FOURIER TRANSFORM MID-INFRARED PHOTOACOUSTIC SPECTROSCOPY (FTIR-PAS) COUPLED WITH CHEMOMETRICS FOR NON-DESTRUCTIVE DETERMINATION OF OIL CONTENT IN RAPESEED

摘要

This article presents the application of Fourier transform mid-infrared photoacoustic spectroscopy (FTIR-PAS) coupled with chemometric techniques for rapid and non-destructive determination of oil content in rapeseed. The oil content of 150 rapeseed samples, as measured by Soxhlet extraction, were used as reference values for model calibration by partial least squares (PLS). The full-spectrum based PLS model achieved a root-mean-square error of prediction (RMSEP) of 1.11% and residual predictive deviation (RPD) of 2.20. Improved prediction results were obtained by performing variable selection based on ordered predictors selection (OPS) and competitive adaptive reweighted sampling (CARS). The best accuracy was achieved with the OPS based-PLS model, with RMSEP of 0.87% and RPD of 2.79.