Title: Decision models for process improvement using combined response surface modelling and multi-objective optimisation
Authors: Mohamed H. Gadallah
Addresses: Department of Operations Research, Institute of Statistical Studies and Research (ISSR), Cairo University, Giza 12613, Egypt
Abstract: Response surface methodology (RSM) is often utilised to develop a mathematical model based on experimental results. We particularly investigate the role data sequence on the resulting modelling error. A response surface model based on UL92 (92 data points) is developed. A procedure is given to vary four models (UL8-2, UL27-2, UL25 and UL32 = UL92) in a deterministic manner to quantify the sequence effect on modelling error. Comparison of different models in terms of maximum/minimum error indicates: (a) high accuracy relative to experimental results and (b) the resulting combined models are sequence dependent. Three optimisation formulations are proposed. The first is a multi-objective optimisation problem subject to the natural constraints of process variables. The second is minimisation of the standard deviation of force subject to limits on mean force and natural constraints of process variables. The third is minimisation of difference between maximum and minimum force subject to limits on mean and standard deviation and process limits. The three formulations resulted in operating conditions least sensitive to process variations.
Keywords: response surface methodology; RSM; process modelling; multi-objective optimisation; process improvement; mathematical modelling.
DOI: 10.1504/IJISE.2011.038565
International Journal of Industrial and Systems Engineering, 2011 Vol.7 No.2, pp.165 - 194
Published online: 31 Jan 2015 *
Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article