Issue 78

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

than independently validated touchdown detection. Time–frequency analysis provides a suitable framework for characterizing such signals [17] because it captures the evolution of spectral content across the event rather than compressing it into a single stationary spectrum. While the short-time Fourier transform (STFT) is commonly used for this purpose, classical single-window spectral estimates can be sensitive to spectral leakage and estimator variance when the required near real-time updates necessitate short analysis windows. The multitaper approach introduced by Thomson addresses these limitations by utilizing multiple orthogonal tapers (DPSS/Slepian sequences) to optimize the bias–variance trade-off while controlling leakage [18, 19]. Multitaper methods have been widely validated in challenging high-frequency transient contexts, demonstrating significantly improved stability relative to single-taper estimates [20]. Extensions and refinements have also been proposed to reduce curvature-related bias in multitaper spectral estimates, reinforcing the suitability of the approach when accurate spectral tracking is required [21]. Furthermore, multitaper concepts have been integrated into time–frequency distribution constructions for signals embedded in noise, providing additional motivation for their use in nonstationary settings [22]. Recent work has even proposed statistical frameworks based on state-space multitaper modeling, highlighting the ongoing methodological evolution in this field [23]. Recent literature emphasizes the advantages of multitaper spectral estimators, hybrid STFT–reassignment techniques, and synchrosqueezing transforms in stabilizing the analysis of noisy, multicomponent structural signals [24, 25]. In landing-gear monitoring, these tools are relevant because the quantities of interest are not only instantaneous strain amplitudes, but also recurrent frequency components, transient energy redistribution, and sensor-dependent spectral signatures. Contemporary reviews on landing-gear structural health monitoring likewise stress the importance of direct load monitoring and automated landing-event screening, while also highlighting the need for sensing frameworks that can capture short-duration, high energy events in a consistent and interpretable way [10, 15]. When time–frequency representations are employed operationally, a secondary challenge lies in extracting features that remain stable across events and are interpretable with respect to physical mechanisms. Ridge-based analysis addresses this problem by tracking concentrated energy paths in the time–frequency plane. Such paths may correspond to dominant structural or operational components, although their assignment to specific structural modes requires independent validation. Ridge extraction and segmentation strategies based on the reassignment vector have been proposed to detect and estimate AM/FM components while structuring the time– frequency plane into mode-related basins of attraction [26]. Reassignment itself is a well-established post-processing technique used to sharpen time–frequency representations and improve readability under the classical localization/interference trade-off [27]. Efficient reassignment implementations have also been proposed to reduce computational overhead and enhance robustness, which is critical for real-time processing [28]. In parallel, ridge estimation from spectrograms has been treated as a formal estimation problem for multicomponent signals, providing algorithmic support for feature extraction under noise and mode interference [29]. In this context, the present paper is framed as a field demonstration and event-characterization study, rather than as a validated damage-detection, health-state assessment, or hard-landing classification method. Four FBGs were installed on the main landing gear of an Erickson S-64 Skycrane operated by the Italian National Fire Brigade, and two representative field datasets were analyzed. The aim is to evaluate whether imperfect operational FBG recordings can be screened for data integrity, converted into normalized strain, and reduced to compact time–frequency descriptors using STFT, Thomson multitaper estimation, and ridge tracking. The extracted frequency families are therefore interpreted as candidate descriptors of landing-gear response under field conditions. Their attribution to specific structural modes, operational sources, damage states, or hard-landing severity requires future work with synchronized reference measurements, temperature compensation, and a larger statistical dataset. Airframe and operational profile he test article was an Erickson (Sikorsky) S-64 Skycrane, a heavy-lift rotary-wing aircraft equipped with a twin-bogie main landing gear (Fig. 1). The measurement campaign was carried out during routine operational activities, including taxi, approach, repeated touch-and-go maneuvers, roll-out, and braking on a paved surface. Environmental conditions were representative of temperate operations, and the activity was conducted under standard safety procedures without imposing dedicated test-specific flight constraints. The measurements were aimed at recording the operational strain response of the main landing gear during segments expected to include approach, ground-contact interaction, roll-out, and braking. Since no synchronized avionics channels or independent event markers were available during this campaign, the analysis presented in this paper is restricted to response-based event characterization. Accordingly, the detected events are interpreted as high-energy FBG-derived event candidates rather than as independently validated touchdown instants. The T M ATERIALS AND METHODS

521