Simulation Optimization by Genetic Search: A Comprehensive Study with Applications to Production Management

摘要

In this report, a relatively new simulation optimization technique, the genetic search, is compared to two more established simulation techniques- the pattern search and the response surface methodology search. The pattern search uses the Hooke-Jeeves algorithm, and the response surface methodology search uses the computer code of Dennis Smith. The three algorithms are compared for both accuracy and stability. Accuracy is evaluated in terms of how close each algorithm comes to the optimum, the optimum having been previously determined from exhaustive testing. Stability is evaluated using the variance of the response function determined from sample searches-the lower the variance, the more stable the response. The examples tested are an inventory system with integer decision variables, a university time-sharing computer system with two real decision variables, and a job-shop with five decision variables (the number of machines located at each station). The response of interest for each system is the cost of operating the system. The genetic algorithm is shown to be a superior optimization method compared to the two other search techniques.