Structural damage classification in composite materials using the Wigner-Ville distribution and convolutional neural networks

Autor(es)

Monson, Paulo Monteiro de Carvalho ; Conceicao Junior, Pedro de Oliveira ; Dotto, Fabio Romano Lofrano ; de Aguiar, Paulo Roberto ; Rodrigues, Alessandro Roger ; David, Gabriel Augusto

• Tipo de documento: Artigo Científico
• Fonte: Materials Letters; v. 369, p. 4-pg., 2024-05-31.

Resumo

Detecting damage in composite materials is crucial and has received considerable attention in the field of machine learning. However, challenges remain in addressing backbone issues and classifying specific types of damage. This paper presents a novel deep-learning approach for automatically distinguishing delamination and microcracks in Carbon Fiber-Reinforced Polymer (CFRP). The methodology utilizes signals from piezoelectric transducers transformed into time-frequency representations based on the Wigner-Ville Distribution (WVD). Backbone issues were successfully addressed by transitioning the time series classification problem into a computer vision (CV) context through Convolutional Neural Networks (CNN). The analysis included a thorough examination of delamination and microcracks datasets produced experimentally by the National Aeronautics and Space Administration (NASA), focusing on these two types of damage. The proposed methodology achieved a precision range of 98.3 % to 100 % in damage classification, demonstrating its effectiveness for structural health monitoring in composite materials. (AU)

• Processo FAPESP: 23/02413-2 - Sistemas inteligentes para monitoramento e diagnóstico de falhas em ativos industriais baseados no conceito de indústria 4.0
• Beneficiário: Pedro de Oliveira Conceição Junior
• Modalidade de apoio: Auxílio à Pesquisa - Regular