Computational modeling of multiscale flows - C-SWARM
Transcription
Computational modeling of multiscale flows - C-SWARM
Computational modeling of multiscale flows T. Grenga, S. Paolucci Department of Aerospace & Mechanical Engineering Center for Shock Wave-processing of Advanced Reactive Materials (C-SWARM) WCCM XI - ECCM V - ECFD VI - Barcelona 21 July 2014 Motivation “a single overarching grand challenge: the development of a validated, predictive, multi-scale, combustion modeling capability to optimize the design and operation of evolving fuels in advanced engines”, Basic Research Needs for Clean and Efficient Combustion of 21-st Century Transportation Fuels, DOE 2007 Models need to simulate the intricate coupling between the fluid mechanics and chemistry Multi-scales problems require • advanced computational approaches • high performance computing T. Grenga, S. Paolucci — University of Notre Dame (C-SWARM) — Computational modeling of multiscale flows 2 Introduction The model includes detailed chemical kinetics, multicomponent diffusion Problems are typically multidimensional and contain a wide range of spatial and temporal scales An adaptive method is applied to the simulation of compressible adaptive flows “Research needs for future internal combustion engines”, Physics Today, Nov. 2008, pp. 47-52 The method resolves the range of spatial scales present, while greatly reducing computational cost T. Grenga, S. Paolucci — University of Notre Dame (C-SWARM) — Computational modeling of multiscale flows 3 Governing Equations ∂ρ ∂t ∂ρui ∂t ∂ρE ∂t ∂ρYk ∂t ∂ =− (ρui ) ∂xi ∂ ∂τij =− (ρuj ui + pδij ) + ∂xj ∂xj ∂ ∂ =− (uj (ρE + p)) + (uj τji − qi ) ∂xj ∂xi ∂ ∂ji,k =− (ui ρYk ) − + Mk ω̇k , k = 1, . . . , K − 1 ∂xi ∂xi ρ-density, ui -velocity vector, p-pressure, E-specific total energy, Yk -mass fraction of species k, τij -viscous stress tensor, qi -heat flux, ji,k -species mass flux, Mk - molecular mass of species k, and ω̇k -production rate of species k T. Grenga, S. Paolucci — University of Notre Dame (C-SWARM) — Computational modeling of multiscale flows 4 Governing Equations � � � � ∂ul ∂ui ∂uj 2 τij = µ + δij + κ− µ ∂xj ∂xi 3 ∂xl � K � � ∂T RT hk ji,k − qi = −λ + DTk di,k ∂xi Mk X k k=1 ji,k ρYk = M Xk K � j=1 j�=k Mj Dkj di,j DTk ∂T − T ∂xi k=1 ∂Xk 1 ∂p di,k = + (Xk − Yk ) ∂xi p ∂xi � � K I K � � � �� � � � �� � f νki r νki νki − νki ki ω̇k = [Xk ] − ki [Xk ] i=1 1 E = e + ui ui 2 ρRT p= M K � Yk = 1 k=1 k=1 e-internal energy, T -temperature, R-universal gas constant, M -mixture molecular mass, µ-shear viscosity, κ-bulk viscosity, λ-thermal conductivity, hk -enthalpy of species k, Xk -mole fraction of species k, di,k -diffusion driving force, DkT thermal diffusion coefficients, Dij -ordinary multicomponent diffusion coefficients, � �� νki and νki -product and reactant stoichiometric coefficients in reaction i, kif and kir forward and reverse reaction rate constants T. Grenga, S. Paolucci — University of Notre Dame (C-SWARM) — Computational modeling of multiscale flows 5 Wavelet Adaptive Multiresolution Representation (WAMR) Method l - l0 It is a dynamically spatially 6 adaptive multi-scale method Collocation points are 4 associated with wavelets 2 Wavelet amplitudes reflect the local regularity of functions 0 1.0 High resolution grid with small 0.5 number of points y Solutions are automatically verified Method is parallelized with MPI approach 1.0 0.5 0.0 0.0 x S. Paolucci, Z. Zikoski, D. Wiraseat, WAMR: An adaptive method for the simulation of compressible reacting flow. Part I. Accuracy and efficiency of algorithm, J. Comp. Phys., 272 (2014), pp. 814-841 S. Paolucci, Z. Zikoski, T. Grenga, WAMR: An adaptive method for the simulation of compressible reacting flow. Part II. The parallel algorithm, J. Comp. Phys., 272 (2014), pp. 842-864 T. Grenga, S. Paolucci — University of Notre Dame (C-SWARM) — Computational modeling of multiscale flows 6 WAMR Method Sparse wavelet representation (SWR) u (x) = J � � u0,k Φ0,k (x) + J−1 � � dj,λ Ψj,λ (x) j=0 {λ : |dj,λ |≥ε} k �� � uJ ε �� � RεJ Density Pressure Temperature order 1/6 −2 −d/p Max error −3 10 (i) J Dx uε �V,∞ 1 −4 10 −5 10 −6 10 Error of derivative approximation �∂ u/∂x − dj,λ Ψj,λ (x), 10 Number of collocation points i 0 −1 �u − uJε �∞ ≤ C1 ε i � 10 Error in SWR: � j=0 {λ : |dj,λ |<ε} 10 Threshold parameter: ε NE ≤ C2 ε + J−1 � ≤ −6 10 −5 10 −4 10 −3 ε 10 −2 10 −1 10 − min((p−i),n)/2 CNE T. Grenga, S. Paolucci — University of Notre Dame (C-SWARM) — Computational modeling of multiscale flows 7 WAMR: 3D Shock-Bubble interaction (SBI) Problem Description • Domain [0, 3]×[0, 0.4]×[0, 0.4] cm • Ambient Mixture YN2=0.868, YO2=0.232 P=101.3 kPa T=1000 K • Bubble YH2=0.990 r=0.1 cm at x=0.5cm • Loaded by Shock Ms = 1.5 at x=0.3 cm • Wavelet parameter ε=10-2 p=4 [Nx× Ny × Nz]coarse=[60×8×8] J=6 Δxmin= 7.81 µm • Time integration Runge-Kutta-Fehlberg 45 εrel=10-3, εabs=10-10 T. Grenga, S. Paolucci — University of Notre Dame (C-SWARM) — Computational modeling of multiscale flows 8 WAMR: 3D SBI - Results 3 Density [g/cm ] Temperature [K] 1000 1100 1200 1300 1400 1500 4.0e-4 1600 5.0e-4 6.0e-4 7.0e-4 8.0e-4 0µs 2.5µs 5µs 10µs 18.75µs T. Grenga, S. Paolucci — University of Notre Dame (C-SWARM) — Computational modeling of multiscale flows 9 WAMR: 3D SBI - Results HO2 Velocity [cm/s] 20000 40000 60000 80000 4.0e-5 100000 6.0e-5 8.0e-5 1.0e-4 1.2e-4 0µs 2.5µs 5µs 10µs 18.75µs T. Grenga, S. Paolucci — University of Notre Dame (C-SWARM) — Computational modeling of multiscale flows 10 WAMR: 3D SBI - Results 10μs H2 5.00e-02 4.00e-02 3.00e-02 2.00e-02 1.00e-02 H2O2 1.60e-08 1.20e-08 8.00e-09 4.00e-09 T. Grenga, S. Paolucci — University of Notre Dame (C-SWARM) — Computational modeling of multiscale flows HO2 1.00e-05 8.00e-06 6.00e-06 4.00e-06 2.00e-06 H2O 2.80e-05 2.00e-05 1.20e-05 4.00e-06 11 WAMR: 3D SBI - Results 15.5μs H2 2.80e-02 H2 2.40e-02 2.80e-02 2.00e-02 2.40e-02 2.00e-02 1.60e-02 1.60e-02 1.20e-02 1.20e-02 8.00e-03 8.00e-03 4.00e-03 4.00e-03 HO2 1.40e-04 1.00e-04 6.00e-05 2.00e-05 H2O2 1.60e-06 1.20e-06 8.00e-07 4.00e-07 T. Grenga, S. Paolucci — University of Notre Dame (C-SWARM) — Computational modeling of multiscale flows H2O 4.40e-03 3.20e-03 2.00e-03 8.00e-04 12 WAMR: 3D SBI - Results 18.75μs H2 1.00e-02 HO2 1.20e-04 7.00e-03 1.00e-04 4.00e-03 8.00e-05 1.00e-03 6.00e-05 H2O2 8.00e-05 6.00e-05 4.00e-05 2.00e-05 T. Grenga, S. Paolucci — University of Notre Dame (C-SWARM) — Computational modeling of multiscale flows H2O 1.00e-01 7.00e-02 4.00e-02 1.00e-02 13 WAMR: 3D SBI - Results H2 1.0e-4 10µs 1.0e-3 1.0e-2 12.5µs 15.5µs 18.75µs O2 1.6e-1 1.8e-1 2.0e-1 T. Grenga, S. Paolucci — University of Notre Dame (C-SWARM) — Computational modeling of multiscale flows 2.2e-1 14 WAMR: 3D SBI - Results H 1.0e-5 10µs 1.0e-4 12.5µs 1.0e-3 18.75µs 15.5µs O 1.0e-5 1.0e-4 1.0e-3 T. Grenga, S. Paolucci — University of Notre Dame (C-SWARM) — Computational modeling of multiscale flows 1.0e-2 15 WAMR: 3D SBI - Results OH 1.0e-5 10µs 1.0e-4 1.0e-3 12.5µs 1.0e-2 18.75µs 15.5µs H2O 1.0e-4 1.0e-3 1.0e-2 T. Grenga, S. Paolucci — University of Notre Dame (C-SWARM) — Computational modeling of multiscale flows 1.0e-1 16 WAMR: 3D SBI - Results 7 −7 10 10 Grid points Time step [s] 6 10 −8 10 5 10 4 10 0 0.2 0.4 0.6 0.8 1 1.2 Time [s] 1.4 1.6 • Runtime ~2.58 × 10^5 CPU-hr 1024 maximum cores • Equivalent uniform grid 109 points (7.81 µm) −9 10 1.8 2 −5 x 10 T. Grenga, S. Paolucci — University of Notre Dame (C-SWARM) — Computational modeling of multiscale flows 17 Conclusions The WAMR method • • • • provides a robust spatial adaption strategy minimizes the number of degrees of freedom provides a verified solutions MPI parallelization with good scaling up to 1024 cores The SBI problem • has been solved with high accuracy (7.81 µm) • a wide range of spatial and temporal scales has been captured • interaction between advection, diffusion and reaction has been shown T. Grenga, S. Paolucci — University of Notre Dame (C-SWARM) — Computational modeling of multiscale flows 18 Center for Shock Wave-processing of Advanced Reactive Materials (C-SWARM) Computational modeling of multiscale flows T. Grenga, S. Paolucci Department of Aerospace & Mechanical Engineering University of Notre Dame WCCM XI - ECCM V - ECFD VI - Barcelona 21 July 2014