| Ships have a wide range of influence and functions in today’s world and are widely used in international trade,tourism,maritime safety,resource development and other aspects.However,a large number of ship accidents have triggered people’s thinking on the navigation safety of ships.The International Maritime Organization,IMO stated that between 2015 and 2019,human factors have become the most important cause of ship accidents,accounting for 74%.In response to this phenomenon,the International Maritime Organization introduced a formal safety assessment(FSA)to analyze and evaluate the risk factors in ship navigation accidents.The China Maritime Safety Administration also issued a ship accident report,calling on relevant industry practitioners to study safety management countermeasures.Based on the ship accident report issued by China Maritime Safety Administration,this paper combined the Systems-theoretic process analysis(STPA)and Bayesian Network(BN)model,to analyze and evaluate the human risk factors of ship navigation in Chinese waters.Firstly,the ship accident reports were collected.A total of 489 ship accident reports issued by China Maritime Safety Administration between January 1,2017 and December 31,2021 were collected.The general history,types and results of ship accidents were briefly read,as well as the grade evaluation of the accident by the Maritime Safety Administration.The 445 accident reports were screened out and the basic information was classified and summarized.Four kinds of systems-level losses and eight kinds of systems-level hazards in ship accidents are defined by the method of systematic theoretical analysis.After a thorough reading of relevant literature on ship risk factor analysis and some public materials of classification societies,the ship safety control structure diagram is determined,21 kinds of unsafe control behaviors of crew/shore-based personnel are identified from the safety control structure diagram,and further analysis is carried out on the accident process of ship accident report,unsafe behavior summary and responsibility identification.Combined with the extracted unsafe control behaviors,three unsafe control behavior scenarios and three unsafe control behavior types are determined.Then,the unsafe control behavior is summarized and summarized in combination with its scenarios and types,forming 12 kinds of human risk factors in ship navigation,and the human risk factors summarized are the nodes of Bayesian network.Based on the summarized human risk factors,detailed reading of the accident process,unsafe behavior summary and responsibility identification in the ship accident report is conducted again.The causal relationship between all human accident factors is established according to the stage and accident process,and the topological structure of Bayesian network is obtained by combining with the structural learning method of Bayesian network.At the same time,the human risk factors contained in the ship accident report are quantified and made into a data set.When collecting ship accident reports,29 accident reports were selected as validation data set according to the completely random principle,and the remaining 416 data were used as training data.In this paper,the Bayesian visualization software Netica is used to establish and verify the Bayesian network model.The obtained network nodes and topology structure are constructed into the original Bayesian network model in the software,and the training data is input into the network for parameter learning.Then,the risk factors contained in the verification data are input to check the posterior probability of ship accidents.Since all the data involved in the accident report have occurred,the accuracy of the model is verified by comparing this probability with 100% probability of accident occurrence.Then,by using the diagnostic reasoning and causal reasoning functions of Bayesian networks,the posterior probability of the parent nodes of each risk factor in the network model is examined,and seven outstanding human risk factors and five core accident cause chains are obtained.Based on the obtained information,guiding opinions are put forward for the shipping industry and ship related staff.The Bayesian network model of human risk factors for ship navigation in the waters of China built in this paper has high accuracy and good complexity,and can be used to study the internal causal relationship of human risk factors in ship navigation in the waters of China,the influence degree of each risk factor on the occurrence of ship accidents and the cause chain of ship accidents.To provide guidance and data support for the safe navigation of ships.The research method of combining system theory analysis method(STPA)and Bayesian network(BN)model used in this paper also has reference significance. |
