Second-order response surface designs in computer simulation

Publisher Website
Google Scholar
Abstract
Computer simulation models are frequently used to determine the combination of levels for a set of independent variables at which a response variable is optimized. This process leads to problems of experi mental design, i.e., the choice of levels of the independent variables at which the simulation model is to be run and at which the model's responses are to be observed. This paper considers second-order response-surface designs and optimization procedures for use in this situation. Several classes of designs are constructed and applied to surfaces typically generated by computer simulation models. Results are obtained which should assist in the selection of de signs in general or for use on specific surfaces. An application of second-order response-surface de signs to a digital simulator of a street network is presented; the object is to find optimal settings for the traffic signals during the morning peak of traffic flow.