A Fast Transition Prediction Method for High-Speed Boundary Layers Based on Neural Network
Please login to view abstract download link
The accurate prediction of boundary layer transition in high-speed represents a persistent and extensively studied challenge in fluid mechanics and aircraft aerodynamic design. Its prediction accuracy has an important impact on the calculation of drag and heat flux, and is related to the safety and economy of aircraft. At present, the eN method based on linear stability theory (LST) analysis is one of the most widely recognized transition prediction methodologies. However, LST analysis requires calculating the second derivatives of velocity and temperature profiles along the wall normal direction in a body-fitted coordinate system, guessing initial eigenvalues, and solving eigenvalue problems. This process is very cumbersome and hinders its engineering application. Considering the powerful feature extraction and nonlinear fitting capabilities of neural network, it is introduced to replace the traditional eigenvalue solving process. Baseflow of boundary layer and linear stability analysis data are used as samples for training to predict the disturbance amplification factor of the most unstable mode. A large number of numerical examples show that the proposed model can achieve high-precision prediction of the disturbance amplification factor, and the neural network model can be reused infinitely after one-time training. These findings provide an efficient data-driven alternative method for high-speed boundary layer transition prediction and promote the engineering application of LST.
