Design Of Bearing Fault Diagnosis System For Rotor Power System

In recent years,the development of automation and refinement of mechanical equipment has made the connection between all components of the equipment closer.While improving the efficiency of industrial production,it also increases the difficulty of fault diagnosis of parts in mechanical equipment to a certain extent.Bearings,as an important part of the rotor power system,play an irreplaceable role in the entire equipment,but when it operates at high temperature,humidity and serious load,it often produces failure problems such as wear and corrosion,and the probability of failure is much higher than that of other parts.At present,most of the methods of bearing fault diagnosis are regular maintenance and detection,which affects the work efficiency and has the possibility of missing detection.Therefore,the establishment of bearing fault diagnosis system and the timely detection and maintenance of bearing faults are of great significance to reduce economic losses and ensure the safety of users.This thesis takes the rotor dynamic system bearing as the research object,performs data processing and specific analysis on the vibration signal,realizes the detection of the bearing health,and uses the multi-body dynamic model to obtain the bearing vibration simulation data for verification.The field data lays the foundation,and finally a fault diagnosis system is established to provide a basis for the maintenance of the rotor power system.The main work of this thesis is as follows:(1)Aiming at the problem that the traditional empirical wavelet transform produces excessive division when dividing the frequency band,an improved empirical wavelet transform method based on kurtosis and correlation coefficient indexes is proposed.This method utilizes the characteristic that kurtosis is sensitive to shocks to merge and divide the frequency bands,so that the fault shocks are completely preserved in the same frequency band without being scattered.Then,the component with the correlation coefficient in line with the rules and the largest kurtosis value is selected for signal reconstruction.On the premise of reflecting the characteristics of the original signal,the component with the most fault components is selected to complete the extraction of fault features.According to the experimental results,the improved empirical wavelet transform reduces the number of redundant components and can extract fault features more accurately.(2)A fault diagnosis model based on improved empirical wavelet transform GG clustering algorithm is established.First,the multi-scale permutation entropy of the maximum kurtosis component after the decomposition of the improved empirical wavelet transform is calculated,and Then after dimensionality reduction,it is input into the GG clustering algorithm as a feature vector.The fault type of the input signal is judged according to the membership of the eigenvectors to the cluster centers of different fault types.In order to compare the clustering effect,the feature vector is input into FCM clustering and GK clustering algorithms for comparison,and the value of clustering evaluation index division coefficient and average fuzzy entropy are taken as the standard.The experimental results show that GG clustering algorithm has better discrimination and can improve the accuracy of classification when dealing with non-stationary signals such as bearing vibration signals.(3)By summarizing the functional requirements of users,the bearing fault diagnosis system is designed and developed based on QT framework.The system includes modules such as bearing vibration signal analysis,fault diagnosis,bearing data and diagnosis record management,which realizes visual operation and is convenient for inspectors.