A data mining framework for risk-based ship design

This thesis was previously held under moratorium from 31st January 2012 until 31st Janury 2014.The Risk-Based Ship Design (RBD) methodology, advocating the systematic integration of risk assessment in the conventional design process so that ship safety is treated as an objective rather than a constraint, has swept through a wide spectrum of the maritime industry over the past fifteen years. Through this methodology, safety is situated at a central position alongside conventional design objectives, so that wellbalanced design effort could be spent and consequently comprehensive design optimisation can be performed. Despite the recognition and increasing popularity, important factors that could potentially undermine its implementation arise both from qualitative and quantitative aspects. This necessitates the development of an objective, reliable and efficient methodology for risk-based ship design implementation. The research presented in this thesis proposes a formalised methodological framework to fulfil this global objective. It comprises three interrelated stages to be performed during risk assessment, namely the development of next generation marine accident/incident database, risk modelling in Bayesian networks by deploying data mining techniques, and the integration with the framework for risk-based design decision making. Working procedures, techniques, methods and algorithms have been developed and applied to representative examples and case studies to demonstrate the applicability and the potential offered by this framework. Each stage of the framework is a field with vast potential for further research, development and application. The ensuing findings firm the faith that an optimal approach towards risk-based design is achievable and extensive applications need to be conducted before experience and confidence can be gained. It is believed that this research has contributed positively towards the evolvement of risk-based ship design.The Risk-Based Ship Design (RBD) methodology, advocating the systematic integration of risk assessment in the conventional design process so that ship safety is treated as an objective rather than a constraint, has swept through a wide spectrum of the maritime industry over the past fifteen years. Through this methodology, safety is situated at a central position alongside conventional design objectives, so that wellbalanced design effort could be spent and consequently comprehensive design optimisation can be performed. Despite the recognition and increasing popularity, important factors that could potentially undermine its implementation arise both from qualitative and quantitative aspects. This necessitates the development of an objective, reliable and efficient methodology for risk-based ship design implementation. The research presented in this thesis proposes a formalised methodological framework to fulfil this global objective. It comprises three interrelated stages to be performed during risk assessment, namely the development of next generation marine accident/incident database, risk modelling in Bayesian networks by deploying data mining techniques, and the integration with the framework for risk-based design decision making. Working procedures, techniques, methods and algorithms have been developed and applied to representative examples and case studies to demonstrate the applicability and the potential offered by this framework. Each stage of the framework is a field with vast potential for further research, development and application. The ensuing findings firm the faith that an optimal approach towards risk-based design is achievable and extensive applications need to be conducted before experience and confidence can be gained. It is believed that this research has contributed positively towards the evolvement of risk-based ship design