NMR enhances mass-spec FCC feedstock characterization

Hak N. Kim
Stone & Webster Engineering Corp.
Houston

Jan J. Verstraete
Institut Français du Pétrole
Solaize, France

Preetinder S. Virk
Massachusetts Institute of Technology
Cambridge, Mass.

Anne Fafet
Institut Français du Pétrole
Rueil-Malmaison, France

To characterize fluid-catalytic cracking (FCC) feedstock, Stone & Webster Engineering Corp. (SWEC) and Institut Français du Pétrole (IFP) adopted a combination of conventional physical properties coupled with detailed Hreims (high-resolution electron-impact mass spectrometry) and an NMR (nuclear magnetic resonance)-based MIFF (magnetic imaging of FCC feedstock) system.

These combined technologies yielded significantly improved predictions over the conventional approaches for feedstock characterization.

Feedstock characterization is crucial for accurate prediction of FCC yields, especially when processing heavy resids blended with multiple feed streams from various sources.

In general, FCC feedstock properties are dominantly influenced by two factors: crude source and preprocessing history such as atmospheric distillation, vacuum distillation, coking, hydrotreating, hydrocracking, solvent deasphalting, and visbreaking. Combining the effects of these processes with heavy feeds makes reliable feedstock characterization very challenging.

Bulk properties vs. mass spec

FCC feedstock properties can be analyzed many different ways, some of which include bulk physical properties, distillation, type analysis based on molecules, and structural analysis based on atoms.

Some typical bulk properties are API gravity, K-factor, elemental analysis (S, H, C, O, N), metals content (V, Ni, Fe, Na), Conradson carbon residue (CCR), refractive index, and aniline point.

Although these bulk properties are important, they are simply averaged values. Thus, they are unable to reveal the molecular structure or the distribution of properties in different boiling fractions. The distribution of physical properties can be severely distorted by preprocessing or blending with different feedstocks. Distillation data are helpful for estimating the distribution of physical properties, but not for identifying the molecular structure.

Table 1 [51,409 bytes] illustrates why relying on bulk properties alone can be misleading. Feedstocks from three commercial units, A, B, and C, show very similar bulk properties. Feeds B and C are almost indistinguishable. From these bulk properties, one might speculate the product yields of each feed to be similar.

The mass-spec analysis indicates significant differences. Feeds A and B have similar saturates contents but a large difference in monoaromatics and di+ aromatics content. Feed A has 5% more di+ aromatics than Feed B. Di+ aromatics can increase the cycle oil yield considerably and thereby lower conversion.

Feeds B and C differ in aromatics and saturates contents. In addition, the split between monoaromatics and di+ aromatics is different. Feed B has 6.8% more di+ aromatics than Feed C, which may lower Feed B conversion substantially.

Why are the molecular structures important in catalytic cracking? The FCC yields depend on molecular structures. Studies indicate that cracking patterns differ between the large and small molecules, the branched and straight-chain molecules, and the core aromatics side chains.

Lab analyses show that gasoline consists mostly of olefins, monoaromatics, and paraffins of smaller molecular weights. Coke and slurry oil contain mostly tri+ aromatics. An FCC heat balance also depends on molecular-type reactions. Cracking, dealkylation, and dehydrogenation are endothermic reactions, while hydrogenation, cyclization, and condensation are exothermic reactions.

Over the years, SWEC and IFP have accumulated libraries of extensive feedstock properties and operating data from the licensed FCC units, which have processed widely varying crude types such as Mideast, Indonesian, North Sea, Chinese, and domestic Midcontinent crudes. These feedstocks range in character from exceptionally paraffinic (waxy) to naphthenic to aromatic.

Table 2 [42,229 bytes] shows typical mass-spec analyses for vacuum-gas oil (VGO) and vacuum-tower bottoms (VTB) samples. Of major significance in the mass-spec analysis is the subtotal of saturates and monoaromatics, called the conversion precursors. The conversion precursors have the highest probability of cracking and forming materials boiling below 430° F. As a result of hydrogen-transfer reactions, however, this material can also end up with cycle oil and coke through formation of di+ aromatics.

In the mass-spec analysis, the decanted oil (DO) and coke precursors include all the tri+ aromatic compounds. These precursors usually dealkylate, but do not crack, and therefore primarily form DO and coke. The diaromatic compounds are primary contributors to light cycle oil (LCO).

While the mass-spec analytical technique provides vastly better yield-prediction capability over the conventional feedstock characterization techniques (based on bulk properties), it has some drawbacks.

The mass-spec method identifies molecules by "Z-number," and detects homologous series of molecules therein under groupings such as paraffins, cycloparaffins, and aromatics. As a result, under the same Z-number, it cannot differentiate olefins from cycloparaffins, nor can it differentiate aromatics with side chains from aromatics of higher condensation levels. It tends to include potentially crackable side chains attached to aromatics into the aromatics group, thereby understating the cracking potential of the feedstock. It could also overstate the cracking potential of some long-chain paraffins that may be uncrackable.

To overcome such drawbacks, several analytical techniques can be helpful for detecting the core structure of molecules. Some of those techniques are NMR, UV (ultraviolet), and HPLC (high performance liquid chromatography). There are also some correlations available such as n-d-M (refractive index, density, molecular weight) or e-d-M (elemental analysis, density, molecular weight) methods, which are derived from other physical properties.

NMR analysis

SWEC uses a C13-NMR technology called MIFF, which has been developed over the years in collaboration with Massachusetts Institute of Technology. This MIFF method can quantify the aromatic core carbon and thereby augment the mass-spec analysis. A combination of mass spec and MIFF analyses was more powerful than any scheme based on one analytical technique.

In MIFF, the carbon atoms are separated into aromatic carbon and aliphatic carbon, each of which is then divided into more details based on the position and relationship with neighboring atoms, as shown in Table 3 [54,322 bytes]. Combining these analyses, the carbon atoms are characterized into four groups: Cna (carbon associated with n-alkanes), Cma (carbon associated with methyl alkanes), Ccs (carbon associated with aliphatic cyclic and highly substituted atoms), and Car (carbon associated with aromatics).

MIFF also provides an analysis of hydrogen atoms, the distribution of alkanes, and the "length" of n-alkanes, which could relate to the crackability of long paraffinic molecules.

Fig. 1 [80,836 bytes] compares mass-spec vs. NMR analyses of the same sample. According to the mass-spec data, the sum of monoaromatics and di+ aromatics is 46% (13% + 33%); MIFF data suggest that the core aromatic carbons are 17%. Thus, the mass-spec aromatic groups have about 29% (46-17%) aliphatic carbon, which is more than half the total.

The portion of this aliphatic carbon that is associated with the di+ aromatics represents potential FCC cracking material. The NMR technique is especially useful for feedstocks with a large amount of aromatics where the conversion precursors are understated when mass spec alone is used.

Yield estimate comparison

Table 4 [49,729 bytes] illustrates an application of NMR to improve the quality of FCC yields prediction. When the mass-spec data were used alone in the FCC model, the yield estimates were significantly worse than the actual plant data. The information from NMR analysis revealed that a sizable amount of material seemed to be crackable; hence, the fraction of conversion precursors in the feed was modified accordingly.

When this corrected amount of conversion precursors was used in the FCC model, the estimated yields improved considerably and were very close to the actual plant data.

The conversion estimates from mass-spec analysis do not always under-predict but occasionally over-predict. In such cases, the plant data show a much higher yield of DO. This could indicate that some conversion precursors analyzed by mass spec may be uncrackable.

Reliability of mass spec

In view of the importance of mass spec and NMR technology applications in yield predictions, the issue of quality and reproducibility was addressed. A total of 10 feedstock samples were collected from commercial FCC units. The samples included gas oils and resids with gravity ranging from 15 to 26° API and CCR from 0 to 9 wt %. The analytical results were compared between SWEC-MIT and IFP-Total data.

It should be noted, however, that the methods used in the various labs differ, because they have been developed independently. This therefore makes the comparison all the more severe.

Both SWEC and IFP use a mass-spec analysis method that is analogous to the one described by Fisher.1 The analyses of paraffins between SWEC and IFP laboratory show a good agreement (Fig. 2a [112,982 bytes]). The dotted line represents an approximate reproducibility error range.

Fig. 2b compares the conversion precursors, which are key parameters for yield predictions. The overall agreement seems reasonable except for one heavy resid feedstock with 15° API and 9% CCR.

Reliability of NMR

SWEC-MIT and IFP-Total both used the NMR-based technology, but the IFP-Total procedure, which was developed by Bouquet and Bailleul,2 took a more rigorous approach in identifying the peaks in spectrum. As a result, the type of information generated by the two methods should be compared with caution.

Fig. 2c favorably compares the aromatic carbon analyses by SWEC-MIT and IFP-Total. One heavy resid feedstock is far outside the error range.

Fig. 2d compares the fraction of saturated methyl carbon atoms by the two methods. It was generally observed that some other carbon categories did not compare very well as a result of differences in the identification method, but the overall comparison seemed acceptable in most cases.

Because different labs can generate different analytical results for mass spec and NMR, special care must be exercised for correct interpretation and consistency. Areas of improvement in the current feed analyses will include more details of the mono, di, and tri+ aromatics and further refinement in accuracy and consistency of the analysis methods.

Acknowledgment

The authors would like to thank S. Gautier for the NMR analyses at IFP and for his helpful discussions. Preetinder S. Virk is grateful to the staff of the MIT spectrometry laboratory for its help.

References

1. Fisher, I.P., "Residuum catalytic cracking: Influence of diluents on the yield of coke," Fuel, 1986, Vol. 65, No. 4, pp. 473-79.
2. Bouquet, M., and Bailleul, A., "Routine method for quantitative 13C NMR spectra editing and providing structural patterns," Fuel, 1986, Vol. 65, pp. 1240-46.

The Authors