Análise de agrupamento de sinais obtidos por simulação numérica do ensaio ultrassônico para reconhecimento de defeitos de soldagem

Abstract

The non-destructive ultrasound test is widely used to detect and size up discontinuities embedded a material. The TOFD (Time Of Flight Diffraction) was an ultrasound based technique used. However, the classification of the type of the discontinuity from the ultrasound signals acquired from an inspection requires a highly skilled professional. This task can be done by using patter recognition algorithms, which are able to quickly process a great amount of data. In this present work, three types of discontinuities (lack of penetration, porosity and crack) were simulated using the Finite Volume Method (FVM). After that, the method of pattern recognition K- nearest neighbor (K-NN) was used, to verify how these data are divided. This work is an initial step for the use of numerical simulation of the ultrasound simulation to generate simulated signs and use them in the training section of pattern recognition tools that which, in the future, can be tested with experimental ultrasound signals.Obtaining an average success rate of approximately 84%.