Defects Clustering using Kohonen Networks during Ultrasonic Inspection

In Non Destructive Testing (NDT) of materials,the ultrasonic waves propagating in a structure are reflected or refracted from the presented discontinuities. The reached waves are received and converted to electrical signals containing informations about the internal defects. The characterization of those defects is an important task and the use of tools of signal processing gives an appreciated help in the decision making for the human operators. This work is a contribution for the defect characterization. The Neural Networks are used in order to classify reached signals allowing to distinguish between signals reflected from two different defects (cracks and inclusions) included in a welding.