A Discrete Nonlinear Simplex Method for Optimized Engineering Design

This paper discusses the use of a modified nonlinear simplex method for constrained engineering optimization problems in which some or all of the variables can only take on discrete values. The objective function and constraints need not be analytic expressions; it is only necessary that their values be computable. The modifications to the nonlinear simplex method include the incorporation of unidimensional search, new acceleration and regeneration methods, and the exploration of decomposition strategies. The algorithm has been successfully tested using problems selected to represent a variety of engineering design applications.