基于改进的FOA-LSSVM磨矿粒度软测量模型  

Soft-measuring Model for Grinding Size Based on the Improved FOA-LSSVM Model

作　　者：张燕[1,2] 代亚菲 陈玲玲[1,2] 宣伯凯[1] 

机构地区：[1]河北工业大学控制科学与工程学院,天津300130 [2]河北省控制工程技术研究中心,天津300130

出　　处：《矿业研究与开发》2015年第11期97-103,共7页Mining Research and Development

基　　金：国家自然科学基金项目(61203323);河北省自然科学基金项目(F2011202094);河北省高等学校科研项目(Q2012079)

摘　　要：以典型的两段磨矿回路为研究对象,针对磨矿粒度在线检测困难而难以满足实时控制的难题,提出一种新的软测量建模方法。应用最小二乘支持向量机(LSSVM)进行磨矿粒度软测量建模,为解决参数设置的盲目性,利用改进的变步长果蝇优化算法(FOA)较强的寻优能力对LSSVM的惩罚系数和核参数进行优化。对该模型进行预测仿真,同时与网格搜索法、粒子群法和未改进的果蝇算法优化的LSSVM模型进行对比实验。结果表明,相对于其他模型,改进的FOA-LSSVM收敛速度快,预测精度最高,较好地实现了对磨矿粒度的实时检测。Online testing of grinding size is difficult to meet the requirements of real-time control. According to the prob- lem, a new soft-measuring modeling method was put for- ward through taking the typical two-stages grinding circuit as a research object. The soft-measuring model for grinding size was conducted by LSSVM. To solve the blindness of setting parameters, the strong optimization ability of the im- proved FOA was used to optimize the penalty coefficient and kernel parameters of LSSVM. Then, the forecast simulation of models was carried out, meanwhile the contrast experi- ments were proceed with the LSSVM models optimized by grid search method, particle swarm method and original fruit fly algorithm. The results showed that the improved FOA-- LSSVM model, compared with other models, had fast con- vergence speed and high prediction accuracy, which well re- alized the real-time detection of grinding size.

关 键 词：磨矿粒度 软测量 实时检测 改进的果蝇算法 最小二乘支持向量机 

分 类 号：TD921.5[矿业工程—选矿]