在台灣這個多山的地形，建造台灣路網，公路隧道已成為無可避免的設計。事故發生時比一般開放性路段更嚴重。對人員傷亡程度較為廣且威脅性大。國內目前既有之公路隧道大部分為三公里以下之較短隧道，對於未來可能出現的長隧道安全問題，更應該被社會大眾及政府所重視。 目前台灣在這方面有關的研究議題稀少，國外相關的長隧道議題中風險評估、風險分析、風險管理等技術成果斐然，但應用於台灣經驗是否可行，實在需要加以研究。 本研究旨在建立一套貨車通行長隧道風險評估程序，以適當的風險評估方法提升長公路隧道安全，並希望能夠協助決策者用以作為決策依據和提升決策品質。 根據上述主旨，本研究在考量運輸風險影響個體生命安全甚鉅之情況下，嘗試應用FN 曲線分析技術評估運輸系統安全。此外，?了解決觀測隧道缺乏事故資料可提供FN 曲線進行分析，本研究嘗試結合國外開放貨車通行的長隧道事故資料建立一套迴歸模式，用以預測長隧道通行貨車所可能產生的平均死亡人數，再運用卜瓦松分配轉化成適合FN 曲線進行分析的資料，以其解決資料不足的問題。 In order to construct the highway networks, the road tunnels become an inevitable design to accommodate mountainous ranges in Taiwan. However the damage is often much greater than on open road, if an accident occurs in a tunnel. The consequences could be extremely destructive and dangerous. Presently, most part of the road tunnels in Taiwan are less than 3 km. The security problem arisen in long-road tunnel will attract much attention in the future. At present, the researches in this field are still rare. In other countries, there’re plenty of reports concerned with risk assessment, risk analysis and risk management for long-road tunnels. Whether these experiences aboard can be applied in Taiwan are still worth further studies. Taking these concerns into account, this study aims to develop a procedure of risk assessment about truck passing through the long road tunnel and to promote the safety of long road tunnel with appropriate way of risk assessing. We expect we could assist policymaker to make the policy and promote the quality of policy. According to the substance above, this study tries to use the FN curves to assess the safety of transportation system. Besides, in order to solve the lack of the traffic accident data which establishes the foundation for FN curves analysis, this study build a regression model trying to connect the traffic accident data of other countries concerned with the long road tunnel opened for truck passing, to forecast the average mortality caused by the accidents of passing trucks in the long road tunnel. And use Poisson distribution to transform the data which can be used by FN curves to solve the problem with the lack of data.