Micro-Computer Based Decision Support System for Response Surface Methodology.

摘要

Currently it can be difficult to conduct response surface methodology analysis because it requires computer assistance and a thorough understanding of the mathematics and statistical theory involved. The difficulty of RSM analysis would be greatly reduced if a system existed that integrated the RSM mathematics and the RSM decision making process into a single computer program. The purpose of this thesis was to develop PCRSM, a personal computer based decision support system (DSS) for response surface methodology (RSM) analysis. This study focused on the experimental design and regression elements of the RSM analysis process. Experimental design is conducted through a series of questions, choices, and advice regarding the numbers of factors, design choice, resolution, number of center points, and number of design replications. PCRSM includes 60 designs for 2 to 39 factors. It supports six different two and three level experimental designs, including full and fractional designs, Plackett-Dehnken-Burman, designs, central composite designs, and Box-Behnken designs. Least squares linear regression is used to determine the meta-model of the response surface. Output provided by the regression package includes residuals plots, two different ANOVA tables, coefficient tables, variance-covariance matrices, correlation matrices, and the design matrix and responses. (kr)