GC-based prediction of fuel auto-ignition quality

摘要

The early assessment of fuel auto-ignition quality is of paramount importance for any fuel design methodology (cf. discussions in Chapter 1 and in Subsection 2.1.1). Although multiple models for the prediction of pure-component octane number (ON) and cetane number (CN) from molecular structure have been proposed over the course of the last 30 years (DeFries et, al., 1987; Meusinger and Moros, 1999; Yang et al., 2001; Albahri, 2003; Santana et al., 2006; Smolenskii et al., 2008; Lapidus et al., 2008; Katritzky et al., 2010; Creton et, al., 2010; Abdul Jameel et al., 2016), their applicability ranges are largely limited to non-oxygenated hydrocarbon species, i.e., the constituents of fossil fuels. Specifically, the validity of the earlier CN models is restricted to n-alkanes, iso-alkanes and singly substituted alkylbenzenes (DeFries et al., 1987), to alkanes and cycloalkanes (Lapidus et, al., 2008; Smolenskii et al., 2008), or even to iso-alkanes only (Yang et al., 2001). The model of Smolenskii et al. (2008) also returns some strange CN values, e.g., -146.12 for 3-methyl-3-ethylhexane. More recently, Creton et al. (2010) have proposed separate QSPR models for four classes of molecules, i.e., alkanes, cycloalkanes, alkenes and aromatics, and Abdul Jameel et al. (2016) have trained a multiple linear regression model on CN and derived cetane number (DCN) data of 71 hydrocarbons and 54 hydrocarbon blends. The few attempts to extend the range of validity of ON and CN models to oxygenates have generally suffered from a lack of ignition data of adequate quality for the variety of molecular structures. Taylor et, al. (2004), Saldana et al. (2011), Dahmen et al. (2012) and Sennott et, al. (2013a,b) have built QSPR models by gathering data mainly from the experimental CN compendium released by Murphy et, al. (2004), which contains some data on oxygenates. However, only a small set, of CN from this compendium had been measured in an ASTM D613 (2015) cooperative fuels research (CFR) engine, whereas a much larger set of correlated CN had been derived from ertene numbers, from ignition delay data acquired in combustion bomb experiments and from mixture data (Murphy et al., 2004).