一种小波神经网络与遗传算法结合的优化方法  

Optimization method by combination of wavelet neural networks and genetic algorithm

作　　者：王保国[1] 刘淑艳[1] 钱耕[1] 南希[2] 郭宇航[1] 

机构地区：[1]北京理工大学宇航科学技术学院,北京100081 [2]北京航空航天大学航空科学与工程学院,北京100191

出　　处：《航空动力学报》2008年第11期1953-1960,共8页Journal of Aerospace Power

基　　金：国家自然科学基金(50376004);高等学校博士学科点专项基金(20030007028)

摘　　要：提出一种基于小波神经网络(简称WNN)与Pareto遗传算法相结合的优化方法,并用于内流的数值流场优化计算.小波神经网络由输入层、隐含层和输出层组成.在隐含层用Morlet小波母函数取代了误差反向传播(BP)神经网络中常用的Sigmoid激励函数.Pareto遗传算法具有很好的全局寻优能力和良好的优化效率,在通常情况下它总可以得到均匀分布的Pareto最优解集.典型算例表明:该算法快速、高效,能高精度的完成非线性函数的逼近与映射,其泛化能力很强.An optimization method based on the combination of wavelet neural networks （WNN）and Pareto genetic algorithm was proposed, and was applied to the numerical optimi zation in internal flows. WNN is composed of input layer, hidden layer and output layer. It replaces the commonly used Sigmoid activation function in back propagation （BP） neural net works by Morlet wavelet generating functions in hidden layer. Pareto genetic algorithm has great global optimum ability and optimization efficiency. Generally, it can always gain uniformly distributed Pareto optimal solution set. Typical algorithm examples indicate that this algorithm can complete approaching and mapping of non-linear function quickly, efficiently and accurately, with great generalization ability.

关 键 词：小波神经网络 Pareto遗传算法 射流元件 叶轮机械 优化设计 

分 类 号：V231.3[航空宇航科学与技术—航空宇航推进理论与工程]