Prediksi Komponen Minyak Buah Kelapa Sawit Secara Non-Destruktif Menggunakan UV-Vis Spectroscopy

Abstract

Kandungan kimia buah kelapa sawit menentukan kualitasnya dan berubah selama proses pematangan. Prediksi kandungan kimia umumnya dilakukan secara visual, namun cara tersebut kurang akurat. Metode laboratorium lebih akurat tetapi menghabiskan banyak waktu dan biaya. Untuk mengatasi hal tersebut perlu dikembangkan metode cepat untuk memprediksi kandungan kimia buah sawit. Penelitian ini bertujuan memprediksi kandungan kimia buah kelapa sawit menggunakan spektrometer UV-Vis dan metode kalibrasi PLS-R dan MLR. Sebanyak 377 sampel buah sawit diukur reflektansinya menggunakan spektrometer UV-Vis pada panjang gelombang 313–818 nm. Kandungan kimia ditentukan melalui analisis kimia. Data spektrum diberi pra-perlakuan Smoothing Savitzky-Golay dan normalize, kemudian dikalibrasi dengan data kimia menggunakan PLS-R dan MLR dengan software The Unscrambler X 10.4. Hasil penelitian menunjukkan PLS-R memberikan hasil kalibrasi terbaik. Prediksi asam lemak bebas dan kadar air terbaik masing masing pada Smoothing Savitzky-Golay (r=0,85; SEC=0,87%; SEP=1,04%; CV=28,16%; RPD=1,51), dan (r=0,96; SEC=6,21%; SEP=6,50%; CV=11,15%; RPD=3,38), sedangkan kadar minyak terbaik dengan normalize (r=0,96; SEC=5,15%; SEP=5,13%; CV=21,50%; RPD=3,47). UV-Vis dan PLS-R efektif untuk prediksi minyak dan kadar air, namun kurang optimal untuk asam lemak bebas.

The chemical composition of oil palm fruit determines its quality and changes during the ripening process. Chemical composition is generally predicted visually, but this method lacks accuracy. Laboratory methods are more accurate but require considerable time and cost. To address this issue, a rapid method is needed to predict the chemical composition of oil palm fruit. This study aims to predict the chemical composition of oil palm fruit using a UV-Vis spectrometer and calibration methods PLS-R and MLR. A total of 377 oil palm fruit samples were measured for reflectance using a UV-Vis spectrometer at wavelengths of 313–818 nm. The chemical composition was determined through chemical analysis. Spectral data were pre-treated using Savitzky-Golay Smoothing and normalization, then calibrated with chemical data using PLS-R and MLR with The Unscrambler X 10.4 software. The results showed that PLS-R provided the best calibration. The best predictions for free fatty acid and moisture content were obtained using Savitzky-Golay Smoothing (r = 0.85; SEC = 0.87%; SEP = 1.04%; CV = 28.16%; RPD = 1.51) and (r = 0.96; SEC = 6.21%; SEP = 6.50%; CV = 11.15%; RPD = 3.38), respectively. The best prediction for oil content was achieved using normalization (r = 0.96; SEC = 5.15%; SEP = 5.13%; CV = 21.50%; RPD = 3.47). UV-Vis and PLS-R were effective for predicting oil and moisture content but were less optimal for free fatty acid prediction.