From time–frequency decomposition to compressed signatures: an integrated approach for subsurface defect detection in composite materials

This paper presents an integrated signal‑processing method for the identification and quantitative feature-based evaluation of the subsurface defects in the carbon fibre reinforced polymer composites. Artificial defects, including inclusions, voids and rectangular cut, are systematically introduced under controlled experimental conditions, to assess the effectiveness of the proposed framework. Data acquisition using linear frequency‑modulated thermal‑wave imaging begins the workflow, providing input to the processing framework. Baseline correction followed by Coiflet 5 wavelet decomposition is applied to suppress high‑frequency fluctuations while preserving relevant thermal properties of defects. The pre-processed signals are then pulse-compressed to improve temporal localisation. Further, Hilbert-based processing is employed to reduce oscillatory components and enhance the interpretation of the defect signatures. A comprehensive quantitative analysis is performed using various features like signal-to-noise ratio, contrast-to-noise ratio, peak side-lobe ratio, spectral entropy, kurtosis and skewness. Results demonstrate significant improvement in signal localisation, contrast enhancement and statistical separability between sound and defective locations. The effect of defect parameters (size, depth) and metallic backing conditions on the thermal-response properties is also investigated and discussed in detail. The proposed study serves as a controlled experimental testing with artificial defects, while validation on real defect scenarios and probabilistic evaluation methods are identified as important directions for future work.