Prediction of neutral leaching ferric ion concentration based on DGC-XGBoost

Neutral leaching is an important part of wet zinc refining. However, the detection of ferric ion concentration in the neutral leaching process is a great challenge for traditional testing instruments due to the harsh environment and complex working conditions. Data detection has a substantial lag, which greatly increases the difficulty of field control. To solve this problem, the DGC-XGBoost modelling method is proposed in this study. The Dual Gated Convolution(DGC) is used to extract features among variables, and the combination of the original features is used as input of the XGBoost model to predict the ferric ion concentration. The XGBoost adopts the boosted gradient algorithm to train the model, while the traditional gradient calculation can not update the parameters of the DGC model. To address this problem, the prediction results of XGBoost are constructed as fitness functions, and the parameters are updated directly using the Spider Wasp Optimizer (SWO). We use a zinc refinery’s data collected directly from factory to verify the proposed model’s effectiveness. The simulation results show that the proposed method achieves the best prediction results among LSTM, XGBoost, SVM, RF, GBDT, and CNN-BiLSTM.