Exploring relaxation features for rapid identification of olive oil adulteration via low-field nuclear magnetic resonance integrated with chemometrics.

Abstract

Low-field nuclear magnetic resonance (LF-NMR) is a non-destructive analytical technique utilized for relaxation spectral fingerprinting, which shows great potential in detecting commercial food fraud. This study developed an analytic framework to identify olive oil adulteration by combining LF-NMR with chemometrics. Relaxation parameters were extracted from four types of pure oils, three binary blends, and three ternary blends. The differential parameters were visualized using volcano plots and canonical correlation analysis (CCA) to identify critical features. Subsequently, multiple machine learning models were employed to classify the ten types of oil samples and quantify the olive oil content in the mixtures. The extreme gradient boosting (XGBoost) model demonstrated superior performance, achieving an average coefficient of determination (R2) exceeding 0.93. Moreover, the interpretation of the model using the Shapley Additive exPlanations (SHAP) algorithm highlighted the significant contribution of relaxation signals at characteristic peaks to the quantitative analysis. The results confirm that the integration of LF-NMR and chemometrics provides an efficient strategy for the rapid screening of blended olive oils.