Quantitative Analysis of Dynamic Fault Trees by means of Monte Carlo Simulations: Event-Driven Simulation Approach

The reliability analysis of complex and dynamic systems is often achieved by a quantitative analysis of dynamic fault trees (DFT), which model the system failure, i.e. a specific undesired event called top event, in terms of failures of the components of the system. Indeed, DFT takes into account the sequential relationships among events and their statistical dependencies. Given the failure probability of the components, the quantitative analysis aims at numerically evaluating, among other things, the failure probability of the top event. In this paper, we are interested in the Monte Carlo simulation which can consider any kind of failure distribution and is not limited in the DFT representation: it considers DFT with repeated events and shared events, takes into account all dynamic gates (PAND, SEQ, FDEP, and SPARE). However, Monte Carlo simulation encounters some disadvantages: an entirely new simulation must be executed every time a parameter changes and it may be time-consuming when the desired accuracy is high. To address these difficulties, this paper proposes a new dynamic fault tree simulation performed by an event-driven simulator. With this approach, gate simulations that produce no change in the output of a gate are eliminated augmenting the speed up of the simulation. The implementation of our approach uses an event queue data structure and an event-scheduler as alternative to the usual time-driven implementation which is characterized by an iterative loop. Thus, periods of inactivity are omitted. As results, computational efficiency is obtained and the speed-up performance of the Monte Carlo simulation program is improved.