KAOLIN REFLECTANCE SPECTROSCOPY: USING PLS-R TO PREDICT CONTAMINANT CONTENT

Abstract

The present study deals with the prediction of contaminant concentrations in kaolin using Partial Least Squares Regression (PLS-R). The aim is to show that PLS-R method can be used to predict contaminant concentration in kaolin. High level kaolin means a kaolin with high-brightness. Since brightness is directly related to the reflectance spectrum and kaolin contaminants affect the reflectance spectrum it is important to the beneficiation of kaolin relates optical features and contaminants. Depending on the product to be produced, the optical parameters will influence how the kaolin will be processed. High-brightness kaolin and two red and yellow inorganic pigments were used to simulate colours contaminants frequently found in Brazilian kaolins, such as, hematite, goethite, rutile and anatase. By adding different pigment concentrations to the pure kaolin, it was possible to create a small dataset containing the visible reflectance spectrum of each sample with the respective optical quality parameters of the kaolin. Results allow us to conclude that PLS-R can predict through the reflectance spectrum the contaminant concentration of the kaolin with R-squared equal 0.9954 for red content and R-squared equal 0.9973 for yellow one.