Surrogate-Assisted Optimization for Real-World Wet-Process Phosphoric Acid Production

Wet-process phosphoric acid production (WPAP) is a crucial part of modern industry, and its optimization can enhance production efficiency and enable precise utilization of ore resources. However, due to the complexity of its reaction mechanisms, effective simulation and optimization methods remain lacking. To address these issues, this study integrates lab-scale kinetic models, which are based on both the literature and industrial data, into full-process simulation to construct a precise first-principles model. By combining the first-principles model with actual production data, an accurate and computationally efficient surrogate model is established. Three optimization scenarios were developed based on real-world conditions, aiming to minimize the content of nonwater-soluble phosphorus (NWSP). The sequential quadratic programming for machine learning-embedded optimization (MLSQP) algorithm was applied to solve these optimization problems, enabling efficient convergence to the Karush–Kuhn–Tucker (KKT) point of the surrogate model. Validated by the rigorous model, the optimizations under the three scenarios achieved reductions of 20.79%, 8.91%, and 38.61% in NWSP, respectively. Field tests confirmed an 8.3% reduction in total residual phosphorus and an 8.9% decrease in NWSP, without an increase in energy or raw material consumption.<br/><br/>