Utilização de técnicas espectro analíticas para avaliação de açúcar de cana-de-açúcar
Abstract
Brazil is the largest sugarcane producer in the world, and its consolidated industry also
places it in the position of the largest producer and exporter of sugar. With a highly
variable process and also with large transported volumes, the control of tampering and
analysis of its physical chemical parameters becomes a crucial step at various points
in its supply chain, establishing quality standards and its final demand. Given the
importance of control, in the laboratory scope, new techniques and applications are
sought that enable quick, non-destructive analysis, not using harmful reagents and
also easy to operate, and that this allows for quick actions that lead to improvements
in the efficiency of the processes decrease in losses. Thus, the objective of this work
is the evaluation and development of new applications for sugar analysis using the
Laser Induced Plasma Optical Emission Spectrometry (LIBS) and Near Infrared
Region Spectroscopy (NIRS) techniques. In the first part of this study 64 samples of
mixture were prepared, and using chemometric tools, PLS models were developed
and optimized to determine the parameters Sugar (%), Sand (%) and Black Spots
(pp/100g). For performance evaluation, the RMSEP values were obtained for the
prediction models for Sand at 6.48, 1.29 and 5.96 percent, for Sugar at 6.53, 1.23 and
5.99 percent and for Black Spots at 7.74, 5.81 and 10.94 bp/100g of sample, each of
which is obtained with the LIBS, NIR and Data Fusion data respectively. In the second
part of this study, spectroanalytical data from 41,110 samples of Crystal, VHP,
Amorphous Refined and Granulated Refined sugars were collected via NIRS in
different regions of Brazil and mPLS and ANN prediction models were developed for
Ash (% m/m), Color ICUMSA (UI), Polarization (°Z) and Humidity (%). Individual mPLS
and ANN models were compared with a global model (containing all types of sugar),
with the literature and also with a proficiency test applied in industry laboratories. For
the ANN models of the Global calibration, it was possible to obtain RMSEP values of
0.017 for ash, 43 for color, 0.016 for moisture and 0.098 for Pol. Therefore, the LIBS
technique showed promise in the evaluation of contaminants in sugar, and can be
evaluated for development of routine applications obtaining greater spectral variability.
The NIRS technique for sugar evaluation demonstrated an advanced state of
development and robustness due to global variability and performance when
compared to literature and laboratory techniques. Finally, the implementation of global
ANN models provide greater stability, robustness and precision, having several
advantages over local models and mPLS models.
Collections
The following license files are associated with this item: