Gear Fault Diagnosis Based On Transfer Learning And Wavelet Kernel Extreme Learning Machine

As a heavy-duty component in mechanical equipment,the gearbox is of great significance to the safe and reliable operation of the entire mechanical equipment.If a fault occurs,it will affect the operating efficiency of the entire mechanical system,and cause a major safety accident.Therefore,the gearbox fault diagnosis technology is carried out to accurately identify the fault mode and degree of the gearbox,which is of great significance for timely scheduling maintenance,ensuring the safe operation of the mechanical system and avoiding the occurrence of major accidents.In practical applications,identifying the severity of the fault can help the user understand the development trend of the equipment's operating status and arrange appropriate maintenance and maintenance strategies.With the rise of big data,cloud computing and artificial intelligence in recent years,gear fault diagnosis has developed in the direction of intelligence.The diagnosis and identification of gear fault types has been widely studied,but in practical applications,the severity of gear faults is still recognized.A challenge in the field of diagnosis.Because there are many non-linear factors in the high-speed operation of the gear,it is very important to have a monotonic relationship with the degree of failure.Secondly,there is not enough sample data of the failure degree in the experiment.It is necessary to construct a sufficient fault degree sample by means of the simulation model.Diagnose different fault levels,and finally use enough fault degree feature samples to train the diagnostic model to accurately identify the fault level.Firstly,based on the gear model of the Gear Research Center of the Free University of Brussels,this paper analyzes the gearbox components,establishes the fault gearbox model with the help of 3D modeling software SolidWorks,and combines the multi-body dynamics simulation software Adams to simulate the dynamics of the faulty gearbox.Secondly,after simulating the vibration signal and extracting the features,the ordered mutual information(RMI)and the standard mutual information(SMI)are used to select the feature set that has monotonic consistency to the degree of failure and robustness to different working conditions.Thirdly,after the filtered simulation feature set and the experimental feature set are migrated and learned,a expanded sample set is obtained.Finally,a novel wavelet fault learning machine(SAPSO-WK-ELM)intelligent fault diagnosis system based on adaptive inertia weight particle swarm optimization(PSO)is proposed to identify the fault degree of the expansion feature set.Through the simulation analysis of the model,the vibration signal of the gearbox under different fault degrees can be accurately obtained,and the feature sets under different working conditions can be extracted.After the migration learning,the wavelet kernel limit learning(WK-ELM)is input for classification and identification,and the original features are performed.Compared.The comparison results show that the feature set of different working conditions can effectively identify different fault levels after migration learning,and it is better than the original feature set classification,which proves the effectiveness of migration learning.The wavelet kernel learning machine(SAPSO-WK-ELM)with adaptive particle swarm optimization is compared with other optimized wavelet kernel learning machines.The results show that the proposed optimization algorithm can obtain higher classification accuracy and verify The effectiveness of SAPSO-WK-ELM in identifying the extent of gear failure.