b'Figure 4. Cumulative distribution function (CDF) for the PCT.PCT. That is, it is 95% confident thatresults with a slightly different PCT the maximum PCT from the 59limit of 619.65 K. calculations bounds at least 95% of theFor the SA, DAKOTA calculated the PCT population distribution. Figure 2response correlations for the FOM. shows the calculated PCT versus theThe correlation of interest is the partial iteration index (1 through 59) usingcorrelation coefficient. It reflects the the Latin Hypercube sampling method.correlation between two variables (any Figure 3 compares the PCT fromof the sensitivity parameters and the three cases, the maximum PCT, thePCT) after adjusting for (i.e., removing) minimum PCT and the baseline casethe effects of other variables. The partial (no uncertainty).The results indicatecorrelation coefficients indicate the that the chosen sensitivity parametersPCT has a strong positive relation to with their probability distributionscore power, a strong negative rela-do not have significant impact on thetion to the fuel thermal conductivity, PCT. This is not unexpected becausea mildly negative relation to both the of the mild transient as a result of thegap conductance and the clad thermal LOOP (a primary pump trip followedconductivity, and a much smaller by reactor trip). Figure 4 shows therelation to the specific heat of the fuel CDF for the PCT. The 95/95 PCTand the cladding. The SA results suggest limit corresponds to the maximumthat for the assumed uncertainties the PCT calculated for the 59 parametricPCT for a LOOP is influenced mainly by cases and is 619.72 K. A second UQfuel parameters that raise the initial fuel calculation using the Monte-Carlotemperature. The partial rank correla-sampling method resulted in similartions suggest the same conclusion.2019|AFC ACCOMPLISHMENTS 155'