||Jorgen Amdahi, Soren Ehlers, Bernt Johan Leira
- - - - - - download - - - - - -
The purpose of accident modeling is to learn more about accidents in order to prevent them in the future. Probabilistic accident models, depending on the underlying theoretical accident model type used, quantitatively describe accident causes, mechanisms, event chains, or system variability. Such a model could be utilized within a cost-benefit analysis, risk management or safety-related decision making. However, a ship, and further the marine traffic system as a whole, can be considered as a complex socio-technical system. In such a system an accident is hardly ever a result of a single cause or a chain of events. On the other hand, accidents are low probability events and thus relatively little data about accidents exists. Therefore, the lack of data combined to the complexity of the problem might result in unreliable or invalid probabilistic models. This paper discusses the feasibility of ship accident data for probabilistic collision and or grounding modeling purposes. In addition, as incidents or near-misses occur more frequently than accidents but might be partly governed by the same underlying mechanisms and thus could provide additional information about marine traffic accidents, also incident data is considered. The study is based on examining the data itself when available, reviewing relevant literature, and a case study of evaluating accident data feasibility to learning a Bayesian network model of the dependencies between the reported accident causes. The examination is limited to accident databases providing categorical information on the accidents, accident investigation reports, a near-miss reporting database, and Vessel Traffic Service violation and incident reports. Other potential data sources such as Port State Control inspection data, occupational safety data, data from insurance companies or classification societies are not addressed. The systems and practices of accident or incident reporting or the corresponding data formats might differ from country to country. Here the emphasis is on data describing the marine traffic in Finland.