Numerical Approaches to Determine Cetane Number of Hydrocarbons and Oxygenated Compounds, Mixtures, and their Blends

Abstract

In the present work, we report the development and use of models to predict the cetane number of hydrocarbons and oxygenated compounds, mixtures, and their blends. The study is divided in three steps: (i) the prediction of pure compounds' CN using ML-based approaches, (ii) the development and application of mixing rules, and (iii) the external validation of models on a set of real fuels. Experimental CN values for 658 pure compounds are collected from the literature and merged to obtain a consistent and comprehensive database. ML-based models are then trained on the database. A second database is built from the collection of 572 experimental CN values for mixtures. Existing and proposed mixing rules powered by either experimental CN or CN predicted using the ML-based models are then assessed on the basis of the second database. The new mixing rule involving the activity coefficients of mixtures' components shows the best performance. Finally, the application of our predictive numerical approach to 27 real fuels demonstrates its accuracy and relevance, and that it could be further used for testing large numbers of samples.