Presented
by
Dr Ceri
Computational Methods -
Morris Week 6
What is CFD?
2020
What is a computational model?
Mathematical representation of a real life system, that is able to
predict, mimic, or imitate the behaviour of the system within a
set of prescribed boundary conditions
Types of Computational Models
Structural Mechanics - FEA Fluid-Structure Interaction - FSI Fluid Dynamics - CFD
Bridge Tidal Stream Turbine Cyclist
Multiphysics
What is computational fluid dynamics?
• Computational fluid dynamics (CFD) is the science of predicting fluid
flow, heat transfer, mass transfer, chemical reactions, and related
phenomena by solving the mathematical equations which govern
these processes using a numerical process
• The result of CFD analyses is relevant engineering data used in:
• Studies of new designs
• Detailed product development
• Troubleshooting
• Redesign
• CFD complements testing and experimentation
Advantages of CFD
• Relatively low cost
• Relatively inexpensive compared to experimental testing
• Time
• Can be set up in a short period of time compared to
experimental testing
• Ability to simulate real conditions
• Many flow and heat transfer processes can not be (easily)
tested, e.g. hypersonic flow, large scales,
• Ability to simulate ideal conditions
• CFD allows great control over the physical process
• Comprehensive information
• Provides a lot of useful information about the flow
Limitations of CFD
• Physical models
• Can only be as accurate as the physical models on which they
are based
• Numerical errors
• Round-off error: finite word size available on the computer
• Truncation error: approximations in numerical models
Truncation errors decrease as grid is refined
• Boundary conditions
• Accuracy is only as good as the initial/boundary conditions
• Time
• Complex models can take a long time to run
• Validation
• Always up to the user to validate results, hence accuracy is
dependent on expertise
Presented
by
Dr Ceri
Computational Methods -
Morris Week 6
Creating the Geometry
2020
Creating Geometry
• Internal Flow
• e.g. flow in a pipe or duct
• Open Channel Flow
• e.g. flow in a river or irrigation
channel
• External Flow
• e.g. flow around a vehicle or
building
Creating Geometry
• Only need the fluid not any solid parts
Internal flow
• don’t need pipe/duct
thickness
Open channel flow
• need geometry for two
fluid phases
External flow
• need to subtract object of
interest from domain
Geometry Considerations
• Are there any problems with the CAD drawing?
• Gaps between surfaces or edges
• Intersecting surfaces/edges
which should be joined
• Excessive use of points, lines or
surfaces
• Inconsistent orientation
• Surface bodies which should be
volumes
• Wrong shape
Geometry Considerations
• Can the geometry be simplified?
• Small unnecessary details
• 3D/2D
• Symmetry
• Very sharp edges rounded
Geometry Considerations
• Geometry must represent fluid region
• For external flows, the geometry of the object is removed
from the flow domain
• Domain must be large enough to prevent blockage/flow
acceleration
Presented
by
Dr Ceri
Computational Methods -
Morris Week 6
Meshing the Geometry
2020
Meshing
• What is a mesh?
• Control volumes (cells or elements) the domain is split into to
calculate the solution
• One value of each variable for each cell
• Has cells grouped into boundary zones where boundary
conditions are applied
• The mesh has a significant impact on:
• Rate of convergence (or even lack of convergence)
• Solution accuracy
• CPU time required
Terminology
• Cell = control volume into which
domain is broken up
• Node = grid point
• Cell center = center of a cell
• Edge = boundary of a face
• Face = boundary of a cell
• Zone = grouping of nodes, faces,
and cells:
• Wall boundary zone.
• Fluid cell zone
• Domain = group of node, face
and cell zones
Cell/Element shapes
Choice depends on the problem and the CFD solver capabilities
Structured mesh
• Follows a structured i,j (2D) or i,j,k
convention (3D)
• Normally quadrilateral elements (2D) or
hexahedral elements (3D)
• A structured mesh makes writing a CFD code
simpler and can be efficient with memory
• Difficult to conform it to complex shapes
Unstructured mesh
• Elements are not aligned in a regular
fashion, but irregularly across the
domain.
• Indices i,j,k are replaced by node Unstructured Triangular 2D mesh
numbers.
• Unstructured meshes can model complex
and irregular domains often encountered
when solving real world problems
3D unstructured Tetrahedral mesh
Unstructured mesh
• For unstructured grids:
• The cells are arranged in an arbitrary fashion
• No constraints on cell layout
• There is some memory and CPU overhead
Ansys Workbench can produce unstructured meshes
Mesh Density
• Mesh density should be high enough to capture all relevant flow
features, e.g. near boundaries, separated flow regions, wakes
• To ensure the CFD results are mesh independent, the mesh density
should be increased until the changes in the results are negligible
Mesh Density
• If mesh is not fine enough some
important details could be missed
• Mesh does not need to be the
same density throughout the
domain
Mesh Quality
• For the same cell count, hexahedral meshes will
give more accurate solutions
• In boundary layers: quad, hex, and prism/wedge
cells are preferred over tri’s, tets, or pyramids.
• Some measures of quality:
o Skewness
o Smoothness (change in size)
o Aspect ratio
Skewness
• Based on equilateral volume:
• Above definition applies to tris and tets only
• Other definitions of skewness exist
Range of skewness
Smoothness and aspect ratio
• Change in size should be gradual (smooth)
• Aspect ratio = ratio of the largest edge length to the shortest
edge. Ideal = 1 for an equilateral triangle/quad
Ideally, the maximum change in
grid spacing should be <20%
Mesh Quality Analysis in Ansys
• Shows number of cells against a quality measure (mesh metrics)
• E.g. Skewness, orthogonal quality…
• Ideal value is either 0 or 1
• Clicking on each bar displays cells with selected metric value in
Graphics window
Boundary layer mesh
• For flows over walls/solid surfaces, there is always a thin
boundary layer that develops over the body surface as a result
of fluid flow.
• The velocity changes from its value outside the boundary layer
to zero at the wall; so very large changes in ∂u/ ∂y
• To capture the boundary layer near the wall we usually employ
a finer mesh close to boundaries to capture the large changes in
velocity gradients
y u(y)
Solid surface
Boundary layer meshes
Road car
2D wing (aerofoil)
Aircraft body
Pipe/channel
Inside the boundary layer, viscous effects are important, whereas
outside of it such effects are negligible. Cells outside boundary layer can
be larger; this is OK as no significant flow changes/gradients occurring.
Some Ansys Global Mesh Controls
• Sizing
o Max / Min size
- Controls minimum and maximum cell size
o Max face size
- Controls maximum size of cell faces on surfaces
o Growth Rate
- Controls increase in edge length from one cell to the next
• Advanced size Function
o Curvature
- Alters cell density based on curvature of edges/faces
o Proximity
- Alters cell density based on distance
- between two edges/faces
• Defeaturing - Removes details smaller than a set tolerance
Some Ansys Local Mesh Controls
• Body sizing
• Face sizing
• Edge sizing
o Element size / Number of divisions
o Bias
Inflation
• If global settings used, all boundaries are inflated
• If local settings, only specified boundary
• Layer of structured cells (usually in a boundary layer)
Some meshing tips
• How many mesh points to use? As few as possible that do the job. There
should be sufficient number of mesh points to show all the flow features.
• Although improvements have been made in the automatic production of
meshes, the human still guides the process and judges the result.
• Good news! There are grid adaptation algorithms. You provide a starter mesh
and then as the run proceeds the CFD package alters the mesh by adding (or
removing) and moving cells to obtain a better result.
• Bad news – grid adaptation algorithms can be difficult to use
• Try to show that your results are independent of mesh size (grid
independence).
General points
• More cells can give higher accuracy, but increased memory/CPU time. Is it worth
it?
• Only cluster high number of cells where needed, e.g. near boundaries, in wakes..…
• Some major sources of errors with meshes:
• Mesh too coarse
• High skewness
• Large jumps in volumes between adjacent cells
• Large aspect ratios
• Inadequate boundary layer mesh
• When a mesh has been created, the boundaries need to be identified so boundary
conditions can later be specified. In Ansys meshing this is done by creating Named
Selections.
