Fault Diagnosis And Performance Degradation Assessment Of Rolling Bearings

Rolling bearings are very important part in mechanical equipment.It plays a role in transmitting load and bearing stress during mechanical transmission.Due to long-term operation under heavy load and high speed conditions,rolling bearings are the parts that are extremely prone to damage.And the damaged rolling bearings cause overall or partial failure of mechanical equipment.Understanding the performance degradation of rolling bearings,and making accurate diagnosis of early failures of rolling bearings,which is conducive to real-time monitoring and prediction of bearing changes,formulating different maintenance methods,and ensuring the smooth implementation of production plans.Therefore,this article focuses on the fault diagnosis and performance degradation of rolling bearings.The main task is to extract the features by the bearing signal and establish a fault diagnosis model and a degradation assessment model.The following are the main contents:The article introduced the basic composition,vibration mechanism and failure modes.For each part of the rolling bearing,the calculation formula of the characteristic frequency of the failure is listed.The article studied the signal characteristic of the rolling bearing,and discussed the corresponding characteristic index in the time and frequency domains.Because the fault frequency of the early signal is not obvious,this paper has conducted a study on the early failure of rolling bearings.It uses an improved sparse representation method to enhance the failure frequency of rolling bearing signals.Compared with the theoretical failure frequency,it can determine the type of failure and the degree of damage.The introduction of envelope spectral kurtosis enables the algorithm to adaptively retain features with smaller weights,and the non-convex regular term ensures the sparsity of the signal.Finally,the whole life cycle data of the rolling bearing in University of Cincinnati is taken as an example to conduct experiments.The experiment indicates that the suggested method can effectively reduce the signal-to-noise ratio of the signals.The proposed method enhances the fault frequency and performs early fault diagnosis accurately.For the problem of bearing data imbalance in actual production,a fault diagnosis method based on feature generation of bearing imbalance data is proposed.The paper combines CNN and CSMO.For achieving data equalization,CNN extracts features,and CSMO generates new feature data to.Taking the experimental data as an example,the simulation experiment shows that compared with the traditional time-frequency domain features,the proposed algorithm has the ability of adaptive feature extraction.So it can find deep-level features.Compared with other classifiers,SVM has better classification capabilities and is suitable for diagnosing problems in the case of small samples.In addition,the method of feature generation can improve the accuracy of diagnosing imbalance problems.The traditional degradation features have some problems such as flat curve and wrong identification of degradation points.In this paper,a new degradation index is constructed.Firstly,the wavelet packet singular spectrum entropy of the data is obtained.The method calculates the MD between the health data and the test data,and obtains the MD feature set.Then the performance degradation of rolling bearing is evaluated well.Taking the data from the University of Cincinnati as an example,compared with traditional time-domain features,experiments show that Mahalanobis distance can better quantify the degradation process,and singular spectrum entropy can better present the degradation state of bearing signals.The proposed method accurately evaluates the bearing degradation,and the degradation curve is more intuitive.