The present thesis aims to damage detection of composite aeronautical materials with the phased array ultrasonic testing (PAUT) and data analysis artificial intelligence (AI). First, there is an introduction to the source of the project's inspiration and its scientific significance. Secondly, there is a reference on composite materials, how they are used in aeronautical applications, and the damage mechanisms that take place. At that point, after briefly examining some standard non-destructive testing (NDT) methods, the selected technique and its capabilities are analyzed in depth, both at a theoretical and a scientific level. Next, the concept of Artificial Intelligence, Machine and Deep Learning is introduced, and the methods of "Supervised Learning" and "Unsupervised Learning" are analyzed in detail. More specifically, the clustering algorithm K-means and the Deep Neural Networks (DNNs) with emphasis on the convolutional neural networks (CNNs) are examined. Then, after a literature review of the scientific efforts made concerning the analysis of ultrasound data using artificial intelligence, the description of the project is presented, followed by the forming of the research questions and the methodology that was applied in the following research. Subsequently, the experimental setup (materials, equipment), the way of capturing the data, and the pre-processing and post-processing methods are described. Next, the unsupervised machine learning method used for binary classification for two different composites is analyzed, and the respective results are presented. Furthermore, the supervised deep learning method used for binary classification is analyzed, and the relevant results are presented. Finally, the conclusions that emerge from this analysis are presented, and suggestions for improvement and expansion of the work in the future are reported.
