Reconstructing cellular surface of gaseous detonation based on artificial neural network and proper orthogonal decomposition

Abstract

Gaseous detonation has complicated cellular surface, whose comprehensive investigation is critical not only to the detonation physics but also the detonation engine development. Because measuring the high-resolution dynamic surface is beyond the present experimental technical skills, we propose a reconstruction method of detonation wave surface based on post-surface flow field. This method combines two technologies, the proper orthogonal decomposition (POD) in fluid research and the artificial neural network (ANN) in machine learning research. POD is employed to extract the main features of flow fields, and the pre-trained ANN builds up the connection between the reduced coefficients of full flow fields and post-surface flow fields. The reconstruction is tested through the numerical results from one-step irreversible heat release model, displaying a good performance in both cellular normal detonations and unstable oblique detonations. The method may provide a universal frame for the detonation research, and has the potential to be employed in other numerical and experimental results.