INCORPORATING DYNAMIC 3D SIMULATION INTO PRA

Through continued advancement in computational resources, development that was previously done by trial and error production is now performed through computer simulation. These virtual physical representations have the potential to provide accurate and valid modeling results and are being used in many different technical fields. Risk assessment now has the opportunity to use 3D simulation to improve analysis results and insights, especially for external event analysis. By using simulations, the modeler only has to determine the likelihood of an event without having to also predict the results of that event. The 3D simulation automatically determines not only the outcome of the event, but when those failures occur. How can we effectively incorporate 3D simulation into traditional PRA? Most PRA plant modeling is made up of components with different failure modes, probabilities, and rates. Typically, these components are grouped into various systems and then are modeled together (in different combinations) as a “system” with logic structures to form fault trees. Applicable fault trees are combined through scenarios, typically represented by event tree models. Though this method gives us failure results for a given model, it has limitations when it comes to time-based dependencies or dependencies that are coupled to physical processes more » which may themselves be space- or time-dependent. Since, failures from a 3D simulation are naturally time related, they should be used in that manner. In our simulation approach, traditional static models are converted into an equivalent state diagram representation with start states, probabilistic driven movements between states and terminal states. As the state model is run repeatedly, it converges to the same results as the PRA model in cases where time-related factors are not important. In cases where timing considerations are important (e.g., when events are dependent upon each other), then the simulation approach will typically provide superior results and insights. We also couple the state model with the dynamic 3D simulation analysis representing events (such as flooding) to determine which (if any) components fail. Not only does the simulation take into account any failed items from the state model, but any failures caused by the simulation are incorporated back into the state model and factored into the overall results. Using this method we incorporate accurate 3D simulation results, eliminate static-based PRA issues, and have time ordered failure information. « less