列車位置偵測邏輯之安全性與可靠度評估研究

Abstract

列車位置偵測系統是鐵路號誌控制的基礎，為了提升可靠度與安全性，近年來鐵路系統開始採用多組偵測系統同時偵測。考慮到採用不同設計邏輯同時使用多組偵測系統時，可能造成可靠度與安全性的變化，本研究先提出一有效率的分析平台。理想的偵測系統分析架構應考慮股道配置、偵測區間分佈、邏輯設計、通過列車特性（長度、開通進路）等，來決定系統的可靠度與安全性，現今的分析方式大多仰賴人工步驟與判斷，效率較低且較為主觀，因此，本研究的第一個目標是開發一個分析平台，可以自動化產生所有失效情境、計算失效仍然安全 (fail safe) 與會造成危險 (wrong-side failure) 的機率、最終得出該系統設計下的可靠度與安全性指標。除此之外，本研究亦提出一高可靠度與安全性之創新偵測邏輯設計—基於模擬資料之投票邏輯。此邏輯先模擬所有可能情境，產生模擬資料庫，接著當列車進入該站區後，此邏輯隨即開始比對當下列車佔用狀況，與所建立的模擬資料庫是否相符。當系統有可被偵測的錯誤發生時，此邏輯能參考資料庫及自動修正錯誤。由案例分析結果得知，所提出之分析平台能應用於實際之路段，產生之可靠度與安全性指標可用來選擇路段中最佳的偵測設計邏輯。除此之外，將第二部分創新邏輯納入分析平台時，可以發現此邏輯設計，無論是在複雜股道，或是在多樣的車種狀況下，基於模擬資料之投票邏輯都有相當高的可靠度與安全性，應用此邏輯設計能顯著提升偵測系統的可靠度與安全性。

A train detection system is the basis of signaling and control in railway transportation, and there is a trend to adopt multiple train detection systems simultaneously to improve reliability and safety. Considering the change of the reliability and safety performance when adopting different design and logic of multiple train detection systems, this research first proposes an efficient evaluation framework. The ideal evaluation process for train detection systems should be able to determine their reliability and safety by considering the track layout, detection block, design logic, and train characteristics (length, routing). Existing methods still require a few manual interactions and human judgment, which could be inefficient and subjective. Consequently, this research develops a automatic framework to generate all failure scenarios, calculate the probability of fail-safe conditions and wrong-side failures, and finally determine the reliability and safety improvement index of the corresponding logic. In addition, this research also proposes an innovative design logic specially for dual train detection systems, namely, simulated data-based voting (SDBV) logic, in order to improve reliability and safety. The process first generates and stores all possible train detection data in a database, and then executes the standard or special process to trace the passing trains and correct possible detection errors on the basis of the simulated data. Results of case studies demonstrate the applicability of the proposed framework to actual cases. The use of this framework can assist railways in identifying the appropriate design and logic of multiple train detection systems. As to the SDBV logic, it is proved to outperform other existing logics in terms of integrated performance in reliability and safety. Thus, applying this logic can substantially improve the reliability and safety of the dual train detection system.