De-noising Study Of Rolling Bearing Fault Vibration Signal Based On Empirical Mode Decomposition

Rolling bearings are often accompanied with high accident rates and high risk offailure in the industrial site of complex and changeable. It is very important to achieveefficient, fast, accurate identification and diagnosis of rolling bearing failures for theeffective protection of the production efficiency and personnel safety. However,measured vibration signal are annihilated by noise due to the complexity of theoperating environment, which causes great inconvenience to the fault featureextraction, especially in weak fault characteristics or failure early. To achieve faultvibration signal effective de-noising, and find out the de-noising method is suitablefor fault vibration signal de-noising of rolling bearing, which is of great significanceto improve the accuracy and efficiency of status monitoring and fault diagnosis of therolling bearing.Firstly, comparing vibration signals and noise of normal bearing, inner race fault,outer race fault and ball fault, which are measured from rolling bearing, and analyzingthe distribution of bearing vibrate noise and energy characteristics.Secondly, empirical mode decomposition (EMD) de-noising method based onautocorrelation and threshold is proposed. After analyzing the de-noising performanceof EMD, noise autocorrelation characteristics are used for identify noise modes, withthreshold to reconstruct signal to achieve de-noising, which can effectively identifythe noise dominant mode component, and as little as possible the loss of a usefulcomponent. The simulated signals are used to validate de-noising effect of thismethod.Thirdly, autocorrelation ensemble empirical mode decomposition (EEMD)de-noising method and adaptive EEMD de-noising method are proposed. EEMD,which overcomes the mode mixing problem, is taken, combining with autocorrelationsorting and threshold processing to achieve EEMD de-noising; analyzing thecharacteristics of noise energy in model component, adaptive generating threshold toachieve the processing of de-noising. Simulated signals are adopt to verify thede-noising performance of EEMD method.Finally, de-noising analysis of vibration signal of fault inner race and fault outerrace measured from rolling bearing are studied, compared with common de-noisingmethod. These proposed new methods can effectively identify bearing fault characteristic frequency and operating frequency, which proves better than thecommon method of common de-noising method.In this paper, the fault vibration signal characteristics of the noise of rollingbearing are analyzed, and effective de-noising of rolling bearing fault vibration signalusing improved EMD method are carried out, which provides effective signalpreprocessing method for condition monitoring and fault diagnosis of rolling bearing.