Propriedades visuais de espuma para predição dos teores da flotação em coluna do segundo concentrador da Samarco

Abstract

This work studied the relationship between visual properties of the reverse cationic froth flotation of iron ore in columns and the silica concentrate grade and iron tailings grade. In daily practice, experienced operators use flotation circuit characteristics called "aspect" of froth for visual decision-making process and parameterization. This is often described as visual characteristics of the froth. Trying to exploit these visual features, for improvements in the process, was installed a system to collect digital images from the column flotation circuit at Samarco. Through this system, 32,400 images were obtained under various operating conditions and a set of 72 continuous hours of operation was selected for the tests. The image database was processed using routines developed especially for this purpose, using the platform and user interface AForge IPLAB in C #. The main characteristics analyzed were first-order statistics, mean and standard deviation, the horizontal and vertical displacements of the froth and the textures described by matrices GLCM and NGLDM, totaling 11 descriptors. A new algorithm for displacement of the froth based on the translation property of Fourier transform, called phase correlation, was proposed, and its performance and results were compared with standard algorithms in the literature. This algorithm showed good performance and reliability in various lighting conditions and characteristics of the froth. Finally, it was possible to build an ARX mathematical model to predict the silica concentrate grade based on chemical analyses and visual properties of the froth, demonstrating that the aspect, described by operators, can be extracted by digital images system and may be used as a tool in decision-making or operational control.