Feature Evaluation And Model Construction Of Bearing Performance Degradation Evaluation

With the development of science and technology,data-driven intelligent diagnosis has gradually become a research focus.Rolling bearings are necessary parts of rotating machinery,and they are also the most easily damaged parts.It is of great significance to evaluate the performance degradation of them,determine the fault situation in time,and make corresponding maintenance strategies.Therefore,a large number of scholars began to study the evaluation method of bearing performance degradation driven by data.Data-driven intelligent evaluation methods mainly include feature extraction and intelligent models.The operating conditions of rolling bearings are complex.In view of the difficulty of vibration signal feature extraction,a feature extraction method combining wavelet packet transform and information entropy is adopted to effectively deal with the non-stationary and nonlinear characteristics of vibration signals.Utilize the advantages of strong generalization ability and good global convergence of RBF neural network to improve the accuracy and real-time performance of bearing performance degradation assessment.Finally,the life cycle data is used to verify the model’s ability to evaluate the degradation state,and the envelope spectrum is used to verify the accuracy of early failure points.The evaluation results are better than the commonly used monitoring indicators.Aiming at the commonly used alarm threshold 3σ that requires data to obey a specific distribution form,the adaptive alarm threshold box plot is studied.Boxplot uses robust quartile calculation data,which can truly reflect the actual appearance of the data,remove abnormal points,and does not require data form,and the diagnosis effect is objective and reasonable.Aim at the failure of traditional performance degradation assessment methods to determine the type of failure,an intelligent assessment method based on GA-SVM is used.Irrelevant features or redundant features not only fail to provide any valuable fault information,but also exacerbate the computational load of the data set and affect the accuracy and efficiency of intelligent evaluation.In order to select effective feature information,a feature evaluation method for bearing performance degradation is studied.The correlation,monotonicity and robustness are used to evaluate the multi-domain features composed of time-domain features and EEMD energy entropy,and the optimal features are obtained through multi-objective optimization function analysis.Define new degradation indicators and test the feasibility of the model using life-cycle data.The final results show that the intelligent evaluation model based on GA-SVM can not only effectively diagnose the types of bearing faults,the average accuracy rate of diagnosis is as high as 97.69%,but also determine early failure points,and at the same time,can accurately evaluate the degradation status of the bearing performance in combination with degradation indicators.