Using Compact Proton Nuclear Magnetic Resonance at 80 MHz and Vibrational Spectroscopies and Data Fusion for Research Octane Number and Gasoline Additive Determination

Commercial fuels are characterized by parameters, such as research octane number and contents of additives, such as ethanol, ethyl-t-butyl ether, ethyl-tert-methyl ether, olefins, etc. For fast and easy parameter determination without the need for sample preparation, we used compact and benchtop nea...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:Energy & fuels 2020-01, Vol.34 (1), p.103-110
Hauptverfasser: Legner, Robin, Voigt, Melanie, Wirtz, Alexander, Friesen, Anatoli, Haefner, Simon, Jaeger, Martin
Format: Artikel
Sprache:eng
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:Commercial fuels are characterized by parameters, such as research octane number and contents of additives, such as ethanol, ethyl-t-butyl ether, ethyl-tert-methyl ether, olefins, etc. For fast and easy parameter determination without the need for sample preparation, we used compact and benchtop near-infrared (NIR), proton nuclear magnetic resonance (1H NMR) at 80 MHz, and two Raman spectrometers to predict selected relevant fuel parameters of 179 samples known from CFR motor and norm-compliant analyses. Repeatability and reproducibility criteria according to ASTM and ISO norms served as goodness of prediction measures. The prediction relied on partial least squares regression type 1 yielding one target parameter and type 2 yielding simultaneously n target values. While PLS-1 provided more accurate results, PLS-2 might be further applicable to RON and oxygenated additive content determination. Among the methods applied, benchtop Raman and 1H NMR performed best. Low-, mid-, and high-level data fusion were applied to transform pretreated subspectra from up to three individual techniques to result in pseudo-spectra, combined score matrices, or decision models, which further improved the accuracy of the RON prediction. Best results for RON were obtained with mid-level fusion of NIR, NMR, and Raman data yielding 63% of the predicted values within reproducibility of 0.2 and up to 97% within repeatability of 0.7 RON.
ISSN:0887-0624
1520-5029
DOI:10.1021/acs.energyfuels.9b02944