Modelling life extension of safety critical systems

Due to the cost of large-scale infrastructure construction many asset intensive industries operate their plant and systems beyond the originally intended design life. For new systems there is also a strong motivation for the systems to operate for extended lifetimes. Maintenance is used to keep the condition of the infrastructure in an acceptable state. Controlling the asset state using maintenance requires an asset management strategy which is optimised to yield minimal whole life costs. Adequate asset management of aging plant is an important requirement to prevent service disruption through breakdown but also essential for safety critical systems where failure can result in fatalities. Integrated Fault Tree and Event Tree methods are commonly used to evaluate the risk posed by such systems. However, these approaches have limitations. A fundamental assumption of these methods is that component failures occur independently of each other. Features in the system design or the maintenance process frequently introduce dependencies between component states. Commonly the software implementations of the methods will only accept exponentially distributed times to failure and repair which cannot account for the increasing rate of failure experienced as components age. Another concern is that the simple models incorporated for the fault tree basic events cannot adequately represent the details of the complex maintenance strategies employed. This paper presents a new integrated modelling approach based on Petri nets and Bayesian Networks which expands the modelling capability to overcome the deficiencies of the traditional methods.