Fault Diagnosis Of Rolling Bearing Based On Vector CEEMD

With the development of mechanization and automation,rotating machinery is playing a more and more important role.Nearly 30 years of theoretical research and practical application show that the emergence of fault diagnosis technology has opened up a new way to improve the reliability and safety of the system.Its emergence,rise and rapid development are the result of the alternation of practical application demand and the development of multidisciplinary theory.From the perspective of practical application requirements,with the continuous improvement of modern automation technology,the complexity of various engineering systems has greatly increased,and the reliability and safety of the system has become a key factor to ensure economic and social benefits,which has been highly valued by the engineering community.Therefore,in order to ensure the safe and stable operation of mechanical equipment,the development and application of state monitoring and fault diagnosis technology are particularly important.Advanced state monitoring and fault diagnosis technology can not only detect early faults and avoid accidents,but also fundamentally solve the problems of insufficient maintenance and overage in equipment maintenance.In order to ensure the completeness and reliability of the state information collection in the running state monitoring of rolling bearings,several sensors are usually arranged at the measuring points.Most of the collected multi-channel signals show nonlinear and non-stationary characteristics.All-vector spectrum technology can effectively fuse rotor dual-channel information,avoid information omission,and obtain complete signal spectrum.The phase space reconstruction technology can reconstruct the information of the related factors of the complex system and describe the whole picture of the system with limited data.In this paper,the overall average empirical mode decomposition method(CEEMD),permutation entropy and vector spectrum technology are combined in the fault feature extraction of rolling bearings,and the extracted fault features are applied to fault diagnosis with multi-classification limit learning machine.The main research work ofthis paper is as follows:1.According to the traditional empirical mode decomposition(EMD)method for rolling bearing vibration signal is decomposed prone to "modal aliasing phenomenon",and will measure the decomposed components different situation,using the general empirical mode decomposition(EEMD),through analog simulation analysis and experimental measurement signal analysis,proves the EEMD method to solve the shortcomings of the EMD method.2.Aiming at the reconstruction error problem caused by white noise caused by traditional empirical mode decomposition(EMD)and the deficiency of IMF component selection method,a feature extraction method based on overall average empirical mode decomposition(CEEMD)and permutation entropy was proposed.Firstly,multi-channel signals from rolling bearings with different degradation states were decomposed by CEEMD to obtain multiple IMF components.Then,the entropy values of each multi-scale IMF component were calculated.According to the entropy values,signals with high noise content and complex information were selected for noise reduction and signal reconstruction was completed.Finally,simulation analysis and example analysis are carried out to verify the effectiveness of this method.3.This paper proposes a mixed results can be output in the output of multiple classification extreme learning machine machine learning algorithm,the method in addition to keep on rolling bearing fault classification accuracy with high level,with the traditional support vector machine(SVM)classification method,compared to the classification of the time needed for greatly reduced,and reduces the time cost.Based on the above experimental analysis of fault classification of signals extracted by all-vector CEEMD and permutation entropy algorithm,the results show that this multi-classification ELM fault classification method based on all-vector CEEMD has good effectiveness and practicability in practical production and life.