Combinação de ferramentas quimiométricas e infravermelho próximo para análise de material desfibrado de cana-de-açúcar

Abstract

The Monsanto Company presents in one of its segments the development of varieties of sugacane which have high productivity. The productive capacity of these clones is measured using at lest six analytical parameters: brix degree, Pol, purity, fiber, juice pol and cane pol. The determination of these parameters is done by instrumental measurements that require intense attention of the operator. Beyond that, in Monsanto, the cost per analysis is the order to R$7.50. Of course, not only the cost per analysis is the main parameter, but how fast the frequency of results to verify if the variety is or isn’t economically viable. The proposal of this professional master’s degree project is to use the near infrared spectrophotometer (NIR) combined with chomometric tools to propose calibration models. These models would be used to provide faster and low costs of the analytical parameters reported above. In realization of this project were used over 7000 NIR spectra were obtained after defibration sugarcane. Spectra were obtained and constructed multivariate calibration models (6300 spectra) with the chemometric tool PLS (partial least squares). The models were created in the Conchal laboratory and validated with 700 spectra that not were part of the calibration database and the validation errors ranged from 0.57 (for brix degrees) to 3.55 (for pol). Furthmore, the models of calibration were used in two other experimental stations of Monsanto. The first is located in Araçatuba (SP) and the second in Mandaguaçu (PR) At this part were tested 3 strategies of transfer calibration (recalibration, model update and slope and intercept) and slope and intercept had the lowest validation errors (around twice lower when compared with no transfer). The professional master degree represented up to now a saving of financial resources around R$ 100.000,00 per harvest.