Abstract
This study proposes an integrated modeling approach that combines mechanistic and data-driven methods for the hydrogen-based mineral phase transformation (HMPT) process of refractory iron ores. First, a dynamic mechanistic model covering multistage preheaters, the heating furnace, and the restorer is developed, describing the heat transfer, mass transfer, and reaction kinetics in the redox asynchronous compartment system. Second, a parameter identification method combining fuzzy serial-parallel stochastic configuration networks (F-SPSCN) with a grid search is introduced to accurately estimate key unmeasurable parameters. Finally, under typical industrial operating conditions, simulation results show that the system achieves a final Fe(3)O(4) conversion rate of 99.01%, with the temperatures of the heating furnace and the restorer stabilizing at 848 and 863 °C, respectively, and temperature fluctuations of the preheaters remaining within ±2.5 °C. The control variable experiment and disturbance variable experiment further confirm that this mechanistic model is capable of effectively capturing the synergistic effects of multiple parameter changes.