topp (1954) for aircraft design at Boeing.
Argyris (Stuttgart 1955) and Zienkiewicz (Swansea 1964) for Civil
Engineering.
Industrial areas:
Aerospace, Automotive, Mechanical and Civil Engineering.
Nuclear.
Manufacturing.
Biomechanics.
Geomechanics.
MEMS.
Typical applications:
Stress analysis.
Structural design.
Heat and other flow (fluid) problems.
Crash and impact.
Metal stamping and forming.
IFB
Composites modelling:
Introduction to Finite Elements
A.K. Pickett, 2013-2014
Institut fr Flugzeugbau, University of Stuttgart
Another FREE (?)
useful reference
source might be the
web site of Carlos
Felippa:
http://caswww.color
ado.edu/courses.d/IF
EM.d/Home.html
(NB I am not aware
if there are copyright
issues)
Extracted from the
above wesite
IFB
Composites modelling:
Introduction to Finite Elements
A.K. Pickett, 2013-2014
Institut fr Flugzeugbau, University of Stuttgart
FE examples: Typical and advanced applications
Simple (or complex) stress analysis
More advanced topics: e.g. biomechanics
vents
Injection point
Flow problems: e.g. temperature and (as
shown here) composites resin injection
Bird strike simulation with an SPH bird
IFB
Composites modelling:
Introduction to Finite Elements
A.K. Pickett, 2013-2014
Institut fr Flugzeugbau, University of Stuttgart
Evolution of the FE method (and CPU Power): E.g. Crash simulation over
past 25 years
Ca. 300 elements (NonLinear implicit analysis)
1 CPU day
Early (1982) PAM-NL and
PAM-CRASH automotive
component studies
Gordon Moore
(Intel boss 1965):
CPU power
doubles every 18
months
First (1985) PAM-CRASH
full car simulation (VW POLO)
State-of-the-art simulation in 1997 (BMW)
Ca. 5,000 elements
(explicit analysis)
Ca. 300,000 elements
(explicit analysis)
1 CPU day
1 CPU day
Typically up models are 2-3 million elements today (2011) and increasing!!
7
IFB
A.K. Pickett, 2013-2014
Institut fr Flugzeugbau, University of Stuttgart
Composites modelling:
Introduction to Finite Elements
VW Polo The
first car crash
model 1983-85
(5000 elements)
Typical car crash model 2007
(1,000,000+ elements)
Previously models were developed for each crash case (front,
rear, side) nowadays one model is used for all cases
8
IFB
Composites modelling:
Introduction to Finite Elements
A.K. Pickett, 2013-2014
Institut fr Flugzeugbau, University of Stuttgart
CPU, memory and costs developments over 27+ years
2.3 Mflop/sec
2012
3.4 Mflop/sec
Ca. 1985
1985 (Cray 1 supercomputer)
o Ca. 15 Million dollars
o Main memory 5MB
2012 (typical smart phone)
o 33% slower
o 3,200 times more memory
o 30,000 times cheaper
Polo crash
model (1985)
Overbraiding
3-layers
IFB
Composites modelling:
Introduction to Finite Elements
A.K. Pickett, 2013-2014
Institut fr Flugzeugbau, University of Stuttgart
Procedure
1.
2.
3.
4.
5.
6.
7.
Divide structure into pieces (discretization: using bar, beam, 2D, 3D, shell elements).
Describe the behaviour of the physical quantities of each element (thickness, stiffness).
Connect the elements (assembly: done at the nodes).
Apply loading and boundary conditions (usually done at the nodes).
Solve the system of equations for the FE structure (linear, non-linear, dynamic..).
Calculate the desired quantities (nodal displacements, element strains and stresses).
Check the solution (e.g. explicit codes energy balance; implicit codes load balance,
deformations...).
Pressure loading:
Pressure loading
Point load
E.g. structure with hole
modelled as:
approximate pressure as
point loads at nodes
Point load
applied at
node
nodes
2D if axi-symmetric,
plain stress or plain
strain is assumed
3D solid if the structure
10
has finite or variable
thickness
elements
Restraint boundary
conditions
Boundary conditions: Fixation is
enforced at the nodes
IFB
A.K. Pickett, 2013-2014
Institut fr Flugzeugbau, University of Stuttgart
Composites modelling:
Introduction to Finite Elements
Some sources of error (reasons for differences between the FE and
exact (or ideal) solutions)
1. Descretisation (the way the structure is modelled/meshed):
a) Selection of the right element types:
Bar or beam?
2D (plane stress, plain strain, axi-symmetric) or is a full 3D element needed?
Membrane or shell element?
Thin or thick shell element?
b) The element can only approximate the true behaviour.
c) Selection of the right material law (elastic, elasto-plastic, viscous, failure).
d) The mesh scheme. For good resolution:
Many elements are needed in critical areas.
Less elements (coarse mesh) in unimportant areas to save CPU time/storage.
A good grading of meshes between the zones is needed.
2.
Approximation of the loading and boundary conditions.
3.
Accuracy of the computer (its precision) and the algorithms used in the FE code.
Note: Most of these errors can be limited with experience and good engineering judgement
11
IFB
A.K. Pickett, 2013-2014
Institut fr Flugzeugbau, University of Stuttgart
Composites modelling:
Introduction to Finite Elements
Solution steps in more detail (for so-called IMPLICIT FE codes)
Pre-processor: creates the mesh and user defined control data, materials, node and
element definitions, boundary conditions, loads....
Form element [k]s: Read element data and calculate each element stiffness matrix [k].
Analysis
Assemble structure stiffness matrix [K]: The individual element [k]s are assembled;
compatibility of displacements is enforced at the nodes.
Apply displacement boundary conditions modifies [K].
Compute the load vector {P} from pressure and point loads.
Compute structure displacements: From the global structure equations [K]{D}={P},
the displacements are computed using {D}= [K]-1{P}. Inverting [K] is CPU expensive.
Compute results files: E.g. element strains and stresses; output files are written.
Post-processing and validation of results
12
IFB
A.K. Pickett, 2013-2014
Institut fr Flugzeugbau, University of Stuttgart
Composites modelling:
Introduction to Finite Elements
1D elements
Bars: (also called spring and truss) for axial
loading only. They have pin ends (free) and
cannot transfer moments, torsion or shear.
Beams: Are bars that can carry torsion, shear
and bending. The end nodes are not pinned.
Use bars or beams
depending on the
required end (joint)
type.
2D elements
2D solids and shells/plates: Usually used in;
- plane strain,
- plane stress,
- and axi-symmetric loading conditions.
Thin shell
3D elements (have a thickness direction)
Thick shell
3D shells/plates: Usually assumed to be
either thin or thick (different element
formulations).
brick
3D solids
wedge
There are others!!
tetrahedra
13
IFB
A.K. Pickett, 2013-2014
Institut fr Flugzeugbau, University of Stuttgart
Composites modelling:
Introduction to Finite Elements
Example: Application of different element types
Probably 95+% of the structure surface area are shells; but
membranes, solids, bars, beams, spotwelds, contacts, joints, rigid
bodies are also to be found.
Solids for
bumper foams
and engine
Maybe:
Membranes for tires
Shells for wheels
Beams and
bars for
suspension
connections
Membrane
elements:
Airbag
inflation
14
IFB
Composites modelling:
Introduction to Finite Elements
A.K. Pickett, 2013-2014
Institut fr Flugzeugbau, University of Stuttgart
Typical solution steps: 3 stages to running an FE analysis
FE dataset
1. Pre-processing
Preparation of the mesh
(use CAD data or mesh
created from geometry
information), application
of physical properties,
BCs, loading, etc.
FE results files
2. Solution
Check validity of mesh
and data.
Perform solution(s) and
generate results.
3. Post-processing
Visualisation of results,
contour and x-y plots...
Typical Solution packages:
I-DEAS (mainly a pre-processor with limited FE analysis)
Commercial general
purpose for mostly static
problems (implicit codes).
LUSAS (limited pre-and post-processor)
Abaqus (also explicit capabilities) and Marc
Ansys, NASTRAN
15
IFB
PAM-CRASH (Implicit)
Commercial general
purpose for dynamic
problems (explicit codes).
PAM-CRASH, DYNA3D, Abaqus (explicit) and Radioss
A.K. Pickett, 2013-2014
Institut fr Flugzeugbau, University of Stuttgart
Composites modelling:
Introduction to Finite Elements
Linear static analysis
Some basic assumptions used in static linear elastic analysis:
1.
Structural deformations are assumed to be small (deformations proportional to loads).
Lo
Original (undeformed) geometry
L>Lo
L=Lo
2.
Deformed (assumes small strain theory
gives poor results for large deformations)
For correct deformations (large strain theory
and non-linear analysis is required) *
Material behaviour is assumed to be linear elastic (stress is proportional to strain).
Linear elastic
Elastic-plastic, creep *
* For these special non-linear
analysis options must be applied
3.
16
Loading is static (constant) and does not include inertial effects due to dynamic loading.
Note: Analysis of dynamic, large deformations problems with material nonlinear
behaviour is possible (e.g. car crash), but its computationally expensive.
IFB
Composites modelling:
Introduction to Finite Elements
A.K. Pickett, 2013-2014
Institut fr Flugzeugbau, University of Stuttgart
Things to consider in preparing an FE model
Type of analysis:
Static, Dynamic, Buckling, Failure, Nonlinear (which), Thermal, Type of
solution (1D, 2D, 3D...) > Is the FE code / computer suitable?
Time constraints:
CPU time and model building time often limit the model complexity. A
balance must be sought between model complexity and required accuracy.
Is an accurate one off solution or approximate parametric studies needed?
Mesh details:
Critical areas. E.g. locations of high stress gradients, or potential failure,
must have fine meshes. Can symmetry that can be exploited.?
Problem definition: Determine design loads and points of applications, possible constraints, rigid
bodies, appropriate material models, etc.,
First trial model:
A simplified model may help to guide the final analysis model (e.g. 2D
instead of 3D). Construct the mesh from rough sketches and experiment
with element types, mesh densities, loading and BCs. Is a non-linear
solution needed? Is the FE package (pre-, solution and post-) suitable?
Final models:
Now plan a detailed mesh (beware the CPU costs) and construct the FE
model. Perform and validate/assess the results with the post-processor.
Check deformations and stress distributions and compare with the trail
model. Identify areas for potential mesh refinement and repeat the analysis.
17
IFB
Composites modelling:
Introduction to Finite Elements
A.K. Pickett, 2013-2014
Institut fr Flugzeugbau, University of Stuttgart
Finite elements and convergence
Finite elements only approximate true behaviour. It can be expected (and happens) that if
the element type and model are good the solution will converge toward a true solution with
increasing number of elements. A valid element will fulfil something called the Patch test.
B
True
True
A
FE quads
FE quads
FE triangles
FE triangles
A
E.g. built in cantilever with end
load using 2D elements
50
200
50
Number of elements/nodes
200
This convergence should be used to gauge the accuracy of your solution and when you have
reached acceptable accuracy.
Repeat solutions with 2-3 different meshes (or element types) to identify the convergence.
The aim here is not to reach the true solution but to find an acceptable engineering result.
18
IFB
Composites modelling:
Introduction to Finite Elements
A.K. Pickett, 2013-2014
Institut fr Flugzeugbau, University of Stuttgart
Some basic features of simple element types and their applications
The following slides detail the main characteristics of various 1D, 2D and
3D Finite Elements and the different types of stuctural analyses for which
each can be used.
To support this a brief introduction to relationships between material
deformations and material strains is first given. This is important since
Finite Elements are formulated to have only certain deformation modes;
thus only certain materials strains (and stresses) can be represented.
19
IFB
Composites modelling:
Introduction to Finite Elements
A.K. Pickett, 2013-2014
Institut fr Flugzeugbau, University of Stuttgart
The relationship between displacements and material strains
There are many different types of Finite elements; each type can
undergo only certain assumed deformations (displacements) which
are caused by the applied nodal loadings
These displacements lead to strains (and stresses) in the element.
The important point is that the Finite Element should (must) be
able to properly represent the required deformations for the applied
loading.
dy
Average strains of this
finite area (volume) are:
dx
undeformed
dv
du
deformed
u
x
v
y =
y
v u
xy = +
x y
x =
y (v)
x (u)
20
Which depend on the two
permissible displacements
u,v (at each node) in the x,y
directions.
10
IFB
Composites modelling:
Introduction to Finite Elements
A.K. Pickett, 2013-2014
Institut fr Flugzeugbau, University of Stuttgart
Derivation: The relationship between displacements and material strains
deformed element
Strain in x (=x)
= change in length (OA-OA)
original length (OA)
y(v)
V+
Ditto for strain in y (=y)
Shear strain is the change in
angle AOB to AOB
Note if this was an element
each corner must move in x,y
for the 3 strain components
Original element
x(u)
u +
xy
u
dx u
u
x
=
dx
x
= ditto
v
u
v
1
dx
x
x
x
=
+ .... =
u
1 + u 1
dx +
dx
21
Neglect second order terms
for small deflection theory
v
y
+ ... = x
v
x
u
x
x + ... = v + u
x
2
To get unity in the denominator
IFB
Composites modelling:
Introduction to Finite Elements
A.K. Pickett, 2013-2014
Institut fr Flugzeugbau, University of Stuttgart
Element (constant cross
Truss (also called Bar or Spring elements) section) to carry axial tension
Fx
/compression only
Usually the section area and elastic properties are
constant; but must not be.
xL(u)
Pinned
nodes
The element is 1D but it is used to construct 1D, 2D
and 3D structures.
Only axial loads are transmitted (no shear or bending).
Loads and supports can only be applied at nodes.
Typical truss structure
The strain in a member is given by:
change in displacement
original length
The stress-strain relation is simply:
22
= E
Note: Each member in the truss is one
element. Any subdivisions would give a
mechanism and an indeterminate
solution leading to numerical problems
(the stiffness matrix would be singular).
Global No. dof = 3 (x,y,z).
du
dx
But the element has 1 dof
(axial) per node xL(u)
E is usually a constant.
E could be a function of
displacement requiring a
non-linear analysis.
11
IFB
A.K. Pickett, 2013-2014
Institut fr Flugzeugbau, University of Stuttgart
Composites modelling:
Introduction to Finite Elements
Beams
A beam element can carry transverse loads of shear, torsion and moments (unlike a
truss/bar element).
Elements are formulated using classical beam theory which (usually) assume:
1. The cross section is constant along the length and small compared to the length.
2. Cross section strains and stresses vary linearly across the depth on the beam.
3. Deformations and curvatures are small.
But note: Advanced beams are usually available in most FE codes which allow tapered
sections, large deformations and non-constant stress variations across the section.
23
Classical beam theory defines the relations between deformations and the resulting strain
and stress distributions within the beam section and the resulting forces; some indications
are given below:
z
Mz My
Shear e.g. for Vz
Fx
x
Vy
y
Note: Complex sections are
M x Vz
+
replaced by equivalent
Axial
Bending
Forces and moments are
simplified sections (same
for Fx
eg for My
transferred via the nodes to
areas and inertias Iy, Iz, J)
connected elements/restraints
Stress variations
The beam element has 6 dof (x, y, z, x, y, z) per node
IFB
A.K. Pickett, 2013-2014
Institut fr Flugzeugbau, University of Stuttgart
Composites modelling:
Introduction to Finite Elements
Simplification of (some) 3D structures to equivalent 2D problems
E.G. A wide constrained 3D structure
Plain strain
(Strain is zero in z direction)
It does have stress in z
E.G. A thin unconstrained 3D structure
Plain stress
(Stress is zero in z direction)
It does have strain in z
24
Note: The meshing is easier and CPU costs are much reduced compared to 3D solid elements.
Loading cannot be applied in the axial (= out of plane) direction.
12
IFB
Composites modelling:
Introduction to Finite Elements
A.K. Pickett, 2013-2014
Institut fr Flugzeugbau, University of Stuttgart
Two dimensional solids (Plane Stress/Strain)
A two dimensional solid requires the geometry and the loading to
lie in the same plane. There are two limiting cases for 2D solids:
z(w)
y(v)
x(u)
1. Plane stress: (stress in the thickness direction is zero) and occurs
for a thin solid it is free to contract. The transverse strains are
non-zero and depend on the in plane stretching.
2. Plane strain: (strain in the thickness direction is zero) and occurs
for a section from a thick solid not free to contract. The transverse
stresses are non-zero and depend via Hookes law on the in plane
strains. They occurs when transverse dimensions are large compared
to the section, or when the section is constrained in this direction.
For both the straindisplacements are:
du
dx
dv
y =
dy
du dv
+
xy =
dy dx
x =
The stress-strain relations
for plain stress are:
The stress-strain relations
for plain strain are:
x
E
y =
1
2
xy
x
1
E
1
=
(
1
+
)(
1
2
)
0
0
xy
1
1
0 0
0 x
0 y
1
2 xy
0 x
0 y
12
2 xy
Note: Plane stress components z, xz and yz are all zero.
Plane strain components z, xz and yz are zero and the
corresponding stresses are non-zero. The element has 2
dof (x, y) per node
25
IFB
Composites modelling:
Introduction to Finite Elements
A.K. Pickett, 2013-2014
Institut fr Flugzeugbau, University of Stuttgart
Axisymmetric solids: This is also a 2D simplification of a 3D structure
This is a 3D solid of revolution. If the loading is also axisymmetric (constant around the axis
of revolution) then the FE analysis must only consider a 2D section of the 3D solid. Various
pressure vessels, discs, etc., are all examples. The section may be hollow or solid, but the
dimensions cannot vary in the circumferential direction.
du
dr
u
hoop
=
strain
r
dw
=
dz
du
dw
=
+
dz
dr
Axis of revolution
r =
There are 4
strain
components;
the straindisplacement
relations are:
z
rz
The 4 stressstrain relations
are:
26
Any load must act equally
around the full circumference
r(u)
z(w)
Torsion loads are possible but
require special treatment
The axis system is r, z with rotation ;
u, w, v are the respective displacements.
NB the solution is independent of rotation
0 r
r
1
1
Note: out of
0
E
plane shears
=
z are zero
1
0
(
1
+
)(
1
2
)
rz
0
0 1 2 2 rz
0
The element has 2 dof (r,z) per node
13
Composites modelling:
Introduction to Finite Elements
A.K. Pickett, 2013-2014
Institut fr Flugzeugbau, University of Stuttgart
IFB
For arbitrary geometries with arbitrary loading. The
strain-displacement relations are:
Three dimensional solids
z
du
dx
dv
y =
dy
dw
y =
dz
x
y
x =
zx
xz
zy
yz
yx xy
y
yz
x
1
1
y
z
1
E
=
xy (1+)(12) 0 0 0
xz
0 0 0
yz
0 0 0
Solid: Hexahedra,
best performance
but less flexible
x
y
0 0 0 z
12
0 0 xy
2
0 12 2 0 xz
0 0 12 2 yz
0
0
0
0
0
0
The element has 3 dof (x,y,z) per node
27
Composites modelling:
Introduction to Finite Elements
A.K. Pickett, 2013-2014
Institut fr Flugzeugbau, University of Stuttgart
IFB
xz
The stress-strain relations for an elastic isotropic
material are:
The following are the most common
3D elements:
Tetrahedra:
Good performance
and especially
useful for complex
geometries
du dv
+
dy dx
du dw
=
+
dz dx
dv dw
=
+
dz dy
xy =
2D Plate Element
Transverse loads are carried by bending only.
Inplane loads (xy plane) are zero meaning that x=y=xy=0
Similar assumptions to beams apply regarding section
dimensions, small displacements, and strain variations are
linear through-the-thickness.
z(w)
y(v)
Loads (point or pressure) normal to the surface, or moments,
may be applied.
x(u)
As with beams classical plate theory relates deformations to
3 dof at each
node: w, dw/dx plate strains and stresses and resulting forces/moments.
and dw/dy
z(w)
deformed midsurface
(uo=vo=0)
dw
dx
dw
u = z
dx
dw
v = z
dy
w
t
y
28
dx
du
d 2w
= z 2
dx
dx
dv
d 2w
= z 2
y =
dy
dy
x =
The straindisplacement
relations are
xy =
du dv
d 2w
+
= 2 z
dy dx
dxdy
The stress components z, xz, yz
are neglected.
Element has 3 dof (w, dw/dx, dw/dy) per node
14
A.K. Pickett, 2013-2014
Institut fr Flugzeugbau, University of Stuttgart
IFB
Composites modelling:
Introduction to Finite Elements
3D Shell Elements
Shells are similar to plates but can form a curved 3D geometry, rather than just a flat 2D
one; see below. Examples are car bodies, aircraft structures, etc. It is much more versatile
than a plate element.
Both out-of-plane and in-plane loads are allowed leading to a coupled interaction of the two.
There are many shell theories available, for example Mindlin shell theory which is often
used in FE codes. In this case membrane deformations are coupled with bending
deformations.
Shells generally have 6 strain components x, y, z, xy, xz and yz; the corresponding
stresses are x, y, z (=0 because it is thin = plane stress), xy, xz and yz.
There are also thick
shell elements,
which are somewhere between thin
shells and 3D solid
elements.
y
z
x
The element has 6 dof (x, y, z, x, y and z) per node
29
IFB
A.K. Pickett, 2013-2014
Institut fr Flugzeugbau, University of Stuttgart
Composites modelling:
Introduction to Finite Elements
Typical elements type in FE codes
Poor
1D
Better
Less
good
Good
Better
30
Extracts from the LUSAS
elements manual
15
IFB
A.K. Pickett, 2013-2014
Institut fr Flugzeugbau, University of Stuttgart
Typical elements in an FE code
Composites modelling:
Introduction to Finite Elements
Overview of the 100+ LUSAS
elements available
Generally:
B are Bars or Beams
T are Triangular (2D elements)
Q are Quadrilateral
T are Tetrahedral (3D elements)
H are Hexahedral, etc.
The number indicates the number of
nodes. NB higher order elements have
corner, mid-side and additional nodes
to improve their accuracy.
The other letters usually identify other
specific characteristics, e.g. material
types, formulations, etc.
The LUSAS manual gives specific
details for each element, including
formulation, recommendations, etc.
Extracts from the LUSAS
elements manual
31
IFB
A.K. Pickett, 2013-2014
Institut fr Flugzeugbau, University of Stuttgart
Example: Complex FE
structures (e.g. a Human)
mix element types as
needed to approximate the
real thing
Composites modelling:
Introduction to Finite Elements
The head is a layer of stiff shells for
the geometry overlaid with deformable
solids.
Location of
sensor to
measure
deceleration
The neck uses a mixture
of bars, joints, elastic
solids and shells.
View of the complete
(side impact) Human
model: Everything is
modelled as a 3D solid.
The ribs are elastic shells
covered with deformable
(visco-elastic) solids for the
flesh. Bars with dampers join
the ribs to a stiff shell spine
box.
32
16
IFB
Composites modelling:
Introduction to Finite Elements
A.K. Pickett, 2013-2014
Institut fr Flugzeugbau, University of Stuttgart
Exercise: Consider which elements are appropriate for modelling the
different parts
Elements types could include:
Bars (1D)
Beams
2D
Plates
Shells
Membranes
3D
Joints
Contact
Chaindrive
Bonnet
impact
33
IFB
A.K. Pickett, 2013-2014
Institut fr Flugzeugbau, University of Stuttgart
Two extremes for modelling:
Coarse shells/beams versus fine
solids
Coarse modelling is
suitable for crash/impact
and general loadings, or
to see the kinematics of
the dummy.
Composites modelling:
Introduction to Finite Elements
But to properly represent local stresses
and failure would need fine solid
models to capture the complex stress
and physics that occur
34
17
IFB
Composites modelling:
Introduction to Finite Elements
A.K. Pickett, 2013-2014
Institut fr Flugzeugbau, University of Stuttgart
Modelling parts with solid elements: E.g. for Fatigue analysis
Fatigue starts at a microscopic scale as a minute crack (or defect) and develops under
the action of cyclic stresses over many cycles. The stresses are much lower than the
materials failure stress.
Areas of stress concentration are especially susceptible to fatigue crack growth
Process depicted
in a book
Initiation site
Fatigue crack
growth striations
Actual
process on
my bike!
Fast fracture
area
cracks
35
IFB
Composites modelling:
Introduction to Finite Elements
A.K. Pickett, 2013-2014
Institut fr Flugzeugbau, University of Stuttgart
Stress
Fatigue analysis:
1 stress cycle
1. The part must be modelled for accurate
stress distributions (i.e. with 3D Solid
elements).
Range
of stress
2. A typical max load in one cycle is applied.
3. From which the max stress in the part is
obtained.
4. From this stress and the material S-N
experimental curve the number of cycles
(life) of the part may be estimated.
Max
stress
Mean
stress Sm
Min
stress
Fatigue S-N Curves
Stress versus No cycles to failure
curves (= Endurance) are
plotted in different ways (eg.
Linear-linear or linear-log)
Endurance curves(statistical
average from several tests)
Example S-N (linear-log) curves
for: Aluminium Alloy 24S-T3 (x)
and Mild Steel ()
Low cycle fatigue
High cycle fatigue
36
18
IFB
Composites modelling:
Introduction to Finite Elements
A.K. Pickett, 2013-2014
Institut fr Flugzeugbau, University of Stuttgart
Modelling parts with solid elements: E.g. for Failure analysis
For failure a detailed 3D solid element mesh is needed; plus the right materials law: A
typical model is the Gurson model for predicting voids nucleation, growth and coalescence
In this case the load is increases until failure (coalescence) is predicted.
Shell elements have zero stress in the thickness direction and cannot be used for accurate
failure prediction.
1
4
3
From the book Fracture
mechanics, by T.L. Anderson
37
IFB
Composites modelling:
Introduction to Finite Elements
A.K. Pickett, 2013-2014
Institut fr Flugzeugbau, University of Stuttgart
Failure prediction for porous ductile metals: The Gurson model
The Gurson model treats void growth and nucleation
as an elastic-plastic with strain softening process
evp =
f()2
2
Inclusions in a ductile matrix
Void nucleation
Effective stress
+ 2q1f* cosh
3 m
q
1 + q3f*2 = 0
2 2 y
No damage elastoplastic law
Damaged e.p. law due
to voiding
Void growth
Necking between voids
38
Strain localisation between voids
Voids coalescence and fracture
Rate of void
growth
Plastic strain
Plastic strain
Main deficiency: void growth depends only on
hydrostatic stresses; there is no void generation
under shear (of shear failure).
19
IFB
Composites modelling:
Introduction to Finite Elements
A.K. Pickett, 2013-2014
Institut fr Flugzeugbau, University of Stuttgart
Example of failure modelling:
Aluminium welding
Some recent work:
The impact project with AUDI
Two Cranfield MSc thesis projects
Tensile
strength
Strength
Ductility
Ductility
Weld
Original
workhardened
structure
Grain growth
Recrystallization
39
IFB
Composites modelling:
Introduction to Finite Elements
A.K. Pickett, 2013-2014
Institut fr Flugzeugbau, University of Stuttgart
Failure sequence for a typical Butt joint (8 mm notched)
a)
Development of void
growth in triaxial
stress state
b)
Shear bands
formation
c)
Final fracture
4000
3500
Calibrated
numerical
solution
Specimen 1
Specimen 2
3000
Specimen 3
Simulation 4
Force (N)
2500
Comparison of the
simulation with the
experimental results
2000
1500
1000
500
0
0
40
0,5
1,5
2,5
Displacement (mm)
20
IFB
Composites modelling:
Introduction to Finite Elements
A.K. Pickett, 2013-2014
Institut fr Flugzeugbau, University of Stuttgart
MIG-welded S-joint validation study
5.0 m/s
initial
velocity
3500
3000
Force (N)
2500
2000
1500
1000
TEST
500
Contours of
Damage (DcRc
model)
0
0
10
15
20
25
30
35
40
45
Displacement (mm)
S-type T-joint experiment (static)
Simulation
41
21
