A hierarchical method based on improved deep forest and case-based reasoning for railway turnout fault diagnosis. (September 2021)

• Record Type: Journal Article
• Authors: Zhang, Yao; Xu, Tianhua; Chen, Cong; Wang, Guang; Zhang, Zhizhe; Xiao, Tian
• Is Part Of: Engineering failure analysis. Volume 127(2021)
• Publication Date: 2021-09
• Subjects: High-speed railway -- Railway turnout system -- Hierarchical fault diagnosis -- Deep forest -- Case-based reasoning; System failures (Engineering) -- Periodicals; Fracture mechanics -- Periodicals; Reliability (Engineering) -- Periodicals; Pannes -- Périodiques; Rupture, Mécanique de la -- Périodiques; Fiabilité -- Périodiques; Fracture mechanics; Reliability (Engineering); System failures (Engineering); Periodicals; Electronic journals; 620.112
• DOI: 10.1016/j.engfailanal.2021.105446
• Languages: English
• ISSNs: 1350-6307

Abstract

Highlights: A hierarchical fault diagnosis method is proposed for railway turnout. An improved deep forest which embedded priori knowledge is proposed. Case-based Reasoning is utilized to achieve detailed and comprehensive diagnosis. Abstract: Railway turnout system (RTS), which is widely laid along the railway tracks, is one of the most crucial devices in the whole railway infrastructures. Unpredictable outdoor weather, complex site conditions, and mechanical wear make the RTS the most vulnerable asset. Scientific and practical fault diagnosis is of great significance for reducing railway failures, improving operation efficiency, and ensuring transportation safety. In this paper, a hierarchical fault diagnosis method is proposed to realize accurate fault diagnosis of railway turnouts and improve the reliability and safety of railway systems. At the first level, an improved deep forest that embeds priori knowledge is adopted to distinguish the first-class fault sets. The embedded priori knowledge can help to outperform the traditional gcForest with fewer features produced at each level of the cascade forest. At the second level, a Case-based Reasoning component is utilized to classify the sub-class fault sets, which are difficult to be distinguished by the first level fault diagnosis due to their similar features. The hierarchical fault diagnosis method has been validated by using a real-world railway maintenance dataset. Extensive experiments show that the proposed method achieves superior performance compared to the state-of-the-art approaches in diagnostic precision under the constraint of limited data.