Review: CFD Step by Step

CFD Analysis: Basic Steps 1. Define your modeling goals. 2. Identify the domain you will model. 3. Choose an appropriate solver. 4. Design and create the grid. 5. Set up the numerical model. 6. Compute the solution. 7. Examine the results. 8. Adapt the grid (optional). 9. Consider revisions to the model.

Define Your Modeling Goals What results are you looking for, and how will they be used? What degree of accuracy is required? How quickly do you need the results? Do you require a unique modeling capability? User-defined subroutines (written in FORTRAN) in FLUENT 4.5 User-defined functions (written in C) in FLUENT 5

Identify the Domain You Will Model How will you isolate a piece of the complete physical system? Where will the computational domain begin and end? What boundary conditions are needed? Can the problem be simplified to 2D?

Choose an Appropriate Solver (1) FLUENT 5 Incompressible or compressible flow Complex geometry and physics Applications: external aerodynamics, underhood flows, fans, pumps, cyclones, reactor vessels, furnaces, power plants, aircraft, turbomachinery, nozzles, other aerospace problems FLUENT 4.5 Incompressible (or mildly compressible) flow Less complicated geometry, complex physics Applications: risers, fluidized beds, other multiphase and VOF (free surface) problems

Choose an Appropriate Solver (2) FIDAP Incompressible (or mildly compressible) flow Complex geometry and physics Applications: biomedical flows; metal casting, solidification and extrusion; extruders; complex die flows; other materials processing flows

Choose an Appropriate Solver (3) POLYFLOW Viscous, laminar flow Complex rheology (including viscoelasticity) Applications: extrusion, coextrusion, die design, blow molding, thermoforming, film casting, fiber drawing NEKTON Applications: thin film coating flows IcePak Electronics cooling MixSim Mixing tank analysis

Design and Create the Grid Should you use a quad/hex grid, a tri/tet grid, or a hybrid grid? What degree of grid resolution is required in each region of the domain? Will you use adaption to add resolution? How many cells are required for the problem? Do you have sufficient computer memory?

Tri/Tet vs. Quad/Hex Meshes For simple geometries, quad/hex meshes can provide high-quality solutions with fewer cells than a comparable tri/tet mesh. For complex geometries, quad/hex meshes show no numerical advantage, and you can save time by using a tri/tet mesh.

Quad/Hex Mesh Example RAE 2822 airfoil grid Flow gradients predominantly normal to airfoil surface (stretched mesh required). Easy to grid with quads. Quad grid is more efficient than comparable tri grid. Stretched quadrilateral grid for the RAE 2822 airfoil

Tri/Tet Mesh Example Centrifugal blower grid Geometry is complicated. Flow field is complex. Difficult to grid with quads; easy to grid with triangles! Triangular grid for a 2D model of a centrifugal blower

Hybrid Mesh Example Valve port grid Specific regions can be meshed with different cell types. Both efficiency and accuracy are enhanced relative to a hexahedral or tetrahedral mesh alone. Tools for hybrid mesh generation are now available (Gambit and TGrid). tet mesh hex mesh Hybrid mesh for an IC engine valve port wedge mesh

Set Up the Numerical Model For a given solver, you will need to: Select appropriate physical models. Define material properties. Fluid Solid Mixture Prescribe boundary conditions at all boundary zones. Provide an initial solution. Set solver controls. Set up solution convergence monitors (residuals and surface or force monitors).

Compute the Solution: Steady-State Calculation normally requires many iterations before converged solution obtained. Solution considered converged when changes in key solution parameters (convergence monitors) become small. Possible convergence monitors include: Residuals Point values Integrated flux balances Integrated forces Converged solution is not necessarily accurate due to: Grid resolution Accuracy of numerics (discretization error) Accuracy of physical models (e.g., turbulence)

Compute the Solution: Unsteady Unsteady calculations computed by subiterating from one physical time level to next. Subiteration process should be well-converged before proceeding to next time level. Choose time step that can resolve unsteady physics of problem. Determine characteristic time scale T of unsteady physics of problem. Choose time step that is some fraction of time scale. e.g., t = T /100. Generally, adjust time step so calculation requires about 10 - 20 subiterations to converge. For problems with fast “startup” transient, may need to decrease time step during initial phase of solution.

Examine the Results Visualization can be used to answer such questions as: What is the overall flow pattern? Is there separation? Where do shocks, shear layers, etc. form? Are key flow features being resolved? Are physical models and boundary conditions appropriate? Are there local convergence problems? Numerical reporting tools can be used to calculate quantitative results: Lift and drag Average heat transfer coefficients Surface-averaged quantities

Tools to Examine the Results Graphical tools Grid, contour, and vector plots Pathline and particle trajectory plots XY plots Animations Numerical reporting tools Flux balances Surface and volume integrals and averages Forces and moments

Adapt the Grid Locally increase resolution of grid exactly where needed. You can adapt to: Gradients of flow or user-defined variables Isovalues of flow or user-defined variables All cells on a boundary All cells in a region Cell volumes or volume changes y+ in cells adjacent to walls Combinations of the above To assist adaption process, you can: Draw contours of adaption function Display cells marked for adaption Limit adaption based on cell size and number of cells

Adaption Example: Grid 2D planar shell - initial grid 2D planar shell - final grid

Adaption Example: Solution 2D planar shell - contours of pressure initial grid 2D planar shell - contours of pressure final grid

Consider Revisions to the Model Are physical models appropriate? Is flow turbulent? Is flow unsteady? Are there compressibility effects? Are there 3D effects? Are boundary conditions correct? Is the computational domain large enough? Are boundary conditions appropriate? Are boundary values reasonable? Is grid adequate? Can grid be adapted to improve results? Does solution change significantly with adaption, or is the solution grid independent? Does boundary resolution need to be improved? Would another mesh type be more suitable (quad vs. tri, hex vs. tet)?

The Analyst's Paradox Everyone believes a test except the tester. No one believes an analysis except the analyst.