Issue 78

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

Accordingly, the workflow should be considered a field-oriented processing approach for response-based event characterization under non-ideal acquisition conditions. Its extension to validated HUMS or structural-integrity decision making requires synchronized operational references, explicit temperature compensation, and a larger set of repeated measurements. Interpretation limits and recurrence across representative datasets Because synchronized avionics channels, such as weight-on-wheels, airspeed, vertical speed, braking command, accelerometers, video, or flight-log markers, were not available, the interpretation remains response-driven. Consequently, the comparison between Dataset A and Dataset B cannot be used to establish statistically validated repeatability, nor can it provide unambiguous labeling of the operational phase associated with each event candidate. Instead, the comparison should be interpreted as an assessment of whether comparable FBG-derived spectral descriptors can be recovered from two independent representative recordings acquired with the same sensor installation and processing workflow. The addition of quantitative descriptors makes this comparison more traceable than a purely visual assessment of spectrograms or ridge plots. Ridge median frequency, ridge dispersion, ridge persistence within selected frequency bands, and band-limited spectral RMS provide compact descriptions of which frequency families dominate each event-centered window and how stable those families remain over time. These metrics distinguish between windows characterized by a relatively persistent dominant family and windows in which the ridge switches between competing spectral contributions. However, these descriptors should be interpreted as event-level summaries of the measured FBG response, not as validated indicators of damage, landing severity, or health state. Within this framework, Dataset A is characterized by reconvergence of the microstrain traces after each event candidate and by clearer recovery of the pre-event trend. Its quantitative descriptors support a structured but comparatively more transient response, with low-end dominant content in S1–S2, a recurring family near 150 Hz in S3, and a high-frequency family near 429 Hz in S4, particularly in the earlier event-centered windows. Dataset B preserves a comparable sensor-dependent spectral organization, but it also exhibits stronger redistribution in the later part of the record. In particular, Dataset B shows a stable high-frequency family in S4 during Events 1 and 2, whereas Event 3 is associated with a shift of the ridge medians of S3 and S4 toward the low end of the analyzed band and with a marked increase in ridge dispersion for S4. This behavior indicates stronger competition among spectral contributions within the same event-centered window and a less stationary response evolution than that observed in Dataset A. The agreement between the two datasets is therefore not expressed as identical trajectories, identical event responses, or population-level repeatability. Rather, it is expressed by the recurrence of comparable sensor-dependent frequency regions within the analyzed band: low-end dominant content in S1–S2, a mid-frequency family near 150 Hz in S3, and a high-frequency family near 427–429 Hz in S4. Differences between datasets are expressed by changes in persistence, dispersion, and relative spectral weighting of these families across the event-centered windows. This interpretation also clarifies the limits of the ridge-based analysis. Around high-energy event transitions, pre-event and post-event response components may contribute to the same analysis frame because of the finite 2.048 s frame duration used in the main multitaper configuration. Under these conditions, the ridge should be interpreted as a compact dominant-frequency descriptor of the evolving response rather than as an instantaneous modal identification. Without independent modal data, finite-element correlation, accelerometers, or synchronized operational references, the observed frequency families cannot be assigned definitively to specific structural modes, operational sources, or damage states. Overall, the recurrence of comparable frequency families indicates that the descriptors support response-based comparison under field-acquired, non-ideal conditions. The use of multitaper estimation and quality-gated ridge extraction improves the traceability of this comparison by reducing sensitivity to leakage, residual noise, and repaired signal segments. Physical interpretation, event-level spectral summaries, and implications for future digital-engineering workflows The recurrence of specific ridge families across the two representative datasets suggests that the measured landing-gear response is not spectrally random, but structured into a limited number of dominant frequency regions within the analyzed band. As summarized in Table 4, low-end families near 10–20 Hz are primarily shared by S1 and S2, whereas S3 and S4 emphasize more localized mid- and high-frequency families near 150 Hz and 427–429 Hz, respectively. This distinction suggests that some components may reflect broader structural participation, while others remain more strongly dependent on sensor position and local load transfer. Shared families observed across multiple sensors may therefore be interpreted as candidate broader response features of the landing-gear structure, whereas sensor-specific families are consistent with local installation position, local stiffness distribution, strain-transfer conditions, and load-path sensitivity during transient loading. These interpretations should be regarded as physically plausible but not definitive. In the absence of an independent modal model, finite-element correlation, ground-vibration testing, accelerometers, synchronized avionics, or other operational ground-truth channels, the observed frequency families cannot be assigned unambiguously to specific structural modes,

537