A GUIDED SIMULATION METHODOLOGY FOR DYNAMIC PROBABILISTIC RISK ASSESSMENT OF COMPLEX SYSTEMS

Abstract

Probabilistic risk assessment (PRA) is a systematic process of examining how engineered systems work to ensure safety. With the growth of the size of the dynamic systems and the complexity of the interactions between hardware, software, and humans, it is extremely difficult to enumerate the risky scenarios by the traditional PRA methods. Over the past 15 years, a host of DPRA methods have been proposed to serve as supplemental tools to traditional PRA to deal with complex dynamic systems. A new dynamic probabilistic risk assessment framework is proposed in this dissertation. In this framework a new exploration strategy is employed. The engineering knowledge of the system is explicitly used to guide the simulation to achieve higher efficiency and accuracy. The engineering knowledge is reflected in the "Planner" which is responsible for generating plans as a high level map to guide the simulation. A scheduler is responsible for guiding the simulation by controlling the timing and occurrence of the random events. During the simulation the possible random events are proposed to the scheduler at branch points. The scheduler decides which events are to be simulated. Scheduler would favor the events with higher values. The value of a proposed event depends on the information gain from exploring that scenario, and the importance factor of the scenario. The information gain is measured by the information entropy, and the importance factor is based on the engineering judgment. The simulation results are recorded and grouped for later studies. The planner may "learn" from the simulation results, and update the plan to guide further simulation.

SIMPRA is the software package which implements the new methodology. It provides the users with a friendly interface and a rich DPRA library to aid in the construction of the simulation model. The engineering knowledge can be input into the Planner, which would generate a plan automatically. The scheduler would guide the simulation according to the plan. The simulation generates many accident event sequences and estimates of the end state probabilities.