Local Fault Diagnosis Of Rolling Bearing Based On Composite Sparse Filtering

As the core component of rotating machinery,rolling bearings are widely used in high-end manufacturing fields such as aerospace,rail transit and wind power generation due to their high efficiency,low frictional resistance and easy installation and lubrication.The health of the bearing during operation directly determines the stability and reliability of the equipment.According to statistics,the proportion of rotating machinery failure due to rolling bearing failure is 45%~55%.Therefore,the research on bearing fault diagnosis has reusable academic value and engineering significance.Considering the difficulty of bearing vibration feature extraction and selection under the constraints of small samples,this paper focuses on bearing vibration feature extraction,feature selection and fault diagnosis.The main research work includes the following aspects.(1)Aiming at the problem of bearing vibration feature extraction under small samples,this paper proposes a composite sparse filter based on sparse filtering,which is used to extract the linear and non-linear features of bearing vibration signals.Meanwhile,the feature mapping matrix regular term is introduced into the objective function of sparse filtering to accelerate the convergence of the algorithm.Experiments verify the superiority of composite sparse filtering in feature extraction.(2)For the problem of determining the number of linear filter and nonlinear filter in composite sparse filtering,this paper proposes a feature selection method based on improved genetic algorithm,and determines the number of features extracted by composite sparse filtering through feature selection.Aiming at the problem that the standard genetic algorithm is easy to fall into the local optimum in the early stage of evolution and the search efficiency is low in the late evolution stage,an improved genetic algorithm is proposed to improve the population initialization and genetic operators.Under the premise of small samples,the verification set or test cannot be used.The recognition accuracy of the set is used as the evaluation index of the feature subset.In this paper,the feature subset evaluation index of the distance between the multiplicative penalty factors and the intra-class distance is proposed.The index considers the feature weight and the feature number.(3)Based on composite sparse filtering and feature selection based on improved genetic algorithm,this paper proposes a new method for local fault diagnosis of rolling bearings.The flow of fault diagnosis method is introduced,and the feature map matrix adjustment based on feature selection is proposed.Through experiments,the effects of training sample size,sample length and environmental noise on the diagnosis results are studied.By comparison,the effectiveness of composite sparse filtering in feature extraction is verified by comparison with other existing methods.The method of this paper is superior in bearing fault diagnosis under small sample complex conditions.