基于径向基多小波神经网络的水电机组故障诊断  

Fault diagnosis of hydro-turbine generating unit based on radial basis multiwavelet

作　　者：卢娜[1] 肖志怀[1] 曾洪涛[1] 符向前[1] 

机构地区：[1]武汉大学动力与机械学院,湖北武汉430072

出　　处：《武汉大学学报（工学版）》2014年第3期388-393,共6页Engineering Journal of Wuhan University

基　　金：国家自然科学基金项目(编号:51179135;51379160);中央高校基本科研业务费专项资金资助项目(编号:201120802020004)

摘　　要：针对常规神经网络收敛速度慢,难以实现水电机组故障在线学习的不足,提出基于径向基多小波神经网络的水电机组故障诊断方法.采用多小波尺度函数作为径向基多小波神经网络的核函数,建立网络模型.利用水电机组振动信号频谱分量的幅值作为特征向量,对网络进行训练,实现特征样本集到振动故障集的有效映射,达到水电机组故障诊断的目的.实验结果表明:与常规神经网络的诊断方法相比,径向基多小波神经网络水电机组故障诊断方法具有较快的收敛速度和较好的泛化能力,为水电机组故障在线学习和诊断提供了有效的解决途径.The traditional neural networks suffer from deficiency of slow convergent speed; therefore, they cannot be trained online when used as classifiers in fault diagnosis of hydro-turbine generating unit. In or- der to overcome this disadvantage, a novel radial basis multiwavelet net method is investigated. Multiscal ing functions are used as the kernel functions of the proposed neural network. Spectrum analysis is applied to extract the features of the vibration signals in frequency domain; and then these features are taken as learning samples to train the network and realize the mapping relationships between the spectrum features and the faults. The diagnosis results show that, comparing with traditional neural network methods, ra- dial basis multiwavelet has faster convergence speed and better generalization ability. Consequently, it can be trained online and provides an effective solution for fault diagnosis of hydro-turbine generating unit.

关 键 词：水电机组 故障诊断 径向基多小波神经网络 径向基神经网络 

分 类 号：TM312[电气工程—电机]