Process-Oriented Basis Representations for Multivariate Process Diagnosis and Control

在传感设备的复杂性的进步导致生产系统，同时监测许多产品的质量特征。这些数据的可用性为过程监视和控制提供了很好的机会。这项研究是宾夕法尼亚州立大学和波多黎各大学合作进行的，研究如何将多变量质量数据中的模式与与某些生产问题或工艺参数相关的模式联系起来。研究人员将开发新的统计过程控制工具，以帮助质量工程师直接诊断生产不规范的可能原因。他们的方法论采用过程导向的基础方法，即质量测量向量与制造过程中的假设问题相关。对于给定的测量向量，该方法将识别一小组潜在原因，即具有大系数的基本元素。然后，流程工程师可以手动或自动方式选择适当的控制操作。所有制造业都认识到质量监测和控制是实现快速高效生产的关键因素。随着生产过程变得越来越复杂，不仅监控产品质量变得越来越重要，而且在质量指标下降时快速准确地调整工艺参数也变得越来越重要。本研究解决了多变量质量控制和过程整改中的基本统计问题，预计将对工程师准确诊断由过程变量引起的生产问题的能力产生相当大的影响。研究人员将与三家电子元件和/或系统制造商密切合作进行这项研究，从而确保研究与行业需求相关。

9700330 Barton Advances in the sophistication of sensoring devices have led to production systems that simultaneously monitor many product quality characteristics. The availability of these data provide a great opportunity for process monitoring and control. This research, a collaboration between the Pennsylvania State University and the University of Puerto Rico, investigates ways to link patterns in multivariate quality data with patterns associated with certain kinds of production problems or process parameters. The researchers will develop new statistical process control tools to assist quality engineers in directly diagnosing the likely causes of production irregularities. Their methodology uses a process-oriented basis approach, whereby quality measurement vectors are correlated with hypothesized problems in the manufacturing process. For a given measurement vector, the methodology will identify a small set of potential causes, i.e. those basis elements with large coefficients. Process engineers can then select appropriate control actions in either a manual or automated fashion. All manufacturing industries recognize that quality monitoring and control are critical elements in enabling fast and efficient production. As production processes become more sophisticated, it becomes increasingly important not only to monitor product quality, but also to quickly and accurately adjust process parameters when quality metrics degrade. This research addresses fundamental statistical problems in multivariate quality control and process rectification, and as such is expected to have considerable impact on the ability of engineers to accurately diagnose production problems caused by process variables. The investigators will work closely with three manufacturers of electronic components and/or systems in the conduct of this research, thus ensuring that the research is relevant to industry needs.