Fitting a mixture-based response surface using computer simulation

In this paper, an augmented simplex-lattice mixture design is used to fit a mathematical metamodel for predicting work in process in a flexible manufacturing cell. The average work in process is modeled as a response variable by varying the incoming percentages of different products. Computer simulation is utilized to run different experiments to obtain the required data for fitting the metamodel. The accuracy of the metamodel predictions warrant its use in place of the simulation model, thus eliminating the need for lengthy simulation runs and complicated output analysis. This work provides an example of the utility of mixture experiments in a discrete manufacturing environment.