Data-driven lightning-related failure risk prediction of overhead contact lines based on Bayesian network with spatiotemporal fragility model. (March 2023)

• Record Type: Journal Article
• Authors: Wang, Jian; Gao, Shibin; Yu, Long; Zhang, Dongkai; Xie, Chenlin; Chen, Ke; Kou, Lei
• Is Part Of: Reliability engineering & system safety. Volume 231(2023)
• Publication Date: 2023-03
• Subjects: Overhead contact lines -- Lightning strike -- Failure risk prediction -- Bayesian network -- Spatiotemporal fragility curve; Reliability (Engineering) -- Periodicals; System safety -- Periodicals; Industrial safety -- Periodicals; Fiabilité -- Périodiques; Sécurité des systèmes -- Périodiques; Sécurité du travail -- Périodiques; 620.00452
• DOI: 10.1016/j.ress.2022.109016
• Languages: English
• ISSNs: 0951-8320

Abstract

Highlights: Propose a probabilistic lightning model, depicting uncertain features in occurrence and intensity of lightning strike. Propose a spatiotemporal fragility model, investigating spatiotemporal dependencies between LI and failure probability of OCLs. Develop a data-driven BN approach for predicting lightning-related failure risk of OCLs. Abstract: Lightning-related failures are of great concerns for the reliable performance of overhead contact lines (OCLs) of high-speed railway. Predicting lightning-related failure probability is valuable to capture the recurrence of OCL failures due to lightning strike and enable predictive maintenance decision-making. In this paper, a data-driven Bayesian network (BN) approach with spatiotemporal fragility model is developed to investigate the dependencies between lightning strike and OCL failures, and predict lightning-related failure risk of OCLs. It consists of three critical components, (1) a probabilistic lightning model that integrates multiple key lightning parameters is proposed to capture the uncertainty in the occurrence and intensity of lightning strike; (2) a spatiotemporal fragility model of OCL corridor is presented to examine the impacts of lightning strike on OCL failure probability; (3) furthermore, the Bayesian network is embedded with above-mentioned two models to predict lightning-related failure risk of OCLs, improving its robustness. Compared with other advanced prediction methods, the proposed approach achieves better prediction performance with high accuracy over imbalanced dataset. In addition, it can still work acceptably on noisy lightning data with a signal-to-noise ratio of 15dB or higher.