, on September 07, 2023, 01:41 PM

AISSI: Addressing Semiconductor Manufacturing Challenges

In an era driven by digitalization and the ever-increasing demand for microchips, SYSTEMA has taken a proactive stance by investing in groundbreaking research and development initiatives. One such initiative is the “Autonomous Integrated Scheduling for Semiconductor Industry” (AISSI) project. Using state-of-the-art AI methods, AISSI aims to revolutionize semiconductor manufacturing and improve efficiency, cost-effectiveness, and quality for European semiconductor manufacturers.

Empowering SMEs and Global Competitiveness

AISSI aims to strengthen the global position of European semiconductor manufacturers by addressing the growing demand for microchips. The project also sets the stage for the rise of tech-savvy Small and Medium Enterprises (SMEs) by facilitating solution validation, research access, and a path to large-scale commercial growth.

Integrating Manufacturing, Maintenance, and Customer Fulfillment 

The key innovation of the project was the development of a new approach to semiconductor production planning that effectively converges the needs of manufacturing, maintenance, and customer fulfillment requirements. This approach, supported by an innovative interaction between human expertise and AI, increased the flexibility, quality, and efficiency of the production system.

Creating an Industry-Driven AI Ecosystem

In partnership with leading European semiconductor manufacturers, data science experts, and innovative SMEs, AISSI set out to build an industry-driven AI ecosystem. This collaborative effort resulted in the creation of state-of-the-art integrated scheduling solutions that align with various European smart industry transition roadmaps.

SYSTEMA and Partners: Advanced Meta-Heuristics in Practice

SYSTEMA, in collaboration with Bosch, Karlsruhe Institute of Technology, Nexperia, and D-SIMLAB Technologies, implemented advanced meta-heuristics within the AISSI project to increase production efficiency at Nexperia Hamburg. This solution included a digital twin in bottleneck areas, an event-driven dispatcher, a constraint solver, and a variety of strategies such as combinatorial optimization and reinforcement learning. The outcome was a robust production planning testbed, that improved AI-based production optimization strategies.

Compliance and Transparency in AI Methodologies: Digital Twin as Angle Point

In line with GDPR, SYSTEMA’s work also involved the development of AI methodologies for complete production planning, ensuring transparency and explainability of the AI methods used. In general, we conceptualize the digital twin as based on two pillars: a) complex fab simulations and b) a holistic digital twin. Practical tasks encompassed the adaptation of the holistic digital twin (HDT) to factory conditions, developing a prototype, testing KPIs, and detailing requirements for model extensions.

Holistic Digital Twin: Enhancing Production Processes 

The holistic digital twin HDT is an informationally transparent and physically “speaking”, multi-granular information module. It holds at the same time – by implementing an informational model that is based on the ai-grounding manifold hypothesis – an isomorphism between the physical fab states and the current “internal” complex configuration of ai-tools such as neural nets. Significantly, the HDT interface facilitated transparency, explainability, and data-driven decision-making to optimize production processes. Additionally, a production planning methodology based on information quality was explored, which required a unified simulation of the production process.

A New Level of Continuous Production Optimization

In conclusion, this project has led to considerable advancements in the handling of information processing, creating a holistic approach with real-time capabilities and consistency, comprising a new level of overall informational quality and system performance. This enables us to step into the world of continuous production optimization. The intent to efficiently incorporate cutting-edge AI methodologies may herald a new path to manufacturing optimization while supporting and enabling ad-hoc decisions for specific scenario simulation and validation and complex scenarios to continuously train and monitor, steer, and transparently optimize ai-based components (such as the AISSI reinforcement agent). To learn more about production optimization/scheduling, please contact us.


No comments published.
Be the first to comment.