GOALI: Going Beyond Design of Experiments (DOE) - Process Improvement by Joint Adjustment and Monitoring

在许多情况下，过程控制的挑战是，如果一个过程被单独留下，它将倾向于漂移和偏离目标，结果，质量将受到影响。在过程输入和输出不断变化的工业环境中尤其如此。传统的过程控制集中于检测特殊原因，这些特殊原因是由数据中的重要模式所暗示的，这些模式指出存在意想不到的变化或信号。然而，在许多情况下，由于某些过程信号是系统或操作的特征，因此可以预期它们。在这个项目中使用的方法是导航到一个良好的操作区域，使用实验设计（DOE）来评估一组变量，然后将控制工作集中在关键变量上。过程传递函数将用于优化这些变量的控制调整，并使用Cuscore统计数据通过利用预期信号的结构来有效地监控过程。对于过程调整和监控，将调查以下问题：(1)通用过程特征的适当控制器和Cuscore统计数据是什么？(2)新的监测算法在检测信号方面的效果如何？(3)控制器和监视器对其关键参数的逼近（或偏离）有多鲁棒？本研究的贡献将包括经验和理论的发展，通过更好的过程理解，监测和控制来提供改进的输出。特别是，该研究有望提供一种将实验设计知识与过程改进联系起来的方法，提供一整套调整和Cuscore监控算法及其性能的统计评估，生成工具和图表算法以改进日常过程操作和控制，并证明联合调整和监控在实际工业应用中的可行性。这项工作包括与Springs Window Fashions Division， Inc. （Middleton， WI）的合作。斯普林斯是一家领先的家居窗处理制造商，他们的加工操作将作为一个试验台，以实施结果。它还包括一个研究生研究助理参与项目并与公司合作的机会。

The challenge for process control in many settings is that if a process is left alone it will tend to drift and wander away from target and as a result, quality will suffer. This is especially true in an industrial setting where both process inputs and outputs continuously change. Traditional process control has centered on detecting special causes that are suggested by significant patterns in the data which point to the existence of unexpected changes or signals. Many situations occur, however, where certain process signals are anticipated because they are characteristic of a system or operation. The approach to be used in this prjoect is to navigate to a good operating region using design of experiments (DOE) to evaluate a set of variables, then focus the control effort on the critical variables. Process transfer functions will be used to optimize the control adjustment of these variables and use Cuscore statistics to efficiently monitor the process by capitalizing on the structure of the anticipated signal. For process adjustment and monitoring, the following questions will be investigated: (1) What are the appropriate controllers and Cuscore statistics for common process characterizations? (2) How well will the new monitoring algorithms work in detecting signals? (3) How robust are the controllers and monitors to approximations of (or deviations from) their key parameters? The contributions of this research will include both empirical and theoretical developments to deliver improved output through better process understanding, monitoring, and control. In particular, the research is expected to provide a methodology for connecting the knowledge on experimental design and process improvement, provide a full set of adjustment and Cuscore monitoring algorithms and statistical evaluation of their properties, generate tools and charting algorithms to improve daily process operation and control, and demonstrate the feasibility of joint adjustment and monitoring for real industry applications. This work includes a partnership with Springs Window Fashions Division, Inc. (Middleton, WI). Springs is a leading manufacturer of home window treatments and their processing operations will serve as a test bed for implementation of the results. It also includes an opportunity for a graduate research assistant to work on the project and with the company.