Research On The Degradation State Recognition Technology Of Rolling Bearings

As an important component of rotating machinery— Rolling bearing, its condition directly affects the rotor it supports and even the the run of the whole machine. The default of rolling bearing is a development process from mild to severe, accurately and timely identification the running state of rolling bearing can reasonable guidance the performance inspection and replacement maintenance plan,and it is of great importance to improve the reliability of the mechanical equipment operation.A hybrid method combining EEMD method which has been improved and Support Vector Machine(SVM) technique was proposed to identify the degradation state of rolling bearing. EEMD method can decompose signals to the different scale adaptively, and especially suitable for the unsteady and nonlinear signals.This method select the EEMD parameters according to the characteristics of signal and the added white noise of EEMD. It can avoids the problem of inaccurate decomposition results because of the traditional EEMD method selection parameters according to the experience of artificial. In order to reduce the interference of noise, through thresholding and reconstructing every intrinsic mode function which has been acquire d through EEMD.The envelop spectrum of the signal was analyzed,extracted the fault characteristics of the rolling bearings,and then we can acquire the failure location of rolling bearing.The key problem of assess the degradation state of rolling bearing accurately is how to establish proper model under limited state data. And SVM is a kind of machine learning algorithms which used to solve the classification and prediction problem of small sample, the algorithm is based on the VC dimension theory and structure risk minimum principle of statistical learning theory.Finding the best compromise in the complexity of the model and the learning ability according to the limited sample information, then can obtain better classification accuracy. Using the SVM and GA to extract the identification characteristics of degradation status of rolling bearing according to the different fault types on the basis of EEMD decompositio in this paper.Then the degradation state probability distribution and historical remnant life of rolling bearing are calculated to determine the optimal number of degradation state, which is employed to construct the SVM model for degradation state recognition. Finally,using the test data of different degradation states as the input vector of SVM which characteristics and parameters have been optimized by GA to identify andclassification the degradation status of rolling bearing.The analytical results for full lifetime datasets of a certain bearing demonstrate the validity of the method.