Response surface kullanan sağlamcı eniyileme

Abstract: During the last couple of decades, optimization methods have become one of the most important research fields and received increasing attention from engineers, product designers and researchers especially in the field of production and related industries.The classical optimization approaches assume that the input data of the optimization model is known with certainty. However, in many areas of application of real world problems like inventory management, portfolio selection, supply chain optimization and production planning, it?s needed to integrate the uncertainty of the input data into the optimization model which refers to optimization under uncertainty.The increasing interest in simulation optimization for problems that arise in practical applications becomes relevant where explicit mathematical formulations are too restrictive. Therefore, for many practical cases one cannot obtain an analytical solution through those kind of methods. Indeed, simulation optimization has led to the numerical solution of large-scale, real-world decision-making problems.A simulation optimization method, Response Surface Methodology (RSM); aims to achieve optimum operation conditions while minimizing the variability in order to produce high quality and reliable products and services at the lowest possible cost. As an extension of Robust Parameter Design (RPD), RSM is a combination methodology of mathematical and statistical techniques in problem modeling and analysis.The risk-neutrality problem of the classic simulation optimization problems can be handled by the Dual Response Surface (DRS) approach within RSM and combining it with the Taguchi?s RPD enables researchers to cope with the unknown environments.In this work, a risk-averse approach to Response Surface Methodology, which explicitly deals with random environments, is presented. The main contribution of this thesis is to adapt Taguchian RSM to discrete-event simulation studies.The thesis introduced the steps of this Taguchian RSM approach and then an application of these steps to an inventory optimization is provided. The computer program is coded in Matlab 7.6. , and the optimization is performed through the built-in function fmincon in Matlab.Furthermore, Taguchian RSM method is applied to a more complex example which is a call center problem modeled in Arena. The results taken from the execution of the model is used in our optimization algorithm coded in Matlab.Although it?s usefull to increase the number of decision variables in the example, because of the version limits of Arena, an example with an additional environmental factor is provided in order to expand the original example. Thus we left this issue as a future research.For the future work, this study can be extended to an iterative approach, or the proposed approach can be developed to handle multiple random responses.Keywords: Robust parameter design, Discrete-event simulation, Dual response surface