The transition between trickle flow and pulse flow regimes in cocurrent trickle-bed reactors is not properly predicted by existing phenomenological, semi-theoretical and empirical tools. Based on the most complete flow regime transition data base (700 measurements, 30 gas-liquid systems, 18 columns diameters, 38 packing materials, high pressure conditions, coalescing, non-coalescing and pseudoplastic non-Newtonian aqueous and organic liquids), a state-of-the-art explicit correlation of trickle-to-pulse flow changeover was derived relying on neural network modeling. Robustness of the proposed correlation was verified, and the limitations of the literature correlations and models were demonstrated through systematic statistical testing over the constructed data base. The overall result was a net improvement in predicting the trickle-to-pulse flow regime transition.