An Assessment of Prediction Methods of CHF in Tubes with A Large Experimental Data Bank

Abstract

An assessment of prediction methods of CHF in tubes has been carried out using an expanded CHF data bank at Chalk River Laboratories (CRL). It includes eight different CHF look-up tables (two AECL versions and six USSR (or Russian) versions) and three empirical correlations. These prediction methods were developed from relatively large data bases and therefore have wide range of application. Some limitations, however, were imposed in this study, to avoid any invalid predictions due to extrapolation of these methods. Therefore, these comparisons are limited to the specific data base that is tailored to suit the range of an individual method. This has resulted in a different number of data used in each case. The comparison of predictions against the experimental data is based on the constant inlet-condition approach (i.e., the pressure, mass flux, inlet fluid temperature and tube geometry are the primary parameters). Overall, the AECL tables have the widest range of application. They are assessed with 21 771 data points and the root mean-square error is only 8.3%. About 60% of these data were used in the development of the AECL tables. The best version of the USSR/Russian CHF table is valid for 13 300 data points with a root-mean-square error of 8.8%. The USSR/Russian table that has the widest range of application covers a total of 18 800 data points, but the error increases to 9.3%. The range of application for empirical correlations, however, are generally much narrower than those covered by the CHF tables. The number of data used to assess these correlations is therefore further limited. Among the tested correlations, the Becker and Persson correlation covers the least amount of data (only 7 499 data points), but has the best accuracy (with a root-mean-square error of 9.71%).