Response Surface Based Aerodynamic Shape Optimization of High Speed Train Nose

Abstract

The aim of this work is to demonstrate the possibility of using adaptive response surface for optimization problems which require expensive computations. In this paper and attempt is made to address the robustness of the optimization algorithms and make improvements. A new optimization algorithm, hybrid GA PSO algorithm was introduced and was found to be robust and accurate than the original GA or PSO algorithms. The new hybrid algorithm was used to identify the optimum nose shape of the train. The prediction of aerodynamic drag requires computational uid dynamic simulations. To limit the number of computer simulations required to perform the optimization, a response surface identical to kriging model with accuracy check was combined with the shape optimization algorithm. Initial results show that the combined shape optimization algorithm requires a small number of simulations to nd the optimum design compared with other methods. The suggest method not only requires a small number of simulations but is also robust. The assumption that it requires around 10 simulations for each of the design variables in the optimization problem needs further study to further reduce the number of simulations without loosing the eectiveness of the algorithm.