Quantitative infrared detection methods for debonding in concrete-filled steel tubes during the hydration heat phase

Abstract

Conducting rapid quantitative debonding detection after the completion of Concrete-filled steel tube (CFST) pouring is crucial for the timely identification and repair of potential structural issues. Currently, there is a lack of an infrared detection method that can perform quantitative detection specifically during the construction phase of CFST. This study proposes a Discreteness-Based Image Preprocessing (DBIP) method, combined with Otsu’s and K-means image segmentation methods, to explore its effectiveness in detecting debonding in CFST during the hydration heat phase. A full-scale CFST model was used to simulate debonding areas of different sizes, and infrared thermal imaging data were collected. The results show that the DBIP method significantly improved detection accuracy, and the DBIP+K-means combination can effectively quantify debonding areas with a minimum side length of 126 mm (10 % debonding rate). The study also reveals that the correlation between the F1-score and thermal contrast is linear when the thermal contrast is between 0°C and 0.18°C. When the thermal contrast exceeds 0.18°C, the F1-score stabilizes at approximately 0.8. The finding clarifies the detection accuracy range under different thermal contrast conditions and suggests potential optimization directions for the quantification of CFST debonding in practical application via infrared thermography.