Quantifying isomeric effects on metric for fuel impact

Kurzfassung

The aeronautical sector, responsible for about 2 to 3 % of global CO2 emissions, has set an ambitious goal: achieving carbon neutrality by 2050 in order to limit its impact on climate change. In this perspective, sustainable aviation fuels appear as a key solution, offering a renewable alternative capable of significantly reducing emissions while integrating with existing infrastructure. To maximize the efficiency and minimize the environmental impact of SAF, it is essential to understand in detail how the molecular composition, and more particularly isomeric variations, influence their combustion properties and emissions. With this in mind, the DLR is currently building a tool called simfuel to analyze, design and optimize fuels for aeronautics. One of the main features of simfuel is to be a database consisting of complex fuel but also pure molecule data. We will therefore only use the database composed of pure molecules as part of this report. This database is based on experimental data collected in the University of Washington database but also with inter- nal data from the DLR. Our study is based on the following physico-chemical properties: density, kinematic viscosity, surface tension and cetane number. Since all these properties are not available in the initial databases, using a model to predict these properties allows us to solve this problem. Thus the method ’Mean Quantitative Structure-Property Relationship method’ which will be called MQSPR allows to overcome this problem. This method allows starting from the numbering of 49 atomic particulate structures which we will call MQSPR representation to predict a possible value of the different missing proper- ties. We also need to know the evaporation rate of the different molecules studied; for this we used an internal calculation code at DLR: Spraysim. Spraysim allows calculating an evaporation time of our molecules when we insert them into the combustion chamber, this allows calculating an evaporation rate if we know the atomization diameter of our fuel at the injection of the combustion chamber. Now that all our fundamental properties have been estimated, we can detail the models used to address our problem of quantifying the effects of isomers. Our different models are: the sauter mean diameter which allows estimating the atomization diameter, the fuel consumption limit at lean burn, the maximum distance that our aircraft can travel, water emissions, carbon dioxide emissions and soot emissions, the effects of condensation streaks and finally the total radiative effects due to combustion. We have therefore developed a python code allowing to calculate these different models from fundamental properties of each molecule in the simfuel database. The preliminary results of these models allow us to realize that aromatic families should be avoided to reduce pollutant emissions. To deepen this result, we calculated a Pearson correlation between the data of these models and the MQSPR representation. This highlights the atomic structures that contribute the most to each of our models. Thus the presence of a ring and aromatic carbon (carbon in a ring that has a single bond and a double with other carbon atoms) are the two main structures that contribute most to pollutant emissions. Finally, we looked at the influence of isomer structures on our models within a hydrocarbon family and a fixed carbon number. We compared this gap to the average value with the gap between the families of hydrocarbons as well as between the numbers of carbons. It follows that the type of hydrocarbon influences the models more than the structures of isomers. However, the types of isomers influence nearly twice as much as the number of carbon. Thus, when possible, knowledge of the structure of the isomer provides real additional information. However, knowing the isomer structure of each molecule in a multi-molecule fuel remains very complex to obtain experimentally. The objective of the simfuel project is ultimately to combine this structural knowledge with machine learning approaches to accelerate the rational design of sustainable and efficient fuels.