Research On The Methods And Applications For Incipient Fault Feature Extraction Of Rolling Bearing

The high-grade CNC equipment’s functions increasingly rich and complex, andonce a certain key component appears weak fault at running time, which may have animpact on the whole equipment’s safety operation. Therefore, study of the complexmechanical equipment’s running state monitoring and fault diagnosis theory andapplications to ensure its safe, reliable, continuous operation.Rolling bearing is one of the key components with high failure rate. Therefore,this article is researching on the fault feature extraction methods of rolling bearing.The empirical mode decomposition method can effectively deal with thenon-stationary and nonlinear dynamic signal, but it appears mode mixing with theinterference of noise and shock pulse signal. Therefore, the introduction of noiseassisted ensemble empirical mode decomposition (EEMD) treatment for bearing’sfault dynamic signal, separating and extracting signals which contained with the faultsignal and the noise signal.According to the noise reduction characteristics of EEMD, researching on therolling bearing fault diagnosis using adaptive envelope analysis method based onEEMD de-noising and the spectral kurtosis theory. Through the EEMD method for theoriginal dynamic signal decomposition, selecting the useful mode components basedon kurtosis value index and correlation analysis and reconstruction, and thencombined with spectral kurtosis which in advantage on transient signal detection toadaptively determine optimum band-pass filter parameters of square envelope analysisfor de-noised signal. Application the envelope analysis techniques to identify incipientfault feature frequency, and using the simulation fault data and bearing faultexperiments to verify the validity of this method.Further to the effective anti-alias decomposition and feature extraction of EEMDmethod. The EEMD noise reduction effect is not ideal under stronger noiseinterference, then researching on the weak feature extraction method based oncascaded-bistable stochastic resonance pretreatment and EEMD decomposition fornon-stationary dynamic signal. Decomposition and extraction the low-frequency faultfeature component of the stochastic resonance output signal contains, which willreduce the decomposition layers and enhancement the diagnosis effect, finally through the simulation signal and experimental data of rolling bearing to verify its validity ofthis method.Research and development of the machine tool’s key components’dynamic signalmonitoring and diagnosis system based on the internet on the basic of the abovetheoretical research, adopts to the modularized design idea using Visual C++andMatlab mixed programming technology to realize the system construction andalgorithm integration. Exposition and realization of each module function from dataacquisition layer, information transmission layer and remote monitoring and diagnosisanalysis layer, finally, verification of the accuracy of this system through the CNCsystem’s screw bearing loosening experiment and realization of the machine tooloperation state’s dynamic signal real time display, monitoring, remote diagnosis andmaintenance.