A Dynamic Multi-Zone Partitioning Scheme for Solving Detailed Chemical Kinetics in Reactive Flow Computations

Abstract

A dynamic multi-zone (DMZ) partitioning scheme is proposed for enhancing the computational efficiency associated with solving detailed chemistry in reactive flow calculations. The DMZ scheme employs a novel and highly flexible evolutionary data clustering algorithm to partition a group of computational cells into zones for solving the chemistry source terms at each hydrodynamic time step. The algorithm dynamically determines the optimal number of zones for each time step of the calculation based on the thermochemical conditions in the cells and user-specified dispersion thresholds. After solving the chemistry equations on zonal level, the DMZ scheme maps the updated solution back to the cells while preserving the initial parameter stratification. The DMZ scheme is implemented into a KIVA3V-CHEMKIN code and benchmarked against fully resolved (non-partitioned) calculations using direct-injection compression-ignition engine cases. The benchmarking results show that the DMZ scheme accurately reproduces the pressure and key species profiles obtained from the fully resolved calculations in simulations that range from homogeneous to highly stratified. Using a computational mesh containing 15,000 cells, the DMZ scheme achieved a 20- and 8-fold speedup for the moderately and severely stratified operating conditions, respectively. Based on the benchmark calculations, recommended values of the user-specified thresholds are proposed.

Keywords:

• Data clustering
• Detailed chemistry
• Dynamic multi-zone model
• Engine combustion simulation