Dakota-SAS4A/SASSYS-1 coupling for uncertainty quantification and optimization analysis

Argonne has been leading the safety analysis of Sodium-cooled Fast Reactors under protected and unprotected transient events by developing and maintaining the system code SAS4A/SASSYS-1. In these analyses, some assumptions must be made for the models because of the uncertainties related to the heat transfer systems; the propagation of nuclear data uncertainties also has significant impact on the reactivity coefficients. In order to estimate the safety related parameters (e.g. coolant boiling margin, peak fuel temperature) more accurately, SAS4A/SASSYS-1 has recently been coupled with uncertainty quantification toolkits, including Dakota and RAVEN. The objectives of this paper are to present the coupling and demonstrate the capabilities of the uncertainty quantification and design optimization. The unprotected transients of Argonne's Advanced Burner Test Reactor (ABTR) were examined in this study. A sensitivity analysis was conducted within the uncertainty domain considering five uncertain parameters. It was found that core radial expansion has the most significant impact on the safety performance because of the large negative feedback during the Unprotected Loss of Heat Sink (ULOHS). The uncertainty quantification of the minimum boiling margin and the peak fuel temperature is discussed in response to the uncertainty propagation. Beyond the conventional sampling techniques, two additional advanced sampling techniques (i.e. Importance Sampling and Reliability Method) were tested. Another objective of the coupling is to optimize the design for the potential to improve the safety performance. The peak fuel temperature during the ABTR ULOHS is minimized by perturbing the Doppler feedback coefficient and radial expansion feedback coefficient. Three optimizers supported by Dakota have been tested and finally a hybrid method is recommended for the SAS4A/SASSYS-1 system optimization in the future due to its efficiency for the global optimum. (authors)