Physics-Information-Driven Intelligent Fault Diagnosis for Rotating Machinery

Rotating machinery, including critical components such as gears, bearings, and planetary systems, serves as the backbone of modern industrial equipment, from wind turbines and aerospace actuators to high-speed railways. The reliable operation of these transmission systems is paramount for safety and economic efficiency. Over the past decades, research has extensively focused on high-fidelity dynamics modeling to understand complex vibration behaviors, as well as on signal processing techniques for fault detection. However, traditional model-based methods often struggle with unmodeled complexities, while purely data-driven intelligent diagnostics face challenges of poor interpretability and limited generalization under varying operating conditions. This has created a critical gap in achieving truly reliable and explainable fault diagnosis. Consequently, there is a growing impetus to bridge the gap between physical principles and artificial intelligence, paving the way for a new generation of hybrid methodologies.Despite significant progress in both dynamics modeling and intelligent fault diagnosis, a critical disconnect remains. High-fidelity physical models often fail to generalize to real-world, variable operating conditions due to unmodeled complexities and uncertainties. Conversely, purely data-driven AI methods, while powerful in pattern recognition, typically operate as "black boxes" lacking physical interpretability, requiring massive labeled datasets and struggling with extrapolation. This Research Topic aims to bridge this gap by fostering the development of hybrid methodologies that synergistically integrate physical knowledge with data-driven intelligence. Recent advances in physics-informed machine learning, symbolic regression, and digital twin technologies now make this integration achievable. By embedding governing equations (e.g., dynamics of gear meshing or bearing raceways) into neural network architectures, or by using physical priors to guide feature extraction, we can achieve diagnostics that are both accurate and interpretable. The goal is to solicit contributions that leverage these advances to address challenges such as cross-condition generalization, early fault detection under strong noise, and remaining useful life prediction. We seek to establish a new paradigm where models not only learn from data but also "understand" the underlying physics of rotating machinery, leading to more trustworthy and robust condition monitoring systems.This Research Topic welcomes contributions on hybrid methodologies that integrate physical principles with data-driven techniques for rotating machinery condition monitoring. Themes include, but are not limited to:1. Physics-Informed Machine Learning: Neural networks incorporating governing equations (e.g., gear/bearing dynamics) as physical constraints.2. Hybrid Digital Twins: Integration of sensor data with dynamic models for enhanced diagnostics and prognostics.3. Interpretable Fault Diagnosis: Methods revealing physical meaning behind AI diagnostic decisions.4. Dynamics Modeling of Transmission Components: High-fidelity models for gears, bearings, planetary roller screws, and gear trains considering time-varying parameters, nonlinear contacts, and multi-source excitations.5. Dynamics-Guided Signal Processing: Vibration analysis techniques informed by system dynamic characteristics.6. Physics-Aware Life Prediction: Hybrid models for RUL estimation considering dynamic loads and fault evolution.We welcome Original Research and Case Studies with experimental validation or industrial applications. Topic editors: Zhifeng Shi, Xingkai YANG, Zheng Cao, Liming Wang, Xinbin Li Source: https://www.frontiersin.org/research-topics/80857/physics-information-driven-intelligent-fault-diagnosis-for-rotating-machinery