Sensitivity Analysis on the DYNASTY Facility 1D Modelica Models with the DAKOTA-FMU Approach

An in-depth understanding of the natural circulation phenomenon is essential for developing and operating future generation nuclear reactors, which will mostly rely on natural circulation safety systems. For example, the molten salt fast reactor concept will adopt passive systems for decay heat removal in nominal and accidental scenarios. Due to the unique features of this reactor design, including internal heat generation and coupled free convection–driven loops, the DYnamics of NAtural circulation for molten SalT internally (DYNASTY) heated facility was built at the Energy Labs of Politecnico di Milano. Several experiments and numerical studies have been made at the facility; however, an in-depth sensitivity analysis on all parameters influencing natural convection has not yet been performed. Indeed, the complete understanding of the natural circulation phenomenon in such a complex system requires a detailed study of the influence of the input parameters on the expected outputs of the system. This work aims to fill this gap by proposing a sensitivity analysis framework for the DYNASTY facility by coupling the DAKOTA Sandia code with the one-dimensional Modelica models developed and validated in previous works by the authors. In particular, this study includes both a global sensitivity analysis based on the Sobol indices and a local sensitivity analysis. At first, the analysis considers the two inputs of the facility, namely, the heating power and the cooler air flow rate, and their interaction under different heating configurations. Then other parameters of interest, such as the ambient temperature, the friction factor, and the heat transfer coefficient, are added to the analysis. The results will show how the heating power is the most significant control variable for the mass flow rate; conversely, parameters such as the cooler air flow rate and the heat transfer coefficient seem to have the highest influence on the fluid temperature. This analysis will allow for further refinement of the numerical models for the facility to achieve a better representation. Additionally, the framework has been developed in such a way to be easily extendable to other system codes, such as RELAP5 and TRACE.