Scientific Article

An optimization method for material sound absorption performance based on surrogate model

¹ Systems Engineering Research Institute, Beijing 100036, China
² School of Naval Architecture and Ocean Engineering, Jiangsu University of Science and Technology, Zhenjiang 212000, China

^* Corresponding author: songhaox@163.com

Received: 13 July 2021
Accepted: 6 April 2022

Abstract

In some complex engineering design problems, the use of numerical simulation methods to solve the target value often consumes several hours or even longer, which limits the real-time response to the model. The surrogate model can solve the above-mentioned shortcomings because of its use of statistical ideas to link the design variables with the target value. Kriging model has been widely used in other fields due to its simple algorithm compilation and good stability of calculation results, but there is little research in the field of silencing structure optimization. In order to study the optimization efficiency and optimization effect of the surrogate model in the optimization design of the anechoic structure, combined with the surrogate model and the multi-point plus point criterion, a set of general optimization algorithm framework suitable for the surrogate model and the gradient-enhanced Kriging model (GEK) was developed. Based on this framework, the evolution of the sound absorption coefficient of the anechoic structure under three different working conditions (100–10 000 Hz, 100–1500 Hz, 100–10 000 Hz frequency under static pressure) was compared. The gradient enhancement Kriging model and the gradient optimization algorithm were compared and studied. The results show that under the assumption that the gradient of the objective function and the objective function have the same amount of calculation, the optimization times obtained by the Kriging model with gradient enhancement are better than those obtained by the Kriging model and the gradient optimization algorithm in most cases, and the optimization results of GEK and Kriging models are better than those of gradient optimization.