基于多尺度小波支持向量机的脉冲漏磁缺陷三维轮廓重构

针对目前使用的支持向量机(Support vector machine,SVM)核函数在缺陷轮廓重构问题中不能逼近任意目标函数的问题,将小波理论与支持向量机核方法进行结合形成小波支持向量机。同时,根据多分辨率逼近思想引入多尺度小波支持向量机回归模型并将其运用到脉冲漏磁缺陷的三维轮廓重构中。实验中,将缺陷漏磁信号水平分量Bx作为多尺度小波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,通过对样本的训练建立了由缺陷的漏磁信号到缺陷三维轮廓图的映射关系,实现了缺陷的三维轮廓重构。实验结果表明该方法具有小波良好的抗噪能力、多尺度逼近方法较高的精度以及SVM很好的泛化能力。

3-D Defect Profile Reconstruction from PMFL Signals Based on Multi-scale wavelet SVM

The wavelet and the support vector machine(SVM) are combined to form wavelet SVM(WSVM) based on the idea of multi-resolution approximation.Then,the method is introduced into a 3-D defect reconstruction.In the experiments,the horizontal component of magnetic flux density Bxis chosen as the input of WSVM nets,the defect geometric parameters,including length,width and depth are output.A mapping of pulsed magnetic flux leakage(PMFL) response signals onto 3-D profiles of defects is established,and the inversion of 3-D profiles of defects from magnetic flux leakage inspection signals is achieved.Experimental results show that the proposed method can combine the advantages of SVM and wavelet and has a high precision,a good generalization and tolerance of noise.