以數據驅動貝葉斯網絡建立鐵道行車風險評估框架

Abstract

在普悠瑪事故和太魯閣事故接連發生之後，鐵路行車安全的風險管理成為大眾關注的焦點。面對鐵道行車的不確定性，本研究提出了一個結合人因分析與分類系統和貝葉斯網路的六階段步驟，建構鐵道行車風險評估框架，全面檢視風險並提前辨識潛在風險因子。首先，透過事故分析識別和定義潛在風險事件，並進行因果分析，確定導致風險事件的原因。接著，基於過往行車數據構建貝葉斯網路，量化各風險因子之間的關係和影響，並利用貝葉斯推斷計算風險事件的發生機率，進行事故後果分析，評估每個風險事件的潛在影響。在案例分析中，以實際的路線和行車資訊進行路段分析，結果顯示瑞芳到雙溪的風險相對較高，這表明除了人為失誤外，平交道和施工區域造成的風險也必須重視。此結果也說明了本研究提出的框架，能根據每班列車的行車資訊，提前預測路段風險，若能在行車前通知司機員各路段風險差異，使其提前準備，達到事前預防的效果，更可提升鐵道行車的整體安全性

Following the consecutive occurrences of the Puyuma and Taroko accidents, public attention has been drawn to the importance of risk management in train operations. To address operational uncertainties, this research proposes a six-phase approach that integrates the Human Factors Analysis and Classification System with Bayesian networks to develop a risk assessment framework. Initially, potential risk events are identified through accident data, followed by causal analysis. A Bayesian network, constructed from historical data, quantifies relationships and impacts of various risk factors. Bayesian inference calculates the probabilities of risk events and assesses their potential impacts. A case study applying this framework to real routes reveals higher risks between Ruifang and Shuangxi, emphasizing the risk in level crossings and construction areas. The results also demonstrate that the framework proposed in this research can predict section risks in advance based on the train's operational information and all the latest details about the route it travels through. By informing drivers of the varying risks of different sections before their journeys, they can be better prepared, achieving proactive prevention. This approach can possibly enhances the overall safety of train operations.