Research On Vibration Characteristics Analysis And Weak Fault Diagnosis Method Of Rolling Bearings

Rolling bearings are the key components in the operation of mechanical equipment.Their operating environment is complex,the application range is wide,and they are also the part that is more prone to failure.Once it fails,that will affect the normal operation of other components and even paralyze the entire system.Therefore,it is very important to monitor the condition of the bearings during operations.When the bearing fails,repairs or replaces it in time will improve the production efficiency of the system and the production quality of the entire production line.Especially for the diagnosis of early faults,it has become a research hotspot in the field of mechanical fault diagnosis.Early faults characteristics are generally weak,and they are often overwhelmed when the external noise is strong.Traditional signal processing methods are difficult to diagnose it.To this end,this paper takes rolling bearing as the research object,and studies the fault generation mechanism by establishing the excitation signal model under various operating conditions of the equipment.Then,the fault diagnosis related theory and modern signal processing methods are combined,and the research is conducted from the aspects of signal noise reduction,feature extraction and intelligent state classification.Each key issue was analyzed and a corresponding solution was proposed for the existing shortcomings.The signal model of bearing fault point and the transmission path between the vibration source and sensor is analyzed,and the influence of the range of each parameter of the model is examined by using the research method of vibration signal excitation response analysis.The excitation signal models of the bearing under diverse operating conditions are established to calculate the bearing fault quantitative analysis model.The correctness and effectiveness of the model are verified by the test bench data analysis.A method is named as the maximum correlation kurtosis deconvolution refinement narrowband signal denoising with parameter optimization is proposed.Based on the established bearing vibration signal response model,the periodicity,impact and modulation components of the fault signal are analyzed.Improved correlation kurtosis indexes are proposed to search the optimal filter length and filtering period to achieve the optimal filtering effect.The wavelet packet binary tree algorithm is defined to limit the fault signature signal to a narrow band,and the frequency domain correlation kurtosis index is used as the optimal frequency band selection basis to further achieve the filtering purpose.Compared with the customary signal analysis method,the effectiveness of the method in the noise reduction is verified.An algorithm for extracting features utilizing the optimal wavelet-scale cyclic frequency is proposed.Based on the above correlation kurtosis in characterizing the bearing impact fault characteristics.Taking the correlation kurtosis as the analysis target,the analysis range in the wavelet time-frequency diagram is limited to the optimal scale range with large correlation kurtosis value.The modulus or envelope value of the wavelet coefficient corresponding to the time axis is analyzed as a cyclic statistic,and the sensitive component related to the faulty feature can be extracted.The effectiveness of the algorithm in bearing faulty feature extraction is verified by the test bench data.Based on the distribution characteristics of the wavelet-scale spectrum,an early fault warning method based on wavelet sparse representation is proposed.In the wavelet time-frequency diagram,the wavelet coefficients of the wavelet-scale spectrum along the frequency axis are analyzed at each time point.By rearranging the wavelet-scale spectrum and performing synchronous averaging,then extracting the wavelet ridge line,calculating its amplitude and performing spectrum analysis.The local apex of the frequency domain amplitude of the bearing is extracted,and the proportion of the fault characteristic value of the bearing corresponding to the frequency value of the apex is taken as an early warning indicator to realize the early fault alarm of the bearing.The proposed algorithm is used to analyze the data of the planetary bearing test bench,and the effectiveness of the algorithm is verified.In addition,for the large data volume,single analysis fault data brings a large amount of computation.An intelligent state classification algorithm based on deep self-encoding to deal with the correlation kurtosis value in the frequency domain small-scale iteration period is proposed.According to the characteristics that bearing impact fault and noise interference suppression can be well characterized by the correlation kurtosis.By calculating the correlation kurtosis values of the original vibration signal of different frequency ranges within a small frequency range as new classification data.And then,inputting them into the deep sparse self-coding neural network for classification.The results show that the proposed algorithm can more accurately and quickly divide the test data of bearings under diverse operating conditions.The signal response model of bearing is established.The influence of numerous factors and parameters on the model is considered,and the signal characteristic analysis is reasonable.It breaks the limitation of the customary signal analysis method to weak signal analysis and realizes the external noise.The noise filtering and fault characteristics are enhanced when the interference is large and the initial fault features are weak,and the purpose of real-time fault alarm is achieved.For the large amount of experimental data and various types of cases,the fault intelligent classification research is conducted.Accurate and fast identification results are obtained.The research work in this paper provides theoretical support and technical support for fault diagnosis of rolling bearings based on vibration signal analysis.