Análise da performance dos sensores óptico e por transmitância de raios-x no processo de classificação de calcário em mina de Caçapava do Sul/RS

Abstract

The automated material selection technology has its first researches dating back to the 1930s (SALTER & WYATT, 1991; NIENHAUS et al., 2014). However, only from the 1980s, with the development of larger processors, has it become possible to use it in practice (PRETZ, 2012). The objective of this study was to evaluate the performance of the sensor - based separation technology in the classification of the limestone ROM of Caçapava do Sul mine, as an alternative to the manual selection process currently employed. These dolomite marbles, as well as their syngeneic casts, are framed in the Passo Feio Complex (HOERLLE et al., 2010, FABRIS, 2015, CPRM, 2017). The formation characteristics of limestone occurrences in the region of Caçapava do Sul are quite particular, making the mining process very complex, due to the series of lithotypes found and the way in which they are organized. As main types of sterile are syngeneic pelitic intercalations, mafic intrusions and granitic apophyses. In the methodology, the sampling was carried out with the help of drone, contemplating all the variants of ore and sterile existing in the deposit. Three granulometry (-25 + 9.5mm, -50 + 25mm, -70 + 50mm) and three material conditions were tested (dirty-dry, clean-wet and clean-dry). In the analysis of results were used massic recovery and metallurgical recovery indices and other indicators such as: accuracy, sensitivity, specificity. As results, the droneassisted sampling generated an outcrop virtual model that served as the basis for the collection of representative samples. In the sorting tests with optical sensor, metallurgical recovery values were between 35 and 85% in the lowest sieve range and between 91 and 98% in the larger range. The accuracy had values between 43 and 85% in the lowest mesh and 97 and 98% in the largest, which also occurred with the behavior of the ROC graphs, which had better separations while increasing the particle size. It was concluded that the X-ray sensor showed no affinity to classify the materials under test. The optical sensor was shown to be applicable in all tested situations, with metallurgical recoveries between 35% and 98%, depending on the condition and grain size of the feed. The granulometry was the major factor influencing the separation performance, implying variations of up to 30% in the accuracy indexes and directly affecting the positioning of the points in the ROC graphs (receiver operating characteristic), where the largest grain size range had the best results, regardless of the simulation performed.