Data-driven super-resolution reconstruction of supersonic flow field by convolutional neural networks

The pursuit of high-resolution flow fields is meaningful for the development hypersonic technology. Flow field super-resolution (SR) based on deep learning a novel and effective method to provide HR in scramjet isolator. Single-path multiple-path network models convolutional neural networks (CNNs) have been developed augment spatial resolution experimental supersonic field. single-path model uses simple layer fully connected serial architecture, increases branch path by adding pooling layers achieve fusion structure architecture. Ground experiments isolator at various working conditions are conducted establish an dataset. trained CNNs compared with traditional interpolation SR reconstruction accuracy. results demonstrated that certain ability, but accuracy not satisfactory; significantly improve SR, CNN one achieves best performance.