Authors: Mohamed H. Gadallah
Addresses: Department of Operations Research, Institute of Statistical Studies and Research (ISSR), Cairo University, Giza, Cairo 12613, Egypt
Abstract: A Response Surface Methodology (RSM) is developed from experimental results. Standard and fractional orthogonal arrays are employed to develop one model out of 100 data points. Another modified model is developed based on the splitting approach and iterative process involving analysis of variance and significance levels. The objective is to improve the model in certain domains instead of making global changes. The two models are compared via several statistical measures. The two models are called UL100 and UL100-s. The results indicate the validity of the splitting approach to achieve higher accuracy than the regular RSM models.
Keywords: RSM; response surface methodology; process modelling; ANN; artificial neural networks; orthogonal arrays; splitting; ANOVA.
International Journal of Mechatronics and Manufacturing Systems, 2010 Vol.3 No.1/2, pp.58 - 78
Available online: 02 Dec 2009 *Full-text access for editors Access for subscribers Purchase this article Comment on this article