Guided Probabilistic Simulation of Complex Systems Toward Rare and Extreme Events

Abstract

Simulation-based or dynamic probabilistic risk assessment methodologies were primarily developed for proving a more realistic and complete representation of the complex systems accident response. Such simulation-based methodologies have proven to be particularly powerful for systems with control loops and complex interactions between its elements, be they hardware, software, or human, as they provide a natural probabilistic environment to include physical models of system behavior (e.g., coupled neutronics and thermal-hydraulic codes for nuclear power plants), mechanistic models of materials or hardware systems to predict failure, and those of natural hazards. Despite the advancements in simulation-based methodologies, the fundamental challenge still persists as the space of possible probabilistic system trajectories is nearly infinite in size in simulating even systems of relatively low complexity.Existing methodologies that analyze and predict these scenarios, some of which may involve rare events of interest, are either computationally prohibitive for the high-dimensional systems of interest or are only effective for low-dimensional dynamics and cannot capture the large-scale, nonlinear behavior of interconnected complex systems.In this paper, a framework is developed to identify rare and extreme events and enabling the use of reverse trajectories to trace failures (or other system states) to causes for potential mitigation actions. This framework consists of an Intelligent Guidance module, Trajectory Generation module and Physical Simulation module. The Intelligent Guidance module provides planning information to the Trajectory Generation module that creates scenarios by interacting with the Physical Simulation in its environment. In turn, system trajectories or scenarios are created and post-processed to provide updating information to the Intelligent Guidance module or aggregate the results when stopping criteria are met. The objective of guided simulation is to control the growth of the “scenario tree” and to efficiently identify important scenarios that meet single or multiple criteria. We present several solution strategies, both qualitative and data-driven for each module.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 13 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.