In recent years, GPFs (Gasoline particulate filters) have been installed in gasoline engines to comply with stricter environmental regulations in China and Europe. In particular, coated-GPFs having a catalytic purification function are required to have high conversion performances, high filter efficiencies in the sense of a high collection efficiency, and low pressure loss. It is not easy to design a filter that satisfies all these parameters. Experimental studies are being conducted, but it is costly to study in trial productions. In this technical paper, a GPF design optimization method will be proposed that combines multi-scale simulation, surrogate models by machine learning, and an optimization algorithm. By using this method, a GPF design that minimizes pressure loss while providing high conversion performance and particle collection rates that satisfy current regulations can be created. In addition, the examination period could be shortened by 97% compared to experimental verifications.
