| About this Abstract |
| Meeting |
2025 TMS Annual Meeting & Exhibition
|
| Symposium
|
Aluminum Reduction Technology
|
| Presentation Title |
Optimization of Alumina Feeding in Electrolysis Cells using Multiphysics Modeling and Deep Learning Surrogate |
| Author(s) |
Kévin Patouillet, Nadia Chailly, Bertrand Allano, Alan Clark, John Perry, Matias Vasquez |
| On-Site Speaker (Planned) |
Kévin Patouillet |
| Abstract Scope |
In the current electrolysis process, all feeders inject the same alumina mass flow rate which is regulated according to the overfeed and underfeed cycles. However, this approach does not prevent local changes in dissolved alumina content in the electrolytic bath, which can lead to anode effects or sludge formation. To enhance homogenization of the dissolved alumina concentration, we propose to customize the injection frequency of each feeder, taking into account the physical phenomena involved in the process. The Alucell software is first used to describe MHD flows, as well as the injection, transport, dissolution, and consumption of alumina in the bath. Then, a deep learning model is trained from the numerical results, providing instantaneous predictions of the dissolved alumina distribution for a wide range of feeder parameters. Ultimately, this innovative approach helps identify the optimal feeder parameters for maximizing alumina homogeneity and thus reducing anode effects. |
| Proceedings Inclusion? |
Planned: Light Metals |
| Keywords |
Aluminum, Computational Materials Science & Engineering, Machine Learning |