Avaliação da liberação mineral por análise de imagens digitais obtidas emmicroscópio ótico de luz refletida

• Ano de defesa: 2013
• Autor(a) principal: Edgar Rodo Mantilla
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Minas Gerais; UFMG
• Palavras-chave em Português: Análise de imagens digitais; Microscopia ótica; Grau de liberação; Minério de ferro; Engenharia metalúrgica; Tecnologia mineral; Minérios de ferro
• Link de acesso: http://hdl.handle.net/1843/CMBC-8UHPH2

This work consists in the application of an automatic routine for determination of degree of liberation in Samarco Mineração S.A.s iron ore routine samples through the use of digital image processing and analysis techniques obtained by reflected light optical microscope. Presently, the Gaudin traditional method is used by Samarco to evaluate the mineral liberation. A Repeatability & Reproducibility test was performed to evaluate this method and then compared with degree of liberation values established as reference. Therewith, it was possible to evaluate the accuracy and precision of mineral liberation analyses currently determined by Samarco. An automatic image processing and analysis routine was developed in the software AxioVision, which Samarco has license to use, for mineral liberation analyses of the routine samples of Samarco´s iron ore. By comparing the results of current methodology used in determination of degree of liberation vs. the reference values, it was possible to evaluate that the current methodology meets the expectations of precision and accuracy of the analyses. With the Repeatability & Reproducibility test it was possible to measure the capacity of the laboratorist to repeat the same measurement contributes in 9.6% in the total analyses variation, and the capacity of different laboratorists in performing the same measurement contributes in 16.5% in the total analyses variation, in other words, was identified the possibility of gains in accuracy in order of 26.1% by using the automatic analysis software. With the automatic routine tests results it was possible to reduce the mean absolute error by 82%, reduce the analyses variability by 21% (nullifying the influence/subjectivity of laboratorians in total analyses variation), reduce the analyses average time by 77%, from 14min 26s to 3min 19s, and increase the data correlation against the reference values by 11%. Finally since a controlled environment is available, the same conditions/parameterizations of the images acquisition and the same conditions of polished sections preparation are kept, it´s possible to obtain gains by the application of the digital images processing and analyses techniques through an automatic routine of determination of degree of liberation at Samarco Mineração S.A.