bachelorThesis
Análise espectroscópica de amostras de trigo por infravermelho médio FTIR e NIR e construção de modelos multivariados de calibração por PLS
Fecha
2016-06-13Registro en:
ROSSI, Andressa Cristina. Análise espectroscópica de amostras de trigo por infravermelho médio FTIR e NIR e construção de modelos multivariados de calibração por PLS. 2016. 65 f. Trabalho de Conclusão de Curso (Graduação) - Universidade Tecnológica Federal do Paraná, Pato Branco, 2016.
Autor
Rossi, Andressa Cristina de
Resumen
Wheat flour production requires several laboratory tests developed not only for the wheat grain, but also for flour, which are (the tests, right?) extremely important to ensure the quality of products made from this raw material. These analyses require large sample volume, a longer time analysis and specific equipment for each kind of analysis. The purpose of this study was to develop multivariate calibration models for regression by partial least squares (PLS) for the determination of falling number (FN), alveography (W), wet gluten (GU) and reason elasticity for extensibility (P/L) in wheat flour samples. It was also developed models of defects content and triguilhos (DT), impurities, humidity and hectoliter weight (PH) of wheat grain samples from the infrared technique in regions near (NIR) and mid (MIR). It was used 170 samples of grain and flour available in Pato Branco, PR. The acquisition of the spectra was performed in duplicate, in the spectral range 12.500 to 4.000 cm-1 for NIR and 4.000 to 400 cm-1 for MIR, reference analyses provided by the mill itself. Of the eight best models developed for samples, four (GU, P/L, humidity and PH) obtained good average evaluation criteria, with validation mean square error (RMSEV) and calibration (RMSEC) below 1.30 and 1.02, respectively, and validation of R2 above 0,77. Two models (FN and W) despite good R² above 0,81, obtained high values of RMSEV and RMSEC (below 24,49 and 5,94). The model showed good DT evaluation criteria, R² of 0,94, RMSEC of 0,11 and RMSEV of 0,26, but the set of spectral data was not representative of the sample, and it is not suitable for quantifying the DT content in unknown samples. The model of impurities did not obtained satisfactory criteria with any of the techniques used. In general, the regression models for MIR were more effective than the models for NIR, demonstrating the possibility of determining both the GU, P/L flour samples and humidity and pH in wheat grain samples, in a fast and economical fashion. They have also shown that it is possible to improve the models DT and impurities to increase the sensitivity of the method. However the MIR and NIR models are not suitable for quantifying the levels of FN and W.