結合故障樹和Petri Net的列車行車風險評估

Abstract

風險管理在鐵道系統中扮演關鍵角色，若缺乏完善機制，往往因忽略細節而導致重大事故。近年台鐵普悠瑪與太魯閣事故的發生，使社會大眾對鐵道安全風險管理更加關注。然而，鐵道營運係由人員、車輛與基礎設施構成，伴隨高度不確定性與複雜交互作用。為因應此挑戰，本研究提出一套整合Petri Net與故障樹分析的風險評估架構。首先透過人為因素分析與分類系統（HFACS）從歷史事故資料中識別潛在風險因子，並依其類別建構初步故障樹網路架構；考量傳統故障樹無法動態表達因子間的依存關係與非線性互動，本研究進一步將部分網路轉換為Petri Net，強化模型對實際風險因子的互動模式。接續以台鐵實際行車紀錄、路線環境資訊與設備可靠度資料為基礎，搭配事故傷亡與延誤，根據列車與環境差異變化，損失進行不同路段間的風險量化。結果顯示，在不同路段間的風險原因會因為不同的輸入而異，如平交道事故的影響會隨著路段是否有平交道有顯著的差異；而人為失誤機率的高低也會根據不同路段的環境差異有不同的起伏，在大部分情形仍然屬於事故主因。整體而言，本研究所提出之架構可有效根據每班列車行車條件預測風險差異，若能事前提供司機員警示，有助於提早因應與預防，有效提升整體鐵道行車安全。

Risk management is essential in railway systems, as inadequate measures can lead to severe accidents. Recent train accidents in Taiwan, like the Puyuma and Taroko incidents, have heightened concerns about railway safety. Due to the complexity and uncertainty of railway operations, this study proposes a risk assessment framework that combines Petri Nets (PNs) with Fault Tree Analysis (FTA). The framework starts by using the Human Factors Analysis and Classification System (HFACS) to identify risk factors from historical accident records, forming preliminary FTA structures. However, to better capture dynamic interdependencies, these FTA networks are converted into PNs. This approach integrates train operation records, operating environment factors, and equipment reliability data from Taiwan Railways, along with details on accidents and delays. We quantify potential losses across various sections to support risk assessment. The results show that risk sources differ across sections based on input conditions. For example, the impact of level crossing accidents depends on the presence of level crossings, while human error likelihood varies with operating environment characteristics. Nevertheless, human error remains the main cause of accidents in most cases. This framework enables proactive risk evaluation for individual train services, providing early warnings to drivers and enhancing overall railway safety.