Loading...

Two New Meta-Model Based Artificial Neural Network Algorithms for Constrained Simulation Optimization Problems with Stochastic Constraints

Mohammad Nezhad, Ali | 2011

702 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 41786 (01)
  4. University: Sharif University of Technology
  5. Department: Industrial Engineering
  6. Advisor(s): Mahlooji, Hashem
  7. Abstract:
  8. Following the recent developments in the field of decision making, a considerable number of problems involved with stochastic systems can be thought of whose analysis depends on a set of intricate mathematical relations. In such cases, simulation is one of the most popular tools that can be applied toward analysis of behavior of such stochastic systems. Not only does not the simulation model rely on such intricate mathematical relations, it also enjoys the added advantage of being free of any restricting assumptions which may normally be considered in a stochastic system.To analyze such problem, one may aim at determining the best combination of input variables to optimize the system performance criterion. Doing so when dealing with a stochastic system bears a striking resemblance to the common practice in the jargon known as simulation optimization. The simulation optimization employs an algorithm to determine the optimal values of parameters of a stochastic system.This work presents two new algorithms for solving continuous constrained simulation optimization problems with stochastic constraints. These algorithms especially address the simulation optimization problems for which the objective function and the stochastic constraints are evaluated by running expensive simulation experiments. The proposed algorithms are developed based upon a sequential approach to approximate the simulation optimization problem by a deterministic mathematical programming problem. In fact, this approach aims at estimating the global optima through solving a number of deterministic models. A nonlinear optimization algorithm is developed to tackle the aforementioned deterministic models.The proposed simulation optimization algorithms rely on very few number of simulations runs. This issue can be a crucial factor for a simulation optimization algorithm as the most real world problems have expensive simulation models
  9. Keywords:
  10. Simulation Optimization ; Artificial Neural Network ; Sequential Experimental Design ; Revied Particle Swarm Optimization Based Discrete Lagrange Multipliers Method (RPSO-DLM)

 Digital Object List

 Bookmark

No TOC