Understanding the deterioration and predominant damage mechanisms of structures is highly relevant, especially for safety-critical components. Non-Destructive Testing (NDT) plays a crucial role in assessing and monitoring their integrity by evaluating damage evolution. However, when it comes to complex structures, the existing NDT methods face challenges in their application, as is the case with sandwich structures. This study employs two NDT methods to analyze the initiation of damage in such structures during a fatigue test. The Acoustic Emission (AE) technique utilizes membrane-free microphones with a broad bandwidth to capture acoustic events from difficult-to-access areas. A machine learning algorithm is used to classify these events to determine their source and associated damage mechanisms. Additionally, Digital Image Correlation (DIC) is employed to measure strain evolution without contact during the test. This method is particularly relevant due to the complex and thin geometry of sandwich structures, where other methods are not applicable. Strain redistributions are considered relevant damage indicators. The results indicate that Acoustic Emission serves as an early indicator of damage, with the cumulative number of events and peak frequency correlating well with the severity of the damage. In contrast, DIC revealed clear indications of damage or deterioration, albeit at a later stage compared to AE.

Aircraft heat exchangers play a crucial role in maintaining thermal balance and ensuring that essential components operate efficiently and safely. In this context, it is highly relevant to gain knowledge about the deterioration and the predominant damage mechanisms of these components. In this study, two different non-destructive methods are employed to analyze the damage initiation of a heat exchanger part during fatigue test. On the one hand, Acoustic Emission technique is employed using membrane-free microphones capable of capturing a broad bandwidth. The acoustic events were classified by a machine learning algorithm to determine their source and damage mechanisms. On the other side, Digital Image Correlation (DIC) allowed the measurement of the strain evolution along the test and more precisely, redistributions were considered as damage indicators. As a result, it was identified that AE served as early damage indicator as the cumulative number of events was in good agreement with the severity of the damage. With respect to classification two clear clusters ascribed to different type of events were identified. In the case of DIC, strain redistributions gave clear indications of damage or deterioration but at a later stage compared to AE