Key takeaways at the s-X-AIPI Month 30 Consortium Meeting in Athens, Greece
It is hard to mention any other topic after such an exciting time in Athens on October 16th and 17th. The s-X-AIPI team enjoyed a productive time in the Greek capital at CORE Innovation Centre offices, meeting colleagues to share research updates and to plan the s-X-AIPI’s final steps as it approaches completion. Over two action-packed days, including presentation sessions, vivid discussions and strategic planning, several key-takeaways emerged. Here’s what you need to know:
Key Achievements to date
Completion of the Beta Version (Final) of the Autonomic Manager (AM)
The final beta version of the Autonomic Manager has been successfully completed, incorporating substantial improvements. A transition from Airflow to Dagster has enhanced task execution speed and reliability, and the integration of Redpanda, a high-performance messaging system, has streamlined communication while reducing system complexity.
Enhanced Self-X Solutions Across Use Cases
All four use cases (asphalt, steel, pharmaceutical, and aluminum) have integrated advanced self-X solutions. These solutions focus on anomaly detection, predictive maintenance, and optimization algorithms to support enhanced control and adaptability of industrial processes based on real-time and historical data.
In-Depth Analysis of Use-Case Sessions
Asphalt Use Case
Self-X Solutions: Predictive and anomaly detection models have been implemented to optimize asphalt mix quality. Advanced models monitor material component sensors and mixing times, alerting operators to any deviations that might impact quality or operational efficiency, including predictive maintenance for key equipment.
Human-in-the-Loop (HITL): Real-time dashboards now enable operators to monitor the mixing process, label anomalies, and fine-tune models based on updated conditions. This HITL setup ensures responsiveness to changing material conditions, enhancing decision-making capabilities in mix production.
Steel Use Case
Self-X Solutions: Optimization algorithms support more efficient scrap utilization in steel production. Self-X components assist in managing scrap mixing, predicting final product chemistry, and detecting anomalies in the Electric Arc Furnace (EAF) process.
HITL: Through a custom interface, operators can view scrap properties, suggested recipes, and predicted compositions, allowing for real-time adjustments. Alerts generated by the Autonomic Manager (AM) provide guidance on necessary adjustments to maintain desired quality and cost efficiency.
Pharmaceutical Use Case
Self-X Solutions: Self-X functionalities focus on supporting the electrochemical conversion process in cortisone production, with real-time monitoring of electrode positioning and process parameters. Anomalies in voltage, current, and chemical concentrations are autonomously detected to maintain product quality.
HITL: The setup includes an Optical Coherence Tomography (OCT) imaging system that monitors electrode positioning and corrosion in real-time. HITL integration enables operators to correct electrode alignment when shifts are detected, ensuring data integrity and quality for each experimental run.
Aluminum Use Case
Self-X Solutions: The Aluminum use case has developed an Intelligent Decision Support System (IDSS) with predictive models for chemical composition based on raw material combinations. Self-X algorithms generate recipe recommendations optimized for cost and material usage.
HITL: Operators receive up to three optimized recipe suggestions for each batch. A user interface allows feedback on recipe quality, considering cost and composition predictions. Anomalies in material consumption and process duration are monitored, with alerts ensuring consistent process efficiency.
KPI Framework for Self-X Systems
A comprehensive set of KPIs using the ECOGRAI methodology has been established to validate the effectiveness of the MAPE-K loops in the different self-X solutions in areas such as latency, resource consumption, and anomaly detection accuracy and so on. These KPIs will provide essential data on system performance and validate improvements as self-X components are integrated into the final testing phase.
Testing and Validation of Advanced Prototypes
Final testing is now in place to validate the self-X AI solutions developed. These prototypes are being assessed for their impact on operational efficiency, process optimization, and quality control in real-world industrial conditions, setting the stage for a final evaluation and performance documentation towards the end of the project.
In conclusion, we extend our sincere gratitude to all partners, whether present in-person or online, for their contributions, engaging discussions, and for fostering an environment rich in information and collaboration. Of course, a special thank you goes to CORE Innovation Centre for organizing the meeting and for their hospitality!
