基于小波神经网络的水轮机叶片裂纹源的定位技术  

Source Location of Cracks of Turbine Blades Based on Wavelet Neural Network

作　　者：王向红[1] 朱昌明[1] 毛汉领[2] 黄振峰[3] 

机构地区：[1]上海交通大学机械与动力工程学院,上海200240 [2]广西广播电视大学,南宁530004 [3]广西大学机械学院,南宁530004

出　　处：《上海交通大学学报》2008年第8期1301-1304,1309,共5页Journal of Shanghai Jiaotong University

基　　金：国家自然科学基金资助项目(50465002)

摘　　要：针对水轮机结构复杂等特点,传统的时差定位及模态定位方法不能满足其裂纹水轮机叶片定位要求,提出利用小波神经网络对水轮机转轮叶片的裂纹进行定位.训练采用标度共轭梯度算法(SCG),并对输出结果采用竞争处理方式.结果表明,与BP网络相比,小波神经网络提高了定位的准确度,所确定的裂纹位置最大误差仅为4.2%,是一种适合复杂结构的定位方法.Turbine runner has a complex structure. The traditional source location methods, such as time of arrival and modal analysis, can not satisfy the accuracy. This paper described source location of cracks of turbine blades using a wavelet neural network （WNN）. The scaled conjugate gradient algorithm （SCGA） was used in the training iteration. The outputs were dealt with competing mode. The results show that the WNN can improve the accuracy of location compared with BP neural network, which indicates the WNN algorithm is a suitable source location method for complex structures.

关 键 词：小波神经网络 标度共轭梯度算法 源定位 声发射 

分 类 号：TB52[理学—物理] TP183[理学—声学]