Research On Fault Diagnosis Method Of Planetary Gear Train With Variable Speed Based On Deep Learning

In the wind turbine,the planetary gear train is an important component of the speed increase gearbox.Under the normal working state,the working load and speed of the system change with time due to the influence of irregular wind field and frequent start and stop and other factors,and the vibration response has complex non-stationary impact characteristics.Traditional fault diagnosis methods mostly rely on expert knowledge and empirical knowledge,which have great limitations,while machine learning method can quickly and efficiently extract sample information without relying on expert knowledge and empirical knowledge.In machine learning,the traditional shallow though neural network fault diagnosis problems in processing equipment has a good diagnosis recognition rate,but it also has certain defects,under the condition of the sample size is bigger,the operation speed is very slow,diagnosis time will be very long,is not conducive to the engineering application,and deep learning can from the mass of the sample data to extract the sample characteristics of fast and efficient.Therefore,based on deep learning theory,fault diagnosis methods of planetary gear train with variable speed are studied in this paper.The main research contents are as follows:1.To solve the problem of whether the fault diagnosis recognition rates of planetary gear train fault sample sets in different domains are equal,fault sample sets in time domain,frequency domain and time frequency domain are made respectively.The characteristics of vibration signals of planetary gear train were extracted,and the simulation signals and measured signals were analyzed by time domain analysis,frequency domain analysis and time-frequency domain analysis,respectively.The complexity of measured signals was compared,and the fault identification and diagnosis could not be carried out without relying on expert knowledge and experience.The characteristic atlas of vibration signals of planetary gear train under constant speed and variable speed conditions are further studied.From the atlas analysis,it can be concluded that the vibration signals under variable speed conditions are more complex.2.In this paper,a variable speed planetary gear train based on convolution neural network fault diagnosis methods,writing font photo collections by player training and testing,verify the integrity of the network model and the feasibility of using convolution neural network model is set up respectively for images of the time domain waveform,frequency domain waveform photo collections and time-frequency domain waveform atlas is used to identify the fault diagnosis,By comparing the recognition rates of the three image sets,it can be concluded that the recognition rate of time-frequency waveform atlas is higher than that of time-domain waveform atlas and frequency-domain waveform atlas.The time-frequency atlas covering more information is more conducive to fault identification and diagnosis under the method proposed in this paper.3.Proposed a convolutional neural network method based on particle swarm optimization to identify and diagnose the faults of planetary gear train,established a convolutional neural network model based on particle swarm optimization,and verified the superiority of the optimized convolutional neural network by comparing the recognition rates before and after handwritten fonts through simulation.The validity of the method is further verified for the fault identification and diagnosis of planetary gear train.