Kompozitlerdeki hasar teşhisi için akustik emisyon (AE) sınıflandırmasında derin öğrenmenin uygulanması

Composites offer high strength and stiffness with lightweight, making them attractive for a wide range of engineering applications. However, complex fracture mechanics of composites leading high maintenance costs and safety concerns especially in safety-critical applications. In particular, some of damage modes such as matrix cracking, delamination and fiber breakage may cause critical failures in specific applications. Therefore, development of effective and reliable Structural Health Monitoring (SHM) systems which are capable of detecting and identifying damage is necessary. Acoustic Emission (AE) is particularly suitable to be used in SHM systems due to its passive nature which enables in-situ and in real-time monitoring. Additionally, deep learning-based methods such as Convolutional Neural Networks (CNN) improved the state-of-the-art in image classification tasks and are used in different domains successfully. In contrast, a little attention has been paid to apply this method on characterization of damage in composites. In this study, CNN based classification scheme is used to classify AE measurements and dependence of classification performance on different propagation distances is investigated experimentally. Different damage modes are represented by the dominant fundamental Lamb wave modes which are generated using artificial AE source. The generated labelled waveforms are measured by sensors mounted at different distances to AE source. Two different architectures of CNN model are used to classify AE measurements and the performance of CNN models is evaluated. Finally, investigation of transfer learning concept for the sensors mounted at different positions is conducted.