Date of Award


Document Type


Degree Name

Master of Science


Department of Operational Sciences

First Advisor

James S. Shedden, PhD


Simulated Annealing was used to optimize three constrained simulation models. For each of these models, seven different acceptance functions were evaluated and compared against the performance of Local Search. These comparisons demonstrated the affect that different acceptance functions have on the performance of the algorithm. The performance was measured by the average solution quality and average efficiency obtained from several runs. The first model facilitated the implementation of Simulated Annealing using the SLAM simulation language. The configuration space was small, described by only two decision variables. It demonstrated the viability of using Simulated Annealing to optimize the variable settings in a simulation model. The second model, with six decision variables, provided greater insight to the advantages and limitations of Simulated Annealing. This model was implemented as an open queuing network. The third model, similar to the second, was implemented as a closed queuing network. The results from this variation were completely unexpected. They showed a wide performance separation among the different acceptance functions that was not present in the first two models. No attempt was made to justify the use of Simulated Annealing from a theoretical perspective. Rather, empirical results from the three models were used to infer the practical utility of the algorithm.

AFIT Designator


DTIC Accession Number



The author's Vita page is omitted.