PI-DeepONet-H2DR: code, data and trained model for constraint-enforcing operator learning of hydrogen direct reduction

Code, data, and trained model for the article

M. M. Teymoori, M. A. Bijarchi, H. Afshin. Constraint-enforcing physics-informed machine learning for hydrogen-based ironmaking: a deep operator network surrogate applied to energy-optimal reduction. Submitted to Journal of Cleaner Production.

The release contains the pellet-scale solver (three-interface shrinking-core model of hydrogen reduction) with its calibrated parameters, the scripts that generated the 15,000-trajectory dataset, the dataset itself, the constraint-enforcing PI-DeepONet with its training scripts and trained weights, the evaluation scripts, and the energy optimization.