Research On Feature Extraction Methods Of Rolling Bearing Fault Signals

The rolling bearing,as the most widely used component in the rotating machinery,usually malfunctions because of overloading,abrasion,uneven heating,insufficient lubrication and so on,which results in failures and even paralyses of rotating mechanical system,causing enormous economic and human capital losses.According to the statistics of rolling bearing,about 33.3% of rotating mechanical faults is caused by the rolling bearing faults.Therefore,the health status of rolling bearings will greatly affect the production efficiency and life cycle of the whole rotating mechanical system,which makes the fault diagnosis of rolling bearing very crucial.However,the feature extraction of rolling bearing's fault signals has always been a difficulty as well as a focus in the fault diagnosis of rolling bearing,in addition,the correctness of fault detection and fault classification of rolling bearing is directly affected by the effectiveness of fault features extraction.Therefore,this paper takes the feature extraction algorithms of rolling bearing's fault signals as its research focus.This paper takes the rolling bearing's vibration signals as the research objects,and the fault characteristics of vibration signals are analyzed.Based on this,an integrated fault feature extraction method for rolling bearing based on Multivariate Empirical Mode Decomposition(MEMD)and Permutation Entropy(PE)as well as a fault feature extraction method for rolling bearing based on Spectral Kurtosis(SK)and improved correlation dimension are put forward in this paper by using hybrid signal analysis methods.The main research work is as follows.(1)Some signal analysis methods,which are applied to process the vibration signals of rolling bearing currently,are studied.Which include time-frequency analysis method,such as MEMD and SK,as well as statistical analysis method,such as PE,correlation dimension and so on.These signal analysis methods mentioned above have been used to extract the fault features from the simulated and experimental fault signals of rolling bearing.And it has verified the applicability and effectiveness of these methods.(2)An integrated fault feature extraction method for rolling bearing based on MEMD and PE is put forward to solve the inference of noise.In the proposed method,an optimal band-pass filter determined by SK is firstly used to remove the interference frequencies which in different frequency band range of fault frequency components.Meanwhile the interference frequencies which in the similar frequency band range of fault frequency components is wiped off by the integrated method combining MEMD with PE.Then,the envelope spectrum of the reconstructed fault signal is calculated and the obvious fault frequencies are obtained.Finally,a quantization evaluated value,namely fault frequency ratio is introduced to evaluate the performance of the feature extraction methods.In addition,the performances of other existing fault feature methods and the improved method based on MEMD and PE are compared by the simulated and experimental fault signals of rolling bearing,and the result demonstrates that the improved fault feature extraction method based on MEMD and PE has advantage of accuracy and applicability comparison with these existing methods.(3)A fault feature extraction method for rolling bearing based on SK and improved correlation dimension is proposed.The proposed method takes the unique advantage of correlation dimension,which uses the quantization value to describe the fault features of rolling bearing of different working state.Compared with these fault feature extraction methods which use the feature frequencies to describe the fault feature of rolling bearing,the improved method reduces the influence of noise effectively by using the correlation dimension value as fault feature.Finally,the correctness and effectiveness of proposed method based on SK and improved correlation dimension are verified by rolling bearing's signals of different fault types and rolling bearing's signals of same fault type and different fault degrees.