Detection method for automatically identifying and detecting weld defects

Abstract

The invention relates to a detection method for automatically identifying and detecting weld defects. Weld defects are detected based on a deep learning algorithm of a Faster R-CNN network. The methodcomprises the following steps of firstly, preprocessing collected weld defect images, randomely extracting 20% of the weld defect images for testing, manually storing residual images and adding labels for training; inputting a training sample into a Faster R-CNN neural network for training, then testing the obtained model, and finally achieving a detection effect. The Faster R-CNN network is adopted to perform feature extraction on the weld defect images, so that low efficiency and uncertainty of traditional manual feature extraction are avoided, and meanwhile, the detection process also hasrelatively high robustness. Deep learning technology and image processing technology are combined. A deep learning method is used for detecting the weld defects. When learning rate is initially set tobe 0.001, model classification accuracy reaches 99.3%, detection precision is obviously improved and detection speed is increased.

Description

technical field [0001] The invention belongs to the field of welding detection, and in particular relates to a detection method for automatically identifying and detecting weld defects. Background technique [0002] Welding is the most important processing method in today's manufacturing industry. With the rapid increase in the number of welded products, the complexity of the product structure is getting higher and higher, and how to inspect the weld quality after welding is becoming more and more stringent. The weld defects after welding such as slag inclusions, pores, cracks Weld defects such as , incomplete penetration, and incomplete fusion all have an extremely important impact on subsequent production efficiency. Therefore, it is necessary to identify and detect defects on all welds of welding products to improve product quality. [0003] The traditional seam defect detection method requires professionals to judge and analyze the defects of the film with the help of ...

AI Technical Summary

Problems solved by technology

However, the time overhead of training and testing is large, and the detection speed is slow, which cannot meet the real-time industrial requirements.

Images