A branching search approach to safety system design optimisation

Safety systems are designed to prevent or mitigate the consequences of potentially hazardous events. In many industries the failure of such systems can result in fatalities. Current design practice is usually to produce a safety system which meets a target level of performance that is deemed acceptable by the regulators. However, when the system failure will result in fatalities it is desirable for the system to achieve an optimal rather than adequate level of performance given the limitations placed on available resources. The unavailability of safety systems can be predicted using fault tree analysis methods. Formulating an optimisation problem for the system design has features which make standard mathematical optimisation techniques inappropriate. The form of the objective function is itself a function of the design variables, the design variables are mainly integers and the constraint forms can be implicit or non-linear. This paper presents a Branching Search algorithm which exploits characteristics common to many safety systems to explore the potential design space and deliver an optimal design. Efficiency in the method is maintained by performing the system unavailability evaluations using the Binary Decision Diagram method of fault tree solution. Limitations are placed on resources such as cost, maintenance down-time and spurious trip frequency. Its application is demonstrated on a High Integrity Protection System.