Please use this identifier to cite or link to this item: http://repository.ipb.ac.id/handle/123456789/165165
Title: Model Prediksi NIR untuk Kuantifikasi Sukrosa, Glukosa dan Fruktosa pada Batang, Nira dan Gula dari Batang Sawit
Other Titles: 
Authors: Hunaefi, Dase
Budi, Faleh Setia
Sani, Sharly Claudia Alghai
Issue Date: 2025
Publisher: IPB University
Abstract: Kelapa sawit merupakan komoditas utama Indonesia dengan produksi limbah batang sawit yang sangat melimpah, terutama dari proses replanting. Limbah ini berpotensi dimanfaatkan sebagai sumber bahan baku nira untuk produksi gula merah. Namun, pengrajin gula masih menghadapi kendala dalam mengukur kandungan gula secara akurat karena keterbatasan akses terhadap metode analisis seperti High-Performance Liquid Chromatography (HPLC), yang bersifat destruktif dan biayanya mahal. Penelitian ini bertujuan mengkaji potensi penggunaan Near-Infrared Spectroscopy (NIR) sebagai metode alternatif yang cepat, non-destruktif, dan ekonomis untuk analisis kuantitatif kandungan total padatan terlarut, sukrosa, glukosa, fruktosa, dan total gula dalam batang sawit, nira, dan produk gula merah. Penelitian ini juga bertujuan untuk mengembangkan dan memvalidasi model prediksi kadar gula berbasis data NIR yang mempertimbangkan berbagai faktor yang memengaruhi kualitas nira. Hasil dari penelitian ini diharapkan dapat membantu pengrajin gula meningkatkan efisiensi produksi dan pemantauan kualitas bahan baku secara real-time serta menekan biaya produksi dengan tetap menjaga akurasi analisis. Penelitian ini dilaksanakan di Desa Pegajahan, Kecamatan Pegajahan, Kabupaten Serdang Bedagai dan Laboratorium Departemen Ilmu dan Teknologi Pangan, Fakultas Teknologi Pangan, Institut Pertanian Bogor. Bahan yang digunakan dalam penelitian ini adalah batang, nira, juruh dan gula merah dari batang sawit. Penelitian ini terdiri dari lima tahapan, yaitu persiapan sampel, proses pengukuran destruktif dengan menggunakan HPLC, pengukuran total padatan terlarut dengan refractometer, dan pengukuran non destruktif dengan menggunakan NIR. Hasil data yang diperoleh dikalibrasi dengan metode Partial Least Square Regression (PLSR) menggunakan perangkat bawaan MicroNIR OnSite dan Microsoft Office Excel. Model yang diperoleh dievaluasi menggunakan parameter statistik diantaranya nilai koefisien determinasi (R2), konsistensi, standard error calibration (SEC), standard error prediction (SEP), root mean squares error of calibration (RMSEC), root mean squares error of cross validation (RMSECV), limit of detection (LOD), dan limit of quantification (LOQ). Model Partial Least Square Regression (PLSR) berbasis NIR menunjukkan potensi untuk memprediksi kadar sukrosa, glukosa, fruktosa, dan total padatan terlarut (TPT) pada batang, nira, dan gula merah dari batang sawit. TPT pada nira dan juruh memberikan hasil terbaik dengan nilai R² sebesar 0,9988 dan Standard Error of Prediction (SEP) sebesar 0,7440 °Brix, menunjukkan bahwa NIR sangat efektif untuk analisis TPT secara non-destruktif. Gula monosakarida seperti glukosa dan fruktosa pada nira juga dapat diprediksi dengan akurasi moderat atau R² lebih besar dari 0,85, meskipun kesalahan prediksi (SEP) meningkat pada konsentrasi tinggi akibat interferensi senyawa non-gula. Terdapat keterbatasan model yang diperoleh pada batang sawit dan gula merah. Sampel batang memiliki akurasi model yang rendah yaitu R² kurang dari 0,7 karena heterogenitas sampel dan dominasi senyawa structural seperti lignin dan selulosa yang mengganggu spektrum NIR. Berbeda dengan sampel gula merah, meskipun nilai R² yang diperoleh tinggi untuk glukosa yaitu 0,99128, kesalahan prediksi (SEP) = 7,10252 g/100g dan saturasi spektrum pada sukrosa (SEP) = 14,4452 g/100g menunjukkan keterbatasan instrumen dalam mendeteksi matriks padat berkadar gula tinggi. Hal ini dapat disebabkan oleh rentang panjang gelombang spektrometer yang terbatas yaitu 950–1650 nm yang tidak mencakup puncak absorpsi kunci ikatan C-O dan O-H pada rentang 1700–2500 nm.
Oil palm is a major commodity in Indonesia with abundant production of palm trunk waste, especially from the replanting process. This waste has the potential to be utilized as a source of nira (juice of palm trunk) for palm sugar production. However, sugar artisans still face obstacles in accurately measuring sugar content due to limited access to analytical methods such as High-Performance Liquid Chromatography (HPLC), which are expensive and destructive. This study aimed to assess the potential use of Near-Infrared Spectroscopy (NIR) as a fast, non-destructive, and economical alternative method for quantitative analysis of total soluble solids, sucrose, glucose, fructose, and total sugar content in palm trunks, sap, and brown sugar products. The research also aimed to develop and validate a sugar content prediction model based on NIR data that considers various factors that affect nira quality. The results of this research were expected to help sugar craftsmen improve production efficiency and real-time monitoring of raw material quality and reduce production costs while maintaining analysis accuracy. This research was conducted in Pegajahan Village, Pegajahan Subdistrict, Serdang Bedagai Regency and the Laboratory of the Department of Food Science and Technology, Faculty of Food Technology, Bogor Agricultural University. The materials used in this study were palm trunks, sap, juruh and brown sugar from palm trunks. This research consists of five stages, namely sample preparation, destructive measurement process using HPLC, measurement of total soluble solids with a brix refractometer, and non-destructive measurement using NIR. The data obtained will be calibrated using the Partial Least Square Regression (PLSR) method using MicroNIR OnSite and Microsoft Office Excel. The model obtained will be evaluated using statistical parameters including the coefficient of determination (R2), consistency, standard error of calibration (SEC), standard error of prediction (SEP), root mean squares error of calibration (RMSEC), root mean squares error of cross validation (RMSECV) limit of detection (LOD), and limit of quantification (LOQ). The NIR-based Partial Least Square Regression (PLSR) model showed potential for predicting sucrose, glucose, fructose, and total soluble solids (TPT) levels in trunk, juice, and palm sugar. TPT in juice and juruh gave the best results with an R² value of 0.9988 and Standard Error of Prediction (SEP) of 0.7440 °Brix, indicating that NIR is very effective for non-destructive analysis of TPT. Sugars (mono saccharide) such as glucose and fructose in the juice could also be predicted with moderate accuracy or R² greater than 0.85, although the prediction error (SEP) increased at high concentrations due to interference from non-sugar compounds. There are limitations to the models obtained for palm trunks and brown sugar. The stem samples had low model accuracy of R² less than 0.7 due to sample heterogeneity and the dominance of structural compounds such as lignin and cellulose that interfere with the NIR spectrum. In contrast to the brown sugar samples, although the R² value obtained was high for glucose at 0.99128, the prediction error (SEP) = 7.10252 g/100g and spectrum saturation at sucrose or SEP = 14.4452 g/100g showed the limitations of the instrument in detecting high sugar content solid matrices. This could be due to the limited spectrometer range of 950-1650 nm which does not include the key absorption peaks of C-O and O-H bonds in the 1700-2500 nm range.
URI: http://repository.ipb.ac.id/handle/123456789/165165
Appears in Collections:MT - Agriculture Technology

Files in This Item:
File Description SizeFormat 
cover_F2501231054_722135166dc046c6858719c2fcd29a2a.pdfCover597.17 kBAdobe PDFView/Open
fulltext_F2501231054_6ea178dc8ed64d8c81f5ce259cb5b216.pdf
  Restricted Access
Fulltext1.41 MBAdobe PDFView/Open
lampiran_F2501231054_625083d19435432ab6ea2110ffe071b5.pdf
  Restricted Access
Lampiran970.9 kBAdobe PDFView/Open


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.