Estimation of middle distillate fuel properties by FT-IR

As an alternative to established laboratory protocols, partial least-squares (PLS) regression models based on Fourier transform infrared (FT-IR) spectra were developed for the rapid and simultaneous determination of several middle distillate fuel properties. Requiring less than 2 mL of sample, the PLS models use a single wavelength range of baseline-uncorrected FT-IR spectra. Several preprocessing parameters were investigated, including mean-centering of the raw spectra and mean-centering of first- and second-difference spectra. In addition, modification to the calibration parameters included varying the number of cross-validation rotation samples from 5 to 50, applying a baseline correction to the spectra, and using restricted wavelength regions. In order to obtain an accurate estimate of the prediction error, we applied the method provided by Faber and Kowalski [Appl. Spectrosc. 51, 660 (1997)] that corrects for the measurement error in the reference values. On the basis of this work, the following middle distillate fuel properties may be confidently estimated by FT-IR: gravity, API degrees; density, g/mL; kinematic viscosity, cSt, 40 degrees C; boiling point at 50% recovery, degrees C; cetane index; carbon, wt %; hydrogen, wt %; carbon-to-hydrogen ratio, C/H; net heat of combustion, MJ/Kg (or BTU/1b); monocyclic aromatics, wt %; dicyclic aromatics, wt %; polycyclic aromatics, wt %; and total aromatics, wt %. Estimation of several other properties is also possible where reduced precision is acceptable.