Scholarlink

<p>The purpose of this paper is to study the quality control technology of multi-layer weaving welding (MLWW) process for V-shaped grooves of medium and thick plates. Firstly, an improved P-SegNet semantic segmentation network is proposed to achieve accurate recognition of swing welding pool images. Then, taking the size of the melt pool and welding process parameters as input conditions, a prediction model for weld width and sidewall penetration based on GWO-BPNN network was constructed. Furthermore, a control strategy is proposed for correcting the deviation in the swing welding process using the prediction results of weld forming dimensions. The experimental results show that the recognition rate of the molten pool using the algorithm proposed in this article reaches 98.85%, and the average prediction error of weld size is less than 5%, which can meet the accuracy requirements of sidewall penetration control and quality evaluation. Finally, a swing welding process correction experiment was conducted using the weld forming size.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Extraction of Weld Pool Contours and Quality Prediction of Fillet Welds in Fillet Welding Based on Deep Learning with Welding Deviation Correction

  • 期刊论文

  • 发表日期：

    2025年2月27日

查看全文
  • Tianqi Wang,
  • Jiaxin Hao,
  • Junjie He,
  • Lifeng Du

摘要

The purpose of this paper is to study the quality control technology of multi-layer weaving welding (MLWW) process for V-shaped grooves of medium and thick plates. Firstly, an improved P-SegNet semantic segmentation network is proposed to achieve accurate recognition of swing welding pool images. Then, taking the size of the melt pool and welding process parameters as input conditions, a prediction model for weld width and sidewall penetration based on GWO-BPNN network was constructed. Furthermore, a control strategy is proposed for correcting the deviation in the swing welding process using the prediction results of weld forming dimensions. The experimental results show that the recognition rate of the molten pool using the algorithm proposed in this article reaches 98.85%, and the average prediction error of weld size is less than 5%, which can meet the accuracy requirements of sidewall penetration control and quality evaluation. Finally, a swing welding process correction experiment was conducted using the weld forming size.