基于近似模型的透平机械多目标优化

Multi-Objective Optimization on the Rotor of a Transonic Axial Compressor

本文研究了采用多目标优化算法NSGA-Ⅱ与前馈神经网络的耦合优化方法的收敛性问题。提出了一种改进的拥挤距离和coarse-to-fine的迭代策略,有效地解决了原耦合方法不收敛的问题。在此基础上,提出了基于数值模拟和耦合优化方法的多目标气动设计框架,并对转子37进行了三目标优化。优化结果表明,多目标优化结果的气动性能优于单目标优化结果。同时,多目标优化可以提供多种可选择方案,具有很高的实用性。

The convergence of the coupled optimization method based on a multi-objective genetic algorithm NSGA-II and back propagation neural network is analyzed in the presented paper. An improved crowding distance and a coarse-to-fine iteration strategy are proposed to overcome the less convergence of the coupled method. An aerodynamic design framework based on the improved coupled method and CFD simulations is proposed and applied to the triple-objective optimization on the rotor 37. The aerodynamic performance of the multi-objective optimization result is better than that of the single objective optimization. Meanwhile, the multi-objective optimization can provide more alternative solutions, which is highly useful for practical applications.