Prediction of cobalt ions concentration in zinc smelting purification section II based on classification modeling

The data in the zinc hydrometallurgy process usually have multimodal characteristics due to the change of working conditions, which brings difficulties to the online estimation of cobalt ions. Existing machine learning algorithms may not be able to capture the complex relationships between variables under multimodal condition. In this paper, a classification modeling idea is proposed to solve the multimodal problem of cobalt removal process in zinc hydrometallurgy. To be specific, the preprocessed data are first classified into n classes, each class is reconstructed by the AutoEncoder. The reconstructed data are utilized to capture the complex nonlinear relationships between the data in different modes. The Support Vector Regression(SVR) is used to model each class of data, and the parameters of the SVR are optimized by the hippopotamus optimization(HO) algorithm. In the testing phase, the test set data are compared with the class centres of the training data to determine the category which it belongs to and the corresponding model will be used to obtain cobalt ion concentration predictions. The experimental results in the purification section of a zinc smelter show that the proposed method is superior and can provide guidance for the optimisation of process parameters in the purification process.