Issue 78

C. Vendittozzi et alii, Fracture and Structural Integrity, 78 (2026) 519-540; DOI: 10.3221/IGF-ESIS.78.30

Figure 4: Dataset A: Raw Bragg-wavelength vs. time histories from two of the four installed FBGs (S1 and S2). Abrupt vertical excursions and gaps indicate invalid interrogator samples, including dropouts and saturation artifacts, rather than physical strain excursions.

Figure 5: Dataset B: Raw Bragg-wavelength vs. time histories from two of the four installed FBGs (S1 and S2). Abrupt vertical excursions and gaps indicate invalid interrogator samples, including dropouts and saturation artifacts, rather than physical strain excursions. Normalized microstrain time histories and segmentation into dominant events After integrity screening, baseline normalization, wavelength-to-microstrain conversion, detrending, and high-pass filtering, the four channels exhibit a clearer event structure for both datasets (Figs. 6 and 7). This processed representation removes the visual dominance of invalid wavelength samples and makes the transient strain response suitable for response-based event segmentation . For consistent segmentation, candidate events were detected using the short-time energy proxy computed from the multichannel response, namely the moving average of the sum of squared microstrain across the four FBG channels over a 250 ms window. A minimum separation constraint was applied to prevent closely spaced peaks belonging to the same excitation from being treated as independent events. The three most energetic separated peaks were retained for each dataset and used to define event-centered windows spanning (-10) s to (+20) s around each candidate. These windows were then applied consistently in the subsequent time–frequency analysis and descriptor extraction. In Dataset A (Fig. 6), the three retained event candidates are visible as intervals of increased oscillation amplitude and pronounced inter-sensor divergence, followed by progressive reconvergence of the signals. In Dataset B (Fig. 7), three high energy event candidates are also identified; however, the post-event behavior differs from Dataset A, with a more persistent elevated response after the final event window. Because the segmentation is derived from the FBG signals themselves, Events 1–3 should be interpreted as algorithmically selected high-energy FBG-derived event candidates, not as independently validated touchdown instants or pre-classified operational phases. Their labeling reflects the adopted energy based selection and separation criteria. Response-based interpretation of the strain-response sequence (Dataset A vs Dataset B) The normalized strain responses shown in Figs. 6 and 7 support a response-based interpretation of the event sequence, derived from (i) clustering versus divergence of the four FBG channels and (ii) the three dominant energy peaks used to define the event-centered windows. Because synchronized avionics or independent operational markers, such as weight-on wheels, vertical speed, airspeed, braking command, acceleration, video, or flight-log data, were not available, the interpretation is intentionally restricted to strain-consistent response phases rather than uniquely identified maneuvers.

529