Influência da preparação metalográfica e das condições de aquisição das imagens na quantificação automática de fases secundárias

Abstract

In this work, the influence of the metallographic preparation and the image acquisition conditions on the quantification of secondary phases by a multilayer perceptron neural network was investigated. The images were obtained by scanning electron microscopy (SEM) from Ni-based superalloy coatings on C-Mn steel deposited by gas tungsten arc welding (GTAW) cold wire process. The results showed that excessive chemical etching of the microstructure (inadequate metallographic preparation) and/or inadequate brightness and contrast adjustment during the SEM image acquisitions interfere in the segmentation results and consequently on the secondary phase quantifications.