Temporal Interpolation of Flow Fields for Knocking Combustion Using CNN

Internal combustion engines suffer from the sudden unwanted sound, formally known as knocking. Knocking is detrimental to engine efficiency and degrades its life. This makes it essential to understand knocking in its precise form. However, the problem usually faced by researchers is that knocking is a spontaneous process, posing a problem for high-end resolution cameras to capture the phenomenon in its entirety. The situation only gets exacerbated in resource-constrained research. One possible workaround is to utilize the video frame interpolation (VFI) method to enhance the simulation’s intermediate frame, hence assisting the researcher in having better insight into the knocking phenomenon. This paper attempts to study the various VFI methods for different scenarios. The study includes the performance of a few state-of-the-art VFI with respect to the frame rate and type of species. We will establish the methodology wherein, given a dataset of simulated video, it attempts to find the best model with its structural similarity measure index (SSIM) and interpolated frames.