摘要
针对密闭鼓风炉故障诊断中难以获得大量故障数据样本以及特征提取和诊断知识获取困难等不足,提出了应用支持向量机(SVM)进行故障诊断的新方法.采用改进"1对其余"算法构建多个SVM,利用可靠性数据分析技术中一些基本概念处理原始样本数据作为特征向量,输入到由多个SVM构成的多类分类器中进行故障分类.经实验证明,该方法简单,重复训练量少,训练、分类速度快,准确度高.
Aiming at the difficulty in getting adequate fault samples, extracting eigenvectors and acquiring diagnosis knowledge in fault diagnosis for the lead-zinc imperial smelting furnace, a novel method for the furnace fault diagnosis based on support vector machine （SVM） is put forward. An improved ＇one to others＇ algorithm is introduced to construct the multi-class SVM classifier. Some basic conceptions of data reliability analysis are adapted to preprocess the data as the input of the multi-class classifier to identify faults. The method is simple and has little repeated training amount. And the excellent performance on training speed and accuracy has been verified in the real application.
出处
《小型微型计算机系统》
CSCD
北大核心
2008年第4期777-781,共5页
Journal of Chinese Computer Systems
基金
国家“九七三”项目(2002CB312200)资助
关键词
支持向量机
可靠性数据分析
故障诊断
SVM多类分类器
support vector machine
data reliability analysis
fault diagnosis
multi-class SVM classifier
作者简介
蒋少华，女，1966年生，博士研究生，高级工程师，研究方向为智能控制、复杂过程故障诊断等；E-mail：sgjsh66@hotmail．com
桂卫华，男，1950年生，研究方向为复杂过程的建模、优化及故障诊断等；
阳春华，女，1965年生，教授，博士生导师，研究方向为复杂过程优化控制、故障诊断等．
