Traditionally, regulation of design and operation of nuclear power plants (NPPs) have been based on deterministic safety assessment (DSA) methods to verify criteria that assure plant safety in a number of postulated design basis accident scenarios. These criteria also allow identifying which plant structures, systems, and components (SSCs) and activities are important to safety. Design, operation, and maintenance of these “safety-related” SSCs and activities are controlled through regulatory requirements and supported by extensive probabilistic safety assessment (PSA). Compliance with the evolving regulatory requirements will require innovative deterministic and probabilistic approaches of safety assessment for the existing NPPs and their life extension. In this respect, a medium-term challenge is to combine the use of deterministic and probabilistic methodologies in an integrated deterministic and probabilistic safety analysis (IDPSA) and tackle the related problems of addressing aleatory (stochastic aspects of accident scenarios) and epistemic (model and parameters) uncertainties in a consistent manner. Open issues and challenges of IDPSA include the following.

• Increased complexity of the thermal-hydraulic (TH) models, with prohibitive computational costs for running hundreds/thousands of transients simulations
• Difficult quantification of the uncertainties of the TH models adopted for accident analysis
• Nontransparency of complex PSA models, when attempting at resolving time dependent interactions between physical phenomena, control logic, operator actions, software/firmware, and equipment failures
• Increased reliance on expert judgement in providing conservative assumptions about uncertainties in time dependent scenarios
• Comparison of IDPSA outcomes with available safety test results (especially when dealing with fleet of reactors or attempting at justifying long-term operation of earlier NPP designs)
• Complexity in the assessment of the impact of human operator actions on time dependent scenarios.

Potential topics include, but are not limited to:

• Reduction of the computational burden of TH models, thanks to larger employment of metamodels for multiparameter and nonlinear modelling, and exploration of the space of combinatorial plant scenarios
• Consistent treatment of different sources of uncertainties
• Identification and characterization of a priori unknown vulnerable scenarios, or "sleeping threats"
• Description of the time-dependent interactions between physical phenomena, equipment failures, safety and nonsafety systems interactions, control logic, and operator actions
• Reduction of reliance on expert judgment and simplifying assumptions about interdependencies