基于平滑先验分析改进散布熵的滚动轴承故障诊断

Improved Rolling Bearing Fault Diagnosis Based on Smoothing Prior Analysis

运用SPA散布熵与SG聚类方法对滚动轴承故障进行智能诊断。先通过平滑先验分析(SPA)的过程处理轴承振动信号获得趋势项与波动项;再对散布熵值进行计算得到故障特征向量,最后利用SG分类器对特征向量聚类识别。研究结果表明:通过分析系统参数对DE值的影响确定最优的取值:高嵌入维数3,类别6,时间延迟1。利用SG模糊聚类方法处理SPA-DE,不同的故障类型聚集于靠近聚类中心的区域,相邻聚类中心间形成了明显间隔,并未出现不同故障问题的相互交叉混叠,因此采用本文方法对各类轴承故障进行诊断时可以实现优异故障分类性能。相对EMD-DE-SG聚类诊断模式,本文算法具备更大分类系数,与1达到了更大的接近程度,大幅降低了平均模糊熵,判断本文算法对于各损伤程度的轴承故障具备更优的诊断性能。根据滚动轴承测试参数可知可以采用此方法快速准确识别轴承各类故障问题及其严重程度,可以实现故障的可靠诊断效果。

The intelligent diagnosis of rolling bearing faults is carried out by using smoothing prior analysis(SPA)dispersion entropy and SG clustering method.Firstly,the trend term and wave term of bearing vibration signal are obtained by SPA.The fault feature vector is obtained by calculating the dispersion entropy value,and finally the feature vector is identified by clustering using the SG classifier.The results show that the optimal values are determined by analyzing the influence of system parameters on DE value:high embedding dimension 3,category 6,time delay 1.The SG fuzzy clustering method is used to deal with SPA-DE.Different fault types cluster in the area close to the cluster center,and obvious intervals are formed between the adjacent cluster centers.There is no cross-aliasing of different fault problems,so the proposed method could achieve excellent fault classification performance when diagnosing various bearing faults.Compared with emD-DE-SG clustering diagnosis mode,the proposed algorithm has a larger classification coefficient,which is closer to 1 and significantly reduces the average fuzzy entropy.Therefore,it is judged that the proposed algorithm has better diagnostic performance for bearing faults with various damage degrees.According to the test parameters of rolling bearings,this method can be used to quickly and accurately identify all kinds of bearing faults and their severity,which can achieve reliable fault diagnosis effect.