A real-time weld defect detection process allows dynamic corrective measures to overcome the defect or halt the welding process to avoid further wastage . These are essential steps to avoid damage to the welded structure, thus possibly avoiding critical safety hazards.
Different welding parameters influence the weld quality, and unbalanced samples in a dataset will affect weld defect detection. The annotation software ‘ labelimg ’ was used to annotate the image sample into classes 1, 2, and 3 for the task of defect detection in the USW defect detection dataset. 3.2.2. Deep Learning Models Used
This article proposes a lightweight deep-learning algorithm called MGNet for detecting welding defects in the current collectors. We introduce a lightweight MDM module based on multiscale channels, which utilizes deep dynamic convolutions as its basic structure to extract compelling features while reducing computational complexity.
Along with weld quality prediction, high-accuracy weld defect detection is also proposed in this work. For this task, a dataset of weld joint images showing the welding defects (the USW weld defect dataset) is set up using offline image augmentation techniques.
The focus of research in weld defect detection is to develop a non-destructive testing method for weld quality assessment based on observing the weld with an RGB camera. Deep learning techniques have been widely used in the domain of weld defect detection in recent times, but the majority of them use, for example, X-ray images.
For weld defect detection based on images, complex welding processes have a certain effect on the robustness of such algorithms. Hence, image-based techniques are developed for specific use cases and applications in welding . Manual annotation of datasets for deep learning is a laborious and time-consuming task.