Monte Carlo approach to PSA for dynamic process systems

The methodologies employed in the Probabilistic Safety Assessment of technological systems have nowadays reached a high level of maturity. However, there are still some important issues that need further investigations, in order to increase the confidence in the results obtained. Important areas such as human factors, expert judgement and physical modelling are considered to introduce large uncertainties in the probabilistic safety assessments and strong efforts are being devoted lately to the improvement of these aspects within the PSA methodology. Recently, a more ‘dynamic’ approach to PSA analysis has been proposed, with the intention of addressing those problems related to the mutual interactions of the hardware components and the physical evolution of the plant. In the present paper we outline this approach within the framework of the Monte Carlo simulation. The method is shown to provide a flexible environment in which many important issues of the plantʹs dynamics can be considered, such as the action of control/protection systems and its failures on demand, the effects of the process variables on the system stochastic characteristics and the delay with which a top event can occur. We refer, for our examples, to a simple case study taken from the literature. The limitations of the Monte Carlo method in treating rare-events is also discussed and we address the problem of biasing the probabilities of the control/ protection systems failing on demand.