Title: Techniques for Highorder Adaptive Discontinuous Galerkin Discretizations in Fluid Dynamics Dissertat
1Techniques for High-order Adaptive Discontinuous
Galerkin Discretizations in Fluid
DynamicsDissertation Defense
- Li Wang
- PhD Candidate
- Department of Mechanical Engineering
- University of Wyoming
- Laramie, WY
- April 21, 2009
2Outline
- Introduction
- Objective
- Steady Flow Problems
- High-order Steady-State Discontinuous Galerkin
Discretizations - Output-Based Spatial Error Estimation and Mesh
Adaptation - Unsteady Flow Problems
- High-order Implicit Temporal Discretizations
- Output-Based Temporal Error Estimation and
Time-step Adaptation - Conclusions and Future Work
3Introduction
- Computational Fluid Dynamics (CFD)
- Computational methods vs. Experimental methods
- Indispensible technology
- Inaccuracies and uncertainties
- Improvement of numerical algorithms
- High-order accurate methods
- Sensitivity analysis techniques
- Adaptive mesh refinement (AMR)
L. Wang, transonic flow over a NACA0012 airfoil
with sub-grid shock resolution (2008)
M. Nemec, et. cl., Mach number contours around
LAV (2008)
D. Mavriplis, DLR-F6 Wing-Body Configuration
(2006)
4Introduction
- Why Discontinuous Galerkin (DG) Methods?
- Finite difference methods
- Simple geometries
- Finite volume methods
- Lower-order accurate discretizations
- DG methods
- Solution Expansion
- Asymptotic accuracy properties
- Compact element-based stencils
- Efficient performance in a parallel environment
- Easy implementation of h-p adaptivity
5Introduction
- High-order Time-integration Schemes
- Explicit schemes (e.g. Explicit Runge-Kutta
scheme) - Easy to solve
- Restricted time-step sizes
- Run a lot of time steps
- Implicit schemes
- No restriction by CFL stability limit
- Accuracy requirement
- Accuracy
- Computational cost
- Efficient Solution Strategies
- Required for steady-state or time-implicit
solvers - p- or hp- nonlinear multigrid approach
- Element Jacobi smoothers
6Introduction
- Sensitivity Analysis Techniques
- Applications
- Shape optimization
- Output-based error estimation
- Adaptive mesh refinement
- Adjoint Methods
- Linearization of the analysis problem
Transpose - Discrete adjoint method
- Reproduce exact sensitivities to the discrete
system - Deliver Linear systems
- Simulation output L(u), such as lift or drag
- Error in simulation output e(L) (Adjoint
solution) (Residual of the Analysis Problem)
7Objective
- Development of Efficient Solution Strategies for
Steady or Unsteady Flows - Development of Output-based Spatial Error
Estimation and Mesh Adaptation - Investigation of Time-Implicit Schemes
- Investigation of Output-based Temporal Error
Estimation and Time-Step Adaptation
8Model Problem
- Two-dimensional Compressible Euler Equations
- Conservative Formulation
9Outline
- Introduction
- Objective
- Steady Flow Problems
- High-order Steady-State Discontinuous Galerkin
Discretizations - Output-Based Spatial Error Estimation and Mesh
Adaptation - Unsteady Flow Problems
- High-order Implicit Temporal Discretizations
- Output-Based Temporal Error Estimation and
Time-step Adaptation - Conclusions and Future Work
10Discontinuous Galerkin Discretizations
- Triangulation Partition
- DG weak statement on each element, k
- Integrating by parts
- Solution Expansion
- Steady-state system of equations
11Compressible Channel Flow over a Gaussian Bump
- Free stream Mach number 0.35
- HLLC Riemann flux approximation
- Mesh size 1248 elements
Pressure contours using p0 discretization and
p0 boundary elements
Pressure contours using p4 discretization and
p4 boundary elements
12Compressible Channel Flow over a Gaussian Bump
- Spatial Accuracy and Efficiency for Various
Discretization Orders
Error convergence vs. Grid spacing
Error convergence vs. Computational time
13Compressible Channel Flow over a Gaussian Bump
- Element Jacobi Smoothers
- Single level method
- p-independent
- h-dependent
14Compressible Channel Flow over a Gaussian Bump
- p- or hp-multigrid approach
- p-independent
- h-independent
15Outline
- Introduction
- Objective
- Steady Flow Problems
- High-order Steady-State Discontinuous Galerkin
Discretizations - Output-Based Spatial Error Estimation and Mesh
Adaptation - Unsteady Flow Problems
- High-order Implicit Temporal Discretizations
- Output-Based Temporal Error Estimation and
Time-step Adaptation - Conclusions and Future Work
16Output-based Spatial Error Estimation
- Some key functional outputs in flow simulations
- Lift, Drag, Integrated surface temperature, etc.
- Surface integrals of the flow-field variables
- Single objective functional, L
- Coarse affordable mesh, H
- Coarse level flow solution,
- Coarse level functional,
- Fine (Globally refined) mesh, h
- Fine level flow solution,
- Fine level functional,
17Output-based Spatial Error Estimation
18Output-based Spatial Error Estimation
- Discrete adjoint problem (H)
- Transpose of Jacobian matrix
- Delivers similar convergence rate as the flow
solver - Reconstruction of coarse level adjoint
- Estimates functional error
- Indicates error
distribution and drives mesh adaptation
- Approximated fine level functional
19Refinement Criteria
- is used to drive mesh adaptation
- Element-wise error indicator
- Flag elements required for refinement
20Mesh Refinement
- h-refinement
- Local mesh subdivision
- p-enrichment
- Local variation of discretization orders
- hp-refinement
- Local implementation of the h- or p-refinement
individually
21Additional Criteria for hp-refinement
- For each flagged element
- How to make a decision between h- and
p-refinement?
- Local smoothness indicator
- Element-based Resolution indicator Persson,
Peraire - Inter-element Jump indicator
Krivodonova,Xin,Chevaugeon,Flaherty,
22Subsonic Flow over a Four-Element Airfoil
- Free-stream Mach number 0.2
- Various adaptation algorithms
- h-refinement
- p-enrichment
- Objective functional drag or lift (angle of
attack 0 degree) - Starting interpolation order of p 1
- HLLC Riemann solver
- hp-Multigrid accelerator
Initial mesh (1508 elements)
23Subsonic Flow over a Four-Element Airfoil
Mach number contours
Flow and adjoint problems target functional of
lift
Adjoint solution, ?(2)
Comparisons on hp-Multigrid convergence for the
flow and adjoint solutions
24h-Refinement for Target Functional of Lift
- Fixed discretization order of p 1
Final h-adapted mesh (8387 elements)
Close-up view of the final h-adapted mesh
25h-Refinement for Target Functional of Lift
- Comparison between h-refinement and uniform mesh
refinement
Error convergence history vs. degrees of freedom
Error convergence history vs. CPU time (sec)
26p-Enrichment for Target Functional of Drag
- Fixed underlying grids (1508 elements)
Final p-adapted mesh discretization orders p14
Spatial error distribution for the objective
functional of drag
27p-Enrichment for Target Functional of Drag
- Comparison between p-enrichment and uniform order
refinement
Error convergence history vs. CPU time (sec)
Error convergence history vs. degrees of freedom
28Hypersonic Flow over a half-circular Cylinder
- Free-stream Mach number of 6
- Objective functional surface integrated
temperature, - hp-refinement
- Starting discretization order of p0 (first-order
accurate) - hp-adapted meshes
Final hp-adapted mesh 42,234 elements.
Discretization orders p03
Initial mesh 17,072 elements
29Hypersonic Flow over a half-circular Cylinder
- Final pressure and Mach number solutions
30Hypersonic Flow over a half-circular Cylinder
- Convergence of the objective functional
31Outline
- Introduction
- Objective
- Steady Flow Problems
- High-order Steady-State Discontinuous Galerkin
Discretizations - Output-Based Spatial Error Estimation and Mesh
Adaptation - Unsteady Flow Problems
- High-order Implicit Temporal Discretizations
- Output-Based Temporal Error Estimation and
Time-step Adaptation - Conclusions and Future Work
32Implicit Time-integration Schemes
- Time-Implicit System
- First-order accurate backwards difference scheme
(BDF1) - Second-order accurate multistep backwards
difference scheme (BDF2) - Second-order Crank Nicholson scheme (CN2)
- Fourth-order implicit Runge-Kutta scheme (IRK4)
33Convection of an Isentropic Vortex
- Initial condition
- Isentropic vortex perturbation Periodic boundary
conditions - HLLC Flux approximation
- p 4 spatial discretization
- ? t 0.2
BDF1 (First-order accurate)
IRK4 (Fourth-order accurate)
34Convection of an Isentropic Vortex
- Temporal accuracy and efficiency study for
various temporal schemes
Error convergence vs. time-step sizes
Error convergence vs. Computational time
35Shedding Flow over a Triangular Wedge
- Free-stream Mach number 0.2
- Unstructured mesh with 10836 elements
- Various spatial discretizations and temporal
schemes
Unstructured computational mesh with 10836
elements
36Shedding Flow over a Triangular Wedge
- Free-stream Mach number 0.2
- Unstructured mesh with 10836 elements
- Various spatial discretizations and temporal
schemes
Density solution using p 1 discretization and
BDF2 scheme
37Shedding Flow over a Triangular Wedge
- t 100
- Various spatial discretizations and temporal
schemes
p 1 and BDF2
p 1 and IRK4
38Shedding Flow over a Triangular Wedge
- t 100
- Various spatial discretizations and temporal
schemes
p 1 and BDF2
p 3 and IRK4
39Outline
- Introduction
- Objective
- Steady Flow Problems
- High-order Steady-State Discontinuous Galerkin
Discretizations - Output-Based Spatial Error Estimation and Mesh
Adaptation - Unsteady Flow Problems
- High-order Implicit Temporal Discretizations
- Output-Based Temporal Error Estimation and
Time-step Adaptation - Conclusions and Future Work
40Output-based Temporal Error Estimation
- Same methodology can be applied in time
- Global temporal error estimation and time-step
adaptation - Implementation to BDF1 and IRK4 schemes
- Time-integrated objective functional
- Unsteady Flow solution
- Unsteady adjoint solution
- Linearization of the unsteady flow equations
- Transpose operation results in a backward
time-integration
41Output-based Temporal Error Estimation
- Two successively refined time-resolution levels
- H coarse level functional
- h fine level functional
- Approximation of fine level functional
- Localized functional error (for each time step i)
- Local time-step subdivision if
42Shedding Flow over a Triangular Wedge
- Implementation for BDF1 scheme ( p 2)
- Validation of adjoint-based error correction
- Objective function Drag at t 5
- Error prediction for two time-resolution levels
Refined time-resolution levels
Computed functional error
(Reconstructed adjoint) (Unsteady residual)
43Shedding Flow over a Triangular Wedge
- Adaptive time-step refinement approach vs.
Uniform time-step refinement approach - Objective functional
Error convergence vs. computational cost
Error convergence vs. time steps (i.e. DOF)
44Outline
- Introduction
- Objective
- Steady Flow Problems
- High-order Steady-State Discontinuous Galerkin
Discretizations - Output-Based Spatial Error Estimation and Mesh
Adaptation - Unsteady Flow Problems
- High-order Implicit Temporal Discretizations
- Output-Based Temporal Error Estimation and
Time-step Adaptation - Conclusions and Future Work
45Conclusions
- High-order DG and Implicit-Time Methods
- Optimal error convergence rates are attained for
the DG discretizations - Perform more efficiently than lower-order methods
- Both h- and p-independent convergence rates
- An attempt to balance spatial and temporal error
- Perform more efficiently than lower-order
implicit temporal schemes - h-independent convergence rates and slightly
dependent on time-step sizes
- Discrete Adjoint based Sensitivity Analysis
- Formulation of discrete adjoint sensitivity for
DG discretizations - Accurate error estimate in a simulation output
- Superior efficiency over uniform mesh or order
refinement approach - hp-adaptation shows good capturing of strong
shocks without limiters - Extension to temporal schemes
- Superior efficiency over uniform time-step
refinement approach
46Future Work
- Dynamic Mesh Motion Problems
- Discretely conservative high-order DG
- Both high-order temporal and spatial accuracy
- Unsteady shape optimization problems with mesh
motion
- Robustness of the hp-adaptive refinement strategy
- Incorporation of a shock limiter
- Investigation of smoothness indicators
- Combination of spatial and temporal error
estimation - Quantification of dominated error source
- More effective adaptation strategies
- Extension to other sets of equations
- Compressible Navier-Stokes equations (IP method)
- Three-dimensional problems