基于BP人工神经网络的离心压气机叶轮多目标优化设计方法

利用Concept NREC软件建立离心压气机叶轮设计样本库，借助BP（back propagation）人工神经网络建立样本库中各设计参数与压气机性能之间的关系，接下来以多目标遗传算法寻找Pareto解，从而获得离心压气机叶轮最佳设计参数.将该方法应用于Krain叶轮设计工况，所得叶轮的效率、压比较Krain叶轮原型分别提高1.4%和10.9%.通过对人工神经网络模型可靠性的讨论、多目标优化模型的主成分分析和所设计叶轮性能的CFD验证，证明了所构建的目标函数与所获得的Pareto解集的合理性，说明本方法可以有效应用于在离心压气机设计、选型. 

Multi-objective optimization design of centrifugal compressor impeller based on BP artificial neural network

The database of the centrifugal compressor impeller performance was given by Concept NREC software. The relationship of the impeller''s main design parameters and performance was established by BP (back propagation) artificial neural network. Then the optimal array of main design parameters was acquired by the multi-objective genetic algorithm. This design method was used to satisfy Krain impeller''s performance objective; the optimized impeller''s efficiency and pressure ratio increased by 1.4% and 10.9%, respectively. Through discussion of artificial neural network model''s reliability, multi-objective optimization model''s principal component analysis and the impellers'' CFD numerical simulation verification, the validity of objective functions and Pareto optimal solutions was proved, the effectiveness of the present centrifugal compressor design method was well confirmed.