波动工况下点焊质量在线预测及模型解释

吕天乐; 齐苗苗; 闫德俊; 黎书华; 夏裕俊; 李永兵

doi:10.12073/j.hjxb.20220702002

波动工况下点焊质量在线预测及模型解释

吕天乐^{1, 2,},
齐苗苗³,
闫德俊⁴,
黎书华⁴,
夏裕俊^1, ,,
李永兵^{1, 2,}

1.
上海交通大学, 上海市复杂薄板结构数字化制造重点实验室, 上海, 200240
2.
上海交通大学, 机械系统与振动国家重点实验室, 上海, 200240
3.
上海航天设备制造总厂有限公司, 上海, 200245
4.
中船黄埔文冲船舶有限公司, 广东省舰船先进焊接技术企业重点实验室, 广州, 510715

基金项目: 国家自然科学基金项目(52025058)；国防基础科研项目(JCKY2021203B074)；工信部高技术船舶项目(MC-201704-Z02)

详细信息

作者简介:
吕天乐，硕士研究生；主要研究方向为铝合金电阻点焊及在线监测；Email: ltl1715@sjtu.edu.cn

通讯作者:
夏裕俊，助理研究员；Email: xyjdbgt6509@sjtu.edu.cn.

中图分类号: TG 453.9
计量
- 文章访问数: 369
- HTML全文浏览量: 86
- PDF下载量: 63
出版历程
- 收稿日期: 2022-07-01
- 网络出版日期: 2022-10-11
- 刊出日期: 2022-11-24

Online prediction of resistance spot weld quality and model explanation under fluctuating conditions

LV Tianle^{1, 2,},
QI Miaomiao³,
YAN Dejun⁴,
LI Shuhua⁴,
XIA Yujun^1, ,,
LI Yongbing^{1, 2,}

1.
Shanghai Key Laboratory of Digital Manufacture of Complex Thin Plate Structures, Shanghai Jiao Tong University, Shanghai 200240
2.
State Key Laboratory of Mechanical Systems and Vibration, Shanghai Jiao Tong University, Shanghai 200240
3.
Shanghai Aerospace Equipment Manufacturing Plant Co., LTD, Shanghai 200245
4.
Guangdong Provincial Key Laboratory of Marine Advanced Welding Technology Enterprise, CSSC Huangpu Wenchong Shipbuilding Co., LTD, Guangzhou 510715

摘要

摘要: 基于电阻点焊过程的多传感信号特征，面向多种板材组合建立焊点质量在线预测模型，研究了异常工况波动对四类机器学习回归模型的影响，分析了不同模型和输入变量对含异常工况试验数据集的适应性，并采用Shapley值、t-SNE等方法对波动工况下的模型性能进行解释.结果表明，高斯过程回归模型和电阻 + 力信号具有最佳的熔核直径预测性能，焊接电流、热输入能量和电极位移峰值特征对于波动工况具有良好普适性.此外，异常工况引起的信号特征分布差异会显著影响回归预测模型的泛化性能，应尽量减少训练集与数据集差异以提高焊点质量预测的准确性.
- 电阻点焊 /
- 波动工况 /
- 多传感器融合 /
- 质量在线检测 /
- 机器学习
Abstract: Based on the features of multi-sensing signals in resistance spot welding process, online prediction models were established for the spot weld quality of different stack-ups in this paper. The influence of fluctuating welding conditions fluctuation on four machine learning regression models was studied, and the adaptability of different models and input variables on the database containing data of abnormal conditions was analyzed. Shapley value, and t-SNE methods were used to explain the model performance under fluctuating conditions. The results show that the Gaussian process regression model and resistance + force signal input had the best prediction performance of nugget diameter. Features of welding current, heat input and peak value of electrode displacement had good universality under fluctuating conditions. Besides, the difference of feature distribution caused by condition fluctuation could significantly influence the generalization performance of regression models. Thereby, the reduction of the difference between training set and test set could improve the prediction accuracy.
- resistance spot welding /
- fluctuating condition /
- multi-sensor fusion /
- online quality inspection /
- machine learning

HTML全文

图 1 标准、倾斜、边距和间隙工况的试样尺寸及设置方法示意图(mm)

Figure 1. Schematics of the specimens and setup methods of standard, electrode angle, edge proximity and initial gap conditions (mm). (a) standard condition; (b) electrode angle condition; (c) edge proximity condition; (d) initial gap condition

下载: 全尺寸图片幻灯片

图 2 不同板材组合的熔核直径

Figure 2. Nugget diameter of different stack-ups

下载: 全尺寸图片幻灯片

图 3 不同工况的熔核直径

Figure 3. Nugget diameter of different conditions

下载: 全尺寸图片幻灯片

图 4 回归预测建模流程图

Figure 4. Flowchart of regression prediction modeling

下载: 全尺寸图片幻灯片

图 5 过程信号特征提取示意图

Figure 5. Schematics of feature extraction methods of dynamic resistance, electrode force and electrode displacement signals. (a) dynamic resistance features; (b)electrode force features; (c) electrode displacement features

下载: 全尺寸图片幻灯片

图 6 不同机器学习回归模型的预测—实测值

Figure 6. Prediction-response results of different machine learning regression models. (a) multiple linear regression; (b) Gaussian process regression; (c) support vector regression; (d) multilayer perceptron regression

下载: 全尺寸图片幻灯片

图 7 输入特征值对GPR模型的Shapley值分布

Figure 7. Shapley value distribution of input features on GPR models. (a) model trained on ST dataset; (b) model trained on ALL dataset

下载: 全尺寸图片幻灯片

图 8 GPR模型泛化性测试的预测—实测值

Figure 8. Prediction-response results of GPR model generalization test. (a) training result in ST dataset; (b) training result in ALL-ST dataset

下载: 全尺寸图片幻灯片

图 9 不同工况下的特征分布

Figure 9. Feature distribution of in different welding conditions

下载: 全尺寸图片幻灯片

表 1 DP590和BUFD的化学成分(质量分数, %)

Table 1 The chemical components of DP590 and BUFD

材料	C	Si	Mn	P	S	其它	Fe
DP590	0.055	0.507	1.616	0.010	0.004	0.048	余量
BUFD	0.002	—	0.012	0.003	0.006	0.030	余量

下载: 导出CSV

表 2 DP590和BUFD的力学性能

Table 2 The mechanical properties of DP590 and BUFD

材料	屈服强度R_eL/MPa	抗拉强度R_m/MPa	断后伸长率A(%)
DP590	357	627	24.9
BUFD	162	288	49.0

下载: 导出CSV

表 3 焊接方案设计

Table 3 Welding schedule design.

序号	上板	下板	焊接电流 I/kA	电极力 F/kN	焊接时间t/ms
1	0.8 mm厚 DP590	0.8 mm厚 DP590	5 ~ 10	2.6 ~ 3.6	100 ~ 200
2	0.8 mm厚 BUFD	1.6 mm厚 BUFD	5 ~ 10	2.6 ~ 3.6	100 ~ 200
3	0.8 mm厚 DP590	1.6 mm厚 BUFD	5 ~ 10	2.6 ~ 3.6	100 ~ 200
4	1.6 mm厚 BUFD	1.6 mm厚 DP590	5 ~ 10	2.6 ~ 3.6	100 ~ 200

下载: 导出CSV

表 4 线性回归模型的预测性能对比

Table 4 Performance comparison of multiple linear regression models.

模型	A_{±1 mm} (%)	RMSE/mm
标准 MLR	86.82	0.695
含交互项 MLR	11.75	35.657
稳健性 MLR	85.67	0.728
逐步 MLR	89.68	0.665

下载: 导出CSV

表 5 高斯过程回归模型的预测性能对比

Table 5 Performance comparison of Gaussian process regression models.

模型	A_{±1 mm} (%)	RMSE/mm
二次有理GPR	90.54	0.600
指数GPR	91.12	0.571
平方指数GPR	90.54	0.905
可优化GPR	91.12	0.570

下载: 导出CSV

表 6 支持向量回归模型的预测性能对比

Table 6 Performance comparison of support vector regression models.

模型		A_{±1 mm} (%)	RMSE/mm
线性SVR		86.53	0.696
多项式SVR	二次	89.40	0.676
多项式SVR	三次	83.09	0.831
高斯SVR	粗略	86.25	0.725
	中等	89.40	0.637
	精细	75.07	0.966
可优化SVR		87.96	0.659
混合核函数SVR		87.11	0.761

下载: 导出CSV

表 7 神经网络回归模型的预测性能对比

Table 7 Performance comparison of multilayer perceptron regression models.

模型			A_{±1 mm} (%)	RMSE/mm
MLP		节点数	A_{±1 mm} (%)	RMSE/mm
单层MLP	10节点		81.95	0.907
	50节点		83.89	0.875
	150节点		86.24	0.768
	10节点		83.67	0.828
双层MLP	50节点		83.95	0.818
	150节点		85.57	0.765
	10节点		82.23	0.876
	50节点		85.96	0.777
三层MLP	150节点		86.25	0.728
	500节点		88.25	0.713
	1000节点		87.38	0.749
ResNet-MLP	150节点		17.23	3.151
ResNet-MLP	500节点		87.39	0.736

下载: 导出CSV

表 8 优化后回归模型的预测性能及稳定性

Table 8 Prediction performance and stability of optimized regression models.

模型	A_{±1 mm} (%)	RMSE/mm
MLR	86.61 ± 1.13	0.777 ± 0.038
GPR	90.13 ± 0.51	0.614 ± 0.023
SVR	88.08 ± 0.67	0.671 ± 0.014
MLP	87.45 ± 0.90	0.743 ± 0.037

下载: 导出CSV

表 9 不同信号特征输入下GPR模型的预测性能

Table 9 Prediction performance of GPR models with different input signals.

输入信号	训练集	A_{±1 mm} (%)	RMSE/mm
工艺参数	ST	88.30 ± 2.12	0.706 ± 0.045
工艺参数	ALL	76.68 ± 0.84	0.951 ± 0.010
电阻	ST	92.71 ± 1.95	0.604 ± 0.036
电阻	ALL	85.80 ± 2.68	0.729 ± 0.040
电极力	ST	90.11 ± 3.90	0.660 ± 0.173
电极力	ALL	85.27 ± 2.01	0.751 ± 0.027
位移	ST	94.01 ± 2.60	0.635 ± 0.035
位移	ALL	88.56 ± 0.84	0.716 ± 0.029
电阻 + 电极力	ST	92.06 ± 3.90	0.607 ± 0.054
电阻 + 电极力	ALL	91.24 ± 1.51	0.601 ± 0.026
电阻 + 位移	ST	90.32 ± 3.90	0.657 ± 0.135
电阻 + 位移	ALL	89.26 ± 2.18	0.628 ± 0.055
电极力 + 位移	ST	93.23 ± 3.25	0.628 ± 0.060
电极力 + 位移	ALL	88.32 ± 1.68	0.683 ± 0.057

下载: 导出CSV

表 10 GPR模型对波动工况的泛化性测试

Table 10 Generalization test of GPR model on welding condition fluctuation.

训练集工况	A_{±1 mm} (%)	RMSE/ mm	测试集工况	A_{±1 mm} (%)	RMSE/ mm
标准	94.10	0.567	倾斜 + 边距 + 间隙	79.18	0.833
标准 + 倾斜 + 边距	92.24	0.590	间隙	92.39	0.581
标准 + 倾斜 + 间隙	88.89	0.585	边距	51.21	1.361
标准 + 边距 + 间隙	91.60	0.583	倾斜	69.99	1.045

下载: 导出CSV

参考文献(20)

[1]	王敏. 电阻焊在汽车工业中的应用[J]. 电焊机, 2003, 33(1): 1 − 6. Wang Min. Application of resistance welding in automotive industry[J]. Electric Welding Machine, 2003, 33(1): 1 − 6.
[2]	Cho Y, Li W, Hu S J. Design of experiment analysis and weld lobe estimation for aluminum resistance spot welding[J]. Welding Journal, 2006, 85(3): 45s − 51s.
[3]	中国机械工程学会焊接学会电阻焊专业委员会(III). 电阻焊理论与实践[M]. 北京: 机械工业出版社, 1994. Professional Committee of Resistance Welding (III) of Welding Society of China Mechanical Engineering Society. Theory and Practice of Resistance Welding[M]. Beijing: China Machinery Industry Press, 1994.
[4]	夏裕俊, 李永兵, 楼铭, 等. 电阻点焊质量监控技术研究进展与分析[J]. 中国机械工程, 2020, 31(1): 100 − 125. doi: 10.3969/j.issn.1004-132X.2020.01.011 Xia Yujun, Li Yongbin, Lou Ming, et al. Recent advances and analysis of quality monitoring and control technologies for RSW[J]. China Mechanical Engineering, 2020, 31(1): 100 − 125. doi: 10.3969/j.issn.1004-132X.2020.01.011
[5]	Dickinson D W, Franklin J E, Stanya A. Characterization of spot welding behavior by dynamic electrical parameter monitoring[J]. Welding Journal, 1980, 59(6): 170s − 176s.
[6]	Chang H S, Kwon H C. In-process monitoring of micro resistance spot weld quality using accelerometer[J]. Korea Journal of Welding and Joining, 2011, 29(1): 115 − 122. doi: 10.5781/KWJS.2011.29.1.115
[7]	Zhang H Y, Hu J S. Resistance welding: Fundamentals and applications[M]. London: CRC Press, 2011.
[8]	Chen G, Sheng B, Luo R, et al. A parallel strategy for predicting the quality of welded joints in automotive bodies based on machine learning[J]. Journal of Manufacturing Systems, 2022, 62: 636 − 649. doi: 10.1016/j.jmsy.2022.01.011
[9]	Gavidel S Z, Lu S, Rickli J. Performance analysis and comparison of machine learning algorithms for predicting nugget width of resistance spot welding joints[J]. The International Journal of Advanced Manufacturing Technology, 2019, 105: 3779 − 3796. doi: 10.1007/s00170-019-03821-z
[10]	Zhao D, Bezgans Y, Wang Y, et al. Performances of dimension reduction techniques for welding quality prediction based on the dynamic resistance signal[J]. Journal of Manufacturing Processes, 2020, 58: 335 − 343. doi: 10.1016/j.jmapro.2020.08.037
[11]	El-Sari B, Biegler M, Rethmeier M. Investigation of the extrapolation capability of an artificial neural network algorithm in combination with process signals in resistance spot welding of advanced high-strength steels[J]. Metals, 2021, 11(11): 1874. doi: 10.3390/met11111874
[12]	Zhou B F, Pychynski T, Reischl M, et al. Comparison of machine learning approaches for time-series-based quality monitoring of resistance spot welding[J]. Archives of Data Science: Series A, 2018, 5(1): 1 − 17.
[13]	Xia Y J, Su Z W, Li Y B, et al. Online quantitative evaluation of expulsion in resistance spot welding[J]. Journal of Manufacturing Processes, 2019, 46(10): 34 − 43.
[14]	American welding society. Specification for automotive weld quality-resistance spot welding of steel[S]. AWS/ANSI Standard AWS D8.1M, 2013.
[15]	Zhao D W, Wang Y X, Liang D J. Correlating variations in the dynamic power signature to nugget diameter in resistance spot welding using Kriging model[J]. Measurement, 2019, 135: 6 − 12. doi: 10.1016/j.measurement.2018.11.025
[16]	Simončič S, Podržaj P. Resistance spot weld strength estimation based on electrode tip displacement/velocity curve obtained by image processing[J]. Science & Technology of Welding & Joining, 2014, 19(6): 468 − 475.
[17]	Wang L, Hou Y, Zhang H, et al. A new measurement method for the dynamic resistance signal during the resistance spot welding process[J]. Measurement Science and Technology, 2016, 27(9): 095009. doi: 10.1088/0957-0233/27/9/095009
[18]	Min J. Real time monitoring weld quality of resistance spot welding for the fabrication of sheet metal assemblies[J]. Journal of Materials Processing Technology, 2003, 132: 102 − 113. doi: 10.1016/S0924-0136(02)00409-0
[19]	Lee J, Noh I, Jeong S I, et al. Development of real-time diagnosis framework for angular misalignment of robot spot-welding system based on machine learning[J]. Procedia Manufacturing, 2020, 48: 1009 − 1019. doi: 10.1016/j.promfg.2020.05.140
[20]	Zhou L, Xia Y J, Shen Y, et al. Comparative study on resistance and displacement based adaptive output tracking control strategies for resistance spot welding[J]. Journal of Manufacturing Processes, 2021, 63(3): 98 − 108.

施引文献(8)

期刊类型引用(5)

1.	王通，孟惠民，葛鹏飞，李全德，巩秀芳，倪荣，姜英，龚显龙，戴君，隆彬. 2Cr-1Ni-1.2Mo-0.2V钢在NH_4H_2PO_4溶液中的电化学腐蚀行为研究. 中国腐蚀与防护学报. 2022(04): 551-562 . 百度学术
2.	苍雨，黄毓晖，翁硕，轩福贞. 环境变量对核电汽轮机转子钢焊接接头电偶腐蚀性能的影响. 中国腐蚀与防护学报. 2021(03): 318-326 . 百度学术
3.	黄毓晖，张建辉，胡语林，孙雯暄，徐宇斌，汪毅豪. 核电汽轮机转子堆焊焊接接头的电偶腐蚀行为及有限元仿真. 焊接学报. 2021(08): 33-39+99 . 本站查看
4.	陈君，康凯，冯钜，熊长奇，曹丹，尹杰. 压水堆核电站结构材料的腐蚀行为研究进展. 西华大学学报(自然科学版). 2020(03): 104-112 . 百度学术
5.	黄毓晖，司晓法，翁硕，轩福贞. 疲劳损伤对核电汽轮机焊接转子接头应力腐蚀开裂敏感性的影响. 焊接学报. 2020(04): 12-19+37+97-98 . 本站查看