Modeling Data from Electronics Manufacturing Processes

这个校企合作研究项目的目标是创造一个环境，培养具有解决实际应用问题经验的学生。建议的技术项目有可能在分析许多工程应用中的大量（非标准）数据方面产生广泛的影响。NSF和西门子都提供了三名学生研究助理奖学金的一半，西门子还提供工程专业知识、过程数据和实验设备。一半的学生时间将用于理解在西门子遇到的产品、过程、问题和数据，另一半的学生时间将用于开发解决类似于西门子遇到的问题的通用方法。随着自动化数据收集过程中更先进的工具的开发，许多像西门子这样的公司有机会探索从这些数据中提取知识的方法，以提高他们的决策质量。综合从西门子印刷电路板组装过程中收集的数据所面临的挑战是数据的大尺寸和大量的零缺陷计数。为了解决这些问题，本提案制定了以下几项具体任务：(1)提取均值和方差变化点特征（利用数据分割和小波处理）；(2)制定多层流程改进规则；(3)研究自动数据建模流程；(4)对零膨胀数据进行贝叶斯分析，提高生产质量。有很多机会发展统计理论来理解所提出的方法，所有的任务都涉及大量的计算和数据分析。西门子已经为学生提供了试验数据，以便他们在研究培训项目中获得实践经验。该提案首先制定了一个由大学教师、学生和公司经理组成的工作组，以组织我们的资源和努力，在拟议的技术项目中开发数学工具。然后，委员会将确定学生论文研究培训的短期和长期目标，并制定具体问题。最后，学生将进行实践项目，记录研究结果并在会议上展示研究结果。该项目由MPS多学科活动办公室（OMA）和数学科学司共同支持。

The goal of this university-industry cooperative research project is to create an environment for training students with experience of solving real-life application problems. The proposed technical projects have potentials to make a broad impact in analyzing large volumes of (non-standard) data in many engineering applications. Both NSF and Siemens provide half of the three student research assistantships, and Siemens also provides engineering expertise, process data and facility for experimentation. Half of the student time will be spent on understanding the product, process, problem and data encountered at Siemens and the other half of the student time will be involved in developing general methodologies that solve problems similar to what Siemens encountered.With more advanced tools developed in automating the data collection process, many companies like Siemens have an opportunity to explore ways to extract knowledge from this data for improving their decision-making quality. The challenges in synthesizing data collected from Siemens' printed circuits board assembly processes are the large sizes of data and the numerous numbers of zero-defect counts. This proposal formulates several specific tasks outlined as follows to solve these problems: (1) Extract mean and variance change-point features (using data segmentation and wavelet procedures), (2) Formulate multi-layer process improvement rules, (3) Research automatic data modeling procedures, (4) Conduct Bayesian analysis of zero-inflated data for improving production quality.There are many opportunities for developing statistical theory to understand the proposed methods, and all tasks involve heavy computing and data analysis. Siemens has located trial-out data for students to gain hands-on experience in their research training programs. This proposal first formulated a task-force team consisting of university faculty members, students and company managers for organizing our resources and efforts to develop mathematical tools in the proposed technical projects. The committee will then identify short- and long-term goals and formulate specific problems for student thesis research training. Finally, students will conduct hands-on projects, document study results and present findings in meetings. This project is jointly supported by the MPS Office of Multidisciplinary Activities (OMA) and the Division of Mathematical Sciences.