Early artificial neural networks diagnosis using admittance model of a three-phase induction machine with inter-turn short-circuit fault

This paper proposes the application of an admittance model of a three-phase induction machine (IM) to the diagnosis of inter-turn short-circuit (ITSC) faults using artificial neural networks (ANN). The aim of this work is the early detection of this type of faults, essential to avoid damage to machines and ensure operational reliability. The first objective is the development of a simplified model based on impedances giving a new vision of the behavior of this type of machine in the presence of this type of fault. The second objective, based on this simplified admittance model, is the early diagnosis at zero rotation speed allowing to take a decision before launching a direct start or a vector speed control. The various proposed diagnostics use artificial neural networks to authorize or not the start of the machine and then estimate the position and the rate of this type of fault. Simulation results are presented in this article to show the validity of the model and the proposed diagnostics.