RESUMEN
Palm oil is widely used in the food industry, and its quality is associated with the free fatty acids (FFA) content. Determination of FFA in oil is time-consuming, requires chemicals and generates residues. There is a trend of applying process analytical technologies (PAT) for fast and nondestructive determination of oil parameters. Portable near-infrared (NIR) spectrometers are cheaper than bench top equipment, and have been used for several tasks in the food processing industry, as it provides fast and reliable data for inline measurements. This study investigated the use of NIR spectra using a portable equipment, combined with both unsupervised and supervised multivariate analyses for identification of palm oil samples with different levels of FFA. Soft independent modeling of class analogy , k-Nearest Neighbors, and linear discriminant analysis models were able to correctly identify 100% of the studied samples with selected wavelengths from NIR spectra. Calibration models were performed for acidity prediction, achieving R2 = 0.97, with root mean square error of prediction = 4.37 for partial least squares model using most relevant wavelengths. These results demonstrate the feasibility of applying a low-cost portable NIR spectrophotometer to predict quality parameters of palm oil. PRACTICAL APPLICATION: This work presents results that show the feasibility of using a low-cost portable near-infrared spectrophotometer for the classification of raw palm oil samples according to free fatty acids contents. Regression models are presented as a fast and nondestructive alternative to classify samples for acidity, which is an important quality parameter and that directly affects the market value of crude palm oil.
