Human Error Analysis For Inland Waterway Ship Collisions By A Bayesian Network Approach

Inland waterway transport(IWT),as characterized by its relatively lower cost,higher capacity,and apparently less environmental impact,has been playing an increasingly important role in the integrated ground-water transportation system.However,a number of serious ship accidents have attracted public concerns to IWT safety and reliability.Accidents records and statistics have indicated that ship collision is the most frequent occurrence in inland waterway areas,and about 80% of ship accident are caused by human errors.Due to frequent occurrence and serious consequences of ship accidents,the analysis of accidents causation has always been the research hotspot in IWT field.Focusing on the impacts of human errors on ship navigation safety,this thesis takes the ship collision as a typical accident type and aims to make human reliability evaluation for ship navigation safety by proposing a BN model to depict causal relationships.By collecting and analyzing 123 ship collision accident investigation reports,this thesis summarizes the types of human errors concerned with ship collisions and establishes the ship collision model to present the accident mechanism by the information processing procedure “perception-decision making-action” under the guidance of Swain & Guttmann model.Then a Bayesian network reflecting the causality of included human errors is established with the combining of domain expert knowledge and data learning results.The qualitative part of the BN is structured by three expert group.The quantitative part is based on correlation analysis of human related factors from 123 accident investigation reports.The validity of BN model is verified by sensitivity analysis and case analysis of real ship collisions.Then this thesis uses the BN model to analyze the main causation errors to find the most critical human errors by Bayesian inference,including diagnostic inference and causal inference.In view of the analyzing results,some specific suggestions are offered,such as professional training on ship handing skills,reasonable shift arrangement,better ship traffic service,and so on.The established model could directly reflect the correlations among the human errors and their impacts on a collision.Findings of this thesis have certain reference value for prevent or decrease ship accidents in inland waterway.And the selected Swain & Guttmann model and Bayesian network model can be used for reference in risk analysis of other types of transportation accidents.