Advances of deep learning based ultrasonic nondestructive testing for CFRP structures: a review.

Abstract

Carbon fiber reinforced polymers (CFRP) exhibit exceptional properties such as low density, high strength, corrosion resistance, and design flexibility, making them widely employed in aerospace, automotive, and biomedical applications. However, various types of damage may occur during manufacturing, assembly, or service loading, potentially compromising structural integrity. Therefore, the development of efficient nondestructive testing (NDT) techniques is essential to ensure the reliability of CFRP-based systems. Ultrasonic NDT (UNDT) offers long-range detection, high sensitivity, and cost efficiency for damage evaluation, yet conventional UNDT methods rely heavily on expert interpretation and manual analysis, which limits their accuracy and efficiency. With the rapid progress of deep learning (DL), deep learning-based ultrasonic NDT (DL-UNDT) methods have emerged as a promising alternative, capable of automatically learning complex signal-damage relationships from ultrasound data. This paper provides a comprehensive review of recent advances in DL-UNDT for CFRP structures, summarizes key developments across damage detection, localization, and imaging tasks, and discusses prospective research directions for addressing current challenges.