多元统计理论在整流电路故障诊断中的应用

Application of Multivariate Statistical Theory in Fault Diagnosis of Rectification Circuit

在研究三相晶闸管整流电路模型特点的基础上,提出了一种基于主元分析法和决策树理论的整流电路故障诊断方法。首先对整流电路的输出波形采样,并对其进行快速傅里叶变换,提取关键频率点的功率谱值组成特征向量,再应用主元分析法提取特征分量的主元,得到降维的特征空间,最后建立决策树对本例中的故障进行准确分类,最终实现了若干典型故障的分离。仿真实验及结果分析证明,此方法可以提高识别速度和测试速度,且不会改变样本的分布特性,保持样本的分类信息。

On the basis of the research on the characteristics of the three-phase thyristor rectification circuit, we adopt principal component analysis and decision tree to diagnose faults in rectification circuit. Firstly, we sample the output waveforms of the rectification circuit ,apply fast Fourier transform （FFT） and extract values corresponding to key frequency points to constitute feature vectors. Based on principal component analysis method, the principal components and compacted feature space are obtained ,then several typical fault patterns are separated successfully by decision tree . Simulation result proves that this method can improve identification speed and test speed, simultaneously, distribution of the sample is unchangeable and classification information of the sample is retained.