Because of the importance of molten salt as a heat transfer fluid within the energy industry, the thermo-physical properties of the salts and the corresponding uncertainties must be well characterized. Prior work measuring the liquid density of FLiBe has focused on identifying the linear correlation between density and temperature as well as the thermal expansivity. Aggregation of past experimental efforts show a 11% variability among prior measurements of FLiBe liquid density, and a variability exceeding 40% for the thermal expansivity. Furthermore, neither the uncertainty of the thermal expansivity and density are consistently reported in the literature. To address this gap a new series of FLiBe density measurements and analysis procedure were designed such that the impact of measurement errors was propagated throughout the density and thermal expansivity calculations. The experiments were carried out according to the hydrostatic method from 400°C to 820°C with increments of temperature of 50 C. The temperature of the salt and the mass of a solid weight immersed in FLiBe were measured continuously; the density was calculated from the mass and temperature at each 50 C after the system reached thermal equilibrium. An average measurement at equilibrium accounted for measurement fluctuations. Neither simple linear regression nor weighted regression can incorporate the measured temperature error, therefore York regression was employed to account for error in both density and temperature. This is done iteratively by updating individual sample weights when determining the density and thermal expansivity of FLiBe by least squares. Welch’s T-test is employed to identify statistically significant differences between the regression parameters of the experiments to check for result consistency prior to aggregation or averaging. Comparing to prior data the run-to-run variability for density between the experiments was 0.5% and for thermal expansivity it was 8.5%. Averaging over the experiment results provides the following mean values and standard deviations for the intercept and thermal expansivity of 2245±7 kg/m<sup>3</sup> and 0.424±0.017 kg/(m<sup>3</sup>°C) which correspond to a percent deviation of 0.3% and 4.1%. In summary, given quantified experimental error statistical methods, such as Welch’s T-test and York regression, can be leveraged to provide more representative estimates and uncertainties for relevant FLiBe properties when compared to standard regression techniques.

Incorporating experimental uncertaintiy in density and thermal expansivity regression estimates of molten 2LiF-BeF2 (FLiBe)

ACS Fall 2021

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.