Quality Control Parameters in the Roasted Coffee Industry: a Proposal by Using MicroNIR Spectroscopy and Multivariate Calibration

Abstract

This research presents the use of near-infrared (NIR) spectroscopy as a fast alternative of conventional methods, suitable for in-line use for quality control in the industrial coffee processing. Chemometric models were constructed from spectra of 217 coffees sampled during the period of 5 months of coffee processing, which demonstrates the viability of the NIR spectroscopy in terms of prediction, since it provides realistic variability in chemometric models. In this work, granulometry, moisture content, color, and the infusion time of commercially roasted coffee beans were evaluated by reference methods, NIR spectroscopy, and partial least squares (PLS). Four models were built using 217 roasted coffee samples, differentiated by type of beverage and roast degree. After the model’s optimization by the outlier evaluation, parameters of merit such as residual prediction deviation, sensitivity, analytical sensitivity, the inverse of analytical sensitivity, limits of detection and quantification, adjust, and linearity were calculated. The models showed predictive capacity and high sensitivity in determining the analyzed ones. All these quality control parameters were predicted with adequate correlation coefficient, which ranged from 0.73 to 0.84 and presented satisfactory performance. The parameters of merit presented promising results, indicating that the prediction models developed for granulometry, color, moisture content, and infusion time can be safely used in the coffee industry as an alternative to reference methods. Moreover, the proposed methodology is sample preparation free, fast, simple, efficient, and suitable for in-line uses.