支撑向量机回归在炼焦配煤优化中的应用  

Pattern recognition classification and support vector regression and their applications to optimize the coal blend

作　　者：陆军[1] 姚莉秀[1] 陈念贻[1] 马智明 

机构地区：[1]上海交通大学图像处理与模式识别研究所,上海200030 [2]河南安阳钢铁集团技术中心,河南安阳455004

出　　处：《计算机与应用化学》2006年第8期703-706,共4页Computers and Applied Chemistry

基　　金：国家自然科学基金资助(50174038)宝钢集团联合资助(50174038)

摘　　要：针对7个不同产地煤的煤质特性参数(X,Y,G)和配比,在数据结构类型未知的情况下,本工作融合了基于线性变换的PCA、PLS和基于非线性变换的Lmap方法,即在各算法的空间变换基础上,同时考虑每个二维投影的分类效果(计算各二维投影图上的各类点的类间距和类内距),从中获得一个各类之间分类效果最好的最佳投影。由此定性研究了上述煤质特性参数和配比对焦炭机械强度和推焦电流的影响。然后在良好分类效果基础上,用支撑向量回归方法分别总结了该组参数与其对应的目标之间的定量关系,建立了预测模型,得到了准确的回归预测结果,由此可有目的地调节配煤比例以更好地提高炼焦效果。The influence of coal blending on the corresponding coke quality has been investigated. First, pattern recognition methods are used to analyze the qualitative relation between the coke quality and 28 parameters, X, Y, G and ratio of every coal from seven different producing areas. It can be concluded that the samples with good coke quality can be elearly separated from that with bad coke quality on the space spanned by these 28 parameters. Then, support vector regression is used firstly to investigate the quantitative relation between all these parameters and the corresponding coke quality and the power for coke pushing. And a mathematical model is built to predict the coke quality and the power for coke pushing. So the coke making could be optimized by adjusting the ratio of coke blending purposefully. The obtained good results showed that support vector regression could be used in the optimization of such complex industry processing efficiently.

关 键 词：配煤 焦炭强度 模式识别 支撑向量回归 

分 类 号：TQ522.16[化学工程—煤化学工程]