Based On RSM And BP Neural Network To Predict The Amount Of Reagent Added In The Concentrator

The market is more and more strict to the concentrate product quality,the concentrate grade and recovery index will change accordingly due to the change of the original ore property.The flotation process is the core part of the beneficiation process.The purpose and significance of this study is to improve the economic benefits.When the flotation index reaches a certain degree,there is a contradiction between the grade and recovery of the concentrate.When the grade of concentrate decreases,the recovery rate of concentrate increases.With the change of copper concentrate price in the market,there will be an optimal balance point of economic value.According to the change of ore properties,the most reasonable grade and recovery rate of copper concentrate will be determined by predicting the amount of flotation agent.So this article is based on the BP neural network and the RSM surface response method to predict reagent adding research,put forward neural network input data to predict reagent adding amount-test index analysis of factors influencing academic ideas.For the flotation process of agents intelligent addition,and in the mineral processing model is not effective and practical.In this paper,the study is carried out from two aspects: the BP neural network modeling to predict the amount of agent and the experimental model to optimize the flotation index.Based on collected actual production data in a Sichuan copper mine,the BP neural network contain adaptive self organization,fault tolerance and strong characteristics,based on the BP neural network.Input layer establish 4,hidden layer 7 and output layer is 2 neural network structure,in 2015-2017 years ago 30 months of actual production data as the training sample data,the neural network training largest number 5000 times,the neural network learning rate 0.05,neural network target error of 0.65 * 10-3,input layer type contain ore grade,oxidation rate,concentrate grade and concentrate recovery,the output layer contain the data of xanthine dosage and pine oil dosage,the BP neural network model with the optimal prediction effect is selected from the four algorithm models through the comparison of learning and training analysis results by BP neural network algorithm,the linear regression R value is 0.99904 > 0.99,and the additive system of agents is predicted.In order to verify that under the conditions of different reagent combinations,the linear regression R values of the neural network for lime,light diesel and sodium sulfide were greater than 0.99927,0.99899 and 0.99923,respectively.The BP neural network model is of good quality,and it can predict the prediction effect of the dosage of combined agents,a good prediction result is obtained.The application of this model will provide broad space for the development of intelligent mines in the future.In order to further study on basis of the interaction effect of indicators,through the response surface method RSM for the experiment design optimization,according to the arrangement of the CCD experimental Design,the design contain 3 factors 3 levels experiment,three factors influencing the oxidation rate of ore grade and recovery.concentrate grade and recovery rate p values were 0.0400 and 0.0200 are less than 0.05,the index of the experimental results show that factors affecting test has a significant effect.Under the three-dimensional surface diagram,it can be analyzed that when the oxidation rate is fixed,the concentration grade increases significantly with the increase of raw ore grade and the amount of xanthine added.The F values of concentrate grade and recovery were calculated to obtain A52.94,B8.9,C18.43,A16.91,B1.65 and C5.35.The primary and secondary factors affecting concentrate grade and recovery were obtained as follows: A raw ore grade > Cxanthine addition > B oxidation rate,and there was an obvious interaction between the three factors.Finally,under the optimal condition of RSM index,the optimal ratio of influencing factors under the optimal concentrate grade and concentrate recovery was calculated and the obtained.The optimal ratio of influencing factors was obtained for A raw ore grade 0.846% B oxidation rate 0.968% C xanadin amount 23.686g/t,so as to obtain the optimal index of concentrate grade 25.152 % and recovery rate 94.410%.It provides a convenient and efficient experimental design method for mineral processing researchers.