Integrative risk-based assessment modelling of safety-critical marine and offshore applications
This research has first reviewed the current status and future aspects of marine and offshore safety assessment. The major problems identified in marine and offshore safety assessment in this research are associated with inappropriate treatment of uncertainty in data and human error issues during the modelling process. Following the identification of the research needs, this thesis has developed several analytical models for the safety assessment of marine and offshore systems/units. Such models can be effectively integrated into a risk-based framework using the marine formal safety assessment and offshore safety case concepts. Bayesian network (BN) and fuzzy logic (FL) approaches applicable to marine and offshore safety assessment have been proposed for systematically and effectively addressing uncertainty due to randomness and vagueness in data respectively. BN test cases for both a ship evacuation process and a collision scenario between the shuttle tanker and Floating, Production, Storage and Offloading unit (FPSO) have been produced within a cause-effect domain in which Bayes' theorem is the focal mechanism of inference processing. The proposed FL model incorporating fuzzy set theory and an evidential reasoning synthesis has been demonstrated on the FPSO-shuttle tanker collision scenario. The FL and BN models have been combined via mass assignment theory into a fuzzy-Bayesian network (FBN) in which the advantages of both are incorporated. This FBN model has then been demonstrated by addressing human error issues in a ship evacuation study using performance-shaping factors. It is concluded that the developed FL, BN and FBN models provide a flexible and transparent way of improving safety knowledge, assessments and practices in the marine and offshore applications. The outcomes have the potential to facilitate the decision-making process in a risk-based framework. Finally, the results of the research are summarised and areas where further research is required to improve the developed methodologies are outlined.