全位置焊接熔池的深度学习检测方法

doi:10.3901/JME.2023.12.272

机械工程学报 ›› 2023, Vol. 59 ›› Issue (12): 272-283.doi: 10.3901/JME.2023.12.272

扫码分享

全位置焊接熔池的深度学习检测方法

马晓锋^1,2, 夏攀^1,2, 刘海生^1,2, 史铁林³, 王中任^1,2

1. 湖北文理学院机械工程学院襄阳 441053;
2. 智能制造与机器视觉襄阳市重点实验室襄阳 441053;
3. 华中科技大学机械科学与工程学院武汉 430074

收稿日期:2022-07-30 修回日期:2023-04-09 出版日期:2023-06-20 发布日期:2023-08-15
通讯作者: 王中任(通信作者),男,1974年出生,博士,教授,硕士研究生导师。主要研究方向为智能制造、机器视觉与智能焊接。E-mail:wzrvision@hbuas.edu.cn
作者简介:马晓锋,男,1997年出生。主要研究方向为机器视觉与智能焊接。E-mail:mxf15571177591@126.com
基金资助:
襄阳市重点科技计划(2020ABH002033)和国家自然科学基金(11902112)资助项目。

Deep Learning Detection Method of All-position Welding Pool

MA Xiaofeng^1,2, XIA Pan^1,2, LIU Haisheng^1,2, SHI Tielin³, WANG Zhongren^1,2

1. School of Mechanical Engineering, Hubei University of Arts and Sciences, Xiangyang 441053;
2. Xiangyang Key Laboratory of Intelligent Manufacturing and Machine Vision, Xiangyang 441053;
3. School of Mechanical Science and Engineering, Huazhong University of Science and Technology, Wuahan 430074

Received:2022-07-30 Revised:2023-04-09 Online:2023-06-20 Published:2023-08-15

摘要/Abstract

摘要： 在熔化极活性气体保护焊(Metal active gas arc welding，MAG)焊接过程中，传统机器视觉熔池检测方法难以适应剧烈变化的强弧光和飞溅等噪声的干扰。为此，提出一种结合GAN与改进PSPNet的熔池深度学习检测方法，实现对熔池的动态跟踪与精准检测。首先，通过生成对抗网络(Generative adversarial networks，GAN)中的解码器、编码器及小步数的反卷积模块，对焊接熔池图像进行增强，解决因强弧光和飞溅等噪声导致的图像模糊。然后，对不同位置的熔池形态进行力学现象分析，得出不同位置的熔池形成规律。最后，结合GAN网络与全位置熔池变化规律，对PSPNet网络进行改进，利用多层金字塔池化模块进行信息提取、空洞卷积大视野的获取全局信息、以及引入的Swish激活函数，对熔池进行动态分割。试验结果表明，该方法平均像素准确率和平均交并比等指标提高至88.03%和85.45%，网络模型训练时间比改进前提升19.05%，能够实现熔池动态跟踪，对机器人智能焊接有重要意义。

关键词: 熔化极活性气体保护焊, GAN, 改进PSPNet, 熔池动态分割

Abstract: In the process of MAG Welding, the traditional Welding pool detection method is difficult to adapt to the intense Arc light due to the serious interference of arc light and splash light. A fusion pool de-noising detection method combining GAN and improved PSPNet is proposed to realize dynamic tracking and accurate detection of the fusion pool. Firstly, the image of welding pool is denoised by generating decoder, encoder and small step deconvolution module in adversity-network (GAN), and the image blur caused by shielding gas and arc light is solved. Subsequently, the formation regularity of the molten pool at different positions is obtained by analyzing the mechanical phenomena of the molten pool at different positions. Finally, combined with GAN network denoising and all-position molten pool change law, PSPNet network is improved. Multi-layer pyramid pooling module is used to extract information, global information is obtained by cavity convolution large field, and Swish activation function is introduced to dynamically segment molten pool. Experimental results show that the average pixel accuracy and average intersection ratio of the method are improved to 88.03% and 85.45%, and the training time of the network model is improved by 19.05% compared with that before the improvement. Dynamic tracking of weld pool is realized which is of great significance to of robotic intelligent welding.

Key words: metal active gas arc welding, GAN, improvement of PSPNet, weld pool dynamic segmentation

中图分类号:

TP39

马晓锋, 夏攀, 刘海生, 史铁林, 王中任. 全位置焊接熔池的深度学习检测方法[J]. 机械工程学报, 2023, 59(12): 272-283.

MA Xiaofeng, XIA Pan, LIU Haisheng, SHI Tielin, WANG Zhongren. Deep Learning Detection Method of All-position Welding Pool[J]. Journal of Mechanical Engineering, 2023, 59(12): 272-283.

参考文献

[1] 陈华斌,陈善本. 复杂场景下的焊接智能制造中的信息感知与控制方法[J]. 金属学报,2022,58(4):541-550. CHEN Huabin,CHEN Shanben. Information perception and control method for intelligent welding manufacturing in complex Scene[J]. Acta Metallurgica Sinica,2022,58(4):541-550.
[2] 冯静,薛龙,黄军芬,等. 高压环境下湿度对低碳钢熔化极气体保护焊焊接质量的影响[J]. 上海交通大学学报,2016,50(10):1631-1634. FENG Jing,XUE Long,HUANG Junfen,et al. Effect of humidity on welding quality of low carbon steel under high pressure[J]. Journal of Shanghai Jiaotong University,2016,50(10):1631-1634.
[3] 李海超,刘景风,谢吉兵,等. 基于卷积神经网络的GTAW熔透预测[J]. 机械工程学报,2019,55(17):22-28. LI Haichao,LIU Jingfeng,XIE Jibing,et al. Prediction of GTAW penetration based on convolutional neural network[J]. Journal of Mechanical Engineering,2019,55(17):22-28.
[4] LI R,Gao H. Denoising and feature extraction of weld seam profiles by stacked denoising autoencoder[J]. Welding in the World,2021,65(9):1725-1733.
[5] 张楚昊,赵壮,陆骏,等. 基于边缘导向算子模板匹配的熔池轮廓提取方法[J]. 焊接学报,2022,43(2):67-74,118. ZHANG Chuhao,ZHAO Zhuang,LU Jun,et al. Welding pool contour extraction method based on edge guide operator template matching[J]. Journal of Welding,2022,43(2):67-74,118.
[6] 刘天元,鲍劲松,汪俊亮,等. 受限解空间下粗粒度正则化的熔池边缘自适应检测方法[J]. 焊接学报,2020,41(12):49-54,100. LIU Tianyuan,BAO Jinsong,WANG Junliang,et al. A coarse-grained regularized adaptive detection method for weld pool edge under constrained solution space[J]. Journal of Welding,2020,41(12):49-54,100.
[7] 王奉涛,杨守华,吕秉华,等. 金属增材制造过程熔池动态监测研究综述[J/OL]. 计算机集成制造系统:1-23[2021-11-03]. https://kns.cnki.net/kcms2/article/abstract?v=3uoqIhG8C45S0n9fL2suRadTyEVl2pW9UrhTDCdPD65zPhUaeS——apF3-t-awBJCPUDZbKzzn2naYwM9Q8vHhTUreKDWcitV&uniplatform=NZKPT. WANG Fengtao,YANG Shouhua,LÜ Binghua,et al. Review on dynamic monitoring of molten pool in metal additive manufacturing process[J/OL]. Computer Integrated Manufacturing Systems:1-23[2021-11-03]. https://kns.cnki.net/kcms2/article/abstract?v=3uoqIhG8C45S0n9fL2suRadTyEVl2pW9UrhTDCdPD65zPhUaeS——apF3-t-awBJCPUDZbKzzn2naYwM9Q8vHhTUreKDWcitV&uniplatform=NZKPT.
[8] 廉艳平,王潘丁,高杰,等. 金属增材制造若干关键力学问题研究进展[J]. 力学进展,2021,51(3):648-701. LIAN Yanping,WANG Panding,GAO Jie,et al. Research progress on some key mechanical problems in metal additive manufacturing[J]. Advances in Mechanics,2021,51(3):648-701.
[9] 陈波,孟正,马程远,等. 扫描振镜激光TC4钛合金焊接性能及熔池流动行为[J]. 航空学报,2022,43(4):438-450. CHEN Bo,MENG Zheng,MA Chengyuan,et al. Welding performance and pool flow behavior of titanium alloy by scanning galvanometer laser TC4[J]. Aeronautical journal,202,43(4):438-450.
[10] 王鹏飞,杨坤,陈明智,等. 倾斜基板激光熔覆GH3536熔池形貌的仿真与实验研究[J]. 中国激光,2021,48(10):210-221. WANG Pengfei,YANG Kun,CHEN Mingzhi,et al. Simulation and experimental study on molten pool morphology of GH3536 by laser cladding on inclined substrate[J]. Chinese Journal of Lasers,2021,48(10):210-221.
[11] 谢非,朱腾飞,杨继全,等. 基于边缘夹角附加损失函数的熔池形貌检测方法[J]. 焊接学报,2021,42(7):82-90,103-104. XIE Fei,ZHU Tengfei,YANG Jiquan,et al. The weld pool morphology detection method based on edge angle additional loss function[J]. Transactions of the China Welding Institution,2021,42(7):82-90,103-104.
[12] YAN L,LIU D,XIANG Q,et al. PSP net-based automatic segmentation network model for prostate magnetic resonance imaging[J]. Computer Methods and Programs in Biomedicine,2021,207:106211.
[13] 武花,张新长,孙颖,等. 融合多特征改进型PSPNet模型应用于复杂场景下的建筑物提取[J]. 测绘通报,2021(6):21-27. WU Hua,ZHANG Xinchang,SUN Ying,et al. Integrated multi-feature improved PSPNet model for building extraction in complex scenes[J]. Bulletin of Surveying and Mapping,2021(6):21-27.
[14] 肖进胜,申梦瑶,雷俊锋,等. 基于生成对抗网络的雾霾场景图像转换算法[J]. 计算机学报,2020,43(1):165-176. XIAO Jinsheng,SHEN Mengyao,LEI Junfeng,et al. Image transformation algorithm of haze scene based on generative adversarial network[J]. Chinese Journal of Computers,2020,43(1):165-176.
[15] LIU F,Jiao L,TANG X. Task-oriented GAN for PolSAR image classification and clustering[J]. IEEE Transactions on Neural Networks and Learning Systems,2019,30(9):2707-2719.
[16] WEN P,GUAN Y,LI J,et al. A-PSPNet:A novel segmentation method of renal ultrasound image[C]//2021IEEE International Conference on Systems,Man,and Cybernetics (SMC). IEEE,2021:40-45.
[17] 张省,李山山,魏国芳,等. 面向精细化多尺度特征的遥感图像目标检测[J]. 遥感学报,2022,26(12):2616-2628. ZHANG Sheng,LI Shanshan,WEI Guofang,et al. Remote sensing image target detection for fine multi-scale features[J]. Journal of Remote Sensing,2022,26(12):2616-2628.
[18] 宋益,赵宁雨,颜畅,等. 隧道衬砌裂缝实时分割的Mobile-PSPNet方法[J]. 铁道科学与工程学报,2022,19(12):3746-3757 SONG Yi,ZHAO Ningyu,YAN Chang,et al. Mobile PSPNet method for real-time crack segmentation of tunnel lining[J]. Journal of Railway Science and Engineering,202,19(12):3746-3757.
[19] 孟琭,徐磊,郭嘉阳. 一种基于改进的MobileNetV2网络语义分割算法[J]. 电子学报,2020,48(9):1769-1776. MENG Lu,XU Lei,GUO Jiayang. An improved semantic segmentation algorithm based on MobileNetV2[J]. Acta Electronica Sinica,2020,48(9):1769-1776.
[20] 胡明洪,李佳田,姚彦吉,等. 结合多路径的高分辨率遥感影像建筑物提取SER-UNet算法[J/OL]. 测绘学报:1-14[2022-10-22]. https://kns.cnki.net/kcms2/article/abstract?v=3uoqIhG8C45S0n9fL2suRadTyEVl2pW9UrhTDCdPD66ECMZxmrdk9XP5_xaD5JA2z8dA3gOyNSzp9cvUqObrSYdiwowBHjpN&uniplatform=NZKPT. HU Minghong,LI Jiatian,YAO Yanji,et al. SER-UNet algorithm for building extraction from high resolution remote sensing image combined with multipath[J/OL]. Journal of Surveying and Mapping:1-14[2022-10-22]. https://kns.cnki.net/kcms2/article/abstract?v=3uoqIhG8C45S0n9fL2suRadTyEVl2pW9UrhTDCdPD66ECMZxmrdk9XP5_xaD5JA2z8dA3gOyNSzp9cvUqObrSYdiwowBHjpN&uniplatform=NZKPT.

全位置焊接熔池的深度学习检测方法

Deep Learning Detection Method of All-position Welding Pool

PDF

可视化

摘要/Abstract

引用本文

使用本文

参考文献

相关文章 1

编辑推荐

Metrics

本文评价