An approach to simulation optimization of industrial systems with discrete design variables and stochastic behavior

A sequential approximate optimization approach is proposed for simulation models with discrete design variables and stochastic behavior. Linear response surface approximations of objective function and constraints are built in a search subregion of the design space based upon simulation experiments according to a D-optimal experimental design. An integer linear programming algorithm is used to calculate the approximate optimum design in the search subregion. The approach is illustrated for a non-convex analytical test problem and a simulation model of a four-station production flow line