摘要
针对超声相控阵(NDT)缺陷检测,本文提出一种利用多任务深度学习网络进行A扫缺陷分类的方法。运用超声相控阵探伤仪检测物体缺陷,得到A扫缺陷波形图;将得到的A扫图像经过预处理,运用MATLAB将波形图像中曲线上的数据进行提取转为一维时序数据;利用小波变换(WCT)将得到的一维时序数据转为二维时频图像;最后,利用深度学习网络Resnet进行训练。实验结果表明,利用小波变换结合深度学习的缺陷识别方式有较高的准确率。
For ultrasonic phased array(NDT)defect detection,this paper proposes a method to classify A-scan defects using a multi-task deep learning network.The ultrasonic phased array detector is used to detect object defects and obtain A-scan defect waveform images;the obtained A-scan images are pre-processed and the data on the curve in the waveform images are extracted and converted into one-dimensional time-series data using MATLAB;the obtained one-dimensional time-series data are converted into two-dimensional time-frequency images using wavelet transform(WCT);finally,the deep learning network Resnet is used for training.The experimental results show that the defect recognition method using wavelet transform combined with deep learning has a high accuracy rate.
作者
邹宸玮
王狄飏
ZOU Chenwei;WANG Diyang(Aviation Portage College,Shanghai University of Engineering and Technology,Shanghai 201620,China)
出处
《智能计算机与应用》
2023年第5期117-121,125,共6页
Intelligent Computer and Applications
关键词
时频图像
缺陷识别
A扫
深度学习
小波变换
time-frequency images
defect identification
A-scan
deep learning
wavelet transform
