基于小波和马氏距离的电力电子电路故障诊断

Fault diagnosis of power electronic circuit based on wavelet and mahalanobis distance

基于小波分析和马氏距离,提出了一种电力电子电路故障诊断的方法,并应用于DC-DC变换电路。首先仿真所有故障模式下的电路,对采样所得输出数据进行小波分析,提取特征向量以简化数据,并计算特征向量的重心及协方差矩阵的逆作为最终的故障特征建立故障字典。然后在诊断时,测量未知故障电路数据,采样并进行小波分析,只要计算未知故障与故障集之间的马氏距离就可以判断待测电路的故障。这种方法覆盖了参数性故障及结构性故障这两方面,并且通过实例表明可以达到较高的故障检测率。

Based on wavelet analysis and the Mahalanobis distance theory,a fault diagnosis method for power electronic circuits is presented and used in the DC-DC circuit.Firstly,use wavelet to decompose the output voltage data that sampled with Monte Carlo simulation,which is made for both fault-free and faulty circuits with tolerance.Secondly,we get matrixes as characteristic vectors,and calculate the focuses and covariance matrixes of these characteristic vectors.Thirdly,a circuit with an unknown fault is simulated.Mahalanobis distances between current state and each fault are calculated to judge which fault current the circuit is.This method can diagnose both parametric faults and structural faults.The example shows that this method can also get a high diagnosis correct rate,has a high ability to test circuit fault.