Detailed Transport and Performance Optimization for Massively Parallel Simulations of Turbulent Combustion with OpenFOAM

This work describes the implementation of two key features for enabling high performance computing (HPC) of highly resolved turbulent combustion simulations: detailed molecular transport for chemical species and efficient computation of chemical reaction rates. The transport model is based on an implementation of the thermo-chemical library Cantera [1] and is necessary to resolve the inner structure of flames. The chemical reaction rates are computed from automatically generated chemistry-model classes [2], which contain highly optimized code for a specific reaction mechanism. In combination with Sundials’ [3] ODE solver, this leads to drastic reductions in computing time. The new features are validated and applied to a turbulent flame with inhomogeneous mixing conditions on a grid with 150 million cells. The simulation is performed on Germany’s fastest supercomputer “Hazel Hen” [4] on 28,800 CPU cores, showing very good scalability. The good agreement with experimental data shows that the proposed implementations combined with the capabilities of OpenFOAM are able to accurately and efficiently simulate even challenging flame setups.