Damage detection of Carbon Fibre Reinforced Polymer (CFRP) tubes subjected to low-velocity transverse impact

Abstract

CFRP composites are extensively used in aerospace, mechanical, and civil infrastructure due to their high strength-to-weight and stiffness-to-weight ratios. However, they are vulnerable to impact-induced damage such as delamination, matrix cracking, and fiber breakage. This study presents a machine learning-based approach to classify damage in CFRP tubes subjected to low-velocity impact. A series of drop-hammer tests and validated numerical simulations were conducted to generate a comprehensive dataset comprising 81 samples, with an 80:20 train–test split for model development and evaluation. The dataset includes impact parameters — specifically, impact mass and drop height — as input features for training the models. The output consists of binary labels representing the presence or absence of five distinct damage types: compressive fiber damage, tensile fiber damage, compressive matrix damage, tensile matrix damage, and shear damage. Three supervised learning algorithms — Random Forest (RF), Decision Tree (DT), and Multi-Layer Perceptron (MLP) Neural Network — were employed for classification. The RF model achieved the highest classification accuracy (0.93), outperforming DT (0.89) and MLP (0.91). Precision–recall analysis showed excellent performance for most damage types, including perfect scores (AP = 1.00) for tensile fiber damage, compressive matrix damage, and shear damage. The results highlight the effectiveness of machine learning models in identifying impact-induced damage, enabling real-time structural health monitoring of CFRP tubes. © 2025 Elsevier B.V., All rights reserved.