Research On Rolling Bearing Fault Diagnosis Based On VMD And IFWA-SVM

The operation and maintenance of mechanical equipment is the basis for ensuring the normal operation of equipment.Rolling bearings are used in almost all large-scale machinery,and their role in reducing energy loss is indispensable.Monitoring rolling bearings is not only conducive to equipment maintenance,but also conducive to safe production.Therefore,this thesis studies bearing status detection.This thesis analyzes the vibration characteristics,decomposes vibration signals,obtains the hidden information contained in vibration signals in different bearing states,and then identifies the characteristic information of different bearing states,so as to achieve accurate classification of bearings in different states.First of all,from the three aspects of signal processing,feature selection and recognition model,the current research situation is sorted out,and the research methods to be adopted in this thesis are determined.In addition,the vibration mechanism of rolling bearings is sorted out.Based on this,the bearing state is divided,and the change characteristics of vibration signals when damage occurs in different positions are analyzed.In terms of signal processing,a VMD algorithm that is not easy to generate modal overlap is used to realize signal decomposition,and the original signal is divided into multiple subsignals to reduce noise interference caused by the vibration of other equipment.In order to effectively quantify different bearing states,calculate the SE value of the decomposed signal,and select the representative time-domain indicators of the original signal as a supplement,TDI-VMD-SE feature data is constructed to pave the way for the data input of the identification model.Aiming at the problem that it is difficult to obtain a variety of state bearing data in the actual collection process of bearing data,an SVM model close to the classification of small samples is selected for fault identification,and a model based on improved fireworks algorithm optimization SVM is built.In view of the excessive randomness of the roulette selection of offspring in the fireworks algorithm,this thesis introduces the adaptability value in the original offspring selection process and combines it with the method of selecting offspring according to spark distance in the original algorithm,so as to reduce the randomness of offspring selection and improve the optimization efficiency of the algorithm.The proposed improved fireworks algorithm is used to search for the optimal parameters of SVM,so as to improve the diagnostic accuracy of SVM and provide a basis for efficient fault identification.Finally,through the status identification of rolling bearings with single variables such as different fault types,different defect sizes and different motor loads,the validity of the proposed model under given conditions is verified.Then,the impact of the control variables is comprehensively analyzed,and the feasibility of the model is verified by mixing fault identification from the multi-defect size and multi-fault state.In addition,the diagnostic accuracy under different eigenvalue selection,different algorithm optimization support vector machines and neural network models are compared,which proves the feasibility and advantages of the diagnostic methods proposed in this thesis.In this thesis,the TDI-VMD-SE feature selection method is combined with the IFWA-SVM model to verify the efficiency of this feature extraction method,and finally realize the effective diagnosis of rolling bearing faults,and solve the problems of low accuracy of single eigenvalue multi-fault identification and low calculation efficiency of excessive eigenvalues.And broaden the optimization method of SVM parameters,which is conducive to the safe production of the factory.