An integrated predictive model with an on-line updating strategy for iron precipitation in zinc hydrometallurgy

• We propose the idea and framework of the integrated model for ferrous ion concentration prediction.
• A mechanism model is built with some parameters to be identified.
• A new particle swarm optimization algorithm IEDPSO is proposed to identify the parameter.
• A data-driven error compensation model is established to minimize the output prediction error of the mechanism model.
• The integrated model is calibrated based on sensitivity analysis and proved effective.

Iron precipitation by goethite plays an important role in zinc hydrometallurgy. The ferrous ion concentration, which is a key index for assessing the iron removal rate and process control results, cannot be measured on-line. In this study, an integrated predictive model of the ferrous ion concentration is established by integrating the mechanism model and error compensation model, which is based on data identification. The mechanism model is proposed based on an analysis of the process reaction and considering the reaction unit as a continuous stirred tank reactor model. For unknown parameters in the mechanism model, a double-particle swarm optimization algorithm based on information exchange and dynamic adjustment of the feasible region is developed for optimal selection. To improve the adaptive capability of the integrated model, we propose a model-updating strategy and parameter calibration method based on a sensitivity analysis to accomplish on-line adaptive updating of the predictive model. The simulation results demonstrate that the proposed model can effectively track the variation tendency of the ferrous ion concentration and successfully improve the adaptability of the integrated model.