智能焊接检测技术创新以及在机械制造中的应用比较

打开文本图片集

中图分类号：TQ037;TP311 文献标志码：A 文章编号：1001-5922（2025）12-0032-04

Innovation of intelligent welding detection technology and comparison of its application in mechanical manufacturing

MA Long （Shaanxi Vocational College of Industry and Commerce，Xi'an 71O119，China）

Abstract：In order to improve the accuracy of welding quality detection in mechanical manufacturing，this study proposes an inteligent welding detection method based on improved YOLOv8 algorithm，which aims to solve the problemsof low efficiency，high missed detection rateand inaccurate positioning of rotating defects intraditional welding detection.Firstly，the problem of category imbalance is solved by using a composite data enhancement strategy，which lays adata foundation for subsequent model training and detection.Secondly，on the model structure，the series combination CBS module and multi-branch pooling path of convolution，normalization and activation functionare introduced into thebackbone network to enhance the multi-scale feature fusionability.The CBAM module of mixedattentionmechanism of channeland spatial attention isembedded inthe neck network to improve the dynamic modeling abilityof the modelto the relationship between channeland space.In thepartof the detection head，the modeof combining the horizontal detection frameandthedirectional detection frame isadopted to improve thepositioningaccuracyof therotation defect.Atthesame time，inorder toverifythe performanceof the model，the research compares the detection efectof the YOLOv8 algorithm on welding defects before and after the improvementthrough experiments.The experimental results show that the average detection accuracy，recall rate and F1 score of the improved model for four typical defects of crack，hole，splash and welding line are 90.4% ， 88.5% and 89.4% respectively，which are better than those before the improvement.This study can provide a reliable technical solution for welding quality inspection in the process of mechanical manufacturing，and has certain reference value for the algorithm optimization of industrial inspection tasks.

Key words：mechanical manufacturing；intelligent welding technology;deep learning;welding defect detection; YOLOv8 algorithm

在机械制造领域，焊接技术为其核心工艺之一，焊接的质量将直接关系到产品的结构强度与使用寿命。（剩余4858字）