Research On Bearing Fault Diagnosis Method Based On Feature Network

With the development of IoT,the massive data generated by mechanical equipment has driven fault diagnosis into the era of "big data".Intelligent fault diagnosis becomes a future development trend in a data-driven manner.Rotating machinery is widely used in the field of mechanical engineering and plays an important role in engineering applications.As one of the most significant parts of rotating machines,rolling bearing has a great influence on operating status of mechanical equipment.So it is of importance to monitor the running state of the bearings to ensure the safe operation of the rotating machinery.How to improve the bearing fault diagnosis accuracy through feature selection,and fuse multi-source information to ensure the diagnosis result more reliable when facing multi-sensor problem are the focus of this article.The main research work is described as follow.First,this paper proposed a novel feature selection method based on bearing fault diagnosis called Feature-to-Feature and Feature-to-Category-Maximum Information Coefficient(FF-FC-MIC),which considers the relevance among features and relevance between features and fault categories by exploiting the nonlinearity capturing capability of maximum information coefficient.In this method,a weak correlation feature subset obtained by a Feature-to-Feature-Maximum Information Coefficient(FF-MIC)matrix and a strong correlation feature subset obtained by a Feature-to-Category-Maximum Information Coefficient(FC-MIC)matrix are merged into a final diagnostic feature set by an intersection operation.Second,on the foundation of FF-FC-MIC,a novel feature selection net is presented.In the net,the feedback conditions are set with the feature selection dimensions and running time of feature selection algorithm after utilizing the FF-FC-MIC.When the feedback conditions are met,the feature weights are calculated through CART,selecting the features whose weight are not 0 for the first.After that,features selected by CART are selected again by FF-FC-MIC for the second time,constructing a feature net owing a feedback effect.Comparing to the FF-FC-MIC method,the feature network not only reduces the feature dimension,but also greatly reduces the time used for feature selection and realizes online feature selection.Third,an IDSbKC(Improved Dempster-Shafer Evidence Theory based on Kappa Coefficient)information fusion method is proposed for multi-sensor information in bearing fault diagnosis to deal with multi-source sensor data consistency problem.In the method,consistency of diagnostic results for different sensor data is verified by calculating the Kappa coefficient for the purpose of revising the weight of evidences.And then,the weight of sensors are revised through the average diagnosis accuracy for different sensors.Based on the revised weight of evidences and sensors,the DS evidence theory is improved.Comparing to traditional DS evidence theory and the DS evidence theory based on distance function,IDSbKC can achieve higher average fusing accuracy and promote the reliability of diagnostic results.The framework proposed in this study provides a new idea for the bearing fault diagnosis feature selection method,which is helpful for online fault diagnosis of bearing faults,and has certain practical application value.