Research On Rolling Bearing Fault Based On Combined Feature And IPSO Optimized SVM

Rolling bearing is one of the most important parts in rotating machinery,and it is vulnerable to damage.Its operating state will affect the performance of the whole mechanical and equipment,Timely detection and diagnosis of rolling bearing can effectively avoid serious accidents.Therefore,fault diagnosis analysis of rolling bearing has important significance and practical value.For the rolling bearing vibration signal,two different methods are used to diagnose the rolling bearing fault,One is to extract the fault characteristic frequency for different fault type diagnosis,and the other is to extract the different feature parameters to form the feature vector input the support vector machine(SVM)with optimized parameters to diagnoses different types of faults and different degrees of damage.The main research contents and research methods of this paper have the following aspects:(1)Aiming at the nonlinear characteristics of rolling bearing vibration data and being vulnerable to environmental noise,a new wavelet packet threshold denoising method is proposed by introducing permutation entropy(PE)as the adjustment factor.The original vibration signal is denoised and preprocessed,and combined with complete ensemble empirical mode decomposition with adaptive noise(CEEMDAN)to extract the characteristic frequencies of the envelope spectrum of the correlation components.The different types of fault analysis are performed on the rolling bearing vibration signal.The validity of the method was verified by using the rolling bearing data of Case Western Reserve University.(2)The SVM model of radial basis kernel function is used to solve the nonlinear and small sample pattern recognition problem of rolling bearing faults.The effects of penalty factors and kernel parameters on the performance of SVM model are analyzed.Various optimization algorithms are used to optimize the parameters.For the deficiency of particle swarm optimization,the key parameters of particle swarm optimization and related improvement methods are studied and analyzed.An improved particle swarm optimization(IPSO)algorithm is proposed,which is improved in many aspects.Ten standard test functions are used to simulate and compare the results.The experiments show that the improved particle swarm optimization proposed in this paper has obvious advantages,and establishes the IPSO optimization SVM model.(3)Using the rolling bearing data of Case Western Reserve University to analyze the faults of rolling bearings with different types and different degrees of damage,using the rolling bearing data of the actual test bench to analyze the failure of different types of rolling bearings;Four single features of the original vibration signals of rolling bearings are extracted,including time domain feature,frequency domain feature,energy feature of wavelet packet nodes and permutation entropy feature of CEEMDAN components.After data normalization,the eigenvectors of eight SVM models are optimized for rolling bearing fault diagnosis.For further fault diagnosis analysis,four single features are combined to form a variety of combined features.After data normalization,principal component analysis(PCA)is used to reduce dimension for dimension problem,and feature of dimension reduction is used for fault diagnosis of rolling bearings.The experiments show that the IPSO optimized SVM model is feasible for fault diagnosis of rolling bearings,and in most cases,it has certain advantages over other optimization models.