基于RSM和RA-BPNN的锌窑渣中铜浮选试验优化

云南某地锌窑渣矿样,原矿为高碱性矿石,矿石含铜1.38%,氧化率为30.12%。以磨矿细度、戊基黄药用量、碳酸钠用量、硫酸铜用量和硫化钠用量作自变量,以浮选回收率为因变量,分别建立了CCD响应曲面设计模型(RSM)和基于回归分析-BP神经网络(RA-BPNN)的浮选预测优化模型。根据两种模型的优化能力、优化精度进行分析和对比,结果表明,基于RA-BPNN模型进行预测及试验验证,铜回收率达到了64.06%,误差为0.74%,浮选回收率较RSM模型提高了1.54个百分点,且误差明显小于RSM模型,这表明RA-BPNN模型的优化能力高于RSM。根据试验结果,确定了锌窑渣浮选回收铜的最佳浮选条件为:磨矿细度90%、戊基黄药用量370g/t、碳酸钠用量720g/t、硫酸铜用量1080g/t、硫化钠用量870g/t。通过“一次粗选、三次精选、两次扫选、中矿顺序返回”的闭路浮选工艺流程,获得了品位为6.58%,回收率为55.98%的铜精矿。

Optimization of Copper Flotation Conditions from Zinc Kiln Slag Based on RSM and RA-BPNN Model

A sample of zinc kiln slag in Yunnan province,the ore is a highly alkaline ore,it’s contains 1. 38% copper,and the oxidation rate is 30. 12%. Based on the fineness of grinding,the amount of amyl xanthate,the amount of sodium carbonate,the amount of copper sulfate and the amount of sodium sulfide as the independent variables,and the flotation recovery rate as the dependent variable,the CCD response surface design model( RSM) was established and based on Regression analysis-BP neural network( RA-BPNN) flotation prediction optimization model. According to the optimization ability and optimization precision of the two models,the results show that the prediction and experimental verification based on RA-BPNN model,the copper recovery rate reached 64. 06%,the error is 0. 74%,and the flotation recovery rate is higher than the RSM model. The 1. 54 percentage points and the error are significantly smaller than the RSM model,which indicates that the RA-BPNN model has better optimization ability than RSM. According to the test results,the optimal flotation conditions for copper recovery from zinc kiln slag flotation are as follows: grinding fineness 90%,amyl xanthate 370 g/t,Sodium carbonate 720 g/t,copper sulfate 1080 g/t,Sodium sulfide dosage 870 g/t. Through the closed-loop flotation process of"one rough selection,three cleanning,two sweeps,and medium-sequence return",copper concentrate with a grade of6. 58% and a recovery rate of 55. 98% was obtained.