Early detection and diagnosis of air-gap eccentricity faults based on Signal-Energy-Image conversion algorithm for the Deep learning in induction motor

This research introduces a new and effective Deep learning technique for diagnosing air gap eccentricity faults using a transfer learning approach. The study used an improved version of SqueezNet, which is enhanced with a novel attention mechanism module specifically designed for fault feature identification. Additionally, the paper introduces a Signal Energy-image transformation method based on Variational mode decomposition (VMD) and Hilbert transform technique to estimate the severity of the faults. Through extensive experimentation and performance evaluations, the proposed method demonstrates remarkable accuracy in fault diagnosis, only for a single training process