Proposição de modelos de classificação e de calibração para polímeros presentes em resíduos eletrônicos empregando laser-induced breakdown spectroscopy (LIBS) e ferramentas quimiométricas

Abstract

This thesis proposes the development of analytical methods for the analysis of polymers from waste electrical electronic equipment (WEEE) by laser-Induced breakdown spectroscopy (LIBS). The results presented in the chapters 2 and 3 of this thesis were developed in collaboration with a recycling company of WEEE called REVERSA. In the first method proposed LIBS was used for the identification and classification of six polymers and a total of 477 samples were analyzed. Classification models using k-nearest neighbors (KNN) and soft independent modeling of class analogy (SIMCA) were used to classify the polymers. The proposed models presented satisfactory results for the validation samples, with accuracy of 98% for KNN and 92% for SIMCA. In addition, other validation parameters such as sensitivity, false alarm rate and specificity showed adequate. In the second method, the LIBS was employed to determination of the constituents polycarbonate (PC) and acrylonitrile-butadiene-styrene (ABS) in PC/ABS blends. For this were built multivariate calibration models using partial least squares (PLS). The PLS models obtained for the PC and ABS presented good analytical parameters as: i) standard error calibration (SEC) of 3.4 (w/w%); (ii) standard error of cross-validation (SECV) of 5.6 (w/w%); iii) linear correlation coefficient of 0.996; iv) number of latent variables equal to 2; v) explained variance of 70%. For the validation of the PLS models PC/ABS blends with unknown concentrations of PC and ABS were analyzed, and the information provided by LIBS was in agreement with the reference technique. A third method was developed in collaboration with a plastics recycling company of car batteries. These plastics were contaminated with Pb and the objective was to evaluate different calibration strategies in LIBS for the quantitative determination of this analyte. Multivariate calibration strategies and a new univariate strategy were proposed. The new univariate calibration strategy was called a two-point calibration transfer and the results presented were satisfactory. The chemometrics tools including experimental desing, supervised methods for classification, unsupervised methods and multivariate calibration were employed during the development of the methods proposed in this thesis.