Scholarlink

<p>Friction stir welding (FSW) is a promising technique for joining aluminum alloys used in sectors requiring both lightness and high mechanical performance. However, optimizing welding parameters remains challenging to ensure optimal joint properties. The objective of this study is to develop reliable predictive models for anticipating the mechanical performance of AA6082-T6 alloy welded by FSW. A three-factor experimental design (rotation speed, feed rate, penetration depth) generated 27 configurations analyzed by the response surface method (RSM), while an additional 9 independent experiments were conducted to validate the RSM predictions. All 36 experimental data points were then used for artificial neural network (ANN) modeling. The ANN models were trained using three backpropagation algorithms Stepped Conjugate Gradient, Levenberg Marquardt, and Bayesian Regularization. Bayesian Regularization provided the best training performance, achieving an R² greater than 0.9998 for both ultimate tensile strength and elongation. The networks were further evaluated using K-fold cross-validation and 5 additional independent experiments, confirming their robustness and generalization ability. The network architecture consists of a single hidden layer with a sigmoid activation function and a linear output layer. Compared to conventional RSM models, the ANN reduced prediction errors by 94.6%, demonstrating the value of advanced neural approaches for accurate FSW process optimization and highlighting their potential for industrial application.</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

Prediction and optimization of mechanical properties in FSWed AA6082-T6 joints using RSM and artificial neural networks

  • 期刊论文

  • 发表日期：

    2025年12月23日

查看全文
  • Mohammed Abdelghani Ben Messaoud,
  • Mohammed Nadhir Djamel Eddine Cherief,
  • Abdelkader Miloudi,
  • Abdrrahim Belloufi,
  • Amina Belaribi,
  • Ismail Chekalil,
  • Abdelhalim Zoukel,
  • Mohammed Mousaab Blaoui

摘要

Friction stir welding (FSW) is a promising technique for joining aluminum alloys used in sectors requiring both lightness and high mechanical performance. However, optimizing welding parameters remains challenging to ensure optimal joint properties. The objective of this study is to develop reliable predictive models for anticipating the mechanical performance of AA6082-T6 alloy welded by FSW. A three-factor experimental design (rotation speed, feed rate, penetration depth) generated 27 configurations analyzed by the response surface method (RSM), while an additional 9 independent experiments were conducted to validate the RSM predictions. All 36 experimental data points were then used for artificial neural network (ANN) modeling. The ANN models were trained using three backpropagation algorithms Stepped Conjugate Gradient, Levenberg Marquardt, and Bayesian Regularization. Bayesian Regularization provided the best training performance, achieving an R² greater than 0.9998 for both ultimate tensile strength and elongation. The networks were further evaluated using K-fold cross-validation and 5 additional independent experiments, confirming their robustness and generalization ability. The network architecture consists of a single hidden layer with a sigmoid activation function and a linear output layer. Compared to conventional RSM models, the ANN reduced prediction errors by 94.6%, demonstrating the value of advanced neural approaches for accurate FSW process optimization and highlighting their potential for industrial application.