مدل شبکه های بیزی مدیریت ایمنی دریایی

This paper presents a model of maritime safety management and its subareas. Furthermore, the paper links the safety management to the maritime traffic safety indicated by accident involvement, incidents reported by Vessel Traffic Service and the results from Port State Control inspections. Bayesian belief networks are applied as the modeling technique and the model parameters are based on expert elicitation and learning from historical data. The results from this new application domain of a Bayesian network based expert system suggest that, although several its subareas are functioning properly, the current status of the safety management on vessels navigating in the Finnish waters has room for improvement; the probability of zero poor safety management subareas is only 0.13. Furthermore, according to the model a good IT system for the safety management is the strongest safety-management related signal of an adequate overall safety management level. If no deficiencies have been discovered during a Port State Control inspection, the adequacy of the safety management is almost twice as probable as without knowledge on the inspection history. The resulted model could be applied to performing several safety management related queries and it thus provides support for maritime safety related decision making.

Safety management is a subarea of organizational management. Its aim is to develop, plan, realize and follow operations for preventing accidents and minimizing risks related to the safety of people, environment or property. In the maritime domain, the International Safety Management (ISM) Code provides requirements for safety management systems (International Maritime Organization, 2013). The ISM Code is mandatory for all ships belonging to the scope of the International Convention for the Safety of Life at Sea (SOLAS) of International Maritime Organization (IMO), that is, the majority of internationally trading ships. To support the ISM Code, some particular sectors of the maritime transportation have launched their own additional safety management guidelines such as the Tanker Management Self-Assessment (TMSA) by The Oil Companies International Maritime Forum (OCIMF, 2008). Safety management is a broad topic and covers several subareas. A model describing the elements of maritime safety management, how these elements interact, and how strongly safety management and safety are linked could provide useful information about the functioning of the safety management. It could serve as an assessment and monitoring tool and aid in continuous improvement and decision making when managing maritime traffic safety. Previously several frameworks for assessing the effects of organizational aspects on risk or safety have been published within different domains (Embrey, 1992, Paté-Cornell and Murphy, 1996, Øien, 2001, Mohaghegh et al., 2009 and Roelen et al., 2003) and in the maritime transportation field (Trucco, Cagno, Ruggeri, & Grande, 2008). Components and the component connections of concepts closely related to safety management, such as the safety culture, have been mathematically described in other domains (dos Santos Grecco, Vidal, Cosenza, dos Santos, & de Carvalho, 2014) and on a coarse level in the maritime traffic (HÅvold, 2005, Oltedal and McArthur, 2011 and Ek et al., 2014). However, these have not addressed safety management per se. While studies such as Le Coze (2013) have investigated the elements of safety management and even qualitatively modeled their interdependencies (Hale, Heming, Carthey, & Kirwan, 1997), detailed quantifications of the maritime safety management subarea interactions seem to be lacking; the existing models provide rather limited means to comprehensive reasoning about the maritime safety management mechanisms and to the related decision-support. Furthermore, to the authors’ knowledge, no safety management models based on the established safety management norms or standards have been published. To address the lack of quantitative decision-support tools for maritime safety management in the existing literature, this paper presents a Bayesian network based expert system which models the maritime safety management subarea qualities and their dependency patterns. In addition, the paper links the safety management to three maritime traffic safety indicators: maritime traffic accident involvement, conducting a violation or incident that has been reported by a Vessel Traffic Service (VTS) center, and deficiencies discovered in Port State Control (PSC) inspections. The aim of the model is to describe the current status of safety management on board ships navigating within Finnish waters. More specifically, it provides an expert-based probabilistic representation of maritime safety management norms while also considering the uncertainty related to the subareas and their links and the one between the experts. Moreover, by connecting the current status of safety management to the aforementioned accident and incident data, the model provides means to evaluate how informative these safety indicators are for assessing the state of the safety management and vice versa. The featured aspects the authors believe to be novel in this paper can be summarized as (1) the utilization of current norms and standards in establishing a maritime safety management model, (2) the application of Bayesian networks as the modeling technique to describe and reason about safety management, and in this case, the specifics of maritime safety management, and (3) the linking of an expert-based maritime safety management model to three data-based maritime safety indicators. The rest of the paper is organized as follows. Section 2 describes the applied methods and input data in building the model. The resulting model and some findings derived with it are presented in Section 3, Appendix A and Appendix B. The Section 4 discusses the results further while also presenting an evaluation of the model validity. Finally, conclusions are drawn in Section 5.

This study has introduced a Bayesian network based expert system of safety management to the maritime domain. It is the first step towards a comprehensive decision-supporting expert system for improving the maritime traffic safety through safety management. In fact, the domain experts involved in the model building process showed keen interest towards having such a system as a tool for their work. The constructed model can be utilized in probabilistic reasoning about the current situation of maritime safety management within the Finnish waters, the examination of the dependency patterns between several safety management subarea qualities and the overall safety management level. Furthermore, the model provides information between safety management and accident involvement, VTS reported incidents and Port State Control inspection results of ships inspected in Finnish ports. As the model includes the views of multiple experts with different backgrounds while including the possibility to analyze the problem based on only one expert, it could be applied to analyzing the safety management within one shipping company, or it could be used for the entire maritime traffic within the Finnish waters. On the other hand, the expert-knowledge based description of the safety management subarea relationships within the existing safety management frameworks provides additional insight on the contents of the norms, which can be considered useful for managing safety in any sea area, perhaps even in other safety-critical domains. While providing a visually understandable representation of the maritime safety management and its subareas for a person without experience on probabilistic modeling or joint probabilities, the applied Bayesian network approach also takes into account the uncertainty involved in the modeling. It can be concluded that while the proposed model might require further validation, it seems to adequately well reflect the views of several maritime experts on the matter and can serve as a starting point for analyzing the safety management situation on board vessels visiting Finnish ports. While the presented BN model can already support decision-making, it still has room for improvement. In regards to expanding the model contents, in future the model could be developed to mimic the consequences of implementing certain changes in the safety management procedures. This enhancement could provide means to the consideration of which elements of safety management should be tackled in order to increase the maritime safety as well as possible, for example. On the other hand, the development of the modeling approach should be further studied. First, the utilization of dynamic Bayesian networks for capturing the continuous improvement aspect of safety management is worth evaluating. Second, methods for enhancing and improving the expert knowledge elicitation would be needed. Third, the potential of using statistical safety management data in learning the safety management subarea patterns instead of only relying on the expert knowledge could be studied. More specifically, the model could be augmented with additional safety management indicator variables which would be based on the data gathered from sources such as internal and external audits. Then the actual safety management subareas could be treated as hidden BN variables. While providing additional knowledge inferred from the data, such a model would also better reflect the safety management philosophy of using measurable indicators for the safety performance assessment.