Modeling Cycle-to-Cycle Variations of a Spark-Ignited Gas Engine Using Artificial Flow Fields Generated by a Variational Autoencoder

A deeper understanding of the physical nature cycle-to-cycle variations (CCV) in internal combustion engines (ICE) as well reliable simulation strategies to predict these CCV are indispensable for development modern highly efficient engines. Since process ICE strongly depends on turbulent flow field cylinder and, spark-ignited engines, especially around spark plug, prediction using computational fluid dynamics (CFD) is limited modeling flows. One possible way determine by applying large eddy (LES), whose potential this has already been shown despite its drawback requiring considerable time and resources. This paper presents a novel strategy based unsteady Reynolds-averaged Navier–Stokes (uRANS) CFD combination with variational autoencoders (VAEs). VAE trained data from presimulated cycles at specific crank angle. Then, can be used generate artificial fields that serve initialize new simulations process. With approach, high number individual simulated fraction LES needs same amount cycles. cycles, information transferred generated