Stochastic wavenumber estimation: Damage detection through simulated guided Lamb waves

Structural Health Monitoring (SHM) is a field of research that seeks to identify the onset of damage in infrastructure systems such that catastrophic failures can be averted. Acoustic Wavenumber Spectroscopy (AWS) is an emerging SHM technique that can identify and locate change of thickness resulting from damage in thin plates through the estimation of the characteristic wavenumber. AWS measures the ultrasonically induced vibrations of thin plates by laser scanners and monitors the propagation of Rayleigh-Lamb waves through the structure. This method is particularly suitable for thin-walled structures such as those contained in pipes, airplanes, and wind turbine blades.;While AWS measurements may successfully locate damage in a structure, the severity of damage cannot be quantified without the assistance of an accurate guided Lamb wave propagation (LWP) model. Given material properties of the structure (elastic modulus, density, and Poisson's ratio), the LWP model uses Lamb wave equations to relate the local wavenumber to the effective thickness. Reduction in the effective thickness of these structures is then used as an indicator of damage, for instance due to corrosion for metal structures or delamination of the internal layers for composite structures. Successful determination of thickness from the measurements using the LWP model relies on two aspects: uncertainties regarding material properties of the system (referred to herein as parametric uncertainty) and uncertainties regarding data collected in the field under less than ideal conditions (referred to herein as experimental uncertainty).;Current state of the art in AWS estimates wavenumber based on the maximum data fit of the wavenumber dispersion curve and derives the thickness deterministically through the Lamb wave equations. This deterministic technique often leads to large false positives due to the parametric and experimental uncertainties.;The focus of this thesis is to develop a stochastic approach for inferring thickness from the measurements in which both parametric and experimental uncertainties are accounted for, henceforth referred to as Bayesian Wavenumber Estimation. Herein, parametric uncertainty is dealt with by calibrating material properties using wavenumber measurements. Experimental uncertainty is dealt with by incorporating expert judgment through an elicited prior uncertainty of thickness. The technological advancement produced in this study is demonstrated on a case study application of an aluminum plate with imposed thinning.