CFX Simulations for 3 AIAA Drag Prediction Workshop...ANSYS-CFX-3rdDPW.ppt Author Joseph Morrison...
Transcript of CFX Simulations for 3 AIAA Drag Prediction Workshop...ANSYS-CFX-3rdDPW.ppt Author Joseph Morrison...
-
© 2005 ANSYS, Inc. 1 ANSYS, Inc. Proprietary
CFX Simulations for 3rd AIAA DragPrediction Workshop
Robin Langtry; Florian MenterANSYS Germany, 83624 Otterfing
[email protected]@ansys.com
-
© 2005 ANSYS Germany2June, 2005
Outline
• CFX-10 solver technology• CFX-10 mesh strategy• Results 3rd AIAA DPW
-
© 2005 ANSYS Germany3June, 2005
CFX-10
• Finite volume method for mixed unstructured meshes• Fully conservative vertex based discretisation• Co-located variable arrangement (pressure based)• Rhie & Chow velocity-pressure coupling• Fully coupled equation system (mass and momentum
coupling)• Implicit formulation – 1st and 2nd order backward Euler• Algebraic multigrid solver• Scalable parallelisation• Second order time- and space discretisation• Entire Re and Mach number range
-
© 2005 ANSYS Germany4June, 2005
Turbulence Models
• Wide range of turbulence models– One-equation KE1E– Two-equation (k-ε, k-ω, SST ..)– RSM (LRR, SSG, SMC-ω,…)– LES, DES, SAS
• AIAA drag prediction based onSST model:– Reliable separation prediction– high accuracy near walls
(automatic wall treatment) –heat transfer validation
– Robustness
Automatic Wall Treatment
-
© 2005 ANSYS Germany5June, 2005
Goals
• Good results wereobtained for the 2ndDrag PredictionWorkshop using purelyHexahedral grids
• Goal of the presentwork was to investigatethe accuracy andrequired grid sizes forunstructured tet/prismgrids.
Drag Polar from the 2ndDPW using purelyHexahedral grids
-
© 2005 ANSYS Germany6June, 2005
Grid CFX-Mesh (Tet/Prism) without Fairing
-
© 2005 ANSYS Germany7June, 2005
Grid CFX-Mesh (Tet/Prism) withFairing
-
© 2005 ANSYS Germany8June, 2005
Grid CFX-Mesh (Tet/Prism)without Fairing
Coarse: 3 Million Nodes Medium: 8 Million Nodes Fine: 18 Million Nodes
-
© 2005 ANSYS Germany9June, 2005
Grid CFX-Mesh (Tet/Prism) withFairing
Coarse: 3.2 Million Nodes Medium: 8.3 Million Nodes Fine: 20.5 Million Nodes
-
© 2005 ANSYS Germany10June, 2005
Grid Hexahedral Mesh
-
© 2005 ANSYS Germany11June, 2005
Time Integration
• For small time steps (Δt~1.x10-5) unsteadyoscillations observed at wing-body separated zone(no fairing case).
• Computations carried out in unsteady mode– 3 coefficient loops– Start with a small time step (Δt~1.x10-5) and slowly ramp up to a
large time step (Δt=2.x10-4) to damp unsteadiness• Convergence reached in ~80-150 time steps• Computing times ~20-40h
– 20 million nodes (45 GB memory)– 21 Dual Core Nodes 2.4GHz Opteron HP Proliant Linux cluster.
• Note that steady state simulations are factor 3faster (no coefficient loops).
-
© 2005 ANSYS Germany12June, 2005
Convergence History
• Unsteadiness damped by large time step Δt=2x10-4 s• Good convergence in forces after 75-150 time steps
-
© 2005 ANSYS Germany13June, 2005
Grid Convergence/RichardsonExtrapolation
Coarse
3 million nodes
Medium
8 million nodes
Fine
20 million nodes
Hexahedral GridIndependent
Mesh 14 millionnodes
-
© 2005 ANSYS Germany14June, 2005
Drag Polar
-
© 2005 ANSYS Germany15June, 2005
Lift Curve
-
© 2005 ANSYS Germany16June, 2005
Lift Curve
-
© 2005 ANSYS Germany17June, 2005
Cp Distributions WB no FairingEffect of Grid Refinement
0.377 Span
-
© 2005 ANSYS Germany18June, 2005
Upper Surface Flow Vis.
Hexahedral Fine Grid
Tet/Prism Fine Grid
-
© 2005 ANSYS Germany19June, 2005
New ICEM Hybrid MeshingApproach
-
© 2005 ANSYS Germany20June, 2005
Summary
• High grid requirements for tet/prism mesh– Grid independence not achieved for 20
million nodes• Richardson extrapolation performed
Proper grid refinement achieved?• Hexahedral grid independent at approx. 12
million nodes(?)• Future goal: Hybrid approach using
hexahedral in BL, tetrahedral every whereelse (see next slide)