引用本文:刘长鑫,丁进良,姜波,柴天佑.选矿过程精矿品位自适应在线支持向量预测方法[J].控制理论与应用,2014,31(3):386~391.[点击复制]
LIU Chang-xin,DING Jin-liang,JIANG Bo,CHAI Tian-you.Adaptive online support vector regression prediction model for concentrate grade of the ore-dressing processes[J].Control Theory and Technology,2014,31(3):386~391.[点击复制]
选矿过程精矿品位自适应在线支持向量预测方法
Adaptive online support vector regression prediction model for concentrate grade of the ore-dressing processes
摘要点击 2396  全文点击 1987  投稿时间:2013-05-16  修订日期:2013-10-10
查看全文  查看/发表评论  下载PDF阅读器
DOI编号  10.7641/CTA.2014.30491
  2014,31(3):386-391
中文关键词  自适应参数  在线预测  混合核函数  支持向量机  精矿品位
英文关键词  adaptive parameters  online prediction  mixture kernel  support vector regression  concentrate grade
基金项目  国家自然科学基金资助项目(61273031, 61134006); 新世纪人才计划资助项目(NCET–12–0104); 辽宁省优秀人才支持计划资助项目 (LJQ2012020); 国家支撑计划“选矿过程全流程先进控制技术”资助项目(2012BAF19G01).
作者单位E-mail
刘长鑫* 东北大学 cxliu@mail.neu.edu.cn 
丁进良 东北大学  
姜波 东北大学  
柴天佑 东北大学  
中文摘要
      本文提出了一种基于支持向量回归的选矿过程精矿品位自适应在线预测方法, 通过使用新的混合核函数 和参数在线更新机制提高了精矿品位的预测精度. 在分析经典核函数特性后, 构造了一种混合核函数以兼顾模型的 学习能力与泛化能力, 同时为了提高预测方法对选矿生产动态过程的适应性, 模型依据新工况样本对现有样本集统 计特性的影响, 引入了模型参数自适应调整机制, 并采用在线迭代学习机制更新模型, 提高了模型的计算速度. 使用 某选矿厂生产实际数据进行实验分析, 结果表明本文方法比现有方法在计算时间和预测精度上都有明显优势, 适合 应用于动态变化的选矿生产过程.
英文摘要
      An online prediction approach is proposed for concentrate grade of ore-dressing processes based on support vector regression (SVR). The algorithm combines a new mixture kernel function and an updating technique for adaptive online parameters. In this context, different type of kernel functions is analyzed, and a new mixed kernel is developed to reach a compromise between learning capability and generalization. Furthermore, to improve the adaptability, an adaptive parameter scheme for online updating is introduced to match the changing process conditions, which considers that each new sample can change statistic properties of the training data set. Experiment results demonstrate its higher accuracy and faster computation speed than existing methods, which suggests that it is appropriate to be applied to dynamic ore-dressing production processes.