Proceedings of the Eighth International Conference on Probabilistic Safety Assessment & Management (PSAM)
173 Solution of Phased-Mission Benchmark Problem Using the SimPRA Dynamic PRA Methdology (PSAM-0345)
Download citation file:
This paper describes the application of the Simulation-based Dynamic PRA methodology (SimPRA)  to the benchmark problem proposed for an invited session on advanced PRA methods. The benchmark problem  is a multi-phase mission involving an ion propulsion system needed for a science mission to the outer solar system. The propulsion system is needed only in some of the phases, during which thrust is continually provided.
SimPRA is an adaptive-scheduling dynamic PRA environment, where prior knowledge of the systems and knowledge gained during simulation is used to dynamically adjust the scenario exploration rules. SimPRA modeling environment has three key ingredients. The planner captures any prior knowledge about the possible behavior of the system, and is used to provide high level guidance for the simulation and generation of risk scenarios. The actual scenario generation is governed by the scheduler, generating branch points based on probabilistic and deterministic rules. The system behavior itself is coded in form of a Simulation Model built from models of system components and their functional and physical properties.
A simulation model was built for the benchmark exercise to represent the physical characteristics of the propulsion system at the same level of detail as specified in the problem definition. SimPRA environment provides a Dynamic PRA library to model different failure modes and dynamic features, such as those described in the benchmark problem. A separate high level state-based model feeds the planner. For this relatively simple system a simple plan was sufficient. The simulation result, i.e., risk scenarios leading to predefined End States are in form of specific realizations of time dependent sequences of events. End State probabilities are based on grouping of such realizations.
In order to show the full capabilities of SimPRA, the benchmark problem is extended in following ways: 1) use of more diverse time-to-failure distributions, such as Weibull, 2) addition of physical process controlled common cause failures in addition to standard parametric models, and 3) inclusion of a control software model with malfunction failure modes. This more complex case has also been solved with SimPRA.