Fault Diagnosis Of Gearbox Vibration Signal Based On Stochastic Configuration Network

With the continuous advancement of China’s industrialization process,rotating machinery has been applied in more and more fields,and it also faces with more harsh working conditions and operating environment.As one of the basic components of rotating machinery,the condition of gearbox is directly related to the operation of the whole rotating machinery equipment.Once the gearbox breaks down,the result will be heavy economic losses in the lightest,and catastrophic in serious cases.In recent years,with the development of machine learning technology,the method based on neural network has been widely used in vibration signal fault diagnosis.Taking the gearbox and its main component bearings of rotating machinery as the research object,this paper studies the vibration signal fault diagnosis of gearbox based on stochastic configuration network.The main contents are as follow:(1)For the gearbox vibration signals collected in the laboratory,the statistical feature extraction method and the Gaussian model curve fit ting method are used to perform preprocessing and feature extraction work respectively,and statistical features and Gaussian model parametric features of gearbox vibration signals are extracted.(2)Stochastic Configuration Network is used for the classification task of gearbox fault diagnosis,and a feature stacking method is proposed,which combines statistical features and Gaussian model parametric features for gearbox fault diagnosis.The results show that the stacking features greatly impro ve the performance of stochastic configuration network classification model.(3)Several other common fault diagnosis classification algorithms are introduced.Statistical features,Gaussian model parametric features and stacking features are used for fault diagnosis to experimentally validate the comparative impact of both feature extraction and classification algorithm on prediction performance,and the results have a certain reference value.(4)A method based on the sliding window and 3σ interval to jointly judge the vibration signal initial fault point of the the whole lifetime cycle data is proposed,then the exponential models are used to complete the prediction of the remaining useful life prediction for rolling bearing.The result shows that the method proposed in this study can be explained better.