Scholarlink

<p>The quality of resistance spot welding is particularly crucial for the welding manufacturing industry. Inspecting the quality of welding joints efficiently and accurately is a significant process. When inspecting high strength steel welding joints, traditional regression models exhibit limited generalization capabilities. In this paper, we propose a CSSA-XGBoost algorithm to predict the tensile shear strength of welding joints by extracting representative features from the dynamic resistance signals during the welding process as model inputs. Specifically, an improved sparrow search algorithm for XGBoost optimization is used, which introduces a chaotic map and Levy flight strategy to improve the quality of hyperparameter search. We conduct comprehensive experiments on the self-built dataset WeldResDB to demonstrate the state-of-the-art of our approach. The root mean square error (RMSE) in prediction reaches 0.3423&#xa0;KN and the mean absolute percentage error (MAPE) is only 2.25%. The prediction accuracy outperforms other algorithms and provides a reliable method for weld quality inspection in industrial applications.</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

CSSA-XGBoost: a novel algorithm for inspecting spot welding quality based on dynamic resistance signal

  • 期刊论文

  • 发表日期：

    2025年2月9日

查看全文
  • Rui Wang,
  • Ruichen Mi,
  • Hao Xu,
  • Zhonglin Gao,
  • Weipeng Liu,
  • Tao Liang,
  • Kun Liu

摘要

The quality of resistance spot welding is particularly crucial for the welding manufacturing industry. Inspecting the quality of welding joints efficiently and accurately is a significant process. When inspecting high strength steel welding joints, traditional regression models exhibit limited generalization capabilities. In this paper, we propose a CSSA-XGBoost algorithm to predict the tensile shear strength of welding joints by extracting representative features from the dynamic resistance signals during the welding process as model inputs. Specifically, an improved sparrow search algorithm for XGBoost optimization is used, which introduces a chaotic map and Levy flight strategy to improve the quality of hyperparameter search. We conduct comprehensive experiments on the self-built dataset WeldResDB to demonstrate the state-of-the-art of our approach. The root mean square error (RMSE) in prediction reaches 0.3423 KN and the mean absolute percentage error (MAPE) is only 2.25%. The prediction accuracy outperforms other algorithms and provides a reliable method for weld quality inspection in industrial applications.