基于LS-SVM的超声导波管道缺陷二维重构

针对当前超声导波检测中的缺陷成像技术难点,提出了基于支持向量机的缺陷轮廓重构方法.通过实验和有限元仿真相结合的方式,获得不同大小缺陷的检测信号.采用最小二乘网络学习算法,选取缺陷回波数据作为支持向量机的输入,缺陷轮廓数据作为输出.建立了缺陷回波到缺陷二维轮廓的非线性映射,实现了缺陷轴向宽度和径向深度的二维轮廓重构,并与径向基神经网络重构效果进行了对比.实验结果表明,该方法速度快、精度高、泛化能力好,是管道超声导波定量化、可视化检测的一种可行方法.

2-D reconstruction of the pipeline defects by means of ultrasonic guided wave based on LS-SVM

An ultrasonic guided wave pipeline defect 2-D reconstruction method based on least squares support vector machine(LS-SVM) is presented for solving the difficulties existing in ultrasonic guided wave defect detection.The echo signals of the pipeline defects of different sizes were obtained by combining experiments with finite element analysis techniques.The nonlinear mapping from the echo signal of defect to the 2-D outline of its axial width and redial depth is established by using the echo signals as the input of the SVM net and the defect outline data as its output and using least squares as the learning algorithm of the network.Finally,the reconstruction effect of the LS-SVM network was compared with that of the RBF neural network,and it is shown that the former has higher speed and precision and better generalization ability.Therefore,the ultrasonic guided wave pipeline defect 2-D reconstruction method based on(LS-SVM) is effective to the quantitative and visual detection of pipeline defect.