Research on supersonic flow field reconstruction using physics-informed neural networks

The shock wave system in supersonic flow fields is a critical factor influencing the quality of chemical laser beams. This study develops a comprehensive interpretation technique for supersonic flow field information by utilizing a deep learning method incorporating physical explanations. A compressible flow analysis framework of Physics-Informed Neural Networks (PINNs) is established for supersonic jets with a maximum Mach number of 10, combined with Planar Laser-Induced Fluorescence (PLIF) technology. This approach enables the reconstruction of velocity and temperature fields from concentration field measurement data. The proposed method achieves high accuracy in reconstructing data in the shock wave region, with relative L2 errors of 20.56% and 23.78% verified by Computational Fluid Dynamics (CFD) data. Experimental demonstrations showcase the ability of this method to reconstruct measured shock waves, providing an effective technical means for the comprehensive analysis of supersonic shock wave characteristics.