代理模型在管道减噪优化设计中的比较研究

摘要

The Kriging model is widely used in many engineering optimization problems and improves the efficiency of the optimization design.However,the efficiency of the traditional Kriging model method is not high enough for the acoustic noise reduction optimization problems with large amount of design variables and large amount of calculations.In order to further improve the design efficiency,this paper develops a more efficient and highly accurate gradient enhanced Kriging model and a set of noise reduction optimiza-tion framework for duct Acoustics problems based on Kriging surrogate model and multi-point addition criterion.Based on this optimization framework,two standard optimization functions are first used to com-pare the optimization efficiency of the gradient enhanced Kriging model and the Kriging model.Then,the Kriging model and Kriging model are used to optimize a typical intake to reduce the noise radiated from the intake,respectively.The optimization results show that the results obtained by gradient enhanced Krig-ing model are better than those obtained by Kriging model,and the computational efficiency is significant-ly improved,which confirms the effectiveness of gradient enhanced Kriging model applied in noise reduc-tion optimization for duct acoustic problems.%Kriging代理模型方法广泛应用于许多工程优化设计问题中,提高了优化设计的效率.对于设计变量多,计算量大的管道声学降噪优化设计问题,传统的Kriging代理模型方法的效率不高.为了进一步提高设计效率,结合Kriging代理模型、多点加点准则,发展了更高效、预测精度更高的梯度增强的代理模型和一套适用于管道声学问题的减噪优化设计算法框架.基于框架,首先采用两个标准优化函数,比较了梯度增强的Kriging模型与Kriging模型的优化效率.针对典型进气道算例,分别采用梯度增强的Kriging模型与Kriging模型进行降噪优化设计.优化结果显示,采用梯度增强的Kriging模型得到的优化结果优于采用Kriging模型得到的结果,且优化计算效率显著提高,证实了梯度增强Kriging模型在管道声学减噪优化问题中的有效性.