Formal safety analysis methods and their application to the design process

Abstract

The work described in this thesis is concerned with formal safety analysis methods and their application to the "design for safety" process of marine and other large Made-To- Order (MTO) products with particular reference to the incorporation of safety aspects into the design process from the initial stages. Large MTO products are complex assemblies of components for which building and testing of prototypes is not usually possible. This thesis proposes a "design for safety" methodology for large MTO products based upon the general spirit of the recommendations from recent government reports including the Cullen and Carver reports. Such a methodology, consisting of five phases, namely problem definition, risk identification, risk estimation, risk evaluation and design review, is used as the basis for the development of more scientific and objective safety analysis methods and techno-economic modelling techniques applicable to the control of major accidents of large MTO products. An analysis of the input requirements and the outcomes of the typical safety analysis methods is conducted to identify their possible inter-relationships within the "design for safety" process in order to make full use of the advantages of each method. The selection of these safety analysis methods is discussed in the context of large MTO products. Problems concerned with failure and repair data collection programmes are studied and some typical failure and repair data sources are described. In order to systematically and effectively identify and estimate risks of large MTO products, an inductive bottom-up Modified Boolean Representation Method (MBRM) is developed to directly make use of the information produced using Failure Mode, Effects and Criticality Analysis (FMECA) to identify and estimate all possible system failure events and respective causes. Such a method can be used to analyse any engineering system which is capable of being broken down into subsystems and components. The overall model and the algorithms are described and tested in association with appropriate computer software. A modified qualitative reasoning method is developed to describe the behaviour of a large complex system. Such a modelling method can be used for failure propagation analysis. The proposed qualitative modelling method is further combined with the MBRM to form a flexible mixed safety modelling methodology. In this methodology, the MBRM is used to process the information produced from the qualitative reasoning analysis at the component level to obtain a description of the total system behaviour. This methodology allows a bottom-up safety analysis approach to be taken even in those cases where it is difficult to obtain complete input-output relations for all the components of the system. Two general simulation models are developed to process the information produced using FMECA and the MBRM. Such simulation models can be used as a quantitative safety analysis tool to simulate system availability, component/subsystem failures, and the probability of occurrence of each identified system failure event. These two models are developed in an Object-Oriented Programming (OOP) environment. This thesis also presents a new safety analysis and synthesis methodology involving the use of fuzzy Set modelling and evidential reasoning, where fuzzy set modelling is used to describe each failure event and an evidential reasoning approach is then employed to synthesise the information produced to assess the safety of the whole system. This subjective reasoning methodology can be used as an alternative approach by safety analysts to carry out analysis particularly in those situations where mostly nonnumerical safety data is available or where there is a lack of information regarding distributions of variables for use in probabilistic risk studies. A techno-economic modelling methodology is also developed to determine where reasonably practicable design actions are required. The proposed methodology brings together risk and cost objectives into the decision making process for the improvement of design aspects and maintenance policies. Information produced using the safety analysis approaches developed in this thesis can be utilised to construct a technoeconomic model. Multiple Objective Decision Making (MODM) techniques are then employed to process the constructed model. The results produced can assist designers in developing good compromise designs that take into account risks, their possible consequences, maintenance cost, repair cost and design review cost. A hydraulic transmission system of an offshore pedestal crane is used to demonstrate the methodologies developed in this thesis. Finally, the results of the research project are generally summarised and the areas where further effort is seen to be required to improve the developed methodologies are outlined.