Center For Compressible Multiphase Turbulence Overview Principal Investigator: S. Balachandar Department of Mechanical and Aerospace Engineering University of Florida
Center for Compressible Multiphase Turbulence Purpose of the Center To radically advance the field of CMT To advance predictive simulation science on current and near-future platforms with uncertainty budget as backbone To advance a co-design strategy that combines exascale emulation, exascale algorithms, exascale CS To educate students and postdocs in exascale simulation science and place them at NNSA laboratories T1-2
Demonstration Problem Integrated simulations Experimental measurements for validation T1-3
How Different Pieces Fit Exascale Emulation Exascale UQ UR Proxy-apps Energy-efficient algorithms UQ: Uncertainty Quantification UR: Uncertainty Reduction DB: Dakota Bundles CS: Concurrent Simulations UQ UR Uncertainty Budget DB CS Load balance, etc. Simulation Roadmap Computer Science T1-4
Multiscale Coupling Strategy Atomistic Quantum and MD EOS, Thermodynamic and transport properties, shock Hugoniot Macroscale > O(10 9 ) particles Macro LES of turbulence Point-particle approximation Continuum Scale Modeling and Simulations Microscale O(1) O(10 4 ) particles Fully resolved, DNS Multiphase LES closure models for interface & particulate turbulence Mesoscale O(10 5 ) O(10 8 ) particles Well resolved interface turbulence Unresolved particulate turbulence (Meso-LES) Particle-flow mass, momentum and energy coupling models T1-5
3D-Cylindrical Explosive Dispersal of Particles up to 200μs Features: 30 Millions computational cells 5 Millions computational particles r max = 0.30cm 6
2D-Cylindrical Explosive Dispersal of Particles up to 1ms Features: 2.5 Millions computational cells 1 Millions computational particles r max = 0.60cm 7
2D-Cylindrical Explosive Dispersal of Particles - IC study No Perturbation Charge Perturbed Particle Vol. Frac. Perturbed Later time, looking only at The gas phase Charge Perturbed Particle Vol. Frac. Perturbed A small perturbation in the charge has no influence on the particle dispersal 8
Shock-Particle Modeling Include the divergence term of F IU (New Model) F t = 6πμa u S u P +V ρ Du Dt t + ξ= K IU t ξ C 0 a V d 2 V xρ dt 2 t + V ξ= C0 a dξ K IU t ξ C 0 a V dρu dt C0 a dξ T1-9
Multi-Particle (FCC) / Shock Interaction 10
Multi-Particle Force Analysis Inviscid flow with a shock Mach number of 1.22 10% volume fraction (FCC arrangement) Front Plane Shoc k Middle Plane Back Plane Force histories 11
Shock Interaction with Array of Particles Spacing one d p C D,max for 5 th particle: 15% increase at M = 1.22 48% increase at M = 2.0 Spacing one d p / 2 C D,max for 5 th particle: 8% increase at M = 1.22 20% increase at M = 2.0 12
Shock/Multi-Particle Simulations FCC 20% Volume Fraction FCC 30% Volume Fraction FCC 40% Volume Fraction 13
Decision Making with Uncertainty Budget T1-14
Validation and UQ Experiments Validation Numerical Simulation Mimic Surrogate Model Measured Input Measurement Uncertainty in Input Measured Prediction Metrics Measurement Uncertainty in PM Prediction Metrics Propagated Uncertainty Physical Model Error Numerical Model Error Numerical Solution Error
time (sec) Model Error of Shock-Tube Simulation Model error = Physical model error + Numerical model error 95% confidence interval of particle curtain edge location due to model error for the nominal input Reducing uncertainty in the model error 8 x 10-4 6 4 2 0 0 0.02 0.04 0.06 0.08 0.1 Edge location (m) 100% 0% 100% 0% % of total uncertainty Downstream time (sec) Upstream Input Measurement uncertainty error in input Measurement uncertainty error PM
Cd M 1.2 1 0.8 Extrapolation with the Method of Lines Drag force for Mach number and Reynolds number Extrapolation drag in high Mach number and high Reynolds number range 1.6 True Value=0.97 Extrapolation=0.91±0.54 1.4 0.6 0 0.2 0.4 0.6 0.8 1 α 1 (Line1) 1.5 1.4 1.3 1.2 1.1 1 0.9 1.8 1.6 1.4 1.2 True Value=0.97 Extrapolation=0.94±0.20 0.8 0 0.2 0.4 0.6 0.8 1 1 α 2 (Line2) Line 1 2 2.5 3 3.5 4 4.5 5 5.5 log(re) 0.98 0.96 0.94 0.92 0.9 Line 2 Inaccessible domain Line 3 0.88 True Value=0.97 0.86 Extrapolation=0.97±0.00 0.84 0 0.2 0.4 0.6 0.8 1 α 3 (Line3)
Overview of Year 1 Achievements: Code run on all DoE platforms available Code has been tested with 16k cores Signature problem has been run (2D & 3D) Study of the physics of the signature problem has started Upcoming: Test with larger core count in preparation Mesoscale problems to start running shortly CoProcessing capability being implemented Improved models being implemented 18
Do you have any questions?