Tackling the complexity of PRA Models: measure a Fault Tree

Probabilistic Risk Assessment (PRA) models are becoming increasingly complex. This growing complexity is due to the expanding coverage perimeter (level 1, 2 and 3 PRA, hazards, combination of hazards, etc.) as well as the need for enhanced safety and finer levels of detail. This increased complexity of PRA models makes them challenging to manage, verify, maintain, and quantify. Many stakeholders in PRA domain are working over this issue, exploring potential model simplifications without compromising modelling details. Several publications on this topic can already be found in the literature. In 2023, we initiated an R&D project to tackle this complexity. This work consists of several steps: defining « model complexity » and its different types, identifying the sources of such complexities and, eventually, addressing them. The initial task of defining and identifying the different types of complexity have already been presented in previous works. In this study, we focus on computational time. One potential cause of long computational times might be the complexity of Fault Trees (FTs). To assess this complexity, the computational time should be put in relation with the size and the structure of FTs. But which metrics can be used to characterize and measure it? This paper aims at answering this question. The final goal of the activity being to determine the impact of the FT metrics on the computational time.

Keywords: structure, fault trees, complexity, PRA/PSA.