基于高斯混合模型的滚动轴承故障预警

Rolling Bearing Fault Warning Based on Gaussian Mixture Model

滚动轴承的早期故障预警一直是研究人员和相关行业关注的问题,及时发现滚动轴承的早期故障并预警有助于降低生产中因零件损坏引发的损失。在分析了主流故障预警方法后提出一种基于高斯混合模型(gaussian mixture model, GMM)的轴承故障预警方法;通过GMM对轴承的振动信号建模,描述其不同阶段的分布情况,提出一种新的基于KL散度的轴承健康指标(bearing health index based on KL divergence, BHI-KL),用来描述轴承劣化过程;利用3σ准则提取出健康指标中的异常值,实现故障预警。利用轴承寿命加速试验数据对所提方法进行验证,并通过包络谱验证其精确性。结果表明,该方法较常用的故障特征具有良好的时效性,可以实现对轴承故障进行有效预警。

Early fault warning of rolling bearings has always been a concern of researchers and relevant industries.Timely detection and early warning of rolling bearings′early fault can help reduce the loss caused by parts damage in production.After analyzing the mainstream fault early warning methods,a bearing fault early warning method based on gaussian mixture model(GMM)is proposed;The vibration signal of bearing is modeled by GMM to describe its distribution in different stages.A new bearing health index based on KL divergence(BHI-KL)is proposed to describe the bearing deterioration process;The abnormal values of health indicators are extracted by using the three sigma criterion to achieve fault early warning.The proposed method is verified by using the accelerated bearing life test data,and its accuracy is verified by envelope spectrum.The results show that the method has good timeliness compared with common fault features,and can achieve effective early warning of bearing faults.