Design of experiments (DOE) is an effective method providing useful information about the interaction of operating variables and the way the total system works by using statistical analyses. However, its industrial application is limited because it is almost difficult to maintain variables in DOE matrix at desired constant levels in industrial environment. Thus, this paper aims to present a new mixed modeling method which is a combination of fuzzy logic and design of experiments methods to overcome such practical limitations. The method first uses a fuzzy model which is trained by practical data gathered from industry to predict DOE response corresponding to each run in DOE matrix. Then, a statistical parametric model is constructed for the prediction of process response to any change of operating parameters under real industrial conditions. The proposed mixed method was successfully validated by using data obtained from a coal hydraulic classifier at Zarand Coal Washing Plant (Kerman, Iran). The method also seems to be a promising tool for modeling all devices and processes in real industrial environment and allows researchers to benefit from all the advantages of experimental design and fuzzy logic methods simultaneously.