Series Dynamic Hybrid Model Based on Material Conservation and Neural Networks for Oxidation Leaching Process

Oxidation leaching process is an important stage of the hydrometallurgy of zinc, and the establishment of its mathematical model is of great significance to the implementation of optimal control and monitoring. As a result, a series dynamic hybrid model based on material conservation and neural network for oxidation leaching process is developed in this work. On the on hand, in order to eliminate the shortcoming that traditional first principle models are easy to fall into the local working condition, based on the traditional mechanism framework of material conservation, this paper proposes the solid phase mass conservation equation. On the other hand, for obtaining the reaction rate model more conveniently and discovering the deep information which is difficult to explain by mechanism, the neural network is introduced to replace the original complex kinetic reaction rate equation. In addition, for the purpose of reducing the influence of noise amplification and transmission induced by derivative operation, this paper introduces integration regularization method to estimate the reaction rate, thus it further increases the accuracy and anti-noise ability of the proposed model. Finally, extensive experiments are conducted to verify the accuracy and efficiency of the proposed model, the numerical results indicated that the proposed method is superior than the traditional first principle model and data-driven model.