A spectral-mathematical strategy for the identification of edible and swill-cooked dirty oils using terahertz spectroscopy

• Different types of oils were measured by THz-TDS.
• Features can be observed from the absorbance, while the waveforms were very similar.
• The use of mathematical methods can clearly classify the edible and dirty oils.
• THz-mathematical detection can be a promising selection in food safety field.

Effective technique and procedure based on the terahertz (THz) spectroscopy and mathematical methods, named THz Mathematics (T-Math), have been developed for the precise and rapid identification of edible oil from a series of typical swill-cooked dirty oils in daily life. Differences can be observed in the absorbance spectra of edible and dirty oils, while the waveforms were very similar. Principal component analysis (PCA) and support vector machine (SVM) were employed for the identification of edible or swill-cooked oil classification. Compared with dirty oils, normal edible oil had the largest first principal component (PC 1) score with a contribution rate of 97.4%. The PC 1 score deviations of edible oil and dirty oils varied from ∼0.5 to 2.3. Meanwhile, by using SVM leave-one-out cross-validation, the swill-cooked oils can be directly identified with the precision of 100%. This study proved that the developed method was very suitable for the rapid determination of swill-cooked oils in food safety field.