The primary goal of this collaborative GOALI research is to develop and validate a novel non-threaded advanced process control (APC) framework for next-generation high-mix semiconductor manufacturing. Semiconductor technology lies at the heart of the revolution in computing, communications, consumer electronics, transportation and health care. In the last decade, diversified demand from consumers has been pushing semiconductor industry to produce many differentiated products. As a result, multi-product-multi-tool ('high-mix') manufacturing has become increasingly the standard manufacturing model, which poses many challenges that the current APC framework cannot address. The PIs plan research in the fields of run-to-run (RtR) control, control performance assessment (CPA) and statistical process monitoring (SPM) to meet the emerging needs in high-mix production. Intellectual Merit: The research will create a non-threaded paradigm for high-mix semiconductor manufacturing by breaking from the current tradition of threaded APC, and provide new theories and techniques to address the challenges posed by high-mix production. By sharing information among different threads and different APC components, monitoring and control performance will be greatly improved and the number of required models will be significantly reduced. Specifically merits of each project are summarized below. Project 1: State estimation and control model update: It will provide theoretical analysis on the non-threaded state estimation problem; in addition, it will develop a systematic approach for non-threaded state estimation and control model update for high-mix production, which handles large-scale nonlinear systems through a linear regression formulation. Project 2: Control performance assessment and diagnosis (CPA/CPD): Instead of comparing the actual control performance against a theoretical benchmark, the proposed framework explicitly estimates model-plant mismatch and disturbance dynamics to achieve CPA/CPD simultaneously. In addition, it will provide the first non-threaded CPA/CPD tools for RtR controllers in high-mix fabs. Project 3: Statistical process monitoring: Analyzing the pattern of batch statistics instead of the pattern of process variables for SPM is planned. The approach eliminates data pre-processing required by threaded methods, greatly improves monitoring performance, and significantly reduces the number of required models. Broader Impact: This research will have an immediate impact on the industrial practice of semiconductor manufacturing, as it specifically addresses emerging industrial needs. Due to the complexity of semiconductor processes and the critical role of APC in fab-wide monitoring and control, the problem addressed in this research has the potential to transform the way industry performs process control. Because few restrictions were posed during the framework development, the proposed framework is not limited to the semiconductor processes, instead, it can also be applied to the batch-oriented pharmaceutical, specialty chemical, and polymer industries and could inspire new solutions and research directions in general batch process monitoring and control. This research promotes the education of control engineers for semiconductor manufacturing at both graduate and undergraduate levels. Currently, U.S. semiconductor companies are facing challenges in sustaining a well-qualified semiconductor workforce, including engineers in the area of process control. Therefore, the three universities are committed to the continuing education and training of students and professionals in semiconductor manufacturing process control. Moreover, these projects are potential resources for involving minorities and giving them research experience in semiconductor process control. Finally, the PIs will offer short courses on the new process control paradigm to mid-career professionals in the semiconductor industries.
Semiconductor technology lies at the heart of the revolution in computing, communications, consumer electronics, transportation and health care. Driven by diversified consumer demand, high-mix, i.e., multi-product-multi-tool manufacturing has increasingly become the standard manufacturing model, which poses many challenges that the existing advanced process control (APC) framework widely deployed in semiconductor fabs cannot address. In this project, we have developed a novel non-threaded APC framework for next generation high-mix semiconductor manufacturing, which provides a paradigm shift on how manufacturing data can be shared among different machines, different products, etc. The developed APC framework consists of three components: state estimation module, process monitoring module, and control performance monitoring module. The developed algorithms can effectively share information among different threads and different APC components, therefore, the overall production efficiency can be greatly improved. Besides new algorithms, we have also developed new theories to guide the practitioners in the field. Our results drew a lot of interests from control engineers in semiconductor manufacturing, not only from US, but also from other countries all over the world, including Germany, Singapore and Taiwan. Finally, this project provided training opportunities for three graduate students at Auburn University and several undergraduate students, including two from Tuskegee University, a local HBCU. In addition, one PhD and one MS were graduated in 2012 and 2013, respectively.
