Prediction of mechanical properties of poly(ethylene terephthalate) using infrared spectroscopy and multivariate calibration

Authors

  • Viviane Fonseca Caetano,

    1. Departamento de Engenharia Química, Universidade Federal de Pernambuco, CTG, 50740-521 Recife, Pernambuco, Brasil
    Search for more papers by this author
  • Glória Maria Vinhas,

    Corresponding author
    1. Departamento de Engenharia Química, Universidade Federal de Pernambuco, CTG, 50740-521 Recife, Pernambuco, Brasil
    • Departamento de Engenharia Química, Universidade Federal de Pernambuco, CTG, 50740-521 Recife, Pernambuco, Brasil
    Search for more papers by this author
  • Maria Fernanda Pimentel,

    1. Departamento de Engenharia Química, Universidade Federal de Pernambuco, CTG, 50740-521 Recife, Pernambuco, Brasil
    Search for more papers by this author
  • Simone da Silva Simões,

    1. Departamento de Engenharia Química, Universidade Federal de Pernambuco, CTG, 50740-521 Recife, Pernambuco, Brasil
    Search for more papers by this author
  • Mário César Ugulino de Araújo

    1. Departamento de Química, Universidade Federal da Paraíba, CCEN, Caixa Postal 5093, 58051-970 João Pessoa, Paraíba, Brasil
    Search for more papers by this author

Abstract

This work describes a method to determine mechanical properties (tensile strength at break and tensile modulus) of poly(ethylene terephthalate) using median infrared spectroscopy and multivariate calibration. Infrared spectroscopy is very promising for polymer process control and final product analysis because it is rapid and nondestructive. The spectra of the films were obtained using two techniques: attenuated total reflection and direct transmission. The spectra were subjected to various preprocessing procedures, such as smoothing and derivative using the algorithm Savitzy-Golay, standard normal variate, multiplicative scatter correction and, as well, combinations of some of these preprocessing techniques. The predictive ability of the regression models were evaluated using an external validation set. The regression techniques used, partial least square and multiple linear regression, showed, in general, comparable results with root mean square error of prediction similar to the repeatability of the conventional method used to determine these mechanical properties (1.3 kgf/mm2 for tensile strength at break and 29.6 kgf/mm2 for tensile modulus). © 2012 Wiley Periodicals, Inc. J. Appl. Polym. Sci., 2013

Ancillary