Machine learning-driven screening of fuel additives for increased spark-ignition engine efficiency

Fuel design aims to develop and optimize fuels to meet specific performance, environmental, and economic objectives. It encompasses a range of considerations, including selecting appropriate feedstocks, adjusting molecular structures, and incorporating additives to achieve desired characteristics. One critical aspect of fuel design is its relevance to addressing environmental concerns, such as reducing greenhouse gas emissions and air pollutants. In a spark-ignition engine, increasing engine efficiency leads to a reduction in CO2 emissions. The composition of the fuel plays a vital role in enhancing engine efficiency. Anti-knock properties, latent heat of vaporization (HoV), and laminar flame speed (LFS) are some of the fuel properties that can influence engine operating regimes. In the current study, we explore additives that can improve the efficiency of spark-ignition engines. Machine learning-based quantitative structure-property relationship (QSPR) models are developed to predict research and motor octane numbers, HoV, and LFS of 379,500 hydrocarbons containing only carbon, hydrogen, and oxygen atoms. The molecules are ranked based on an established merit function, and the top five candidates are selected. Methanol is the most promising additive candidate, allowing for the highest degree of efficiency enhancement among the screened candidates. Other potential candidates are substituted furans and tetrahydrofuran.