Ted R. Aulich, Xinming He, Curtis L. Knudson
Energy & Environmental Research Center
Grand Forks, N.D.
With restrictions on gasoline vapor pressure tightening in the U.S., and increasingly the world over, refiners are facing the possibility, of having their gasolines tested for Rvp, oxygenate content, or other qualities, while in transport.
A recent study by the Energy & Environmental Research Center-a nonprofit, contract research arm of the University of North Dakota-showed that sampling technique and location can affect whether the gasoline in a tanker truck meets specifications.
Given the quality discrepancies found among gasoline samples taken at different points in the transportation process, refiners should make sure that adequate testing is performed and documented before a fuel shipment leaves the refinery. Additionally, refiners can involve themselves in the regulatory process and inform legislators of the importance of sampling technique to testing accuracy.
Since these tests were performed, the U.S. Environmental Protection Agency (EPA) has issued two recommended sampling methods. These Guidelines also will be important to refiners outside the U.S., who may face this sort of sampling now, or in the future.
GASOLINE TESTING
An analysis of regular unleaded gasoline sampled from the top and bottom of commercial bulk transport tanks showed that the Rvp of the top samples was higher than that of the bottom samples. Moreover, the lighter (more volatile) gasoline components were more concentrated in the top samples than in the bottom samples.
The findings of the investigation--funded by the U.S. Department of Agriculture--may impact how fuels are sampled for the purpose of ensuring compliance with reformulated gasoline Rvp specifications, as mandated by the Clean Air Act Amendments.
Gasoline samples were obtained from commercial transport trucks about 30 min after the tanks wei,loaded at a pipeline terminal and again from the same tanks after being transported 60 miles to a gasoline station. Three tanks were sampled:
- Tank 1-3,200 gal, 98.4% full of unleaded regular
- Tank 2-2,000 gal, 90.0% full of unleaded regular
- Tank 3-4,200 gal, 83.3% full of 10 vol % ethanol/90 vol % unleaded regular.
Two samples were obtained from each tank at each sampling: a "top" and a "bottom" sample. The top samples were taken through the fill hole at the top of the tank by submerging a quart jar encased in a metal frame to a level just below the surface of the gasoline. The bottom samples were taken from the tank drain after flushing approximately 1 gal of gasoline from the tank.
The National Alternative Fuels Laboratory at the Energy & Environmental Research Center at Grand Forks, N.D., analyzed the samples. The analyses were performed as part of an investigation of the effects of bulk transport on gasoline composition.
The before and after-transport samples were compared, as were the top and bottom samples, on the bases of Rvp and gas chromatograph (GC) component analysis.
Tables 1 and 2 list temperature and Rvp data for the samples. In the tables, "Pipeline" refers to samples obtained from the transport tanks after being filled at the pipeline terminal. "Station" refers to samples obtained from the transport tanks after they arrived at the gas station and before they were unloaded into underground storage tanks.
All Rvp determinations were made using a Grabner CCA-VPS portable vapor pressure tester. The ethanol content difference between the top and bottom samples (Table 2) is caused by the splash blending procedure used, which consisted of pumping 350 gal of denatured ethanol into the 4,200-gal tank, followed by 3,150 gal of unleaded regular gasoline.
Because the ethanol was added first, its concentration in the 4-in. tank drain pipe (in which mixing is likely inhibited, relative to the main tank compartment) was much higher than in the main compartment. The higher ethanol content in the station top sample, as compared to the pipeline sample, is caused by mixing during transport.
Because the density of ethanol is higher than that of gasoline (0.79 g/ml and 0.71 g/ml, respectively), 9.3 wt % ethanol (Table 2) is roughly equivalent to 8.5 vol %. Because sufficient ethanol was added to yield a 10 vol % mixture, this low result indicates that, following the 60-mile transport, the blending of ethanol and gasoline was incomplete.
Fig. 1 compares the Tank 1 top and bottom samples in terms of the concentrations of their major components. While the concentration values were calculated based on the total sample detected, including minor components, only components present in concentrations greater than 0.5% (on a GC area-percent basis) are shown.
"GC boiling point index" values, which normally vary 1-3 C. from the actual boiling points, were calculated using a simulated distillation technique.1 This technique uses the boiling points of the series of C4-C15 normal alkanes to relate GC retention time to boiling point.
For example, a compound with a retention time halfway between the retention times of pentane and hexane is assigned a boiling point index of 52.5 C., which is halfway between 36 C. and 69 C. --the boiling points of pentane and hexane, respectively.
The technique yields fairly accurate boiling points for non-highly branched aliphatics and olefins, which are similar to normal alkanes in chromatographic behavior. It yields slightly less accurate boiling points for aromatics and highly branched aliphatics, which behave a little less like normal alkanes in the chromatograph.
For example, 3-methyl-pentane- a single-branch aliphatic with a boiling point of 63.3 C.-is assigned a boiling point index of 62.6 C. On the other hand, benzene-an aromatic with a boiling point of 80.1 C. -is assigned a boiling point index value of 83.0 C.
Fig. 2 shows that the top and bottom samples from Tank 2 have a similar relationship to their Tank 1 counterparts. The figure also indicates that the gasoline at the top of Tank 2 appears to contain slightly lower concentrations of highly volatile components.
Figs. 3 and 4 compare the Tank 2 top samples obtained before and after the 60-mile transport from the pipeline terminal to the gas station. Fig. 3 compares the two samples on the basis of relative difference in component concentration. It also includes all components present in concentrations greater than 0.1 GC area %.
Each symbol on the graph represents the difference between a component's concentration in the top and bottom samples, divided by its top concentration, expressed as a percent. Value that diverge significantly from the trend line are derived from data on components present in small concentrations, for which GC quantitation errors can "magnify" relative concentration differences.
While the magnitudes of the differences between the two samples are only slightly outside the error range of the analysis technique, Figs. 3 and 4 show the effect of evaporation of the lighter components during transport.
The top and bottom samples of the ethanol-gasoline blend in Tank 3 were compared by the same methods described above. Because the large difference in ethanol concentration between the two samples would have overshadowed any other differences, ethanol was "subtracted" from the chromatograms of the two samples.
The remaining data were then renormalized to enable comparing the samples on the basis of their gasoline fractions. When comparing top and bottom samples, the gasoline fraction of the ethanol blend displays a trend similar to the nonethanol gasoline.
The before and after-transport samples from the top of Tank 3 were also compared. The results showed that the concentration of all components (excluding ethanol) is essentially the same in both samples because the differences were within the range of analytical error.
In comparing the effect of transport on the evaporation of ethanol blend vs. nonethanol gasoline, however, the ethanol content at the top of the tank ranged from about 4.4 wt % at the pipeline terminal to 9.3 wt % at the gasoline station.
Tables 1 and 2 show that, for a given sample point (top or bottom), Rvp normally decreased as a result of transport from the pipeline to the station. This was true in all cases except for the samples taken from the bottom of Tank 2.
Table 1 shows that the Rvp values for the pipeline and station samples were 10.5 psi and 10.7 psi, respectively. This volatility increase was confirmed by a GC analysis comparing the two samples on the basis of major component concentrations.
A probable reason for the volatility difference involves the location of the tanks on the semitrailer and the configuration of their drain pipes. Tanks 1, 2, and 3 are, respectively at the front, middle, and rear of the trailer.
Each tank is drained from the rear by a 4-in. diameter pipe. Because the drain pipes for Tanks 1 and 3 extend toward the Tank 2 drain (which extends straight down), the pipes for Tanks 1 and 3 are significantly longer than the Tank 2 pipe. And because Tank 3 is the largest tank, it has the longest drain pipe.
It is likely that the shorter length of the Tank 2 drain pipe enabled more mixing of the drain pipe contents with the gasoline in the main tank compartment.
To further investigate the observed top vs. bottom difference in Rvp, another set of samples was collected from a bulk transport tank. Table 3 shows the Rvp values of three samples collected from a full 1,000-gal tank before it was drained into an underground storage tank.
The "Bottom, absolute" sample was collected from the drain pipe immediately after opening the drain valve; the "Bottom, after 20 gal" sample was collected after draining about 20 gal from the tank; and the "Top" sample was collected from the fill hole at the top of the tank.
The table shows a small but significant difference in Rvp between the top and bottom samples. The fact that the difference is smaller than observed with the larger tanks in previous testing may be because the air temperature was much colder during the second sampling.
Table 4 shows the results of a study designed to evaluate the precision of the GC technique used to determine the component concentration data in this report. The table shows variations in the concentration of selected components in a reference gasoline that was analyzed at regular intervals throughout a set of 89 GC auto-sampler injections.
The relative standard deviation (calculated as the standard deviation divided by the mean, converted to a percentage) of the listed components indicates the precision of the technique. Under the GC conditions used, about 240 components were detected in the reference gasoline.
Table 4 shows that, for quantitation of the 12 components monitored, the relative standard deviation values for 10 compounds ranged from 1.69 to 3.48, with benzene and ethanol outside of this range.
The fact that these two components display the largest relative standard deviation values is not surprising because they differ the most chemically from the other 10 components monitored. Ethanol is a polar molecule and benzene is a pure aromatic; the other compounds are either aliphatic or have some aliphatic character.
Based on these results, it seems reasonable to discount as analytic error comparative differences in overall gasoline composition that are less than 5%. In other words, if the trend line for a comparison does not diverge by more than 5% from the X axis, the two gasolines are assumed to be the same.
IMPLICATIONS
The data presented here show that gasoline is subject to composition and Rvp changes through commercial handling and transportation. Although more sampling is needed to validate and quantitate these changes, the results of these tests indicate that gasoline may not always behave as a homogeneous mixture and that the composition of a gasoline sample from a commercial truck transport tank may depend on where the sample is collected.
And although contamination cannot be ruled out as a contributing factor to these findings, the consistency of the data and the correlation between the GC and Rvp data indicate that the observed trends are not sampling or analytical artifacts.
In a recent rule, the EPA provides for possible enforcement action if a fuel is found to exceed the applicable Rvp standard by more than 0.3 psi.2 The two techniques recommended by EPA for tank truck sampling involve moving a glass beaker or bottle vertically through the tank contents at a rate that results in a 70-85% full bottle.
Briefly, a "running" sample is obtained by lowering, then raising, an open container at a uniform speed, An "all levels" sample is obtained by lowering a stoppered container, removing the stopper, and raising the container at the variable rate required for proportionate filling. 3
In addition, if stratification is a possibility, EPA recommends taking upper, middle, and lower spot samples.2 Proper execution of EPA-recommended sampling techniques may involve some practice, and possibly, correlation of the results with another party.
While the data presented here show a definite difference between tank-top and tank-bottom Rvp, they were insufficient to investigate possible causes of the difference. The migration of volatiles to the liquid/vapor interface and the evaporation of volatiles into the vented head space above the liquid, however, may be involved.
In the case of the splash-blended ethanol fuel, the top-to-bottom Rvp difference could be related to the top vs. bottom ethanol concentration difference. (Tests have indicated that ethanol's impact on Rvp elevation is greatest at concentration of 5-10%, and that a fuel containing 25-40% ethanol will have roughly the same Rvp as its base gasoline.)
It should be noted that the data showing the larger Rvp difference in unleaded regular were derived from high-Rvp winter gasoline sampled during a sunny, unseasonably warm (61 F.) day in October in Grand Forks, N.D. It is therefore possible that a smaller difference would be observed for lower-Rvp summer -gasolines.
Energy & Environmental Research Center plans to perform additional truck sampling and analysis work on EPA-recommended running and all levels-type samples, in addition to experimental samples.
Other work under the National Alternative Fuels Laboratory Testing Program involved correlation of Rvp and GC data with fuel evaporation data to investigate the possibility of estimating gasoline evaporative losses from tanks based on before and after-evaporation Rvp values.
The work includes laboratory tests to determine evaporation rates for at-the-pump and lab-blended fuel samples, and monitoring of the Rvp and volatile component concentrations of gasoline in commercial 7,000-gal underground tanks.
REFERENCES
- Knudson, C.L., Mathsen, D.V., and Aulich, T.R., "Chemical and Performance Relationships in Ethanol and Nonethanol Fuels," U.S. Department of Agriculture report, Contract No. 88-38819-6044, October 1991.
- Federal Register, 40 CFR Part 80, Volatility Regulations for Gasoline and Alcohol Blends, Vol. 58, No. 50, Mar. 17, 1993, p. 14476.
- Code of Federal Regulations, 40 CFR Part 80, Regulations of Fuels and Fuel Additives, July 1, 1992, p. 370.
Copyright 1993 Oil & Gas Journal. All Rights Reserved.
