Automatic procedure for thermal NDE of delaminations in CFRP by using neural networks

This work is a first step in detecting and characterizing defects in an automatic way by using artificial intelligence. Transient thermal NDE by IR thermography is the method used for such a purpose. Data are processed by Neural Networks. Objectives: 1) avoiding participation of an expert operator in selecting a reference zone (non-defect area) 2) improvement in dealing with the problem of uneven heating (both the different absorptivity and the non uniformity of the heating source) 3) taking into account the case of a strong 2D diffusion, as it is for CFRP where the thermal diffusivity along the carbon fiber texture (generally parallel to the surface of the object) assumes values two times greater than the in depth diffusivity.