Optimization Techniques in Response Surface Methodology for Quality Improvement

9988563戴尔·卡斯蒂略这项研究的目标是开发用于响应面方法(RSM)的新的高效优化技术。RSM是一套统计和优化技术，旨在通过顺序应用设计的实验和建模技术来改善制造过程的质量特征。这项研究的具体目标包括1)在大样本变异性存在的情况下开发新的统计搜索方法；2)针对RSM研究中经常出现的二次规划问题，包括多个二次响应的情况，开发新的全局优化算法。将研究为使用多项式回归技术建模的制造过程寻找最佳操作设置的置信度区域的方法。最后，将开发一种快速响应面方法，允许快速优化多个响应过程。这项研究的结果将是一套新的工业实验优化技术，将受到目前使用现有RSM优化技术的工艺和质量工程师的欢迎。这将通过考虑工业实践中实验优化的特殊特点来实现，即其序贯性质、高抽样变异性的存在、多个响应的存在以及在昂贵的过程中快速优化的需要。后者将引起资本密集型制造商的兴趣，在这些制造商中，鉴定或优化工艺所需的实验次数应保持在最低限度。与工业研究人员(朗讯技术公司和SmithKline Beecham)以及宾夕法尼亚州立大学纳米制造设施的合作将为该项目中开发的技术提供试验台。宾夕法尼亚州立大学IE部门的新总部莱昂哈德大楼的制造实验室将允许在更传统的制造过程中进行测试。将开发易于使用的软件，以促进技术转让。

9988563Del CastilloThe objective of this research is to develop new and efficient optimization techniques for use in Response Surface Methodology (RSM). RSM is a set of Statistical and optimization techniques aimed at improving the quality characteristics of a manufacturing process via the sequential application of designed experiments and model building techniques. Specific goals of this research include 1) the development of new statistical search methods under the presence of large sampling variability; 2) development of new algorithms for the global optimization of the type of quadratic programming problems frequently arising in RSM studies, including the case of multiple secondary responses. Methods will be studied for finding a confidence region for the best operational settings of a manufacturing process that is modeled using polynomial regression techniques. Finally, 3) a Rapid Response Surface Methodology will be developed that will allow for fast optimization of multiple response processes. The outcome of this research will be a new set of optimization techniques for industrial experimentation that will be well-received by Process and Quality Engineers who currently use existing RSM optimization techniques. This will be accomplished by taking into consideration the particular characteristics of experimental optimization in industrial practice, namely, its sequential nature, the existence of high sampling variability, the presence of multiple responses, and the need for rapid optimization in expensive processes. This latter will be of interest to capital-intensive manufacturers where the number of experiments required to qualify or optimize a process should be kept to a minimum. Collaboration with industrial researchers (Lucent Technologies and SmithKline Beecham) and with Penn State's Nanofabrication facility will provide a testbed for the techniques developed in this project. The manufacturing laboratories at the Leonhard building, the new home of the IE department at Penn State, will allow testing in more traditional manufacturing processes. Easy to use software will be developed that will facilitate technology transfer.