Vehicle Aerodynamic Optimization: On a Combination of Adjoint Method and Efficient Global Optimization Algorithm 06-12-02-0011
This also appears in
SAE International Journal of Passenger Cars - Mechanical Systems-V128-6EJ
This article presents a workflow for aerodynamic optimization of vehicles that for the first time combines the adjoint method and the efficient global optimization (EGO) algorithm in order to take advantage of both the gradient-based and gradient-free methods for aerodynamic optimization problems. In the workflow, the adjoint method is first applied to locate the sensitive surface regions of the baseline vehicle with respect to the objective functions and define a proper design space with reasonable design variables. Then the EGO algorithm is applied to search for the optimal site in the design space based on the expected improvement (EI) function. Such workflow has been applied to minimize the aerodynamic drag for a mass-produced electric vehicle. With the help of STAR-CCM+ and its adjoint solver, sensitive surface regions with respect to the aerodynamic drag are first located on the vehicle. Then the design samples are determined with the help of the uniform design (UD) method and trained by the EGO algorithm (written by MATLAB) for searching the optimum. The optimization results are analyzed to validate the workflow and provide more insights into the improvement of the flow field.