| Rolling bearing is an important part of rotating machinery,and its health status is directly related to the overall performance of mechanical equipment.Bearing failure will not only lead to mechanical failure in engineering applications,but also lead to fatal accidents.Moreover,bearings usually undergo the degradation process from normal state to final failure in practice.Therefore,the installation of sensors to monitor the running state of bearings,extraction of fault features from these monitoring signals,health assessment and fault identification of bearings have been extensively studied in the past few decades.Based on the analysis of vibration signal,the bearing is studied by data-driven method in this paper.Firstly,the feature extraction method of bearing vibration signal is studied in this paper.In view of the common phenomenon that vibration signals contain noise components,firstly,the wavelet packet decomposition(WPD)method is introduced to reduce the noise of the original vibration signals,and then the feature extraction is realized by the Auto-Regression(AR)spectral analysis.Finally,the three-dimensional visualized scatter plot proves that the proposed method has better advantages than the traditional time domain feature extraction method.Secondly,the bearing fault diagnosis method is studied.The fault diagnosis model based on WPD-AR and DBN is established,considering that the shallow model is difficult to effectively characterize the complex mapping relation between the high-dimensional characteristic space of the vibration signal and the bearing failure.Based on the bearing data of Case Western Reserve University,the influence of variable load and different sample size on the algorithm is discussed.At the same time,by comparing with other common fault diagnosis methods,it is proved that the proposed method has better accuracy.Then,based on the life cycle data of University of Cincinnati,the assessment method of bearing health is studied.In order to solve the problem that a single feature is difficult to represent the healthy status of the bearing,this paper first calculates the cosine similarity between the whole life cycle signal and the healthy state signal based on the feature extraction,and then the Exponential Weighted Moving Average(EWMA)algorithm is adopted to obtain the healshow th index for evaluating the health status of bearings.The experimental results that the proposed method is more sensitive than the traditional time domain parameters.Finally,the rolling bearing fault diagnosis system is designed and developed.Based on the theoretical research of data processing and fault diagnosis algorithm,a rolling bearing fault diagnosis system based on the hybrid programming of C#and MATLAB is developed,and the bearing data of Case Western Reserve University are used to verify the system functions.Software functions include vibration signal analysis,data processing,fault identification,data storage and historical data viewing. |
