Synthetical Model Generator for Probabilistic Risk Assessment Tools: Enhancing Testing, Verifying and Learning

Improving a probabilistic risk assessment (PRA) quantification engine requires a deep
understanding of how to solve PRA models and access to a wide variety of models for testing. A quantification
engine parses a model's text file, generates cut sets, and produces numerical results. However, access to diverse
models is limited due to the constraints of the nuclear industry. To address this challenge, we propose a synthetic
model generator capable of creating event trees, fault
trees, or a combination of both in various configurations. For event trees, configurations include options such as
the number of functional events and the number and types of end states. In contrast,
for fault trees, the configurations are more complex, offering researchers more flexibility to explore different
scenarios. Additionally, this work establishes an educational platform for junior PRA practitioners. By
generating and solving various types of models, they can gain hands-on experience
in model building, understanding expected outcomes, and developing problem-solving techniques
The synthetic model generator utility is built on four key pillars. The first pillar involves defining
model configurations, where users provide input through a text file specifying all necessary parameters. The
second pillar focuses on generating the model based on user-defined expectations.
The third pillar verifies the generated model adheres to the defined schema. Finally, the fourth pillar involves solving
the generated model using a quantification engine for completeness.
For this research, the Open-PSA is used as the model format, and SCRAM is used as the
quantification engine for demonstration purposes. However, the methodology can be extended to support other
model formats and quantification engines as needed.