الفهرس 
IntroductionOnly 14 pages are availabe for public view
 |  | from | 96 |  |  |
| | | from | 96 | | |
Abstract
Simulation systems are used extensively as models of real systems to evaluate output responses. The choice of optimal simulation parameters can lead to improved operation, but configuring them well remains a challenging problem. Actually, creating applicable stochastic optimization methods has been revived to confront designing simulation software since all recent simulation software packages contain optimization procedures. This may explain the common use of the term “simulation-based optimization” instead of “stochastic optimization”. Simulation-based optimization (SBO) is an emerging field which integrates optimization techniques into simulation analysis. Optimization problems that arise in SBO are formulated as stochastic programming problems whose objective functions are an associated measurement of an experimental simulation. The classical non-linear programming techniques may fail to solve such problems, due to the complexity of the simulation and the uncertainty of the problem. Actually, the objective function usually has the properties: (a) subject to various levels of noise, (b) not necessarily differentiable, and (c) computationally expensive to evaluate. In this study, new hybrid versions of Evolutionary Algorithms (EAs) are proposed as promising solvers for the SBO problem. The proposed methods aim to overcome the drawbacks of slow convergence and random constructions of EAs. In those hybrid methods, different diversification and intensification schemes are inlaid in order to compose more intelligent global search methods. Specifically, the diversification schemes aim to escape from the detected local minima and to direct the search process to explore new regions. On the other hand, the intensification schemes aim to accelerate the search process by using faster local search strategies. The proposed methods, as evolutionary-based methods, are derivative-free methods. Moreover, direct search methods, which are also derivative-free methods, are invoked to play the role of local search in the proposed hybrid methods. Therefore, the hybrid methods proposed in this study confront SBO problems in which the gradient information is not available. Another important feature of the proposed methods is to use replicated samples to reduce the noise level. Several experiments had done to test the performance of the proposed methods through several benchmark problems.