PSI - Issue 52

Ilias N. Giannakeas et al. / Procedia Structural Integrity 52 (2024) 655–666 Author name / Structural Integrity Procedia 00 (2019) 000 – 000

657

3

Fig. 1:Illustration of the objectives for a SHM system. It is understood that the signals and the extracted damage related features cannot measure damage directly. Additional operations are required to translate the SHM indications to information relating to the physical damage (Axiom IVa in (Worden et al. 2007)). There has been considerable scientific effort in composites to link the Health Indicators (HI) extracted from the SHM system to estimations relating to the residual life or strength, utilizing various modelling approaches (Tao et al. 2017; Banerjee et al. 2019; Milanoski et al. 2022; Peng et al. 2015; Wang et al. 2018; Lee et al. 2022). In a previous study by the authors (Giannakeas et al. 2023), a framework was developed to estimate delamination damage size between the skin and the stringer of a stiffened composite panel using HI extracted from GWSHM systems. Due to the unavailability of a large experimental dataset, a multi-fidelity model was used that fuses numerical and experimental results. This allowed to characterize the damage and subsequently estimate the residual strength of the structure given a damage size. Although the methodology had an intrinsic stochastic character; it was assumed that perfect information was available. In this study the effect of uncertainty in the inputs on the estimation of the delamination size is studied. To estimate the delamination damage size, the methodology required information from both the detection and the localization modules. First, the detection module processes the signals to extract damage sensitive features and compute a HI. If the HI is above the threshold value, the status of the structure is indicated as faulty and the localization module provides an estimate of the possible damage location. Then the characterization module operates on the HI and location values to provide an estimate on the delamination size. Both the HI and the location however are prone to uncertainties that can affect the accuracy of their estimation. These inaccuracies are propagated to the characterization of the damage (i.e. size estimation) and subsequently will also be transferred to the residual life estimations. To understand the behaviour of the SHM system, it is imperative to both study the sources of these uncertainties and their effect on the subsequent estimations. Therefore in this study, the multi fidelity model presented in (Giannakeas et al. 2023) is utilized to propagate uncertainties associated with diagnostic and localization modules and study the response of the estimates for the delamination damage size. This contribution is structured as follows: in section 2 the methodology of the study is presented along with a brief description of the experimental and numerical campaigns using in (Giannakeas et al. 2023) to train the multi fidelity model. In section 3, the results from the uncertainty propagation study are presented. The effect of uncertainties arising from both the diagnostic and the prognostic module are included. Lastly, concluding remarks are included in section 4 2. Methodology 2.1. Signal Processing and Definition of SHM Health Indicator Interrogation of a structure using guided waves for damage detection and localization is carried out by comparing two signals, ̂ ( ) and ̂ ( ) that correspond to the signals recorded between the th pair of sensors at the pristine and current state, respectively. The existence of damage will change the propagation characteristics as well as produce reflections. To generate guided waves, a narrowband excitation has been selected due to the dispersive nature of wave propagation in layered composites. The excitation is given as: ( ) = 0.5 0 [ ( ) − ( − / )] sin(2 ) (1 − cos(2 / )) , (1) where, (∙) is the Heaviside function, =5 is the number of cycles and =50kHz is the central frequency. To reduce the effect of boundary reflections, a window function ( ) is applied as: ( ) = ̂ ( ) ( ) (2)