Simulação de propagação de ondas ultrassônicas sob diferentes regimes de espalhamento

Abstract

The ultrasonic non-destructive evaluation is vastly used to identify and measure discontinuities, to characterize material, to measure grain’s size and microestructural degradation present in the material. Heterogen medium cause ultrassonic wave scattering during its propagation due the constant change of the accoustic impendance inside the medium. The type of scatter acting during the wave’s propagation relies on the relation between the mean grain size composing the medium and the ultrassonic wave lenght. Therefore, waves propagating inside materials containing differents grains sizes, will be under differents scatterings. Differents bidimensional heterogen media were acquired from simulation of grain growth by the Potts model and then used in the simulation of the ultrasonic wave propagation, which was developed according to the finite volume method. The data obtained from the ultrasonic wave propagation were submited to the detrended fluctuation analysis preprocessing. After this preprocessing, Karhunen-Loève (along with the nearest mean class rule algorithm) and neural networks algorithms were used to classify the scattering of the wave in each modeled media, reaching respectively, 97,50% and 95,16% of mean hit ratios. Such numbers show that the simulation of the ultrassonic wave propagation in different heterogen media was successful.