Research On Fault Diagnosis Method Of Rolling Bearing Under Varying Loads Based On Broad Transfer Learning Network

As an important part of mechanical equipment,rolling bearing is very applicable in aviation,transportation,mechanical production and other fields.Effective fault diagnosis of bearing is helpful to prevent equipment accidents.Because rolling bearings are deeply buried in various machinery in actual work,it is difficult to diagnose their faults directly.In addition,the location of the bearing is complex and demanding,and the workload often changes,making fault diagnosis become a hard task.So,the research on the fault diagnosis method of rolling bearings under varying loads can realize intelligent multi-state fault diagnosis,which will have a far-reaching impact.Taking the broad learning network and transfer learning network as the core technology,the paper establishes a multi-state fault diagnosis model of rolling bearing to complete the fault diagnosis task under varying loads.To solve the issues with high time cost and low efficiency of rolling bearing fault diagnosis by deep neural network,an improved broad learning network is proposed.This is proposed a diagnosis method in bearings under varying loads based on the network.This method uses frequency domain data as input,and uses an improved broad learning network to extract broad features to construct a sample set of broad features.At the same time,the chicken swarm optimization selects the appropriate parameters of the network,reduce the influence of human factors,and improve network performance.The experimental results show that the improved broad learning network has fast training time and shows good fault diagnosis accuracy.To solve the problems with large divergences in data distribution of rolling bearings under varying loads,scarce vibration data with labeled information,and unbalanced distributions of multiple-state data,a new broad transfer learning network is proposed.This is proposed alongside a method in bearing diagnosis based on the network.This broad learning system is used to extract data features enabling the construction of feature sample sets.An unsupervised balanced distribution adaptation method in transfer learning is adopted to reduce this divergence in data distribution.Moreover,the chicken swarm optimization selects the appropriate parameters,and a network model is established.Finally,broad transfer learning network is applied in the intelligent fault diagnosis of rolling bearings under varying loads.The experimental results show that the broad transfer learning method is high effective and accurate.