Issue 78

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

Workflow item

Parameter / criterion

Frame-level SNR gate Maximum interpolated fraction per frame

SNR ≥ 10 dB

≤ 5%

Main Hann STFT FAST Hann STFT

(L = 2048), (H = 512), 75% overlap (L = 1024), (H = 256), 75% overlap (L = 2048), (H = 512), (NW = 3), (K = 5)

Thomson multitaper estimate

Main analysis band Descriptor bands Event-energy proxy Event-energy window

9–500 Hz

20–60 Hz, 60–120 Hz, 120–220 Hz, 350–500 Hz

Moving average of the sum of squared microstrain across the four FBG channels

250 ms

Event selection

Three most energetic separated event candidates retained for each representative dataset

Event-centered analysis window

(-10) s to (+20) s around each FBG-derived event candidate

Per-event spectra

Two-dimensional periodograms computed over event-centered windows; duration capped at 60 s Ridge median frequency, ridge interquartile range, ridge persistence, band-limited RMS Spectral centroid, spectral kurtosis, coherence, and asymmetry indices were not used for decision making in this study Response-based event segmentation only; no synchronized avionics, accelerometers, video records, weight-on-wheels signals, or braking-command references Field demonstration and event-characterization workflow; not a validated damage-detection, health state assessment, or hard-landing classification method

Main descriptors reported Metrics not used in the present results

Validation status

Interpretation scope

Table 1: Main acquisition, preprocessing, event-segmentation, and time–frequency analysis parameters used in the workflow The parameters summarized in Table 1 define the complete workflow used for the two representative datasets analyzed in this study. The event candidates are derived from the FBG response itself and should therefore not be interpreted as independently validated touchdown instants.

R ESULTS

Raw wavelength measurements and data integrity igs. 4 and 5 report the raw Bragg-wavelength time histories from two of the four FBG channels, S1 and S2, for Dataset A and Dataset B, respectively. These two records correspond to the representative field datasets analyzed in this study: Dataset A has a duration of approximately 130 s, whereas Dataset B has a duration of approximately 70 s. They are used here to illustrate the data-quality conditions encountered during field acquisition and to demonstrate the subsequent processing workflow, not to provide population-level statistics of landing events. In both recordings, the traces contain short discontinuities, saturated segments, and dropouts that appear as abrupt vertical excursions and gaps. These portions are interpreted as invalid interrogator samples, associated for example with temporary loss of optical return or saturation during acquisition under highly dynamic conditions, and should not be interpreted as physical strain excursions of the landing gear. Instead, they motivate the integrity screening, short-gap handling, and frame-level exclusion criteria used before strain conversion and spectral analysis. To preserve transparency about acquisition quality, the wavelength axis in Figs. 4 and 5 is intentionally kept wide, so that corrupted samples remain visible rather than being hidden by axis zooming. On this scale, the physically meaningful Bragg-wavelength variations around the nominal sensor wavelengths are not visually apparent because they are much smaller than the excursions produced by invalid points. After integrity screening, baseline normalization, wavelength-to-microstrain conversion, detrending, and high-pass filtering, the signals are brought to a scale suitable for response-based event segmentation and time–frequency descriptor extraction. Sensor color coding is as follows: S1 (green, approximately 1530 nm), S2 (red, approximately 1550 nm), S3 (blue, approximately 1540 nm), and S4 (black, approximately 1570 nm). F

528