自适应Kriging近似模型及其在二维扩压器优化设计中的应用

Application of adaptive Kriging approximation model in two dimensional diffuser aerodynamic optimization design

将均匀设计方法、CFD技术、Kriging近似模型及小生境微种群遗传算法相结合发展了一种自适应全局优化设计方法。优化过程中综合考虑Kriging模型的预测值与预测标准差,引入了EI(Expected Improvement)函数得到校正点,解决了采用近似模型最优策略得到校正点带来的局部收敛问题。分别采用该方法和小生境微种群遗传算法进行扩压器气动优化设计,以扩压器平均静压恢复系数为目标函数并采用Nurbs曲线完成几何参数化建模。优化后扩压器平均静压恢复系数提高了8.5%,结果表明本文方法比随机性优化算法更为有效。

An adaptive global optimization method was developed coupled with uniform experimental design,CFD analysis,Kriging approximation model and niching micro genetic algorithm.In the optimization procedure,EI function was introduced to identify the next sampled point by considering the prediction and mean squared error of Kriging model to decrease the risk of trapping into the local optimum when the optimal strategy was used.The proposed method and niching micro genetic algorithm have been applied to the diffuser optimization design respectively,average static pressure recovery coefficient is selected as objective function and Nurbs curve is used to parameterize the geometric model.8.5% improvement of average static pressure recovery coefficient is obtained and the result shows that the method is more effective than stochastic optimization algorithm.