基于离散小波变换和随机森林的轴承故障诊断研究

针对不同工况下数据特征选择困难和单一分类器在滚动轴承故障诊断中识别率较低等问题,提出了一种基于离散小波变换和随机森林相结合的滚动轴承故障诊断方法。该方法首先利用离散小波变换分解振动信号,得到n层近似系数;然后创新性地采用sigmoid熵构造出n维特征向量,sigmoid熵能较好地提取非平稳信号的特征,提高诊断准确率;最后采用随机森林对滚动轴承不同故障信号进行分类。实验采用西储凯斯大学轴承数据中心网站提供的轴承数据,与传统分类器(KNN和SVM)以及单个分类回归树CART进行对比分析,结果表明该方法具有更好的诊断效果。

Research on bearing fault diagnosis based on discrete wavelet transform and random forest

Aiming at the difficulty of data feature selection under different working conditions and the low recognition rate of single classifier in rolling bearing fault diagnosis,this paper proposed a rolling bearing fault diagnosis algorithm based on discrete wavelet transform and random forest.Firstly the proposed method decomposed the vibration signal by discrete wavelet transform to get n-layer approximate coefficients.Then,it used the sigmoid entropy to construct n-dimensional eigenvectors innovatively.The sigmoid entropy could extract the features of non-stationary signals better and improve the diagnostic accuracy.Finally this paper used random forest to diagnose different fault signals of rolling bearing.It used the bearing data provided by the bearing data center website of Case Western Reserve University for experiments.Comparing with the results of traditional classifier(KNN and SVM)and single classification regression tree CART,this method has better diagnostic results.