Biofuels are increasingly gaining importance as the demand for fuels increases in the transportation and energy sectors. For example, heavy-duty, aviation, and maritime transportation systems face the most significant challenges, as demand must be balanced with sustainable options. Therefore, biofuels can serve as an alternative for these sectors. A prominent issue is that biofuel molecules lack experimental property data required for systematic evaluation. Additionally, determining the safety and performance of fuels is equally crucial. Properties like boiling point and flash point are primary indicators and one of the most critical properties for determining an engine's performance, efficiency, and the safety of a fuel. To help with the lack of experimental property data, we will develop a machine learning framework for accurately predicting our selected key fuel properties based on molecular structure. The methodology involves a supervised machine learning approach to address the complex relationships between molecular structure and fuel performance characteristics. Machine learning methods are commonly used for their high predictive accuracy and to prevent overfitting when working with complex chemical datasets. By constructing various sub-samples of data and aggregating their outputs, the model effectively reduces variance and improves generalization to tested molecules. The model will be trained on a comprehensive dataset of known compounds and their experimentally verified properties, sourced from the Gelest Chemical Catalogue, the DIPPR Database, Lange’s Handbook of Chemistry, the Hazardous Chemicals Handbook, and the PubChem Chemical Database, to predict critical fuel metrics such as flash point and boiling point. The datasets consist of 10,575 and 6,000 unique chemical molecules, respectively. Ultimately, this predictive tool can find the boiling point and flash point, which represent performance and safety, respectively. The model can design new biofuel candidates that are sustainable and efficient, as well as safe and practical for widespread use.
REFERENCES
Eugene D. Nikitin (2024). Thermophysical properties of the biofuel components: A mini-guide to the critical properties, heat capacities, and thermal conductivities. Fluid Phase Equilibria, 580, 114035.
Restrepo-Flórez, J. M., Ryu, J., Witkowski, D., Rothamer, D. A., & Maravelias, C. T. (2022). A systems level analysis of ethanol upgrading strategies to middle distillates. Energy and Environmental Science, 15(10), 4376–4388. https://doi.org/10.1039/d2ee02202h
Sun, X., Krakauer, N., Politowicz, A., Chen, W., Li, Q., Li, Z., Shao, X., Sunaryo, A., Shen, M., Wang, J., & Morgan, D. (2020). Assessing Graph-based Deep Learning Models for Predicting Flash Point. Molecular Informatics, 39(6).