A possible application of FT-NIR spectroscopy was tested in order to evaluate different roasting degrees in coffee beans. Relationships between some physico-chemical changes during coffee roasting and near infrared spectra of raw and roasted coffee were investigated in order to test the effectiveness of non-destructive NIR measurements for the prediction of roasting degree. Pure and blend samples of several Arabica and Robusta coffee varieties were roasted in a pilot plant roaster for 3, 4, 5, 5.5, 6 and 6.5 minutes and analysed for conventional roasting parameters (weight loss, colour, moisture and density) and NIR spectroscopy. Data results of conventional roasting analysis were used as reference method for the calibration of the NIR system. Using the partial least square algorithm, good statistics were obtained for the calibration of moisture, weight loss and density (R2 > 0.9) with root mean square error of cross-validation ranging from 0.02 to 1.32. Furthermore, it was possible to discriminate the different roasting degrees of coffee beans by cluster analysis on the basis of the obtained NIR spectra. Validation results allowed to well predict the data obtained by conventional measurements.
Prediction of coffee roasting degree by near infrared spectroscopy / Alessandrini L.; Romani S.; Dalla Rosa M.. - ELETTRONICO. - (2006), pp. 290-293. (Intervento presentato al convegno 21st International Conference on Coffee Science tenutosi a Montpellier nel 10 – 15 septembre 2006).
Prediction of coffee roasting degree by near infrared spectroscopy
ALESSANDRINI, LAURA;ROMANI, SANTINA;DALLA ROSA, MARCO
2006
Abstract
A possible application of FT-NIR spectroscopy was tested in order to evaluate different roasting degrees in coffee beans. Relationships between some physico-chemical changes during coffee roasting and near infrared spectra of raw and roasted coffee were investigated in order to test the effectiveness of non-destructive NIR measurements for the prediction of roasting degree. Pure and blend samples of several Arabica and Robusta coffee varieties were roasted in a pilot plant roaster for 3, 4, 5, 5.5, 6 and 6.5 minutes and analysed for conventional roasting parameters (weight loss, colour, moisture and density) and NIR spectroscopy. Data results of conventional roasting analysis were used as reference method for the calibration of the NIR system. Using the partial least square algorithm, good statistics were obtained for the calibration of moisture, weight loss and density (R2 > 0.9) with root mean square error of cross-validation ranging from 0.02 to 1.32. Furthermore, it was possible to discriminate the different roasting degrees of coffee beans by cluster analysis on the basis of the obtained NIR spectra. Validation results allowed to well predict the data obtained by conventional measurements.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
