Research On Rolling Bearing Fault Diagnosis Method Based On Deep Learning

Rolling bearings are one of the most important components of modern rotary machinery and equipment,playing an important role in the operation of machinery and are widely used in modern industrial equipment.According to statistical data,nearly half of all mechanical failures are caused by the rolling bearing fault,bearing failure makes normal and efficient operation of the machinery and equipment,security is reduced,shorten the service life.Timely prevention and diagnosis of rolling bearing faults will greatly reduce the failure of rotating machinery,reducing safety accidents and avoid economic loss.Therefore,research and development of rolling bearing fault diagnosis methods fault diagnosis has very important.In real industrial applications,the data generated by the change of working conditions and data acquisition environment are different,which often leads to poor performance of existing failure diagnostic methodology.This paper deals with the difficulty of fault diagnostics for rolling bearings in complex and variable working conditions and noisy environments,the application of convolution block attention mechanism and leakage threshold module combined with one-dimensional residual contraction neural network is explored in fault diagnosis network.First of all,the deep learning approach maps the rolling bearing vibrations into a non-linear spatial field,combining feature extraction in classifying,to automatically from the original vibration signal analysis and relevant information bearing running condition,for the small scale and displacement characteristics of the input signal,the depth of the learning method still can be done to properly identify,Adaptive extraction of rolling bearing fault features under variable working conditions from input signals.Secondly,on the basis of fully considering the fault features of rolling bearings,the convolutional block attention mechanism was designed to solve the problem of insufficient extraction of comprehensive features of rolling bearings fault diagnosis.Combined with the residual module,an improved residual neural network diagnosis method was designed.This method enhances the fault related features and ignores the irrelevant features to extract feature information effectively.Through experiments,the accuracy of fault diagnosis in different working conditions is compared.With the support of the attention mechanism,the diagnostic accuracy is improved.Finally,in order to improve feature learning ability and accurately diagnose rolling bearing faults under strong background noise,a new leaky threshold function of shrinkage function was proposed,and an improved leaky threshold module was designed based on residual block model to replace the soft threshold in residual shrinkage network.In the comparative experiment in this paper,this method combined with the improved deep residual contraction network diagnosis method can more effectively eliminate the influence brought by the noise of signal characteristics,thus improving the diagnosis accuracy and performing stably in various noise environments.