Predicting the physical and chemical properties of sustainable aviation fuels using elastic-net-regularized linear models based on extended-wavelength FTIR spectra

FUEL(2024)

引用 0|浏览1
暂无评分
摘要
A Fourier transform infrared (FTIR) spectra-based prescreening approach was developed for estimating nine important physical and chemical properties of sustainable aviation fuels (SAFs): molecular weight (MW), hydrogen-to-carbon (H/C) ratio, density, net heat of combustion (NHC), derived cetane number (DCN), threshold sooting index (TSI), initial boiling point (IBP), flash point, and kinematic viscosity (KV). A training dataset containing the vapor-phase FTIR absorption spectra of pure hydrocarbons spanning four molecular classes, blends of neat hydrocarbons, conventional jet fuels and SAFs, across the 2-15.38 & mu;m wavelength range, was compiled. The physical and chemical properties of the fuels in the training dataset were sourced either from experimental data reported in the literature or calculated using relevant property blending correlations. Elasticnet-regularized linear models were trained for each property by optimizing the model parameters using a crossvalidated grid search. The results from these models were compared with the results from previous models (Lasso-regularized linear models developed previously by Wang et al. at Stanford), which were trained on FTIR absorption spectra across the limited wavelength range of 3.3-3.55 & mu;m. Use of the extended wavelength range and the new model-parameter optimization strategy results in significant improvement in predictive performance of the current models compared to the previous models for all nine properties. The model performance is also evaluated on three candidate SAFs. The current models are found to achieve higher property prediction accuracy than the previous models for these test fuels. Overall, this work demonstrates the utility of predictive models based on extended-wavelength FTIR spectra and their potential use as a low-volume prescreening method for characterizing properties of new sustainable aviation fuel candidates.
更多
查看译文
关键词
Sustainable aviation fuel (SAF),Prescreening,FTIR spectroscopy,Machine learning,Property prediction
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络
Chat Paper
正在生成论文摘要