Scholarlink

<p>Wire arc additive manufacturing (WAAM) is an emerging technology that enables efficient production of complex metal parts with high deposition rate and reduced material waste. This study aims to optimize deposition efficiency (DE) for ER70S-6 material in WAAM using six machine learning (ML) models, the models were trained on process parameters such as travel speed (TS), wire feed rate (WFR), and distance from the torch to the base (DFTB). Among the models, the Polynomial Regression (PR) model provided the best performance with a Coefficient of Determination (<i>R</i><sup>2</sup>) value of 0.97 and minimal error metrics root mean square error (RMSE) of 2.24, mean squared error (MSE) of 5.03, and mean absolute error (MAE) of 1.92. K-fold cross validation (<i>k</i> = 5) results confirmed the robustness of the PR model with a mean <i>R</i><sup>2</sup> of 0.91. SHAP analysis revealed that TS had the most significant impact on DE. In addition, the PR outperformed the traditional Taguchi method with lower error % of 0.84. This demonstrates that PR-based ML prediction and optimization offers better accuracy in real-world applications.</p> Graphical abstract <p></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

Machine learning-based prediction and optimization of WAAM process parameters for enhanced deposition efficiency

  • 期刊论文

  • 发表日期：

    2025年8月4日

查看全文
  • P. Saravanamuthukumar,
  • Ahmad Baharuddin Abdullah,
  • Zarirah Karrim Wani

摘要

Wire arc additive manufacturing (WAAM) is an emerging technology that enables efficient production of complex metal parts with high deposition rate and reduced material waste. This study aims to optimize deposition efficiency (DE) for ER70S-6 material in WAAM using six machine learning (ML) models, the models were trained on process parameters such as travel speed (TS), wire feed rate (WFR), and distance from the torch to the base (DFTB). Among the models, the Polynomial Regression (PR) model provided the best performance with a Coefficient of Determination (R2) value of 0.97 and minimal error metrics root mean square error (RMSE) of 2.24, mean squared error (MSE) of 5.03, and mean absolute error (MAE) of 1.92. K-fold cross validation (k = 5) results confirmed the robustness of the PR model with a mean R2 of 0.91. SHAP analysis revealed that TS had the most significant impact on DE. In addition, the PR outperformed the traditional Taguchi method with lower error % of 0.84. This demonstrates that PR-based ML prediction and optimization offers better accuracy in real-world applications.

Graphical abstract