Modification of the simulated annealing optimization technique and its application to both a dynamic control and gas tagging problem

Advanced optimization techniques, based on analogies related to physical systems rather than on classical mathematical theory, are becoming more widely used than ever before. One such type of technique is simulated annealing, a Monte Carlo (stochastic) method. Although it has been used primarily for the solution of combinatorial optimization problems, it is just starting to be applied to problems with continuous domains as well as to linear programming problems.This dissertation investigates the simulated annealing technique and its application to problems other than those of the combinatorial optimization type. The first problem implements a modified simulated annealing type algorithm for the solution of a dynamic control problem in position and velocity space. The second one involves a nuclear engineering cost minimization problem for gas tagging, a promising way of detecting leaks while on line. The specific problem solved is a linear one. These implementations allow the technique to be benchmarked for both problem types and, thereby, to check the robustness of the algorithm.The acceptance criteria of the algorithm, namely, the use of the Boltzmann distribution, is also investigated. A modification to this criteria is accomplished by implementing a two-tailed alternative distribution, the Fermi-Dirac distribution.U of I OnlyETDs are only available to UIUC Users without author permissio