Lamb waves based active sparse tomography for damage size quantification in composite structures : data-driven and parameter inversion methods

This thesis concerns the structural health monitoring of aeronautical composite structures (SHM) using guided waves generated by piezoelectric transducers. The general objective is to propose and validate experimentally several methods to quantify the size of a delamination type damage present in the structure to be monitored. The first contribution of this thesis is the creation and publication of an open access and reusable database of SHM measurements from fatigue tests of composite structures representative of aeronautics components. The second contribution is the development and validation on damage of unknown size of a purely data-driven quantification approach that consists in training a mathematical model from labeled data and an original mathematical descriptor computed from localization images. The third contribution is the development and validation of a quantification method that relies on the estimation of damage-related parameters from a physical model and does not require reference data. To solve this inverse problem, an analytical model is developed to predict the signal measured by a piezoelectric sensor at the passage of a wave generated by a piezoelectric actuator and scattered by a damage. This model is limited to quasi-isotropic composite material and has been validated by comparison to finite elements method simulations. Based on this theoretical model, an identification tool is proposed to estimate the damage size and severity. These two quantification methods, one based on data and the other on a physical model, have been validated on signals from numerical simulations and on experimental signals from structures with realistic damages.