Class-8 Heavy Truck Duty Cycle Project Final Report - Center for ...

uns 91 to 100. Just by chance, the combination of the amount of fuel in the tank at the beginning and
end of the 4,000 mile trip for any of these runs was such that the observed discrepancy in fuel
consumption would have been positive. In this case, the conclusion would have been that the databus
overestimated fuel consumption even when in reality it could have underestimated it or measured it
with total accuracy.
% Difference in Total Fuel Consumption Estimation
15%
10%
5%
0%
0 10 20 30 40 50 60 70 80 90 100
-5%
-10%
-15%
-20%
Perfect Accuarcy 5% Underestimation Error 5% Overestimation Error
Run Number
Fig. 62. Fuel Consumption Estimation Errors
Even though it was not possible to calibrate the gathered fuel information because of the reasons
stated above, a literature review on this topic showed that the errors introduced by measuring fuel
consumption with databus information are small. For example, Mats Bohman (2006) reports that
databus fuel consumption estimation was within 1% of the readings obtained using a fuel
measurement tube (i.e., a two-meter long tube with known diameter that is used instead of the fuel
tank). Similarly, a 2002 EPA study indicates that the fuel consumption obtained from databus
information is accurate, as long as severe “spikes” in the data stream are capped (EPA, 2002). For the
present study, this precaution was followed by restricting the maximum rate of change for fuel flow to
approximately 0.018 (gal/sec)/sec.
Browand, Radovich, and Boivin (2005) indicate that the use of the databus fuel rate signal is accurate
and reliable particularly if differences in fuel consumptions are measured rather than absolute values.
In other words, systematic errors in the estimation of fuel consumption are not important (and in fact
are irrelevant) when comparing fuel economies to, for example, determine if a given type of tires is
more efficient than another type. Consider, for example, that the databus always overestimates fuel
consumption by a given percentage p>1 and that this factor is intrinsic to that particular databus (i.e.,
any truck with the same databus would always overestimate fuel consumption by p). Suppose that
truck T has the fuel consumption curve shown in Fig. 63 for a given trip t. In that figure, the real fuel
75