NIR Spectroscopy-Based Qualitative and Quantitative Detection of Adulteration of Peanut Oil

Abstract

In this study, qualitative and quantitative analyses of peanut oils adulterated with waste oil, soybean oil, corn oil and canola oil were performed by combining near infrared (NIR) spectroscopy with chemometrics methods. With NIR spectrometer, the spectra of 108 adulterated peanut oil samples were collected. The collected data were preprocessed with three different preprocessing approaches, including standard normal variate (SNV) transformation, de-trending (DT), and orthogonal signal correction (OSC). Backward interval partial least squares (BiPLS) was used to extract the characteristic wavelengths of the preprocessed data. On the intervals of the full spectrum and the characteristic wavelengths, support vector classification qualitative discriminant models and support vector regression quantitative analysis models were established. Experiments demonstrated that the established qualitative models could accurately determine the type of adulterant in peanut oil, with the accuracies on both the calibration set and the prediction set reaching 100%. With the quantitative models, the percentage of adulteration (3% ~ 55%) could be determined accurately, with correlation coefficients as high as 99.10%. All the models returned prediction root mean square errors lower than 6.96E-4. It was validated that the combination of NIR spectroscopy with chemometrics methods can realize the qualitative and quantitative detection of adulteration of peanut oil.