基于PCA-RF的永磁电机故障诊断

Fault diagnosis of permanent magnet motor based on PCA-RF

针对永磁电机振动信号非周期、非线性,特征提取困难且复杂,故障难以辨识的问题,提出了一种基于主成分分析(PCA)与随机森林(RF)的故障诊断方法。本文进行了台架实验,提取正常工况、转子偏心、定子短路、轴承内圈故障的振动时域数据,以15个转速周期划分数据段,提取每段数据共13个典型时域特征和数学统计特征,引入主成分分析法对特征降维去噪、计算方差贡献率,将得到的二维特征用随机森林进行故障分类。结果表明,与传统分类算法相比,基于PCARF的特征信息提取更加全面,具有更高的诊断精度、更快的诊断速度。

Aiming at the problems of non-periodic and nonlinear vibration signals of permanent magnet motors,difficult and complex feature extraction,and difficult identification of faults,a fault diagnosis method based on principal component analysis(PCA)and random forest(RF)was proposed.A bench experiment was performed to collect time-domain data of normal operating condition,rotor eccentricity,stator short circuit,and bearing inner ring failure.Fifteen cycles were merged into one data segment.A total of 13 typical time-domain features and mathematical statistical features were extracted from each data segment.The principal component analysis method was introduced to reduce the dimensionality of features and noise,and the variance contribution rate was also calculated.Random forest was used to classify the obtained two-dimensional features.The results show that,compared with traditional classification algorithms,the feature extraction algorithm based on PCA-RF is more comprehensive,with higher diagnostic accuracy and faster diagnosis speed.