基于ELM的硫熏强度软测量模型  

ELM-based Soft-sensing Modeling for Intensity of Sulfitation

作　　者：宋绍剑[1] 刘小兵[1] 林小峰[1] 

机构地区：[1]广西大学电气工程学院,南宁530004

出　　处：《计算机测量与控制》2015年第3期713-715,共3页Computer Measurement &Control

基　　金：国家自然科学基金重点项目(610034002);桂科攻(14122007-33)

摘　　要：硫熏强度是亚法糖厂澄清工段的关键工艺参数之一;硫熏强度过低会影响澄清效果,过高会造成成品糖二氧化硫残留过高;目前由于缺乏合适的硫熏强度在线测量装置,该参数主要采用人工取样和离线化验的检测方式,化验滞后时间较长,难以根据该指标及时指导实际生产的问题;为此文章采用极限学习机(ELM)方法建立了硫熏强度软测量模型,并与基于支持向量机(SVM)、径向基函数(RBF)神经网络和反向传播(BP)神经网络的硫熏强度软测量模型进行对比分析;结果显示,基于ELM的硫熏强度软测量模型具有训练收敛速度快、模型精度高和泛化性能好等优点,可以满足实际糖厂澄清工段的要求。The intensity of sulfitation is one of the key process parameters of clarification process of Cane sugar factory. If the intensity of sulfitation becomes too low, as a result, the clarification effect will be influenced. In contrast, if the intensity of sulfitation becomes too high, it will bring high sulfur dioxide residue in the final while sugar. Due to lack of suitable on--line measuring devices currently, the intensity of sulfitation mainly takes manual samp!ing and laboratory offline testing mode. As testing lag time is long, it is difficult to guide the practical production problem timely according to the index. This paper takes use of Extreme Learning Machine （ELM） method to establish the intensity of sulfitation soft--sensing model, and contrastively analyses the model based on Support Vector Machine （SVM）, Radial Basis Function （RBF） neural network and Back--Propagation （BP） neural network. The result shows that the ELM--based soft--sensing modeling for the intensity of sulfitation has the advantages of training convergence speed, high model precision and good generalization performance, etc. It can meet the requirements of clarification process of actual sugar mill .

关 键 词：澄清过程 硫熏强度 软测量 神经网络 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]