FT-IR Spectroscopy Determination of Sucrose in Sugarcane Juice Using Standard Solutions

──Comparison of CLS, MLR, and PLSR Calibration Models──

Abstract

　We compared the performances of classical least squares (NAS-CLS), multiple linear regression (MLR), and partial least squares regression (PLSR) for the measurement of sucrose in sugarcane juice using Fourier-transform infrared (FT-IR) spectroscopy and standard sugar solutions. In the range of 1500-1200 cm⁻¹, all methods showed high accuracy, with R² values of 0.87-0.95 and root-mean square errors of prediction (RMSEP) of 0.49 %-0.63 %. However, the prediction error increased for samples containing more than 1.0 % fructose. This indicates that the standard solutions used to develop calibration models should consider deteriorated samples. The three methods yielded similar accuracy. Considering calibration model maintenance, the MLR method is useful because it allows explicit selection of informative wavenumbers, and the NAS-CLS method is advantageous because its model can be more easily adapted to account for interfering components, such as glucose and fructose.