Research on Rolling Bearing Fault Diagnosis Based on Distributed Fuzzy Broad Learning System Under Imbalanced Data

As a novel fast and effective intelligent algorithm, Broad Learning System (BLS) is popularly used in data classification task. However, the real data are often imbalanced, which makes BLS have limited performance in dealing with imbalanced data and cannot achieve good classification results. In view of this, a method based on Distributed Fuzzy Broad Learning System (DFBLS) under imbalanced data is proposed in this paper. The model assigns a fuzzy affiliation value to each training point, reducing the effect of imbalanced samples on the model, and DFBLS executes fuzzy function in parallel, thus improving computational efficiency and processing speed. Moreover, DFBLS learns the features in the input data by mapping features and enhancement features, which solves the deficiency of fuzzification that leads to the loss or confusion of data samples. Finally, it is validated by two rolling bearing experimental datasets, and the research results show that this technology has demonstrated excellent performance in various evaluation indexes and provides an efficient solution to the fault diagnosis problem in imbalanced data sets.