用BP神经网络建立脉动高梯度磁选过程模型  

AODELING OF PULSATING HIGH GRADIENT MAGNETIC SEPARATION WITH NEURAL NETWORK

作　　者：何平波[1] 向发柱[1] 陈荩[1] 

机构地区：[1]中南工业大学矿物工程系,长沙410083

出　　处：《有色金属》1998年第4期34-39,共6页Nonferrous Metals

摘　　要：用BP神经网络建立脉动高梯度磁选过程模型．对不同隐含层节点数的神经网络模型预测性能进行了评价，隐含层节点为13的神经网络模型选择为最佳模型．利用选择的最佳模型，对黄铜矿高梯度磁选进行模拟研究．模型研究结果表明，在相当宽操作的范围内，模型能够很好地预测磁选精矿中铜的品位和回收率．这说明建立的高梯度磁选模型合理可行．A sigmoid backpropagation neural network model for pulsationg high gradient magnetic separation is developed in this paper. The performance of various neural network models containing 4 to 14 hidden nodes is estimated. The network containing 13 hidden nodes which performs best on a test data set is selected as the most adequate model of Pulsating high gradient magnetic separation process. The simulation studies on the application of backpropagation neural network model to the high gradient magnetic separation processes of chalcopyrite are carried out. The simulation results show that the model is capable of making a good prediction of the copper grades and recoveries to magnetic concentrates over a broad range of operating conditions, and it is demonstrated that the developed model is reasonable and feasible.

关 键 词：磁选 神经网络 数学模型 选矿 黄铜矿 

分 类 号：TD924[矿业工程—选矿] TD952.1