A novel intelligent-mechanistic model was developed to understand the behavior of multiphase chemical reactors. Computational fluid dynamics (CFD) and an intelligent algorithm were combined to predict different levels of 3D cylindrical bubble-column reactors. An adaptive neuro-fuzzy inference system (ANFIS) was used as the intelligence algorithm, and different ANFIS parameters were evaluated. With about one third of the training data the method can predict the overall behavior of the gas fraction in the reactor. A number of rules significantly influence the accuracy of the ANFIS method. After finding appropriate parameters, the method is applied for prediction of points which are not simulated with CFD, representing ANFIS mesh refinement. Also, bubble-column reactors without training of exact values of measured data or numerical results can be predicted. Main advantages are time savings and reduction of computational expenses.