Model-based fault diagnosis and monitoring of induction machine bearing fault

• Published in: Mechanical systems and signal processing Vol. 238; no. 113245; p. 113245
• Main Authors: Kavugho, S. Moloverya, Ngandu Kalala, G., Rasolofondraibe, L., Kilundu Y'Ebondo, B., Chiementin, X.
• Format: Journal Article
• Language: English
• Published: Elsevier Ltd 01.09.2025; Elsevier
• Subjects: Bearing fault; Computer Science; Engineering Sciences; Induction motor; Mechanical engineering; Mechanics; Modelling; Physics; Signal and Image Processing; Stator current; Induction motor; Modelling; Stator current; Bearing fault
• ISSN: 0888-3270; 1096-1216
• DOI: 10.1016/j.ymssp.2025.113245

This paper proposes a novel approach for the diagnosis of bearing faults in the presence of coexisting electrical anomalies. A simplified dq-model is developed to simulate localized spalling on the outer race, with torque disturbances explicitly incorporated to represent the mechanical defect. Unlike finite element and magnetic models, which are often too computationally intensive for real-time industrial use, the proposed approach captures bearing fault transmission to the stator current with reduced computational cost. This trade-off between accuracy and efficiency allows the model to be extended to multifault scenarios, including broken rotor bars and inter-turn short circuits. The system equations are numerically integrated using the fourth-order Runge–Kutta method to ensure both stability and computational precision. To address fault detection under multifault conditions, a diagnostic strategy based on stator current analysis is introduced. The method combines frequency-domain analysis, used to identify characteristic fault frequencies, with time-domain processing based on newly proposed features specifically designed for bearing fault identification. These features, not previously reported in the literature, provide a robust basis for the development of data-driven classification algorithms capable of distinguishing concurrent faults. Experimental validation is performed on a dynamic test bench equipped with an induction motor. The results demonstrate the feasibility of the approach for real-time condition monitoring, although further validation is required to fully assess its robustness and generalization across industrial operating conditions.