Scholarlink

<p>This study explores the integration of machine learning (ML) models within a digital twin (DT) framework to predict and correlate the fatigue life of CFRP-to-aluminum adhesive joints exposed to hygrothermal aging. By combining experimental testing with advanced ML techniques, the research addresses the relationship between natural and accelerated aging effects on adhesive joints. The joints were fabricated and subjected to natural aging for periods ranging from 1 to 3&#xa0;years. In parallel, accelerated aging was conducted under hygrothermal conditions for durations between 4 and 50&#xa0;days. Fatigue life was assessed using three-point bending tests. To predict the effects of natural aging based on accelerated aging data, five ML algorithms were employed: Support Vector Regression (SVR), Artificial Neural Network (ANN), linear regression, Random Forest (RF), and Extreme Gradient Boosting (XGBoost). Among these, XGBoost provided highly accurate predictions, while SVR and linear regression showed weaker performance with notable prediction errors. The fusion of digital twin technology with machine learning demonstrated strong potential for real-time modeling and accurate fatigue life prediction, enhancing the durability and reliability of composite structures. This work emphasizes the promise of advanced ML and DT methods in optimizing the performance and maintenance strategies of adhesive joints.</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 digital twin and machine learning approach for real-time fatigue life prediction of CFRP-to-aluminum adhesive joints subjected to environmental aging

  • 期刊论文

  • 发表日期：

    2025年7月11日

查看全文
  • Ardavan Anvari,
  • Sajjad Karimi,
  • Amin Latifi

摘要

This study explores the integration of machine learning (ML) models within a digital twin (DT) framework to predict and correlate the fatigue life of CFRP-to-aluminum adhesive joints exposed to hygrothermal aging. By combining experimental testing with advanced ML techniques, the research addresses the relationship between natural and accelerated aging effects on adhesive joints. The joints were fabricated and subjected to natural aging for periods ranging from 1 to 3 years. In parallel, accelerated aging was conducted under hygrothermal conditions for durations between 4 and 50 days. Fatigue life was assessed using three-point bending tests. To predict the effects of natural aging based on accelerated aging data, five ML algorithms were employed: Support Vector Regression (SVR), Artificial Neural Network (ANN), linear regression, Random Forest (RF), and Extreme Gradient Boosting (XGBoost). Among these, XGBoost provided highly accurate predictions, while SVR and linear regression showed weaker performance with notable prediction errors. The fusion of digital twin technology with machine learning demonstrated strong potential for real-time modeling and accurate fatigue life prediction, enhancing the durability and reliability of composite structures. This work emphasizes the promise of advanced ML and DT methods in optimizing the performance and maintenance strategies of adhesive joints.