报告名称：Development of an AI-Based Model for Optimizing Screw Configurations and Operating Conditions in Twin-Screw Extrusion Under Limited Data Availability（有限数据条件下双螺杆挤出中基于AI的螺杆配置和操作条件优化模型的开发）
报告人：Yuan Yao
时间：2025 年7月20日周日下午 14:30-15:30
地点：东校区会议中心第一会议室
报告简介：
双螺杆挤出机（TSEs）在聚合物加工中至关重要，能够在制造过程中处理多种材料。然而，TSE的性能在很大程度上依赖于优化的螺杆配置和操作条件。由于两个关键因素，TSE优化面临挑战：材料-工艺组合中的数据有限性以及螺杆配置、操作条件与产品质量之间复杂且非线性的关系。
为了解决这一问题，我们提出了一种递归深度嵌入网络（RDEN）——一种面向有限数据下TSE优化的数据驱动建模框架。受到自然语言处理进展的启发，RDEN利用自编码器学习螺杆配置的紧凑表示，这些表示随后与操作条件输入相结合，并通过基于GRU的递归神经网络处理，从而将螺杆元件的位置信息纳入质量预测。
我们进一步开发了一个优化框架，应用RDEN共同设计螺杆配置和操作条件。这种方法提高了过程仿真精度，整合了定性和定量变量，并提高了设计效率——即使在数据受限的情况下。
Twin-screw extruders (TSEs) are essential in polymer processing, enabling the handling of diverse materials in manufacturing. However, TSE performance heavily depends on optimized screw configurations and operating conditions. Such optimization is challenging due to two key factors: limited data availability across material–process combinations and the complex, nonlinear relationship between screw configuration, operating conditions, and product quality.
To address this, we propose a recurrent deep embedding network (RDEN)—a data-driven modeling framework designed for TSE optimization under limited data. Inspired by advances in natural language processing, RDEN uses an autoencoder to learn compact representations of screw configurations, which are then combined with operating condition inputs and processed by a GRU-based recurrent neural network to incorporate positional information of screw elements for quality prediction.
We further develop an optimization framework that applies RDEN to jointly design screw configurations and operating conditions. This approach improves process simulation accuracy, integrates both qualitative and quantitative variables, and enhances design efficiency—even in data-constrained scenarios.
报告人简介：
Yuan Yao获得了浙江大学控制科学与工程专业的学士和硕士学位（分别于2001年和2004年），并于2009年获得香港科技大学（HKUST）化学与生物分子工程博士学位。
2009年至2011年，他在香港科技大学高分子加工与系统研究中心担任研究员。2011年，他加入清华大学（台湾）化学工程系，成为助理教授，2015年晋升为副教授。自2019年8月起，他担任清华大学化学工程系的正教授。自2022年起，他担任《定量红外热成像》期刊的副主编。他已发表约140篇SCI期刊论文，撰写了三章书籍章节，并持有13项专利。他的研究兴趣包括过程数据分析、无损检测数据处理和智能过程控制。
他担任过80多个项目的负责人或共同负责人，包括众多的产学合作项目。近期的工业合作伙伴包括台塑集团、台化公司、长春集团、中华工程公司、先进半导体工程公司（ASE）、真鼎科技控股有限公司、友达光电、德硕科技、瑞萨电子、Swancor Advanced Materials Co., Ltd.等领先的研究机构如工业技术研究院（ITRI）。
Yuan Yao received his Bachelor's and Master's degrees in Control Science and Engineering from Zhejiang University in 2001 and 2004, respectively, and his Ph.D. in Chemical and Biomolecular Engineering from the Hong Kong University of Science and Technology (HKUST) in 2009.
From 2009 to 2011, he worked as a Research Associate at the Center for Polymer Processing and Systems, HKUST. He joined the Department of Chemical Engineering at National Tsing Hua University (NTHU), Hsinchu, Taiwan, as an Assistant Professor in 2011 and was promoted to Associate Professor in 2015. Since August 2019, he has served as a Full Professor. Since 2022, he has been serving as an Associate Editor of the Quantitative InfraRed Thermography Journal. He has published approximately 140 SCI journal papers, authored three book chapters, and holds 13 patents. His research interests include process data analytics, nondestructive testing data processing, and intelligent process control.
Yuan Yao has served as the principal investigator (PI) or co-PI on more than 80 projects, including numerous industry-academic collaborations. Recent industrial collaborators include Formosa Plastics Corporation, Formosa Petrochemical Corporation, Chang Chun Group, CTCI Corporation, Advanced Semiconductor Engineering, Inc. (ASE), Zhen Ding Technology Holding Limited, Unimicron Technology Corporation, Delta Electronics, DELmind Inc., Swancor Advanced Materials Co., Ltd., and CoreTech System Co., Ltd., as well as leading research institutions such as the Industrial Technology Research Institute (ITRI).
