Network-combined broad learning and transfer learning: a new intelligent fault diagnosis method for rolling bearings

We propose a network combining broad learning with transfer learning for problems with large divergences in data distribution between source and target domains associated with rolling bearings under varying loads, scarce vibration data with labeled information, unbalanced distributions of multiple-state data, and low efficiency from model training. This is proposed alongside an intelligent method in rolling 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 method is introduced to optimize the parameters of the network, and a network model is established. Finally, a network combining broad learning with transfer learning is applied to the intelligent fault diagnosis of rolling bearings under varying loads. The experimental results verify the high effectiveness and accuracy of the proposed method.