Application of hybrid neural network in data-driven flow field simulation

Due to the strong nonlinearity of navier stokes equation, it is difficult to solve the hydrodynamics simulation problem. As a century problem, it is still a major difficulty in the academic community. With the improvement of computer ability and the development of data platform, some new changes have taken place in the research direction and content of turbulence model. The data-driven machine learning method is different from the traditional approximate equation solving method in physics, and shows its application potential in highly complex flow fields. In this study, convolution cyclic hybrid neural network is used to predict the complex flow field, and the generated confrontation network is used to generate the simulated flow field.