Abstract: Combining with the actual working conditions of flotation process in concentrator， the key process variables and process indexes of flotation process in site were collected for a long time， and a prediction model of flotation reagent flow based on LM-BP neural network was put forward. The results of data cross-validation show that this method can predict the flotation reagent scheme reasonably make the flotation pulp reach the optimum mineralization state， and then optimize the flotation indicators on the premise that the concentrate grade， recovery and other indicators meet the production requirements. It has a certain reference value for reducing the production cost of flotation process in the plants.

Key words: LM-BP neural network, Flotation reagent flow prediction model, Flotation reagent scheme