Uso de espectroscopia LIBS acoplados a rotinas de aprendizado de máquina para diferenciação de minerais

Abstract

Minerals are defined as natural inorganic substances that are formed due geological process with a regular atomic lattice and solid format in room temperature. Their chemical structure are formed between charged particles, where they are grouped into classes based of their predominant anion group. The correct classification of minerals samples plays an impotant role in several fields of study. Minerals samples can be mistakenly classified into different groups, if analyzed only visually, due to similarities and/or differences in visual characteristics between samples of the same mineral groups. The most commonly used tests, apart from the visual analysis that requires skillful employees, are analitical techniques such as X-ray spectroscopy ou Inductively Coupled Plasma Optical Emission Spectroscopy (ICP). But, these analitical techniques are not advantageous because they require the use of ionizing radiation and no possobility to perform in situ analysis, respectively. The LIBS technique is an alternative to be used in situ, that provides fast and efficient elementar analysis, what is of grear importance to discriminate and classify mineral samples. In the present study, the use of the spectroscopy technique Laser Induced Breakdown Spectroscopy (LIBS), combined with chemometrics methods as principal component analysis (PCA) and machine learning algorithms, k-nearest neighbour (KNN), support vector machine (SVM) and discriminant analysis (DA), were evaluated in the differentiation and classification of minerals samples from the classes silicate, oxide, phsphate, carbonate and fluoride. The results showed that the Fine KNN algorithm were able to correctly classify all samples in the external validation scheme, for the silicate class, reaching accuracy of 100% in three differents approachs implemented. There are still some limitations in the use of this proceedment, mainly related to the differentiation of the non-silicate minerals samples.