Deep learning enhanced metamodel design based on reduced mode shapes for delamination identification in composite structures

Abstract

Structures requiring robust maintenance and accessibility pose challenges for structural integrity monitoring. This monitoring, crucial in various engineering applications, enables corrective actions to ensure machinery and equipment safety. Laminated composite materials, prevalent in aeronautical structures, often exhibit imperceptible delaminations or cracks, demanding the use of artificial intelligence for damage characterization. This study employs a numerical-experimental methodology, integrating artificial neural networks, to identify delaminations. Delamination, a structural damage form, was intentionally induced in epoxy-fiberglass composite structures. Modal tests captured characteristics in damaged and undamaged structures, revealing artificial intelligence effectiveness in pinpointing damage location and size. Damage significantly altered beam mode shapes, generating complex modes. Induced delaminations led to a measurable 15% reduction in natural frequencies, and damping ratios increased by approximately 10% in damaged structures. Artificial neural networks achieved an impressive 92% accuracy in predicting delamination location and size. Quantitative analysis of mode shape alterations validated the proposed methodology’s efficacy. This study provides valuable quantitative insights into damage detection in laminated composite structures, highlighting AI's potential for enhanced structural health monitoring. The numerical–experimental approach, supported by quantitative results, emphasizes the importance of advanced techniques for precise and efficient damage characterization in critical engineering applications.