Scholarlink

<p>Monitoring the structural integrity of laminated composite materials, such as those used in aerospace structures, presents significant challenges due to the complexity of failure mechanisms like delaminations, which are often invisible on the surface. This study proposes an innovative methodology for identifying multiple delaminations using a combination of optimization techniques and artificial intelligence, addressing a highly complex problem in both fields. The Sunflower Optimization Algorithm (SFO) was explored to solve the inverse problem, while Machine Learning (ML) models and Artificial Neural Networks (ANN) were employed for classifying the number of delaminations and regressing their parameters. The damages were parameterized as ellipses in a square composite laminate plate modeled using finite elements. Strain data, obtained from the forward problem, were used to evaluate the performance of the SFO, which demonstrated high efficiency in converging on cases with 1 and 2 induced damages in the plate. Additionally, a new strain data dataset was developed for training and testing ML and ANN models to assess the efficiency of regression techniques. The results showed that both classification and regression performed at a high level, with approximately 100% accuracy in classifying the different quantities of damages. In the regression of a single damage, an average <InlineEquation ID="IEq1"> <EquationSource Format="TEX">\(R^2\)</EquationSource> <EquationSource Format="MATHML"><math> <msup> <mi>R</mi> <mn>2</mn> </msup> </math></EquationSource> </InlineEquation> of 99% was achieved for position parameters and 49% for its size. However, for multiple damages, especially in cases with 3 damages for SFO and 2 and 3 damages for ML models, the regression of parameters proved challenging, reflecting the inherent complexity of the problem addressed. This work also proposes a new dataset as a resource for future evaluations of regression efficiency, contributing to advancements in the field of structural integrity monitoring.</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

Strain-based identification of multiple damages in plate-like structures using artificial intelligence and metaheuristic optimization

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年2月3日

查看全文
  • Guilherme Ferreira Gomes,
  • Vinicius Dammenhain Takano

摘要

Monitoring the structural integrity of laminated composite materials, such as those used in aerospace structures, presents significant challenges due to the complexity of failure mechanisms like delaminations, which are often invisible on the surface. This study proposes an innovative methodology for identifying multiple delaminations using a combination of optimization techniques and artificial intelligence, addressing a highly complex problem in both fields. The Sunflower Optimization Algorithm (SFO) was explored to solve the inverse problem, while Machine Learning (ML) models and Artificial Neural Networks (ANN) were employed for classifying the number of delaminations and regressing their parameters. The damages were parameterized as ellipses in a square composite laminate plate modeled using finite elements. Strain data, obtained from the forward problem, were used to evaluate the performance of the SFO, which demonstrated high efficiency in converging on cases with 1 and 2 induced damages in the plate. Additionally, a new strain data dataset was developed for training and testing ML and ANN models to assess the efficiency of regression techniques. The results showed that both classification and regression performed at a high level, with approximately 100% accuracy in classifying the different quantities of damages. In the regression of a single damage, an average \(R^2\) R 2 of 99% was achieved for position parameters and 49% for its size. However, for multiple damages, especially in cases with 3 damages for SFO and 2 and 3 damages for ML models, the regression of parameters proved challenging, reflecting the inherent complexity of the problem addressed. This work also proposes a new dataset as a resource for future evaluations of regression efficiency, contributing to advancements in the field of structural integrity monitoring.