An end-to-end welding defect detection approach based on titanium alloy time-of-flight diffraction images

The weld defect recognition of titanium alloy is of great significance for ensuring the safety and reliability of equipment. This study proposes a method based on the enlighten faster region-based convolutional neural network (EFRCNN) to recognize titanium alloy weld defects. First, by designing defect test blocks and using probes with different frequencies, a dataset of time-of-flight diffraction (TOFD) weld defect detections is constructed. Next, to overcome the problems of high data noise and low recognition accuracy, a parallel series multi-scale feature information fusion mechanism and a channel domain attention strategy are designed, and a deep learning network model based on the faster region-based convolution neural network (Faster R-CNN) is constructed. Finally, the proposed method is verified by the TOFD test data of titanium alloy welds. The results show that the proposed method can achieve a defect type recognition accuracy of more than 92%, especially in detecting cracks or a lack of fusion.