Considerations for comparing CHF correlations

Accuracy is a key consideration when selecting a critical heat flux (CHF) correlation for use in DNB safety analyses of pressurized water reactors (PWR). The desire is to minimize the amount of power margin which must be set aside to cover the uncertainty of the correlation. The accuracy of a CHF correlation for a particular fuel design may be determined by evaluating its ability to predict DNB conditions in tests performed with rod bundle geometries which simulate the production fuel assemblies. The results of such evaluations are typically tabulated in terms of the average measured-to-predicted (or predicted-to-measured) critical heat flux ratio (CHFR) or critical power ratio (CPR) for the data, and the corresponding sample standard deviation. PWR fuel vendors have historically chosen to express the accuracy of their correlations using CHFR statistics. This is a logical choice, since DNB safety analysis are performed using critical heat flux ratios. Recently a number of investigators have attempted to compare the accuracies of different correlations for predicting publicly available DNB test data. In a few cases a comparison of CHFR statistics was used to rank the various correlations. It will be shown in this paper that such an approach can be misleading, and will not always lead to the selection of the most accurate correlation. (author)