基于支持向量机的水电机组故障诊断

Support Vector Machines Based Approach for Hydroelectric Generating Unit Fault Diagnosis

针对水电机组故障样本少的问题,将支持向量机引入水电机组故障诊断研究,提出一种结合小波频带分解与最小二乘支持向量机的水电机组故障诊断模型。基于机械设备"能量-故障"映射关系,运用小波分解提取机组振动信号各频带能量特征值,然后将能量特征值输入到多分类的支持向量机,实现对机组不同故障类型的识别。通过实验信号分析,表明将小波能量提取与支持向量机结合进行水电机组故障诊断是可行有效的,并具有较高的故障分辨能力,为水电机组故障诊断提供了新的方法和思路。

To solve the small-sample problem in the hydroelectric generating unit （HGU）, SVM （Support Vector Machines） is introduced, and a new diagnosis model of least squares support vector machines （LS-SVM） based on wavelet decomposition with energy extraction is proposed in this paper. Based on the mapping of energy and fault, the energy features of different frequency band of vibration signal are extracted through wavelet decomposition. These extracted feature vectors are input into the multi classification LS-SVM to detect different abnormal cases. Testing results show that the model is feasible and effective in the HGU fault diagnosis and has higher accuracy of classification, which provides a new method for fault diagnosis of HGU.