Abstract: The echo signal components of Carbon Fiber Reinforced Plastics（CFRP） acquired by ultrasonic nondestructive testing are often complex, which results in the unconspicuous flaw characteristic. Therefore, traditional wavelet based methods are failed to extract features of CFRP ultrasonic flaw signals. To solve the problem, the local energy feature extraction method based on dual tree complex wavelet packet transform is proposed to obtain the original feature vector for CFRP flaw signals. After that, the ε-reduct method based on rough set is used for feature dimension reduction. The experimental results show that the proposed methods are effective, which would provide new approach to automatically and correctly recognize different kinds of flaws for CFRP.

Key words: ultrasonic flaw signals, feature extraction, feature reduction

摘要： 在对碳纤维复合材料进行超声无损检测时获取的回波信号往往构成复杂，某些缺陷特征不明显，使用传统小波方法对这类信号进行特征提取时效果并不理想。为解决上述问题，提出基于双树复小波包变换的频带局部能量特征提取方法以获取碳纤维复合材料超声缺陷信号的初始特征向量；在此基础上，使用基于粗糙集的ε-约简方法完成特征降维。实验结果验证了所提出方法的有效性，为实现碳纤维复合材料缺陷的自动和准确识别提供了新途径。

关键词: 超声缺陷信号, 特征提取, 特征降维