Scholarlink

<p>It is argued unfavorably that the exploitation on sample features in existing reliability analysis method based on active learning surrogate model is inadequate, resulting in a notable decrease in computational efficiency. This study proposes an active learning polynomial chaos expansion model with adaptive weightage (AW-APCE) for structural reliability analysis. Firstly, the weight matrix of the training samples is defined in conjunction with the Gauss-Markov theorem, eliminating the influence of heteroscedasticity on accuracy of polynomial chaos expansion (PCE) model. A method is then proposed to estimate the local prediction error of PCE model by analyzing the differences between the basic functions. Finally, an adaptive weighted learning function is constructed, which adaptively updates the sample weights based on the interrelationships among the training samples. The proficiency of the proposed method is demonstrated through two numerical analysis examples and a solid rocket motor with an emphasis on reliability analysis of noisy data.</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

Active learning PCE method with adaptive weightage for reliability analysis and its application in chamber of solid rocket motor

  • 期刊论文

  • 发表日期：

    2025年7月1日

查看全文
  • Tianci Wang,
  • Hongfei Zhang,
  • Meide Yang,
  • Dapeng Wang,
  • Fang Wang

摘要

It is argued unfavorably that the exploitation on sample features in existing reliability analysis method based on active learning surrogate model is inadequate, resulting in a notable decrease in computational efficiency. This study proposes an active learning polynomial chaos expansion model with adaptive weightage (AW-APCE) for structural reliability analysis. Firstly, the weight matrix of the training samples is defined in conjunction with the Gauss-Markov theorem, eliminating the influence of heteroscedasticity on accuracy of polynomial chaos expansion (PCE) model. A method is then proposed to estimate the local prediction error of PCE model by analyzing the differences between the basic functions. Finally, an adaptive weighted learning function is constructed, which adaptively updates the sample weights based on the interrelationships among the training samples. The proficiency of the proposed method is demonstrated through two numerical analysis examples and a solid rocket motor with an emphasis on reliability analysis of noisy data.