Health assessment of a brushless direct current motor stator using a physics-informed long short-term memory network

Abstract

The brushless direct-current (BLDC) motor is widely regarded as a promising and versatile electronically controlled motor. Ensuring its reliable operation is crucial for maintaining production efficiency and safety. The stator is essential to the functioning of BLDC motors, and the non-invasive health assessment of the stator can minimise operational disruptions to the motor while ensuring reliability. In this study, we propose a physics-informed long short-term memory network for the non-invasive assessment of BLDC motor stator health conditions. The proposed framework combines physics-informed neural networks (PINNs) with long short-term memory networks by optimising the structure and loss functions of the PINNs to thereby leverage the complementary strengths of neural networks and physical models to enhance their predictive capabilities. A case study involving a BLDC motor system validated the feasibility and effectiveness of the proposed method and yielded precise and reliable results. These results indicate that integrating physical information into deep learning methods is a promising approach by which to assess BLDC motor stator health.

Keywords

• Physics-informed neural network
• BLDC motor
• Health assessment
• Electromagnetic simulation
• Stator fault