Comparison of response surface model and Taguchi methodology for robust design

• Main Author: Sudasna-na-Ayudthya, Prapaisri
• Other Authors: McDowell, Edward D.
• Language: en_US
• Published: 2013
• Subjects: Response surfaces (Statistics); Taguchi methods (Quality control); Experimental design
• Online Access: http://hdl.handle.net/1957/37196

The principal objective of this study was to compare the results of a proposed method based upon the response surface model to the Taguchi method. To modify the Taguchi method, the proposed model was developed to encompass the following objectives. The first, with the exception of the Taguchi inner array, was obtain optimal design variable settings with minimum variations, at the same time achieving the target value of the nominal-the best performance quality characteristics. The second was to eliminate the need for the use of a noise matrix (that is, the Taguchi outer array), resulting in the significant reduction of the number of experimental runs required to implement the model. The final objective was to provide a method whereby signal-tonoise ratios could be eliminated as performance statistics. To implement the proposed method, a central composite design (CCD) experiment was selected as a second-order response surface design for the estimation of mean response functions. A Taylor's series expansion was applied to obtain estimated variance expressions for a fitted second-order model. Performance measures, including mean squared error, bias and variance, were obtained by simulations at optimal settings. Nine test problems were developed to test the accuracy of the proposed CCD method. Statistical comparisons of the proposed method to the Taguchi method were performed. Experimental results indicated that the proposed response surface model can be used to provide significant improvement in product quality. Moreover, by the reduction of the number of experimental runs required for use of the Taguchi method, lower cost process design can be achieved by use of the CCD method. === Graduation date: 1993