Experimental results on the flooding capacity of randomly dumped packed beds were collected from the literature to generate a working database. The reported measurements were first used to review the accuracy of existing predictive tools in that field. A total of 14 correlations were extracted from the literature and cross-examined with the database. Many limitations regarding the level of accuracy and generalization came to light with this investigation. Artificial neural network modeling was then proposed to improve the broadness and accuracy in predicting the flooding capacity, which is an important design parameter for packed towers. A combination of six dimensionless groups, namely, the Lockhart-Martinelli parameter (chi); the liquid Reynolds (Re-L), Galileo (Ga-L) and Stokes (St(L)) numbers; the packing sphericity (phi); and one bed number (SB) outlining the tower dimensions were used as the basis of the neural network correlation. With an initial database containing 1019 measurements, the correlation yielded an absolute average relative error (AARE) of 16.1% and a standard deviation of 20.4%. Another database containing over 100 measurements on the flooding capacity was used to validate the correlation. The prediction based on these results yielded an AARE of 11.6% and a standard deviation of 13.7%. Through a sensitivity analysis, the Stokes number in the liquid phase was found to exhibit the strongest influence on the prediction, while the liquid velocity, gas density, and packing shape factor were determined to be the leading physical properties defining the flooding level. As a matter of fact, the neural correlation remains in accordance with the design recommendations and trends reported in the literature.