基于主成分SVM的防腐层缺陷分类识别算法研究

Research on Classification and Recognition Algorithm of Anticorrosive Layer Defects Based on Principal Component SVM

管道防腐层缺陷类型识别是管道防腐层健康检测和安全运行的重要研究方向。利用非线性超声导波获取管道防腐层导波缺陷信号并采用局域小波分解方法获得回波信号,针对回波信号峰度系数、偏度系数、离散系数、形状系数和小波包能量系数对于不同类型防腐层缺陷的敏感程度不同,提出一种基于主成分分析的SVM防腐层缺陷分类方法,同时引入动态扰动粒子群算法,获得SVM最优的惩罚系数C和RBF核函数g值。结果表明:算法能够实现孔洞、裂纹和凹坑三类管道防腐层缺陷的有效分类,与传统SVM分类相比,准确率提高了12.9%,为管道防腐层缺陷检测提供了有效的分类方法。

Defect type identification of pipeline anticorrosion layer is an important research direction for health inspection and safe operation of pipeline anticorrosion coating.The guided wave defect signal of pipeline anticorrosion coating is obtained by nonlinear ultrasonic guided wave,and the echo signal is obtained by local wavelet decomposition method.In view of the different sensitivity of kurtosis coefficient,skewness coefficient,dispersion coefficient,shape coefficient and wavelet packet energy coefficient of echo signal to different types of anticorrosion coating defects,a SVM anticorrosion coating defect classification method based on principal component analysis is proposed,and the dynamic disturbance particle swarm optimization algorithm is introduced to obtain the optimal penalty coefficient C and RBF kernel function G value of SVM.The results show that the algorithm can effectively classify three kinds of pipeline coating defects:holes,cracks and pits.Compared with the traditional SVM classification,the accuracy is improved by 12.9%,which provides an effective classification method for pipeline coating defect detection.