改进共空间模式与多源特征融合的轴承智能诊断方法

Intelligent Bearing Fault Diagnosis Using Modified Common Spatial Pattern and MultiSource Feature Fusion

针对传统方法难以有效提取滚动轴承的故障特征,以及单向振动信号易存在故障信息丢失等问题,将共空间模式(CSP)算法引入滚动轴承故障诊断领域,提出了一种改进共空间模式(MCSP)算法与多源特征融合的轴承智能诊断方法。首先,利用CSP算法将轴承3个方向的振动信号进行数据层与特征层的深度融合,充分提取故障信息的同时缩短了数据处理时间;然后,采用子类空间滤波器集合策略改进决策层,构建最优空间滤波器,有效提取轴承不同健康状态的空域特征;最后,通过支持向量机(SVM)实现不同故障类型及程度的识别。使用SLIET轴承公开数据集和实验室轴承数据集进行验证,结果表明:与未改进共空间模式方法相比,改进后的方法在2个数据集上的轴承平均识别准确率均达到99%以上,且分别提升了3.04%和6.26%,平均耗时均在0.1 s以下,验证了所提方法具有更好的故障识别准确率和运行效率。

The common spatial pattern(CSP)algorithm is modified and introduced into the field of diagnosing rolling bearing faults in order to address the issues that the traditional method faces a challenge in effectively extracting the features of rolling bearing fault features and the singledirection vibration signal is susceptible to missing fault information.An intelligent diagnosis method for rolling bearings with a modified common spatial pattern(MCSP)algorithm and multisource feature fusion is proposed.Firstly,CSP is used to combine threedirection vibration signals of bearings at the data and feature levels,realizing full extraction of fault information while shortening data processing time.Then,the strategy of subclass spatial filter ensemble is improved at the decision level to construct the optimal spatial filter,so as to effectively extract the space domain features under different bearing health states.Finally,a support vector machine(SVM)is utilized to identify faults of various types and degrees.A bearing dataset of the laboratory and a publicly available one from SLIET and are used for test verification.The results show that,compared with the CSP approach,the proposed method improves the average recognition accuracy of bearing faults by 3.04%and 6.26%for both datasets,respectively.It achieves an average recognition accuracy of over 99%and an average time of less than 0.1s.This verifies the proposed method has better fault identification accuracy and running efficiency.