Noncompliant gasoline compromises engine performance, durability, and emissions. In this study, a virtual electronic tongue combining electrochemical impedance spectroscopy (EIS) and artificial neural networks (ANNs) was applied to identify and quantify common gasoline adulterants, namely, n-hexane, toluene, and mineral turpentine, in single- and multiadulterant systems. Measurements were performed using a glassy carbon electrode with platinum counter and pseudoreference electrodes. Single-adulterant systems exhibited increasing Nyquist semicircle diameters in the order n-hexane < mineral turpentine < toluene, while binary and ternary mixtures showed nonmonotonic impedance behavior, reflecting concentration-dependent intermolecular interactions. ANN models trained with the imaginary impedance component (Z″) demonstrated improved performance when restricted to the semicircle region of the Nyquist plots. This approach resulted in no misclassifications in the test set for adulterant type and enhanced regression performance for individual adulterants, even in mixed systems (test-set R2 = 0.846 for n-hexane, 0.965 for toluene, and 0.742 for mineral turpentine). These results show that focusing on the semicircle domain concentrates the most informative impedance features and enables reliable identification and quantification of gasoline adulteration while also revealing the inherent complexity of multiadulterant fuel systems.
Cola et al. (Thu,) studied this question.