Adaptive wavenumber variation mapping for full-field ultrasonic wavefield-based damage detection

This study introduces a data-driven damage-detection method that leverages full-field, steady-state ultrasonic wavefields measured using a scanning laser Doppler vibrometer. Recent work has shown that steady-state wavefields can be used to estimate local wavenumber variations for damage detection. However, such approaches have rarely been utilized on tapered or geometrically nonuniform structures, where spatial variations in stiffness induce nonstationary wave propagation. Thus, practical implementation remains hindered by direction-dependent wave dispersion, mode coupling, and sensitivity to parameter selection. To address these challenges, we propose a data-driven wavenumber-analysis framework tailored for complex geometries. First, dominant wavenumber components across multiple modes are analyzed from a reference dataset to design a band-stop filter whose rejection bandwidth varies adaptively with the propagation direction and wave mode. Second, a local window-based wavenumber analysis is performed, and a damage index quantifying deviations from the reference wavenumber spectrum is introduced to achieve spatially resolved and noise-robust damage detection. Third, the behavior of steady-state wavefields in tapered structures is characterized, with the findings demonstrating that the proposed data-driven approach remains effective despite spatial variations in dispersion induced by geometric nonuniformity. The proposed method was experimentally validated on a mock-up unmanned aerial vehicle horizontal stabilizer and a linearly tapered aluminum plate. It achieved reliable damage detection in both isotropic and anisotropic structures with varying geometric profiles.