Content uploaded by Evangelos. G. Tzirakis
Author content
All content in this area was uploaded by Evangelos. G. Tzirakis on Mar 07, 2017
Content may be subject to copyright.
Proceedings of the 13th International Conference on Environmental Science and Technology
Athens, Greece, 5-7 September 2013
CEST2013_0489
ON-ROAD REAL DRIVING AND ROAD GRADIENT DATA PROCESSING
METHODOLOGIES TO FORM DRIVING CYCLE COMPLETE DYNAMOMETER
TESTS
E.G. TZIRAKIS1 and F.E. ZANNIKOS1
1Laboratory of Fuels and Lubricants Technology, School of Chemical Engineering,
National Technical University of Athens, Iroon Polytechniou 9 15773 Zografou Athens.
e-mail: vtziraks@central.ntua.gr
EXTENDED ABSTRACT
When a Driving Cycle (DC) is applied on a laboratory chassis dynamometer, it simulates
the driving conditions in the road network of modern cities. These models for
dynamometer control software are used for exhaust emission and fuel consumption
measurements and they find broad application in research for environmental pollution,
energy conservation and alternative automotive fuels (biofuels). Furthermore, DC tests on
chassis dynamometer simulate a flat road ignoring the effect that Road Gradient (RG)
has on exhaust emission and fuel consumption.
This work is a part of a project which involves the development of ‘Real World’ DCs from
on-road driving and road gradient data from the greater area of Athens, Greece. Two
different methodologies of DC development are presented here, based on Matlab code.
Driving patterns from various test vehicles were processed to form DCs of specific
number of driving periods (phases). The first method (A΄) was applied for short duration
and limited number of driving periods. The second method (B΄) was suitable for long
duration and multi-driving period cycles. The main criterion for the design and acceptance
of each method was the correlation of the resulted DCs with the corresponding
characteristics of the processed road data. Time is an important issue when processing
data. Method A΄ is time consuming but gives very accurate results requiring small
amounts of data. In opposition, method B΄ a larger amount of data is needed for
acceptable results but the processing time is extremely short. Applying both methods on
the same set of data, method A΄ needed 5 hours to complete the processing instead of
the 9 seconds of method B΄. The corresponding average accuracies were 99.9 and 97.2
respectively.
RG is a parameter that seriously affects vehicle’s exhaust emissions and fuel
consumption especially in a city like Athens with unique road network topography. RG
was taken into account for the development of the DCs for both methods. For method A΄,
RG was included in the final result by intervention on the mean positive and negative
accelerations by the one that results from gravity on an inclined road. For method B΄, RG
was assimilated by a “Load Cycle” expressed in kW and used in a combination with the
developed DC, thus forming a complete chassis dynamometer test. “Load Cycle” (LC)
was produced by calculating the power related to vehicle’s mass that holds or assists it
when driving uphill or downhill. Complete test that include RG were developed for both
motorcycles (Greek Urban Driving Cycle for Motorcycles) using method A΄ and passenger
cars (Greek Urban Driving Cycle) using method B΄.
KEYWORDS: Driving cycles, road gradient, chassis dynamometer, emissions, fuel
consumption.
CEST2013_0489
1. INTRODUCTION
In order to evaluate the environmental impact from vehicle emissions, researchers have
developed Driving Cycles (DC) used for laboratory chassis dynamometer emission
testing. DCs are speed-time profiles simulating driving conditions on road networks,
either urban or not. Road network topography and driving style are important factors
directly associated with driving conditions. Road Gradient (RG) should be taken into
account when estimating exhaust emissions and fuel consumption, as well as the driving
behaviour which is mostly aggressive in modern cities (OECD, 2004), (Tzirakis et al.,
2007). In addition, motorcycles constitute a significant share of vehicle fleet in southern
European countries and especially Greece. Due to their differences in speed profiles
compared to cars, driving cycle development for motorcycle emission testing is
considered essential (Chen et al., 2003). The vehicle fleet is changing in terms of
technology as manufacturers need to continuously reduce their fleet average CO2
emissions to the levels of 95gr/km by 2020 (European Commission, 2010) (measured on
NEDC). All these impose the need for driving cycles updating in order to review
legislation cycles. Legislative DCs, are used for Vehicle Emission Certification and are
imposed by governments. Non-legislative cycles such as the Hong Kong driving cycle
(Hong et al., 1999), and the Athens driving cycle (ADC) (Tzirakis et al., 2006) are a useful
tool in research for energy conservation and pollution evaluation through the vehicle
testing on exhaust emissions and fuel consumption (Andre et al., 2006) as well as in the
field of vehicle design tooling and marketing. Since ‘Real world’ driving cycles reflect
more accurately the urban driving conditions, they support research on fuel behaviour
comparison, alternative and renewable fuels (i.e. Biodiesel), fuel reformulation, fuel
additive development and engine modification (Karavalakis et al., 2007). Each city has a
unique driving profile and the collected traffic data result in a different driving cycle
depending on characteristics such as the road network, the driving behaviour and the
vehicle fleet potentiality and number (Andre et al., 2006). As recently mentioned, in lab
real driving emissions are determined in order to establish the emission limits for EURO
VI (Hausberger et al., 2012). Comparison of results from chassis dynamometer tests with
NEDC and ‘Real World’ driving cycles based on traffic data from European and non-
European cities showed significant differences in emission and fuel consumption levels
(Karavalakis et al., 2007). Real world driving data and chassis dynamometer tests
contribute on developing models which calculate CO2 emissions and fuel consumption
directly using a set of vehicle fleet characteristics. There are a series of well known
developed models that can be applied in various countries and cities around the world
(Robin and Ntziachristos, 2012). Furthermore, aggressive driving is as fuel consuming
and emission surcharging as driving uphill as the use of excessive throttle accelerator is a
necessity on both driving conditions (Tzirakis et al., 2007). The DC construction
methodologies include techniques such as the chase car method using instrumented
vehicles, route selection or complete road network coverage of specific areas and cycle
construction where a great number of DC characteristics should meet in the largest
percentage possible, the corresponding characteristics of the on road data. Sometimes
the process was to the extent of the whole trip a vehicle was performing and some others
to the extent of its stop to stop driving phases (Hung et al., 2007). For the monitoring of
European traffic characteristics and the formation of representative driving cycles, top-
notch institutions are involved (Andre, 2004).
2. DATA COLLECTION PROCEDURE
A significant parameter when developing a real world DC for a specific city or area is
whether the road gradient is included in the bench test. DCs are usually developed
assuming that test vehicles are driven on zero gradient and therefore, when a vehicle is
tested on a chassis dynamometer it is assumed that it is driven on a flat road. In cases
CEST2013_0489
where the topography plays a significant role on the speed profile of the vehicles, it is
essential that the road gradient is included in the final dynamometer test, through the DC.
(a)
(b)
Figure 1. Change of the average value of average speed (a) and stop % time (b), during
data collection progress as well as of the reliability level of 95%.
For a general work frame, the peculiar area of Athens was selected for the collection of
on-road data with the aid of instrumented vehicles. The procedure followed for the
development of DCs comprises of three major components, the logging technique, the
route or area selection and the cycle construction methodology (Tzirakis and Zannikos,
2011). They include the use of sophisticated equipment which is rearranged in order to
record the desired vehicle data and the routes or area covered in the Attica basin for the
data collection. It was mandatory to ensure that the samples collected for every driving
cycle to be developed, are statistically sufficient. For this purpose graphs were created
displaying the sample collection sequence for two of their main characteristics (Figure 1).
3. DRIVING CYCLE DEVELOPMENT METHODS
The concept of developing a complete DC dynamometer test is described in this paper.
Two methods were determined using mainly Matlab code. This paper describes how road
gradient is taken into account when developing a DC for both of the methods by using a
different approach. Method A΄ (Tzirakis and Zannikos, 2011), (Tzirakis et al., 2008), was
used to develop the Greek Urban Driving Cycle for Motorcycles (GUDCM) and method B΄
for developing the Greek Urban Driving Cycle (GUDC) for passenger cars.
3.1. Method A’
The first three steps of the methods are identical. Data which were collected in second
intervals were processed in the form of stop-drive-stop phases. A marginal 5% of the total
number of phases was removed considering 5 basic characteristics. Phases were
separated in groups according to their duration. The procedure followed according to
method A΄ can be seen in Figure 2(a). A capable number of phases were selected from
each group according to their duration starting from the mean value of the duration of
each group. The number of phases selected depends upon the accuracy and the number
of the input criteria and the capability of the computer used to do the combinations. The
selected phases are then used for a very large number of combinations, which depends
on the number of the groups and phases selected from each group, which are performed
in order to have the desired result of which the basic characteristics (input criteria) will be
as close to the characteristics of the road data as possible. The basic criteria where
based upon a number of average values of the on-road data which must agree with those
of the final cycle. Those criteria are: % stop time, % positive acceleration, % negative
acceleration, average speed without the stops, average positive acceleration and
average negative acceleration.
CEST2013_0489
(a)
(b)
Figure 2. Methods developed for DC construction (a) Method Α΄ (b) Method Β΄.
RG can be included in the results when using method A΄ by incrementing to mean values
of positive and negative acceleration and gives accurate results even though the
acceleration values are changed in order to include road gradient. The additional
acceleration (αadd) that a vehicle wastes when driven on an inclined road could be easily
calculated. It is the component of acceleration of gravity parallel to the road surface:
αadd = g (H2-H1)/d (1)
Where g is the acceleration of gravity, d is the distance covered for 1 second (data
logging interval=1Hz) and H2-H1 is the altitude difference between the logs.
Road data in the form of driving phases (microtrips)
Removal of 5% of the
marginal phases
according to the criteria:
Duration
Distance
Average speed
Maximum
speed
Average
acceleration
Separation in equal, as
far as the number of
phases is concerned,
groups according to the
duration of the phases
Selection of capable
number of phases
according to duration,
around the mean value of
each group
Criteria input for
the driving cycle
selection
depending on the
needs
Combinations with the phases
on from each group at a time
for the formation of the cycle
and checking according to the
criteria
Phases
combination
=1
Phases
combination =0 or
Phases
combination >1
Change of
the input
criteria
depending on
the result
Final Dynamometer test (Driving Cycle-road
gradient included by incrementing on positive and
negative acceleration)
Selection of the specific phases
which constitute the Average
Driving Cycle of all data used
in the process
Estimation according to the
minimum summation of the
deviation percentages from the
mean value of the groups for
each criterion
Road data in the form of driving phases (microtrips)
Removal of 5% of the
marginal phases
according to the criteria:
Duration
Distance
Average speed
Maximum
speed
Average
acceleration
Separation in equal, as
far as the number of
phases is concerned,
groups according to the
duration of the phases
Final Dynamometer test (Driving Cycle-road
gradient included
Driving Cycle
- no road
gradient
included
Test
vehicle
mass
Assimilation of
the RG on
chassis
dynamometer
(“Load Cycle”)
CEST2013_0489
3.2. Method B’
This method is a very quick statistical evaluation of the driving phases according to the
input criteria, which can be performed only by using programming. It gives only one result
(see step 6 Figure 2(b): “Driving cycle – no road gradient included”) which is a
combination of phases of different duration. The basic characteristics (input criteria) of the
result match, in the greatest percentage possible, the corresponding characteristics of the
on-road data. The accuracy increases as the population of the data increases. This
method also gives the capability of including road gradient through positive and negative
acceleration by incrementing on those criteria. This however, reduces the accuracy of the
resulted driving cycle in relation to the rest of the criteria that come from on-road
recordings. In addition to that the driving cycle is becoming more aggressive. The
accuracy can be increased either by collecting more data to be entered to the programme
or by including road gradient in the chassis dynamometer test using the method
described in Figure 3 and in the lines of the next paragraph.
Figure 3. Procedure followed for the assimilation of road gradient.
In order to assimilate the load on a chassis dynamometer, the recorded altitude that
relates to the selected phases that will be used to form the final driving cycle were
isolated. Proper modification was made in order that the gradient to match the mean
gradient of all data recorded. In addition to that, attention was paid to a great extend on
the increases and decreases of gradient, so that the final driving cycle will be easy driven.
For the development of a “Load Cycle” that will be used on a chassis dynamometer in
order to assimilate the road gradient, the power of the vehicle that relates to its mass due
to gravity on a road with slope must be calculated. This power can be either positive
(downhill) or negative (uphill). Power (Pgradient) can be calculated as:
Pgradient = g v sinφ mvehicle (2)
Where g is the acceleration of gravity, v is the velocity of the DC, φ is the road gradient of
the specific phases of the derived driving cycle and mvehicle the mass of the test vehicle.
For gradients below 5o it is assumed that: sin(φ) = tan(φ)=φ. Therefore, the gradient can
be written as: φ = (H2-H1)/d (3)
Where d is the distance covered for 1 second (data logging interval=1Hz) and H2-H1 is
the altitude difference between the logs. For the data logging of 1Hz stands: d=v.t = v so
(1) becomes: Pgradient = [(H2-H1)/v] g v mvehicle=(H2-H1) g mvehicle (4)
3.3. Comparison of methods
For comparison purposes, both methods were applied on the same set of data creating
two simple speed-time profiles, without using the option of RG (Tzirakis et al., 2008). As
seen in Figure 2, the methods differ from each other only in the final stage of driving cycle
formation, through the phase combination. The phase groups used are the same for both
methods. The resulted cycles have both similarities and dissimilarities. Time is an
important issue when processing data. Method ‘A’ is time consuming (5 hours) but gives
Impedance and assistance
rolling estimation in kW.
Assimilation of road
gradient on the chassis
dynamometer:
(“Load Cycle”
For Method B΄)
Isolation of the
recorded altitude that
relates to the phases
of the driving cycle
Estimation of
the road
gradient in %.
Adjustment of the altitude
characteristics in terms of
the on road data:
% uphill
% downhill
Mean gradient ascending
Mean gradient descending
CEST2013_0489
very accurate results. Many of the characteristics of the DCs, shown on Table 1, match
over 99% the corresponding characteristics of the on-road data. Using the ‘B’ method, the
resulting cycle is acceptable since the smallest match percentage is 95.57%. Despite the
acceptable percentage the time needed for the computer to process the data was
extremely small (9 seconds).
Table 1. Comparison of matching percentages in 6 characteristics of DCs with the
corresponding of the on-road data, developed through two different methods using the
same set of data.
Method ‘A’
Method ‘B’
Time consumed for processing (computer)
5 hours
9 seconds
Cycle duration
100%
97.44%
Average speed
99.76%
97.09%
Average speed (without stops)
99.79%
96.95%
% stop time
99.56%
99.26%
Average positive acceleration
99.93%
97.30%
Average negative acceleration
100%
97.85%
4. RESULTS
Figure 4 illustrates the Greek Urban DC for Motorcycles (GUDCM) compared to a DC
developed without taking into account the road gradient. These cycles were developed
using method A’. When road gradient is included through acceleration the cycle is more
aggressive resulting in higher travelling speeds.
Figure 4. GUDCM compared to DC with no road gradient taken into account
Table 2. Basic DC characteristics compared to those of the recorded data.
Characteristic
Data Average
Values
GUDCM
DC – No Road
Gradient
Duration(s)
822
822(100%)
822 (100%)
Moving time (s)
723
723(100%)
723 (100%)
Average speed (km/h)
29.04
28.86 (99.38%)
29.05 (99.96%)
Stop time (%)
12,07
12.04 (99.75%)
12.04 (99.75%)
Av. positive acc. (ms-2)
0.6045 (0.5398)
0.6065 (99.67%)
0.5411 (99.76%)
Av. negative acc.(ms-2)
0.6497 (0.5598)
0.6349 (97.72%)
0.5629 (99.45%)
Table 2 shows the basic characteristics of GUDCM and the one developed without taking
road gradient into account as well the corresponding values of the recorded data.
Average positive and negative acceleration without road gradient taken into account is
written inside the brackets of Data Average Values column. When no RG is included all
basic characteristics match over 99.5 % with the data average values. When RG is
included the corresponding percentages are also over 99.5 % with the exception of
negative acceleration (97.7%). This is due to the changed values of positive and negative
acceleration in the programme’s input data in order to include RG, which tries to combine
the new values of acceleration with the rest unchanged characteristics. Method B΄ was
CEST2013_0489
used for the construction of GUDC. The complete driving cycle test that includes the “load
cycle” is illustrated in Figure 5.
Figure 5. GUDC and assimilation of road gradient by creating a “Load Cycle”.
Table 3. Speed and load points of GUDC for a ten second section of the cycle.
Time
speed (km/h)
power (kW)
Time
speed (km/h)
power (kW)
…
…
…
1044
24.96
-3.673
1047
34.47
-1.476
1045
28.26
-3.133
1048
37.87
0.000
1046
28.69
-2.079
1049
35.78
1.364
…
…
…
Table 4 shows the main characteristics of GUDC compared to other driving cycles.
Table 4. GUDC compared against recorded data and against other driving cycles.
Characteristic
Data Av.
Values
GUDC
NEDC
ECE-
15
Artemis
Urban
Duration (s)
1180
1175(99.58%)
1180
195
993
Average Speed (km/h)
18.94
18.67(98.57%)
33.6
18.4
17.65
Av. Speed-no stops (km/h)
27.76
27.31(98.84%)
44.8
26.5
24.66
Stop time (%)
31.76
31.66(99.69%)
25.42
30.8
28.4
Positive Acceleration (%)
33.89
33.89(100%)
20.93
21.5
34.64
Av. pos. acc.(ms-2)
0.686
0.692(99.13%)
0.726
0.642
0.732
Distance (m)
6207
6094(98.18%)
11007
990
4870
No of Phases
19.62
20(98.10%)
13
3
22
5. CONCLUSIONS
Method A΄ has the advantage of giving acceptable results with a small amount of
recorded data. Additionally it is not affected when changing the average values of positive
and negative acceleration in order to include road gradient in the resulted driving cycle.
On the other hand, the time needed to give results is mainly affected by the number of
phases of the cycle to be developed so it is suitable for cycles with small number of
phases. The accuracy in Method B΄ is greatly affected when road gradient is included in
the process by incrementing on the criteria of acceleration. The computer however needs
only a few seconds to give the final result. That is the reason why the process of including
the road gradient using the “Load Cycle” was created. More criteria could be added in the
programs in order for the resulting driving cycles to be more representative in relation to
the characteristics of the road data. However, this could lead to their low accuracy which
can be improved with a larger set of data grows. Gathering of on-road data can be done
remotely through GPS technology and telemetry or with the help of really fast data
transferring 3G and lately, 4G mobile networks. A new application that can be run in a
smart phone or tablet is now under development, will log the vehicle’s operation
parameters directly from the CAN BUS and send it in real time at the laboratory’s data
base. Real world driving cycles are extremely useful for policy makers and scientists on
CEST2013_0489
environmental and technico-economic applications and need to be constantly and
frequently updated through new road data recordings, set by the changes in traffic
conditions which are the result of the development of road networks, the growing and
change of the car fleet, the traffic adjustments or the changes in driving behaviour, in
order to be reliable and to meet up to date traffic conditions. Also the latest European
crisis has led former passenger car drivers to park or even sell their cars and use a more
cheap and efficient transportation such as public means or a smaller car or motorcycle or
a bicycle. This fact also changes the traffic in the cities of Europe and mostly in countries
such as Greece where the problem is deeper. It is also clear that evaluation of the CO2
emissions from a newly designed vehicle, based on results from a chassis dynamometer
test operated under driving cycles which do not represent the driving conditions of
modern cities (eg. NEDC) could be completely misleading. Since emissions differ from
cycle to cycle, new technology “green” car solutions should be tested on driving cycles
reflecting real world driving conditions.
REFERENCES
1. OECD (2004), Can Cars Come Clean? – Strategies for Low-Emission Vehicles. 2, rue Andre
Pascal, 75775 Paris: OECD Publications, Cedex 16, ISBN-92-64-10495-X.
2. Tzirakis E, Zannikos F, Stournas S. (2007), Impact of driving style on fuel consumption and
exhaust emissions: defensive and aggressive driving style, Proceedings of the 10th
International Conference on Environmental Science and Technology, 5-7 September 2007,
Kos island, Greece.
3. Chen KS, Wang WC, Chen HM, Lin CF, Hsu HC, Kao JH, Hu MT. (2003), Motorcycle
emissions and fuel consumption in urban and rural driving conditions, The Science of the
Total Environment, 312, 113–122.
4. European Commission (2010), Reducing CO2 emissions from passenger cars,
http//ec.europa.eu/clima/policies/transport/vehicles/cars_en.htm.
5. Tong H. Y., Hung W. T., Cheung C. S. (1999), Development of a driving cycle for Hong Kong,
Atmospheric Environment, 33, 2323-2335.
6. Tzirakis E, Pitsas K, Zannikos F, Stournas S. (2006), Vehicle Emissions and Driving Cycles:
Comparison of the Athens Driving Cycle (ADC) with ECE-15 and European Driving Cycle
(EDC), Global NEST Journal, 8, 3, 282-290.
7. Andre M, Joumard R, Vidon R, Tassel P, Perret, P. (2006), Real-world European driving
cycles, for measuring pollutant emissions from high- and low-powered cars, Atmospheric
Environment, 40, 5944–5953.
8. Hausberger S, Luz R, Rexeis M. (2012), New Emission Factors for EURO 5 & 6 Vehicles,
19th International Transport and Air Pollution Conference 2012, 26-27 November 2012,
Thessaloniki, Greece.
9. Karavalakis G., Tzirakis E., Zannikos F., Stournas S., Bakeas E., Arapaki N., Spanos A..
(2007), Diesel/Soy Methyl Ester Blends Emissions Profile from a Passenger Vehicle Operated
on the European and the Athens Driving Cycles, Ref No 2007-01-4043, SAE 2007
Transactions, Journal of Fuels and Lubricants, ISBN: 978-0-7680-1983-4, ISSN: 0096-736-X.
10. Robin S, Ntziachristos L. (2012), COPERT Australia: Developing Improved Average Speed
Vehicle Emission Algorithms for the Australian Fleet, 19th International Transport and Air
Pollution Conference 2012, 26-27 November 2012, Thessaloniki, Greece.
11. Hung WT, Tong HY, Lee CP, Ha K, Pao LY. (2007), Development of a practical driving cycle
construction methodology: A case study in Hong Kong, Transportation Research Part D, 12,
115–128.
12. Andre M. (2004), The ARTEMIS European driving cycles for measuring car pollutant
emissions, Science of the Total Environment, 334/335, 73–84.
13. Tzirakis E, Zannikos F. (2011), Development of passenger car and motorcycle driving cycles
for Athens, 12th International Conference on Environmental Science & Technology 2011,
Greece 8-10 September 2011, Rhodes island, pp. A1947-A1954.
14. Tzirakis E., Kyriakidis A., Zannikos F. (2008), ‘Methodologies for driving cycle development,
using on-Road data from Athens’, Proceedings of: Transport Research Arena Europe 2008,
Ljubljana, Slovenia, 21-24 April 2008.
