PSI - Issue 77

Andrzej Katunin et al. / Procedia Structural Integrity 77 (2026) 18–25 Author name / Structural Integrity Procedia 00 (2026) 000–000

22

5

allowing for reliable identification and quantification of damage in laminated composite plates using the SHVT technique. More details for this algorithm and examples of its implementation can be found in (Amraei et al., 2025). The application of this algorithm resulted in the selection of the best raw thermograms for the considered scenarios, which are presented in Fig. 2.

(a)

(b)

(c)

Fig. 2. The best raw thermograms selected based on the developed algorithm for damage scenarios: (a) 10%, (b) 25%, (c) 50% of thickness reduction in the damaged area. The location of damage is recognizable in each case; however, the boundaries of damage are well-distinguishable only for the scenario with 50% of thickness reduction in the damaged area. 3.2. Enhancement algorithm To improve the detectability of boundaries of damaged areas and suppress measurement noise, which influencing proper identification of boundaries, and further quantification, the enhancement algorithm has been developed. During selection of a thermogram, the compromise between the sharpness of the boundaries and the amount of delivered energy, which influences the detectability of damage, needs to be reached. For this reason, the wavelet-based fusion was implemented according to the flowchart presented in Fig. 3. The implementation was performed in the MATLAB environment.

Fig. 3. The flowchart of the proposed enhancement algorithm.

To maintain sharpness of the edges of damage, a sequence of 5 thermograms was empirically selected, where the starting thermogram is the best raw thermogram selected based on the algorithm presented in section 3.1. Each thermogram in the sequence was subjected to filtering using non-local means denoising algorithm, which allows preserving edge detectability, while effectively removing noise from thermograms. Next, the fusion of pre-filtered thermograms was performed using 2D discrete wavelet transform (DWT) with the fusion criterion of energy maximization. For this purpose, the single-level DWT-based decomposition was performed and the maximal energy was computed for each set of approximation and detail coefficients. After this, the sets of coefficients were reconstructed using inverse DWT. For this study, the biorthogonal wavelet 3.5 was selected. The decomposition filter with 5 vanishing moments provides effective local approximation, while the reconstruction filter with 3 vanishing moments results in smoother output due to better energy concentration within the compact support. This, in turn, results in good compromise between edge preservation, distortion, and quality of noise removal. For improving