The vast generation of silica‑rich iron ore tailings (IOT) poses major environmental and economic challenges. This study presents a hydrometallurgical route for selective silica recovery using ammonium bifluoride (NH4HF2) to produce high‑grade nanoscale amorphous silica. Near‑complete extraction (~ 98%) is achieved within 3 h at 80 °C with 30 wt% NH4HF2, and kinetic analysis indicates a mixed‑control mechanism (Ea = 16–18 kJ/mol). To model the recovery as a function of time, reagent concentration, and temperature, three machine learning (ML) techniques were compared: fully connected artificial neural network (ANN), random forest (RF), and least squares regression (LS). The ANN outperformed the others (R2 = 0.995, RMSE = 0.028), providing a robust predictive tool that also enables flexible selection of conditions for 98% recovery with reduced reagent use. With the potential to recycle up to 95% of NH4HF2, the ANN‑optimized process enhances the economic feasibility of the industrial‑scale IOT valorization. Selective silica removal enriches the residual valuable elements and simplifies their downstream separation by circumventing silicate interference. By combining systematic experiments with ML, this work offers a data‑driven, scalable approach to transforming mining waste into a valuable resource, thereby reducing environmental burdens.


扫码关注我们
求助内容:
应助结果提醒方式:

