Research On Fault Feature Extraction And Diagnosis Of Rolling Bearing Based On QH-ITD

Rotating machinery is very important in modern industrial production and operation,and rolling bearing is one of the most important parts of rotating machinery and equipment,which is directly related to the operation of the entire production chain.Serious failure will even cause huge property losses and casualties.If the early failure of rolling bearing can be detected,it is of great significance for the factory to arrange maintenance plans,prepare spare parts and ensure the normal operation of mechanical equipment.Therefore,this paper focuses on early fault diagnosis of rolling bearings and conducts the following research from fault feature extraction and fault diagnosis classification:Firstly,feature extraction is carried out on the vibration signal of rolling damage fault processed by human.A Complementary Ensemble Empirical Modal Decomposition(CEEMD),Fast Independent Component Analysis(Fast ICA)and Teager-Kaiser Energy Operator(TKEO)were used to extract rolling bearing fault features.The method uses CEEMD algorithm to decompose the original fault vibration signal,and obtains a series of Mode components(Intrinsic Mode Function,IMF).Then the effective IMF was selected to complete the reconstruction of the observed signal according to the kurtosis criterion,and the remaining IMF was used to complete the reconstruction of the virtual noise channel signal.The reconstructed signal is denoised by Fast ICA method.TKEO is introduced to demodulate the signal after noise reduction.Finally,Fast Fourier Transform(FFT)is performed on the demodulated signal to analyze the spectrum characteristics of the transformed signal.The method is applied to the simulated experimental data of rolling bearing damage and failure at case western reserve university in the United States.The analysis results show that the method can reduce the influence of noise and highlight the impact characteristics of fault signals.Secondly,it is difficult to extract the characteristic information of the early weak fault signal of rolling bearing.Improved Intrinsic Time-scale Decomposition(Quartic Hermite interpolation-the Intrinsic Time-scale Decomposition,QH-ITD)and Adaptive Maximum Correlated Kurtosis Decomposition(AMCKD)of rolling bearing fault feature extraction method are proposed.This method uses QH-ITD algorithm to decompose the original rolling bearing fault signal,and then according to the kurtosis criterion and correlation index,the Proper Rotation(PR)signal is selected to reconstruct the signal after the new denoising,then processed by AMCKD algorithm,and finally processed by TKEO demodulation.The method was applied to the rolling bearing data of the university of Cincinnati in the whole life cycle to verify that the basic frequency and frequency doubling information of the early weak fault signal of the rolling bearing could be extracted.Finally,the classification and identification method of early weak faults of rolling bearings was studied,and an improved multi-scale approximate entropy and GG(GathGeva)clustering method for early fault diagnosis of rolling bearings based on QH-ITD was proposed.First use of the above mentioned QH-ITD decomposition algorithm for the rolling bearing fault signal decomposition algorithm,and then extract the improved multi-scale approximate entropy of the components in the eigenvector matrix,GG clustering algorithm is used to change the normalized eigenvector matrix of fault diagnosis classification,clustering center of the different fault types,finally,the fault types are classified and identified according to the Euclidean close degree in the approximate principle.This method is applied to the early fault data of the full life cycle accelerated life experiment of rolling bearings in Xi 'an Jiaotong university.The analysis results show that this method can effectively classify and identify the early weak faults of rolling bearings.