Faculty Early Career Development: Multivariate Process Monitoring with Correlated Data

9501903本研究的重点是理解过程的质量控制问题，其行为受到多个自相关变量的影响。传统的统计过程控制（SPC）是基于单一变量和正态性、独立性和均匀方差的假设。本研究利用了这些假设的不适当性，并使用单一变量作为理解和控制过程的基础，而实际上在今天的工业环境中可能存在多个自相关变量。研究的目的是系统地探讨在存在自相关多变量数据时，几种多变量质量控制方法的行为、灵敏度和一致性。这些方法包括多元EWMA、主成分分析和多时间序列。这项研究的动机是缺乏足够的分析工具来处理统计过程控制情况，其中多个相关因素影响过程的行为。该研究计划使用不同形式的模型来解决这个问题。研究的教育部分展示了研究者的教学理念，跟踪教学绩效的方法，以及学生对所教材料的理解和参与。研究者将通过开发一门新的统计学和质量控制相结合的课程，将研究成果应用到课堂上，重点是使用这些技术来解决统计过程控制中与行业相关的问题。此外，本研究以改善教学过程为目的所研究的技术，也将用于提高教学效果。这项研究的潜力是好的。本文的研究成果可以促进多元过程控制领域的知识体系的发展。这将导致对过程行为和控制方法的更好理解。改进的控制与生产率呈正相关，因此也与竞争力呈正相关。通过课堂教学传播研究成果，学生将了解过程行为如何受到过程控制参数的影响。他们还将获得如何在实际工业环境中监测和控制过程的宝贵经验。

9501903 This research is focused on the quality control problem of understanding processes whose behaviors are affected by multiple, autocorrelated variables. Traditional statistical process control (SPC) is based on a single variable and the assumption of normality, independence, and homogeneous variance. This research capitalizes on the inappropriateness of these assumptions and the use of a single variable as a basis to understand and control a process when indeed in today's industrial environment there could be multiple, autocorrelated variables. The objective of the research is to systematically explore the behavior, sensitivity, and consistency of several multivariate quality control methods when autocorrelated multivariate data is present. These methods include the multivariate EWMA, principal component analysis, and multiple time series. The research was motivated by the lack of sufficient analytical tools to handle statistical process control situations in which multiple correlated factors influence the behavior of a process. The research plans to pursue the use of different forms of models to address the problem. The educational component of the research presents the investigator's teaching philosophy, methods of tracking teaching performance, and students understanding and involvement in the materials being taught. The investigator will deploy and integrate the research outcome into classroom by developing a new combined course in statistics and quality control with emphasis on the use of these techniques to solving industry-relevant problems in statistical process control. In addition, the techniques investigated in the research for the purpose of improving processes will be used to improve teaching effectiveness. The potential for the research is good. The outcome of the research can advance the body of knowledge in the broad area of multivariate process control. This will lead to a better understanding of process behavior and method of control. Improved control has a positive relationship to productivity and consequently, on competitiveness. By disseminating the research outcome through classroom instructions, students will gain an understanding of how process behavior is affected by process control parameters. They will also gain valuable experience on how to monitor and control processes in actual industrial settings.