基于泡沫图像特征的浮选槽液位智能优化设定方法  

An Intelligent Optimal Setting Approach Based on Froth Features for Level of Flotation Cells

作　　者：赵洪伟[1] 谢永芳[1] 蒋朝辉[1] 徐德刚[1] 阳春华[1] 桂卫华[1] 

机构地区：[1]中南大学信息科学与工程学院,长沙410083

出　　处：《自动化学报》2014年第6期1086-1097,共12页Acta Automatica Sinica

基　　金：国家自然科学基金重点项目(61134006);国家创新研究群体科学基金(61321003);高等学校博士学科点专项基金博导类资助课题(20120162110076);高等学校博士学科点专项科研基金优先发展领域资助课题(20110162130011);中央科研基本业务费中南大学国家杰青培育专项(2011JQ009)资助~~

摘　　要：浮选生产过程中浮选槽液位通常根据经验人工设定,具有主观随意性﹑液位波动大,使精/尾矿品位不满足要求.为此,提出一种基于浮选泡沫图像多特征的浮选槽液位智能优化设定的方法.在浮选槽工作原理以及液位与泡沫图像特征间关系的分析基础上,将基于案例推理的浮选槽液位预设定﹑基于多泡沫图像特征的改进LS-SVM(Least squares support vector machine)品位预测及基于BP神经网络的自学习模糊推理智能补偿等模型有机集成,提出了充分利用泡沫图像特征的液位智能优化设定方法.将该方法在某铝土矿浮选生产过程进行应用验证,可使粗选槽液位波动减小,提高了粗选精/尾矿品位合格率、总精矿品位合格率及回收率.In the flotation production process, the liquid level of flotation cells is usually set by on experiences. The liquid level can fluctuate in a large range such that the concentrate grade and tailings grade may not meet the requirement. In this paper, an intelligent optimal setting approach based on forth image features is proposed. On the basis of analysis of flotation cells/ working principle and relationship between level and froth image features, the pre-setting model based on CBR, the improved least squares support vector machine （LS-SVM） grade prediction model based on multiple froth features, and the self-learning fuzzy reasoning intelligent compensation model based on BP neural network are integrated together. This method is tested on a bauxite flotation process. The level fluctuation of rougher flotation cells decreases. The pass rates of concentrate and tailings grade in rougher flotation and the pass rate of the concentrate grade and recovery of the overall flotation increase.

关 键 词：浮选液位 泡沫特征 案例推理 品位预测 BP神经网络模糊推理 优化设定 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]