¶ UC13 - Intelligent Sensors Improving Robustness of Automated Wafer Transportation and Storage Systems
This use case develops intelligent sensor-based predictive maintenance solutions for automated wafer transportation and storage systems. The goal is to improve uptime and robustness by detecting mechanical wear, faults, malfunctions, or contamination risks before they cause breakdowns. Sensor-on-chip systems, mobile and stationary sensor platforms, data fusion, supervised learning, damage assessment, and reinforcement learning for autonomous navigation are considered. Since wafer transport and storage are central to highly automated fabs, the use case can significantly improve productivity, reduce downtime, lower maintenance impact, and reduce the environmental footprint of wafer handling.
| # |
Technology Name |
Description |
| 1 |
AMHS |
A novel hardware and software architecture is established for demonstrators of high-volume Automated Material Handling Systems, including overhead transportation systems in Dresden and Villach. The goal is to reliably predict and detect hardware malfunctions from sensor data, support predictive maintenance for AMHS components, improve OEE and cycle times, and increase the uptime of wafer transportation and storage systems. The technology also includes vision-guided object detection and classification on embedded AI hardware using video frames from industrial-grade cameras. |
| 2 |
Tool and Person Detection with Classification for Mobile Systems |
The baseline system using ToF camera and radar sensor is extended with an RGB camera and NVIDIA Jetson-based processing. A deep neural network is fine-tuned on a custom dataset for tool and person detection and classification. The enhanced system detects and classifies tools and people, estimates their distance and velocity, and uses radar noise-filtering techniques so that the system can be deployed on mobile platforms. |
| KPI # |
Related Technology |
KPI Target |
| 1 |
AMHS |
First machine models installed to detect hardware malfunctions in Y1. |
| 2 |
AMHS |
Test with defective hardware within Y2, with validation to be completed within Y3. |
| 3 |
AMHS |
First new automation components with lower footprint installed at IFD. |
| 4 |
AMHS |
Creation of more than 1,000 sensor data samples, with separation into good and bad data ongoing. |
| 5 |
AMHS |
Validation of fault-detection patterns for bearings using laboratory setups. |
| 6 |
AMHS |
Successful demonstration of anomaly detection on OHT vehicles using new hardware systems, supporting improved uptimes of wafer transportation and storage systems. |
| 7 |
AMHS |
Fast detection of faults in high-volume transport systems to reduce material cost and time losses by avoiding unnecessary replacement of expensive parts. One 300 mm vehicle costs approximately 30,000 EUR. The target also supports a reduced environmental footprint for wafer transport, handling, and storage. |
| KPI # |
Baseline |
| 1 |
Manual inspection and control of all parts by highly skilled maintenance personnel. |
| 2 |
High manufacturing losses occur in case of unplanned hardware failures. |
| 3 |
No sensor systems are available for predictive maintenance actions. |
| 4 |
Preventive maintenance is performed on all parts of automated material handling systems; predictive maintenance is not possible. |
| 5 |
No existing industrial dataset utilizes sensing technologies such as ToF cameras, RGB cameras, and radar. |
| 6 |
Existing detection systems are primarily based on person detection and are limited to stationary setups. |
| KPI # |
Validation Method |
| 1 |
Define the new infrastructure and perform requirement analysis, working in parallel on a strategy for monitoring and controlling vehicles on the OHT system. |
| 2 |
Design and demonstrate an experimental workflow using a tracking system with two ToF cameras and one RGB camera to monitor a swarm of small robots, enabling detection and analysis of abnormal behavior and deviations from expected patterns. |
| 3 |
Develop machine learning algorithms to anonymize collected visual data, ensuring privacy protection through two different methods. |
| 4 |
Develop two sensor-box demonstrators integrating ToF camera, radar, and RGB camera. Use them to collect datasets, followed by data extraction, preprocessing, noise removal, visualization, and interpretation. |
| 5 |
Build a customized dataset for tool and person detection and use it to fine-tune object detection deep neural networks. Validate the mobile person and tool detection/classification system, including handling of noisy radar signals caused by platform movement. |
| 6 |
Validate fault-detection patterns for bearings using laboratory setups, with ongoing experiments using alternative sensor setups to increase robustness. |
- Industrial process phases made more effective and precise through integrated and tailored AI algorithms, methods, and tools.
- 20% improvement in prediction precision.
- 25% improvement in classification precision.
- Improved overall productivity of the manufacturing factory by approximately 20%.
- Improvement of uptimes of wafer transportation and storage systems by 20%.
- Providing documented methods for dynamic application of tools to improve sustainability.
The use case contributes directly to improving the uptime of wafer transportation and storage systems by developing intelligent sensor-based fault detection, anomaly detection, and predictive maintenance technologies for AMHS and OHT systems. Sensor data, audio-signal-processing feature extraction, time-series analysis, supervised and unsupervised learning, and predictive-maintenance models support earlier detection of hardware malfunctions and abnormal behavior. The tool and person detection system contributes to improved classification precision through mobile, sensor-fusion-based object detection and classification. By reducing unplanned failures, unnecessary part replacement, and manual inspection effort, the use case supports higher productivity, reduced material losses, and a lower environmental footprint for wafer transport, handling, and storage.
| AI Toolbox Tool |
Use within the Use Case |
| Object detection tool |
Used to detect and classify tools, people, and relevant objects from industrial camera data in automated wafer transportation and storage environments. It supported the vision-guided monitoring functions deployed on embedded AI hardware. |
| Object detection with YoloV8 |
Used for real-time tool and person detection on RGB camera data in mobile monitoring systems. It supported the fine-tuned deep neural network pipeline for detecting objects and people in wafer transport areas. |
| Image Classification Tool |
Used to classify detected tools, people, and system states from visual data. It supported the mobile tool and person detection component by distinguishing relevant object classes in factory environments. |
| Multi-Object Tracking Tool |
Used to track moving objects, people, or mobile systems across video frames. It supported the monitoring of dynamic behavior and deviations in automated material handling and mobile-system scenarios. |
| Time-Series Anomaly Detection Tool |
Used to detect abnormal temporal patterns in sensor data from AMHS and OHT vehicles. It supported predictive maintenance by identifying early signs of hardware malfunction, bearing faults, or abnormal vehicle behavior. |
| Tabular Prediction Tool |
Used to train predictive models on structured sensor and maintenance data. It supported the prediction and detection of hardware malfunctions from collected AMHS sensor data. |
| Explainability Tool |
Used to make predictive-maintenance and anomaly-detection outputs more interpretable for maintenance experts. It supported understanding of which sensor features or operating patterns indicated possible equipment faults. |
| Device-Level Quality Detection Tool |
Used to connect machine-signal-related features with equipment-level faults and root causes. It supported fault detection and diagnostic analysis for wafer transportation and storage systems. |