Intelligent and adaptive asset management model for railway sections using the iPN method

Saleh, Ali; Remenyte-Prescott, Rasa; Prescott, Darren; Chiachío, Manuel

Abstract

The maintenance strategy in railway transportation is crucial in ensuring safety, availability, and reducing operating costs. However, finding the optimal maintenance plan that takes into account the complex relationships between railway assets can be a challenging task. To address this challenge, this study introduces an Intelligent Petri Net (iPN) model to effectively consider the maintenance and operation of railway sections with a focus on optimizing ballast maintenance. The iPN model merges Petri net (PN) with Reinforcement Learning (RL) to create a model that is able to simulate and learn at the same time. The model is able to use diverse information, including usage, degradation rates, maintenance effectiveness, fault probabilities, and maintenance time, to simulate and learn at the same time. By considering the interconnections between these factors, the model found that reducing unnecessary maintenance actions increases the age of railway sections and leads to higher net profits. The study also introduced a method to reduce computational effort by dividing the PN into subnets and another method to make learning faster by using multiple RL environments. In conclusion, the developed iPN model presents a promising solution for optimizing ballast maintenance within railway operation.

Citation

Saleh, A., Remenyte-Prescott, R., Prescott, D., & Chiachío, M. (2024). Intelligent and adaptive asset management model for railway sections using the iPN method. Reliability Engineering and System Safety, 241, Article 109687. https://doi.org/10.1016/j.ress.2023.109687

• Journal Article Type: Article
• Acceptance Date: Sep 23, 2023
• Online Publication Date: Oct 6, 2023
• Publication Date: 2024-01
• Deposit Date: Sep 27, 2023
• Publicly Available Date: Oct 7, 2024
• Journal: Reliability Engineering and System Safety
• Print ISSN: 0951-8320
• Electronic ISSN: 1879-0836
• Publisher: Elsevier
• Peer Reviewed: Peer Reviewed
• Volume: 241
• Article Number: 109687
• DOI: https://doi.org/10.1016/j.ress.2023.109687
• Keywords: Petri net; Reinforcement learning; Q-learning; railway; maintenance modelling; degradation models
• Public URL: https://nottingham-repository.worktribe.com/output/25377972
• Publisher URL: https://www.sciencedirect.com/science/article/abs/pii/S0951832023006014