基于模糊c均值聚类算法的焊缝缺陷等级磁记忆定量识别

Quantitative Metal Magnetic Memory Identification of Weld Defect Levels Based on Fuzzy c-means Clustering Algorithm

针对磁记忆检测技术在焊缝缺陷等级定量识别中模糊性和分散性的难题,提出以模糊c均值聚类算法为理论基础的焊缝缺陷等级磁记忆定量识别模型。以Q235焊接试件为试验材料,进行疲劳拉伸试验。对比X射线检测定量标准,确定磁记忆焊缝缺陷量化等级,提取不同等级的磁记忆信号征参数向量,通过训练样本的学习选取最优模糊加权指数m,建立了基于模糊c均值聚类算法的焊缝缺陷等级磁记忆定量识别模型。模型验证结果表明,预测损伤等级的准确率达到了90%,为实际工程中准确区分焊缝缺陷等级提供理论依据和新的思路。

In order to overcome the fuzziness and dispersion difficulties of metal magnetic memory （MMM） testing in weld defect level identification, a qualitative MMM identification model based on the fuzzy c-means clustering algorithm was presented. The fatigue tensile test was carried out with Q235 weld- ing specimen as the experimental material .By comparing with the quantitative standard of X-ray detection, the MMM quantification levels of weld defect were determined, and different levels of MMM signal charac- teristic parameters were extracted. The optimal fuzzy weighted exponent m was selected through learning of training samples, and quantitative MMM identification model of weld defect levels based on fuzzy c-means clustering algorithm was established. The verification results show that the accuracy of damage prediction level reached 90% ,which provides a theoretical basis and a new idea for accurate distinguishing of welddefect levels in practical engineering.