Efficient background removal and defect identification in infrared thermography of CFRP composites using adaptive fixed-rank kriging

Abstract Carbon fiber reinforced polymers (CFRPs) perform safety-critical functions in advanced engineering applications. However, the structural integrity of CFRP plays a crucial role in ensuring operational safety. This study presents an efficient and robust framework for defect detection in infrared thermography of CFRP composites based on adaptive fixed-rank kriging (AutoFRK). The method models the thermal field as a low-rank spatial process, effectively separating diffusion-consistent background patterns from localized defect-induced anomalies with minimal manual tuning. The residuals obtained from AutoFRK are further analyzed by principal component thermography (PCT) for enhanced defect visualization. Experimental validation on CFRP specimens demonstrates that the proposed AutoFRK–PCT approach achieves clear and stable defect identification across various depths and model resolutions, while significantly reducing computation time compared to multi-dimensional ensemble empirical mode decomposition and physics-informed neural network methods. The results highlight AutoFRK’s potential as a practical, interpretable, and scalable tool for real-time thermographic inspection.