PSI - Issue 77

João Queirós et al. / Procedia Structural Integrity 77 (2026) 475–483

481

speckle noise is found in the highest spectral range. By viewing the raw phase map as a combination of signals with different frequency spectra, this method effectively isolates the damage-related signals using a technique analogous to a band-pass filter. This filter allows us to isolate the middle range of frequencies, revealing the subtle variations caused by internal damage. The process uses a two-step filtering approach. First, a low-pass filter with a higher cut-off frequency is applied to the raw phase map to eliminate the high-frequency speckle noise. Next, a second low-pass filter with a lower cut-off frequency is applied to remove both the high-frequency noise and the subtle, localized variations caused by internal damage. The next step is to subtract the second filtered map from the first, which effectively isolates the desired damage-related signals. These filters are applied in the spatial domain using the sine-cosine average filter (Aebischer and Waldner, 1999). This technique involves using a moving average window with different sizes and multiple passes to achieve the desired filtering effect. The final step involves summing up a sequence of acquired filtered phase maps acquired over time. This process is used to highlight damage-related signals and reduce background noise, thereby enhancing the overall SNR ratio. This same method was applied to the phase maps captured during the cooling stage of the plate specimen following the rectangular pulsed thermal excitation. The main steps of this post-processing technique are outlined in the flowchart in Figure 5. First, one of the phase maps is selected to define the cut-off frequencies for the band-pass filter. This filter is then applied to the sequence of acquired phase maps. Finally, the resulting filtered phase maps are summed to enhance the damage signatures.

Selecting one of the phase maps for analysis

Adjusting the cut-off of band-pass filter

Phase maps data

Applying the band-pass filter to the phase map sequence

Summing the sequence of the filtered phase maps

Fig. 5. Shearography data post-processing procedure.

5. Results and Discussion A full experimental campaign was conducted to optimize the thermal excitation, data acquisition settings, and post processing parameters. For thermography, the best results were obtained using 0.5 seconds of pulsed excitation (PT) and a sinusoidal excitation frequency of 1/8 Hz (LT). For post-processing, a fifth-order polynomial was used for PT signal reconstruction, and the best PCA-filtered thermograms were derived using components 4 through 10 for both PT and LT. For DS, the optimal results were achieved with a 10-second pulsed excitation duration, with the band pass filter implemented by applying 10 passes each of a 15×15 and an 85×85 moving average windows to the raw phase maps. The best image result from thermography is compared with shearography in Figure 6. From the different thermography measurements and the post-processing methodologies, we were able to identify the PT image phase as the one that led to the best damage identification of the two internal damages of the CFRP plate.