Research On Rolling Bearing Fault Diagnosis Based On Bayesian Optimization Svm And Wavelet Packet

Rolling bearing is one of the core parts of industrial rotating equipment such as motor.Its performance directly determines whether the whole production process can be carried out safely and reliably.However,because it often operates under bad working conditions such as high temperature,high load and long-time operation,rolling bearing faults occur frequently,accompanied by the damage of mechanical devices,and even personal safety accidents and property losses.Therefore,the fault diagnosis of rolling bearing has high research value.This paper analyzes the vibration signal of rolling bearing and the fault diagnosis of rolling bearing is studied by using wavelet packet transform and support vector machine.This paper firstly proposes a fault diagnosis method of rolling bearing based on wavelet packet transform and spectral kurtosis,aiming at the problem that the vibration signal of rolling bearing is non-stationary and noisy,which makes it difficult to diagnose the bearing fault accurately.The original vibration signal of bearing is reconstructed by wavelet packet transform and kurtosis criterion,and then the filtering parameters are determined by Fast Spectral Kurtosis Algorithm.The reconstructed signal is band-pass filtered and envelope demodulated to obtain the envelope spectrum.Finally,the fault characteristic frequency obtained by the envelope spectrum is compared with the actual fault characteristic frequency.The experimental results show that this method can well screen out the fault characteristic information and realize the rapid fault diagnosis of rolling bearing.For the problem that the traditional support vector machine model is easy to fall into local optimization and over fitting in the classification process,a rolling bearing fault diagnosis method based on wavelet packet permutation entropy and Bayesian optimization support vector machine is proposed.The fault feature of bearing vibration signal is extracted by wavelet packet transform and permutation entropy,and then the Bayesian optimized support vector machine model is used for fault classification.Finally,the comparison experiment is carried out with other optimized support vector machine models.The experimental results show that Bayesian optimized support vector machine model has better classification and recognition effect,and is an effective fault recognition method.For the problem that the single-scale permutation entropy can only select the entropy on a single scale and can not extract the fault features to the greatest extent,a rolling bearing fault diagnosis method based on wavelet packet multi-scale permutation entropy and Bayesian optimization support vector machine is proposed.Using the multi-scale characteristics of multi-scale permutation entropy,the permutation entropy under the optimal scale factor of each node after wavelet packet decomposition is selected to construct the fault feature vector.The experimental results show that compared with the fault feature extraction method of single-scale permutation entropy,this method can extract the fault feature of bearing vibration signal more effectively,and has certain engineering practical value.