SDSB 79-28
Technical Report
Fuel Consumption Measurements-
Carbon Balance vs Flow Meter
by
Dale Turton
July 1979
NOTICE
Technical Reports do not necessarily represent final EPA decisions
or positions. They are intended to present technical analysis of
issues using data which are currently available. The purpose in the
release of such reports is to facilitate the exchange of technical
information and to inform the public of technical developments which
may form the basis for a final EPA decision, position or regulatory
action.
Standards Development and Support Branch
Emission Control Technology Division
Office of Mobile Source Air Pollution Control
Office of Air, Noise and Radiation
U.S. Environmental Protection Agency
-------
-2-
I. Introduction
In a recent EPA experiment, fuel consumption data was obtained by
both the carbon balance method and the fuel flow method. Since vehicle
fuel economy is an area of present concern, this report was prepared to
describe and compare the two methods of measurements.
II. Background
EPA fuel economy measurements are currently obtained by the carbon
balance method. This method, in effect, assumes that the carbon atoms
emanating in the vehicle exhaust came from the fuel going into the
carburetor. Therefore, by measuring the carbon in the exhaust the
quantity of fuel consumed can be determined. This approach is used by
EPA because all the equipment required is also required for the exhaust
emission measurements. The fuel economy can therefore be obtained
concurrently with the emissions data at essentially no extra cost.
Fuel consumption can also be measured by several other methods,
such as volumetric, fuel flow meters, or gravimetric methods. Any
difference between fuel consumption measurements by the available
methods is small compared to the total fuel consumed. Therefore,
accurate comparison between methods must be done statistically and
requires a large data base.
A recent EPA program which investigated the accuracy of the dyna-
mometer simulation of the speed of a vehicle consisted of numerous
repetitive steady state dynamometer tests. It provided an ideal oppor-
tunity to obtain sufficient fuel consumption measurements by both
methods for the statistical analysis necessary to detect any difference
between the methods.
III. Discussion
The experimental program consisted of numerous steady state tests
of approximately 5 minutes duration each. During each test the fuel
consumption data was obtained by both the carbon balance and fuel flow
meter methods. The paired data was then statistically analyzed to
compare the two measurement methods. The following subsections describe
the experimental program and data analysis in greater detail.
A. Experimental Program
A 1976 Mercury Montego with a 351 y-8 engine was used throughout
the experiment. Fuel consumption measurements were obtained at all the
combinations of three different dynamometer power absorbtions (10.4,
11.4, 12.4 HP) and three different speeds (40, 50, 55 mph). Thus, nine
test points were obtained with this sequence. This sequence was repeated
twice, once with the dynamometer rolls coupled and once with the rolls
uncoupled. These tests provided 18 fuel consumption data points over a
wide range of values. Six different sets of tires, including both
radials and bias, were tested increasing the total data base and expanding
-------
-3-
Figure 1
135.00 +
•i
*
UO.GO •»• * *
2
•f
*
.i.;.'s. oo •(• * *
* *
/—N
^ * 2 2*"* * * *
O * 3 *
vS 120.00 + *
* * 2
to •-,
j_i •»
*
115.00 + *****
* *
4-
* *
2
1.1.0.00 +
*
* * *
•f
* *
4 *
100.00 + * * * *
K
+
2
V *"'i 0 '* I 0 "f"
.!..-._. _«.!... w. -.1 „- .1. —._ —.— .L...M. .« V. !._•_«_ .L— __«..!..«__„. .L «..»».«. 4,_,_ ™,— . L«_M. ^.X •*.«.»•. >i. «~— «. « .1. .-.!-•. ... 1^ X IJlTllMll.- -J ...... i- T - - ..I ... -i I
95.000 ' 105.00 115.00 . .125100 135.00
J.00.00 110.00 120.00 130.00
Fuel Flow Measurements (cc/km)
-------
-4-
the range over which fuel consumption was observed. The same test
sequence was used each time.
Each day a dynamometer and vehicle warm-up was conducted prior to
the testing. This was considered necessary to minimize variations due
to test equipment from day to day. Each test point consisted of a 300
second steady state. After stabilizing at the steady state speed, the
fuel flow meter and the exhaust sampling bag were simultaneously turned
on. After about 300 seconds both were turned off completing the sampling
interval. The fuel meter paper tape, carbon balance data, and dynamometer
roll revolutions were then recorded for the data reduction and evaluation.
B. Data Reduction
The data reduction process simply transformed the carbon balance
and flow meter data to a common system of units for comparison. In this
report fuel consumption, in cc/km, was used for all comparisons. For
the carbon balance method the exhaust gases are collected, and sampled.
The hydrocarbons, carbon monoxide, and carbon dioxide are then measured
from this sample. The amount of carbon in these gases and therefore,
the amount of gasoline used is then calculated in the manner typically
used in EPA fuel consumption calculations. These results are expressed
in a fuel volume correct to a standard fuel temperature of 60 F. This
fuel volume was then divided by the distance travelled, as measured from
the dynamometer rear roll, to obtain a fuel consumption rate in cc/km.
The flow meter measurements of fuel consumption were obtained from
the printed tape output of the flow meter. The fuel temperature at the
beginning and end of each 5 minute test was also obtained from the
printed paper tape. The average of these temperatures was used to
correct the measured consumption to the standard temperature of 60 F.
Again, the distance travelled as measured by the dynamometer roll revo-
lutions was used to compute fuel consumption per kilometer placing the
measurements on a equal basis.
C. Data Analysis
The scatter plot in Figure 1 presents the carbon balance fuel
consumption plotted vs. the fuel consumption measured by the flow meter.
The plot shows an apparent linear relationship between the two methods.
To check this linearity the data was analyzed by regressions of the
form.
Fp = a + bF-
CB fm
and:
C B f m f m
where:
F_ = carbon balance fuel economy
F
fm = fuel fl°w meter fuel economy
a,b,c constants
-------
-5-
Figure 2
I
0
o
§
M
to
CO
01
s
-------
-6-
The regression using equation (2) indicated little confidence that
the coefficient c^was different from zero. Therefore, as observed
visually, there is no evidence of a non-linear component of the re-
lationship between the measurement methods.
The regression using equation (1) indicated that there was an 85%
confidence that the constant term (a) was greater than zero. This
suggests that there may be a constant offset between results of the two
methods, but there is relatively little confidence in this statement.
Since there was little confidence in the constant term of the
previous regression it was judged more appropriate to consider the
results of the two methods to be proportional. A regression deleting
the constant term yielded a slope of .969 implying a 3.1% difference
between the results of the two methods. This equation is plotted in
Figure 2 next to the line which would occur if the carbon balance
measurements were equivalent to the fuel flow measurements.
Once it was established that the two results can be best treated- as
proportional, the following alternate data analysis methods can be used
to investigate this difference in greater detail:
The precent difference can be defined as:
F - F
% Difference = -Q= — x 100
CB
Computing a mean and standard deviation of this variable:
Mean % Diff. = 3.12%
Standard Dev.=1.83
A students "t" test of the mean indicates that there is a virtually
100% confidence that the fuel meter measures higher fuel consumption
than does the carbon balance method. In addition, this type of analysis
indicates that there is over a 90% confidence that the true percent dif-
ference is between 2.9% and 3.5%; and there is over a 95% confidence
that the value is between 2.8% and 3.6%.
III. Conclusions
It is concluded that the carbon balance method of measuring vehicle
fuel consumption yields higher fuel economies than are obtained by the
direct volumetric measurement of fuel flow. Under the conditions of
this experimental program; that is, one vehicle and one test cell, the
difference was about 3% and was very consistent. There is no reason to
believe that the percent difference is vehicle dependent although it
could be affected by analyzer calibrations or other test cell related
parameters.
-------
-7-
The flow meter had a rated accuracy of 0.5% and calibrations,
immediately before the program confirmed that the measurement accuracy
was within this range. In comparison, the carbon balance method requires
several indirect calibrations and relies on several assumptions about
the fuel composition. Therefore, the flow meter approach is probably
the more accurate method of determining fuel consumption. Consequently,
it is concluded that the carbon balance method overestimates the fuel
economy of a vehicle by a small, but consistant amount. This conclusion
is consistant with earlier EPA observations.(1)
IV. Recommendations
The observed difference between the carbon balance and fuel flow
measurements, 3%, may seem small, however for the tested vehicle used it
represents a composite fuel economy effect of about 0.5 mpg. Also,
since the difference appears to be proportional to the fuel consumed,
the difference between the two methods would be expected to be over 1
mpg in the case of a fuel efficient vehicle.
The first recommendation is to verify, by an independent measure-
ment, which fuel consumption measurement approach is "correct". It is
recommended that gravimetric measurements be obtained in parallel with
the carbon balance and fuel flow measurements for this verification.
If it is verified that the carbon balance approach overestimates
fuel economy, then the reason for the discrepancy should be identified.
If it is a random calibration problem, not affecting all test cells,
then it may be possible to easily correct the problem and reduce test
variability. If it is a systematic problem, such as variations in fuel
composition, then a change would require rulemaking and involve EPCA
constraints. However, even in this case, it may be relatively easy to
compute one set of fuel economies for EPCA compliance and a second set
for the EPA label and buyer's guide.
As a first step in identifying the case of the discrepancies, it is
recommended that measurements be obtained over a wide range of fuel
consumptions to resolve if the differences are truly proportional, as
indicated by this experiment, or if some constant offset may also
exist.
1 Christopher Marzonie, Memorandum to Charles Gray, "Accuracy of Fuel
Economy Measurements", August 1977.
------- |