Deep Learning-Based Depth Estimation of Flat Bottom Holes in CFRP Using Active Infrared Thermography

It is crucial to accurately detect hidden defects in the fibers of carbon fiber-reinforced polymer (CFRP) composites when these materials are used in the aerospace or automotive industries. Active infrared thermography (AIT) is a suggested nondestructive evaluation (NDE) method that is contactless, quick, and provides full-field detection, making it capable of identifying such defects. Unfortunately, traditional analytical methods struggle to precisely determine defect depths due to the complex thermal behavior of CFRP materials. Deep learning models, such as Convolutional Neural Networks (CNNs) and Gated Recurrent Units (GRUs), have been successfully employed to interpret thermal responses and characterize defects, as reported in existing literature. In this study, we introduce a deep learning framework to estimate flat-bottom hole (FBH) depth in CFRP composites from simulated active flash thermography thermal data. Finite Element Method (FEM) simulations were conducted to generate time-temperature profiles for FBHs of various sizes and depths. To enhance the model’s robustness, the dataset was augmented with synthetic variations. A 1D-CNN architecture was trained on this dataset to analyze the temporal patterns in thermal decay signals. The model’s performance was evaluated on a reserved test set using mean absolute error (MAE) and root mean square error (RMSE), showing promising accuracy in predicting defect depth. The findings demonstrate the potential of combining deep learning and AIT to precisely locate defects in CFRP materials. This approach ultimately aims to advance NDE, improve structural health monitoring, and enhance maintenance workflows and strategies.