Issue 78

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

observation should not be interpreted as statistical repeatability over multiple landings, nor as validation of independently identified operational phases. Differences in relative peak amplitude are expected and are consistent with event-to-event variability in excitation intensity, interaction duration, local strain transfer, and redistribution of dominant spectral content within the same event-centered interval. Comparison between Dataset A and Dataset B shows that some dominant peaks occur in comparable frequency regions, supporting the presence of shared spectral families across the two representative recordings. At the same time, differences in peak prominence and spectral redistribution indicate that the event responses are not identical and remain dependent on the specific response evolution captured in each record. Accordingly, the per event periodograms should be interpreted as compact event-level frequency-domain summaries that complement ridge based tracking, rather than as standalone fingerprints for event classification, modal identification, or structural-integrity decision-making.

Figure 13: Dataset A: Per-event power spectra (periodograms) for sensor S3. Panels (a–c) correspond to Events 1–3 using FBG-derived event-centered windows ( − 10 s to +20 s) trimmed to a maximum duration of 60 s. (a) Event 1; (b) Event 2; (c) Event 3.

Figure 14: Dataset B: Per-event power spectra (periodograms) for sensor S3. Panels (a–c) correspond to Events 1–3 using FBG-derived event-centered windows ( − 10 s to +20 s) trimmed to a maximum duration of 60 s. (a) Event 1; (b) Event 2; (c) Event 3.

D ISCUSSION

Field tolerance of the workflow under acquisition imperfections he results indicate that operational FBG recordings acquired on an in-service helicopter landing gear can be transformed into interpretable time–frequency descriptors even when the raw Bragg-wavelength histories contain localized corruption. This observation should be interpreted as case-study evidence of workflow tolerance to field acquisition imperfections, rather than as a statistically validated demonstration of robustness over a population of flights or landings. The usable response descriptors obtained from the two representative datasets rely on the combined application of integrity screening, baseline normalization, wavelength-to-microstrain conversion, detrending, high-pass filtering, and frame-level quality control before descriptor extraction. The integrity-screening procedure was essential to prevent corrupted wavelength samples from driving the subsequent descriptor extraction. After wavelength-to-microstrain conversion, detrending and high-pass filtering reduce slow drift and provide a more stable signal domain in which high energy FBG-derived event candidates can be segmented using the multichannel strain-energy proxy. However, this filtering step does not constitute temperature compensation and does not allow quantitative interpretation of long-duration or quasi static strain trends. Time–frequency maps, ridge trajectories, and per-event spectra then condense the nonstationary response into compact descriptors of the dominant spectral organization within the event-centered windows. These descriptors enable comparison of sensor-dependent frequency families across the two analyzed records, but they do not provide independent touchdown identification, modal validation, damage detection, or health-state assessment. T

536