Pendugaan Indeks Fermentasi, pH, dan Kadar Air pada Biji Kakao Kering Terfermentasi Menggunakan Portable NIR Spectrometer

Abstract

Biji kakao banyak yang belum melewati proses fermentasi setelah pemanenan. Biji tersebut dicuci dan langsung dikeringkan. Oleh karena itu fermentasi perlu dilakukan pada biji kakao yang sudah kering agar mutu biji kakao kering dapat meningkat. Parameter penting seperti indeks fermentasi, pH, dan kadar air menunjukkan mutu biji kakao. Portable NIR Spektrometer SCiO dapat memprediksi kandungan kimia dengan biaya rendah serta persiapan sampel yang minimal. Penelitian ini bertujuan menduga kandungan pada biji kakao kering terfermentasi dengan menggunakan Portable NIR Spectrometer dan membandingkan pra-pengolahan data spektra NIR [De-trending (DT), The Savitzky-Golay first and second derivative (SG1 & SG2), Linear Baseline Correction (LBC), Standard Normal Variate (SNV) dan Multiplicative Scatter Correction (MSC)] dengan metode kalibrasi Partial Least Square (PLS) berdasarkan parameter koefisien korelasi (r), standard error prediction (SEP), standard error calibration (SEC), coeficient of variation (CV), ratio of prediction to deviation (RPD). Pengambilan data spektrum NIR dan pengukuran data destruktif untuk 44 sampel biji kakao kering terfermentasi dengan variasi lama fermentasi (0, 24, 48, 72 jam). Pengukuran indeks fermentasi, pH, dan kadar air menggunakan UV-Vis, pH meter, dan metode oven. Fermentasi biji kakao selama 48-72 jam menghasilkan biji dengan mutu terbaik dalam hal indeks fermentasi dan pH. Hasil penelitian menunjukkan bahwa model terbaik untuk memprediksi pendugaan indeks fermentasi menggunakan metode PLS dengan pra-pengolahan LBC pada faktor 13 (r = 0,821; SEC = 0,097; SEP = 0,102; CV = 7,48 %; RPD = 1,562 dan konsistensi = 95,09 %). Pendugaan pH terbaik dengan SG2 dengan polinomial order 7 dan faktor PLS 6 (r = 0,907; SEC = 0,193; SEP = 0,212; CV = 3,40 %; RPD = 2,129 dan konsistensi = 90,93 %). Pendugaan kadar air terbaik dengan MSC pada faktor PLS 15 (r = 0,847; SEC = 0,293%; SEP = 0,353%; CV = 6,48 %; RPD = 1,534 dan konsistensi = 83,18 %). Metode pra-pengolahan data spektra NIR dengan DT, SG1 & SG2, LBC, SNV, dan MSC terbukti memperbaiki model kalibrasi dibandingkan data original.

Many cocoa beans have not gone through the fermentation process after harvesting. The seeds are washed and dried immediately. Therefore, fermentation needs to be done on dried cocoa beans so that the quality of dried cocoa beans can increase. Important parameters such as fermentation index, pH, and moisture content indicate the quality of cocoa beans. The SCiO Portable NIR Spectrometer can predict chemical content at a low cost as well as minimal sample preparation. This study aims to estimate the content of fermented dried cocoa beans using a Portable NIR Spectrometer and compare pre-processing NIR spectra data [De-trending (DT), The Savitzky-Golay first and second derivative (SG1 &SG2), Linear Baseline Correction (LBC), Standard Normal Variate (SNV) and Multiplicative Scatter Correction (MSC)] with Partial Least Square (PLS) calibration method based on the parameter of coefficient of correlation (r), standard error prediction (SEP), standard error calibration (SEC), coefficient of variation (CV), the ratio of prediction to deviation (RPD). NIR spectrum data collection and destructive data measurement for 44 fermented cocoa bean samples with fermentation duration variations (0, 24, 48, 72 hours). Measurement of fermentation index, pH, and moisture content using UV-Vis, pH meter, and oven method. Fermentation of cocoa beans for 48-72 hours produces the highest quality beans in terms of fermentation index and pH. The results showed that the best model to predict the estimation of fermentation index using the PLS method with LBC preprocessing at a factor of 13 (r = 0.821; SEC = 0.097; SEP = 0.102; CV = 7.48 %; RPD = 1.562 and consistency = 95.09%). Best pH estimator with SG2 with polynomial order 7 and PLS factor 6 (r = 0.907; SEC = 0.193; SEP = 0.212; CV = 3.40 %; RPD = 2.129 and consistency = 90.93%). Best moisture content estimator with MSC at PLS factor 15 (r = 0.847; SEC = 0.293%; SEP = 0.353%; CV = 6.48 %; RPD = 1.534 and consistency = 83.18%). The method of pre-processing NIR spectra data with DT, SG1 &; SG2, LBC, SNV, and MSC has been shown to improve calibration models compared to the original data.