Scholarlink

<p>This study compiled experimental data from the public literature and employed six machine learning algorithms to predict and compare the tensile strength of 6061-T6 aluminum alloy friction-stir-welded joints under various process parameters. The models evaluated include eXtreme Gradient Boosting (XGBoost), Random Forest (RF), Support Vector Regression (SVR), k-Nearest Neighbors (KNN), Gradient Boosting Decision Tree (GBDT), and Multilayer Perceptron (MLP). Feature selection was optimized using the Pearson correlation coefficient and dimensionality reduction techniques, reducing the number of input features from seven to five. Furthermore, SHapley Additive exPlanations (SHAP) were employed to interpret the relationships between disparate features and the tensile strength of the welded joints. The prediction results show that the RFE + XGBoost model achieved a mean absolute percentage error (MAPE) of 2.98%, a coefficient of determination (R<sup>2</sup>) of 0.973, a mean absolute error (MSE) of 45.31, and a mean absolute error (MAE) of 5.25, indicating its effectiveness in predicting the mechanical properties of 6061-T6 aluminum alloy welded joints. Further analyses showed that the features affecting the performance of welded joints were ranked in the following order of importance: welding speed (24.84), rotational speed (10.86), plate thickness (7.93), stirring needle length (7.55), and stirring needle shoulder shaft diameter (5.48).</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

A Machine Learning-based Approach to Predicting Tensile Strength of 6061-T6 Aluminum Alloy Friction Stir Welded Joints

  • 期刊论文

  • 发表日期：

    2026年1月21日

查看全文
  • Shuo Yang,
  • Mengqi Cong,
  • Xiaowei Zhuang,
  • Yang Zhang,
  • Xiaorui Zheng,
  • Bin Shi,
  • Jianhua Wang

摘要

This study compiled experimental data from the public literature and employed six machine learning algorithms to predict and compare the tensile strength of 6061-T6 aluminum alloy friction-stir-welded joints under various process parameters. The models evaluated include eXtreme Gradient Boosting (XGBoost), Random Forest (RF), Support Vector Regression (SVR), k-Nearest Neighbors (KNN), Gradient Boosting Decision Tree (GBDT), and Multilayer Perceptron (MLP). Feature selection was optimized using the Pearson correlation coefficient and dimensionality reduction techniques, reducing the number of input features from seven to five. Furthermore, SHapley Additive exPlanations (SHAP) were employed to interpret the relationships between disparate features and the tensile strength of the welded joints. The prediction results show that the RFE + XGBoost model achieved a mean absolute percentage error (MAPE) of 2.98%, a coefficient of determination (R2) of 0.973, a mean absolute error (MSE) of 45.31, and a mean absolute error (MAE) of 5.25, indicating its effectiveness in predicting the mechanical properties of 6061-T6 aluminum alloy welded joints. Further analyses showed that the features affecting the performance of welded joints were ranked in the following order of importance: welding speed (24.84), rotational speed (10.86), plate thickness (7.93), stirring needle length (7.55), and stirring needle shoulder shaft diameter (5.48).