Overheating Fault Diagnosis of Rail Transit Facilities Based on PINN

DOI: https://doi.org/10.62517/jes.202602104

Author(s)

Shuoxiang Ma

Affiliation(s)

Hebei University of Geosciences, Shijiazhuang, Hebei, China

Abstract

With the rapid development of rail transit system, overheating fault has become an important problem affecting the safety and reliability of rail transit facilities. Traditional fault diagnosis methods have the limitations of strong data dependence and low prediction accuracy. So this paper presents a method of overheating fault diagnosis based on physical information neural network (PINN). This method combines physical models and deep learning techniques to learn the thermal behavior rules of the device through the neural network to realize the early diagnosis of overheating failure of the device. The experimental results show that the proposed method can improve the diagnosis accuracy of overheating faults with less data. Specifically, the mean square error (MSE) of the model in the test set is 0.023, and the identification accuracy of the overheating fault reaches 98.6%, which is significantly improved compared with the traditional method. In addition, the PINN model can provide real-time early warning under complex working conditions to enhance the safety and reliability of the rail transit system.

Keywords

Physical Information Neural Network (PINN); Rail Transit; Fault Diagnosis; Overheat Fault; Deep Learning; Heat Conduction

References

[1] X. Zhang, L. Liu, and X. Xu, "Physics-Informed Neural Networks: A Review of Methodologies and Applications in Fault Diagnosis," IEEE Transactions on Industrial Informatics, vol.19, no.4, pp.2391-2402, Apr.2023. [2] Y.Wang, H.Li, and T.Yang, "Application of Deep Learning in Power Equipment Fault Detection," IEEE Access, vol.11, pp.4562-4573, 2023. [3] J.Zhang, M. Liu, and Y. Guo, "Intelligent Fault Diagnosis of Railway Power Systems Based on PINN and Sensor Data," International Journal of Electrical Power & Energy Systems, vol.137, pp.105-118, Oct.2024. [4] K. Kim, J. Lee, and S.Kim, "Data-Driven Approaches for Fault Diagnosis of Power Equipment," IEEE Transactions on Power Systems, vol. 38, no.2, pp.1135-1144, 2024. [5] H. Chang, L. Chen, and D. Zhang, "Improving Fault Prediction Efficiency Using PINN in Industrial Systems," Journal of Machine Learning in Industry, vol.22, no.6, pp.200-210, 2023. [6] P. Rao, X. Liu, and Z. Wu, "Hybrid Models for Temperature Fault Detection in Railway Power Systems," Energy Reports, vol.10, pp.1458-1469, 2025. [7] Y. Li, L. Zhao, and T. Zhang, "Integration of CNN and LSTM for Power Equipment Fault Diagnosis," Journal of Computational and Applied Mathematics, vol.98, pp.178-189, 2023. [8] Z. Liu, W. Yang, and D. Sun, "Predictive Maintenance of Electrical Equipment Using Physics-Informed Neural Networks," IEEE Transactions on Industrial Applications, vol.60, no.9, pp.1137-1146, Sept.2024. [9] T. Zheng, X. He, and R. Wu, "Application of PINN for Predicting Overheating Faults in Electric Motors," Journal of Electrical Engineering & Technology, vol.10, no.12, pp.2215-2223, 2024. [10] X. Chen, S. Zhou, and L. Li, "Temperature Fault Detection Using Physics-Informed Neural Networks for Industrial Systems," Computers in Industry, vol.118, pp.49-56, 2023. [11] J. Li, Y. Zhang, and W. Yang, "Improved Deep Learning Models for Predicting Electrical Equipment Failures," IEEE Transactions on Smart Grid, vol.14, no.3, pp.1532-1540, May 2025. [12] H. Wu, R. Zhang, and K. Shen, "Enhancing Fault Diagnosis with PINN for Complex Industrial Equipment," Industrial Engineering Journal, vol.39, pp.77-88, 2024. [13] P. Li, W. Yao, and J. Zheng, "An Enhanced Physics-Informed Neural Network for Fault Detection in Complex Systems," IEEE Transactions on Neural Networks and Learning Systems, vol.35, no.7, pp.2302-2313, July 2023. [14] G. Li, Z. Wang, and X. Zhang, "Integrating PINN and Transfer Learning for Fault Diagnosis in Industrial Applications," Journal of Artificial Intelligence in Engineering, vol.45, pp.125-136, 2023. [15] M. Liu, Y. Zhao, and L. Zhou, "A Hybrid Model Based on PINN for Predicting Faults in Electric Power Equipment," IEEE Transactions on Automation Science and Engineering, vol.22, no.5, pp.1345-1357, Oct.2025.