Development of innovative procedures and algorithms for non-destructive quantitative evaluation of damage in materials for aeronautics by means of active thermography techniques

The aim of this thesis is to develop innovative procedures for quantitatively evaluating defects or, more in general, damage in materials. These procedures will be outlined to reduce both the experimental tests and the computational time with the final goal of promoting active thermographic techniques and more in general, Non-destructive Testing (NDT) in industrial applications. Therefore, the collaboration with Diagnostic Engineering Solutions (DES) S.r.l, a spin-off of the Polytechnic of Bari, has been fundamental for customizing suitable procedures to satisfy the needs of the company. In this regard, all the thermography techniques, such as lock-in, pulsed, and step have been considered for inspecting and characterizing the defects/damage within a wide range of materials by adopting both traditional methods and artificial intelligence algorithms (machine and deep learning). More in detail, a one-dimensional convolutive neural network (CNN) has been adopted and thermal temporal data has been provided as an input to train the net. The adoption of a deep learning algorithm has been useful in achieving a double goal: characterizing the defect and speeding up the test phase for inspecting the component. The powerful but also the limits of using the artificial intelligence approach with respect to traditional techniques will be critically discussed.