A neural network-based simulation metamodel for a process parameters optimization: A case study

Simulation is a widely accepted tool in complex systems design and analysis. However, it is essentially a trial-and-error approach, and is therefore, time-consuming and does not provide a method for optimization. Metamodeling techniques have been recently pursued in order to tackle these drawbacks. The main objective has been to provide robust, fast decision support aids to enhance the overall effectiveness of decision-making processes. This paper proposes an application of simulation metamodeling through artificial neural networks (ANNs). The proposed approach is composed of two main steps assisted successively by the ©Neuro Software. The first one consists of building the appropriate ANN model over second-order linear regression model. Based on this model, the second step is a reverse simulation metamodeling that will be used as an optimization tool. To validate the proposed approach, a real case study is adopted from literature (Yang and Tseng [1]). It concerns an anonymous integrated-circuit (IC) packaging company in Taiwan. The case study problem is to maximize throughput performance. The comparative results with Response Surface Methodology (RSM) based metamodel results; illustrate the efficiency and effectiveness of the proposed approach.

[1]  R. D. Hurrion.  Using a Neural Network to Enhance the Decision Making Quality of a Visual Interactive Simulation Model , 1992 .

[2]  Russell R. Barton,et al.  Chapter 18 Metamodel-Based Simulation Optimization , 2006, Simulation.

[3]  Alberto Prieto.  Artificial Neural Networks , 1991 .

[4]  R. D. Hurrion,et al.  A comparison of factorial and random experimental design methods for the development of regression and neural network simulation metamodels , 1999, J. Oper. Res. Soc..

[5]  Larry J. Shuman,et al.  Computing confidence intervals for stochastic simulation using neural network metamodels , 1999 .

[6]  Taho Yang,et al.  Solving a multi-objective simulation model using a hybrid response surface method and lexicographical goal programming approach—a case study on integrated circuit ink-marking machines , 2002, J. Oper. Res. Soc..

[7]  Adedeji B. Badiru,et al.  Neural network as a simulation metamodel in economic analysis of risky projects , 1998, Eur. J. Oper. Res..

[8]  Daniel J. Fonseca,et al.  Simulation metamodeling through artificial neural networks , 2003 .

[9]  Simon Haykin,et al.  Neural networks , 1994 .

[10]  Mansooreh Mollaghasemi,et al.  The development of a methodology for the use of neural networks and simulation modeling in system design , 1999, WSC '99.

[11]  Lefteri H. Tsoukalas,et al.  Fuzzy and neural approaches in engineering , 1997 .

[12]  Mary Lou Padgett,et al.  Neural networks and simulation: Modeling for applications , 1992, Simul..

[13]  Kurt Hornik,et al.  Multilayer feedforward networks are universal approximators , 1989, Neural Networks.

[14]  Robert Allen Kilmer.  Artificial neural network metamodels of stochastic computer simulations , 1994 .

[15]  George Chryssolouris,et al.  The use of neural networks in determining operational policies for manufacturing systems , 1991 .