Reducing Rework and Boosting Fab Capacity with Edge AI and Metrology-Driven Predictive Analytics

Abstract

The presentation provides an overview of the transition from traditional real-time AI systems to modern Edge AI approaches specifically for defect detection and classification tasks. It explores both the technical and operational challenges involved in implementing such solutions, including considerations around hardware deployment, data management, and integration with existing infrastructure. Additionally, the presentation highlights several key benefits of Edge AI, such as reduced latency, lower bandwidth requirements, increased reliability, and enhancements in accuracy, purity and defect classification.
The second part of the presentation delves into the utilization of sensor and inline wafer metrology data to enhance the prediction of defect anomalies and optimize both manufacturing processes and overall yield. By systematically incorporating this information into an edge-driven solution, semiconductor fabs can significantly boost efficiency and productivity. The session will include a detailed case study demonstrating how accurate prediction and management of feedback loops for etch process control can meaningfully reduce the amount of rework required, ultimately decreasing time and resource wastage while improving fab capacity. These insights are crucial for understanding how Machine Learning based data analytics can transform modern manufacturing environments.

Learn More.