About this Abstract |
| Meeting |
2027 TMS Annual Meeting & Exhibition
|
| Symposium
|
AI-Enabled Materials Processing: Integrating Accelerated Experimental Workflows and Processing-Aware Machine Learning
|
| Presentation Title |
A Cost-Aware Contextual Bandit Framework for Autonomous Optimization of Titanium Leaching from Iron-Depleted Red Mud Slag |
| Author(s) |
Dilara Hasil, Irmak Sargin |
| On-Site Speaker (Planned) |
Dilara Hasil |
| Abstract Scope |
Recovering titanium from iron-depleted red mud valorizes residues and reduces reliance on primary resources. Because slag composition varies, no single leaching recipe performs optimally across feedstocks. This work introduces a cost-aware contextual bandit for autonomous leaching optimization. Within this framework, agent serves as decision-maker evaluates candidate operating conditions and selects the optimal action. Its predictive core is an offline Gaussian Process Regression model, estimates recovery and uncertainty, achieving Rē of 0.96 and RMSE of 6.7%. Also, Apriori extracts interpretable rules linking slag properties and process parameters to high recovery. Two policies then compete: a rule-based policy searches within these ranges, while an exploration policy searches physical bounds. The highest-scoring action, balancing recovery, uncertainty, and resource use, is selected for experimentation. Once experimental feedback becomes available, a reward function evaluates observed outcome in terms of achieved recovery and resource usage, closing the loop and allowing agent to learn directly from process. |
| Proceedings Inclusion? |
Planned: |
| Keywords |
Other, Sustainability, Hydrometallurgy |