Characterization of unconventional fuels from car residue processing: uncertainty quantification of composition prediction

The variations of the raw matter used to produce an unconventional fuel could cause an enginefailure due to a property out of range. Checking the octane numbers over time with standard testmethods is too expensive for a local producer due to the small-scale production and a tight budget.To work around this issue, methods based on an intermediary estimation of the fuel compositionare particularly appealing as they allow to evaluate with good accuracy properties such as octanenumbers from properties easier to measure, e.g. the density, the viscosity and the refractive index[1].As no method exist for unconventional fuels, we developed a model especially calibrated for a fuelproduced from automotive shredder residues [2]. This work aims at investigating the range ofapplicability of this model.2000 simulated fuels were generated according to the latin hypercube sampling (LHS) method andthe uncertainties of the predicted compositions were quantified. The range of applicability of themodel considering all the hydrocarbon groups together was defined by carrying a principalcomponent analysis. The model is accurate in a wide range of hydrocarbon fractions.Further work could be leading to strengthen the validation of the model. A first improvement couldbe achieved by investigating several distributions within the hydrocarbon groups. A secondenhancement aiming at increasing the hydrocarbon group space coverage could be reached withanother sampling method.[1]: Riazi, M.R. “Characterization and Properties of Petroleum Fractions,” (2005),soi:10.1017/CBO9781107415324.004.[2]: Tipler, S. Parente, A. Coussement, A. Contino, F. et al. SAE Technical Paper 2018-01-0905, 2018, doi:10.4271/2018-01-0905.