基于超声检测的构件层间粘接缺陷识别方法

Recognition Method of Bonding Defects Between Component Layers Based on Ultrasonic Testing

中文关键词:  超声检测  多特征融合  主成分分析  BP神经网络

英文关键词:ultrasonic testing  multi feature fusion  principal component analysis(PCA)  BP neural network

中文摘要:

      超声无损检测广泛用于检测界面粘接缺陷，然而粘接缺陷类型的识别一直是检测的难点。因此提出了一种基于多特征融合和主成分分析提取界面粘接状况回波信号特征的方法。首先通过对缺陷信号回波进行消噪处理，提取了缺陷信号时域和时频域的特征参数，并构成联合特征向量。随后，经过主成分分析消除联合特征向量的冗余信息并降低特征向量之间的相关性，实现降维，选取累计贡献率超过95%的主成分作为粘接类型的融合特征向量。最后用BP神经网络实现缺陷类型识别分类。实验结果表明，这种方法可以有效地识别出粘接缺陷类型，识别率优于单独时频域特征提取方法，为粘接缺陷的分类识别和无损评价提供了技术参考。

英文摘要:

      Ultrasonic non destructive testing is widely used to detect interface bonding defects.However,the identification of the types of bonding defects has always been a difficult point in detection.Therefore,a method based on multi feature fusion and principal component analysis(PCA) is proposed to extract the echo signal characteristics of the interface bonding condition.Firstly,by performing noise reduction processing on the defect signal echo,the characteristic parameters of the time domain and time frequency domain of the defect signal are extracted,and a joint feature vector is formed.Then,through PCA,the redundant information of the joint feature vector is eliminated and the correlation between the feature vectors is reduced to achieve dimension reduction.The principal component with cumulative contribution rate of more than 95% is selected as the fusion feature vector of bonding type.Finally,BP neural network is used to automatically identify and classify the defect types.The experimental results show that this method can effectively identify the types of bonding defects,and the recognition rate is better than the single time frequency feature extraction method,which provides technical reference for the classification and non destructive evaluation of bonding defects.