Caracterização de defeitos de soldagem através da análise fractal de sinais ultra-sônicos obtidos pela técnica TOFD

Abstract

The present work aims to characterize welding defects (lack of fusion, incomplete penetration and porosity) inserted in specimens during the welding process, by means of the fractal analysis of ultrasonic signals obtained by using the TOFD technique. In order to eliminate the noise, the signals have been treated by two different filtering methods, namely, band pass and Savitzky-Golay. In the fractal characterization of the signals, three methods have been used: Hurst’s analysis, detrended fluctuation analysis and minimal cover analysis. In all three methods, filtered isolate signals and filtered concatenate signals have been considered. The characteristic parameters of the scaling regions, for each analysis, have been calculated by using linear regression, and the standard deviation of the signals has also been determined. The characterization of all defects has been determined from the parameters obtained for the signals filtered by using Savitzky-Golay and for the concatenate signals filtered by using band pass. It is also shown that, ´within this fractal analysis, the best discrimination is obtained by using Savitzky-Golay method.