Simple and effective mathematical models for cellulose water content calculation from absorbance/wavenumber shifts in NIR spectrum

Abstract

This study presents the evaluation of the near infrared (NIR) spectroscopy utility in the quantitative analysis of the moisture content in cellulose (from 0.5 to 11.0 wt.%). Being the first statistic-based work examining the relations between the wavenumber/absorbance and moisture content, this research innovatively describes the variations in the X- and Y-coordinates of 9 absorption bands plotted as a function of water amount established with Karl-Fischer titration. Simple linear, semilogarithmic and power models have been investigated. While model exhibited R2>90%, the standard error of calibration (SEC) has been shown and an external validation was carried out. Power model has been successfully applied to describe the changes in the intensity of the peak situated from 5400 to 5000 cm−1 (R2 = 90%) with a standard error of prediction of approx. 17.2%. Furthermore, absorption bands that wavenumber shifts could provide significant information on the behaviour of the active centres in cellulose have been favourably defined.