一种基于混合神经网络的浮选pH值预测模型  

Prediction Model for pH Value in Flotation Process Based on Hybrid Neural Network

作　　者：唐朝晖[1] 杜金芳[1] 陈青[1] 

机构地区：[1]中南大学信息科学与工程学院,湖南长沙410083

出　　处：《控制工程》2012年第3期416-419,共4页Control Engineering of China

基　　金：国家自然科学基金(61071176);国家自然科学基金(60874069)

摘　　要：矿物浮选过程中,矿浆pH值作为影响浮选效果的一个重要因素,是实现浮选过程监视及优化控制的一个重要参量。目前的pH值测定仪存在交叉污染、测量滞后等问题,难以获得实时准确的pH值。为使浮选运行在最优状态,在泡沫图像特征提取的基础上,提出一种基于自适应遗传混合神经网络的预测模型,该模型首先利用主元分析(PCA)方法对提取的多个图像特征进行降维,然后采用自适应遗传混合神经网络(AGA-HNN)建立pH值预测模型。最后将该模型应用于浮选现场,预测结果能够实时跟踪实际值,根据预测值实时调整工况条件,改善了浮选效果,提高了浮选效率。In mineral flotation process, pH value is one of the flotation elements which affect the flotation performance significantly. It is very important for flotation process monitoring and optimized controll. At present, pH determinator has the problem of cross contami- nation, measurement lag, and so oh. So it is difficult to obtain real time and accurate pH value. To make flotation running in an optimal state, a novel prediction model is proposed in this paper based on adaptive genetic hybrid neural network after extracting several image features. Firstly, feature dimension reduction is done by principal component analysis（PCA）. Then prediction model is built through a- daptive genetic hybrid neural network（AGA-HNN）. Finally, the model is applied to flotation field. Predicted value can well trace the actual value. At the same time, working condition is adjusted according to the predicted value. As a result, the flotation performance and efficiency are improved obviously.

关 键 词：矿物浮选 泡沫图像 预测模型 主元分析(PCA) 自适应遗传混合神经网络 (AGA-HNN) 

分 类 号：TP27[自动化与计算机技术—检测技术与自动化装置]