Bayesian network modeling of potential patterns in maritime safety performance

摘要: Aalto University, P.O. Box 11000, FI-00076 www.aalto.fi Author Maria Hanninen Name of the doctoral dissertation Bayesian network modeling potential patterns in maritime safety performance Publisher School Engineering Unit Department Applied Mechanics Series University publication series DOCTORAL DISSERTATIONS 13/2015 Field research Marine Technology and Naval Architecture Manuscript submitted 10 October 2014 Date defence 6 March 2015 Permission to publish granted (date) 15 December Language English Monograph Article (summary + original articles) Abstract Although major accidents occur rather rarely, they can produce severe consequences. Safety management aims at preventing such accidents. For monitoring improving safety, requires knowledge on performance. Information various variables which are potentially related a description how interact, or types interactions constitute, could be beneficial for increasing that knowledge. However, these might not apparent, as traffic its complicated systems.Although systems. Utilizing approach, this thesis explores between safety. The decision-makers then exploit information starting point analyses mechanisms have generated what tell about models safety-performance from different, complementing angles. work begins with examining feasibility accident causation pattern exploration This includes analyzing an existing causal collision model assessing incident data grounding cause learning. focus shifts present multiple indicator data, before analysis is extended dependencies While found questionable, Port State Control inspections act valuable source information. Finnish ports resulted few deficiencies thus contains only weak patterns. It worth evaluating whether developed so would more detailed differences different ships. On other hand, seems tightly coupled system several properly functioning subareas but inadequate whole. Regarding application networks problem, concludes their capability express uncertain, complex combine expert knowledge, offer attractive tool task. As amount collected shared slowly within community, easily updated new information, quality decision-support improved.