论文部分内容阅读
针对石油套管缺陷超声无损检测(NDT)中缺陷回波的特点,提出了一种基于小波包分解和支持向量机(SVM)的缺陷智能识别新方法。分析了Gabor、小波和小波包3种信号时频变换分解方法的特点,并进行了基于3种方法生成的特征数据可分性比较,确定了小波包分解方法效果最好。根据SVM解决分类问题的原理,采用SVM法对3种时频分解提取的缺陷信号特征数据进行识别。试验表明,基于小波包分解局部熵的特征提取结合SVM模式智能识别的组合方法,可应用于石油套管上的4种典型缺陷的识别。
Aiming at the characteristics of defect echoes in ultrasonic nondestructive testing (NDT) of petroleum casing defects, a new method of defect intelligent recognition based on wavelet packet decomposition and support vector machine (SVM) is proposed. The characteristics of the time-frequency transform decomposition method of Gabor, wavelet and wavelet packet are analyzed, and the separability of feature data generated by the three methods is compared. The wavelet packet decomposition method is the best. According to the principle of SVM to solve the classification problem, the SVM method is used to identify the characteristic data of the defect signal extracted from the three kinds of time-frequency decomposition. Experiments show that the combination of feature extraction based on partial entropy of wavelet packet decomposition and intelligent recognition of SVM mode can be applied to identify four typical defects on oil casing.