Scholarlink

<p>In Structural Health Monitoring (SHM) of polymer laminates, precisely monitoring defects is crucial for long-term integrity. However, conventional SHM methods often lack explainability and transparency, limiting their reliability in practical damage detection. This research brings forth a hybrid deep learning framework for both accurate and interpretable damage detection in polymer composite structures. The proposed approach is validated on damage identification and quantification in a polymer composite. The effective identification of damages with limited data is achieved using hybrid methods, numerical simulation, and experimental analyses. The dataset includes various defect types such as pinhole, porosity, tarnish, blister, void, crack, and corrosion. The validation of damage of the Carbon Fibre Reinforced Polymer (CFRP) material is highly affected by porosity, showing 94% accurate deficiency. The outcome features are well matched with the defect related waveforms, pointing to the damage categorization is more accurate and reliable, displaying 1.02 dB and 1.99 dB amplitude of vibration before and after optimization. A multi-factor defect classification strategy is formulated based on their size, depth, structure, and severity to predict the defect more accurately. Moreover, the performance of LEOPDOC-AFFO-LIME module is validated on a actual idustrial dataset. The results highlighted the supreme performance of the network, even under data scarcity, and the integrated interpretability mechanisms significantly improve the transparency and credibility of the model’s predictions. Thus, this research offered a significant approach on damage identification, highlighting its potential over CFRP products.</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

Lightweight explainable photonic conviformer: a deep learning approach for reliable damage detection in polymer composites

  • 期刊论文

  • 发表日期：

    2025年11月18日

查看全文
  • P. Saravana Kumar

摘要

In Structural Health Monitoring (SHM) of polymer laminates, precisely monitoring defects is crucial for long-term integrity. However, conventional SHM methods often lack explainability and transparency, limiting their reliability in practical damage detection. This research brings forth a hybrid deep learning framework for both accurate and interpretable damage detection in polymer composite structures. The proposed approach is validated on damage identification and quantification in a polymer composite. The effective identification of damages with limited data is achieved using hybrid methods, numerical simulation, and experimental analyses. The dataset includes various defect types such as pinhole, porosity, tarnish, blister, void, crack, and corrosion. The validation of damage of the Carbon Fibre Reinforced Polymer (CFRP) material is highly affected by porosity, showing 94% accurate deficiency. The outcome features are well matched with the defect related waveforms, pointing to the damage categorization is more accurate and reliable, displaying 1.02 dB and 1.99 dB amplitude of vibration before and after optimization. A multi-factor defect classification strategy is formulated based on their size, depth, structure, and severity to predict the defect more accurately. Moreover, the performance of LEOPDOC-AFFO-LIME module is validated on a actual idustrial dataset. The results highlighted the supreme performance of the network, even under data scarcity, and the integrated interpretability mechanisms significantly improve the transparency and credibility of the model’s predictions. Thus, this research offered a significant approach on damage identification, highlighting its potential over CFRP products.