A Fault Diagnostic Method for Induction Motors Based on Feature Incremental Broad Learning and Singular Value Decomposition

The occurrence of fault in induction motors is dangerous in our daily life. It is significant to diagnose motor component faults accurately and quickly. In this paper, we propose an efficient and responsive motor fault diagnostic method based on Feature Incremental Broad Learning (FIBL) and Singular Value Decomposition (SVD). Firstly, we extract fault features from raw signals with Particle Swarm Optimization-Variation Model Decomposition, Sample Entropy and Time Domain Statistical Features. Secondly, these features are input into a broad learning system to train a network. Then we use FIBL to retrain the network if the diagnosis accuracy is unsatisfactory. Finally, SVD is used to further simplify the system structure to reduce diagnostic errors. In order to evaluate the performance of the diagnostic system, experiments are conducted. Experimental results show that with the proposed diagnostic method, motor component faults detection is quicker and more accurate.