Dynamic Load Balancing of a Coupled Lagrange Particle Tracking Solver for Direct Injection Engine Application

A dynamic load balancing technique for multiphysics simulation methods based on hierarchical Cartesian meshes for a direct injection internal combustion engine application is presented. A finite-volume method for the large-eddy simulation of the turbulent in-cylinder flow field is two-way coupled to a Lagrange Particle Tracking algorithm for the liquid fuel phase. Additionally, a semi-Lagrange level-set solver is used to track the location of the moving engine parts. The joint Cartesian mesh is used for the domain decomposition, which allows an efficient redistribution of the computational load using a space filling curve. The simulations are based on meshes with approximately 155 million cells and 1 million embedded spray parcels. Due to significant load changes, created by the solution adaptive mesh refinement, the necessity of a dynamic load balancing technique is demonstrated. The optimal load balancing interval is computed and different weighting methods for the domain decomposition are compared. The simulation results show a strong influence of the in-cylinder flow field on the fuel vapor distribution at start of ignition.