Intelligent Fault Diagnosis Of Bearing Based On Stacked Denoising Autoencoders

Rolling bearing is one of the most commonly used parts in rotating machinery.Its working condition is directly related to the performance of the whole unit and even the safety of the whole production line.Therefore,it is very important to study the fault diagnosis technology of rolling bearing.Generally,In the fault diagnosis of rolling bearing,the fault feature extraction,fault feature selection and state identification are needed.Fault feature extraction is the key to fault diagnosis,and its effect directly affects the diagnosis results.However,there is a strong dependence of the fault characteristics,so it is necessary to grasp a large number of signal processing methods and diagnostic experience,and increase the difficulty and uncertainty of the analysis.In order to achieve a better diagnosis effect,the pattern recognition model based on the fault feature is becoming more and more complex,and the application condition of the model is often neglected,which leads to the difficulty of feature learning.In this paper,the concept of deep learning is introduced.And the feasibility of stacked denosing autoencoders in the fault diagnosis of rolling bearing is studied,in order to solve the problem of feature learning difficulty and complex model building in traditional fault diagnosis.This paper probes into the feasibility of the stacked denosing autoencoders in the fault diagnosis of rolling bearings,and aims to provide a new method and technology for bearing fault diagnosis.The time domain vibration signal characteristic parameter is used as input of the stacked denosing autoencoders,discussed the different hidden layers combination on the recognition rate.The feasibility of using the stacked denosing autoencoders method to realize the intelligent identification of bearing faults are studied;Using the original vibration data to train the stacked denosing autoencoders fault diagnosis model,and discussed the influence of the load on the recognition result;The effects of the network layers,iterations,the proportion of training samples,and the Batchsize on the classification performance of the model are analyzed by comparison experiments;Experimental study was carried out,and the fault types and fault degree were simulated by using QPZZ II rotating machinery fault simulation experiment platform.The vibration signals are collected under the condition of idler and load,and the influence of the position of the measuring points on the identification results is studied,The effectiveness of the stacked denosing autoencoders fault diagnosis method is verified.