Influence of Risk Parameters on Marine Accidents Using a Data-Driven Bayesian Network- a Case Study in Taiwan(I)

View Statistics Email Alert RSS Feed

Information

Details

Project title

Code/計畫編號

MOST109-2410-H019-016-MY2

Translated Name/計畫中文名

資料驅動式貝葉式信賴網路模型之建立及其在船難因素影響程度分析-以台灣各港口及其鄰近海域為例(1/2)

Project Coordinator/計畫主持人

Shuen-Tai Ung

Funding Organization/主管機關

National Science and Technology Council

Department/Unit

Department of Merchant Marine

Website

https://www.grb.gov.tw/search/planDetail?id=13531383

Year

2020

Start date/計畫起

01-08-2020

Expected Completion/計畫迄

31-07-2021

Bugetid/研究經費

964千元

ResearchField/研究領域

經濟學

海洋運輸是國際貿易重要一環，台灣港口及鄰近海域商船海難事故發生件數近年來逐步上升。本計畫將以貝葉式網路為基礎建構船舶航行風險評估模型，提供即時風險預測。除提供台灣海域現況分析外，藉由分析導致船難各變數之風險影響程度，可對特定船舶在進港前預測其可能發生的海難種類、嚴重程度及發生機率，研究成果可提供交通部參考。交通部航港局海難資料庫之變數將成為貝葉式網路船舶航行風險評估模型母節點與子節點，包括船齡、船旗國、船型、總噸位、海難種類以及嚴重程度，2014至2019年變數資訊將經過編碼與統計，成為各節點先驗與條件機率來源。為求風險評估模型完整性，風、浪與能見度亦納入考量，然因能見度資料缺乏，專家意見為該因素主要資訊來源，並以模糊集合與德菲爾法轉換成機率性質之信賴程度。此外，為提升子節點之條件機率與母節點狀況程度之邏輯一致性，本研究另建構條件機率發展模式，產生之風險結果更合理。風險評估模型將會實施合理性驗證，以確認實用性。研究成果亦將與國外相關文獻比較，探討與他國差異程度與原因。由於具備隨機針對特定船舶在其進港前實施即時情境分析之優點，本計畫之風險評估模型及研究成果將可成為交通部航行安全與海難預防重要參考。Frequency of the merchant ship accidents occurring in the Taiwanese ports and the surrounding waters has a gradual increase rising from 67 to 107 within the period between 2013 and 2016. A ship navigation risk assessment model is established in this study of which the objective is to provide real time risk predictions in a proactive manner. Apart from the current status risk analysis, the proposed model will be equipped with the capability of estimating the likely accident type, severity levels and probability of the accident occurrence for a specific ship prior to entering into a specific port or area. This is achieved by evaluating the risk influence by the parameters including Ship Age, Ship Type, Gross Tonnage, Flag and Weather Conditions. Bayesian Network (BN) mechanism is adopted in this study given its ability of manipulating multiple-state parameters, considering the interdependence between elements and handling uncertainty using probabilistic values. The statistical data collected from the Taiwan Ministry of Transportation and Communications (MOTC) will be the important references in this study.Parameters based on the MOTC database for each marine accident will be treated as the parent and child nodes in the constructed BN. These include Ship Age, Flag, Ship Type, Gross Tonnage, Accident Type and Accident Severity. The information of such knobs during the period between 2014 and 2019 is subsequently compiled and forms the basis for the prior and conditional probability calculations. Considering the integrity of the proposed risk-based BN, the meteorological factors are also contemplated, namely, Wind, Wave and Visibility. Due to the lack of information regarding Visibility, expert judgment will be the source for such a node. Fuzzy set and fuzzy Delphi methods are adopted for equipping such elements with a belief degree nature. Furthermore, a mapping mechanism capable of producing consistent conditional probability entries based on the state of each parent node is proposed. Such a feature increases the rationality degree of the conditional probabilities and therefore solves the obstacle of providing logical conditional probability outputs encountered by traditional BN studies.The BN model validation will be analyzed based on the principles of the Face, Content and Predictive validities. Rationality validation of the risk-based BN methodology will also be conducted. This is achieved by examining the effects of Accident Frequency, Accident Severity and Overall Navigation Risk given small alternations of the probability magnitude in the parent node parameters. In addition, the risk results of the Overall Navigation Risk based on Accident Frequency and Accident Severity for the current status in Taiwan will also be compared with those from relevant safety studies and the IMO reports for further validation. The methodology proposed and outcomes generated will be the useful references for the MOTC in terms of navigation risk prediction and accident prevention due to the feature of providing real time risk profiles for each ship prior to entering the Taiwanese territorial waters.

Keyword(s)

海難事故分析
貝葉氏網路
模糊理論
風險預測
機率式風險評估
Marine Accident Analysis
Fuzzy Bayesian Network
Risk Prediction
Probabilistic Risk Assessment

DSpace CRIS

Influence of Risk Parameters on Marine Accidents Using a Data-Driven Bayesian Network- a Case Study in Taiwan(I)

Details

Description