Proposição de modelos de regressão para previsão de propriedades de tintas da cadeia produtiva da indústria gráfica

• Autor(a) principal: Oshiro, Eduardo Hideki
• Data de Publicação: 2019
• Tipo de documento: Dissertação
• Idioma: por
• Título da fonte: Repositório Institucional da UFSCAR
• Texto Completo: https://repositorio.ufscar.br/handle/ufscar/11975
• topic: Tinta de Impressão; Plsr; Quimiometria; Propriedades da Tinta; Printing Ink; Chemometrics; Ink properties; CIENCIAS EXATAS E DA TERRA::QUIMICA::QUIMICA ANALITICA::INSTRUMENTACAO ANALITICA; CIENCIAS EXATAS E DA TERRA::QUIMICA::QUIMICA ANALITICA

Quality control of inks is very important during and after the manufacturing process and is also meaningful in the consumer industries or paint applicators. For this reason, an opportunity was noticed for the development of a method of determining the properties of a solvent based ink of the printing industry. The analytical method proposed was through the infrared spectroscopy and multivariate calibration. The ink used in the project is employed in the rotogravure printing process. This process is used for printing many quantities of printing materials at high speed. It is usually applied in the production of packaging and magazines. The fallowing quality control parameters: viscosity, density and solids of black and white inks were analyzed by the reference methods according ASTM standards. Prediction models for these parameters were developed through Fourier transform infrared spectrometer with attenuated total reflectance sampling accessory (FT-MID-IR ATR) and the multivariate calibration, the partial least squares regression (PLSR). Pre-treatments in the spectra (1st and 2nd derivatives with Savitzky-Golay) were tested and also the models with spectra without treatments. The best results were found according to the parameter and ink tested, for viscosity of black ink the model without treatment was used, for viscosity of white ink the 2nd derivative was used. And for density and solids, unique models were developed to be used in black and white inks, the 1st derivative was the best option in both cases. The best models and latent variables were selected through the evaluation of the root mean square errors values of calibration, validation and cross validation (RMSEC, RMSEV and RMSECV), as well as the determination coefficients (R2) obtained from predicted x reference results. The set of validation samples showed a good quality of the results, according to the p-value. Although there are many articles related to prediction of chemical physics characteristics of fuels and derivatives using infrared spectroscopy and the chemometric tool, few studies with this approach to inks have been found in the literature. The following advantages of the developed method can be highlighted: low sample volume (0.1 mL - 1 drop), rapid analysis (2 minutes) and the low waste amount.