Abstract
•New RON and MON measurements of gasoline surrogates/ethanol blends are reported.•An octane predictive model of ethanol blended to gasoline surrogates is proposed.•The proposed model depends on base fuel properties.•The model can be coupled to the design process of ethanol-based gasoline surrogates.
Ethanol is widely used as a gasoline octane booster, and yet, its blending response is not fully understood. Blending ethanol with gasoline most often leads to a synergistic effect (higher than expected relative to linear blending), but can also lead to a linear or antagonistic blending (lower than expected relative to linear blending). To address the knowledges gap in ethanol blending, this study provides new research octane number (RON) and motor octane number (MON) measurements of ethanol blended with various gasoline surrogates. The first set of blends are ternary mixtures, designed to study the interactions between primary reference fuels (PRFs) and certain gasoline components when blended with ethanol. These gasoline components are 1-hexene, 1,2,4-trimethylbenzene and cyclopentane, which, represents the olefin, aromatic and naphthenes classes in commercial gasoline fuels, respectively. The second set represents multicomponent surrogates for Fuels for Advanced Combustion Engines (FACE) gasolines, developed in our previous work (Badra et al., Applied Energy 2017 p. 778-793). This study developes an octane blending model of ethanol/gasoline surrogates that utilize the new measurements along with datasets available in literature. The model consists of the conventional linear by ethanol molar fraction correlation, with the addition of non-linearity terms that depend on base fuel properties, namely, the octane sensitivity and the mole fraction of gasoline components. The model can predict the octane numbers (RON and MON) of blends containing gasoline surrogates composed of n-heptane, n-pentane, iso-octane, iso-pentane, toluene, 124-trimethylbenzene, cyclopentane, cyclohexane (for RON only) and 1-hexene. To ensure the generality of the developed model and avoid over-fitting, the model is trained using 85% of the available dataset while the remaining measurements have been used to test the model. The proposed model outperforms many ethanol blending models available in literature with 87% of the RON and 83% of the MON measurements being within the reproducibility limits. This model can be integrated to the process of designing gasoline surrogates that contain ethanol.