In view of practical significance of a correlation of critical heat flux (CHF) in the aspects of engineering design and prediction, this study is aiming at evaluation of existing CHF correlations for flow boiling of water with available databases taken from small-diameter tubes, and then development of a new, simple CHF correlation. Available CHF databases in the literature for flow boiling of water in small-diameter tubes (0.33 < D-h < 6.22 mm) are collected, covering wide parametric ranges. Three correlations by Bowring, Katto and Shah are evaluated with the CHF data for saturated flow boiling, and three correlations by Inasaka-Nariai, Celata et al. and Hall-Mudawar evaluated with the CHF data for subcooled flow boiling. The Hall-Mudawar correlation and the Shah correlation seem to be the most reliable tools for CHF prediction in the subcooled and saturated flow boiling regions, respectively. In order to avoid the defect of predictive discontinuities often encountered when applying previous correlations, a simple, nondimensional, inlet conditions dependent CHF correlation for saturated flow boiling has been formulated. Its functional form is determined by the application of the artificial neural network and parametric trend analyses to the collected database. Superiority of this correlation has been verified by the database. The new correlation has a mean deviation of 16.8% for this collected databank, smallest among all tested correlations. Compared to many inordinately complex correlations, this new correlation consists only of a single equation. (c) 2005 Elsevier Ltd. All rights reserved.