Systems' optimization through simulation

Abstract

The problem of systems' optimization through simulation was investigated. An external modular optimization procedure applicable to the simulation model of a multiple-variable, single-response, system whose overall performance is defined in terms of cost or profit in designed. The procedure consists of four modules: The simulation model, A sequential search technique, A "Sampling Routine", and A "Control Module". The first two modules were treated as "black-boxes" and were assumed to be given. The third module controls the sampling process. This module offers dynamic adjustments. of the necessary sample size at any point in the course of the search for an optimal system's design. The fourth module controls the optimization process. This module utilizes a statistical inferential technique in determining the duration of the optimization process, such as to maximize the net return from its application.

The procedure offers the possibility of utilizing various sequential search techniques, requires no modifications of the simulation model, and can be applied to a reasonably broad class of optimization problems.

Descriptions of the procedure's modules, as well as illustrative examples of their implementation are included.