Desenvolvimento de algoritmos e ferramentas computacionais para suporte na caracterização microestrutural de materiais através de imagens de microscopia eletrônica

• Autor(a) principal: Bonotto, Edison Luiz
• Data de Publicação: 2021
• Tipo de documento: Tese
• Idioma: por
• Título da fonte: Biblioteca Digital de Teses e Dissertações da UFPB
• Texto Completo: https://repositorio.ufpb.br/jspui/handle/123456789/22227
• topic: Processamento de imagens; Clusterização; Microestruturas de materiais; Microscopia eletrônica de varredura; Análise de textura; Meta-heurísticas; Image processing; Clustering; Materials microstructures; Scanning electron microscopy; Texture analysis; Meta-heuristics; CNPQ::ENGENHARIAS

The development of functional materials with specific properties is intrinsically dependent on microstructural characteristics. Especially with the increase in the production of materials that use nanotechnology, where the effects of interface, distribution and phase composition on properties are striking, the precise and fast analysis of materials on a microscopic scale becomes imperative in the development process. In this sense, the analysis of microscopic images requires methods that incorporate computational tools that minimize subjective interpretations due to the complexity of the images and, at the same time, the high number of data generated. Although there are several computer programs available for image analysis, there are subtlety of microstructural analysis that are deeply connected with the nature of the materials, so that the image analysis response can incorporate chemical information, such as phase composition. . In fact, considering the scope of the literature review of this work, no computer program automatically addresses the chemical nuances referred to a priori. The thesis defended in this work is that it is possible to obtain automatic quantitative analyzes in polycrystalline materials from scanning electron microscopy images in backscattered mode. Therefore, a computational tool capable of quantitatively determining the degree of dispersion of the constituent phases as well as their compositions was developed. Therefore, an algorithm was developed that estimates the average gray tone of a phase based on the coefficient of retransmitted electrons and incorporated a clustering method using the K-Means optimization routine. Two particle scattering indicator algorithms based on entropy and co-occurrence were also incorporated. The validation of the routine was obtained from simulated images (benchmarks) with which the recovery of the values imposed on the images and other efficiency parameters of the automatic analysis process was verified. With the validation, the program was applied to the analysis of the phase distribution in polycrystalline rocks and in phase images present in a template sample, where its chemical composition is known a priori. The implemented algorithm for the identification of the gray level of the phases proved to be efficient and robust, estimating with good precision the expected average tone for each phase of the samples. The clustering method using the ZK-Means optimization routine presented excellent segmentation, recovering the values imposed on the simulated images in the analysis process. Particle dispersion indicator algorithms generated 2D and 3D graphics, which can be used to observe their relationship with material properties. The tools proved to be promising, and it was possible to verify their effectiveness in determining the phases and their distributions.