SMART-IC: Smart Monitoring and Production Optimization for Zero-waste Semiconductor Manufacturing

The Industry 4.0 revolution introduced decentralized, self-organizing, and self-learning systems for production control. New machine learning algorithms are getting increasingly powerful to solve real-world problems, like predictive maintenance and anomaly detection. However, many data-driven applications are still far from being optimized to cover many aspects and the complexity of modern industries; correlations between smart monitoring, production scheduling, and anomaly detection/predictive maintenance have only been partially exploited. This paper proposes to develop new data-driven approaches for smart monitoring and production optimization, targeting semiconductor manufacturing, one of the most technologically advanced and data-intensive industrial sectors, where process quality, control, and simulation tools are critical for decreasing costs and increasing yield. The goal is to reduce defect generation at the electronic component level and its propagation to the system- and system-of-systems- level by working on (1) enhanced anomaly detection, based on the human-in-the-loop concept and on advanced treatment of multiple time-series and of domain adaptation, (2) smart and predictive maintenance based on both objective data traces and simulated ones, to mitigate the risk of degrading product quality, and (3) the construction of an extended manufacturing software stack that allows anomaly- and maintenance-aware policies to enhance production line scheduling and optimization.