High-fidelity multiphysics model of a permanent magnet synchronous motor for fault data generation

Abstract

This study proposes a high-fidelity multiphysics model of a permanent magnet synchronous motor (PMSM) for generating fault data. The proposed model helps to overcome the challenges encountered in real-world applications of data-driven prognostics and health management, where such data is often limited. To achieve this goal, i.e., generating high fidelity vibration data at a fault condition of a PMSM, the proposed model efficiently incorporates three multiphysics characteristics of a PMSM. First, a simple yet effective two-dimensional electromagnetic model is proposed to calculate electromagnetic force, which consistently induces vibrations in the PMSM. Next, the contact force within the bearing is obtained by solving analytical equations based on a rolling element bearing theory, enabling the representation of the force variation at the spall fault on the outer race. Finally, an effective modeling method for the three-dimensional PMSM structure is introduced, which is capable of generating vibration response on its surface when the Maxwell stress tensor and bearing contact force are imposed to the PMSM. The physical properties of the proposed model are estimated through extensive experiments and modal tests. The vibration data generated by the proposed model is also validated through vibration experiments, confirming the high-fidelity nature of the proposed model. The measured and generated data are also used for spall size estimation in the consideration of PMSM dynamic characteristics to demonstrate the effectiveness of the proposed model. The proposed model not only overcomes the limitation of data-driven diagnosis methods, which often lack sufficient data, but also provides a design enabling solution. As a result, the proposed high-fidelity model makes a significant contribution to the field of data-driven fault diagnosis for electric motors, offering reliable and predictive insights into bearing failures.