Vista: A Multi-field Object Oriented CFD-package

Transcription

1 Vista: A Multi-field Object Oriented CFD-package T. Kvamsdal 1, R. Holdahl 1 and P. Böhm 2 1 SINTEF ICT, Applied Mathematics, Norway 2 inutech GmbH, Germany

2 Outline inutech & SINTEF VISTA a CFD Solver VISTA a CFD Toolbox Extrusion example: Extrud3D

3 inutech - Our Objectives R&D of of Numerical Methods Sales and and Support of of Software Consulting to Solve Engineering and Mathematical Problems Closing the Gap between Engineering and Mathematics Seminars and and Training

4 The Team Univ. of Oslo Univ. of Bayreuth Univ. of Erlangen (resellers world-wide)

5 VISTA A CFD-Solver

6 VISTA: A Multi-field Object Oriented CFD-package + Heat-Problem + ALE Vista Extrud + Free-surface flow + Turbulence Vista Marine Incompressible Navier-Stokes equation: + Fluid-structure Interaction Bloodflow-Problem 1 (based on VISTA) 1 phd-thesis Runar Holdahl

7 VISTA Marine Viscous 3D flow around a cylinder, Re=265, velocity isosurface Large-eddy simulation of flow around two cylinders in a tandem arrangement, Re=22000, velocity isosurface and pressure distribution

8 Computational Facts Block structured, 8 node hexahedral FE-elements LES-turbulence model (Smagorinsky, Van Driest) Number of time steps: 5000 Number of grid points: Total no. of degrees of freedom: Number of processors: 96 Parallel computers necessary for addressing this case!

9 Feature List Incompressible flow Stokes/Navier-Stokes 2D/3D stationary and time dependent variable density Turbulence modeling LES (Smagorinsky with van Driest damping) RANS (k-e: High Reynolds model) Reference system and boundary conditions ALE formulation Moving object (FSI) Free surface Periodic boundary conditions General boundary conditions (Dirichlet/Neumann) Numerical approximation (Mixed) finite elements (Taylor-Hood, equal order) Continuous projection method (including incremental pressure formulation) Semi-implicit BDF time discretisation Parallel Krylov solver reconditioned with two-level domain decomposition Software issues Usage Object-oriented (c++) code based on Diffpack Platform independent (running on Unix/Linux) Parallelised with MPI Preprocessor: GRIDDLER Postprocessor: GLview, VTK, Matlab (2D) NTNU/SINTEF: Free usage for education and research

10 VISTA A CFD-Toolbox

11 The VISTA Vision Navier-Stokes related problem VISTA Plug and Play related problem + = (+ (+ Diffpack) own coupled CFD solver Heat-Problem Turbulence Free-Surface Fluid-Surface- Interaction

12 The Diffpack Vision s = F t ( v, s, t ) arbitrary PDE, ODE, or or other numerical problem [ κ u] f in = Ω u = g on Ω Diffpack Kernel + = own Diffpack Toolboxes (further extensions) own solver

13 Philosophy: Plug and play Divide et Impera! Problem1 Problem2 SimCase Data transfer by: get -functions attach -functions CoupledSolver Solver1 Solver2 Solver3

14 Example 1: Free-surface flow InterfaceProblem ALEproblem FluidProblem FreeSurfaceSolver SimCase InterfaceSolver FluidSolver MeshMover

15 Vista Software - Current and Future Modules

16 Example 2: Fluid-structure interaction FSIsolver FluidSolver StructureSolver MeshMover

17 VISTA Extrud3D

18 Extruded Products Building industries Automotive industry

19 The Extrusion Press Ram Container Billet Die holder Profile Die

20 Flat Strip: Measurement points P T3 Die 15 4 T2 Sharp inlet 1 Slip T1 Profile

21 Mathematical Formulation The Extrusion Problem Governing Equations Boundary Value Problems The Extrusion Solver

22 Extrusion Problem - The Domain

23 Governing Equations Heat conduction: non-newtonian fluid: Zener-Hollomon material model

24 Extrusion Problem - Boundary Value Problem Coupling through temperature dependent viscosity Coupling through deformation energy Coupling through advection

25 Extrusion Problem - Boundary Value Problem (cont d)

26 Extrusion Problem - Weak Formulation

27 Extrusion Problem - Weak Formulation (cont d)

28 Extrusion Problem - Linearized Fluid Flow Equations

29 Extrusion Problem - Linearized Fluid Flow Equations (cont d)

30 Extrusion Problem - Computer Implementation

31 Extrusion Problem - Model Parameter

32 Flat Strip: 2D-mesh Container Detail Die Ram Billet (Aluminium) Number of elements = 504 Profile

33 Flat Strip: 3D-mesh Solve for temperature T Grids do not necessarily match Solve for velocity u pressure p temperature T

34 Bearing Channel Full stick Slip point Friction Full slip Viscoplastic

35 Comparison with Experiments Extrusion of a Flat Strip Temperature measurements Aluminum Die Both 2D and 3D analyses B 1 5 A 6 6 T T 9 3 T P 1 T B A

36 Finite Element Model Full 3D-model Close up Mesh Temperature

37 Extrusion Problem - The Solver

38 Flat Strip: Temperature t = 4.4 s t = 16.7 s

39 Flat Strip: Velocity t = 11.9 s t = 16.7 s

40 Flat Strip: Temperatures in T y T1 T4 z x TEMPERATURE IN T1 ( C) T1_v09_480_stick T1alu_v09_480_exp rampos(mm)

41 Flat Strip: Temperatures in T y T3d T2d z TEMPERATURE IN T2 ( C) x T2_v09_480_stick T2die_v09_480_exp rampos(mm)

42 Thank you for attention!

43 Feature List Incompressible flow Stokes/Navier-Stokes 2D/3D stationary and time dependent variable density Turbulence modeling LES (Smagorinsky with van Driest damping) RANS (k-e: High Reynolds model) Reference system and boundary conditions ALE formulation Moving object (FSI) Free surface Periodic boundary conditions General boundary conditions (Dirichlet/Neumann) Numerical approximation (Mixed) finite elements (Taylor-Hood, equal order) Continuos projection method (including incremental pressure formulation) Semi-implicit BDF time discretisation Parallel Krylov solver reconditioned with two-level domain ecomposition Software issues Usage Object-oriented (c++) code based on Diffpack Platform independent (running on Unix/Linux) Parallelised with MPI Preprocessor: GRIDDLER Postprocessor: GLview, VTK, Matlab (2D) NTNU/SINTEF: Free usage for education and research

44 inutech - Our R&D Experience Structural Optimization Topology Optimization Shape Optimization Sizing- and and Parameter Optimization Large Scale Optimization Multi-Disciplinary Optimization Design of of Experiments Biological Growth Free Free Material Optimization (Composite Design)

45 inutech - Our R&D Experience Differential Equations inutech develops and and markets the the Diffpack Product Line Linefor for the the Numerical Modeling and and Solution of of Differential Equations inutech offers Consulting Services around Diffpack; we we can can deliver customized turn-key solutions for for specialized simulation problems Attention: Diffpack is is a development environment, not not a program! It It allows you you to to generate a simulator!

46 The Diffpack Distribution Diffpack Development Framework Diffpack Utility Programs Third-Party Pre-/Postprocessors Diffpack Kernel (computing/numerics) Diffpack Toolboxes OpenSource Pre-/Postprocessors Books, Manpages, Reports Support phone/

47 The Challenge Speed-Up 1,000,000 over last 50 years Numerical Methods Scientific Computing Hardware Power Speed-Up 1,000,000 over last 50 years Software Development

48 The Diffpack Vision (cont d) Structural mechanics Porous media flow Water waves FEM FDM I/O Grid Field Aerodynamics Electro magnetics Observation: Methodology basis independent of applications Heat transfer Vector Matrix Ax=b Other PDE applications Incompressible flow

49 Diffpack Implementation - use weak formulation [ κ u] f in = Ω u = g on Ω solve on operator level Use weak formulation m κ N N dω + u = f N dω, i = 1,..., m in Ω u = g on Ω j i j i j= 1 Ω Ω

50 Diffpack Implementation - important member global_menu.multipleloop (S) FEM::makeSystem S::adm S::define S::scan FEM::calcElmMatVec FEM::numItgOverElm S::solveProblem S::integrands S::resultReport

51 ConvDiff:: integrands (numerical kernel) m k N N dω+ j i j= 1 Ω Ω Au α=0 v N jnidω u Weak Formulation ( ): j = = Ω f N dω, i = 1,..., m i b

52 ConvDiff :: fillessbc (apply essential boundary Conditions) Apply Boundary Conditions: u ( x, y) = 0, x, y Γ 1 essential conditions