About this Abstract |
| Meeting |
2026 TMS Annual Meeting & Exhibition
|
| Symposium
|
Electrode Technology for Aluminum Production
|
| Presentation Title |
AI-Powered Control Strategy for Predicting Pitch Demand and Enhancing Anode Density |
| Author(s) |
Sujit Anandrao Jagnade, Amit Agashe, Saurabh Kawale, Rajshri Magdewar, Parthprasoon Sinha, Praveen Kapse |
| On-Site Speaker (Planned) |
Sujit Anandrao Jagnade |
| Abstract Scope |
In aluminum smelting, optimizing pitch addition in green anode production is vital for consistent anode density and quality. This study presents a data-driven model to predict pitch demand using variations in calcined petroleum coke (CPC) properties and process parameters. It addresses challenges from fluctuating CPC real density, pitch quality, and their interactions during mixing and compaction. Historical plant data covering CPC porosity, particle size, pitch properties, and real-time process variables—was analyzed using machine learning. Among several models, Random Forest showed the best predictive accuracy, capturing non-linear relationships effectively. Partial Dependence Plots (PDP) aided interpretation of key drivers such as kneader pitch temperature and paste temperature. Model deployment reduced anode density standard deviation to 0.005, enhancing process control and consistency. This approach improves operational efficiency and supports sustainability by optimizing raw material use. It highlights the potential of AI/ML in transforming process control and quality assurance in the aluminum industry. |
| Proceedings Inclusion? |
Planned: Light Metals |
| Keywords |
Aluminum, Machine Learning, Other |