Knowledge Transfer. From Electronic Design Automation to Advanced Process Manufacturing

On the path towards more complex systems that deliver more value at lower cost in semiconductor products, the industry had to develop frameworks to support the rapid design of very complex systems giving rise to widespread adoption of VLSI. A very similar situation exists for the advanced manufacturing process supporting the semiconductor industry. In order to keep the pace of innovation in manufacturing technology, the concept of a Digital Twin has been the main target of multiple studies and proposals. In this work we demonstrate that by learning from the lessons of VLSI, we aim to construct a digital twin of the manufacturing process that permits the monitoring and optimization of the most important aspects of semiconductor manufacturing, while at the same time demonstrating that such agnostic approach can also be applied to other manufacturing operations. A key component to the digital twin is the creation and deployment of reliable, experimentally verified machine learning models. In this work we will present necessary model characteristics to meet such requirements, under data conditions that are atypical for traditional machine learning training methods.