RESPONSE SURFACE METHODS (RSM) ACHIEVE （响应面方法(RSM)实现）.pdfVIP

RESPONSE SURFACE METHODS (RSM) ACHIEVE DESIGN FOR SIX SIGMA (DFSS) GOALS FOR MEDICAL DEVICE MANUFACTURING By Mark J. Anderson, PE, CQE, Stat-Ease, Inc. (mark@) INTRODUCTION Using response surface methods (RSM) for statistical design of experiments (DOE),1,2 engineers at a major medical device manufacturer (USA-based) successfully modeled a key process for their flagship product. The RSM model then became the foundation for development of robust specifications to ensure quality at six sigma levels. Furthermore, RSM optimization tools pinpointed potential for doubling the production rate while halving the variation in a critical attribute, thus meeting objectives of the company’s design for six sigma (DFSS) program. DETAILS ON THE EXPERIMENT DESIGN The manufacturing team performed a 50-run, Box-Behnken design (BBD) on five critical process factors known to affect their results. For proprietary reasons, the factors cannot be revealed, but they were selected from experiments that screened such things as pressures, speeds, distances, flows and environmental factors that influenced product performance. The BBD is a popular template for RSM because it requires only three-levels of each process factor and only a fraction of all the possible combinations. Details on RSM, and the BBD in particular, can be found in reference 1. Figure 1 shows a BBD on three factors. Figure 1: Box-Behnken design on three factors The BBD is comprised of planes intersecting at a center point with outlying points at the extreme vertices in each of the factor dimensions. This selection of points suffices to fit a second-order polynomial equation, called a “quadratic model” in the parlance of RSM. The quadratic generally provides an adequate fit of data from experiments done for purposes of process optimization. ANALYZING THE RESULTS AND INTERPRETING THEM Four key responses, coded as “Y ”, were measured on a number of parts made at each proc