Quantitative evaluation for the position of a defect based on Lamb wave by using Faster R-CNN

Abstract

A deep convolutional neural network method based on Faster R-CNN was proposed to achieve a quantitative evaluation of the position of a defect within a plate-like structure. In this study, a defective plate-like structure was detected using a lamb wave. To reduce the complexity of the Lamb wave response, only the A0 mode was excited and received, and the defect image was obtained by extracting the defect echo signal combined with the total focusing method (TFM). Then the defect non-destructive evaluation within a plate-like structure is transformed into an image target detection problem by using machine learning, and the images of the defect in a plate-like structure are augmented with Open Source Computer Vision Library (OpenCV) to complete the construction of the dataset and solve the problem of insufficient data samples. Subsequently, sufficient data samples were trained by the Faster R-CNN model for quantitative evaluation of the position of the defect. The results show that the Faster R-CNN model can achieve a more precise location and size message of the defect within a plate-like structure. The proposed method has the potential to broaden the application prospects of defect detection.