基于蚁群初始化小波网络的水电机组振动故障诊断  

Vibration fault diagnosis of hydroelectric generating unit based on ACO-initialized wavelet network

作　　者：卢娜[1] 肖志怀[1] 符向前[1] 

机构地区：[1]武汉大学水力机械过渡过程教育部重点实验室,武汉430072

出　　处：《水力发电学报》2014年第2期251-258,共8页Journal of Hydroelectric Engineering

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

摘　　要：针对传统小波网络在进行故障诊断时存在收敛速度慢,对初始参数敏感的缺陷,提出了基于蚁群初始化小波网络的水电机组振动故障诊断方法。该方法采用蚁群算法对小波网络的参数进行初步寻优,将优化后的参数作为小波网络的初始化参数;利用水电机组振动信号频谱分量的幅值作为特征向量,对蚁群初始化小波网络进行训练,实现振动特征集到故障集的有效映射,达到故障诊断的目的。实例诊断结果表明:与传统小波网络及蚁群优化小波网络相比,基于蚁群初始化小波网络的水电机组振动故障诊断方法具有较快的收敛速度和较强的泛化能力,为水电机组振动故障在线诊断提供了有效的解决方案。Traditional wavelet network suffers from slow convergent speed and high sensitivity to initial parameters when used as classifiers for the fault diagnosis of hydroelectric generating unit. In order to overcome these disadvantages, an ACO-initialized wavelet network was investigated. In this method, ACO algorithm was applied to initialize the parameters. Firstly, features were extracted from vibration signals in frequency domain by spectrum analysis. Then, these features were taken as learning samples to train the wavelet network and realize the mapping relationships between the spectrum features and the vibrant faults. The experimental results show that, comparing with traditional wavelet network and the wavelet network optimized by ACO algorithm, the proposed method has faster convergence speed and better generalization ability. Therefore, it provides an effective solution for online fault diagnosis of hydroelectric generating unit.

关 键 词：动力机械工程 故障诊断 小波网络 水电机组 蚁群算法 

分 类 号：TM312[电气工程—电机]