Abstract
Copper flotation process is a complex industrial process. Concentrate copper grade (CCG) is one of the important production indexes of flotation process, and it is also one of the important factors to determine the economic benefit. Therefore, the accurate prediction of CCG can not only provide operational advice for copper flotation process, but has important economic value and practical significance. Due to the long process of copper flotation, the existence of time delay and the serious coupling of variables, the accurate prediction of CCG is challenging. Based on the analysis and study of copper flotation process, this paper adopts partial least squares (PLS), kernel partial least squares (KPLS), least squares support vector machine (LSSVM) and long short term memory (LSTM) methods to predict the CCG for a copper flotation process. All of the proposed prediction models are validated by the field data in this paper and it can be learnt from the prediction evaluation indexes for root mean square error (RMSE), mean absolute error (MAE) and mean absolute percentage error (MAPE) that the prediction accuracy based on LSTM model is higher and suitable for the prediction of CCG for the copper flotation process.