摘要
为准确识别摩擦焊试件接头主要缺陷类型,采用小波包变换对一维超声信号进行处理,利用能量故障"法提取信号特征值,并将特征值引入一对多"支持向量机进行分类识别.通过验证发现,支持向量机较好解决了小样本、非线性和高维数问题,准确率高、容易在线实施,具有较强推广能力.
For identificating main kinds of defi- ciencies in the friction welded joints accurately, then using wavelet packet tansform to analyse de- tectional signal of one - demension in the experi- ence,extract the eigenvalues with ’energy-fail- ure' method,then cite these eigenvalues in ‘1 -v- r' support vector machine to classficate and identi- float. The experiment shows that, support vector machine works accuractly and can be implied online easily, solve relatively well less samples, nonlinear, high dimension problems and is practical.
出处
《机械与电子》
2010年第4期36-39,共4页
Machinery & Electronics
关键词
摩擦焊
无损检测
超声波
小波包
支持向量机
friction welding
NDT
ultrasonic
wavelet packet
support vector machine(SVM)