Research On Early Fault Diagnosis Of Rolling Bearing Based On Vibration Analysis

As the core component of transmission mechanisms, runningstate of rolling bearing affects the accuracy, reliability and lifetime ofequipment directly. Due to factors such as structure characteristics andworking environment, rolling bearing is prone to having failure. Since therelationship between feature vector and pattern recognition of bearing fault iscomplex and nonlinear, how to extract effective fault information fromnon-stationary and nonlinear vibration signal is the key to quantitativediagnosis and prediction for early weak fault and complex fault of rollingbearing, which has great theoretical and realistic significance in mechanicalfault diagnosis. Main research content of this paper is as follows:Firstly, on the basis of comprehensive analysis of failure mechanism andfailure forms and causes of rolling bearing, main fault of rolling bearing wassimulated, data of normal and fault vibration signal was acquired through therolling bearing vibration detection and diagnosis test system, and vibrationcharacteristics were analyzed through time domain parameter statistics andtime-frequency domain processing of measured signal.Secondly, the paper studied early fault identification method of rollingbearing based on stochastic resonance, analyzed scale-transform cascadedeffect of mono-stable stochastic resonance model, verified the feasibility andpracticability of stochastic resonance in noise restrain of background, earlyfault feature extraction through simulation and measured data of normal andearly fault of the outer ring.Thirdly, the feature extraction method EEMD with noise reduction bystochastic resonance of the rolling bearing was proposed, advantage ofEEMD method in adaptive decomposition, resisting mode mixing wasdiscussed, combined with envelope demodulation technique, and the methodwas applied to feature extraction of early single and combined fault of rollingbearing successfully.Finally, on the basis of fault feature vector built by SR-EEMD, training and prediction of rolling bearing state sample set was realized by BP and RBFnetwork, and parameter optimization of RBF network was performed throughgenetic algorithm to improve the network performance.