| 王志颖,孙凯华,何华彬.基于深度分类网络与激光超声的焊深测量方法[J].,2025,44(5):1150-1158 |
| 基于深度分类网络与激光超声的焊深测量方法 |
| Welding depth measurement method based on depth classification network and laser ultrasonic |
| 投稿时间：2024-04-07 修订日期：2025-08-29 |
| 中文摘要: |
| 焊深是衡量焊接质量的重要参数，其对焊接产品的性能有极大影响，因此发展焊深的无损测量方法尤为重要。在众多无损测量方法中，激光超声凭借其无耦合剂、高时空分辨率等特点，在小结构焊深测量中具有独特的优势。由于小尺寸结构中存在超声模式混叠且焊缝的衍射波微弱的问题，该文提出将激光超声与深度学习相结合的焊深测量方法，在线切割模拟焊接样品上进行激光超声扫描实验。将超声A扫信号压缩为矩阵格式可加速深度网络收敛、减少网络参数规模。与反射横波模式方法相比，训练得到的深度分类模型能够实现焊深测量误差小于0.1 mm。最终的结果表明，深度网络在焊深测量中具有良好的发展前景。 |
| 英文摘要: |
| Welding depth is an important parameter for measuring welding quality and has a great impact on the performance of welded products. Therefore, it is particularly important to develop non-destructive measurement methods for welding depth. Among many non-destructive measurement methods, laser ultrasonic has unique advantages in weld depth measurement of small structures due to its characteristics of no coupling agent and high spatial and temporal resolution. Due to the problems of ultrasonic mode mixing in small-size structures and weak diffraction waves of welds, this paper proposes a welding depth measurement method that combines laser ultrasonic and deep learning to conduct laser ultrasonic scanning experiments on wire-cut simulated welding samples. Compressing the ultrasonic A-scan signal into a matrix format accelerates the convergence of deep networks and reduces the scale of network parameters. Compared with the reflected shear wave mode method, the trained depth classification model can achieve a welding depth measurement error of less than 0.1 mm. The final results show that the depth network has good development prospects in welding depth measurement. |
| DOI：10.11684/j.issn.1000-310X.2025.05.006 |
| 中文关键词: 激光超声 深度学习 焊深测量 |
| 英文关键词: Laser ultrasonic Deep learning Welding depth measurement |
| 基金项目: |
|
| 摘要点击次数: 702 |
| 全文下载次数: 324 |
|
查看全文
查看/发表评论 下载PDF阅读器 |
| 关闭 |
