Study On Gear Fault Diagnosis Based On Improved BP Neural Network

Gear transmission is the most important part of mechanical equipment,its working condition affects the efficiency of the whole equipment directly,therefore,in order to improve the efficiency of the equipment,it is of great significance to improve the efficiency of the whole machine.The fault diagnosis of gears is the effective identification of the fault modes.In this paper,four kinds of gears are studied,which are the normal gear state,the tooth root crack fault state,the tooth wear fault state and the fault state,and use the combination of theoretical and experimental analysis,to collecte and analyze the vibration signals and acoustic emission signals of the gear under four conditions and study fault diagnosis of gear.In this paper,the dynamic model is established based on the theory of gear transmission,and select the signals which are suitable for gear fault diagnosis and the vibration signal and acoustic emission signal as the research object,BP neural network for fault identification method to establish the overall structure of the paper,last introduce the basic theory.Secondly,the analysis methods of vibration signal and acoustic emission signal and signal processing method are studied,to explore the problems of weak signal feature and large disturbance noise preprocess the signal by the method of wavelet denoising and wavelet packet decomposition,then the fault feature signal is extracted by means of signal expansion.The fault simulation experiment is carried out on the transmission gear of the JZQ 250 type reducer,and design and build the signal acquisition experiment platform.The vibration displacement signal of the vibration signal,the vibration acceleration signal,the vibration velocity signal and the acoustic emission signal are extracted by the circuit control of the control cabinet.Through the time domain,frequency domain and time-frequency domain,analyze the fault characteristics of each signal,and the spectrum energy of effective signal is compared and analyzed.According to the characteristics of the energy of each frequency band,the characteristic sample library is established.The fuzzy membership function of four kinds of working conditions is established based on the characteristic sample database to determined the data of the sample values in the sample database precisely.The vibration displacement amplitude,the amplitude of vibration acceleration,the amplitude of acoustic emission energy and the peak value of acoustic emission signal are taken as the characteristic parameters,Using MATLAB software to design and create BP neural network model to Complete gear fault diagnosis.The training samples are substituted into the model to train the optimization of BP neural network model based on PSO algorithm,and the experimental results are compared with the test results:The fault diagnosis data of acoustic emission signal and vibration signal can be used to diagnose the fault diagnosis accuracy rate reached 75%,the accuracy of BP neural network optimized by PSO algorithm is 97.5%.and the improved model of PSO algorithm can correct the relative error of the original 0.12 to 0.015,it is closer to the expected value of the system and comparing with the ordinary BP neural network model,the accuracy is improved by 87.5%.It can be proved that this method has a good application prospect and practical value for gear fault diagnosis...