PSI - Issue 54

Vasiliki Panagiotopoulou et al. / Procedia Structural Integrity 54 (2024) 482–489 Vasiliki Panagiotopoulou/ Structural Integrity Procedia 00 (2023) 000 – 000

484

3

numerical models or on analytical and empirical models, where, in both cases, the presence of the damage is considered in the model to simulate the behavior of the structure in more realistic operative conditions. SAMAS 2 offers a significant advantage in that, once the sensor network is installed and the Digital-Twin model of the structure is rapidly self-updated, inspections become feasible without the need to disassemble parts. Furthermore, this system allows for continuous monitoring in real-time, enabling the detection of corrosion degradation and ballistic impact damage, as well as the prognosis of their impact on the overall reliability of the helicopter. The objective of this paper is to detail the ongoing developments within the SAMAS 2 project, with a particular focus on the critical areas requiring further advancement. The paper is structured as follows: Chapter 3 outlines the SHM system designed for ballistic impact monitoring, while Chapter 4 delves into the SHM system specifically designed for corrosion monitoring. Chapter 5 concludes by summarizing the preliminary work accomplished thus far and outlining forthcoming tasks. 3. SHMP system for ballistic impact monitoring The SHM method for ballistic impact monitoring proposed here combines simulated data derived from detailed Finite Element Models (FEM), experimental data collected from both full-scale and simplified versions, and intelligent algorithms capable of learning the underlying relationships between specific damage-sensitive parameters and accurate damage characterization. It's worth noting that it would be feasible to establish a damage detection system solely based on data-based algorithms capable of distinguishing between a healthy case and a critical one using predefined thresholds. However, it would be challenging to gain a physical understanding of the damage characteristics responsible for the alterations in the collected sensor data. Furthermore, it is essential to recognize that damage should not only be identified for its presence but also for its type, location, and extent in order to predict its criticality and establish an adaptive maintenance schedule. Hence, the introduction of FEM, as a source of simulated reference data, help the algorithms in associating a certain physical meaning to the generated alarm. Through simulations under various damage conditions, a comprehensive database can be established, encompassing diverse types of damage, different positions, and varying degrees of criticality. This database serves as a low-cost basic information upon which algorithms are trained to interpret real sensor data. Nonetheless, databases created using model-based approaches are susceptible to errors and biases due to the inherent assumptions and simplifications made in the modeling process. The inclusion of experimental data not only accelerates the algorithm training process, thereby enhancing its performance, but also accounts for unforeseen, unmodeled effects that may impact the system's dynamics. By superposing the numerical database with the experimental dataset, time-consuming experimental activities are limited to only cover the main operational scenarios yet considering environmental and operational parameters not including in simulations, while the database is enriched by simulated data, creating examples corresponding to extreme operational conditions of the helicopter. Thus, the output of this hybrid data-based and model-based system enables another family of category algorithms, namely the prognostic unit, to estimate the Residual Useful Life (RUL) of the monitored component. Obviously, the statistical propagation of the uncertainties relative to the diagnostic level inside the prognostic unit must be carefully considered, to keep the reliability of the whole system at the desired level. 3.1. Choice of damage sensitive parameters In the most general terms, damage to a structure is essentially any change that hampers its current or future performance. To identify damage, we need to compare two states of the system, one of which representing its original, and undamaged condition. In cases of impact damage, it alters the structure's geometry, affecting its stiffness and leading to mass loss. The harmful effects of this damage can occur immediately or gradually, depending on factors like projectile size, velocity, impact angle, location, and applied loads. Damage can accumulate over time, as seen in fatigue or corrosion damage, where cracks slowly form from notches and progressively spread. In the context of SAMAS 2, the objective is to establish a diagnostic and prognostic system based on changes in vibration signals resulting from ballistic impact events. Research by (J.-J. Sinou, A. W. Lees, 2007) and (J.-J. Sinou,