基于支持向量机的超声检测缺陷识别研究  

Support vector machine based flaw recognition in ultrasonic testing

作　　者：袁腾月[1,2] 夏筱筠[2] 郭烽[3] 

机构地区：[1]中国科学院研究生院,北京100049 [2]中国科学院沈阳计算技术研究所,辽宁沈阳110168 [3]中航工业沈阳黎明航空发动机(集团)有限责任公司,辽宁沈阳110043

出　　处：《计算机工程与设计》2013年第4期1260-1264,1301,共6页Computer Engineering and Design

基　　金：国家科技重大专项-开放式数控系统支撑技术创新平台建设基金项目(2011ZX04016-071)

摘　　要：为解决超声检测领域传统人工神经网络方法对于小样本进行缺陷识别时存在的泛化能力差和过学习等问题,提出了一种基于支持向量机的超声检测缺陷识别方法。先使用小波分解对信号进行降噪,再使用小波包变换提取特征值,构造多类分类支持向量机进行缺陷识别。实验结果表明,支持向量机方法具有识别率高、泛化能力强等优点,能够应用于超声检测缺陷识别领域。In order to solve the problems of poor generalization ability and over learning that the traditional artificial neural net- work suffers in flaw recognition in ultrasonic testing field when the sample sets are small, a new flaw recognition method based on support vector machine is proposed. Wavelet decomposition is applied to de-noising of the signals. Then wavelet packet trans- form is utilized in feature extraction. At last a multi-class support vector machine is constructed to identify flaws. The experi- mental result suggests that the support vector machine method has advantages of high accuracy ratio of flaw recognition and strong generalization ability and it can be applied in flaw recognition in ultrasonic testing.

关 键 词：超声检测 小波分解 小波包变换 支持向量机 缺陷识别 

分 类 号：TP274.53[自动化与计算机技术—检测技术与自动化装置]