Combining two methods of Computational fluid dynamics ((CFD)) and design of experiments (DOE) was proposed in modeling to simultaneously benefit from the advantages of both modeling methods. The presented method was validated using a coal hydraulic classifier in an industrial scale. The effects of operating parameters, including feed flow rate, solid content and baffle length, were evaluated based on the classifier overflow velocity and cut-size as the process responses. The evaluation sequence was as follows: the variation levels of parameters was first evaluated using industrial measurement, and then a suitable experimental design was carried out and the DOE matrix was translated to (CFD) input. Afterwards, the overflow velocity values were predicted by (CFD), and the cut-size values were determined using industrial and (CFD) results. The overflow velocity and cut-size values were statistically analyzed to develop the prediction models for DOE responses; and finally, the main interaction effects were interpreted with respect to DOE and (CFD) results. Statistical effect plots along with (CFD) fluid flow patterns showed the effects of type and magnitude of operating parameters on the classifier performance, and visualized the mechanism by which those effects occurred. The suggested modeling method seems to be a useful approach for better understanding the real operational phenomena within the fluid-base separation devices. Furthermore, the individual interaction effects can also be identified and used for interpretation of responses in nonlinear processes.