Scholarlink

In mechanical engineering, fatigue life is a material’s ability to withstand cyclic loading without failure and is crucial for designing reliable structures and components. Traditional approaches for estimating fatigue life often fall short due to oversimplified assumptions and are often expensive to conduct. By using machine learning (ML) algorithms, these difficulties are overcome, and fatigue life is accurately predicted. This research explores the application of ML in predicting the fatigue life of four materials: steel and three additive-manufactured alloys: Ti-6Al-4V, IN718, and AlSi10Mg. Various factors have been considered, such as temperature, heat treatment process, and chemical composition. The study implements and compares the performance of three ML models: gradient boosting machine (GBM), extreme gradient boosting (XGBoost), categorical boosting, and evaluation metrics such as R-squared (R2) and root mean square error. The results show that XGBoost performs the best for steel and IN718 with R2 of 0.98311 and 0.87836, respectively, while GBM performs the best for Ti-6Al-4V and AlSi10Mg with R2 of 0.80542 and 0.72513, respectively.

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

Fatigue Life Prediction of Different Materials Using Machine Learning

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Satyam Mishra,
  • Apar Thakur,
  • Siddharth Majumdar,
  • Yash Verma,
  • Aditya Kumar

摘要

In mechanical engineering, fatigue life is a material’s ability to withstand cyclic loading without failure and is crucial for designing reliable structures and components. Traditional approaches for estimating fatigue life often fall short due to oversimplified assumptions and are often expensive to conduct. By using machine learning (ML) algorithms, these difficulties are overcome, and fatigue life is accurately predicted. This research explores the application of ML in predicting the fatigue life of four materials: steel and three additive-manufactured alloys: Ti-6Al-4V, IN718, and AlSi10Mg. Various factors have been considered, such as temperature, heat treatment process, and chemical composition. The study implements and compares the performance of three ML models: gradient boosting machine (GBM), extreme gradient boosting (XGBoost), categorical boosting, and evaluation metrics such as R-squared (R2) and root mean square error. The results show that XGBoost performs the best for steel and IN718 with R2 of 0.98311 and 0.87836, respectively, while GBM performs the best for Ti-6Al-4V and AlSi10Mg with R2 of 0.80542 and 0.72513, respectively.