AI-Based Shape Optimization for Drag Reduction of a High Speed Train Head/tail

This work aims to develop an AI-based shape optimization method for the aerodynamic drag reduction of a highly streamlined high speed train (HST) head/tail. This method includes five major modules, i.e., parameterization, design space pre-optimization, reduced order model, machine-learning-based optimization of control variables and adjoint optimization. The aerodynamic performances, such as drag and lift coefficients, of the HST are evaluated numerically via Reynolds-averaged Navier–Stokes equations. This method may deal with a very large number of control points, in the order of 1500, and control variables, up to 40. It has been demonstrated that, given a speed of 400 km/h, the method may reduce the total drag of the HST with 3 carriages up to 9.36%, exhibiting a great potential in the shape optimization for enhancing aerodynamic performance of HSTs.