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 |
Causal Intervention Mapping from Process Intensification to Pellet Architecture Design in Hydrogen Based Direct Reduction for Green Steelmaking |
| Author(s) |
Anurag Bajpai, Barak Ratzker, Pasquale Cavaliere, Dierk Raabe |
| On-Site Speaker (Planned) |
Anurag Bajpai |
| Abstract Scope |
Hydrogen-based direct reduction can decarbonize primary ironmaking, but its industrial efficiency is governed by the late-conversion kinetic tail, where furnace residence time, hydrogen recycle load and energy demand are set. Here we establish a physics-constrained interpretable machine learning framework that converts measured reduction trajectories into metallurgical design rules. Conversion-resolved deconvolution reveals a transition from early thermochemical control to late internal-transport limitation, governed by pellet diameter, pore size, porosity and tortuosity. Causal intervention mapping then shows that process intensification is not universally beneficial. We reveal that temperature, pressure and reducing-gas potential deliver high early-stage returns, but their late-stage effectiveness collapses when pore topology becomes restrictive. This defines process-controllable, mixed and topology-locked regimes, identifying when furnace severity should be increased and when pellet architecture must instead be redesigned. The framework links reaction kinetics, pore-network evolution and causal decision maps, providing an actionable route to faster, hydrogen-efficient direct reduction for scalable green ironmaking. |
| Proceedings Inclusion? |
Planned: |
| Keywords |
Machine Learning, Computational Materials Science & Engineering, Sustainability |