基于稀疏化LS-SVM的漏磁缺陷三维轮廓重构

3-D Defect Profile Reconstruction from Magnetic Flux Leakage Signals Based on Sparsity LS-SVM

漏磁缺陷轮廓重构是指由检测到的漏磁信号重构缺陷轮廓及参数,是实现漏磁反演的关键。将最小二乘支持向量机(LS-SVM)应用于漏磁缺陷的三维轮廓重构中,并对LS-SVM采取了稀疏化处理,将漏磁信号磁通密度法向分量Bx作为支持向量机网络的输入,缺陷的几何参数长度、宽度、深度作为输出,由实验测量数据和三维有限元仿真计算得到的仿真数据组建样本库。建立了由缺陷的漏磁信号到缺陷三维轮廓图的映射关系,实现了缺陷三维轮廓的重构。实验结果表明:该方法具有很高的精度和很好的泛化能力,同时对噪声也有一定容忍能力。

The reconstruction of magnetic flux leakage （MFL） defect profiles means the reconstruction of defect profiles and parameters from MFL inspection signals. It is the key for the inversion of MFL inspection signals. 3-D defect reconstruction from MFL signals was studied by the least squares support vector machine （LS-SVM）, on which the pruning algorithm imposes the sparseness. In the network training, the normal component of magnetic flux density Bx was chosen as the input data of SVM nets and the defect geometric parameters： length, width and depth as the output data. The sample data library was constructed by the experimental data and simulated data obtained by 3-D finite element method. A mapping from MFL signals to 3-D profiles of defects was established, and inversion of 3-D profiles of defects from magnetic flux leakage inspection signals was achieved. Experimental results show that the LS-SVM has high precision, a good generalization ability and the capability of tolerating noise.