Combining probabilistic and deterministic methods for accident analysis

• Autor(a) principal: GOMES, DANIEL de S.
• Data de Publicação: 2020
• Outros Autores: GIOVEDI, CLAUDIA, INTERNATIONAL NUCLEAR ATLANTIC CONFERENCE
• Tipo de documento: Artigo de conferência
• Título da fonte: Repositório Institucional do IPEN
• Texto Completo: http://repositorio.ipen.br/handle/123456789/30725
• topic: cladding; data covariances; deterministic estimation; f codes; fuel rods; loss of coolant; probabilistic estimation; reactivity; reactor accidents; reactor cores; reactors; safety analysis; sensitivity analysis; transients

This study describes a practical method applied to nuclear reactor safety analysis (NRSA), based on an approach so-called best estimate plus uncertainty (BEPU). The innovative analysis approach involves statistical methods integrated with deterministic rules to fuel licensing code (FLC). The goal of NRSA is to improve safety margins in the nuclear reactor operation, which has partially achieved with uncertainty treatment. Previously, BEPU analysis was widely used to study the loss of coolant accident (LOCA), via inclusion in thermal-hydraulic codes (THC). The systems can measure the impact caused by uncertainties spread in core reactors with a coupling of THC and optimization packages. This paper shows the result of applying the UA/SA technique to FRAPCON, joined with DAKOTA toolkit. This integration will offer the probabilistic analysis coupled with empirical rules. A perfect fusion of the concepts permits the exploration of parametric uncertainties and calibration of physical models. We can use the combined utilization of FLC systems and the DAKOTA toolkit to produce sensitivity analysis. The first step in this approach is to identify all uncertainty sources of the physical models, the reactor design, and manufacturing parameters. It is subsequently used into an FLC, such as FRAPCON, as input parameters. The uncertainties usually distributed using the Wilks formula, which determines the number of samples required for unilateral tolerance. According to Wilks' method, it needs 59 data samples to achieve a confidence level of 95%. Results from Wilks formula found via Monte Carlo simulation, which applies to FLC coupled with sensitivity analysis.