Study On Gearbox Fault Diagnosis Based On Particle Swarm Optimization And System Performance

The research is originated by National Natural Science Fund Project"complex ggear fault diagnosis of early based on particle swarm optimization and filtering technique"(project NO: 50875247)and Natural Science Fund Project of Shanxi province"Study on Gearbox Fault Diagnosis based on Particle Swarm Optimization"(project NO: 2007011070).Gearbox as the most common dynamical transfer components of mechanism equipment,in the long-term working,bearing and gears are the inevitable some failures,because of manufacture error,impulsion load and work environment factors and fatigue,aging and other effects of the presence.For resolve the problem of vibration signal is difficult to be extrac feature and influence by undulation of input shaft speed signal and torque signal,the paper put forward the method of ARX time series model based on the input shaft speed and torque signal and output vibration signal,characterized by the corresponding ARX model under different condition of the gearbox system characteristics,analyzes the state of gearbox form the aspect of system characteristics,then diagnose.This paper carried out experimengal research work include:the speed signal pick out aberrant signal points,and smooth,for the vibration response signal is proceeding by subtracting mean of signal,filtering,resamping, pick out aberrant signal points and EMD decomposition;establish ARX model make use of input-output data sets;extract the regression coefficients of model and the model error as part of model features;analyse model of time-domain and frequency domain characteristics;the 16 statistical feature parameters of the amplitude frequency response curve used as model characteristics;all these constitute the complete feature set of gearbox fault diagnosis;discrete particle swarm optimization algorithm to establish experimental model,and programming with MATLAB,classification results of RBF neural network as the objective function,choose the best features of the state classification of a subset,making the highest rate of correct diagnosis.All the research production formed a set of gearbox diagnosis theory and method.The effective establiahed of the ARX model by the measured signal,extracted feature,achieved the pattern recognition of fault diagnosis in the optimization process,enhanced the precision for gearbox fault diagnosis and had important significance for drive the develop of gearbox diagnosis technology.