捷運系統鋼軌磨耗預測模式開發

Developing Rail Degradation Model for Metro System

林映均(Ying-Chun Lin)

指導教授 ： 賴勇成

國立臺灣大學/工學院/土木工程學研究所/碩士(2020年)

https://doi.org/10.6342/NTU202003997

摘要

鋼軌是軌道運輸中重要的資產，因隨著使用量及環境等因素受到磨耗，須適時地維護與更換。然而，鋼軌的更換也會涉及軌道維護計畫與鋼軌採購計畫，若能準確地預測鋼軌的磨耗，將可提升相關計畫的執行效率與成果。據文獻回顧，多數研究著重於整體軌道的劣化，少部分研究雖有探討鋼軌軌距的磨耗，然而決定鋼軌更換的指標並非視軌距的變化而決定。有鑑於此，本研究提出三階段的模式方法來探討與發展鋼軌踏面與側向磨耗預測模式，逐步提升其預測之準確與模式使用的便利性，並應用於都會捷運系統。首先由依同性質因素所分類的區段著手分析，以時間為變數，初步了解單一線段中踏面磨耗與側向磨耗的特性與趨勢，可提供現有營運路線鋼軌磨耗預測之資訊；第二階段則將所有區段依線段種類分類，藉由多元迴歸分析及類神經網路，發展出各線段種類的鋼軌磨耗預測模式，可提供未來路線鋼軌磨耗的參考依據；最後，第三階段發展全線段通用模式，彙整所有可能影響因素探究其對於鋼軌磨耗的變化。在第一階段模式中，踏面磨耗的模式有很好的預測能力，側向磨耗方面的預測不如預期，但透過換軌資料可以得知，僅少數區段為側向磨耗先達上限，因此第一階段的模式應能有效地提供既有營運路線鋼軌磨耗的預測。第二、三階段發展的模式經驗證後可得知，踏面磨耗之模式皆屬合理的預測範圍內，側向磨耗的預測相對而言不如踏面磨耗來的準確，由於第二、三階段的目的在於提供未來路線的規劃參考，透過第二、三階段模式中的影響因素可推估尚未營運路線的長期鋼軌需求。本研究之成果可供現有捷運路線之鋼軌維護與採購計畫參考，亦可供新建或規劃路線作為規劃參考與推估工具。

關鍵字

鋼軌 ； 磨耗預測 ； 捷運系統 ； 迴歸分析 ； 類神經網路

並列摘要

Rail is an important asset in rail transportation. It may deteriorate with impact of many different factors, such as traffic volume, environmental conditions, etc. The maintenance and renewal (M R) tasks are essential for the safety issue. In addition, the replacement of rail also affects rail maintenance schedules and rail procurement program. Therefore, once the rail degradation is accurately predicted, the schedules will be implemented more efficiently. In literature, most studies are focused on overall track degradation. Although part of research discusses the variation of gauge, gauge is not the only index to determine the replacement of rail. Consequently, this research proposes a three-stage method that yields a rail degradation model for wear of rail heads and width. In the first stage, rail is categorized as the section with the same property. Trend with time for each section is analyzed and compared to the prediction by operators. In the second stage, the sections are categorized to tangent, spiral and circular curves. The wear prediction models of multiple regression and artificial neural networks (ANNs) are developed for each category. In the third stage, a model is developed to identify impact factors affecting wear different segment categorizes. The results verify and show great performance for wear prediction of rail head. The models proposed by this study can be further applied as reference for M R plan of rail or rail procurement program.

並列關鍵字

Rail ； Wear Prediction ； Metro System ； Regression Analysis ； Artificial Neural Network

參考文獻

Ahac, M. and Lakušić, S. (2017). Track Gauge Degradation Modelling on Small Urban Rail Networks: Zagreb Tram System Case Study. Urban Transport Systems.

Alemazkoor, N., Ruppert, C. J. and Meidani, H. (2015). Track Geometry Analytics. In RAS Problem Solving Competition Report. Mumbai.

An, R., Sun, Q., Wang, F., Bai, W., Zhu, X. and Liu, R. (2018). Improved Railway Track Geometry Degradation Modeling for Tamping Cycle Prediction. Journal of Transportation Engineering, Part A: Systems. 144 (7).

Andrews, J., Prescott, D. and Rozières, F. (2014). A Stochastic Model for Railway Track Asset Management. Reliability Engineering and System Safety. 130, 76-84.

Bai, L., Liu, R., Sun, Q., Wang, F. and Xu, P. (2015), Markov-based model for the prediction of railway track irregularities. Proceedings of the Institution of Mechanical Engineers, Part F: Journal of Rail and Rapid Transit. 229 (2), 150–159.