基于模糊隶属度最大似然估计的焊缝隐性缺陷磁记忆信号识别

Hidden defect metal magnetic memory identification for welded joints based on fuzzy membership and maximum likelihood estimation

针对磁记忆信号在焊缝残余应力集中、隐性损伤识别和定位上的难题,建立了模糊隶属度最大似然估计识别模型。通过疲劳载荷作用下的Q235钢板状焊缝试件磁记忆检测试验,研究了焊缝残余应力集中、隐性损伤直至宏观断裂整个过程的磁记忆信号变化规律,提取多种磁记忆特征值:峰峰值ΔHp(y)、梯度最大值K_(max)、梯度均值K_(ave)、极限状态系数最大值mmax,引入正态分布函数计算其概率,结合半梯形模糊隶属度函数,建立了焊缝隐性损伤的模糊隶属度最大似然估计识别模型,判断疑似缺陷点进而识别隐性损伤并定位缺陷。结果表明:基于多种特征参数基础上的最大似然模糊隶属度估计结果与实际缺陷位置完全一致,验证了该模型可以有效避免误判的发生,为实际工程中应用磁记忆技术进行焊缝隐性损伤识别和定位提供了新的方法。

To overcome the difficulty of Metal Magnetic Memory （MMM） technique in identification and location of the hidden defect in welded joints, a method of Maximum Likelihood Estimation （MLE） and fuzzy membership degree is put forward. The feature law of critical hidden damage is studied by fatigue experiment of Q235 steel plate with incompletely penetrated weld joint. Four feature parameters are extracted, that are peak to peak value AHp（y）, mean gradient K maximal gradient K-x, and maximal limit coefficient mmax. The normal dlstr：but：on ：unction ：s employed to calculate the probabilities of the four feature parameters. Then, MLE values are obtained to establish the MLE fuzzy membership model. This model can be used to identify the suspected defect position. Results show that the MLE fuzzy membership model can effectively locate the weld de-ect and provide a new tool of identification and location of weld hidden damage with MMM method.