基于泡沫图像特征的金锑浮选入矿品位估计  

Mineral concentration estimation of feed ore in gold and stibium flotation based on froth image features

作　　者：吴佳[1] 谢永芳[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,湖南长沙410083

出　　处：《控制理论与应用》2015年第2期262-266,共5页Control Theory & Applications

基　　金：国家自然科学基金项目(61134006;61473318);国家杰出青年科学基金(61025015)资助~~

摘　　要：入矿品位是金锑浮选加药量控制的重要依据.针对入矿品位在线检测困难的问题,提出一种基于泡沫图像特征的入矿品位估计方法.该估计方法首先针对样本数据中存在的不确定性,提出一种基于核主元分析(KPCA)和模糊C均值聚类–概率支持向量回归(FCM--PSVR)的建模方法,然后利用泡沫图像特征与加药量等数据建立起金锑入矿品位和精矿品位的估计模型,最后采用基于专家规则的方法对入矿品位估计结果的可信度进行评价.该方法在金锑浮选工艺中进行了工业验证,为指导金锑浮选加药量的控制起到了重要作用.The regulation of reagents heavily depends on the mineral concentration of feed ore in flotation. Since it is difficult to measure the concentration of feed ore online, an estimator based on froth image features is developed in this paper. At first, considering the uncertainty existing in the sample data, we develop a soft-sensor method based on kernel principal component analysis （KPCA） and fuzzy C-means clustering-probabilistic support vector regress （FCM-PSVR）. Then, estimators for the gold and stibium concentration of feed ore and concentrate are built. Finally, the output of the estimators of feed ore is evaluated based on expert rules. The estimator has been validated in the gold and stibium froth flotation, and it plays an important role in guiding the reagent addition in gold and stibium flotation.

关 键 词：浮选 不确定性 支持向量机 模糊聚类 

分 类 号：TD923[矿业工程—选矿] TP391.41[自动化与计算机技术—计算机应用技术]