PSI - Issue 77

H. Lopes et al. / Procedia Structural Integrity 77 (2026) 673–680 H. Lopes/ Structural Integrity Procedia 00 (2026) 000 – 000

678

6

with a normalized cutoff frequency of 0.0102 is obtained by applying a moving average with kernel (13, 13) and 12 multi-passes to the raw phase map . On the other hand, the filtered phase map Δ with a normalized cutoff frequency of 0.0043 is obtained by applying a moving average of (25, 25) and 18 multi-passes to the same raw phase map. Finally, these two phase maps are subtracted to obtain a filtered phase map with a normalized band-pass between 0.0043 and 0.0102. In contrast to the raw phase map and the other two filtered phase maps, the band-pass filtered phase maps clearly show the two impact locations and damaged regions. The visualization of damage signatures can be further enhanced by adding multiple band-filtered phase maps. This technique enables the precise separation of signatures from background noise, thereby improving the signal-to-noise ratio (SNR). 3. Results and Discussion This section reports the results obtained by summing the band-pass filtered phase maps for various shearing amounts and directions. Figure 4 specifically presents the results of incrementally adding these maps, which were acquired during the plate's cooling stage. The data shown corresponds to a 45-degree shearing direction, with 5 mm shearing amount in both the x and y directions. Initially, at the 8 s time instant (Figure 4(a)), the two impact damages are not identified due to the high level of noise. This happens because at the beginning of the plate's cooling stage, the steep gradient of the strain field leads to speckle decorrelation, resulting in a high noise level within the phase maps. By adding the filtered phase maps of four consecutive instants (Figure 4(b)), the damage signatures become more visible, particularly the signature of the higher-energy impact damage located in the bottom right corner of the plate. The two damage signatures become most visible at the 48 s instant (Figure 4(c)), where the impact location and the damage region of the higher impact are clearly identified. However, this image shows some artifacts related to variations in the air's refractive index, likely produced by vertical convection currents. Finally, the image generated by the summation of 12 filtered phase maps (Figure 4(d)) clearly reveals both damage locations and sizes, confirming a correlation between damage size and impact energy. This result validates the ability of the method to isolate persistent damage signatures from random background noise, thus leading to superior damage identification.

iltered phase map Δ ℎ ℎ

=Δ ℎ ℎ −Δ

aw phase map

Fig. 3. Procedure for obtaining the band-pass filtered phase map.

Information classification: Internal