A data-driven Model for Fault Diagnosis of Induction Motor for Electric Powertrain

Abstract

The interest in electric traction has reached a very high level in recent decades, however, to dominate the market, many research efforts are still devoted to this purpose especially on the traction motors, the best way to preserve the integrity of electric motors in electric vehicles is to provide an on-board diagnostic and prognostic tools to ensure the availability. Since induction machine is among the must use motors in electric traction systems. This study presents a technique based on an artificial intelligence approach for the diagnostic and detection of broken rotor under random load. A dataset of a healthy motor and a malfunctioning motor with broken rotor bars were used to get the transient current and voltage signals during the motor starting and steady state. The current data in time domain properties are retrieved then converted using a Fast Fourier Transform to the frequency domain, proceeding to a preprocessing of the converted data, then opting for a supervised machine learning approach to develop a diagnostic model to evaluate whether the motor’s operation is normal or abnormal.