基于QPSO算法的压气机特性代理模型优化

Surrogate Model Optimization of Compressor Characteristics Based on QPSO Algorithm

关键词:  量子粒子群优化  压气机特性  代理模型  Kriging  优化  重构

Key words:  QPSO  Compressor characteristics  Surrogate model  Kriging  Optimization  Reconstruction

摘要:

考虑到小样本特性数据情况下进行部件特性数据的二维线性插值精度低，提出一种基于量子粒子群优化(QPSO)算法的压气机特性代理模型优化方法。针对原始Kriging模型对其相关模型参数的初始值极度敏感以及易限于局部最优解的缺陷，利用QPSO算法对Kriging的相关模型参数进行全局寻优，克服了基于梯度的模式搜索法对于参数初始值的依赖，经测试该方法具有较好的效率以及稳定性。将该优化模型扩展应用于低压压气机特性代理模型建立与重构。经验证，在小样本特性数据下，基于QPSO的压气机特性Kriging模型仍具有较高精度，应用前景可观。

Abstract:

Inspired by the fact that two-dimension linear interpolations of engine component characteristics have a low precision with a small sample of characteristic data ，a method on surrogate model optimization of compressor characteristics based on quantum-behaved particle swarm optimization (QPSO) algorithm is proposed. Firstly，considering the limitations that the original Kriging model is sensitive to the initial value of parameters of its correlation model and extremely easy to fall into local optimal solution，the QPSO algorithm is introduced to search the global optimal solution to the parameters of the Kriging correlation model，and overcome the dependence on initial values of the pattern search method based on gradient information. The experimental results show that this method has good efficiency and stability. Secondly，the optimized Kriging model based on QPSO is applied to the establishment of surrogate model and reconstruction of a low-pressure compressor characteristics. It is proved that even with a small sample of characteristic data，surrogate model of the compressor characteristics based on QPSO still has a high accuracy，and its application prospect will be considerable.