Predicting Fuel Ignition Quality Using 1H NMR Spectroscopy and Multiple Linear Regression

Abdul Gani Abdul Jameel, Nimal Naser, Abdul-Hamid M. Emwas, Stephen Dooley, Mani Sarathy

Research output: Contribution to journalArticlepeer-review

51 Scopus citations

Abstract

An improved model for the prediction of ignition quality of hydrocarbon fuels has been developed using 1H nuclear magnetic resonance (NMR) spectroscopy and multiple linear regression (MLR) modeling. Cetane number (CN) and derived cetane number (DCN) of 71 pure hydrocarbons and 54 hydrocarbon blends were utilized as a data set to study the relationship between ignition quality and molecular structure. CN and DCN are functional equivalents and collectively referred to as D/CN, herein. The effect of molecular weight and weight percent of structural parameters such as paraffinic CH3 groups, paraffinic CH2 groups, paraffinic CH groups, olefinic CH–CH2 groups, naphthenic CH–CH2 groups, and aromatic C–CH groups on D/CN was studied. A particular emphasis on the effect of branching (i.e., methyl substitution) on the D/CN was studied, and a new parameter denoted as the branching index (BI) was introduced to quantify this effect. A new formula was developed to calculate the BI of hydrocarbon fuels using 1H NMR spectroscopy. Multiple linear regression (MLR) modeling was used to develop an empirical relationship between D/CN and the eight structural parameters. This was then used to predict the DCN of many hydrocarbon fuels. The developed model has a high correlation coefficient (R2 = 0.97) and was validated with experimentally measured DCN of twenty-two real fuel mixtures (e.g., gasolines and diesels) and fifty-nine blends of known composition, and the predicted values matched well with the experimental data.
Original languageEnglish (US)
Pages (from-to)9819-9835
Number of pages17
JournalEnergy & Fuels
Volume30
Issue number11
DOIs
StatePublished - Sep 30 2016

Fingerprint

Dive into the research topics of 'Predicting Fuel Ignition Quality Using 1H NMR Spectroscopy and Multiple Linear Regression'. Together they form a unique fingerprint.

Cite this