Analisis Pengaruh Variasi Jumlah Lapisan Biji pada Akurasi Prediksi Kandungan Minor Biji Kopi Arabika Hijau Bondowoso dengan NIR Spectroscopy

Void space in the bean layers will lead to the occurrence of non-fully interacted radiation (NFIR) affecting the reproducibility of NIRS measurements. Void space in addition to being affected by particle size is also influenced by the number/thickness of the bean layers. The objective of this study was to analyze the effect of number of bean layer variation on prediction accuracy of caffeine, chlorogenic acid and trigonelline in Bondowoso green Arabica coffee beans by NIR Spectroscopy (NIRS). The study was conducted using three kind of layers, i.e. 3, 4, and 5 layers, with 100 samples each. Samples were measured by FT-NIR spectrometer in wavelength of 1.0002.500 nm. The pretreatment method used were second derivative (dg2), the combination of first derivative (dg1) and Multiplicative Scatter Correction (MSC), and the combination of dg2 and MSC, while calibration method used was Partial Least Square (PLS). The results shows that the accuracy of 5 layers was better than 3 or 4 layers. The best calibration and validation for caffeine was obtained by dg2 pretreatment and 6 factors of PLS (r = 0.99; SEC = 0.01%; SEP = 0.01%; and RPD = 5.40), for chlorogenic acid was by dg2 pretreatment and 5 factors of PLS (r = 0.99; SEC = 0.09%; SEP = 0.09%; and RPD = 4.76), whereas for trigonelline was by combination of (dg2, MSC) and 5 factors of PLS (r = 0.99; SEC = 0.01%; SEP = 0.01%; and RPD = 4.86). Therefore, the 5 layers can be used as a reference in NIRS measurement of coffee beans.