A Model-Based Damage Identification using Guided Ultrasonic Wave Propagation in Fiber Metal Laminates

Abstract

Fiber metal laminates (FML) are lightweight hybrid structural materials that combine the ductile properties of metal with high specific stiffness of fiber reinforced plastics. These advantages led to a dramatic increase in such materials for aeronautical structures over the last few years. One of the most common and vulnerable defects in FML is impact-related delamination, often invisible to the human eye. Guided ultrasonic waves (GUW) show high potential for monitoring structural integrity and damage detection in thin-walled structures by using the physical phenomena of wave propagation interacting with the defects [1]. The focus of this research project is on describing an inverse solution for the detection and characterization of defect in FML. Model-based damage analysis utilizes an accurate finite element model (FEM) of GUW interaction with the damage. The FEM is developed by project partners from mechanics at Helmut-Schmidt-University in Hamburg, Germany, and will be treated as a black-box for further analysis. A Bayesian approach (Markov chain Monte Carlo) is employed to characterize the damage and quantify its uncertainties. This inference problem in a stochastic framework requires a very large number of forward solves. Therefore, a profound investigation is carried out on different reduced-order modeling (ROM) methods in order to apply a suitable technique that significantly improves the computational efficiency. The proposed method is well illustrated on a simpler case study for the damage detection, localization and characterization using 2D elastic wave equation. The damage in this case is modeled as a reduction in the wave propagation velocity. The inference problem utilizes a parameterized projection-based ROM coupled with a surrogate model [2] instead of the underlying highdimensional model. This research is funded by the Deutsche Forschungsgemeinschaft Research Unit 3022 under grant LO1436/12-1.REFERENCES [1] R. Lammering, U. Gabbert, M. Sinapius, T. Schuster, P. Wierach (Eds)(2018) Lamb-Wave Based Structural Health Monitoring in Polymer Composites, Springer International Publishing. [2] Paul-Dubois-Taine A, Amsallem D. An adaptive and efficient greedy procedure for the optimal training of parametric reduced-order models. International Journal for Numerical Methods in Engineering 2014.