STFT-CNN enabled quantitative detection of liquid ingress in honeycomb composites via infrared thermography

Honeycomb sandwich composites are widely used in various industries due to their exceptional properties, but they face the challenge of liquid infiltration due to their inherent hollow structure. This study proposes a novel method for quantifying liquid volume in honeycomb structures using pulsed thermography and a deep learning network. By combining the powerful image feature extraction capabilities of convolutional neural networks (CNNs) and the time-frequency analysis advantages of the short-time Fourier transform (STFT), the network effectively extracts spatiotemporal features from the data. Finite element simulation, theoretical analysis, and experimental validation demonstrate the effectiveness of the proposed method. In the near-liquid configuration, the proposed STFT-CNN model can predict liquid volumes up to about 22.5% of the full capacity. For far-liquid configuration, the model demonstrates excellent performance in predicting a wide range of liquid volumes, from very small amounts to near-full capacity of a honeycomb core. Our proposed method provides a valuable tool for monitoring the health and integrity of honeycomb structures.