Physical simulations, such as thermal-hydraulic codes and finite element analysis, have been increasingly utilized to support probabilistic risk assessment (PRA) of nuclear power plants (NPPs) for constructing scenarios and quantifying their probabilities and consequences. This trend is driven, on the one hand, by advances in computational power, which enable the execution of physical simulations within the timeframe consistent with the typical PRA development and update processes. On the other hand, there is growing recognition of the value of explicitly integrating physical simulations with PRA. For new reactors, physical simulations can play an important role in augmenting limited operational data and providing an in-depth understanding of system behaviors. Even for existing NPPs with historical operational data, integrating physical simulations into PRA can yield more detailed, in-depth risk insights that facilitate various risk-informed decision-making, including lifetime extension, power uprate, and maintenance efficiencies.
The authors’ experiences highlight the importance of carefully designing the interface between physical simulations and PRA, as it can significantly affect both the accuracy and computational efficiency of the PRA-physics integration. Multiple aspects that can impact PRA accuracy, such as spatial and temporal resolution and dependencies, must be addressed in the PRA-physics interface. Simultaneously, algorithmic and user interface designs strongly affect the practical usability of PRA-physics interface tools, including their computational efficiencies and accessibility.
In this paper, a literature review is conducted on the integration of PRA and physical simulations for nuclear energy systems. The review analyzes key characteristics of the existing integration approaches or tools, including: (i) What modeling and simulation tools are employed across each layer of the system hierarchy? (ii) What computational models or techniques are used to integrate PRA with physical simulations? (iii) How are dependent failures treated? (iv) What uncertainty quantification methods are used? (v) To what extent are artificial intelligence and machine learning techniques incorporated into the PRA-physics interface, and what roles do they serve? The review findings are synthesized to establish a structured understanding of the current state of the art in integration of PRA with physical simulations, identify research gaps in computational interfaces, and recommend future research directions to enhance risk analysis capabilities to support the continued operation of existing plants and the timely deployment of advanced reactors.