Prediction of GTAW penetration state based on enhanced knowledge of dynamic morphology molten pool
-
摘要:
为了实现对焊接熔透状态的在线监测，保障焊缝质量并推动机器人智能化技术发展，提出了一种熔池动态形变知识增强的钨极惰性气体保护焊(gas tungsten arc welding, GTAW)熔透状态预测方法. 采用高速高动态范围工业相机获取熔池图像，利用DeepLabv3 + 语义分割模型进行熔池动态分割以获取精准的熔池区域，在此基础上进行多帧熔池轮廓图像融合，以描述熔池在焊接过程中的动态形变. 将融合后熔池轮廓图像和原始熔池图像合成输入CNN，学习同一位置熔池轮廓像素变化，指导CNN预测熔透状态. 结果表明，基于熔池动态形变知识增强的CNN能够准确识别部分熔透、适度熔透和过度熔透三种典型焊缝状态，分类准确率可达到97.1%，单帧预测时间0.86 ms. 与未融合熔池动态形变特征专家知识的深度学习方法相比，该方法在少样本数据情况下表现出更高的鲁棒性和准确率.
Abstract:To achieve online monitoring of welding penetration status, ensure weld quality, and promote the development of robotic intelligent technology, a knowledge-enhanced prediction method for Gas Tungsten Arc Welding (GTAW) penetration status based on molten pool dynamic deformation is proposed. High-speed, high-dynamic-range industrial cameras are employed to capture molten pool images, and the DeepLabv3 + semantic segmentation model is utilized for dynamic segmentation of the molten pool to obtain precise molten pool regions. On this basis, multi-frame molten pool contour image fusion is performed to describe the dynamic deformation of the molten pool during the welding process. The fused molten pool contour images and original molten pool images are combined and input into a CNN to learn pixel-level changes in the molten pool contours at the same location, guiding the CNN to predict penetration status. Experimental results demonstrate that the CNN enhanced with molten pool dynamic deformation knowledge can accurately identify three typical weld states: partial penetration, adequate penetration, and excessive penetration, achieving a classification accuracy of 97.1% with a single-frame prediction time of 0.86 ms. Compared to deep learning methods without the integration of expert knowledge on molten pool dynamic deformation features, this method exhibits higher robustness and accuracy in scenarios with limited sample data.
-
-
![]()
图 3 熔透状态示意图
Figure 3. Schematic of penetration states. (a) partial penetration; (b) full penetration; (c) excessive penetration
![]()
图 6 知识增强的CNN熔透状态预测架构
Figure 6. Knowledge-enhanced CNN Penetration prediction architecture
![]()
图 7 不同网络熔池分割比较
Figure 7. Comparison results of different model molten pool segmentation
![]()
图 8 不同熔透状态下熔池轮廓融合
Figure 8. Molten pool counter fusion in different penetration states. (a) partial penetration;(b) full penetration; (c) excessive penetration
![]()
图 11 熔透状态预测结果和真实标签(0-部分熔透，1-适度熔透，2-过度熔透)
Figure 11. Penetration prediction results and truth label(0- partial penetration, 1- full penetration, 2- excessive penetration)
![]()
图 12 不同输入下t-SNE可视化
Figure 12. Visualization of t-SNE with different inputs. (a) single-frame raw molten pool image; (b) multi-frame raw molten pool image fusion; (c) knowledge enhancement for single-frame molten pool contour image; (d) knowledge enhancement for multi-frame molten pool contour image fusion
表 1 焊接试验参数
Table 1 Welding experiment paraments
焊件尺寸
(L × W × H)/mm氦气流量
q/(L·min−1)焊接电流
I/A焊接速度
v/(mm·s−1)钨极直径
d/mm300 × 150 × 5 12 155 ~ 175 270 5.0 
下载: 导出CSV
表 2 CNN架构
Table 2 Architecture of the proposed CNN
模型 输入大小 卷积核1 卷积核2 卷积核3 卷积核4 全连接层 CNN-1 360*360 16 32 64 128 [23*23*128,3] CNN-2 224*224 16 32 64 128 [14*14*128,3] CNN-3 224*224 32 64 128 256 [14*14*256,3] CNN-4 128*128 32 64 128 256 [8*8*256,3]
下载: 导出CSV
表 3 不同分割网络性能对比
Table 3 Quantifying the performance of different segmented networks
方法 平均交并比 平均像素精度 检测速度
v/(帧·s−1)FNC 0.722 0.791 22 SegNet 0.754 0.786 25 U-Net 0.783 0.824 24 DeepLabv3 + 0.812 0.836 19
下载: 导出CSV
表 4 数据集分布(个)
Table 4 Distribution of dataset
数据集 部分熔透 适度熔透 过度熔透 总计 训练集 301 623 350 1 274 测试集 129 268 150 547 总计 430 891 500 1 821
下载: 导出CSV
表 5 CNN测试结果
Table 5 CNN test results
输入图像 准确率
A(%)精度
P(%)召回率
R(%)F1值
F1(%)单帧原始熔池图像 83.5 81.4 83.3 83.8 多帧原始熔池图像融合 90.4 91.5 90.4 90.5 单帧熔池轮廓图像知识增强 93.2 93.4 93.2 93.2 多帧熔池轮廓图像融合知识增强 97.1 98.5 97.4 97.3
下载: 导出CSV
-
[1] 周政, 王国庆, 宋建岭, 等. 2219铝合金不同气氛下TIG焊焊接接头组织性能[J]. 焊接学报, 2018, 39(7): 47 − 50. doi: 10.12073/j.hjxb.2018390173 Zhou Zheng, Wang Guoqing, Song Jianlin, et al. Microstructure and mechanical properties of 2219aluminum alloys TIG welding welded joints in differentshielding gases[J]. Transactions of the China Welding Institution, 2018, 39(7): 47 − 50. doi: 10.12073/j.hjxb.2018390173
[2] 张志芬, 杨哲, 任文静, 等. 电弧光谱深度挖掘下的铝合金焊接过程状态检测[J]. 焊接学报, 2019, 40(1): 19 − 25. doi: 10.12073/j.hjxb.2019400005 Zhang Zhifen, Yang Zhe, Ren Wenjing, et al. Condition detection in Al alloy welding process based on deep mining of arc spectrum[J]. Transactions of the China Welding Institution, 2019, 40(1): 19 − 25. doi: 10.12073/j.hjxb.2019400005
[3] 岳建锋, 龙新宇, 黄云龙, 等. 基于电弧声信号的窄间隙脉冲熔化极气体保护焊侧壁熔合状态在线识别研究[J]. 中国机械工工程, 1-8. Yue Jianfeng, Long Xinyu, Huang Yunlong, et al. On-line identification of narrow gap P-GMAW sidewall fusion state based on arc acoustic signa[J]. China Mechanical Engineering, 1–8.
[4] Peng G, Chang B, Wang G, et al. Vision sensing and feedback control of weld penetration in helium arc welding process[J]. Journal of Manufacturing Processes, 2021, 72: 168 − 178.
[5] 肖宏, 宋建岭, 刘宪力, 等. 基于形态学算法的2219铝合金钨极氦弧焊熔池图像特征提取[J]. 宇航材料工艺, 2019, 49(1): 78 − 81. doi: 10.12044/j.issn.1007-2330.2019.01.015 Xiao Hong, Song Jianling, Chang Baohua, et al. Image feature extraction of helium gas tungsten arc welding pool of 2219 aluminum alloy based on morphological algorithm[J]. Aerospace Materials & Technology, 2019, 49(1): 78 − 81. doi: 10.12044/j.issn.1007-2330.2019.01.015
[6] Hong Y, Yang M, Chang B, et al. Filter-PCA-based process monitoring and defect identification during climbing helium arc welding process using DE-SVM[J]. IEEE Transactions on Industrial Electronics, 2023, 70(7): 7353 − 7362.
[7] Chen C, Lv N, Chen S. Welding penetration monitoring for pulsed GTAW using visual sensor based on AAM and random forests[J]. Journal of Manufacturing Processes, 2021, 63: 152 − 162.
[8] 洪宇翔, 杨明轩, 都东, 等. 铝合金爬坡TIG焊熔池失稳状态的视觉检测[J]. 焊接学报, 2021, 42(10): 8 − 13. Hong Yuxiang, Yang Mingxuan, Du Dong, et al. Unstable state vision detection of molten pool during aluminum alloy climbing-TIG welding[J]. Transactions of the China Welding Institution, 2021, 42(10): 8 − 13.
[9] 马晓锋, 夏攀, 刘海生, 等. 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. doi: 10.3901/JME.2023.12.272 Ma Xiaofeng, Xia Pan, Liu Haisheng, et al. Deep learning detection method of all-position welding pool[J]. Journal of Mechanical Engineering, 2023, 59(12): 272 − 283. doi: 10.3901/JME.2023.12.272
[10] Cai W, Jiang P, Shu L, et al. Real-time identification of molten pool and keyhole using a deep learning-based semantic segmentation approach in penetration status monitoring[J]. Journal of Manufacturing Processes, 2022, 76: 695 − 707.
[11] 李春凯, 王嘉昕, 石玗, 等. GTAW熔池激光条纹动态行为与熔透预测方法研究[J]. 机械工程学报: 1-9. Li Chunkai, Wang Jiaxin, Shi Yu, et al. Study on the dynamic behavior of GTAW melt pool laser streak and penetration prediction method[J]. Journal of Mechanical Engineering, 1–9.
[12] Wang Q, Jiao W, Wang P, et al. A tutorial on deep learning-based data analytics in manufacturing through a welding case study[J]. Journal of Manufacturing Processes, 2021, 63: 2 − 13.
[13] Cheng Y, Wang Q, Jiao W, et al. Detecting dynamic development of weld pool using machine learning from innovative composite images for adaptive welding[J]. Journal of Manufacturing Processes, 2020, 56: 908 − 915.
[14] 刘秀航, 叶广文, 黄宇辉, 等. 激光-MIG复合焊根部驼峰缺陷预测[J]. 焊接学报, 2022, 43(12): 47 − 52,99. doi: 10.12073/j.hjxb.20211216003 Liu Xiuhang, Ye Guangwen, Huang Yuhui, et al. Root hump defect prediction for laser-MIG hybrid welding[J]. Transactions of the China Welding Institution, 2022, 43(12): 47 − 52,99. doi: 10.12073/j.hjxb.20211216003
[15] Jiao W, Wang Q, Cheng Y, et al. Prediction of weld penetration using dynamic weld pool arc images[J]. Weld J, 2020, 99(11): 295S − 302S.
[16] Gao P, Wu Z, Wang Y, et al. Method for monitoring and controlling penetration of complex groove welding based on online multi-modal data[J]. Journal of Intelligent Manufacturing, 2023.
[17] Wu D, Hu M, Huang Y, et al. In situ monitoring and penetration prediction of plasma arc welding based on welder intelligence-enhanced deep random forest fusion[J]. Journal of Manufacturing Processes, 2021, 66: 153 − 165.
[18] Liu T, Bao J, Zheng H, et al. Learning semantic-specific visual representation for laser welding penetration status recognition[J]. Science China Technological Sciences, 2022, 65(2): 347 − 360. doi: 10.1007/s11431-021-1848-7
计量
- 文章访问数: 11
- HTML全文浏览量: 0
- PDF下载量: 3
