Estimation of cellulose crystallinity of sugarcane biomass using near infrared spectroscopy and multivariate analysis methods

Abstract

A method for estimation of sugarcane (Saccharum spp.) biomass crystallinity using near infrared spectroscopy (NIR) and partial least squares regression (PLS) as an alternative to the standard method using X-ray diffractometry (XRD) is proposed. Crystallinity was obtained using XRD from sugarcane bagasse. NIR spectra were obtained of the same material. PLS models were built using the NIR and crystallinity values. Cellulose crystallinity ranged from 50 to 81%. Two variable selection algorithms were applied to improve the predictive ability of models, i.e. (a) Ordered Predictors Selection (OPS) and (b) Genetic Algorithm. The best model, obtained with the OPS algorithm, presented values of correlation coefficient of prediction, root mean squared error of prediction and ratio of performance deviation equals to 0.92, 3.01 and 1.71, respectively. A scatter matrix among lignin, α-cellulose, hemicellulose, ash and crystallinity was built that showed that there was no correlation among these properties for the samples studied.