Additive requirement ratio prediction using trend distribution features for hydrometallurgical purification processes

•A general definition of additive efficiency is proposed to amend theoretical additive amount calculated by a process model in the hydrometallurgical purification process.•A novel case based prediction strategy using trend distribution features (CBP-TDF) is developed.•The CBP-TDF is used for pre-estimating the additive efficiency in the purification process.•The effectiveness of the additive efficiency and the proposed pre-estimation method is confirmed through an industrial case study.

A purification process is to remove impurities through a series of reactors with additives. The theoretical calculated amount of additive does not fulfill actual requirements due to variations in the reaction environment. An additive requirement ratio is thus defined to measure the disparity between theoretical calculation and actual requirements. Considering the influence of the process underlying variations, a novel ratio prediction strategy, case-based prediction with trend distribution feature (CBP-TDF), is developed. In the strategy, the trend distribution features are firstly extracted to describe the underlying variations, and an improved case-based prediction algorithm is proposed where the similarity between these features is computed based on Kullback–Leibler divergence. The proposed strategy is applied to a copper removal process of zinc hydrometallurgy. The experiments indicate the accuracy of the ratio prediction, and the industrial application shows its effectiveness in the control of the purification process.