Collaborative and Quantitative Prediction for Reinforcement and Penetration Depth of Weld Bead Based on Molten Pool Image and Deep Residual Network
Weld quality is generally determined by reinforcement and penetration depth of weld bead in arc welding. Penetration depth reflects weld strength and reinforcement reflects weld shape. What's more, there is a strong coupling between them, therefore it is necessary to monitor them collaborativel...
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IEEE
2020-01-01
|
| Series: | IEEE Access |
| Subjects: | |
| Online Access: | https://ieeexplore.ieee.org/document/9134710/ |
| _version_ | 1818323822193934336 |
|---|---|
| author | Jun Lu Yumin Shi Lianfa Bai Zhuang Zhao Jing Han |
| author_facet | Jun Lu Yumin Shi Lianfa Bai Zhuang Zhao Jing Han |
| author_sort | Jun Lu |
| collection | DOAJ |
| description | Weld quality is generally determined by reinforcement and penetration depth of weld bead in arc welding. Penetration depth reflects weld strength and reinforcement reflects weld shape. What's more, there is a strong coupling between them, therefore it is necessary to monitor them collaboratively and quantitatively. In this paper, vision sensor system is utilized to collect non-interfering molten pool images during CMT (Cold Metal Transfer) welding process. Meanwhile, according to the locating system and the metallographic diagram, the position of molten pool image on weld bead is measured, as well as the corresponding reinforcement and penetration. A reinforcement-penetration collaborative prediction network model based on deep residual is designed to quantitatively predict reinforcement and penetration depth. Combining the physical_mechanisms of forming process, we optimize the network structure that the backbone network is Resnet34, the input is frequency domain image of the middle and rear areas of molten pool, and the output is dual and synergetic. The correlation characteristics that affect reinforcement and penetration depth in molten pool images are fully studied. The reinforcement prediction error is less than 0.13 mm and penetration depth is less than 0.09 mm for various welding parameters and workpiece shapes. |
| first_indexed | 2024-12-13T11:18:48Z |
| format | Article |
| id | doaj.art-828de8146b744fae9633d6483eeb6dc0 |
| institution | Directory Open Access Journal |
| issn | 2169-3536 |
| language | English |
| last_indexed | 2024-12-13T11:18:48Z |
| publishDate | 2020-01-01 |
| publisher | IEEE |
| record_format | Article |
| series | IEEE Access |
| spelling | doaj.art-828de8146b744fae9633d6483eeb6dc02022-12-21T23:48:33ZengIEEEIEEE Access2169-35362020-01-01812613812614810.1109/ACCESS.2020.30078159134710Collaborative and Quantitative Prediction for Reinforcement and Penetration Depth of Weld Bead Based on Molten Pool Image and Deep Residual NetworkJun Lu0Yumin Shi1Lianfa Bai2https://orcid.org/0000-0002-6688-4529Zhuang Zhao3https://orcid.org/0000-0002-2052-8633Jing Han4https://orcid.org/0000-0002-1033-566XJiangsu Key Laboratory of Spectral Imaging and Intelligent Sense, Nanjing University of Science and Technology, Nanjing, ChinaJiangsu Key Laboratory of Spectral Imaging and Intelligent Sense, Nanjing University of Science and Technology, Nanjing, ChinaJiangsu Key Laboratory of Spectral Imaging and Intelligent Sense, Nanjing University of Science and Technology, Nanjing, ChinaJiangsu Key Laboratory of Spectral Imaging and Intelligent Sense, Nanjing University of Science and Technology, Nanjing, ChinaJiangsu Key Laboratory of Spectral Imaging and Intelligent Sense, Nanjing University of Science and Technology, Nanjing, ChinaWeld quality is generally determined by reinforcement and penetration depth of weld bead in arc welding. Penetration depth reflects weld strength and reinforcement reflects weld shape. What's more, there is a strong coupling between them, therefore it is necessary to monitor them collaboratively and quantitatively. In this paper, vision sensor system is utilized to collect non-interfering molten pool images during CMT (Cold Metal Transfer) welding process. Meanwhile, according to the locating system and the metallographic diagram, the position of molten pool image on weld bead is measured, as well as the corresponding reinforcement and penetration. A reinforcement-penetration collaborative prediction network model based on deep residual is designed to quantitatively predict reinforcement and penetration depth. Combining the physical_mechanisms of forming process, we optimize the network structure that the backbone network is Resnet34, the input is frequency domain image of the middle and rear areas of molten pool, and the output is dual and synergetic. The correlation characteristics that affect reinforcement and penetration depth in molten pool images are fully studied. The reinforcement prediction error is less than 0.13 mm and penetration depth is less than 0.09 mm for various welding parameters and workpiece shapes.https://ieeexplore.ieee.org/document/9134710/Collaborative and quantitative predictionreinforcementpenetration depthdeep residual networknetwork structureprediction error |
| spellingShingle | Jun Lu Yumin Shi Lianfa Bai Zhuang Zhao Jing Han Collaborative and Quantitative Prediction for Reinforcement and Penetration Depth of Weld Bead Based on Molten Pool Image and Deep Residual Network IEEE Access Collaborative and quantitative prediction reinforcement penetration depth deep residual network network structure prediction error |
| title | Collaborative and Quantitative Prediction for Reinforcement and Penetration Depth of Weld Bead Based on Molten Pool Image and Deep Residual Network |
| title_full | Collaborative and Quantitative Prediction for Reinforcement and Penetration Depth of Weld Bead Based on Molten Pool Image and Deep Residual Network |
| title_fullStr | Collaborative and Quantitative Prediction for Reinforcement and Penetration Depth of Weld Bead Based on Molten Pool Image and Deep Residual Network |
| title_full_unstemmed | Collaborative and Quantitative Prediction for Reinforcement and Penetration Depth of Weld Bead Based on Molten Pool Image and Deep Residual Network |
| title_short | Collaborative and Quantitative Prediction for Reinforcement and Penetration Depth of Weld Bead Based on Molten Pool Image and Deep Residual Network |
| title_sort | collaborative and quantitative prediction for reinforcement and penetration depth of weld bead based on molten pool image and deep residual network |
| topic | Collaborative and quantitative prediction reinforcement penetration depth deep residual network network structure prediction error |
| url | https://ieeexplore.ieee.org/document/9134710/ |
| work_keys_str_mv | AT junlu collaborativeandquantitativepredictionforreinforcementandpenetrationdepthofweldbeadbasedonmoltenpoolimageanddeepresidualnetwork AT yuminshi collaborativeandquantitativepredictionforreinforcementandpenetrationdepthofweldbeadbasedonmoltenpoolimageanddeepresidualnetwork AT lianfabai collaborativeandquantitativepredictionforreinforcementandpenetrationdepthofweldbeadbasedonmoltenpoolimageanddeepresidualnetwork AT zhuangzhao collaborativeandquantitativepredictionforreinforcementandpenetrationdepthofweldbeadbasedonmoltenpoolimageanddeepresidualnetwork AT jinghan collaborativeandquantitativepredictionforreinforcementandpenetrationdepthofweldbeadbasedonmoltenpoolimageanddeepresidualnetwork |