Non-destructive determination of peroxide value in edible oils using capsule-based near-infrared spectroscopy and chemometrics

Peroxide value (PV) is a key indicator of lipid oxidation and edible oil quality; however, conventional analytical methods for PV analysis are time-consuming and require chemical reagents. This study investigates the combination of near-infrared (NIR) spectroscopy with chemometric modeling as a rapid, non-destructive alternative for PV determination and edible oil discrimination. NIR spectra were acquired using vial- and capsule-based sampling modes, and partial least squares regression models were developed for quantitative PV prediction. Both sampling configurations demonstrated comparable predictive performance, achieving high coefficients of determination in prediction (R²_p = 0.97). Method robustness and practical interchangeability of sampling modes were further evaluated using Bland–Altman analysis, which indicated negligible systematic bias. Based on these results, capsule-based measurements were employed for corn, sunflower, and soybean oils. Due to limited sample sizes, cross-validation was employed for individual oil types; a global calibration was further established using combined oil samples for calibration and external prediction, and showed strong predictive capability. Principal component analysis revealed clear clustering trends according to oil type, reflecting compositional differences among oils. All told, the capsule-based NIR approach eliminates cleaning steps, minimizes cross-contamination, and improves analytical throughput, making it well suited for routine quality control and high-throughput industrial applications.