基于機(jī)器視覺的焊接偏弧檢測方法研究

打開文本圖片集

1圖分類號：TG40 文獻(xiàn)標(biāo)志碼：A文章編號：1006-0316（2025）06-0061-07

doi ： 10.3969/j.issn.1006-0316.2025.06.009

Welding Arc Deflection Detection Method Based on Machine Vision

YI Taixun’，WUHaifeng'，WANG Kun1 ，PEI Weiping’，ZHU Gaoyi'，WANG Jie2 （1.Dongfang Turbine Co.，Ltd.， Dongfang Electric Corporation，Deyang 618ooo， China;

2. College ofMechanical Engineering， Sichuan University， Chengdu 610o65， China）

Abstract ∵ Welding arc deflection not only affects the strength of welded joints，but also may lead to welding defects such as porosity and slag inclusion， which willsignificantly reduce the durability and safety of welded structures.In this paper，a welding arc deflection detection method based on machine vision is proposed. Specifically，the method is divided into two main steps.First，the DeepLabV3+ network is used to process the image during the welding proces，accurately segment the area of the welding pool and the welding material，and extracttheir characteristics.Second，on the basis of feature extraction， the relative angle between the welding pool and the welding material is calculated to further judge whether the welding pool deviates from the predetermined welding path.In order to verifythe effectivenessof this method，a large number of experiments are carried out onthe self-built data set.The experimental results show that the proposed method can achieve 82% mIoU in extracting welding pool and welding material features， which reflects the accuracy of image segmentation. The Precision of 91% is achieved in judging whether the weld pool deviates.

Key words ∵ machine vision ; semantic segmentation ; welding arc deflection ; DeepLabV3+

電弧焊接廣泛應(yīng)用于建筑、造船、汽車制造、管道鋪設(shè)和機(jī)械制造等多個(gè)領(lǐng)域。（剩余6849字）