Abstract
Carbon fiber reinforced polymers (CFRP) have been used as one of the options to strengthen steel structures through adhesive bonding, particularly in specific applications where traditional strengthening methods may not be suitable. Therefore, it becomes crucial to perform inspections on the resulting CFRP-steel adhesive structures (CSAS) to ensure their structural integrity and safety. However, the distinct physical properties of CFRP, epoxy resin, and steel pose significant challenges to accurately inspecting bonding interface defects of such special hybrid engineering structures. To address these challenges, a new approach, streamlined one-dimensional convolutional denoising autoencoder-low-power vibrothermography (SOCDAE-LVT), is proposed in this study to enhance the recognition of bonding interface defects within CSAS. This approach utilizes thermal signals from low-power vibrothermography (LVT) to enhance the recognizability of CSAS bonding interface defects. A low-power vibrothermography inspection system was developed to acquire thermal signals on the surface of CSAS samples. A streamlined one-dimensional convolutional denoising autoencoder (SOCDAE) model was designed for robust representation extraction of the thermal signal at each pixel point. The study further investigated the impact of different types of added noise and signal pre-processing approaches on the performance of the SOCDAE-LVT, aiming to optimize its effectiveness. By comparing qualitatively and quantitatively with the state-of-the-art approaches, the results show that the proposed approach can better improve the recognizability of defects. The enhanced recognizability of bonding interface defects enables accurate assessment of the quality of CSAS, thereby contributing to the safety of such structures.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call