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.
出处
《航空动力学报》
EI
CAS
CSCD
北大核心
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)
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
wavelet neural networks （WNN）
Pareto genetic algorithm
fluidic element
turbomachine
optimization design
