Enhanced quantitative detection of delamination defects in large-tow carbon fiber composites using pulsed thermography

Large-tow carbon fiber fabrics generate considerable structured background texture noise in infrared thermographic inspections due to their nonuniform thermal properties. The presence of such noise poses notable challenges to the reliable and precise detection of defects. To address this issue, this article proposes a defect detection method that integrates the frame averaging difference (FAD) algorithm and principal component analysis (PCA). The method utilizes FAD to remove nontime-varying noise, improving the contrast-to-noise ratio of defect signals. It then employs PCA to separate background texture noise. Building upon this foundation, a region growing segmentation (RGS) algorithm based on fuzzy affinity (FA) was introduced. The algorithm integrates a spatial distance function based on both pixel intensity and spatial proximity, significantly enhancing the accuracy of defect region segmentation. To validate the effectiveness of the proposed method, artificial delamination defects were simulated by embedding Teflon films between layers. Large-tow carbon fiber fabric composite laminates were then fabricated as test samples. The experimental results demonstrate that the algorithm effectively reduces the error rate in defect detection for large-tow carbon fiber composite laminates. This study provides a practical technical solution for the nondestructive testing of large-tow carbon fiber composites and shows strong applicability and robustness under complex background noise conditions.