Current challenges, such as the reduction of fuel consumption and CO2 emissions, motivate research on suitable fuel candidates for future engines. The early selection of fuel components from a wide range of candidates is possible only if global performance metrics for the desired applications are available without exhaustive research efforts. For example, the well-established anti-knock indices research octane number (RON) and motor octane number (MON) as well as the related quantities of octane index (OI) and octane sensitivity (OS) are commonly used to estimate the knock resistance of a fuel. These can serve as ranking criteria to select promising fuel subsets for further investigation. In addition to these key performance indicators, sets of constraints on other fuel properties, such as density, boiling point, or melting point, may be necessary to ensure the basic applicability in engines or the compliance with current fuel norms.

To allow for the selection of fuel components for spark-ignition engine applications, a forward fuel selection approach has been applied at the Institute for Combustion Technology. A large fuel database has been established for this purpose [1]. Since certain key fuel properties, such as RON and MON, are usually not available for all fuel candidates, artificial neural networks were applied in form of quantitative structure-property relationship (QSPR) models. These QSPR models were developed based on both the available experimental data, serving as training dataset, as well as the chemical knowledge on fuel combustion based on which tailored molecular descriptors were defined as model input parameters [2]. A Python code for model execution is available here. The models were then used to obtain the property values for all fuel components in the database [1], if no experimental values were available.

The fuel property database is available in the Supplementary Material of the corresponding article [1] and here. In the case of future updates, an updated version of the database will be made available on this webpage.

[1] F. vom Lehn, L. Cai, R. Tripathi, R. Broda, H. Pitsch, A property database of fuel compounds with emphasis on spark-ignition engine applications, Applications in Energy and Combustion Science 5 (2021) 100018.

[2] F. vom Lehn, B. Brosius, R. Broda, L. Cai, H. Pitsch, Using machine learning with target-specific feature sets for structure-property relationship modeling of octane numbers and octane sensitivity, Fuel 281 (2020) 118772.