The application of design of experiments method in modeling for complex system

Industrial production process is usually a complex nonlinear system. An accurate control scheme is based on a feasible mathematic model. In recent years, data-driven modeling method is extensive used. In this method, data selection is the crucial step. In this paper, the statistics knowledge is applied to system modeling. Firstly, design of experiments methods such as uniform design and Latin hypercube design are used in the sample selection; then, mathematical fitting method, neural network method and fuzzy model method are used for nonlinear system modeling based on the sample data respectively. Finally, error analysis shows that design of experiments method can minimize the sample size and guarantee the model precision.