Research and Application of An Automatic Identification Method for Electrofusion Welding Defects
-
摘要: 聚乙烯（Polyethylene，PE）燃气管道的电熔焊接状况可通过相控阵系统拍摄的超声波图像获得，但是否存在焊接缺陷目前都是由专业人员人工识别每张图片中的特征线、电阻丝、底面回波线等与缺陷相关的信息，进而确定缺陷类别和等级，这种方法费时、费力且容易发生漏检和错检现象。本文针对PE管道电熔焊接缺陷的识别问题提出了一种基于图像处理技术的焊接缺陷自动识别方法，同时判断超声图像中存在缺陷的类别和等级。该方法由4个步骤组成：（1）通过数据增强技术扩充已有图片数量，构建数据集；（2）训练图像语义分割模型，对图像进行语义分割；（3）采用数学形态学方法去除分割结果的噪声，并通过连通域分析法得到缺陷相关信息；（4）依据焊接标准及缺陷相关信息识别缺陷类别及等级。实验结果表明，本文提出的缺陷识别方法在准确率、召回率以及运行时间方面均达到工业应用要求。Abstract: The electrofusion welding status of the polyethylene (PE) gas pipeline can be obtained through the ultrasonic pictures taken by the phased array system. However, whether there are welding defects is currently judged by professionals. And the feature line, resistance wire, and bottom in each picture are manually checked for defect-related information. Then the defect categories and levels are determined. The disadvantage of this method is that it is time-consuming and labor-intensive, and is prone to missed detections and false detections. Aiming at the identification of electrofusion welding defects in PE pipelines, this paper proposes an automatic identification method of welding defects based on image processing technology. This method judges the categories and levels of defects in ultrasound images. The method consists of four steps: 1) expanding the number of existing pictures through data enhancement technology to build a data set; 2) training the image semantic segmentation model to perform semantic segmentation on the image; 3) using mathematical morphology to remove the noise of the segmentation result, and obtaining defect-related information through connected domain analysis; 4) identifying defect categories and levels based on welding standards and defect-related information. The experimental results show that the defect recognition method proposed in this paper meets the requirements of industrial applications in terms of accuracy, recall and running time.
-
表 1 标准相关的几何参数
Table 1. Geometric parameters related to the standard
Classes Geometric parameters Meaning of parameter Feature line $f_{\rm{xcen}}\left( k \right) $ the x-coordinate of the centroid of the kth connected domain of the feature line category $f_{\rm{ycen}}\left( k \right)$ the y-coordinate of the centroid of the kth connected domain of the feature line category Resistance wire $r_{\rm{xcen}}\left( n \right) $ the x-coordinate of the centroid of the nth connected domain of the resistance wire category $r_{\rm{ycen}}\left( n \right)$ the y-coordinate of the centroid of the nth connected domain of the resistance wire category $r_{\rm{x\min}}\left( n \right)$ the minimum value of the nth connected domain of the resistance wire category in the x-axis $r_{\rm{x\max}}\left( n \right)$ the maximum value of the nth connected domain of the resistance wire category in the x-axis Bottom $b_{x\min}\left( m \right) $ the minimum value of the mth connected domain of the bottom category in the x-axis $b_{x\max}\left( m \right) $ the maximum value of the mth connected domain of the bottom category in the x-axis 
下载: 导出CSV
表 2 冷/过焊缺陷表征程度
$H_{\rm{cold/over}}$ 与缺陷类型、等级对照表Table 2. Comparison table of cold/over-welding defect characterization degree
$H_{\rm{cold/over} }$ , defect categories and levelsCharacterization degree Defect categories /level $-100\leqslant H_{\rm{cold/over}}\leqslant -40$ Over welding/Ⅲ $-40 < H_{\rm{cold/over}}\leqslant -20$ Over welding/Ⅱ $-20 < H_{\rm{cold/over}} < 0$ Over welding/Ⅰ $H_{\rm{cold/over}}=0$ Normal $0 < H_{\rm{cold/over}} < 10$ Cold welding/Ⅰ $10\leqslant H_{\rm{cold/over}} < 30$ Cold welding/Ⅱ $30\leqslant H_{\rm{cold/over}}\leqslant 100$ Cold welding/Ⅲ
下载: 导出CSV
表 3 孔洞缺陷表征程度
$H_{\rm{hole}}$ 与缺陷类型、等级对照表Table 3. Comparison table of hole defect characterization degree
$H_{\rm{hole}}$ , defect categories and levelsM Characterization degree/% Defect categories/levels 2 $0 < H_{\rm{hole}} < 5$ Single hole/Ⅰ $5\leqslant H_{\rm{hole}} < 10$ Single hole/Ⅱ $H_{\rm{hole}}\geqslant 10$ Single hole/Ⅲ $>2$ $0 < H_{\rm{hole}} < 10$ Multiple holes/Ⅰ $10\leqslant H_{\rm{hole}} < 15$ Multiple holes/Ⅱ $H_{\rm{hole}}\geqslant 15$ Multiple holes/Ⅲ
下载: 导出CSV
表 4 电阻丝连通域最大偏离量
$d_{\max}$ 与缺陷等级对照表Table 4. Comparison table between the maximum deviation
$d_{\max}$ of the connected domain of the resistance wire and defect levels$d_{\max}$ Defect levels $d_{\max} < \frac{1}{2}x'$ Ⅰ $\frac{1}{2}x'\leqslant d_{\max} < x'$ Ⅱ $d_{\max}\geqslant x'$ Ⅲ
下载: 导出CSV
表 6 多类别识别结果
Table 6. Test results of multiple categories
Defect category Recall/% Cold welding 96.7 Over welding 96.7 Misalignment 93.8 Inclusion 100 Hole 100
下载: 导出CSV
表 7 冷/过焊缺陷等级划分结果
Table 7. Results of classification of cold/over welding defects
Levels Recall(Cold welding)/% Recall(Over welding)/% Ⅰ 80.0 83.3 Ⅱ 90.0 80.0 Ⅲ 93.3 94.7
下载: 导出CSV
-
[1] XIA F Y, LI W Z, LIU Y L. Research on ultrasonic testing technology for polyethylene pipe hot-melting welded joints[J]. Applied Mechanics & Materials, 2013, 331: 170-175. [2] TAO Y J, GUO W C, MIAO C J, et al. Study on electrofusion welding pprocess based on real-time ultrasonic phased array video recording[C]//Far East NDT New Technology & Application Forum, Nanjing: IEEE 2020: 66-70. [3] DUAN F, YIN S, SONG P, et al. Automatic welding defect detection of X-ray images by using cascade AdaBoost with penalty term[J]. IEEE Access, 2019, 7: 125929-125938. doi: 10.1109/ACCESS.2019.2927258 [4] 杨志超, 周强, 胡侃, 等. 基于卷积神经网络的焊接缺陷识别技术及应用[J]. 武汉理工大学学报(信息与管理工程版), 2019, 41(1): 21-5. [5] PAN H, PANG Z, WANG Y, et al. A new image recognition and classification method combining transfer learning algorithm and MobileNet model for welding defects[J]. IEEE Access, 2020, 8: 119951-119960. doi: 10.1109/ACCESS.2020.3005450 [6] HE Y, SONG K, MENG Q, et al. An end-to-end steel surface defect detection approach via fusing multiple hierarchical features[J]. IEEE Transactions on Instrumentation and Measurement, 2020, 69(4): 1493-504. doi: 10.1109/TIM.2019.2915404 [7] CHEN F C, JAHANSHAHI R. NB-CNN: Deep learning-based crack detection using convolutional neural network and naive Bayes data fusion[J]. IEEE Transactions on Industrial Electronics, 2018, 65(99): 4392-4400. [8] 渠慧帆. 基于深度学习的焊缝缺陷自动检测研究与实现[D]. 北京: 北京邮电大学, 2019. [9] 李礼彬, 顾震, 周家乐, 等. 自适应聚乙烯管道超声信号线检测算法[J]. 华东理工大学学报(自然科学版), 2020, 47( ). [10] 黄跃鑫, 钟舜聪, 伏喜斌, 等. 基于小波分析的聚乙烯管道电熔接头冷焊缺陷识别[J]. 无损探伤, 2015, 39(1): 6-10. doi: 10.3969/j.issn.1671-4423.2015.01.002 [11] BOHLOULI R, ROSTAMI B, KEIGHOBADI J. Application of neuro-wavelet algorithm in ultrasonic-phased array nondestructive testing of polyethylene pipelines[J]. Journal of Control Science and Engineering, 2012, 2012(3): 1-9. [12] LONG J, SHELHAMER E, DARRELL T. Fully convolutional networks for semantic segmentation[J]. IEEE Transactions on Pattern Analysis & Machine Intelligence, 2015, 39(4): 640-651. [13] RONNEBERGER O, FISCHER P, BROX T. U-Net: Convolutional networks for biomedical image segmentation[C]//Medical Image Computing and Computer-Assisted Intervention. Munich: Springer, 2015: 234-241. [14] BADRINARAYANAN V, KENDALL A, CIPOLLA R. SegNet: A deep convolutional encoder-decoder architecture for image segmentation[J]. IEEE Transactions on Pattern Analysis & Machine Intelligence, 2017, 39(12): 2481-2495. [15] 吴骏逸, 谷小婧, 顾幸生. 基于可见光/红外图像的夜间道路场景语义分割[J]. 华东理工大学学报(自然科学版), 2019, 45(2): 301-309. [16] QIU Y, CHANG C S, YAN J L, et al. Semantic segmentation of intracranial hemorrhages in head CT scans[C]//International Conference on Software Engineering and Service Science. Beijing: IEEE, 2019: 1-4. [17] CINAROGLU I, BASTANLAR Y. Image based localization using semantic segmentation for autonomous driving[C]//Signal Processing and Communications Applications Conference. Sivas: IEEE, 2019: 1-4. [18] KAIMING HE X Z, SHAOQING REN, JIAN SUN. Deep residual learning for image recognition[C]//IEEE Conference on Computer Vision & Pattern Recognition. Las Vegas: IEEE, 2016: 770-778. [19] 陆锋, 张俊生, 黄刚. 焊缝X射线图像的数学形态学分析[J]. 热加工工艺, 2019, 48(9): 220-222, 227. [20] SANDI D. Mathematical morphology in image analysis[C]// International Conference on Applied Mathematics. Budva: Mechanical Engineering, 1996: 1-10. [21] BLOCH I, MAITRE H. Mathematical morphology[J]. Annals of Mathematics & Artificial Intelligence, 1994, 10(1/2): 55-84. [22] 陆宗骐, 朱煜. 使用线段表实现线段编码与种子填充快速算法[J]. 中国图象图形学报, 2009, 14(3): 499-504. doi: 10.11834/jig.20090319 [23] 上海市质量技术监督局. 燃气用聚乙烯(PE)管道焊接接头相控阵超声检测: DB31/T 1058-2017[S]. 北京: 中国标准出版社, 2017. [24] 国家市场监督管理总局, 国家标准化管理委员会. 压力管道规范 公用管道: GB/T 38942-2020[S]. 北京: 中国标准出版社, 2020. [25] 周志华. 机器学习[M]. 北京: 清华大学出版社, 2016: 29-31. -
点击查看大图
图(9) / 表(7)
计量
- 文章访问数: 68
- HTML全文浏览量: 38
- PDF下载量: 8
- 被引次数: 0
