基于机器学习的电弧行为识别与特征分析

肖典; 蒲柯伶; 褚卓楠; 方乃文; 武鹏博; 吴斌涛

doi:10.12073/j.hjxb.20230602001

基于机器学习的电弧行为识别与特征分析

1.
宁夏大学, 材料与新能源学院, 银川, 750400
2.
中国机械总院集团哈尔滨焊接研究所有限公司, 哈尔滨, 150028

基金项目: 国家重点研发计划资助项目(2021YFB3401100)；黑龙江省头雁行动计划-能源装备先进焊接技术创新团队资助项目(201916120)

详细信息

作者简介:
肖典，学士；主要研究方向增材制造；Email: x20020720d@163.com

通讯作者:
吴斌涛，博士，副教授；Email: wubintao@outlook.com

中图分类号: TG 457
计量
- 文章访问数: 248
- HTML全文浏览量: 52
- PDF下载量: 54
出版历程
- 收稿日期: 2023-06-01
- 网络出版日期: 2024-03-08

Arc behaviour recognition and characterization analysis by using machine learning

1.
Ningxia University, School of Materials and New Energy, Yinchuan, 750400, China
2.
Harbin Welding Institute Limited Company, Harbin, 150028, China

摘要

摘要:
电弧熔丝增材制造过程中电弧行为是影响零件成形精度及质量的关键因素之一，针对电弧熔丝增材制造过程中电弧无振荡、摇摆振荡以及圆周振荡3种电弧状态的监测图像，提出一种基于局部二值模式 (local binary pattern，LBP) 与GoogLeNet神经网络结合识别电弧模式的新方法. 结果表明，通过局部二值模式获取电弧形态图像中的纹理特征，然后建立GoogLeNet神经网络模型，相比于直接对原始图像进行神经网络的训练，该方法可有效识别电弧长度、宽度以及左右最大倾角随堆积层数的变化规律，从而精准判别电弧所属状态. 针对常规存在熔池、熔滴以及复杂背景等因素干扰的电弧形态图像，该方法处理后可获得更清晰的电弧边缘轮廓，更有利于将熔池、熔滴和电弧的形态边界进行划分，最终的状态识别准确率可达99.50%，为电弧熔丝增材制造过程中的电弧状态监测提供理论参考.
- 电弧状态 /
- 局部二值模式 /
- GoogLeNet神经网络 /
- 图像处理
Abstract:
In this paper, we propose a new method based on the combination of local binary pattern (LBP) and GoogLeNet neural network to identify the arc patterns in the monitoring images of three types of arc states, namely, stable arc, swinging oscillation, and circumferential oscillation, in the wire arc additive manufacturing process. The results show that obtaining the texture features in the arc pattern image via local binary pattern, and then building the GoogLeNet neural network model can effectively identify the arc length, arc width, and left and right maximum inclination with the number of stacked layers, which can be used to accurately identify the arc state compared with the direct training of neural network on the original image. For the arc morphology images in where are influenced by droplets, complex background and other factors, the proposed method can achieve a clear arc edge, whichbenefics boundary identification of melt pool, droplets and arc morphology. The extract accuracy of arc state is up to 99.50%. The research outcomes will provide a theoretical reference for monitoring arc state during wire arc additive manufacturing process.
- arc state /
- local binary pattern /
- GoogLeNet neural network /
- image processing

HTML全文

图 1 磁弧振荡焊接系统

Figure 1. Magnetic arc oscillation welding system

下载: 全尺寸图片幻灯片

图 2 不同振荡的电弧图像

Figure 2. Arc image of different oscillation. (a) circumferential oscillation; (b) side-to-side oscillation; (c) no oscillation

下载: 全尺寸图片幻灯片

图 3 电弧参数测量

Figure 3. Measurement of arc parameter

下载: 全尺寸图片幻灯片

图 4 电弧参数趋势折线图

Figure 4. Arc parameter trend line chart. (a) length and width of the circumferential oscillation; (b) angle of the circumferential oscillation; (c) length and width of the no oscillation; (d) length and width of the side-to-side oscillation; (e) angle of the circumferential oscillation; (f) comparison of length; (g) comparison of width; (h) right inclination contrast; (i) left inclination contrast

下载: 全尺寸图片幻灯片

图 5 Inception降维结构

Figure 5. Dimensionality reduction of Inception structure

下载: 全尺寸图片幻灯片

图 6 GoogLeNet网络结构

Figure 6. Network structure of GoogLeNet

下载: 全尺寸图片幻灯片

图 7 原始图像训练的结果

Figure 7. Result of the original image training

下载: 全尺寸图片幻灯片

图 8 局部二值模式

Figure 8. Local binary pattern. (a) example of LBP; (b) arc images before and after LBP processing

下载: 全尺寸图片幻灯片

图 9 LBP处理后训练的结果

Figure 9. Results of training after LBP processing

下载: 全尺寸图片幻灯片

表 1 焊接工艺参数

Table 1 Welding process parameters

沉积电流 I/A	移动速度 v₁/(mm·s⁻¹)	送丝速度 v_f/(mm·s⁻¹)	电极和工件的间距d /mm	电极和焊丝的角度θ/(°)	GTAW焊枪的气体流量V₁/(L·min⁻¹)	移动式惰性气体的流量V₂/(L·min⁻¹)	层间停留时间t/s
120	1 000	880	3	60	15	10	120

下载: 导出CSV

参考文献(12)

[1]	Christian G, Edward W R, Jan P, et al. Application of supervised machine learning for defect detection during metallic powder bed fusion additive manufacturing using high resolution imaging[J]. Additive Manufacturing, 2018, 21: 517 − 528.
[2]	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): 17 − 21. Yang Zhichao, Zhou Qiang, Hu Kan, et al. Welding defect recognition technology based on convolutional neural network and aapplication[J]. Journal of Wuhan University of Technology(Information & Management Engineering), 2019, 41(1): 17 − 21.
[3]	蒋伟琪, 黄海鸿, 刘赟, 等. 基于组合神经网络的钨极氩弧焊环境负荷预测[J]. 焊接学报, 2022, 43(10): 77 − 85. doi: 10.12073/j.hjxb.20211104002 Jiang Weiqi, Huang Haihong, Liu Yun, et al. Prediction for emission of environmental burden in GTAW based on combined neural network[J]. Transactions of the China Welding Institution, 2022, 43(10): 77 − 85. doi: 10.12073/j.hjxb.20211104002
[4]	汤泉, 石志新, 毛志伟. 基于多阈值与神经网络的旋转电弧图像飞溅分析[J]. 焊接学报, 2022, 43(12): 41 − 46. doi: 10.12073/j.hjxb.20211219001 Tang Quan, Shi Zhixin, Mao Zhiwei. Spatter analysis of rotating arc image based on multi threshold and neural network[J]. Transactions of the China Welding Institution, 2022, 43(12): 41 − 46. doi: 10.12073/j.hjxb.20211219001
[5]	孙家豪, 张超勇, 吴剑钊, 等. 基于神经网络的316L不锈钢激光焊焊缝形貌预测[J]. 焊接学报, 2021, 42(12): 40 − 47. doi: 10.12073/j.hjxb.20210526003 Sun Jiahao, Zhang Chaoyong, Wu Jianzhao, et al. Prediction of weld profile of 316L stainless steel based ongeneralized regression neural network[J]. Transactions of the China Welding Institution, 2021, 42(12): 40 − 47. doi: 10.12073/j.hjxb.20210526003
[6]	Wang C, Tan X P, Tor S B, et al. Machine learning in additive manufacturing: State-of-the-art and perspectives[J]. Additive Manufacturing, 2020, 36: 101538.
[7]	Qi Xinbo, Chen Guofeng, Li Yong, et al. Applying neural-network-based machine learning to additive manufacturing: current applications, challenges, and future perspectives[J]. Engineering, 2019, 5(4): 721 − 729. doi: 10.1016/j.eng.2019.04.012
[8]	高昶霖, 宋燕利, 左洪洲, 等. 基于动态权重的自适应PSO-BP神经网络焊接缺陷成因诊断[J]. 焊接学报, 2022, 43(1): 98 − 106. doi: 10.12073/j.hjxb.20210515001 Gao Changlin, Song Yanli, Zuo Hongzhou, et al. Cause diagnosis of welding defects based on adaptive PSO-BP neural network with dynamic weighting[J]. Transactions of the China Welding Institution, 2022, 43(1): 98 − 106. doi: 10.12073/j.hjxb.20210515001
[9]	孙伟, 赵玉普. 增强旋转不变LBP算法及其在图像检索中的应用[J]. 计算机科学, 2019, 46(7): 263 − 267. doi: 10.11896/j.issn.1002-137X.2019.07.040 Sun Wei, Zhao Yupu. Enhanced rotation invariant LBP algorithm and its application in image retrieval[J]. Computer Science, 2019, 46(7): 263 − 267. doi: 10.11896/j.issn.1002-137X.2019.07.040
[10]	Szegedy C, Liu W, Jia Y, et al. Going deeper with convolutions[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. 2015: 1 − 9.
[11]	何冠宇, 顾玉芬, 朱明, 等. 单道叠层堆焊中电弧及熔滴过渡行为检测和分析[J]. 热加工工艺, 2016, 45(19): 176 − 178. He Guanyu, Gu Yufen, Zhu Ming, et al. Detection and analysis of arc shape and droplet transfer behavior in laminated single track bead welding[J]. Hot Working Technology, 2016, 45(19): 176 − 178.
[12]	郭玲玲, 苏冬娜, 胡绍彬. 结合LBP圆形算子的CNN面部表情识别研究[J]. 微型电脑应用, 2023, 39(2): 1 − 4. doi: 10.3969/j.issn.1007-757X.2023.02.002 Guo Lingling, Su Donna, Hu Shaobin. Research on CNN facial expression recognition combined with LBP circular operator[J]. Microcomputer Applications, 2023, 39(2): 1 − 4. doi: 10.3969/j.issn.1007-757X.2023.02.002

施引文献(3)

期刊类型引用(2)

1.	陈先锋，覃世军，汪青峰，曹金，卯鑫，刘鹏，冯星. 偏滤器穿管型结构材料的高温可靠性试验分析. 核聚变与等离子体物理. 2023(03): 283-289 . 百度学术
2.	雷玉成，张伟伟，刘丹，李鑫. 氦离子辐照对316L钢焊缝微观结构及性能影响. 焊接学报. 2021(08): 48-53+99-100 . 本站查看