A physics-guided neural network with embedded multi-wheel FRF for robust wheel polygonal wear diagnosis

Puntos clave

• Diagnosis of wheel polygonal wear achieved using a physics-guided neural network approach, enhancing detection reliability.
• The model integrates embedded multi-wheel frequency response functions (FRF) to improve diagnostic accuracy in complex scenarios.
• Assessment utilizes machine learning techniques to refine data-driven approaches for wear diagnosis in various operational contexts.
• Results indicate a potential for this method to enhance predictive maintenance practices in transportation, leading to cost and safety benefits.