The optimization of simulation models by genetic algorithms:a comparative study

This dissertation is a comparative study of simulation optimization methods. We compare a new technique, genetic search, to two old techniques: the pattern search and the response surface methodology search. The pattern search uses the Hooke and Jeeves algorithm and the response surface method se...

Full description

Bibliographic Details
Main Author: Yunker, James M.
Other Authors: Industrial and Systems Engineering
Format: Others
Language:en
Published: Virginia Tech 2014
Subjects:
Online Access:http://hdl.handle.net/10919/38929
http://scholar.lib.vt.edu/theses/available/etd-07282008-135041/
Description
Summary:This dissertation is a comparative study of simulation optimization methods. We compare a new technique, genetic search, to two old techniques: the pattern search and the response surface methodology search. The pattern search uses the Hooke and Jeeves algorithm and the response surface method search uses the code of Dennis Smith. The research compares these three algorithms for accuracy and stability. In accuracy we look at how close the algorithm comes to the optimum. The optimum having been previously determined from exhaustive testing. We evaluate stability by using the variance of the response function as determined from 50 searches. The test-bed consists of three simulation models. We took the three simulation models from text books and modified them to make them optimization models if that was required. The first model consists of a big S, little s inventory system with two decision variables: big S and little s. The response is the monthly cost of operating the inventory system. The second model was a university time-sharing computer system with two decision variables: quantum, the amount of time that the computer spends on a job before sending it back to the queue and overhead, that is the time that its takes to execute this routing operation. The response was the cost of operating the system determined from a cost function. The third model was a job-shop with five decision variables: the number of machines at each of the five work stations. The response was the cost of operating the job-shop again determined from a cost function. The decision variables were integer for the inventory system and job-shop, and were real for the computer system. === Ph. D.