Surrogate models with kriging and POD for the optimization of an intake port

1. Abstract The automotive industry is facing increasing demands in terms of cost reduction and environment responsibility. To comply with new European emission standards, while satisfying the need for higher engine performances, the intake ports have to be optimized with respect to the mass flow rate and the aerodynamic turbulence inside the combustion chamber. In this study, the shape of an intake port is optimized for a gazole engine. CFD calculations are post-processed to maximize two competing objectives: the mass flow and the tumble. Nevertheless, as the evaluation of these criteria is expensive (more than 2 hours of CFD computations for each design), the optimization is split into two phases: first, a series of refined computer experiments is designed and performed to get the velocity field inside the cylinder (from which the mass flow and tumble are calculated); then, the optimization is performed with surrogate models. We investigate two families of reduced models for the multiobjective optimization of the intake port geometry: kriging surrogate surfaces, based directly on the post-processed values, and Proper Orthogonal Decomposition (POD) models to reduce the 3D velocity field. 2.