Skip to main content Skip to secondary navigation

Infrared Spectral Analysis

infrared spectral analysis schematic.
Main content start

A fuel’s infrared absorption spectrum is a unique fingerprint that effectively indicates the various functional groups/bonds present in the fuel.

These functional groups are known to impact the fuel’s physical and chemical properties (e.g., density, boiling point, viscosity, etc.) Consequently, there is a correlation between the properties of a fuel and the presence of certain features in the fuel’s spectrum. Machine learning models excel at identifying patterns and interpreting latent information from raw data, and hence are well-suited for the task of estimating the properties of a fuel directly based on its IR spectrum. We have successfully applied different classes of machine learning models, including regularized linear models [1,2], support vector regression (SVR), and quadratic programming to predict the properties and molecular composition of aviation-relevant fuels.

By fine-tuning the model parameters for specific applications, these techniques can be further extended to predict the combustion and emissions characteristics of sustainable fuel candidates, thereby aiding in the design of novel fuels.

a flow chart for property predictions
Flow chart of Strategy 1, i.e., Lasso-regularized linear model for property prediction based on FTIR spectra [3].

To learn more, check out some of our publications:

[1] Y. Wang, Y. Ding, W. Wei, Y. Cao, D. F. Davidson, R. K. Hanson, “On estimating physical and chemical properties of hydrocarbon fuels using mid-infrared FTIR spectra and regularized linear models,” Fuel, Vol. 255 (2019) 115715. DOI: 10.1016/j.fuel.2019.115715

[2] Y. Wang, W. Wei, Y. Zhang, R. K. Hanson, “A new strategy of characterizing hydrocarbon fuels using FTIR spectra and generalized linear model with grouped-Lasso regularization,” Fuel, Vol. 287 (2021) 119419. DOI: 10.1016/j.fuel.2020.119419

[3] Y. Wang, “On the use of infrared spectroscopy and statistical learning with sparsity to characterize hydrocarbon fuels,” PhD thesis, Stanford University (2020)