基于SVM及电流牵扯效应的金属缺陷分类识别方法

发布时间：2018-04-11 20:17

本文选题：无损检测 + 缺陷分类　；参考：《四川大学学报(工程科学版)》2015年06期

【摘要】：金属在服役期间经常存在一些由应力、腐蚀和疲劳造成的缺陷,在众多的缺陷检测技术中,交流电位法作为一种无损检测技术在检测腐蚀坑和裂纹方面得到了广泛应用。用交流电位法检测不同缺陷时,由于几何形态的差异,缺陷深度的计算方法也不相同。因此需要在计算缺陷深度之前对所检测区域的缺陷类型做出识别。作者目的在于寻找一种高精度的缺陷分类识别方法。并且针对腐蚀坑和裂纹这两种最常见的金属缺陷,根据其对电流的牵扯效应不同,提出利用邻近检测区域的4个牵扯因子作为缺陷区域的特征向量建立分类模型。在大量仿真计算的基础上,分别建立坑蚀和裂纹的特征向量集,并由这些数据集训练得到基于遗传算法(GA)优化的支持向量机(SVM)分类模型。仿真测试结果中数据测试集分类精度较高,平板实验也得到了较高的识别精度。实验结果表明文中提出的缺陷分类识别方法对腐蚀坑和裂纹的分类识别具有很高的精度。
[Abstract]:There are some defects caused by stress corrosion and fatigue during the service of metals. Among many defect detection techniques AC potential method has been widely used in the detection of corrosion pits and cracks as a kind of nondestructive testing technology.When different defects are detected by alternating current potential method, the calculation method of defect depth is different because of the difference of geometry.Therefore, it is necessary to identify the type of defect in the detected area before calculating the depth of defect.The purpose of this paper is to find a high precision method for defect classification and identification.Aiming at the two most common metal defects, corrosion pits and cracks, according to their different effects on the current, a classification model is established by using the four drag factors of the adjacent detection region as the characteristic vectors of the defect region.On the basis of a large number of simulation calculations, the eigenvector sets of pit erosion and crack are established, and the SVM SVM classification model based on genetic algorithm (GA) optimization is obtained by training these data sets.The classification accuracy of the data test set is higher than that of the simulation test results, and the recognition accuracy of the plate experiment is also high.The experimental results show that the proposed method has a high accuracy for the classification and identification of corrosion pits and cracks.
【作者单位】：四川大学制造科学与工程学院;
【基金】：国家自然科学基金资助项目(61271329) 四川省科技支撑计划资助项目(2012GZ0094)
【分类号】：TG115.28

【参考文献】