This use case develops an AI-enhanced Manufacturing Execution System concept for semiconductor manufacturing clusters. It aims to establish real-time, adaptable MES capabilities that can support complex product mixes, improved workflow engines, AI-based production scenarios, and integration with internal cloud and AI applications. The MES becomes a core enabler for automation, digitization, resilient production, and AI-driven operational improvement. The expected benefits include higher throughput, better cycle time, self-adapted manufacturing processes, improved energy and material efficiency, and a scalable foundation for future semiconductor AI use cases.
| # |
Technology Name |
Description |
| 1 |
MES Airbus |
A Digital Twin of an automation solution is created as a precise virtual representation of the real system. The Digital Twin is connected to the actual PLC, higher-level control systems, and, where applicable, the robotics suite, depending on the automation configuration. From the perspective of the connected software systems, the Digital Twin replicates the behavior, responses, and operating logic of the real-world automation setup. This enables engineers and developers to configure, test, and expand software solutions in a realistic environment without requiring access to the physical automation hardware. |
| KPI # |
Related Technology |
KPI Target |
| 1 |
Anomaly detection |
AI-based MES capability above 10%. |
| 2 |
RAMII integration into MES |
RAMII architecture for an AI-enabling version of the Manufacturing Execution System. |
| 3 |
AI-based MES trainings |
AI-based training for adapted or enhanced production scenarios above 10%. |
| 4 |
Enhanced operational efficiency by MES functionality |
Development of new methods and processes as an enabler to enhance operational efficiency above 20%. This will be shown within UC12 and UC17. |
| KPI # |
Baseline |
| 1 |
Anomaly detection is performed manually or semi-automated, for example in production monitoring and quality control. |
| 2 |
RAMII integration is not performed, or only partly performed, within MES in semiconductor manufacturing. |
| 3 |
No AI-based trainings are available for MES. |
| 4 |
Only manual or semi-automated data visualization is currently available in semiconductor factories. No computer-based or AI-based visualization by MES is available. |
| KPI # |
Validation Method |
| 1 |
Validation is performed by systematic testing using one year of production data from the high-volume semiconductor facility at Infineon Dresden. |
| 2 |
Validation is performed by tests in the production environment at IFD. |
| 3 |
Results are validated with factory experts. |
| 4 |
The method is proven on existing bottleneck situations, including tools, product mix, actual uptime restrictions, technologies, basic product types, and OEE data. |
- AI-based MES capability above 10%.
- Self-adjusted/adapted manufacturing processes above 15%.
- Improved overall productivity of the manufacturing factory by approximately 20%.
- Reduced manufacturing costs in the long term.
- Improved performance and faster time-to-market.
- Industrial process phases made more effective and precise through integrated and tailored AI algorithms, methods, and tools.
The use case contributes directly to AI-based MES capability by developing an AI-enabling MES concept, including anomaly detection, RAMII-based MES integration, AI-based MES training, and enhanced operational efficiency methods. The Digital Twin supports the design, testing, and optimization of automation systems in a virtual environment before deployment, reducing the need for costly on-site troubleshooting and helping to shorten commissioning times. By simulating production scenarios and validating new functionalities without disrupting physical operations, the use case supports more adaptive manufacturing processes, improved operational efficiency, and faster deployment of new production capabilities. The connection to UC12 and UC17 further supports the use of MES functionality for bottleneck detection, improved visualization, and more efficient semiconductor manufacturing operations.
| AI Toolbox Tool |
Use within the Use Case |
| Time-Series Anomaly Detection Tool |
Used to support anomaly detection in MES-related production monitoring data. It helped identify unusual temporal patterns in semiconductor production data and supported the development of AI-based MES capabilities. |
| Tabular Prediction Tool |
Used to analyse structured MES, production, tool, product-mix, uptime, and OEE-related data. It supported the development of data-driven methods for detecting bottlenecks and improving operational efficiency. |
| Explainability Tool |
Used to make AI-based MES outputs more interpretable for factory experts. It supported the validation of anomaly detection, bottleneck analysis, and operational-efficiency recommendations by helping users understand the underlying model behaviour. |
| Device-Level Quality Detection Tool |
Used to support quality-oriented MES analysis by linking production and machine-signal-related features to device-level quality issues and possible root causes. |
| LLM Tool |
Used to support AI-based MES training scenarios and human-facing interaction with MES-related knowledge. It helped generate explanations, training material, and scenario-based guidance for adapted or enhanced production situations. |
| Prompt Context Injection Tool |
Used to ground LLM-based MES training and support functions with relevant production, RAMI/RAMII, Digital Twin, and factory-context information. |