Sensitivity and uncertainty evaluation applied to the failure process of nuclear fuel

Abstract

Nuclear power plants must operate with minimal risk. The nuclear power plants licensing process is based on a paired model, combining probabilistic and deterministic approaches to improve fuel rod performance during both steady state and transient events. In this study, performance fuel codes were used to simulate the test rod IFA-650-4, with a burnup of 92 GWd/MTU within a Halden reactor. In a loss-of-coolant test, the cladding failed within 336 s after reaching a temperature of 800 °C. Nuclear systems work with many imprecise values that must be quantified and propagated. These sources were separated by physical models or boundary conditions describing fuel thermal conductibility, fission gas release, and creep rates. These factors change output responses. Manufacturing tolerances show dimensional variations for fuel rods, and boundary conditions within the system are characterized using small ranges that can spread throughout the system. To identify the input parameters that produce output effects, we used Pearson coefficients between input and output. These input values represent uncertainties using a stochastic technique that can define the effect of input parameters on the establishment of realistic safety limits. Random sampling provided a set of runs for independent variables proposed by Wilks' formulation. The number of samples required to achieve the 95th percentile, with 95% confidence, depending on verifying the confidence interval to each output. The FRAPTRAN code utilized a module to reproduce the plastic response, defining the failure limit of the fuel rod.