Scholarlink

Abstract <p>This study utilizes machine learning (ML) methodology to estimate the stress intensity factor (SIF) for multiple edge cracks in a coating-substrate pair. The arbitrarily varying loading function is decomposed into a weighted sum of sine and cosine functions using Fourier series expansion, from which extracted are the characteristic period and harmonic number. A large data set derived from finite element calculation is used to train the ML model. By validation and comparison, it is found that the even extension method offers the highest accuracy in estimating the SIF. For three different loading functions, the predicted results show an average error of less than 1% compared to those by the finite element method. Additionally, the error of the predicted results is less than 3% in comparison with those in two thermal shock scenarios from existing literatures. The findings highlight the potential of ML-driven computational frameworks to achieve efficient and accurate evaluation of SIF for multiple cracks under realistic service conditions.</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

Machine Learning Prediction on the Stress Intensity Factor for Multiple Edge Cracks in Coatings under Arbitrarily Varying Loads

  • 期刊论文

  • 发表日期：

    2025年8月3日

查看全文
  • W. Y. Liu,
  • X. J. Chen

摘要

Abstract

This study utilizes machine learning (ML) methodology to estimate the stress intensity factor (SIF) for multiple edge cracks in a coating-substrate pair. The arbitrarily varying loading function is decomposed into a weighted sum of sine and cosine functions using Fourier series expansion, from which extracted are the characteristic period and harmonic number. A large data set derived from finite element calculation is used to train the ML model. By validation and comparison, it is found that the even extension method offers the highest accuracy in estimating the SIF. For three different loading functions, the predicted results show an average error of less than 1% compared to those by the finite element method. Additionally, the error of the predicted results is less than 3% in comparison with those in two thermal shock scenarios from existing literatures. The findings highlight the potential of ML-driven computational frameworks to achieve efficient and accurate evaluation of SIF for multiple cracks under realistic service conditions.