Issue 78

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

1. Spectral centroid, spectral kurtosis, inter-sensor coherence, and asymmetry indices were not used for decision-making in the present study and are therefore not reported as quantitative results. These metrics remain potentially useful for future HUMS-oriented implementations, but their interpretation requires a larger number of independently labeled events and synchronized operational references. To avoid introducing unsupported results, they are not included in the present analysis. FBG-derived event segmentation and definition of analysis windows Event segmentation was required to identify high-energy portions of the FBG response and to apply the same analysis procedure consistently to both representative datasets. Candidate events were detected from a short-time energy proxy computed from the multichannel microstrain response, rather than from independently synchronized touchdown markers. The energy proxy was computed as the moving average, over a 250 ms window, of the sum of squared microstrain across the four FBG channels: , j   is the microstrain signal measured by sensor (j). Peaks in   E n were used to identify candidate high-energy events. A minimum separation constraint was applied to avoid counting closely spaced peaks belonging to the same excitation as independent events. For each dataset, the three most energetic separated peaks were retained and used to define event-centered analysis windows. Because the segmentation is derived from the FBG response itself, the detected instants are not claimed as independently validated touchdown times. They are treated as FBG-derived event candidates consistent with high-dynamic landing-gear response. The time–frequency response was then examined around these candidates to assess changes in ridge trajectories and the appearance or redistribution of dominant frequency families. This procedure provides internal consistency between time-domain energy and time–frequency response, but it does not replace external validation by synchronized avionics, weight-on-wheels signals, accelerometers, video recordings, or flight-log data. For each detected event candidate, analysis was performed on a fixed window extending from (-10) s to (+20) s around the detected peak. This interval was selected to include the pre-event response, the high-energy transition, and the early post event evolution. In addition to frame-based spectrogram analysis, compact spectral summaries were computed per event centered window using power spectra (periodograms), with the window duration capped at 60 s for numerical stability and consistent frequency resolution. These spectra were used as compact event-level frequency-domain summaries, not as independent validation of the detected events or as proof of statistical repeatability. For reproducibility, the main acquisition, preprocessing, segmentation, and time–frequency analysis parameters used in the workflow are summarized in Table 1.           1 2 2 2 2 1, 2, 3, 4, 0 1 W q W E n [ n q n q n q n q ]                    (9) where   E n is the multichannel short-time energy proxy, (W=250) samples at 1.0 fs  kHz, and

Workflow item

Parameter / criterion Tecnobis Aerogator

Optical interrogator

Sampling rate

f s = 1.0 kHz

Four FBGs in series: S1 = 1530 nm, S2 = 1550 nm, S3 = 1540 nm, S4 = 1570 nm

FBG array

Optical fiber

Single-mode 9/125 µm polyimide-jacketed fiber Native wavelength time series; no resampling 1500–1600 nm; values ≤ 0 nm excluded

Raw-data handling

Valid wavelength interval

Spike suppression Short-gap treatment

Short-window median filtering of wavelength first differences; width 3–5 samples

Interpolation of gaps up to 10 samples

Longer gaps

Retained as excluded intervals Wavelength-to-strain conversion using nominal p e ൎ 0.22 Microstrain, ( ) Detrending and 0.5 Hz high-pass filtering using a fourth-order IIR filter Median wavelength over an initial quiet interval

Baseline definition Strain conversion

Strain output Drift handling

No direct temperature measurement or explicit temperature compensation; low-frequency and quasi static trends were not interpreted quantitatively

Temperature handling

527