Hybrid machine learning-enabled adaptive welding speed control

被引:29
作者
Kershaw, Joseph [1 ,2 ]
Yu, Rui [2 ,3 ]
Zhang, Yuming [2 ,3 ]
Wang, Peng [1 ,2 ,3 ]
机构
[1] Univ Kentucky, Dept Mech Engn, Lexington, KY 40506 USA
[2] Univ Kentucky, Inst Sustainable Mfg, Lexington, KY 40506 USA
[3] Univ Kentucky, Dept Elect & Comp Engn, Lexington, KY 40506 USA
基金
美国国家科学基金会;
关键词
Robotic welding; Adaptive control; Machine learning; Convolutional neural network; Gradient descent; PENETRATION; PREDICTION;
D O I
10.1016/j.jmapro.2021.09.023
中图分类号
T [工业技术];
学科分类号
08 ;
摘要
Industrial robots have become more diverse and common for automating manufacturing processes, such as welding. Existing robotic control, however, is incapable of adaptively adjusting its operation in response to a dynamic welding environment, whereas a skilled human welder can. Sophisticated and adaptive robotic control relies on the effective and efficient processing of perception data, characterization and prediction of highly dynamic systems, and real-time adaptative robotic reactions. This research presents a preliminary study on developing appropriate Machine Learning (ML) techniques for real-time welding quality prediction and adaptive welding speed adjustment for GTAW welding at a constant current. In order to collect the data needed to train the hybrid ML models, two cameras are applied to monitor the welding process, with one camera (available in practical robotic welding) recording the top-side weld pool dynamics and a second camera (unavailable in practical robotic welding, but applicable for training purpose) recording the back-side bead formation. Given these two data sets, correlations can be discovered through a convolutional neural network (CNN) that is good at image characterization. With the CNN, top-side weld pool images can be analyzed to predict the back-side bead width during active welding control. Furthermore, the monitoring process has been applied to multiple experimental trials with varying speeds. This allowed the effect of welding speed on bead width to be modeled through a Multi-Layer Perceptron (MLP). Through the trained MLP, a computationally efficient gradient descent algorithm has been developed to adjust the travel speed accordingly to achieve an optimal bead width with full material penetration. Because of the nature of gradient descent, the robot would change faster when the quality is further away and then fine-tune the speed when it was close to the goal. Experimental studies have shown promising results on real-time bead width prediction and adaptive speed adjustment to realize ideal bead width.
引用
收藏
页码:374 / 383
页数:10
相关论文
共 26 条
[1]  
[Anonymous], 2018, NAT WELD MONTH TIM L
[2]   Penetration quality prediction of asymmetrical fillet root welding based on optimized BP neural network [J].
Chang, Yushuo ;
Yue, Jianfeng ;
Guo, Rui ;
Liu, Wenji ;
Li, Liangyu .
JOURNAL OF MANUFACTURING PROCESSES, 2020, 50 (50) :247-254
[3]   A study on V-groove GMAW for various welding positions [J].
Cho, D. W. ;
Na, S. J. ;
Cho, M. H. ;
Lee, J. S. .
JOURNAL OF MATERIALS PROCESSING TECHNOLOGY, 2013, 213 (09) :1640-1652
[4]   Robot path planning optimization for welding complex joints [J].
Fang, H. C. ;
Ong, S. K. ;
Nee, A. Y. C. .
INTERNATIONAL JOURNAL OF ADVANCED MANUFACTURING TECHNOLOGY, 2017, 90 (9-12) :3829-3839
[5]  
Hastie T., 2009, Springer series in statistics, Vxxii, DOI 10.1007/978-0-387-84858-7
[6]  
Kingma DP, 2015, C TRACK P
[7]   ImageNet Classification with Deep Convolutional Neural Networks [J].
Krizhevsky, Alex ;
Sutskever, Ilya ;
Hinton, Geoffrey E. .
COMMUNICATIONS OF THE ACM, 2017, 60 (06) :84-90
[8]   Data-Driven Approach For Robot-Assisted Multi-Pass-Welding Thick Sheet Metal Connections [J].
Lauer, Sascha ;
Wiese, Philipp ;
Dryba, Steffen ;
Flugge, Wilko .
PROCEEDINGS OF THE 5TH INTERNATIONAL CONFERENCE ON SYSTEM-INTEGRATED INTELLIGENCE (SYSINT 2020): SYSTEM-INTEGRATED INTELLIGENCE - INTELLIGENT, FLEXIBLE AND CONNECTED SYSTEMS IN PRODUCTS AND PRODUCTION, 2020, 52 :95-100
[9]   Robots in the shipbuilding industry [J].
Lee, Donghun .
ROBOTICS AND COMPUTER-INTEGRATED MANUFACTURING, 2014, 30 (05) :442-450
[10]   Model-Based Predictive Control of Weld Penetration in Gas Tungsten Arc Welding [J].
Liu, Yu Kang ;
Zhang, Yu Ming .
IEEE TRANSACTIONS ON CONTROL SYSTEMS TECHNOLOGY, 2014, 22 (03) :955-966