摘要
运用主成分分析方法对MgO-C耐火材料损伤实验获取的声发射参数进行消除相关性降维处理,以构造出的新参数作为损伤信号分类的指标,采用高斯混合模型聚类方法及贝叶斯信息准则将样本信号分为ω1、ω2两类,其权重分别为63%和37%。扫描电镜验证结果表明,MgO-C耐火材料受载下的损伤形式主要为基质相损伤和界面损伤两种,并以基质相损伤形式为主。
Through the principal component analysis （PCA） method ,two parameters were selected to eliminate the correlation with dimension reduction treatment .The new parameters were the linear combinations of the 15 original parameters and taken as the indexes of damage signal classification . The Gaussian mixture model was used together with the Bayesian information criterion to group the sample signals into two categories ,i .e .ω1 、ω2 ,whose weights were 63% and 37% ,respectively .Electronic microscope scanning shows that the damages of MgO-C refractory are mainly matrix phase damage and interface damage with the former accounting for the larger part .
出处
《武汉科技大学学报》
CAS
2014年第4期269-272,281,共5页
Journal of Wuhan University of Science and Technology
基金
国家自然科学基金资助项目(51075310)
教育部博士点专项基金项目(20114219110003)
湖北省自然科学基金重点项目(2012FFA022)
教育部新世纪优秀人才支持计划资助项目(NCET-12-0715)
中国博士后基金项目(2013M542072)
关键词
耐火材料损伤
损伤类别
声发射参数
主成分分析
高斯混合模型
贝叶斯信息准则
扫描电镜
refractory damage
damage category
AE parameter
principal component analysis
Gaussian mixture model
Bayesian information criterion
electronic microscope scanning
