Data-Driven Fault Diagnosis and Localization in Multiphase Induction Drives

This study introduces a neural network-based approach for diagnosing faults in multiphase induction drives. The primary objective is to enhance the computational time, reliability, and operational efficiency of the drive systems. To achieve this, we have developed a neural network trained on a dataset generated with a five-phase field-oriented controlled drive. The dataset includes stator currents of different faulty scenarios along with the speed of the drive system. The simulation model using a five-phase inverter connected with a five-phase machine, recording stator currents under different fault conditions, is used for collecting the data. The main focus is on inverter faults in a drive system, specifically open-circuit (OC) and open-switch (OS) faults.