Prediction of impurities in cocoa shell powder using NIR spectroscopy

Cocoa shell is a by-product from cocoa industry which contains bioactive compounds of high and attractive value for food, pharmaceutical and cosmetics industry. However, cocoa shell can be contaminated by undesirable materials that, even in small amounts, would not change the color, aroma, and taste characteristics of the final product. Identification and prediction of this impurity are performed using expensive methods that require chemicals and produce residues. Thus, this work aims to investigate the performances of benchtop (867–2535 nm) and portable (900–1700 nm) near-infrared (NIR) spectrometer for fast prediction of cocoa shell powder impurities. Mixtures (n = 432) of cocoa shell powders with leaves, pods, stem fragments and nibs at several proportions (0–20 % w/w), were analyzed. Multivariate calibration models were developed using partial least-squares regression (PLSR) with raw spectra and various preprocessing approaches applied to the spectra. The most informative spectral variables were selected by variable importance in projection (VIP) method. Results obtained for the benchtop (R²_P> 0.99 and RMSEP<0.71) and low-cost portable (R²_P> 0.92 and RMSEP<1.70) devices are promising, and portable spectrometer could be used to certify cocoa shell purity.