Research On Early Weak Fault Diagnosis Of Rolling Bearing Based On Manifold Learning

When the rolling bearing is out of order,its vibration signal often exhibits a strong non-stationary and non-linear characteristics.Therefore,it's the key point that how to obtain the fault feature information and identify the fault type from these non-stationary and nonlinear signals.The empirical mode decomposition(EMD)can provide time domain,frequency domain local information of the vibration signal by decomposing the signal with self-adaption,so it is used by this paper to extract the fault feature of vibration signal of the rolling bearing to construct the original feature quantity.However,these original features have higher dimensions and redundant information,making against the classification and recognition.While the manifold learning local tangent space alignment algorithm(LTSA),as a nonlinear signal processing method,has the good capability of dimension reduction.Therefore,this paper proposes a fault diagnosis method of rolling bearing based on the combination of the empirical mode decomposition(EMD)and manifold learning(LTSA).Firstly,the vibration signal of rolling bearing is decomposed adaptively with EMD method to calculate the covariance matrix eigenvalues of each component(IMF),consisting the original feature set of rolling bearings.Then the original feature set is reconciled and extracted further by the LTSA.At last,all the new characteristics are imported into the K-means classifier for fault recognition.The main contents of the paper are as follows:(1)As the the experimental object,the self-aligning ball bearings are fault-processed by SG bimetallic engraver to simulate the early weak pitting failure of roller,outer ring and inner ring.And the acceleration vibration signal of four kinds of self-aligning ball bearings including normal,roller fault,outer ring fault and inner ring fault are collected by the BVT-5 bearing vibration measuring instrument.(2)The availability of the method of self-aligning ball bearing fault diagnosis based on empirical mode decomposition and manifold learning is confirmed by the study of vibration signal of four cases of self-aligning ball bearings.(3)The fault diagnosis experiments of self-aligning ball bearing under different axial loading with 3 different methods including time-frequency feature extraction and manifold learning,phase space reconstruction and manifold learning,empirical mode decomposition and manifold learning are done.The conclusion shows that the method of rolling bearing fault diagnosis based on empirical mode decomposition and manifold learning is the best,especially for the early weak fault diagnosis.