Scholarlink

<p>Multi-parameter characteristics pose significant challenges in the accuracy of fatigue life prediction model for 2024-T3 clad aluminum (Al) alloy. In this study, machine learning (ML) algorithms are combined with physical model to enhance predictive capabilities. Fatigue life of the 2024-T3 clad Al alloy is evaluated, and a physical fatigue life predication model is established. Based on the experimental fatigue life data, a data-driven fatigue life prediction model is developed using support vector machine (SVM) and ant colony optimization (ACO) algorithms. Mean absolute percentage error of the ACO-SVM model decreases by 69.2% compared to that of the SVM model. ACO-SVM and differential evolution (DE) algorithm are determined to solve the physical parameters of the fatigue life predication model. Consequently, physical equation for the fatigue life prediction of 2024-T3 clad Al alloy assisted by ML is established. The results of uncertainty quantification, reliability analysis, and parameter sensitivity demonstrate that the physical equation exhibits strong generalization capability, engineering applicability, and physical interpretability. The hybrid framework of physical model assisted by ML offers a novel approach for calibrating parameters of the physical model utilized in predicting the fatigue life of 2024-T3 clad Al alloy.</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

Physical Equation for Fatigue Life Prediction of 2024-T3 Clad Al Alloy Assisted by Machine Learning

  • 期刊论文

  • 发表日期：

    2026年1月22日

查看全文
  • Kegui Yan,
  • Weibing Dai,
  • Wenliang Li,
  • Quanlong Zhou,
  • Zhaohong Liu,
  • Jinke Li,
  • Haitao Yue,
  • Chenguang Guo

摘要

Multi-parameter characteristics pose significant challenges in the accuracy of fatigue life prediction model for 2024-T3 clad aluminum (Al) alloy. In this study, machine learning (ML) algorithms are combined with physical model to enhance predictive capabilities. Fatigue life of the 2024-T3 clad Al alloy is evaluated, and a physical fatigue life predication model is established. Based on the experimental fatigue life data, a data-driven fatigue life prediction model is developed using support vector machine (SVM) and ant colony optimization (ACO) algorithms. Mean absolute percentage error of the ACO-SVM model decreases by 69.2% compared to that of the SVM model. ACO-SVM and differential evolution (DE) algorithm are determined to solve the physical parameters of the fatigue life predication model. Consequently, physical equation for the fatigue life prediction of 2024-T3 clad Al alloy assisted by ML is established. The results of uncertainty quantification, reliability analysis, and parameter sensitivity demonstrate that the physical equation exhibits strong generalization capability, engineering applicability, and physical interpretability. The hybrid framework of physical model assisted by ML offers a novel approach for calibrating parameters of the physical model utilized in predicting the fatigue life of 2024-T3 clad Al alloy.