| 基于EMD和多重分形谱参数的耐火材料损伤模式识别 |
| |
| 作者姓名： | 苏 涛 王志刚 刘昌明 徐增丙 但斌斌 |
| |
| 作者单位： | 1.武汉科技大学机械自动化学院,湖北 武汉,430081;2.武汉科技大学冶金装备及其控制教育部重点实验室,湖北 武汉,430081,1.武汉科技大学机械自动化学院,湖北 武汉,430081;2.武汉科技大学冶金装备及其控制教育部重点实验室,湖北 武汉,430081,1.武汉科技大学机械自动化学院,湖北 武汉,430081;2.武汉科技大学冶金装备及其控制教育部重点实验室,湖北 武汉,430081,1.武汉科技大学机械自动化学院,湖北 武汉,430081;2.武汉科技大学冶金装备及其控制教育部重点实验室,湖北 武汉,430081,1.武汉科技大学机械自动化学院,湖北 武汉,430081;2.武汉科技大学冶金装备及其控制教育部重点实验室,湖北 武汉,430081 |
| |
| 基金项目： | 国家自然科学基金资助项目(51075310,51405353,51475340)；教育部新世纪优秀人才支持计划项目(NCET-12-0715)；教育部博士点专项基金资助项目(20114219110003)；中国博士后科学基金资助项目(2013M542072)；湖北省自然科学基金资助项目(2012FFA022,2014CFB825). |
| |
| 摘 要： | 考虑到耐火材料损伤声发射信号模式识别困难,提出一种结合经验模态分解(EMD)、多重分形谱参数和支持向量机的耐火材料损伤形式分类方法。首先对耐火材料损伤声发射信号进行EDM分解得到若干本征模态函数(IMF)分量,并取前4个分量作为研究对象,然后将整个信号的多重分形谱宽及各IMF分量的多重分形谱宽组成的特征向量输入支持向量机进行学习训练,最后实现耐火材料损伤模式识别。研究结果表明,采用由原信号及各IMF分量的多重分形谱宽值组成的特征向量能够有效进行损伤信号的特征提取。该方法对耐火材料界面相损伤的分类准确率为99%,对其基质相损伤的分类准确率为89%。
|
| 关 键 词： | 耐火材料 材料损伤 模式识别 声发射 经验模态分解 多重分形谱 支持向量机 |
Refractory damage pattern recognition based on EDM and multi-fractal spectrum parameters |
| |
| Authors: | Su Tao Wang Zhigang Liu Changming Xu Zengbing and Dan Binbin |
| |
| Institution: | 1. College of Machinery and Automation, Wuhan University of Science and Technology, Wuhan 430081,China;2.Key Laboratory of Metallurgical Equipment and Control of Ministry of Education, Wuhan University of Science and Technology, Wuhan 430081,China,1. College of Machinery and Automation, Wuhan University of Science and Technology, Wuhan 430081,China;2.Key Laboratory of Metallurgical Equipment and Control of Ministry of Education, Wuhan University of Science and Technology, Wuhan 430081,China,1. College of Machinery and Automation, Wuhan University of Science and Technology, Wuhan 430081,China;2.Key Laboratory of Metallurgical Equipment and Control of Ministry of Education, Wuhan University of Science and Technology, Wuhan 430081,China,1. College of Machinery and Automation, Wuhan University of Science and Technology, Wuhan 430081,China;2.Key Laboratory of Metallurgical Equipment and Control of Ministry of Education, Wuhan University of Science and Technology, Wuhan 430081,China and 1. College of Machinery and Automation, Wuhan University of Science and Technology, Wuhan 430081,China;2.Key Laboratory of Metallurgical Equipment and Control of Ministry of Education, Wuhan University of Science and Technology, Wuhan 430081,China |
| |
| Abstract: | Considering the difficulty of pattern recognition of the acoustic emission signals of refractory damage, this paper proposes a classification method for refractory damage pattern based on empirical mode decomposition (EMD), multi-fractal spectrum parameters and support vector machine. First, the acoustic emission signals are decomposed into several intrinsic mode function (IMF) components by EDM, and the first four components are taken as the research objects. Second, a feature vector formed by multi-fractal spectrum width values of the entire signal and IMF components is used in learning and training of support vector machine (SVM). Then the refractory damage pattern classification is completed by SVM. The results show that the constructed feature vector is efficient in feature extraction of damage signals. The classification accuracy of this method for interface damage and matrix damage of refractory can reach up to 99% and 89%, respectively. |
| |
| Keywords: | refractory material damage pattern recognition acoustic emission EMD multi-fractal spectrum SVM |
| 本文献已被 CNKI 等数据库收录！ |
| 点击此处可从《》浏览原始摘要信息 |
| 点击此处可从《》下载免费的PDF全文 |
|
