Non-Destructive Prediction of Moisture Content in Cascara Using NIR Spectroscopy with PLS and PCR

Abstract

Coffee cherry pulp is a by-product of coffee processing that has not been optimally utilized. Coffee cherry pulp can be dried to produce a herbal tea product, known as cascara. As an herbal tea product, moisture content is one of the most important quality parameters for assessing the quality of cascara. Therefore, a method is required to measure the moisture content of cascara. One of the methods developed is NIR spectroscopy, which is non-destructive, fast, and does not require chemicals. The purpose of this research is to explore the application of NIR spectroscopy in predicting cascara moisture content using partial least squares (PLS) and principal component regression (PCR) methods and to evaluate the performance of each method in building an optimal calibration model. Pretreatment of the spectrum data was carried out with standard normal variate (SNV), gap-segment 2nd derivative (dg2), and a combination of SNV+dg2. The results showed that the best prediction of cascara moisture content used the PLS calibration technique with dg2 pretreatment and five factors 5. The values obtained were Rc² = 0.96, RMSEC = 0.87 %, SEC = 0.87 %, Rp² = 0.90, RMSEP = 1.22 %, SEP = 1.16 %, and RPD = 3.44. Meanwhile, the PCR method produced good predictions using SNV pretreatment, with a factor of 8. The prediction results were Rc² = 0.89, RMSEC = 1.40 %, Rp² = 0.90, RMSEP = 1.33 %, and RPD = 3.15. NIR spectroscopy can predict the moisture content of cascara nondestructively and rapidly.