Research On Health Status Assessment Method Of Rolling Bearings

Because rolling bearing is an important part of rotating machinery and the main source of failure,traditional performance indicators such as effective value and kurtosis can not fully reflect the degradation of the bearing at all stages,so this paper takes the rolling bearing as the research object,and conducts some research on the division,identification and quantitative performance evaluation of the bearing health status,mainly including the extraction and optimization of bearing characteristics,the division and identification of bearing status,and the construction of quantitative performance indexes of the bearing.The specific work content is as follows:(1)In order to mine the characteristic information of vibration signals to the maximum extent,13 time domain features are extracted,including 7 dimensional features,6 dimensionless features and 5 frequency domain statistical features.EMD decomposition,Hilbert transform and Fourier transform are used to extract the 1,2 and 3 multiples of bearing inner ring,outer ring,rolling body and cage on the envelope spectrum.Wavelet entropy feature is extracted by wavelet packet decomposition,and sorting entropy reflecting the difference of sorting modes of time domain signals is also extracted.The high-dimensional feature matrix is obtained,and 20 features that can reflect the degradation of bearings are optimized by fusion weight method.(2)Aiming at the subjectivity of artificially dividing the bearing health state,the first principal component obtained by PCA dimension reduction method is adaptively unsupervised clustered based on GG clustering algorithm to realize the state division of the whole-life bearing data,and the division result is evaluated: then HMM model is used to identify the data of several states(3)The performance index of rolling bearings based on logarithmic likelihood probability is constructed by using HMM.At the same time,the performance index of rolling bearings based on minimum quantization error is constructed according to SOM network theory.the construction method of HMIM-SOM model is proposed,and the output of the model is used to obtain a new rolling bearing performance index,which integrates the advantages of the two models.Finally,the index is evaluated from monotonicity,trend and robustness.(4)In order to prove the applicability of HMIM-SOM fusion model,this paper selects the rolling bearing fatigue life test data of Cincinnati Intelligent Maintenance Center,Hangzhou bearing test life data and 8 groups of PHM2012 challenge bearing data to verify the method,which fully verifies the effectiveness of HMM-SOM fusion model for the construction of rolling bearing performance indicators.