Performance Degradation Evaluation Of Rolling Bearings Based On Fusion Probabilistic Modeling And Boundary Distance

Along with the development of science and technology,intelligent diagnosis of equipment has become a research hotspot.Rolling bearing as one of the major rotating machinery,its running state has always been a research hot spot.The actual service life of some bearings is much higher than expected.However,the same batch of bearings maybe only work for a short time before it is seriously damaged that have to be replaced.As a consequence of this,the traditional method of periodic maintenance has been unable to make the bearing life to obtain the greatest advantage.It is of great practical significance to evaluate its performance degradation.In this article,a variety of feature extraction methods are used to extract features from the vibration signal.The dimensions of extracted features are reduced by the improved LLE,then inputs them into probabilistic models,distance models,and models based on fusion probability modeling and boundary distance to obtain the performance degradation index,and draw the performance degradation evaluation curve.The details are as follows:The time domain features have the ability to reserve the primitive signal information as much as possible.The AR energy ratio treatment has a good effect on processing non-stationary signals.The LMD energy entropy can reflect the malfunction circumstance of the bearing.Therefore,the article uses these three feature extraction methods,and then with the improved LLE for dimension reduction.In the end,the feature vector after dimensionality reduction is input into the subsequent degradation evaluation model.There prove no defined upper value of boundary distance models based on Mahalanobis distance and Euclidean cosine distance.Besides,the probabilistic models based on HMM and GMM tends to exist the shortcoming of premature saturation.As a result,this paper puts forward a new performance degradation model.After dimensionality reduction,the feature matrix is input into the probabilistic modeling model and the boundary distance model respectively to attained two performance degradation indexes.And then put them as a new performance degradation index input,getting new performance degradation index,drawing performance degradation assessment curve,blending the merits of both boundary distance and probability model to achieve rolling bearing performance degradation assessment process.In order to verify the merits of the model,the article uses the CEEMDAN and Hilbert envelope method to validate the initial point of failure time.A method based on monotony,correlation and robustness is proposed to evaluate the performance of the model.