Determination of sugar in non-alcoholic beverages using near infrared spectroscopy combined with chemometrics

Abstract

The measurement of Near infrared (NIR) spectroscopy, combined with chemometric techniques, has been widely employed for quality control in food products. This study presents a methodology to optimize the calibration models, called “universal model” of NIR spectra of primary condition (glucose solutions) and maintain the accurate prediction of secondary conditions. For instance, the models were designed for determination of glucose concentration in non-alcoholic drinks. Three stages of methodology including pre-processing, feature selection and main component extraction were applied to spectral data in order to obtain the universal calibration model. The simulated NIR spectra with different noise levels were used to ensure that the model from our methods is able to estimate amount of sugar in any conditions with high accuracy. From the analysis, the universal model improves the prediction for the test set (unseen data) for at least 30 percent compared to the other predictions. Then, it was used to quantify amount of glucose in non-alcoholic beverages (tea, cocoa and coffee in the case). The promising value for root mean square error of prediction (RMSEP) were obtained to be 0.72 (r2 = 0.9972), 0.99 (r2 = 0.9965) and 0.54 (r2 = 0.9940) corresponding to tea, cocoa and coffee system, respectively. Therefore, it might be implied that our universal model approach can be used to estimate glucose concentrations in other non-alcoholic drinks without any requirement of a new calibration model.