Using intelligent technology and real-time feedback algorithm to improve manufacturing process in IoT semiconductor industry

• Published in: The Journal of supercomputing Vol. 77; no. 5; pp. 4639 - 4658
• Main Authors: Li, Bin, Chen, Ruey-Shun, Liu, C.-Y.
• Format: Journal Article
• Language: English
• Published: New York Springer US 01.05.2021; Springer Nature B.V
• Subjects: Algorithms; Business competition; Cluster analysis; Clustering; Compilers; Computer integrated manufacturing; Computer Science; Data mining; Data processing; Deep Learning in IoT: Emerging Trends and Applications - 2019; Feedback control; Flexibility; Foundries; Internet of Things; Interpreters; Manufacturing; Process controls; Process parameters; Processor Architectures; Production costs; Programming Languages; Real time; Statistical process control; Technology assessment; Vector quantization; Intelligent technology; Real-time feedback algorithm; Manufacturing process; Internet of things (IoT); Semiconductor industry
• ISSN: 0920-8542; 1573-0484
• DOI: 10.1007/s11227-020-03457-x

To enable the Internet of things, the semiconductor manufacturing process has progressed from the micron to the deep submicron level. Quality improvement is one of the great challenges in wafer fabrication. Computer-integrated manufacturing (CIM) has arisen as a means by which to reduce wafer rework and continuously improve the semiconductor production process. This study uses statistical process control (SPC) and data mining technology to analyze the collected semiconductor process data. A real-time feedback algorithm is employed to ensure that each product lot is manufactured using optimized process parameters. This maximizes production capability, increases the semiconductor yield rate and reduces the cost of manufacturing. This paper focuses on wafer fabrication facilities (often called “fabs,” or “foundries”). The data mining architecture is implemented between CIM and the manufacturing execution system. Association rules and the k -means clustering algorithm are used together with real-time feedback control analysis to extract and analyze each parameter that affects the semiconductor production yield. This combination of real-time feedback and SPC using historical process data allows the system to predict the optimum process parameters for the next lot. The system compensates dynamically to accommodate differences among various machines and products, giving each machine a level of flexibility in manufacturing the product. The proposed semiconductor system can be applied to traditional manufacturing industry process analysis. Our results show how our system can improve the semiconductor manufacturing process in terms of processing capability, yield rate, stability and flexibility, while reducing costs, thus creating a competitive advantage for the wafer fab.