Prediction of Caffeine and Protein of Arabica Coffee Beans Using Near Infrared Spectroscopy (NIRS)

Abstract

Testing the chemical components of coffee beans has so far been carried out conventionally with laboratory analysis which requires a long time and is expensive. Technological advances allow testing of chemical components to be carried out quickly and accurately using the NIRS (Near Infra Red Spectroscopy) method. This research aims to develop a prediction model for caffeine and protein in Solok Radjo coffee beans using the NIRS method. Solok Radjo coffee is a type of Arabica with specialty grade because of its very strong character, aroma and taste. A total of 30 samples with a weight of 6 g per sample were used in this study. This research uses NIRS Type FT-IR IPTEK T-1516 with a wavelength of 1000 - 2500 nm. The partial least squares (PLS) method was used to process the data with several SNV, MN, and MSC pretreatments to improve the model. The research results show that caffeine is found at wavelengths of 1456 - 1475 nm, 1937 - 1974 nm. Proteins 1455 - 1475 nm, and 1935 - 1974 nm. MSC pretreatment is able to improve PLS performance results. Caffeine calibration values are R2 = 0.996 and SEC = 0.002%, validation values R2 = 0.989, SEP = 0.002%, and RPD 11.869 while protein calibration R2 = 0.999 and SEC = 0.004%, Validation values R2 = 0.999, SEP = 0.010%, and RPD 19,943. NIRS can be used to predict the chemical components of Solok Radjo coffee non-destructively using the PLS method. Key work: Caffeine, NIRS, PLS, predict, protein