Data-Based Spatial and Temporal Modeling for Surface Variation Monitoring in Manufacturing.

Spatial and temporal modeling;Surface variation monitoring;Manufacturing;Industrial and Operations Engineering;Mechanical Engineering;Engineering;Mechanical Engineering

Shao, ChenhuiAbell, Jeffrey A ;

University of Michigan

关键词: Spatial and temporal modeling;    Surface variation monitoring;    Manufacturing;    Industrial and Operations Engineering;    Mechanical Engineering;    Engineering;    Mechanical Engineering;   

Others  :  https://deepblue.lib.umich.edu/bitstream/handle/2027.42/120743/chshao_1.pdf?sequence=1&isAllowed=y

瑞士|英语

来源: The Illinois Digital Environment for Access to Learning and Scholarship

【 摘 要 】

Spatiotemporal processes exist widely in manufacturing, such as tool surface degradation in ultrasonic metal welding and surface shape progression in high-precision machining. High-resolution characterization and monitoring of spatiotemporal processes are crucial for manufacturing process control. The rapid development of 3D sensing technologies has made it possible to generate large volumes of spatiotemporal data for process characterization and monitoring. However, critical challenges exist in effectively acquiring and utilizing such spatiotemporal data in manufacturing, e.g., a high cost in the acquisition of high-resolution spatiotemporal data and a lack of systematic approaches for modeling multi-source data and monitoring spatiotemporal processes.To address these challenges, this dissertation carries out three research tasks for the development of collecting, modeling and monitoring spatiotemporal data. Specifically,(1) A novel dynamic sampling design algorithm is developed to efficiently characterize spatiotemporal processes in manufacturing. A state-space model and Kalman filter are used to predictively determine the measurement locations using a criterion considering both the prediction variance and the measurement costs. The determination of measurement locations is formulated as a binary integer programming problem, and genetic algorithm is applied for searching the optimal design. In addition, a new test statistic is proposed to monitor and update the temporal transition parameters in the spatiotemporal model.(2) A new surface modeling approach is devised to cost-effectively assess spatial surface variations by integrating an engineering model with multi-task Gaussian process (GP) learning. Surface variation is decomposed into a global trend which is induced by process variables and a zero-mean GP which shares commonality among multiple similar-but-not-identical processes. An iterative algorithm is developed to simultaneously estimate the process-specific parameters and the GP parameters.(3) A tool condition characterization and monitoring framework is developed for ultrasonic metal welding in lithium-ion battery manufacturing. The geometric progression of the tool surfaces is characterized using high-resolution spatiotemporal data. Classification algorithms are developed with monitoring features extracted from both the space and frequency domains. A novel impression measurement method is designed to effectively measure tool surfaces without the need of disassembling tools for off-line measurement.