基于泡沫尺寸随机分布的铜粗选药剂量控制  

Reagent Dosage Control Based on Bubble Size Random Distribution for Copper Roughing

作　　者：朱建勇[1,2] 桂卫华[1] 阳春华[1] 吴佳[1] 周文振[1] 

机构地区：[1]中南大学信息科学与工程学院,长沙410083 [2]华东交通大学电气与电子学院,南昌330029

出　　处：《自动化学报》2014年第10期2089-2097,共9页Acta Automatica Sinica

基　　金：国家自然科学基金重点项目(61134006);国家杰出青年科学基金(61025015);国家创新研究群体科学基金(61321003);教育部人文社科青年基金项目(13YJCAH089)资助~~

摘　　要：为了稳定铜粗选选矿指标,提高矿产资源的利用水平,根据铜粗选过程中泡沫尺寸分布随药剂量改变而动态变化的特点,提出一种基于泡沫尺寸随机分布的铜粗选过程药剂量控制方法.首先,针对泡沫尺寸分布具有非高斯统计特性,基于方差和均值的统计参量难以表征该分布形态变化的问题,提出了B样条估计方法以描述泡沫尺寸的概率密度函数(Probability density function,PDF);然后,针对B样条权值相互关联的特点,建立多输出最小二乘支持向量机模型(Multi-output least square support vector machine,MLS-SVM)以表征权值和药剂量的动态关系;最后,为减少系统的随机性,采用基于熵的优化算法以确定药剂量,实现对给定泡沫尺寸分布的跟踪控制.工业数据仿真验证了所提方法的有效性,能有效稳定铜粗浮选的生产指标.A bubble size distribution （BSD） based reagent dosage control method is proposed to stabilize the indices and improve the utilization of mineral resources in copper roughing flotation. Firstly, in order to characterize the bubble size structure with non-Gaussian feature, a B-spline estimator is investigated to depict the BSD. Since the weights of B-spline are interrelated with each other and related to the reagent dosage, a multi-output least square support vector machine （MLS-SVM） is hence used to establish their dynamical relationship between B-spline weights and reagent dosage. Finally, an entropy based optimization algorithm is constructed to determine the reagent dosage in order to implement the tracking of the given BSD. Simulation and experimental results show the effectiveness of the proposed method, which has been used to stabilize the product indices for copper roughing.

关 键 词：铜粗选 泡沫尺寸分布 药剂量控制 概率密度函数 最小二乘支持向量机模型 

分 类 号：TD923.1[矿业工程—选矿]