Probabilistic dynamics as a tool for dynamic PSA

The assumptions, scope and achievements of a probabilistic dynamics theory based on a Chapman-Kolmogorov formulation of mixed probabilistic and deterministic dynamics are reviewed. The formulation of the theory involves both physical (or process) variables and (semi-) Markovian states of the system under study allowing the inclusion of human error modelling. The problem of crossing a safety threshold is used to emphasize the role of timing in concurrent sequences. We show how the adjoint formulation can be used to obtain information on the outcomes of transients as a function of its starting characteristics. These outcomes may, for instance, be damage resulting from safety boundary crossing, or reliability functions. A comparison is made between a Monte-Carlo solution and a DYLAM analysis of a simple multicomponent benchmark problem which shows that for the same accuracy a Monte-Carlo method is much less sensitive to the size of the problem. (C) 1996 Elsevier Science Limited.