light duty vehicle test cycle generation based on real ... · through the deviation of the test...
TRANSCRIPT
i
Light Duty Vehicle Test Cycle Generation
Based on Real-World Data
Alexandr Rosca
Thesis to obtain the Master of Science Degree in
Mechanical Engineering
Chairperson: Professor Mario Manuel Gonçalves da Costa
Supervisors: Doctor Gonçalo Nuno Antunes Gonçalves
Doctor Gonçalo Nuno de Oliveira Duarte
Member of
the committee: Professor Tiago Lopes Farias
October of 2013
ii
Table of Contents
1. Introduction ................................................................................................................................. 1
1.1. Motivation ............................................................................................................................ 1
1.2. Current studies and the State of the Art ..................................................................... 4
1.3. Objectives ............................................................................................................................ 6
1.4. Structure of the Thesis .................................................................................................... 7
2. Background ................................................................................................................................. 9
2.1. Emissions Regulations .................................................................................................. 10
2.2. Test cycles ........................................................................................................................ 12
2.3. How measurements are made in laboratory ............................................................ 21
2.4. Parameters that influence consumption and emissions ...................................... 22
3. Methodology ............................................................................................................................. 28
3.1. Vehicle Specific Power (VSP) ....................................................................................... 29
3.2. Approach on generating test cycles .......................................................................... 33
4. Test Cycle Generating Tool (TCGT) .................................................................................... 36
4.1. User Interface and main functionalities ..................................................................... 36
4.2. TCGT algorithm and resources ................................................................................... 41
4.2.1. How TCGT works ..................................................................................................... 41
4.2.2. Which blocks make the database ....................................................................... 49
5. Validation ................................................................................................................................... 53
5.1. Validation of the Test Cycle Generation Tool .......................................................... 53
5.2. Representativeness of the generated Test Cycles ................................................. 57
5.2.1. Validation with VSP distribution as input ......................................................... 57
5.2.2. Validation with Driving Cycle as input ............................................................... 59
5.3. Influence of generated test cycle duration ............................................................... 71
5.4. Influence of the “weight vector” .................................................................................. 73
6. Applications .............................................................................................................................. 75
6.1. Test Cycles based on a combination of different trips ......................................... 75
6.2. Test Cycles based on a population (fleet) ................................................................ 77
7. Conclusions and Future Work ............................................................................................. 79
Bibliography ........................................................................................................................................ 80
iii
List of Figures
Figure 1 - Evolution of Cars Production over last decade [1] ....................................................... 1
Figure 2 - US total emissions from mobile sources 1970-2005 [3] .............................................. 2
Figure 3 - Pollutant emissions for several measured vehicles, on 4 routes and NEDC ........... 4
Figure 4 - Evolution of CO2 levels of emission over the last centuries ........................................ 9
Figure 5 - Emission standards for other countries [3] .................................................................. 12
Figure 6 - NEDC Driving Cycle (4x UDC + 1x EUDC) [17] ......................................................... 13
Figure 7 - "10-15 Japanese Cycle" [17] ......................................................................................... 14
Figure 8 - JC08 Japanese Cycle [17] ............................................................................................. 14
Figure 9 - FTP-75 EPA Federal Test Procedure [17] ................................................................... 15
Figure 10 - HWFET EPA Federal Test Procedure [17] ............................................................... 15
Figure 11 - US06 EPA Federal Test Procedure [17] .................................................................... 16
Figure 12 - SC03 EPA Federal Test Procedure [17] .................................................................... 16
Figure 15 - Artemis Highway Cycle [17] ......................................................................................... 17
Figure 13 - Artemis Urban Cycle [17] ............................................................................................. 17
Figure 14 - Artemis Rural Cycle [17] .............................................................................................. 17
Figure 16 - Comparison of existing test-cycles and a reference real-world measurement .... 18
Figure 17 - WLTP Class 1 cycle (<22kW/ton) ............................................................................... 19
Figure 18 - WLTP Class2 Cycle (22 to 34 kW/ton) ...................................................................... 19
Figure 19 - WLTP Class 3 Cycle (>34 kW/ton) ............................................................................. 20
Figure 20 - Comparison of WLTP test-cycles and a reference real-world measurement ...... 20
Figure 21 - Chassis Dynamometer with Constant Volume Sampler (CVS) gas flow
measurement system [26] ................................................................................................................ 21
Figure 22 – Diesel Engine Fuel Consumption Map [g/kWh] and power curves [horsepower]
[21] ....................................................................................................................................................... 23
Figure 23 - Fuel Efficiency with vehicle age [28] .......................................................................... 24
Figure 24 - CO2 emissions evolution with average speed [6] ..................................................... 25
Figure 25 - Fuel Consumption with acceleration rate (purple - aggressive, red - calm) [27] . 27
Figure 26 - Comparison of drag and rolling friction with speed .................................................. 30
Figure 27 - Typical Urban VSP mode Distribution ........................................................................ 31
Figure 28 – Extra Urban VSP Distribution ..................................................................................... 32
Figure 29 - VSP mode distribution of a racing car on track ........................................................ 32
Figure 30 - Simplistic representation of methodology algorithm ................................................ 35
Figure 31 - TCGT Graphical User Interface .................................................................................. 36
Figure 32 – Dropdown Menu - Fine Tuning of Approximation Criteria (weight vector)........... 37
Figure 33 - Dropdown menu - Test Cycle duration limit .............................................................. 38
Figure 34 - "+info" window with the list of blocks used to generate the test cycle .................. 41
Figure 35 - Block altitude reduction ................................................................................................ 43
Figure 36 - Block duration re-sampling ("stretching") .................................................................. 43
Figure 37 - Road slope calculation ................................................................................................. 46
Figure 38 - Definition of "phi" parameter for driving cycle classification ................................... 48
Figure 39 – Blocks extracted from: WLTP Class 3 Cycle ........................................................... 51
Figure 41 - Blocks extracted from: WLTP Class 2 Cycle ............................................................ 51
Figure 40 - Blocks extracted from WLTP Class 1 Cycle ............................................................. 51
Figure 42 - Blocks extracted from US06 cycle of EPA FTP ........................................................ 52
Figure 43 - Blocks extracted from CSHVC cycle .......................................................................... 52
iv
Figure 44 - TCGT –Calm urban driving cycle ................................................................................ 53
Figure 45 - TCGT - Typical urban ................................................................................................... 54
Figure 46 - TCGT - Rural.................................................................................................................. 54
Figure 47 - TCGT - Aggressive rural driving ................................................................................. 54
Figure 48 - TCGT - Highway driving ............................................................................................... 54
Figure 49 - ADVISOR GUI post simulation window for an Hybrid Vehicle ............................... 60
Figure 50 - ADVISOR GUI post simulation window for conventional IEC vehicle ................... 60
Figure 51 - Fuel Consumption Comparison................................................................................... 64
Figure 52 - Unburned Hydrocarbons Comparison ....................................................................... 66
Figure 53 - Carbon Monoxide Comparison of all 5 vehicles ....................................................... 66
Figure 54 - Carbon Monoxide comparison of 4 vehicles (all except "gasoline 1") .................. 67
Figure 55 - Nitrogen monoxide and dioxide Comparison ............................................................ 68
Figure 56 - Particulate Matter Comparison .................................................................................... 69
Figure 57 - VSP Mode Distribution Deviation vs. cycle duration ................................................ 71
Figure 58 - Energy per kilometer deviation vs. cycle duration .................................................... 72
Figure 59 - VSP mode distribution with "urban" weight vector ................................................... 73
Figure 60 - VSP mode distribution with "high-way" weight vector ............................................. 73
Figure 61- VSP mode deviation with "weight vector" for 4 common "phi" parameters ........... 74
Figure 63 - Test Cycle generated through TCGT ......................................................................... 75
Figure 62 - Original Driving Cycle ................................................................................................... 75
Figure 65 - Acceleration [m/s2] vs. Speed [km/h]; Blue - Generated, Green - Original .......... 76
Figure 64 - TCGT output for the 22000 seconds input driving cycle ......................................... 76
Figure 66 - VSP mode distribution of the 50 drivers .................................................................... 77
Figure 67 - TCGT results for the VSP mode distribution data of 50 drivers ............................. 78
List of Tables
Table 1 - EU Emission Standards for Passenger Cars [17] ........................................................ 11
Table 2 - VSP mode values [3] ........................................................................................................ 30
Table 3 - Weight Vector description for fine tuning of the VSP mode approach ..................... 48
Table 4 - Blocks from which test cycles are created (picked/combined) .................................. 50
Table 5 - Properties of the 5 Generated Cycles for TCGT Validation ....................................... 55
Table 6 - Properties of the 5 Generated Cycles for TCGT Validation (cont.) ........................... 55
Table 7 - Properties of the 5 Generated Cycles for TCGT Validation (cont.) ........................... 55
Table 8 - Fuel Consumption and Pollutant Emission per VSP mode (BMW 114i) .................. 57
Table 9 - Fuel Consumption and Pollutant Emission per VSP mode (Chevrolet Cavalier 2.2)
[15] ....................................................................................................................................................... 58
Table 10 - Main properties of the 5 cars used for Validation through ADVISOR .................... 61
Table 11 - Comparison between original and generated cycles (using ADVISOR) ................ 61
v
Abstract
In order to launch a new car model, a vehicle manufacturer has to ensure that it follows the applicable
regulations and standards. With concern for the welfare of the environment, emission standards
impose maximum allowable amount of pollutants a car can emit. In order to assess the pollutant
emission rate of the certifiable vehicle, it has to perform a type-approval test cycle on a chassis
dynamometer, while tailpipe gases are collected and analyzed. Currently used test cycles for vehicle
certification are not representative of real-world drive, showing lower fuel consumption and pollutant
emission. This issue has been discussed by various investigators, and is currently being addressed
by Agencies and Commissions of the leading motorization countries.
This research work attempts to implement a methodology capable of generating test cycles with
limited duration, suitable for car certification process, which would closer represent the real-world
data. To validate and apply the test cycle generation methodology, a tool was developed that uses
real-world data as input, to generate test cycles with similar characteristics. The generated cycles are
composed of segments of commonly used test cycles.
Generated Test Cycles were validated using vehicle simulation software, ADVISOR, as well as
through the deviation of the test cycle design criteria (average speed and Vehicle Specific Power
modes distribution). On average, with the test cycle limited to ~2100 seconds, the deviation between
the real-world driving data and the test cycle that was generated based on it, is lower than 1% for
average speed, 2% for specific power, 2.5% for fuel consumption and 7% for pollutant emissions.
Keywords
Driving cycle; Real-World Data; Vehicle Specific Power (VSP); Fuel Consumption; Pollutant
Emissions
vi
Resumo
Para lançar um novo veículo no mercado, o fabricante, deve assegurar que este respeita as normas
e regulamentações aplicáveis. Na tentativa de preservar o meio ambiente, os padrões de emissões
impõem os valores máximos para a emissão de poluentes. A taxa de emissão de é determinada ao
efetuar o ciclo teste de homologação, no banco de rolos, enquanto os gases de escape são
recolhidos e analisados. Os ciclos de teste de homologação, atualmente em vigor, não representam
os dados observados na condução no mundo real, mostrando níveis de consumo de combustível e
emissão de poluentes inferiores. Esta incoerência é atualmente tema de estudo de vários
investigadores, também como está a ser discutido por Agências e Comissões dos países lideres no
ramo da indústria automóvel.
No presente trabalho está a ser desenvolvida uma metodologia capaz de gerar ciclos de condução
com duração limitada, adequados para ser usados como ciclos de teste para certificação dos
veículos e ao mesmo tempo permitem uma representação melhor da condução em estrada. Para
aplicar e validar a nova metodologia criada, foi desenvolvida uma ferramenta que utiliza como
entrada dados recolhidos em condução real, para gerar ciclos teste com comportamento semelhante.
Os ciclos teste gerados são validados através do simulador automóvel, ADVISOR, também como
pelo desvio dos parâmetros de geração da nova metodologia (velocidade média e distribuição dos
modos da potência específica do veículo, VSP). Em média, o desvio entre os resultados da condução
em estrada e o ciclo de teste correspondente, estão abaixo de: 1% para velocidade média, 2% para
potência específica, 2.5% para consumo de combustível e 7% para a emissão de poluentes.
vii
Nomenclature
CADC Common Artemis Driving Cycles
CO Carbon Monoxide
CO2 Carbon Dioxide
CVS Constant volume sampler
ECE Urban Driving Cycle
EGR Exhaust Gas Recirculation
EPA Environmental Protection Agency (US)
EUDC Extra Urban Driving Cycle
FID Flame Ionization Detector
FTP Federal Testing Procedure (United States)
HC Unburned Hydrocarbons
JC08 Japanese Cycle 2008
KDC Korean on-road Driving Cycle
MPG Miles per Galon
NDIR Non-Dispersive Infra-Red
NEDC New European Driving Cycle
NOx Nitrogen monoxide and dioxide
OBD On-Board Diagnostics
OICA Organization of Motor Vehicle Manufacturers
PEMS Portable Emission Measuring System
“Phi” Created parameter to classify driving data (by average speed and
aggressiveness)
PM Particulate Matter
PN Number of Particles
SoC State of Charge
SPTF Supplemental Federal Test Procedure (for the FTP)
TCGT Test Cycle Generating Tool (the developed program to implement the
adopted methodology for test cycle generation)
UNECE United Nations Economic Commission for Europe
VSP Vehicle Specific Power
WLTP World Light Test Procedure (new test cycle developed by EU, Japan and
India, planned to be released and implemented in 2014)
viii
Acknowledgments
I would like to address special thanks to my advisors Gonçalo Nuno Antunes Gonçalves and Gonçalo
Nuno de Oliveira Duarte. Their feedback and input along the development of this thesis were
fundamental and decisive. Their know-how in the subject, allowed the correct guidance along the
way, through frequent meetings that we had during the progress of the thesis.
Thanks to Instituto Superior Técnico (IST) for providing the optimal conditions to achieve the level of
knowledge a starting Mechanical Engineer should possess, through the fair yet rigorous evaluation
system. Special thanks to the professors José Mendes Lopes, João Luís Toste de Azevedo and
Tiago Lopes Farias for their distinct point of view at the mechanical engineering subjects, and their
contribute to the growth of my critical approach on analyzing the achieved results.
I feel the need to show gratitude to my family and my girlfriend for the moral support through the
hardest times, as well as for the financial support given, since without this indirect input the working
conditions would not have been the same and the final outcome could have been lesser.
1
1. Introduction
1.1. Motivation
There is no deny or doubt in the usefulness of road transport sector and how currently the society is
dependent of it. The automobile is by far the favorite type of transport on short to medium range and
excluding underdeveloped countries, it is heavily used in day by day routine. [1]
When cars became increasingly more popular (mid XX century), its negative impact on environment
became more obvious. The main environmental problems created by automotive sector can be
grouped in three major areas:
resource depletion
greenhouse gas emissions
local pollution
To address these issues, emission standards were implemented, around 70‟s in US, later in Europe
and Japan and as of mid-90‟s around the most part of the World (The following chapters will refer
which regulation entities exist and which values of pollutant emissions the standards imply).
According to International Organization of Motor Vehicle Manufacturers (OICA) [1] the world cars
production rate is increasing, as can be observed in Figure 1:
Figure 1 - Evolution of Cars Production over last decade [1]
Note: by the name “car”, it is considered a passenger car, that is, a 4 wheel motorized vehicle for
passenger transportation with no more than 8 seats in addition to the driver‟s seat. Cars (or
automobiles) make up approximately 74% of the total motor vehicle annual production in the world.
35
40
45
50
55
60
65
1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011
Mill
ion
car
s p
er
year
Number of Cars Produced WORLDWIDE
2
The chart in Figure 1, uses the same definition of car as described here. It is estimated that over 1
billion passenger cars travel the streets and roads of the world today. [1]
The average increase of the car production volume, on year to year basis, is of 3.7%. It is expected to
see the same rate of increase in annual cars production in the near future, despite the current
economic crisis. Even though the number of cars produced and that circulate on the roads has
increased continuously over the last couple of decades, the pollutant emissions and fuel consumption
from transport sector did not increased by the same rate, and occasionally even decreased from year
to year, as can be observed in Figure 2 [2]
Figure 2 - US total emissions from mobile sources 1970-2005 [3]
This disparity between the increasing number of vehicles sold and the decrease on pollutant
emissions is explained by the ever more rigorous standards that were and are imposed to car
manufacturers [3] [4] [5]. In order to maintain this trend, Emission Standards Entities have to keep
improving the limit levels imposed by the standards (within reasonable engineering, economic and
technological boundaries for the manufacturer) [4].
In order to maintain the lifestyle that the car provides to society, vehicle manufacturers have to keep
up with the continuously increasing demand caused by world population increase, by countries
development, by increasing level of commodity desired by majority and ever more accessible cars
prices on market. The continuous evolution in technology also allow new car models to surface on
market on a regular basis. For each new car or new engine, a test procedure must be performed,
which consists of executing a driving test cycle on a chassis dynamometer while collecting and
measuring tailpipe emissions in order to assess the car‟s performance regarding the fulfillment of the
requirements imposed by the emission standards. A Test Driving Cycle is a speed vs. time dataset
that is supposed to represent a real world driving style, thus allowing to simulate real world conditions
in an enclosed and fully controlled laboratorial environment.
In Europe, the currently used Test Cycle is New European Driving Cycle (NEDC), which is based on
ECE-15 Urban Driving Cycle that was created in 1970 (test cycles will be exposed in more detail in
3
further chapters). Taking into account the evolution of automotive industry technology, and driving
style over more than 30 years since the implementation of NEDC test cycle, it is not surprising that
the results obtained by performing the cycle don‟t match and do not represent real world average
European emissions/fuel consumption, or in other words, under laboratorial certification tests the car
respects the standard homologation requirements, while on the road in many cases it does not [6]. It
has been recognized for some time that the NEDC does not adequately capture actual „real-world‟
driving patterns. [7] [8] [9] [10]
Many characteristics of the NEDC test cycle are not representative for real-life driving behavior (low
accelerations, low maximum speed, high idling time, constant speeds instead of transients, favorable
shifting points, etc.) [5]. For this reason considerable effort has been dedicated to develop alternative
drive cycles that are designed to be more representative of actual conditions e.g the work of. Andre
(2004) [11]. Recently, this motivated European Union in partnership with India and Japan to work on
the development of a new Driving Cycle, World Light Test Procedure (WLTP), which is to be
implemented in 2014. [12]
In support of the inaccuracy of the NEDC test cycle, recent study shows that the difference between
type-approval and real-world CO2 emission levels of new passenger cars in Germany has increased
from about 8% in 2001 to about 21% in 2010. The widening of the gap is especially noticeable since
2007 when mandatory CO2 emission standards for the EU were under development. [5] The analysis
confirms that there was a decrease in the level of CO2 emissions of new passenger cars in Germany
since 2001. However, the magnitude of reduction in reality appears to be only about half of what is
suggested by the type-approval values (about 7% instead of 15% since 2001). [5] Should be
mentioned, that CO2 emission is totally dependent on fuel consumption, as per each liter of diesel and
gasoline there is a specific amount of carbon dioxide emitted, thus making the conclusions above
about CO2 discrepancy applicable as well to fuel deviation (measured/real-world).
United Nations Economic Commission for Europe (UNECE) and United States Environmental
Protection Agency (EPA) are currently involved in development of new methods for a more accurate
and reliable technique of type-approval tests for vehicle certification, recognizing the inaccuracies of
current in-use test cycles and test environments. [12] [13] This clearly supports the interest in tackling
this problem, as the air quality is progressively becoming a more cautious matter and the number of
vehicles that circulate the roads is increasing determinedly.
An adequate test cycle will allow government to have better idea on the expected pollutant emissions
as well as to allow to do better estimations on the level of fuel consumption and GHG emission within
its territory in order to be able to execute and predict national energy plans with higher accuracy.
4
1.2. Current studies and the State of the Art
Various studies are made around the world, by independent companies, agencies and commissions
[12] [13], as well as researchers at various engineering schools. Without exceptions, all agree on the
disparity in the results obtained in the lab, while performing the type-approval test cycle (especially
NEDC, the oldest one) and the results measured on real world driving conditions.
Taewoo Lee‟s research [14], clearly supports the innacuracy of current type-approval test-cycles, and
compares the New European Driving Cycle (NEDC) to the performance of a Korean typical driving
conditions cycle called: Korean on-road Driving Cycle (KDC) which is far more accurate on
representing the measured fuel consumption and NOx emissions of Korean popullation (in this study,
only NOx was mainly focused).
They designed the KDC using two sets of Korean driving data, one for development and the other for
validation. Both datasets were based on the independent measurement of real-world vehicle
operating in Korea. “We constructed candidate KDCs by combining sub-cycles of this dataset
because their drivability has been assessed to be unproblematic.” [14]The usage of sub-cycles from
already existing test driving cycles, to generate the KDC, is similar to the methodology presented later
in chapter 3. It assures that it is realistic to reproduce the test cycle in the laboratory conditions on the
rolling bench, as it is constituted of already existent and performed segments of existing test cycles.
Figure 3 - Pollutant emissions for several measured vehicles, on 4 routes and NEDC
5
The recent work of Martin Weiss, on measuring on-road emission of light-duty vehicles in Europe
once again proves the underestimation of real world result by the NEDC test cycle, as can be seen in
the Figure 3:
In the majority of cases, the measured emissions from the 4 real world routes, with the Portable
Emission Measuring System (PEMS), have higher values compared to the ones resulting from the
NEDC test cycle. In the case of diesel vehicles, the measured emission values of nitrogen monoxide
and dioxide surpass outrageously the maximum admissible levels by the emission regulations (3.5x in
the case of C, Diesel vehicle). Carbon monoxide only surpasses the regulated limit in the case of
gasoline L vehicle, on the motorway route.
The goal of the type-approval test cycle is to generate fuel consumption and pollutant emission rates
similar to the ones encountered in real life driving. Ideally, the NEDC results in the Figure 3 should be
around the weighted average between the 4 routes (averaged by the percentage of time usually spent
in each mode). Judging by the data presented in Figure 3, this is certainly not the case with the NEDC
test cycle.
In the ideal world, each new car model would be driven for a sufficient amount of time/distance, while
tailpipe emissions and fuel consumption would be measured onboard, and statistic data would be
calculated to determine if car complains or not with the regulations, but such challenge is
economically not practicable. While deciding which test cycle to use for a specific car model, the
country, age group (driving style), specific power (power to weight ratio) should be taken into account,
as well as real world driving data of similar models if available.
In order to achieve a correct measurement in the testing procedure, the key is to improve not only the
test cycle, but also the test procedures, namely [5]:
Cold start testing is performed at ambient temperatures close to 30 °C, while real life
temperatures are lower, leading to higher fuel consumption.
The allowed tolerances and flexibilities in the road load test procedure cause the road load of
type-approval vehicles to be lower than that of production vehicles.
At the type approval test the battery is charged to 100% capacity.
The type approval test weight is lower than the real-life average.
Vehicles are type-approved without the air conditioning system turned on (or any other power
consuming equipment).
Potentially increasing exploitation of existing tolerances and flexibilities in the road load
determination test procedure and the type approval test procedure.
The introduction of start/stop systems in recent model years, whose positive influence is likely
more visible in the type approval than in real-life, etc.
The above parameters, that have an impact on the fuel consumption and pollutant emissions while
performing the test cycle on the dynamometer, are studied in depth in many publication researches.
6
This work, is mainly focused on test-cycle itself, and not on the procedures mentioned above (that
should be respected and enhance regardless of the quality and representativeness of the type-
approval test cycle).
Once again is worth mentioning that the United Nations Economic Commission for Europe (UNECE)
and United States Environmental Protection Agency (EPA) recognize the inaccuracies of current in-
use test cycles and test environments. They are currently involved in development of new methods for
a more accurate and reliable technique of type-approval tests for vehicle certification. [12] [13]
1.3. Objectives
According to what was referred in the previous sections, the objectives within this work are
Analyze the current state of certification test cycles
Develop a methodology that allows generation of test cycles based on real world data
Create Test cycles with limited duration and better representation of the real-world driving
conditions
Validate the accuracy of test cycle generating methodology and apply it to specific cases
The main objective of this work is to create and test a methodology that would allow the
creation/generation of test driving cycles that would perform and represent accurately certain given
real world driving data (given by a Vehicle Specific Power distribution, in further chapters) or real
world driving data (speed and altitude). By accurate representation of the desired real-world data, it is
envisioned to have a test cycle that presents fuel consumption and pollutant emissions (NOx, CO, HC
and PM), per kilometer, comparable to the real-world values. The test cycle duration is around 30-40
minutes mark, while the real-world data used can have from hours to months worth of data.
While creating a single and unique, worldwide accepted and overall accurate driving test cycle is
tempting, such goal is impossible. Driving styles of European, Asian, North and South America drivers
are quite different, even from country to country the difference is persistent, on the other hand, car
models specificities, from small economic urban cars to powerful off-roaders or sport cars, turning
impossible the task of creation of a test cycle with global representativeness [15]. As for example a
low average speed, urban test cycle for the simulation of a small economic car with a calm driver, will
not simulate the driving pattern of a powerful sport car by an aggressive driver [16]. This means that
test cycles should be tailored to suit and represent as best as possible the real conditions that
particular car model will encounter in its future.
7
In order to be able to obtain test cycles for this enormous variety of conditions, as well as to have a
test cycles generation procedure that is easy to adjust in future, it was decided to create a tool which
generates test cycles given the input data in the form of real world driving style or real world driving
data (as mention in the beginning of the sub-chapter).
Such a tool would be a future proof method of obtaining driving cycles which, performed under
laboratorial testing conditions, would show levels of pollutant emissions and energy consumption (fuel
consumption) much closer to reality as well as allow the generation of specific driving test cycles for
each car model in particular, if desired by the regulator entity.
Allowing each country in the world, to easily generate a test-cycle that would represent its population
(non-expensive speed data loggers mounted on OBD ports, on typical vehicles for that country, and in
different regions of its territory, can easily provide a high sample of data to create a VSP mode
distribution and average speed of its population), would open the potential of a much more precise
report of global contribution of GHG and pollutant emissions, further allowing the implementation of
suitable measures for corrections or adjustments (VSP is introduced in chapter 3.1).
Having a robust technique of generating test cycles based on real-world input data, and after
generating some for specific cases, next step will be to assure their accuracy and their
representativeness of real-world data used as input. The validation process will be made initially with
the Vehicle Specific Power (VSP) distribution methodology, and afterwards using a vehicle simulation
environment called “ADVISOR”.
1.4. Structure of the Thesis
The Thesis is divided in 7 chapters:
In the second chapter the current state of Emission Regulating entities is presented, as well as the
evolution of the imposed limits over the past couple of decades. The existing, past and future driving
test cycles are presented in this chapter as well.
In the third chapter is presented the criteria by which test cycles are generated, what is Vehicle
Specific Power and why it represents so well the driving style. Also in this chapter we talk about the
approach on generating test cycles based on real world input data.
In the fourth chapter, we talk about the tool created to generate test cycles, based on data from
different sources. How it works in background, user interface explanation and more.
In the fifth chapter, the test cycles are then used in a vehicle simulation tool, Advisor, to verify if they
represent accordingly the input data. Also a VSP based comparison is made in this chapter. Further it
8
will be evaluated how test cycle accuracy changes with test cycle duration, as well as a statistic
analysis on the consistency of generated cycles.
In the sixth chapter, the tool is used on several case studies to generate test cycles representative of
the driving data and comparison to the NEDC and WLTP test cycles are made.
In the seventh and last chapter conclusions are made on the reliability and accuracy of the developed
methodology and the created tool. Also we talk about what future work could be done to improve the
tool as well as what can be done alternatively to assess cars emissions and consumption.
In Annex, a part of the MatLab code of the program (TCGT) is presented, with additional screen
captures of different test cycles generations.
9
2. Background
The combustion of fossil fuel, or Hydrocarbons (HC), inside an internal combustion engine ideally
would release, beside thermal energy, only water (H2O), carbon dioxide (CO2) and inert Nitrogen (N2).
However it is well known that the combustion never occurs in ideal conditions, thus not all the fuel
completely reacts with oxygen (e.g. because of time available for the reaction, local low temperature,
quenching or local unbalanced air to fuel ratio) and as a result it leads to the presence of unburned
HC in exhaust emissions. Unburned hydrocarbons are toxic and carcinogenic to humans.
The high pressure and temperature present in the combustion chamber (inside the cylinder) allow the
dissociation of the molecular Nitrogen and its subsequent reaction with Oxygen to originate NOx
compounds. NOx can react with moisture to form nitric acid, which can penetrate deep into lungs and
originate respiratory diseases; also NOx can react with volatile compounds to form ozone, which also
may cause respiratory difficulties and diseases.
In case of Spark-ignition engines, the Air to Fuel ratio is considerably lower compared to Diesel
engines, thus causing some partial lack of full oxidation of carbon, originating carbon monoxide (CO),
while in the case of Diesel engines, a poor atomization of fuel and quenching creates areas where
air/fuel ratio is low (while overall Diesel engines have lean mixtures) resulting in carbon deposits
(particulate matter (PM) or black smoke).
Regarding the health impacts caused by CO and PM. Carbon monoxide is absorbed by hemoglobin
and in consequence turns difficult the oxygen transportation to vital organs, while particulate matter
increases the chance of cancer risk. All the local pollutants mentioned have a particularly high impact
Figure 4 - Evolution of CO2 levels of emission over the last centuries
10
in jammed urban environments, where the levels of concentration of these pollutants are extremely
high.
As can be verified in the Figure 2, overall CO2 emission is increasing at a steep rate, and the
component coming from petroleum combustion (mainly gasoline and diesel) also increases. Well
known is the fact that CO2 gas contributes to the greenhouse effect, thus making its contribution
towards global warming. The CO2 emission of a modern car is almost a linear function of its fuel
consumption (exceptions to this linearity are the punctual cases when the CO and/or HC
concentrations are very high), and is given by: 2.68 kg of CO2/liter of Diesel and 2.31 kg of CO2/liter of
Petrol. The only way to reduce CO2 emissions is to turn the vehicles more fuel efficient.
European Union legislation adopted in 2009 sets mandatory emission reduction targets. The fleet
average to be achieved by all new cars is 130 grams of CO2 per kilometer (g/km) by 2015 and 95g/km
by 2020 (or using above linear relation, in terms of fuel consumption, the 2015 target is approximately
equivalent to 5.6 l/100 km of petrol or 4.9 l/100 km of diesel and the 2020 target equates to
approximately 4.1 l/100 km of petrol or 3.6 l/100 km of diesel) [5]. Adopted legislation is flexible
towards small or new car manufacturers, and 2020 target might be out of reach for most of the
manufacturers, but overall the legislation is showing a determination in the reduction of fuel
consumption in cars.
2.1. Emissions Regulations
While CO2 emission is fully defined by the fuel consumption, in the case of NOx, PM, HC and CO it is
possible to reduce their concentration in the exhaust gas by fine-tuning the engine operation and with
advancement in car manufacturing technology. In order to reduce car‟s emission of the above
indicated pollutants, Emission Standard regulation entities were created initially in US, Europe and
Japan, and later in China, India etc. Emission Standard defines what is the maximum allowed amount
of carbon monoxide, nitrogen oxides, unburned hydrocarbons and particulate matter emission (CO,
NOx, HC and PM), usually given in mass per distance traveled (grams per kilometer) or in case of
heavy duty trucks in mass per energy (grams per kilowatt*hour). The limits stated by the regulation
are continuously updated forcing car manufacturers to an endless pursue of engineering perfection, in
order to reduce emission levels to the minimum attainable by the current state of the art technology.
New introduced regulations apply only to the new cars yet to be sold and there are certain tolerances
towards older engines and also to cars that were released to manufacturing shortly before the new
regulation came into effect.
11
EU Emission Standards for Passenger Cars
Stage Date
CO HC HC+NOx NOx PM PN
g/km #/km
Compression Ignition (Diesel)
Euro 1 1992.07 2.72 - 0.97 - 0.14 -
Euro 2 1996.01 1 - 0.7 - 0.08 -
Euro 2 1996.01 1 - 0.9 - 0.1 -
Euro 3 2000.01 0.64 - 0.56 0.5 0.05 -
Euro 4 2005.01 0.5 - 0.3 0.25 0.025 -
Euro 5 2009.09 0.5 - 0.23 0.18 0.005 -
Euro 5 2011.09 0.5 - 0.23 0.18 0.005 6.0×10^11
Euro 6 2014.09 0.5 - 0.17 0.08 0.005 6.0×10^11
Positive Ignition (Gasoline)
Euro 1 1992.07 2.72 - 0.97 - - -
Euro 2 1996.01 2.2 - 0.5 - - -
Euro 3 2000.01 2.3 0.2 - 0.15 - -
Euro 4 2005.01 1 0.1 - 0.08 - -
Euro 5 2009.09 1 0.1 - 0.06 0.005 -
Euro 6 2014.09 1 0.1 - 0.06 0.005 6.0×10^11
Table 1 - EU Emission Standards for Passenger Cars [17]
As can be observed in the Table 1, every 3 to 4 year the European Union Emission Standards for
passenger cars are updated, and the evolution from 1992 to 2005 is very significant. While attempting
to comply with the emission standards, it is common for a car manufacturer to reach a relatively close
to the limit value in one of the pollutants (one of the columns in the table) but obtain a order of
magnitude or even lower values of emission of other pollutants. [18] On average, a modern car will
pollute from 3 to 5 times less than a 20 year old car (and if we would compare a modern high-tech
vehicle with the Model-T Ford from 1 century ago, a single Ford Model T would emit as much NOx,
HC and PM as over 100 modern cars). All local pollutants mentioned in the table are sensitive to
aggressive accelerations and engine cold start, so a car that covers daily relatively large distances will
emit much less g/km of pollutants than a car which travels very short distances several times per day.
That being the task of a test cycle, to correctly represent the average of this driving behaviors.
Emission regulation entities are very important in order to sustain a healthy environment (especially in
urban areas) and to reduce fuel consumption (the ever more scarce and expensive fossil). As a
matter of fact, thanks to the continuously more strict regulations, even with the increasing number of
vehicles produced and circulating on roads every year, the total amount of pollutants emitted is
decreasing. This was observed in the “Figure 2”, which represents the reduction in total emissions
from automobiles in US over a 35 year period (from 1970 to 2005). It would be unreasonable to
12
suggest same level of reduction from 2005 to 2040 while still using internal combustion engines as
the main propulsion system.
The U.S. and Japan emission standards are not absolutely equal to EU standards, but they are close
and also have the same trend of reducing the maximum allowed levels with each new released
regulation (the correspondent regulations tables can be consulted at [17]).
As for the rest of the world, standards tend to follow European or American practice (see Figure 5).
There is a considerable delay in the adoption of the new standards by the non-developed or in-
development countries, for example the Euro VI standard is enforced in Europe in 2014, at a point,
where most other countries will still be mostly using Euro I up to Euro IV standards. [3]
A more in depth analysis of the evolution from Euro 4 to Euro 6 can be found in the work of Martin
Weiss and others, found in Bibliography with reference [19].
Figure 5 - Emission standards for other countries [3]
2.2. Test cycles
The maximum levels of emissions allowed are established, but how can be easily and consistently
measured the emissions from the new car models? This can be achieved by performing a test driving
cycle on top of a chassis dynamometer while measuring fuel consumption and pollutant emissions
directly from tailpipe exhaust. Initially the emission testing was made with the engine on idle, but this
13
changed in favor of drive cycle based emissions testing, as the later is closer to real world engine
behavior. Of course using a mobile laboratory and analyze gases emissions in real world driving,
under various condition and over a statistically long enough amount of time would be the best
possible analysis for a car, regarding fuel use and level of pollutant emissions, however such a
complex process would take too much time and would be expensive and difficult to guarantee
consistency from test to test and from car to car.
The currently used Testing Driving Cycle in Europe is the New European Driving Cycle (NEDC). It
consists of 4 ECE (or UDC) driving cycles (1970) and one EUDC driving cycle. ECE driving cycle is
intended to simulate an urban oriented driving style while the EUDC driving cycle represents a more
high-way like driving style. As can be seen in the Figure 6, the driving cycle is very simplistic, and as a
consequence vehicle and engine manufacturers found a way to explore the limited number of
conditions in the cycle by programming their engine management systems to control emissions to
regulated levels at the specific test points contained in the cycle (at 35, 50, 55, 70 km/h etc. “steady
states”) , but they still create a great deal more pollution under conditions experienced in real
operation and not captured within the execution of the test cycle. [2] Putting it by another words, car
manufacturers found a way to manipulate the engine management during the testing procedure in
order to show lower emission values than otherwise encountered afterwards in real world driving. As
a result, the real emissions are higher than the standards allow, thus undermining the regulations and
public health.
Apart from Europe, the other two big automotive industries, Japan and USA, have their own test
driving cycles, and their latest versions are JC08 and FTP-75 respectively.
Figure 6 - NEDC Driving Cycle (4x UDC + 1x EUDC) [17]
14
The JC08 driving cycle, created in 2005 and introduced gradually over 2008-2011 years, comes to
substitute the previous 10-15 cycle, due to the need of having a better reality representing cycle. The
previous, 10-15 driving cycle had a similar style of the NEDC cycle, which is very sketchy/choppy and
simplistic (although without a doubt more friendly and easy for the driver who performs the test). A
comparison between 10-15 and JC08 can be found in the Figure 7 and Figure 8:
As can be observed in the Figure 7 and Figure 8, the 10-15 cycle has 3 urban components and 1
mixed component and the JC08 cycle is more oriented towards urban driving as well, with a mixed-
highway component at the end.
The FTP-75, or EPA Federal Test Procedure, cycle used in the United States suffered several
changes along the years, and as of 2008, it is composed of 4 parts: city driving (FTP-75), highway
Figure 7 - "10-15 Japanese Cycle" [17]
Figure 8 - JC08 Japanese Cycle [17]
15
driving (HWFET), aggressive driving (SFTP US06), and optional air conditioning test (SFTP SC03).
All the additions to the original FTP-75 urban oriented cycle are justified by the need to represent
better the average real world driving conditions, as they are not only composed of city driving. The
FTP-75 cycle is a successor of FTP-72 cycle which as well is an urban oriented test cycle. In Figures
9 to 12 are represented the 4 cycles that makes the current FTP-75 US driving test cycle:
Figure 9 - FTP-75 EPA Federal Test Procedure [17]
Figure 10 - HWFET EPA Federal Test Procedure [17]
16
Figure 11 - US06 EPA Federal Test Procedure [17]
Figure 12 - SC03 EPA Federal Test Procedure [17]
As can be observed in the Figures 9 to 12, the US cycles have a somewhat more similar to real world
driving style behavior, and thus it is not surprising that the consumption and emission values
measured while performing them are closer to the real world, compared to NEDC or 10-15. [13]. All
this cycles suffered some adjustments over the years, especially the conditions at which the test cycle
are initiated, be that a cold start or warmed up start (engine starts a couple of minutes before the test
and it achieves its normal working temperature).
It is of no surprise that the leading countries in car manufacturing and with high cars per capita ratio
are the ones who invest and try to create the emission regulations and the testing procedures to
assess them. In order to give better results, within European Artemis (Assessment and Reliability of
Transport Emission Models and Inventory Systems) Project, the Common Artemis Driving Cycles
(CADC) were constructed. They have a much better representation on real world European driving
data, but are not currently used for car certification. The cycles were made to match as close as
possible European driving speeds, accelerations, gear changes, traffic jams etc. The 3 CADC cycles
are represented in the Figure 13, Figure 14 and Figure 15:
17
Figure 15 - Artemis Highway Cycle [17]
Figure 13 - Artemis Urban Cycle [17]
Figure 14 - Artemis Rural Cycle [17]
18
Later, in chapter 3.1, the concept of Vehicle Specific Power (VSP) will be introduced in detail, but for
now, it is sufficient for the reader to know that VSP is given in kW/ton, and represent the power which
is required from the wheels in order for the car to perform the movement it does, and further the
absolute VSP values are classified in modes: mode 1 and 2 are negative powers (breaking), VSP
mode 3 is null power, stopped or gas pedal lift-off, while VSP mode 4 up to 14 represent increasing
power demand. A similar distribution of VSP values by the 14 modes of two different trips, would
suggest a strong similarity in their average kinematic behavior, thus similar fuel consumption and
pollutant emission (as will be described in more detail in chapters 3, 4 and 5). To have a better
understanding of representativeness of the existing and currently in use type-approval test cycles,
next is presented a chart that shows the VSP mode distribution for each existing cycle, as well as a
measured real-world driving cycle as a reference for comparison.
Figure 16 - Comparison of existing test-cycles and a reference real-world measurement
It is evident that the time distribution by the 14 VSP modes is significantly different from one test-cycle
to another (which already shows a certain level on inconsistency) and there is no test-cycle that
matches well the “Measured” real-world data curve [20]. The US06 cycle, being developed to simulate
high-way situations, has a greater time share in higher VSP modes, and will present higher fuel
consumption than Measured cycle, while all the others, have lower high VSP mode distribution, thus
underestimating the real-world fuel consumption and pollutant emission.
Europe, Japan and India joined their forces to create an International harmonized test cycle, The
Worldwide harmonized Light vehicles Test Procedures (WLTP), which would be both: more accurate,
and allow easier comparison between countries in respect of cars consumptions and emissions
0
5
10
15
20
25
30
35
40
45
1 2 3 4 5 6 7 8 9 10 11 12 13 14
% T
ime
VSP Mode
10-15
FTP-75
SC03
US06
NEDC
Measured
19
(setting an international standard for test procedure). The test cycle is expected to have a final
release version around 2013-1014. As mentioned above, creating a single cycle that would represent
the vast variety of combinations of car/engine types and driving styles is naive and even impossible,
however there is no doubt that it is possible to largely improve on the existent NEDC and JC08 cycles
(current European and Japanese test cycles). In order to better represent the different cars models
and engine types, WLTP has 3 cycles, for 3 different specific power classes. Vehicles with under 22
kW/ton are included in class 1, from 22 to 34 kW/ton in class 2, and class 3 cycle is for vehicles with
more than 34 kW/ton (the vast majority sold on European Market). Interpreting the 3 cycles of WLTP,
it is easy to understand that class 1 and class 2 cycles are oriented towards Indian market cars, class
2 and 3 towards Japanese market vehicles while in Europe, having in mind the car models frequently
bought/used, class 3 cycle will be the most used one. The 3 cycles for the 3 specific power classes of
vehicles are presented in the Figure 17,Figure 18 and Figure 19:
Figure 17 - WLTP Class 1 cycle (<22kW/ton)
Figure 18 - WLTP Class2 Cycle (22 to 34 kW/ton)
20
Figure 19 - WLTP Class 3 Cycle (>34 kW/ton)
The WLTP cycle has a much more complex profile when compared to the NEDC cycle, and this type
of driving cycle is much closer to real world driving. The 3 specific power classes separation is of high
importance, as with current economic crisis, with fossil fuel depletion and the government‟s intention
on CO2 emission reduction, more and more low power engines will appear, and as mentioned above,
the same test cycle can‟t accurately be applied to low and high power vehicles (as the acceleration
rate and maximum speed achieved are completely different). Once again should be highlighted, that
the dedication of different countries in the development of new type-approval test cycles, assures the
discontent with current test cycles and the need in having reliable and realistic type-approval method.
As was presented in Figure 16, a similar comparison is now made between the 3 WLTP cycles and
the same reference real-world measured driving cycle:
Figure 20 - Comparison of WLTP test-cycles and a reference real-world measurement
0%
5%
10%
15%
20%
25%
30%
35%
40%
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
% T
ime
VSP Mode
WLTP - VSP Mode distribution Comparison
class1
class2
class3
Measured
NEDC
21
Even with the new developed, WLTP cycles the VSP distribution of test-cycle and real-world-cycle are
not overlapping, but as the “Measured” trip is just a typical mixed driving cycle, not exactly
representative of all the drivers in the World or all the combinations of vehicles and roads, the
Measured curve is not very different from WLTP class 3 test-cycle (as mentioned before, WLTP class
1 and 2 are for very low power vehicles, of which the Measured typical cycle is not representative at
all, thus resulting in VSP mode curves completely different from class 3) [21]. Steps are made in the
correct direction of achieving a more representative type-approval test-cycle for vehicles certification.
One of the main goals of this work is to provide a methodology that would allow to generate test-cycle
with very good representativeness of specific real-world cases.
In Figure 16 and Figure 20 the “Measured” VSP mode distribution is of the average of 50 Portuguese
drivers, which is better described in 6.2.
2.3. How measurements are made in laboratory
The certification of every new vehicle is made by performing a type-approval standard test driving
cycle, which is performed on a chassis dynamometer. Exhaust gas flow is usually measured using a
constant volume sampler (CVS), either continuously or using a sample storage bag or a filter [6].
Figure 21 - Chassis Dynamometer with Constant Volume Sampler (CVS) gas flow measurement system [26]
22
The vehicle performs the speed cycle defined by the test cycle and the exhaust emissions are
collected in bags. Vehicle is approved if the total mass of pollutants inside the bags is lower than a
certain imposed limit (according to bags size, temperature, etc). Pollutants are analyzed by
conventional methods: infrared absorption NDIR (non-dispersive infra-red) for carbon mono- and
dioxide and methane CH4, flame ionizing FID (flame ionization detector) for total hydrocarbons HC
(heated for diesel), luminescent chemistry for nitrogen oxides NOx. [14] Fuel consumption is
calculated by carbon equivalence. Also the concentrations of NOx, total hydrocarbon (THC), carbon
monoxide (CO), and carbon dioxide (CO2) in the diluted exhaust gas can be measured using an
emission analyzer and then combined with tunnel flow rates to obtain mass emissions per second
[g/s] [14].
As the performed test cycle is not exactly representative of real world driving conditions, the
measured emissions are, as well, non-representative of real world emissions. Also, as the total
emitted mass is measured from the bags, it is impossible to relate higher emission peaks with specific
behavior in driving cycle, gear change, accelerations etc. In this case, a gas emission analyzer can
help, but its delay, which is variable with engine speed (exhaust gas flow changes with engine speed,
thus for same pipe diameter, the gas speed vary, and therefore the time it takes to travel from exhaust
manifold to tailpipe exit fluctuates), is not very helpful either.
Preferable is to have a type-approval test driving cycle that makes the car behave as it will on road,
thus leading to very similar to reality emission rates and fuel consumption.
2.4. Parameters that influence consumption and emissions
The amount of real-world variables that have an influence on the fuel consumption and pollutant
emission rates is countless. Following are some of the most noticeable of them:
Gearbox and Engine Management
Gears shifting usually goes along with higher power requirements, as before shifting there is an
acceleration to give the vehicle enough momentum through the shifting process, and also just after
the gear shift more power is required to increase the engine rotational speed to the comfortable for
passenger and driving safety. If we would measure with high accuracy the pollutant emission rate
during a gear shift (upwards) we would see an increase and a pick just before the change, a drop
during the change and a pick and stabilization after the gear is changed. Of course gear changing is
required on road, and picking constantly the adequate gear for the situation on overall reduces fuel
consumption and pollutant emission (even though there could be frequent gear changes, if the engine
passes to a working state with better conditions, the reduced fuel consumption of the adequate
engine state, when averaged, will compensate for the picks coming from gear changes). Gearbox
management is the parameter with the biggest variation from driver to driver.
23
Aggressive fast drivers with adequate gear shifting can have similar fuel consumption to slower, more
calm, drivers but using the engine in the rotation speed and load combinations with very poor
efficiency. As exemplified in the Figure 22:
Figure 22 – Diesel Engine Fuel Consumption Map [g/kWh] and power curves [horsepower] [22]
The minimum fuel consumption is around the typical engine maximum torque rotation speeds (around
2000rpm). The minimum specific consumption is at the same engine speed, but on high loads. If we
take a example of a hill climbing, where for a steady speed movement from the engine is required 50
horsepower, if the driver uses a low gear (driver might prefer it for higher responsiveness), for around
3500 rpm, the specific fuel consumption is 250 [g/kWh]. If the driver was to change to a higher gear
(some drivers might not like this while climbing, as the engine rumble is loud), going to 2000 rpm, the
specific fuel consumption would decrease to 200 [g/kWh], but for 2000rpm and 50HP, the break mean
engine pressure is almost twice as high as in the previous situation, hence a trade between fuel
economy and engine use (deterioration) is made (as well as NOx concentration increases with higher
pressures).
Engine management varies greatly throughout the large range of internal combustion engines
available. Some sacrifice the conditions inside the combustion cylinder, leading to increased fuel
consumption and pollutant emissions, in favor of high power demand from the driver. Other engine
management systems put as first priority the correct fuel/air ratio, temperature etc. which leads to
lower consumption and emissions. This parameter is of high impact, hopefully an appropriate test
cycle would catch all the states of engine management (certainly not the case of NEDC cycle, as it
24
maintains constant speed for the most of the duration, and the transitions between speed levels is
made gradually, see Figure 6).
Aging and Maintenance
Vehicle aging usually increases the fuel consumption of the vehicle and could increase emissions of
some pollutants, while maintain the emissions of others. By engine aging, it is mainly considered
engine mileage, as the time passing itself does not modify the working properties of the engine, if the
motor oil is substituted accordingly. Aging and maintenance are linked, a well maintained engine with
certain mileage, could have same mechanical deterioration as a poorly maintained engine with half
the mileage. As mechanical parts get loose, the fuel consumption increases, mainly because of
pressure loss in the cylinder (through the piston/cylinder wall gap). Catalytic converter, filters also
suffer degradation in their nominal performance, hence the emission rate increases from engine
aging, and from less filtering/conversion.
In the Figure 23 can be analyzed the fuel consumption (represented as miles per gallon, MPG, it
should be noticed that less MPG, less fuel efficient, thus more fuel consumed, more l/100km). Initially,
vehicles show an increase in fuel economy, which results from reduction in the thickness of lubricants
and the appropriate components dilatations and adjustments that are considered and predicted during
design process of the vehicle. After around 7 years (or corresponding mileage), vehicle fuel economy
goes below the value presented as brand new. Note that engine and vehicle performance change
with aging differently from case to case, although the generic trend is the one presented in Figure 23.
Fuel Type
Fuel type is also important in the emissions and consumption of the engine. In the design stage, every
engine is conceived with a specific fuel type in mind, any deviation will introduce deviations from the
stated technical specifications. Poor quality fuel (especially some decades ago) contained sulfates,
Figure 23 - Fuel Efficiency with vehicle age [28]
25
lead, mercury, between many other toxic chemical compounds, and as can be easily understood, with
the high temperatures and pressures inside the combustion chamber, different poisonous oxides can
originate that heavily worsen the local pollution.
Cold Start
On cold start conditions, when the engine temperature is far lower compared to “steady state”
conditions, the emission of CO and HC (in case of an Gasoline engine) is much higher than normal,
and that is caused by the quenching effect (non-existent or very reduced combustion reactions close
to the cold walls) and increased thermal losses. Also in the first few minutes after starting the engine,
vehicle generates higher emissions because the emissions-control equipment has not yet reached its
optimal operating temperature (as catalytic converter and EGR if existent). While the engine is cold
(by cold is considered below 100ºC), the piston-cylinder gap is larger compared to the steady state
condition, thus motor oil is primarily consumed in this stage. [23]
Dynamics
While all the above parameters influence the fuel consumption and pollutant emission rates, a type-
approval test driving cycle cannot account for them. The main goal of test cycle is to simulate as close
as possible real world vehicle dynamics.
Average Speed
The average speed of the trip has a big impact on the emissions, as can be seen in the Figure 24 [6]:
Figure 24 - CO2 emissions evolution with average speed [6]
26
As it is very well known, CO2 emission and fuel consumption are practically linearly correlated, that is,
as in perfect conditions, combustion would only produce H2O and CO2 (see 2.1), the Figure 24 can
be interpreted as the evolution of fuel consumption versus average vehicle speed (of course with
different values on the y-axis). It can be easily observed that on higher average vehicle speed, the
fuel consumption is lower compared to typical congested urban traffic (low av. Speed), and the
explanation comes from the frequent accelerations and stops of the urban driving cycle (each break,
is an irreversible loss of kinetic energy). With increased average speed of the trip, the percentage of
breaking is lower compared to urban driving, thus the only energy required from the engine is to
overcome the mechanical, aerodynamic and friction losses, while in heavy traffic conditions and/or
typical urban cycle, the vehicle is exposed to continuous accelerations (energy required to overcome
vehicle‟s inertia) and the waiting times at traffic lights and stop signs (while the vehicle is stopped,
almost every time the engine keeps running) which cause fuel consumption without any increase in
the distance travelled.
The Figure 24 goes up to 120km/h average speed (which already are difficult to achieve even on
100% motorway conditions), but if it was tested for higher average speeds, the fuel consumption rate
would start increasing, and that is caused by the very high increase in aerodynamic resistance, which
will be mentioned and compared later in the Vehicle Specific Power parameter definition (3.3).
While the CO2 emission (or fuel consumption) is somewhat easy to predict/explain given the driving
conditions (driving cycle), when it comes to the pollutant emissions (HC, NOx, CO and PM) their rates
hugely depend on the acceleration rates, engine temperature and ambient conditions (beside
obviously engine type, and the type of exhaust recirculation or catalytic converters, if any). [6] [11]
Acceleration
The kinematic quality of the driving cycles (that is, the shape of the cycle speed curve and
consecutively its slope) is a major factor in measured emission representativeness. Standardized
cycles in Europe and the USA underestimate cold unit emissions for petrol engine and diesel vehicles
by 30 or 15% (at the maximum), respectively, and hot unit emissions by 50 and 30% at the
maximum), respectively [6] [14]. It was demonstrated that average speed is not the only parameter
accounting for recorded emission levels [11].
Contrary to some belief, the acceleration rate influence on fuel consumption is very small, that is, a
steep acceleration or accelerating slowly to the cruise speed consumes the same amount of fuel.
27
Computing the integral below the curves presented in the Figure 25, will result almost in the same
value (less than 1%), which means that there is little to no concern how we accelerate from stand still
to cruise speed (note that the author of Figure 25 mentions that the steep acceleration curve, purple
line, was not made with full throttle, and such acceleration could indeed conduct to higher
consumption as Air/Fuel ratio is getting far from desired values). It should be noted that on the Figure
25 the fuel consumption is specified per unit of time, while the author mentions that the equal fuel
consumption is for the same distance traveled [l/100km], which is the value that matters the most.
While the way driver accelerates might have a small influence on fuel consumption, it does have a
considerable impact on emissions. Also one of driver‟s behaviors that could reduce the fuel
consumption is to minimize the overall accelerations made, and that is done by reducing the number
and intensity of braking, by increasing the gap from the car ahead and thinking forward about the
traffic conditions ahead.
On steep accelerations (above 1.5m/s2), the engine is exposed to high loads, leading to higher
Fuel/Air concentrations, with pressure building up in the chamber to very high values (high load
originated from the inertia of the vehicle), all this leading to high concentrations of NOx and CO in the
exhaust gas.
Figure 25 - Fuel Consumption with acceleration rate (purple - aggressive, red - calm) [27]
28
3. Methodology
Using a mobile laboratory to analyze gas emissions in real world driving, under various conditions
(different cars, climate conditions and test drivers, to cover as many variables as possible) and over a
statistically long enough amount of time would result in an accurate representation of the fuel
consumption and the level of pollutant emissions for that specific vehicle. Although such a complex
process would take too much time and would be expensive and difficult to guarantee consistency
from test to test and from car to car. Instead it was considered and later approved the option of using
test driving cycles, on a chassis dynamometer, in laboratorial environment.
The main parameters commonly used, by researchers in the field, to create/develop a test cycle for
light vehicles certification are:
Adjust mean speed
Adjust acceleration [11]
Adjust gearshifts [11]
Adjust Vehicle Specific Power (VSP)
By adjusting the average vehicle speed, the fuel consumption can be quite representative of reality,
that is of course given the test cycle has a “normal” behavior (as matching the average speed with a
cycle can be done in various ways: high speeds for a portion of the cycle, and very low or null speeds
for the remaining portions (the average after all, is average), or just accelerating as quickly as
possible to the desired average speed and maintaining it for the entire duration of the test cycle).
Since the pollutants are far more sensible to acceleration rates, the method of creating a test cycle
based solely on the average speed, is not sufficient, still, matching the average speed parameter is
very good way of developing the test cycle.
Adjustment of the acceleration, can be done: by matching the total intensity and duration of the
accelerations, by matching the number of accelerations (instances of high slopes on the test cycle
graph) or by matching the average positive acceleration. This method of test cycle generation is
usually more accurate in respect to the pollutants emission rates (NOx, CO, HC and PM), but can be
quite offset with regards to the fuel consumption (and consecutively CO2 emissions).
The definition of Vehicle Specific Power, and its importance in the characterization of a driving
style/pattern, or just to summarize a certain trip, and also its significance on aiding the generation of
the test cycles is exposed in following subchapter 3.1.
29
3.1. Vehicle Specific Power (VSP)
In 1998, J. L. Jiménez, at the Massachusetts Institute of Technology, introduced for the first time the
idea of Vehicle Specific Power (VSP). Its usefulness was since increasingly recognized by the
majority working in the vehicle performance field.
As the name suggests, VSP is the power divided by vehicle total mass, only that it is not the
maximum power the car can developed (the one that appears on vehicle specification sheet), but it is
the instantaneous power required for the car to perform a certain speed, on a road with a certain
slope and with certain acceleration. VSP Is defined by: vehicles speed, road slope, acceleration,
aerodynamic drag and rolling resistance. VSP value represents, usually on a second-by-second
basis, the specific power that is applied to the wheels of a specific vehicle at a given time (note that
the losses from camshaft to wheels axes are not considered, that is, friction losses in gearbox,
transmission etc.). The equation used to calculate VSP is:
( ( ) ) Equation 1
where:
- vehicle specific power [kW/ton = W/kg = m2/s
3];
- instantaneous speed of the vehicle [m/s];
- acceleration of the vehicle [m/s2];
- inclination of the road [rad];
- rolling resistance term coefficient [m/s2];
- aerodynamic drag term coefficient [m-1
].
VSP being a specific parameter allows an easier comparison between different vehicles of different
weights, as the power required for similar driving conditions of different vehicles will be proportional to
their masses. The rolling resistance and aerodynamic drag coefficients vary from vehicle to vehicle,
but the values used in the formula try to estimate and cover the vast majority of typical light weight
vehicles (of course a square box as Mercedes-Benz G class, or a slick Mercedes-Benz CLA class will
differ more than a bit, but those are the extreme cases).
As can be easily observed, equal VSP values can be obtained by different ways, that is, a very
intense acceleration on low speeds can require same power applied to wheels as a steady high
speed cruise on a motorway or a steady climb on a steep hill, etc. Hopefully the fuel consumption (or
CO2 emission) is the same in all the situations with similar VSP value (and as later will be shown, it is
indeed the case), however the emissions (NOx, PM, HC and CO) are not solely dependent on the
VSP value alone, having a considerate variation for high VSP values. If a trip has high VSP value
originated from high speeds, the pollutant emission rate will differ from a route with same proportion of
high VSP values, but originated instead from steep accelerations.
30
In order to easier characterize a trip, drive cycle or driving style/pattern, VSP values are grouped in 14
classes (14 VSP modes) as shown in the Table 2:
VSP Mode
VSP value range [kW/ton]
1 VSP < -2
2 -2 ≤ VSP < 0
3 0 ≤ VSP < 1
4 1 ≤ VSP < 4
5 4 ≤ VSP < 7
6 7 ≤ VSP < 10
7 10 ≤ VSP < 13
8 13 ≤ VSP < 16
9 16 ≤ VSP < 19
10 19 ≤ VSP < 23
11 23 ≤ VSP < 28
12 28 ≤ VSP < 33
13 33 ≤ VSP < 39
14 39 ≤ VSP
Table 2 - VSP mode values [3]
The importance of rolling resistance and aerodynamic drag varies with the vehicle speed. Figure 26
presents the rolling and aerodynamic resistance from VSP formula as a function of speed:
Figure 26 - Comparison of drag and rolling friction with speed
0.0
5.0
10.0
15.0
20.0
25.0
30.0
35.0
40.0
0 50 100 150
Re
sist
ance
[kw
/to
n]
Vehicle Speed [km/h]
Aeodynamic and Rolling Resistance evolution with speed
RollingResistance[kW/ton]
AerodynamicResistance[kW/ton]
TotalResistance[kW/ton]
31
From the Figure 26, it can be seen that until about 70 km/h the rolling resistance is higher than the
aerodynamic drag, but since the later one depends on the cubic power of the vehicle speed, it
increases in a very steep manner, and at 120-130 km/h speeds (motorway speed) the total external
resistance imposed to the vehicle is composed 2/3 from aerodynamic drag and 1/3 from rolling
friction.
To get a better understanding of how VSP mode distribution is helpful, and how it can be used to
differentiate distinct driving styles, the Figure 27, Figure 28 and Figure 29 are presented:
Figure 27 - Typical Urban VSP mode Distribution
For a typical urban style driving, it is easy to observe that more than 30% of the time the vehicle is
stopped (a small slice of VSP 3 mode distribution can come from cruising state, with fuel cut-off, but
essentially it represent the vehicle being stopped, at red light or traffic jam). Positive VSP values (VSP
mode 4 and above) are sequentially decreasing in distribution share hold, which indicates the
absence of high power requirements, which are consistent with urban driving, where high speeds
usually are not practiced (of course it is possible to achieve high VSP mode values even in urban
conditions, by undertaking heavy accelerations or going over the speed limits, but that is not “typical”).
VSP modes 1 and 2 represent negative specific power (see table 2), that is, soft breaking/slowing
down for VSP mode 2 or a more intense breaking for VSP mode 1. VSP modes 10 and above are
almost non-existent, that means, that specific power above 20 [kW/ton] are rarely used. Considering
an average car with a mass of 1400 kg, for a typical driver and a typical urban driving style, the
engine power required from the engine is 1.4*20=28kW (that is, 38 horse power engine is enough for
the most of the time, more specific, 95% of the time).
32
Figure 28 – Extra Urban VSP Distribution
Comparing the previous VSP mode distribution of an urban style driving, with the Figure 28‟s
distribution of a highway trip, it is immediate that the patterns are quite different. VSP mode 3 has a
much smaller distribution share now, below 10%, and is mostly no longer caused from the stopped
vehicle condition, but instead it is generated from fuel cut-off state while on a downhill slope or the
cruising time the driver does before breaking (time between lifting the foot from gas pedal and
pressing the brake pedal). High VSP modes values are now more frequent, which is being caused by
the higher speeds practiced on the highway and the high power requirement of accelerating on high
speeds. VSP modes above 10 now are more important and frequent, and the vehicle even achieves
VSP mode 14 about 1.5% of the time (VSP mode 14 has above 39kW/ton, that is, a 75 horsepower
engine is required to achieve this values, for the same vehicle with 1.4 ton). The share of VSP mode 1
is much bigger than VSP mode 2 (both on breaking), which is also explained by the high speeds on
the motorway, which require more intense breaking.
Figure 29 - VSP mode distribution of a racing car on track
33
On the completely opposite side of Urban Driving style is the racing driving profile, where a powerful
car (a Porsche Cayman) on a racing track (Circuito do Algarve, Portimão, Portugal) drives in the way
that gets the minimum lap time possible. That means that all the time the racing car is either
accelerating on its maximum capacity, or breaking on its maximum non-lockable capability. So it is no
surprise to see that the VSP mode distribution is almost binary, that is, the car only drives in VSP
modes 1 and 14. This type of driving is characteristic to any racing car on track.
The 3 examples of VSP mode distributions presented in Figure 27, Figure 28 and Figure 29 should
make the reader able to recognize the way by which the VSP mode distribution can characterize a
driving style and how judging by the VSP mode curve one can understand in which conditions the trip
was made and how aggressive it was performed. Of course, coming from a VSP mode distribution, it
is impossible to reconstruct the original driving cycle from which it was created, but in any way it gives
a good idea of the way the trip was made. And it‟s not hard to understand (as will be explained in
detail later) that different driving cycles, on different cars, on different roads, but with similar VSP
mode distributions, will have similar, per distance traveled, emission of pollutants and fuel
consumption.
3.2. Approach on generating test cycles
In the initial approach to this work, it was thought of the creation of a single test cycle, that would be
more representative and more realistic than NEDC currently is, thus showing that improvement are
possible and can be done. Then the following question arises: “more representative of whom?
average Portugal drivers? Europe? Worldwide?”. This made an impact on the decision to create a
method, and consecutively a tool, that would allow generating test cycles based on input data that
would represent a specific population, fleet or even individual trip. Of course, for vehicle certification, it
is not recommended to use different test cycles for each car/engine, as this would not allow to make
direct comparison between different cars, but still, each country could generate their own test cycles
with the input data that represents the best their population driving behavior.
Taking into account the information, about a specific trip, that VSP methodology allows to summarize,
the process of generating test cycles based on real world data can be facilitated. A strong similarity of
the distribution curves of the VSP modes between the generated test cycle and the original real world
data (on which the generated cycle was based upon), would also suggest that they would have a
similarity in the fuel consumption and pollutant emission, per distance traveled.
Previous attempts on generating test cycles for vehicles certification where based on one of the
following: matching the desired average speed, matching the average positive acceleration, match the
number of gear changes or even empirically, by trial and error, match the fuel consumption and
pollutant emission results.
34
In order to maximize the accuracy of the generated test cycles, it was decided to use two criteria:
match the VSP mode distribution and match the average speed of the source data. The generating
method was conceived within the development of this work, and from a large-scale perspective, it
consists of a combination of several “blocks” of existing driving cycles in order to achieve the desired
criteria. The concept of “blocks” introduced here, refers to sections of existent test cycles, applicable
for vehicles certification or not, that usually have a short duration of 60 to 200 seconds.
To achieve the desired criteria, of matching VSP mode distribution and average speed, the generating
method: picks, removes and modifies, in an iterative way, pre-existent blocks, from an established
block database. After each iteration step, a comparison between current and previous state of the 2
desired criteria mentioned above (VSP distribution and average speed) is made, and only in the case
of an improvement on the deviation from the currently generated and the desired VSP mode
distribution and average speed, the method accepts the new modifications (block added, removed or
modified) and continues the generating process. The concept behind the generating method is
simple, easy to understand, yet not easy to implement, as approaching one of the 2 criteria to the
desired value, can worsen the approximation of the other criteria.
It was decided to use pre-existent blocks in the generating process, instead of a completely random
generation path, to guarantee the possibility of performing the generated test cycle on the test bench
in laboratory. A generating method, that would design (“draw”) the test cycle curve in a completely
free way, to match the 2 criteria, could occasionally generate unpredictable speed patterns (“jolts”),
that could not be possible to perform by the vehicle or by the test driver on the dynamometer.
The usage of a database with blocks for the generating process, allows the final user of this
methodology to add or remove desired/undesired blocks, in order to tailor the generated test cycle
behavior as desired. As example: if a test cycle is designed for the certification or analysis of a low
specific power electrical vehicle, the user can remove from the database the blocks that require
steeper accelerations that the vehicle cannot perform, as well as blocks with high speeds that the
vehicle is unable to achieve. This way, the generated test cycle will match as closely as possible the
desired driving behavior (VSP mode distribution and average speed) but still use only the limited
blocks database (that was edited) as to allow the specific vehicle to perform the test cycle.
This methodology, and the definition of VSP, are independent of the fuel type the vehicle uses,
whether it is gasoline, diesel, hybrid, electric or other. This means that the test cycles generated using
this methodology are suitable to be used and performed by vehicles with different propulsion types. In
the case of electric and hybrid cars, the charge of the battery during breaking can produce different
results on the fuel consumption (and pollutant emission in case of hybrid) which would depend on
initial state of charge of the battery. The test cycle will still require the same specific power input from
the engine/motor, the energy consumed in an internal combustion engine vehicle vs. hybrid
propulsion vehicle is the same, but because of the energy recovery during the breaking of the hybrid
car the final-initial fuel level will vary between the two engine types. For example a hybrid plug-in car,
would not consume fossil fuel during the depletion time (charge-depletion, only uses energy from
35
batteries charged at home) and start the internal combustion engine when the state of charge of the
batteries achieve a certain specified level (charge-sustaining state). Local pollutant emissions from 2
consecutive attempts at performing same test cycle with the same car will generate different results in
the consumed fossil fuel as the initial state of charge differs.
If the test cycle generated using this methodology provides a bigger than expected deviation of VSP
mode distribution, adding new blocks to the database, that are distinctive from the currently existing
ones, could solve this problem. This provides the user with the ability to customize the database used
for test cycles generation, and contribute to the generation of better and more representing cycles, as
well as guarantee that the blocks that appear in the new test cycle are the ones that suit the user.
The tool created to use this methodology to generate test cycles, is presented in next chapter. The
Test Cycle Generating Tool (TCGT) allows generation of test cycles with durations of around half an
hour that are representative, of input driving data. The input data can have duration of many hours or
even in the case of big vehicles fleets, years of driving data.
Figure 30 - Simplistic representation of methodology algorithm
Read and analyze the input data
•Read the inpud data fle of real world driving, and pick the corresponding "blocks" database to start the generation process.
Pick, Drop, Modificate and
Substitute "Blocks"
•From the available "blocks" in the database, the created methodology picks iteratively blocks one by one, modifies them (increase duration or reduce amplitude) verifies if current test cycle improves, and in case it does, repeats the process until the stopping criteria is met. On each step of iteration, beside picking new blocks and their modifications, it also tests if dropping or substituting previously picked blocks improves the test cycle.
Compile the final Test Cycle •From all the blocks that were picked so
far, compile the final test cycle that represents as close as possible the input data from average speed and VSP mode distribution points of view.
36
4. Test Cycle Generating Tool (TCGT)
The tool that allows the generation of test cycles, using the method presented in the previous chapter,
was programmed in MatLab 2012 software. For the ease of use, a Graphical User Interface (GUI)
was developed for the tool, so the user can intuitively and quickly obtain the desired result.
4.1. User Interface and main functionalities
Next follows a description of the program‟s GUI and afterwards, in 4.2, a description of how the
programming was made, and all the implications.
Consulting the screen capture in Figure 31 of the GUI of the program (further called TCGT: Test
Cycle Generating Tool), the marked regions are as follows:
1) Buttons that allow the manipulation of the graphics (plots) areas, specifically they allow to
pan, zoom in, zoom out and pick a point on the line to see its coordinates. This allows for a
more in depth analysis of specific regions of the plots.
3
1 6 2
4
5 7
8 9
10
11
12
13
15
14
Figure 31 - TCGT Graphical User Interface
37
2) A drop down menu that allows the user to choose the “weighting type” of the VSP mode
approximation, that is, allows the user to choose what region of VSP modes will be better
approached. There are 9 types to choose from, and they are as follows:
a. Uniform – in this weighting vector, equal importance is allocated to all 14 VSP modes,
this option will generally (but not always) result in the minimal VSP mode distribution
deviation between the generated test cycle and the original data. This vector is a
good first pick to observe the type of test cycle generated, and then re-calculate with
different weight vector from drop-menu, to better match the desired criteria of the
TCGT user.
b. Low VSP – a higher importance is given to match the low VSP modes (below 5), this
type is suited for urban style driving cycles, where the low VSP modes have much
higher distribution share.
c. Low-medium VSP – more importance given to VSP modes up to 10. Could be
indicated for a slightly aggressive urban trip or calm extra urban driving profiles.
d. Medium VSP – promotes a better approximation of the medium VSP modes (from 4
to 10). Indicated for very long steady medium to high speed trips.
e. Med-High VSP – more importance is given to the VSP modes above 4. Suits better
the typical highway driving style, or a very aggressive urban trip.
f. High VSP – promotes a better match on the high VSP modes (above 10). Useful for
driving profiles with very high speed, or with high accelerations.
g. Highway – tailored to suit better highway trips data. Tries to minimize the VSP mode
distribution deviation for the modes that are typically frequently used on highways
trips (see Figure 28). Similar to “med-high” option.
h. Urban – suits better the urban style trips (see Figure 27). Similar to “Low-Med VSP”
option.
i. Mixed – this weighting vector is probably the best choice after the “Uniform” one (after
first choice), as it tries to give more importance to higher VSP modes as their specific
power classes (see table 2, VSP mode classes) increases. This type of approach will
usually result in a VSP mode distribution of the generated cycle that minimizes the
energy per kilometer differences between the generated and given data.
Figure 32 – Dropdown Menu - Fine Tuning of Approximation Criteria (weight vector)
38
3) Drop down menu that allows the user to define the maximum length allowed for the test cycle.
The values available are: 1500, 1800, 2100, 2400, 3000, 3600, 4800 seconds and unlimited.
The generated test cycle might have a slightly longer duration that the limited one (no more
than 5%), which is related to the tolerance that was given in the program code that allows the
slight overshoot in favor for a better criteria approximation. In general a longer test cycle
duration allowed, the better will be the correspondence of the generated cycle with the given
data, but on certain occasions the approximation is already very good even on short durations
(2100 seconds), so even increasing the allowed total duration (4800s or unlimited), the
generated test cycle will still be short, as it already matches very well the desired criteria.
Using 1500 and 1800 second durations might result in some very inaccurate generated test
cycles with regard to the given data, in such cases it‟s recommended to try different
combinations of duration and “weighting type” (point 2)) and in case of discontent with
achieved result, increase the allowed test cycle duration and repeat.
4) The “start generation” button, for the case where a VSP mode distribution and the average
speed are available. That is, a 1 column, 15 rows excel spreadsheet where in the first row is
the average speed, in km/h, (of the given data of the trip, driving cycle, driving style, or simply
the desired average speed) and from row 2 to 15 are the distribution share of the 14 VSP
modes, the sum of which should be equal to 1. Note that increasingly more varied and
cheaper equipment (data loggers) become available to plug into the OBD port of the car, and
allow to compute and store various data points of the trip, and some compute and store
directly the VSP values or VSP mode distribution, which can be easily introduced in the excel
file with the specified format, and afterwards allow the generation of a test cycle that
simulates this real world data. Note that computation time does vary, and can take from 15
second up to 10 minutes.
5) The “start generation” button, for the case where a driving cycle is available (with or without
altitude data). That is, a 1 (or 2) column per x rows excel spreadsheet is needed as “input
given data”, where x is the duration in seconds of the available data (the excel data has
second by second data, chronologically in row by row order). When altitude data is absent,
the excel file has only 1 column, which has the instantaneous speed, in km/h, for each
corresponding second. When altitude data is available, is provided in the second column, in
meters, in-line with the speed data. With the given driving cycle (and topography if altitude is
Figure 33 - Dropdown menu - Test Cycle duration limit
39
available) the TCGT generates the test cycle that has a similar VSP mode distribution and
average speed. Note that computation time does vary, and can take form 15 second up to 10
minutes.
6) Generated Test Cycle plot. On y axis is the speed in km/h, and on x-axis is the time in
seconds. Here the user can analyze the behavior of the generated test cycle, observe if it has
some inappropriate behavior and repeat the generation process with slightly different
parameters (drop-down menus in 2) and 3)). This plot is blank until the first generation, and
will update on each new generated cycle (as the user presses the buttons in 4) and 5)). Each
test cycle generation time (computation time) can vary from 15 seconds, up to full 5 minutes.
7) Comparison between the VSP mode distribution of the generated test cycle, and the given
VSP mode distribution or the VSP mode distribution of the given driving cycle (depends if
button 4) or 5) was pressed for test cycle generation). In blue is the generated test cycle and
in red is the VSP mode distribution of the input data. Here the user can visually compare how
close the two VSP profiles are one to another, and have a perception of the success of the
generated test cycle on representing the input real world data. A close match in VSP mode
distribution will generally translate into a close match on fuel consumption and pollutants
emission (analyzed in more depth in following chapter). The VSP mode distribution is one of
the two criteria for test cycle generation, and the one with highest impact on the quality of the
obtained cycle.
8) Speed distribution pie chart of the generated test cycle. Here the user can consult how much
time (percentage of total time) was spent in each specific speed range, 0-50, 50-90, 90-120
and over 120 km/h. This can help determining if the generated test has resemblance to a
highway, urban, mixed or other configuration (of course this can also be deduced from VSP
mode distribution curve as analyzed in previous chapters).
9) Speed distribution pie chart of the given driving cycle (not available for the button “4)”, where
only average speed and VSP mode distribution is given as input data, hence impossible to
calculate or deduce the original cycle speed distribution). This contains the same information
as “8)”, but now for the given, real world, input data. A comparison between the two pie charts
can be made, which also hints to the similarities of generated and given cycles. As mentioned
previously, same VSP values (thus same VSP modes) can be achieved by different means,
so even with almost perfect overlap of the VSP distribution curves in “7)”, the speed
distribution can differ, or vice-versa (same VSP values could have been achieved for different
speeds, with appropriate accelerations or road grades).
10) In this area of the GUI the user can consult and compare the average speed of the input data
and of the generated cycle. Average speed is one of the two criteria for the test cycle
generation method, and usually it is well matched (as long as the input data is within usual
values range). The deviation between the two values is presented in the colored background
(green background if deviation is inferior to 5%, yellow if it‟s within 5 to 15% range and red if
deviation is greater than 15%, this is common for the rest of the GUI where colored
background makes appearance). The other parameter that appears here is “phi”, which is
40
better described later. As of now it can be said that “phi” can have an integer value between 1
and 9, and it mainly helps the user to choose the correct “weight vector” from dropdown menu
(2) to achieve the best result (although occasionally, another than the suggested weighting
vector can get better VSP mode approach).
11) In this section the Average Specific Power is presented, in kW/ton, of the given (input) trip
and of the generated test cycle. Using the formula that defines VSP (Equation 1), its value
can be calculated for every second of the input or generated cycle, and then averaged out to
determine the overall watt per kilogram used during that trip. In the case of input data being
VSP mode distribution, a regression is made using the table that specifies the VSP mode
ranges, and then integrated for the 14 modes. Although, this later method of calculating
average specific power is less accurate, as when for example, we make a regression from
VSP mode 10, VSP value that originated it could be any from 19 to 23 kW/ton, even by
picking the medium value of the range of the specific VSP mode, the error will always be
greater compared to average specific power obtained from a speed cycle.
12) Energy per kilometer – in this section, the user of TCGT is presented with the information with
regard to the energy used, per unit distance, for that specific trip. These values are directly
calculated from average specific power, using also the average speed. This value gives to the
user a better understanding of how close fuel consumption is expected to be. Once again, in
the case the test cycle was generated using the VSP mode distribution, the energy per
distance, kWh/km, is expected to be not very accurate for the given data.
13) This section gives the percentage of time of the cycle where the VSP values are negative
(sum of VSP mode 1 and 2), which corresponds to breaking. This value is another way of
characterizing a trip, and gives an idea of how aggressive is the driving/test cycle and in
which conditions it was performed.
14) Time stopped – shows the percentage of time the car was stopped. Always calculated for the
generated cycle, and only calculated for the input data in the case that was a driving cycle
(not the VSP mode distribution). This percentage represents the time during which VSP value
is equal to zero (which is included into the VSP mode 3). In order to achieve a similar average
speed as the input data, using the “blocks” available for test generation, the percentage of
time stopped might occasionally deviate significantly.
15) In this bottom right section of the GUI of the TCGT the user can consult what is the maximum
specific power during the generated test cycle (this one defines what specific power must
possess the tested vehicle to be able to fully perform the test cycle), shown the duration of
the input driving cycle (not available for VSP mode distribution input) and also has four
additional buttons whose functions are as follows:
a. Plot VSP distributions – plots second by second curve of Vehicle Specific Power
value (kW/ton vs. time (seconds)).
b. Plot Test Cycle – plots the speed vs. time graphic of the generated test cycle (the one
that appears initially after generating process).
41
c. Plot acceleration – plots second by second accelerations during the generated test
cycle, to see the higher demand parts, as well as assure the user there are no
abnormal picks occurring during the cycle (m/s2 vs. time).
d. +info – opens a dialog box where the name of each “block” used to build the test
cycle appears (consult Annex for further information on all the blocks used in TCGT).
This can help the user to decide if he wants some blocks removed from the available
database (if they appear too often or are difficult to perform on bench).
4.2. TCGT algorithm and resources
In this subchapter, the operation of the Test Cycle Generating Tool (TCGT) is described in depth but
without going on the detailed code level, describing how it works and which steps it follows to achieve
the desired results. The programming language used is MatLab (MatLab language, initially based on
C language), which is a common used programming, simulation and analysis tool for engineers. The
TCGT therefore has to be used within MatLab workspace (tested on 2012b or newer), and requires all
the “.m”, “.xls” and “.xlsx” files that come with the executable “TCGT.m” on the Data Disk in annex.
The active folder of MatLab should be set accordingly to data files location. Microsoft Office 2007 or
newer is also required for the MatLab to be able to read the “.xlsx” files.
4.2.1. How TCGT works
When the user runs the program, the first window that is presented should be the one that appears in
Figure 31, with exception that all the fields (plots and data on right column) are empty. The “Cycle
Duration” and “Fine Tuning of Approximation Criteria” (VSP mode distribution weighting vector)
dropdown menus are on their defaults value, and user can choose the desired options for the
following test generation. Next, the user has two options for the test generation: either the test cycle
Figure 34 - "+info" window with the list of blocks used to generate the test cycle
42
will be based on a given VSP mode distribution and a mean speed or the input data will be a real
world driving cycle, with or without altitude data.
After picking the option that suits the user (button “4)” or “5)” on TCGT GUI, see Figure 31), a prompt
window appears asking to pick the input data for test generation process to start (pick the data of the
original cycle).
For the correct reading of input data, the two different input files, corresponding to the different
buttons “4)” and “5)” on TCGT GUI (see Figure 31), should be formatted as follows:
VSP mode – In Microsoft Excel 2003 or earlier “.xls” file, where on the first spreadsheet. The
cell A1 contains the average speed of the given trip, in kilometers per hour, while the 14 cells,
from A2 to A15, contain the share of the distribution of the VSP mode, such that the sum of
these cells is equal to one. The content of each cell is given by: the time spent in that VSP
mode divided by the total time of the trip.
Driving cycle – In Microsoft Excel 2007 or later “.xlsx” file, where on the first spreadsheet. The
first column (column A) is composed of the second by second speed data (in kilometers per
hour). The second column (column B) might either be empty, in which case the TCGT
considers the altitude constant and as if cycle was performed on flat horizontal surface, or it
may contain the altitude data, second by second, given in meters. In case the input data has
information about the altitude, the column A and B must have same length (equal amount of
seconds of data) and both should be refer to the same time stamp.
In case of doubt, consult the existing VSP mode distributions and Driving Cycles available in the
corresponding folders within the data files of the TCGT (found on Data Disk in annex).
Before entering into the step by step modeling of the main “TCGT.m” code file, it is convenient to
mention and explain the created external functions that are repeatedly called during its operation.
acel() - calculates the acceleration based on past and future speed data. Contains if()
conditions for the start and the end of the speed data (where previous and future data is not
available).
o Input: speed cycle (vertical vector containing speed in km/h, for every second of the
trip).
o Output: Vector with acceleration during every second of the trip.
cic2pow() - uses the formula of VSP to calculate its value on every second, and then
computes the average of only the positive values of VSP (VSP > 0).
o Input: vector with speed data (1 second resolution).
o Output: average positive specific power of the input speed cycle.
generate1() – Is called when user presses the “generate using VSP” button (generate a test
cycle based on a given VSP mode distribution). Opens an explorer window asking the user to
pick the input file containing the VSP mode distribution, with the according formatting
43
presented above. From here, “generate1()” and “generate2()” functions are equal.
Given the average speed and VSP mode distribution, test cycle generating process
commences. If average speed of the input real world driving data is below 30 km/h, “generate
()” will only use blocks with average speed of less than 45km/h (on 3rd
worksheet of
“blocks.xlsx”), else if average speed of input data is above 65, only “blocks” with higher than
45km/h average speed will be used (on 2nd
worksheet of “blocks.xlsx”), while if the average
speed of the input data is within 30 to 65 km/h range, “generate()” will use all the blocks
available in the database file (1st worksheet of the “blocks.xlsx” file). This pre-determined pool
of available blocks speeds up the computing time, as for typical highway input trips, with high
average speeds, the blocks with low average speed are inadequate to compose the test
cycle, thus saving many iterations where they would appear (or if the test cycle is to represent
a calm urban driving, blocks with high average speeds are also inappropriate to appear, the
“blocks” pool being reduced to only blocks with lower average speeds).
After reading and storing in a matrix the blocks that will be used in the generation process,
the scaling parameters of the blocks are defined. The blocks, beside their original shape read
from the database file, are modified:
o reducing the speed by multiplying by 0.9:
Figure 35 - Block altitude reduction
o re-sampling the duration of the block to 1.05, 1.1 or 1.2 of original duration:
Figure 36 - Block duration re-sampling ("stretching")
This block adjustment is designed to add variety to the generation tool, as to allow more
options for combinations and achieve faster the desired VSP mode distribution. Block
44
amplification (multiply by a factor greater than 1) or reduction in duration by re-sampling, is
not considered, as this would increase the acceleration and could surpass the manageable
performance of the vehicle or of the test driver. Global variables like: cycle duration, “weight
vector”, the cycle itself among other local and global variables are defined (global variables,
are variables accessible and common for all functions).
The iteration stopping criteria is defined, which is the first of the following to occur: pre-
determined number of total iterations is achieved or the VSP mode distribution difference is
sufficiently low (according to imposed cycle duration, for longer test cycles, the stopping
criteria of VSP mode deviation is more meticulous). The iteration is performed in a “while()”
function, with stopping criteria as conditions. A “for()” cycle is performed going through all the
blocks available for generation process. Inside is another “for()” cycle that goes through the 5
iterations of the current block (the original block and the 4 modifications, multiplication by 0.9,
and resample to 1.05, 1.1 and 1.2 of the duration). The current block is temporarily added to
the block list, and the average speed of the so far picked blocks is calculated. Depending on
the deviation of the average speed of the temporary current test cycle from the average
speed of the input data, the blocks are separated by empty (speed = 0km/h) spaces or joined
together, as to attempt matching the target average speed. The average speed adjustment is
made with auxiliary variables and “for()” cycles. The new cycle using the temporary added
block is created, and the condition presented in Equation 2 is tested:
( ( ( ) ) )
( ( ( ) ) )
Equation 2
, where “abs()” command is a MatLab function that returns the absolute value of input, and
“mean()” is the MatLab function that returns the average of the input data. If the above
inequation is satisfied, the temporary picked block from the current “for()” step will be added
to the “picked blocks pool”, a cell array that contains the current picked blocks from database.
Within similar “for()” cycles, now going through the “picked blocks pool”, the generate() “.m”
file will check if by changing one of the picked blocks by another from the database the
Equation 2 is satisfied, or in another cycle, if by removing one of the previously picked blocks,
the temporary cycle has a closer VSP mode distribution to the desired one. In other words,
there are various “for()” cycles performed, in the first one the blocks are picked from
database, and only the ones that “improve” (by improve I mean a better satisfaction of
Equation 2) the cycle are picked into the “picked blocks pool”, in the second cycle, attempt is
made to substitute some of the picked blocks by another from database, and substitutions are
only made if they actually “improve” the cycle, while in the third cycle the “generate()” function
45
attempts to remove some of the picked blocks, and only removes them if the new cycle is
improved. All the cycles are repeated iteratively until the stopping criteria is met or until a
considerable amount of cycles where already performed (to not lead in an endless loop). After
picking the best blocks from the database to generate the test cycle with desired average
speed and VSP mode distribution, another cycle is performed, in this case only removing
blocks, in order to achieve the “cycle duration” that the user picked in the dropdown menu of
the TCGT graphical user interface window. Finally with the remaining “picked blocks pool” the
test cycle is constructed, and is assigned to a global variable that is later used by “menu()”
executable (“.m” file), also known as TCGT. For a full understanding of the function, see the
MatLab code in annex or annex files.
o Input: Average Speed and VSP mode distribution.
o Computes and Writes: test cycle with as close as possible, average speed and VSP
mode distribution, with regard to the input data.
generate2() – Is called when user presses the “generate using Driving cycle” button. Is
executed after the “read_speed_altitude_create_vsp2()” completed, thus having already
generated the VSP mode distribution of the input Driving Cycle. From here, “generate1()”
and “generate2()” functions are equal. See “generate1()” description from here.
o Input: Driving Cycle input data (w/ or without Altitude data), with formatting presented
above.
o Computes and Writes: test cycle with as close as possible, average speed and VSP
mode distribution, with regard to the input data.
read_speed_altitude_create_vsp() – in all equal to “read_speed_altitude_create_vsp2()”
except it writes the output data to an excel file with same name as the input file, but with an
extension to the end of the name. Writes data with an header, for ease of read.
o Input: Driving Cycle input data (w/ or without Altitude data), with formatting presented
above.
o Computes and Writes: for every second of the trip: Speed (km/h), Altitude(m),
Speed (m/s), Acceleration (m/s2), Covered distance so far (m), road slope (grade,
sin(φ)), VSP value (kW/ton) and the VSP mode (1 to 14). While on the second
worksheet of the output file it writes in the first cell the Average Speed (km/h) and the
VSP mode distribution (14x1 vector with time share percentage).
read_speed_altitude_create_vsp2() – When this function is called, first it opens a window
asking for the “.xlsx” input file with the formatting settings presented above (Driving Cycle w/
Altitude). If input file does not respect the expected formatting, the program shows an error,
and a file with correct formatting must be selected. If the input file does not have altitude data,
the altitude is considered as constant. Afterwards the speed conversion to m/s is made,
second by second acceleration in m/s2 is calculated, and the cumulative covered distance for
every second of the duration of the cycle is computed (needless to say that all is supported by
the existent “for()” and “if()” functions of Matlab). Next, the road slope is computed, which is
46
required in the VSP formula (Equation 1), where sin(φ) is the fraction of altitude variation by
the covered distance. For every second of the duration of input driving cycle, the road slope
is calculated using the 50 meters before and 50 meters ahead altitude variation, and divided
by 100 meters span, gives the sin(φ) value. It should be noticed that shorter distance
segments used to compute road grade are more prompt to errors, while a segment that is too
long will result in a misleading low variation of road slope. The user can change the segment
duration, simply by changing the “x=100;” line in the code of this function to any other value).
Due to the requirements, the code implementation of this part was made in a double “for()”
cycle using the “if()” conditions as well as auxiliary elements and counters. Using the VSP
formula (Equation 1), vehicle specific power in kW/ton is calculated for every second of the
trip duration. Using the classification in Table 2, VSP mode is calculated for every second of
the duration of the input cycle, and later in a “if()/elseif()” condition the VSP modes are
grouped in the 14 groups and the VSP mode distribution is finally computed. The average
speed of cycle, and the percentage of time VSP values are negative is also determined (used
later in TCGT main function, “menu()”).
o Input: Driving Cycle input data (w/ or without Altitude data), with formatting presented
above.
o Computes: Average speed (km/h), VSP mode Distribution (14x1 vector with time
share percentage) and the % of total duration in negative VSP value.
read_speed_only_create_vsp() - in all equal to “read_speed_altitude_create_vsp2()”
except does not account for the altitude data, even if it is available in the input data file,
considering as the input speed cycle is performed on a leveled topography. This function was
mainly used while generating the five test cycle for validation using the ADVISOR software,
presented later in chapter 0.
o Input: Driving Cycle input data (w/ or without Altitude data, which is regardless
ignored), with formatting presented above.
o Computes: Average speed (km/h), VSP mode Distribution (14x1 vector with time
share percentage) and the % of total duration in negative VSP value.
Figure 37 - Road slope calculation
47
TCGT or menu() – the main code file, the one that is executed, computes and calls the
external functions to achieve the final goal: generate a representative test cycle and present
to user its properties in a friendly and intuitive way. The working of TCGT is explained in more
detail below this section.
vsp() – using same process as vspv(), calculates the VSP value of the speed cycle, and later
uses a for() cycle to distribute the VSP values and determine the share of each of the 14 vsp
modes (according to Table 2).
o Input: speed cycle (one dimensional vector with speed data of every second of the
trip).
o Output: vector with size 14x1, that has the distribution share of the VSP modes of
the input cycle.
vsp2pow() – function uses the data from: Table 2, to compute the mean power of every VSP
mode, and multiply by its corresponding percentage of time, to compute finally the average
power of the given VSP mode distribution. Omits the VSP mode 1 and 2, thus only averaging
the positive power (similar to cic2pow()).
o Input: vector with size 14x1, that has the distribution share of the VSP modes.
o Output: average positive specific power of the input speed cycle
vspm() - function is mainly based on Equation 1, and has some if() conditions for exceptions
at beginning and ending of data set, to correctly calculate the acceleration. Converts units
accordingly. The VSP values obtained for every second are classified in the 14 VSP modes
as shown in Table 2.
o Input: speed vector (second by second).
o Output: Vector with VSP mode (1-14) at every second.
vspv() – similar to vspm(), only does not sort VSP values in 14 modes. This function is mainly
based on Equation 1, and has some if() conditions for exceptions at beginning and ending of
data set, to correctly calculate the acceleration. Converts units accordingly. Acceleration on
each second is based on past, current and future speed.
o Input: speed vector (second by second).
o Output: Vector with value of VSP at every second.
Without entering too deep into the specifics of the code, the “menu()” function (or main TCGT
function) is running specific commands and calls specific functions according to the user input.
Dropdown menus choices are always read before generating process starts. The definition of the
“weight vector” from the “Fine tuning of the approximation criteria” dropdown menu is given in the
Table 3:
48
VSP Mode
Weight Vector 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Uniform 0.071 0.071 0.071 0.071 0.071 0.071 0.071 0.071 0.071 0.071 0.071 0.071 0.071 0.071
Low-VSP 0.100 0.050 0.300 0.100 0.075 0.075 0.075 0.050 0.045 0.030 0.040 0.030 0.020 0.010
Low-Med VSP 0.090 0.090 0.090 0.090 0.090 0.090 0.090 0.090 0.090 0.090 0.090 0.004 0.004 0.004
Med VSP 0.011 0.011 0.011 0.011 0.011 0.011 0.011 0.222 0.222 0.222 0.222 0.011 0.011 0.011
Med-High VSP 0.007 0.007 0.007 0.007 0.007 0.007 0.007 0.136 0.136 0.136 0.136 0.136 0.136 0.136
High VSP 0.014 0.014 0.014 0.014 0.014 0.014 0.014 0.014 0.014 0.014 0.014 0.282 0.282 0.282
Highway 0.025 0.025 0.025 0.012 0.027 0.042 0.057 0.072 0.086 0.104 0.126 0.153 0.099 0.148
Urban 0.004 0.002 0.002 0.010 0.022 0.035 0.047 0.059 0.072 0.086 0.104 0.127 0.184 0.245
Mixed 0.018 0.004 0.009 0.022 0.024 0.038 0.051 0.064 0.078 0.093 0.113 0.138 0.160 0.187
Table 3 - Weight Vector description for fine tuning of the VSP mode approach
The weighting vector is used in “Equation 2”, to define how the VSP mode distribution comparison
between the original data and generated test cycle is made. As can be seen, the uniform vector gives
same importance to all 14 VSP modes, while the weighting vectors where defined in such a way that
the VSP mode approach would be better in specific parts that are more crucial for a better
approximation for that type of cycle. For example, the “High-VSP” vector gives much more importance
to the deviation between generated and original VSP modes 12, 13 and 14, while the remaining VSP
modes could deviate more (in other words weighting vector specifies which VSP modes approaches
can be “sacrificed” in order to achieve better approximations on other specific VSP modes).
In order to be able to classify a driving cycle by its overall aspect (urban, mixed or highway; calm,
normal or aggressive), a parameter “phi” is introduced, and it is defined as follows in the Figure 38:
Figure 38 - Definition of "phi" parameter for driving cycle classification
49
With the definition from the Figure 38, cycle with phi parameter 1, 2 or 3 is a cycle with calm behavior
(less high vehicle specific power requirements), and cycles with “phi” equal to 3,6 or 9 are for highway
or exclusively extra urban environments. This classification comes in hand in chapter 5.4.
Taking a look at the GUI of TCGT main Window (see 4.1), if the user pushes the button “4)”, the
“generate1()” function is called. If user pushed button “5)”, the “read_speed_altitude_create_vsp2()”
is called, after which the “generate2()” function is executed. In any case, the result will be a generated
test cycle. Next steps of the “menu()” executable consist of using the original data and the just
generated test cycle to construct the plots and the data on the right column of TCGT main window
(Figure 31).
Also, on the end of the generation process, the TCGT will export some of the generated data and add
a new column on three worksheets of the “output.xlsx” file (or create a new file if it doesn‟t already
exist). In the added column from worksheet 1 it has the following information:
1) Original cycle name
2) Desired Maximum Duration
3) Duration of the generated test cycle
4) Fine Tuning “weight vector”
5) Deviation of Average Specific Power (in %)
6) Deviation of Energy per kilometer (in %)
7) Sum of deviations between the 14 VSP modes of original and generated cycle (in %)
8) “Phi” parameter of the original cycle
9) “Phi” parameter of the generated cycle
, on the worksheet 2, the added column will contain second by second data of the generated test
cycle speed (in km/h), which is, the test cycle itself. On the third worksheet is present the average
speed [km/h] and energy per km [kWh/ton/km] for the generated test cycle as well as for the input
data.
For any doubt concerning the operation of the program, see the program code in Annex, or the “.m”
files attached Data Disc or online on the specified address in Annex.
4.2.2. Which blocks make the database
The source for the test cycle generating method, are the blocks that compose the database (blocks
library). These were carefully chosen, to provide sufficient variety while not creating a too extended
database that would heavily increase the computation time of the generation process. Variety should
be provided as in average speed, aggressiveness (given by the average absolute value of the
acceleration) and short durations.
The Table 4 compacts all the main information about the 42 blocks (driving cycles extracts):
50
Block Number
# Block Name
Duration [seconds]
Average Speed [km/h]
Maximum Speed [km/h]
Average Absolute
Acceleration [m/s^2]
Maximum Acceleration
[m/s^2]
1 WLTP C3 P1-1 89 24,8 45 0,45 1,3
2 WLTP C3 P1-2 250 28,9 57 0,41 1,6
3 WLTP C3 P1-3 55 18,0 34 0,49 1,3
4 WLTP C3 P1-4 57 12,7 25 0,50 1,6
5 WLTP C3 P3 427 60,3 97 0,38 1,8
6 WLTP C3 P4 323 92,0 131 0,31 1,1
7 WLTP C1 P2 249 28,9 45 0,14 0,5
8 WLTP C1 P3 55 20,6 33 0,34 0,6
9 WLTP C1 P4 88 32,6 49 0,33 0,6
10 WLTP C1 P5 388 44,2 64 0,17 0,7
11 WLTP C2 P2 248 28,8 50 0,31 1,0
12 WLTP C2 P3 55 16,9 31 0,32 0,7
13 WLTP C2 P4 88 31,3 51 0,50 0,9
14 WLTP C2 P5 387 43,8 75 0,28 1,0
15 WLTP C2 P6 427 57,5 85 0,26 0,9
16 US06 P2 81 72,7 114 0,84 3,8
17 US06 P3 361 100,2 129 0,34 2,0
18 US06 P4 90 26,4 48 1,27 2,9
19 US06 P3 (short) 140 93,3 129 0,59 2,0
20 US06 P3 (short2) 290 101,5 129 0,38 2,0
21 ARB02 P23 169 82,5 110 0,50 1,9
22 ADVISOR_COMP. 292 49,4 55 0,11 0,6
23 CSHVR P1 280 30,6 61 0,32 1,0
24 CSHVR P2 218 31,0 55 0,30 1,0
25 CSHVR P3 109 25,8 44 0,40 0,8
26 CSHVR P5 199 25,6 70 0,33 1,2
27 CSHVR P6 87 7,7 27 0,32 0,8
28 CSHVR P7 163 32,6 54 0,38 1,0
29 CSHVR P8 180 23,7 58 0,31 0,8
30 Highway 1152 74,1 98 0,15 0,7
31 HWFET 762 78,0 96 0,17 1,4
32 Indian Urban 1 236 32,7 45 0,27 0,8
33 Indian Urban 2 256 34,5 63 0,21 0,7
34 LA92 - 1 235 66,6 99 0,43 2,0
35 LA92 - 2 163 39,6 61 0,57 1,9
36 LA92 - 3 167 68,8 108 0,61 3,1
37 LA92 - 4 179 45,2 78 0,46 3,1
38 SC03-1 601 34,3 88 0,42 2,3
39 SC03-2 54 42,0 66 0,70 2,3
40 SC03-3 71 61,1 88 0,71 2,0
41 SC03-4 148 47,0 62 0,37 2,1
Table 4 - Blocks from which test cycles are created (picked/combined)
51
To not overextend over the available space, next are presented only some of the blocks from Table 4,
with the same numeration. For a more detailed information of the blocks, see Annex and the Data
Disc attached in Annex or on the specified online address.
Figure 39 – Blocks extracted from: WLTP Class 3 Cycle
Figure 41 - Blocks extracted from: WLTP Class 2 Cycle
Figure 40 - Blocks extracted from WLTP Class 1 Cycle
52
As can be seen in the Figure 39 to Figure 42Figure 42 - Blocks extracted from CSHVC cycle (as well
as the missing blocks presented in annex), all the blocks are quite distinctive between themselves,
which allows the TCGT, even with a reduced pool of blocks (only 40) still be able to achieve varied
VSP mode distribution, that are similar to the on presented in the input data, as will be better seen in
5.2.1
The “blocks” not represented can be consulted on the data disk attached in the annex.
Figure 43 - Blocks extracted from US06 cycle of EPA FTP
Figure 42 - Blocks extracted from CSHVC cycle
53
5. Validation
5.1. Validation of the Test Cycle Generation Tool
After constructing the methodology for test cycle generation and after implementing it in an
engineering user friendly programming language, to carry on the heavy computation and repeated
iterations required, the following logical step is to validate the capability of TCGT to effectively
generate test cycles with desired requirements and parameters.
Information about the 5 original, and 5 generated cycles can be consulted in the screen captures of
the TCGT windows:
Figure 44 - TCGT –Calm urban driving cycle
This test cycle was generated from a driving cycle that was clearly made in a very calm way, as there
are almost no points above VSP mode 8. It is easy to distinguish the 12 blocks that compose the test
cycle, and judging by the VSP mode distribution the generated cycle matches very well the original
input cycle. The speed distribution (pie charts at the bottom of the windows in the figures), matches
perfectly and the deviations, on the right side of the figures, are very low. For an easier comparison
see Table 5, Table 6 and Table 7.
54
The five generated cycles from Figure 44 to Figure 48 present a distinctive speed profile one from
another. The blocks combined to create each driving cycle are distinctive and as can be seen, were
picked to allow a good approximation of the VSP mode distribution between original and generated
cycle (the blue and red curves on the middle graph, VSP mode distribution, almost overlap in most
cases).
Observing the VSP mode distribution curves, and keeping in mind that the generated cycles were all
made from the same available database of 41 blocks, these results already show the capability of this
methodology to generate cycles with similar specific power requirements. With finer tuning and
increasing the number of available blocks in the database, the tool could be improved to a level where
VSP mode distribution curve of generated test cycle would indistinguishable overlap the one from
original input data. Thus, based on same methodology, the TCGT could still be improved. The Table
Figure 45 - TCGT - Typical urban Figure 46 - TCGT - Rural
Figure 47 - TCGT - Aggressive rural driving Figure 48 - TCGT - Highway driving
55
5, Table 6 and Table 7, summarizes the data presented in the Figure 44 to Figure 48 (for easier
comparison, and ease of read):
Speed Distribution of the Cycles [km/h]
Duration [s] < 50 km/h
50 to 90 km/h
90 to 120 km/h
> 120 km/h
Cycle Generated Original Gen. Org. Gen. Org. Gen. Org. Gen. Org.
Calm Urban 2120 2925 88% 88% 12% 12% 0% 0% 0% 0%
Typical Urban 2142 8245 64% 69% 26% 18% 5% 11% 4% 2%
Rural 2182 2554 78% 78% 19% 21% 3% 1% 0% 0%
Aggressive rural
2201 5418 58% 60% 28% 26% 13% 13% 1% 1%
Highway Driving
2163 2125 41% 41% 33% 34% 22% 20% 4% 5%
Table 5 - Properties of the 5 Generated Cycles for TCGT Validation
Average Speed [km/h]
"Phi" parameter
Average Specific Power [kW/ton]
Cycle Gen. Org. deviation
[%] Gen. Org. Gen. Org. dev. [%]
Calm Urban 25,61 25,6 0% 2 2 4,19 4,07 3%
Typical Urban 39,86 39,82 0% 5 5 5,46 5,72 -5%
Rural 33,4 33,43 0% 2 2 4,69 4,72 -1%
Aggressive rural
46,22 49,19 0% 6 6 6,3 6,33 0%
Highway Driving
59,63 60,5 -1% 9 9 8,67 9,38 -8%
Table 6 - Properties of the 5 Generated Cycles for TCGT Validation (cont.)
Energy per kilometer [kWh/ton/km]
%Time in negative VSP
Time Stopped
[%] Generated Test Cycle
Cycle Gen. Org. dev. [%]
Gen. Org. dev. [%]
Gen. Org. Max VSP [kW/ton]
number Blocks used:
Calm Urban 0,16 0,16 3% 31% 31% -1% 12% 25% 24 12
Typical Urban 0,14 0,14 -5% 26% 25% 2% 18% 19% 30 11
Rural 0,14 0,14 -1% 29% 29% 0% 14% 14% 24 14
Aggressive rural
0,14 0,14 -1% 24% 23% 3% 16% 16% 41 10
Highway Driving
0,15 0,16 -6% 24% 24% 0% 2% 9% 38 10
Table 7 - Properties of the 5 Generated Cycles for TCGT Validation (cont.)
56
From the data and results presented in Figure 44 to Figure 48 and Table 5 to Table 7, it is easy to
understand the overall similarity between the characteristics and parameters of the real-world input
data and the ones resulting from the generated test cycle.
The duration of generated test cycle is set by the user, but it usually is recommended from 30 to 45
minutes (1800 to 2700 seconds). The test cycle duration is convenient to be short, for practicality in
laboratorial reproduction. The duration of input driving cycle can be from half an hour, to months or
years of driving data.
The deviation of speed distribution of the four speed categories considered (below 50; 50 to 90; 90 to
120; above 120 km/h), is convincingly low, with an average difference of 2%. Note that the speed
distribution is not a parameter that the TCGT specifically attempts to match, the good approximation
being a result of matching the average speed and VSP mode distribution.
Average speed of original data and generated cycle are almost identical not only in the 5 cases
discussed here, but as well in the hundreds of other generations made throughout this work. Average
speed is poorly matched mainly in the cases where the original real-world data used has a very high
average speed, thus turning difficult the task for TCGT to pick from the limited database blocks that
satisfy both design criteria (average speed and VSP distribution).
The “phi” parameter is matched in most cases, but could deviate in specific cases where the average
speed of input data is either unusually low or high. As “phi” parameter was defined as a function of the
average speed and percentage in high VSP modes, when these two parameters are in the same “phi”
range of original and generated cycle, as a consequence the “phi” parameter will be the same.
Average specific power is a parameter that essentially reflects the deviation of the VSP mode
distribution curves (it is the average of the Vehicle Specific Power value for all the cycle), as the later
ones are very similar, it is no surprise the average specific power values are very close as well. It is
curious to note that on average a vehicle on a trip uses from 7 to 12 horsepower.
The Energy per kilometer value (or Specific Energy per kilometer, kWh/ton/km) is achieved dividing
the specific power by the average speed, thus the deviation between the original and generated
cycles is explained by the deviations of the previously mentioned parameters.
Percentage of time in negative VSP, essentially represents the percentage of time spent breaking.
Being the VSP distribution well matched, matching the VSP mode 1 and 2 will result in good
approximation of the negative specific power.
The time spent stopped, as the name implies, represents the amount of time during the cycle with the
speed equal to zero. This value is sometimes sacrificed in favor of achieving the desired average
speed.
Maximum VSP specifies what is the vehicle‟s maximum specific power required to perform the
generated test cycle. The number of blocks used to generate the test cycle is merely informative.
57
5.2. Representativeness of the generated Test Cycles
Test Cycle Generating Tool (TCGT) allows the user to easily generate a test cycle that has a similar
average speed and similar Vehicle Specific Power modes distribution of the real-world input data, but
are those two criteria enough to assure that the generated test cycles are indeed similar to the input
data from point of view of fuel consumption and pollutants emission?
5.2.1. Validation with VSP distribution as input
In the case where the test cycle was generated from a given VSP mode distribution and average
speed (button “4)” on Figure 31), the fuel consumption and pollutant emission difference between the
real and generated cycle is given by the difference (deviation) of the VSP mode curves on the graph.
As in this case, the only method to calculate fuel consumption or pollutants emission, for a specific
vehicle, is to multiply the percentage of time share spent in each of the VSP modes by the average
consumption/emission rates, for that vehicle and for that VSP mode. The average values for each
VSP mode can be measured using the PEMS and a OBD data logger. To clarify the idea above, an
example is presented.
In the Table 8, is presented the summarized data of a trip with BMW 114i. For each second of the trip,
instantaneous speed and altitude where registered, as well as fuel consumption was deduced from air
flow and fuel/air ratio while the pollutants where measured with a Portable Emissions Measurement
System (PEMS) analyzing gas samples directly from exhaust pipe. Using VSP formula (Equation 1),
VSP is calculated for every second of the trip. Grouping all the above data for the corresponding 14
VSP modes specified in Table 2, the Table 8 is created.
BMW 114i (102CV 1598cc) [gasoline] euro5 VSP
mode Average Fuel g/s Avg. CO2 g/s Avg. CO g/s Avg. HC g/s Avg. NOx g/s
1 0,19 0,60 1,33E-04 5,50E-06 1,22E-07
2 0,22 0,71 5,03E-04 2,44E-05 1,96E-07
3 0,18 0,56 4,12E-04 6,37E-05 8,03E-08
4 0,53 1,69 7,13E-04 1,94E-05 3,12E-07
5 0,67 2,13 1,10E-03 1,73E-05 8,54E-07
6 0,87 2,75 6,87E-04 1,19E-05 3,93E-07
7 1,10 3,50 3,61E-04 4,13E-05 3,08E-07
8 1,32 4,19 7,15E-04 2,10E-05 3,34E-06
9 1,57 4,98 0,00E+00 1,24E-04 0,00E+00
10 1,89 6,00 2,20E-04 7,51E-05 2,88E-07
11 2,45 7,76 1,01E-03 1,55E-04 5,32E-06
12 2,86 9,06 3,87E-04 1,03E-04 2,60E-07
13 3,08 9,78 0,00E+00 2,87E-04 0,00E+00
14 3,33 10,57 2,60E-03 9,45E-04 9,36E-06
Table 8 - Fuel Consumption and Pollutant Emission per VSP mode (BMW 114i)
58
Having the average fuel consumption and pollutant emission data per VSP mode, allows to easilt and
with good approximation deduction of what would be the fuel consumed and pollutants emitted during
another trip for the same vehicle, for which only VSP mode distribution is known (or speed and
altitude data, from which VSP mode can be derived). As values in the table are specified per second,
knowing how much time was spend in each VSP mode, allows for the simple multiplication of VSP
mode duration and the corresponding per second value.
Taking the example further, assuming this BMW would drive on a racing Circuit, during half an hour,
and its VSP mode distribution would be similar to the one presented in Figure 29, (which for
simplification we assume is 40% VSP mode 1 and 60% VSP mode 14), the total fuel consumption
during this half hour race would be:
[ ] [
] [
] [ ] [
] [
]
[ ] Equation 3
Equation 3 shows method of determining the Fuel Consumption on a trip, using the table of Fuel
Consumption per VSP mode. This same technique can be applied to calculate any emission, for any
driving profile, for which the VSP mode distribution can be computed. Having this data table per VSP
mode allows calculation of the emissions and consumption for any trip, for that specific vehicle,
without having to perform the trip itself.
For comparison is given another example of consumption/emission per VSP mode for a different car
[15] , presented in Table 9.
In other words, VSP methodology, allows creation of tables, similar to a “Engine Map”, where instead
of specifying engine parameters in function of engine rotation speed and engine load, it specifies fuel
consumption and pollutant emission rates, in function of the 14 VSP modes. This solidifies the
assumption and decision made in Chapter 3, where in order to be representative and realistic, the test
generation methodology was based on matching VSP mode distribution and average speed.
Table 9 - Fuel Consumption and Pollutant Emission per VSP mode (Chevrolet Cavalier 2.2) [15]
59
5.2.2. Validation with Driving Cycle as input
On the other hand, when the test cycle is generated from a driving cycle (button “5)” from GUI),
comparison between generated and given cycle fuel consumption and pollutant emission can
additionally be made, beside the comparing method from 5.2.1, by simulating the cycles in a vehicle
simulating software (ADVISOR). ADVISOR (Advanced Vehicle Simulator) is a model written in the
widely used MATLAB/Simulink software environment. It can be used to simulate and analyze
conventional, advanced, light and heavy vehicles, including hybrid electric and fuel cell vehicles.
ADVISOR uses complex systems simulations, which allow for a good approximation of the fuel
economy, performance, or emissions of the simulated vehicle. Advisor simulates a driving cycle as
close as the specified vehicle is able to follow the speed curve, for specified cold or hot start, giving as
results average values for fuel consumed and pollutants emitted on that trip.
Using 5 real world driving cycles (1 very soft urban, 1 average urban, 2 mixed and 1 highway) and 5
corresponding test cycles where generated in TCGT, both original and generated cycles where
simulated in ADVISOR with 5 different cars (2 Diesel, 2 Gasoline and 1 hybrid), in a combination of 25
simulations of the original cycles and 25 of the generated cycles. For each corresponding pair of
generated/original data, for corresponding vehicle. A comparison was made with regards to fuel
consumption (gasoline or Diesel accordingly), nitrogen oxide and dioxide (NOx) emission, carbon
monoxide (CO) emission, particulate matter (PM) emission and unburned hydrocarbons (HC)
emissions.
For an easier visualization of how ADVISOR displays the calculations to the user, in Figure 49 and
Figure 50 are presented screen captures of ADVISOR post simulation windows.
On Figure 49, on the second plot from top is shown the State of Charge (SoC) of the electric battery
of the hybrid car. When the driving cycle begins (1st plot from top) the SoC states starts to deplete, yet
there are no emissions for that initial time (car moves only on electric motors, after a while the Internal
Combustion Engine (ICE) kicks in and pollutant emission starts to appear, although they are
somewhat reduced compared to later state (after approx. 1000 sec) when SoC of the battery reaches
the sustained state.
On Figure 50, the SoC plot is empty, as it‟s a conventional internal combustion engine vehicle, but
instead there is already a variation in the overall ratio engine/wheel, which is caused by the gear shifts
that are now present.
Taking now into account the 3rd
plot in the Figure 49 and Figure 50, we can see that PM emission is
almost nonexistent (as both vehicles are gasoline), NOx emission is high on the accelerating parts,
while the CO emission for the conventional ICE gasoline car are extremely high, which might be
explained by the fact that this car could be Euro 3 or older (comparing with the emission data from
Table 8 of the BMW 114i, it‟s easy to see how technology evolution can and does reduce significantly
pollutants emission).
60
Figure 49 - ADVISOR GUI post simulation window for an Hybrid Vehicle
Figure 50 - ADVISOR GUI post simulation window for conventional IEC vehicle
61
Table 10 summarizes the specifications of the 5 vehicles, from ADVISOR database, used to perform
the original (real-world data) driving cycles and the generated (with TCGT) test cycles:
Vehicle maximum Power [kW] mass (kg) Specific Power [kW/ton]
Gasoline 1 41 984 41,7
Gasoline 2 144 1924 74,8
Diesel 1 52 1914 27,2
Diesel 2 67 1210 55,4
Hybrid 57 1352 42,2
Table 10 - Main properties of the 5 cars used for Validation through ADVISOR
“Diesel 1” is a typical old London taxi, which presents a low specific power ratio, which caused him
some problems following the more demanding parts of the driving and test cycles, still the deviation of
more than 2km/h from the cycle‟s speed curve never exceeded 4% of the duration of the cycles.
The data from the 25 simulations can be found in the Table 11:
Fuel Cons.[l/100km] HC [g/km] CO [g/km] NOx [g/km] PM [g/km]
Car: Cycle Original Generated Original Generated Original Generated Original Generated Original Generated
Gaso
lin
e 1
Calm Urban 6,7 6,5 0,114 0,111 0,563 0,44 0,118 0,115 0 0
Urban 6,5 6,1 0,101 0,096 0,808 0,552 0,123 0,129 0 0
Rural 6,2 6 0,102 0,099 0,542 0,385 0,121 0,121 0 0
Aggressive Rural 5,9 5,8 0,093 0,09 0,617 0,565 0,125 0,122 0 0
Highway 6,3 5,8 0,093 0,087 1,002 0,599 0,121 0,137 0 0
Die
sel
1
Calm Urban 9,9 9,6 0,058 0,058 0,136 0,151 1,271 1,081 0,051 0,048
Urban 10,6 10,1 0,051 0,049 0,118 0,107 2,047 1,72 0,087 0,08
Rural 9,7 9,7 0,039 0,042 0,072 0,08 1,394 1,26 0,058 0,062
Aggressive Rural 10,3 10,1 0,042 0,036 0,07 0,062 1,912 1,842 0,086 0,085
Highway 10,6 10,4 0,035 0,034 0,048 0,052 2,377 2,118 0,099 0,097
Die
sel
2
Calm Urban 6,2 5,9 0,14 0,138 0,692 0,676 0,713 0,646 0,001 0,001
Urban 5,7 5,4 0,117 0,108 0,397 0,361 0,723 0,594 0,001 0,001
Rural 5,6 5,5 0,1 0,099 0,478 0,456 0,664 0,594 0,001 0,001
Aggressive Rural 5,3 5,1 0,092 0,097 0,305 0,292 0,6 0,565 0,001 0,001
Highway 5,4 5,1 0,086 0,083 0,196 0,193 0,78 0,581 0,001 0,001
Gaso
lin
e 2
Calm Urban 15,3 14,9 0,311 0,309 0,779 0,758 0,55 0,528 0 0
Urban 13,5 12,8 0,243 0,239 0,711 0,577 0,557 0,544 0 0
Rural 13,5 13,3 0,263 0,26 0,587 0,633 0,519 0,5 0 0
Aggressive Rural 12,3 12 0,22 0,218 0,561 0,576 0,528 0,51 0 0
Highway 12,2 11,5 0,205 0,197 0,712 0,541 0,549 0,53 0 0
Hyb
rid
Calm Urban 8 6,4 0,169 0,152 0,171 0,165 0,046 0,045 0 0
Urban 7,8 6,4 0,14 0,124 0,16 0,143 0,053 0,045 0 0
Rural 7,3 6 0,138 0,122 0,146 0,134 0,044 0,039 0 0
Aggressive Rural 7,4 6,8 0,127 0,12 0,149 0,141 0,051 0,047 0 0
Highway 6,4 7,6 0,104 0,119 0,121 0,149 0,043 0,055 0 0
Correlation*: 0.984 0.996 0.946 0.997 0.982**
* Correlation between original and generated cycle data, as calculated by “=correl()” command in MS Excel **Only Diesel 1 vehicle was considered for PM correlation computation
Table 11 - Comparison between original and generated cycles (using ADVISOR)
62
The ADVISOR vehicle simulation software, takes into account all the parameters that establish the
working state of the engine (vehicle speed, acceleration, mass, aerodynamic drag, rolling resistance,
engine efficiency/power/torque on different rotation speeds, gear shifts, temperatures etc.), in order to
calculate as accurately as possible the fuel economy and emissions while performing a specific
driving cycle.
The comparison between original and generated cycles behavior is presented further below, while a
comparison between 5 vehicles and the 5 cycle types is easier to be made while interpreting the table
data.
It is of no surprise that the “Gasoline 2” vehicle, with almost 200 horse power, was the one that
consumed most fuel while performing the same cycles as the other 4 vehicles (it is a sport gasoline
car, from 90‟s). The “Diesel 1” vehicle, which is an old London TAXI, also present much higher fuel
consumption than the modern Diesel cars, and it is mainly explained by the older technology used in
the first one and by its large mass (vehicle with almost 2000kg, in a urban environment, where
start/stop situations create many acceleration requirements, which are responsible for the vast
majority of fuel consumption). In the case of Hybrid vehicle, the fuel consumption presented is the
“equivalent to gasoline” fuel consumption, that is, the electric energy consumed from the battery is
also accounted and converted into equivalent amount of gasoline (electric energy from batteries in
kWh and equivalent heating value of gasoline in kWh/liter) that is summed to the actual gasoline
consumed by the internal combustion engine of the hybrid car. The approach that ADVISOR takes on
calculating the fuel consumption of the Hybrid car makes it easier to compare different trips with
different lengths regardless to the initial state of charge of the battery, although, should be kept in
mind, that for the same amount of consumed energy, the useful energy applied to the wheels differs
significantly, as an internal combustion engine has an average efficiency of 25% while the electric
motors have an average efficiency of over 85%. When focusing on a specific vehicle and analyzing
how its fuel consumption vary from cycle to cycle, we can see an increased fuel consumption in urban
environment compared to the highway driving, which is caused, as already mentioned, by the much
more frequent accelerations that compensate and overshoot the increased aerodynamic drag and
rolling resistance on the highway. This is particularly observed in the case of “Gasoline 2” car, which
is clearly designed to travel on higher speed, thus having efficiency peak on higher engine RPMs,
explaining the greater difference in fuel consumption from urban to highway environment.
The CO2 emissions are not represented, as they are directly dependent of fuel consumption. Per each
liter of Gasoline consumed there is 2.31 kg of CO2 emitted, while per each liter of Diesel burned there
will be emitted 2.68 kg of CO2. To be completely accurate, one should discount the carbon included in
unburned hydrocarbons (HC) and carbon monoxide (CO) to obtain the real CO2 emissions, but the
error is much inferior to 1% for modern vehicles. [23]
Unburned Hydrocarbons (HC) are originated by: lack of time for the chemical reactions to fully occur,
lack of oxygen concentration in the mixture or poor turbulence inside the cylinder (non-uniform
mixture), between many other reasons. Poor turbulence inside the engine cylinder is from the lack of
63
rotational speed, while accelerating from stopped position or driving at low speeds, which are more
frequent in urban driving. This is the reason why we can observe a reduction in HC emission, for all 5
vehicles, when we go sequentially from calm urban, to highway cycles (top to bottom in the table). It is
important to note that the emission per kilometer is lower for higher average speeds, but HC emission
per unit of time might be the same or even increase for highway trips compared to urban. For
example, taking the “gasoline 1” vehicle‟s HC emission: Urban= 0.114 [g/km] and Highway= 0.096
[g/km], and considering average speeds of: Urban= 30 [km/h] and Highway= 70 [km/h], multiplying the
emission per distance by the average speed (distance per time) we conclude that the emission per
unit of time are: Urban= 3.42 [g/hour] and Highway= 6.72 [g/hour]. As the main objective of a vehicle
is transportation from A to B, specifying parameters per unit of distance is usually more representative
and useful. Effects: HC react with nitrogen oxides in the presence of sunlight to form ground level
ozone, a primary ingredient in smog. Though beneficial in the upper atmosphere, at the ground level
this gas irritates the respiratory system, causing coughing, choking, and reduced lung capacity. [24]
Carbon monoxide (CO) emissions are created by the same principles as the HC mentioned above,
but in this case, the cold start condition, is affecting greatly the CO emission. All the simulations made
in ADVISOR where made from hot start condition, as to assure that cycle duration doesn‟t affect
average emissions. For a short trip, with cold start condition, the warming up duration takes a much
higher percentage of total duration compared to a more extensive cycle, thus when averaging the
high emission during the warming stage for different durations, and thus different total distances, the
results are inconsistent. In the case of Diesel vehicles, as the Air/Fuel ratio is much greater than
stoichiometric ratio, the main factor which determines the “easiness” of the carbon from fuel to found
oxygen mainly depends on the turbulence and fuel atomization. [23] The later does not depend as
much on engine speed as the turbulence does, so at higher average speeds we see a decrease in
CO emission, which is explained by better air/fuel mixture and thus a more “complete” combustion. In
the case of gasoline vehicle, the Air/Fuel ratio is much closer to stoichiometric ratio, and does not
fluctuate much, the turbulence in Otto engine is also lower than in Diesel, so in this case the main
factor that influences CO emission on already heated up engine is the time available for combustion.
It‟s easy to understand that, with higher speeds experienced on highway, engine rotation will be
higher as well, leading to less available time for the combustion to occur, which leads to more CO not
fully reacting with oxygen to create CO2, but on the other hand, higher speed include higher
turbulence (which in some way, compensates for the reduction of available time). With modern
technology, engines are better tuned and more complex CFD simulations are made in the design
stage that allow for further reduction on pollutant emission and as was mentioned before, a modern
car, pollute one or even two orders of magnitude less compared to a vehicle from 60s, 70s or even
later. Effects: CO is odorless, colorless, and poisonous gas. When inhaled, CO blocks oxygen from
the brain, heart, and other vital organs. Fetuses, newborn children, and people with chronic illnesses
are especially susceptible to the effects of CO. [24]
Nitrogen monoxide and dioxide (NO and NO2, or simply NOx) originates from the dissociation of
nitrogen molecule (N2) and later combination with oxygen. Nitrogen dissociation is caused by high
64
pressure and/or high temperatures, which are present inside the cylinder of the internal combustion
engine. Diesel engines originates much higher pressure inside the chamber, caused by the high
compression ratio which in consequence rises the temperature to higher values than experienced in
Otto engine (with exception of Liquefied Petrol Gas, which also generates higher temperatures
compared to gasoline). [23] This explains the higher NOx emission rates of Diesel vehicles compared
to Gasoline vehicles (“Gasoline 2” sport vehicle, with older technology still generates a significant
amount of NOx) [2]. There seems to be no evident variation of NOx emission per unit of distance
when we go observe a single vehicle throughout the 5 cycles it performed. Effects: These pollutants
cause lung irritation and weaken the body's defenses against respiratory infections such as
pneumonia and influenza. In addition, they assist in the formation of ground level ozone and
particulate matter. [24]
Particulate Matter (PM), despite considerable amount of research, PM‟s origination precise factors are
not well defined, although modern, turbo-charged diesel engine show much lower PM emission rates
compared to older diesel engines. In the case of spark ignition engines PM emission is negligible. PM
have usually higher concentrations at higher engine loads. Taking into account the data from the
Table 11 we see high emission rate for the old diesel taxi, very low emissions for modern diesel
vehicle and no results were presented by ADVISOR for gasoline vehicles (also expected to be very
low). Effects: These particles of soot and metals give smog its dark color. Fine particles, less than
one-tenth the diameter of a human hair, pose the most serious threat to human health, as they can
penetrate deep into lungs. [25] [24]
When it comes to the comparison between the behavior of the original and the generated driving
cycles, with regards to fuel consumption and pollutant emission, for an easier understanding, the
following plots (Figure 51 to Figure 56) visually represent the data:
Figure 51 - Fuel Consumption Comparison
y = 0.9787x - 0.2095 R² = 0.9687
0
2
4
6
8
10
12
14
16
0 2 4 6 8 10 12 14 16
Gen
erat
ed
Original
Fuel Consumption [l/100 km]
65
The Figure 51 represents a scatter graph of the fuel consumption. Each point on the plot corresponds
to 2 values from the “Table 11”, being on x-axis represented the fuel consumption that resulted on
performing the original cycle, while on y-axis is the fuel consumption that resulted from the simulation
of the generated test cycle. In other words, an original Driving Cycle is available (from real world
driving while recording the speed), this cycle is used as input for an ADVISOR simulation, with one of
the available vehicles in its database, the resulting fuel consumption of this trip is then used to define
the x coordinate on the scatter graph in Figure 51. Taking the same driving cycle, it is used as input
for the Test Cycle Generation Tool (TCGT) to generate a test cycle, with reduced duration, but similar
VSP mode distribution, which is later used as input for ADVISOR simulation, for the same database
vehicle as previously and the resulting fuel consumption is defining the y coordinate of the point. This
process is repeated for the 25 combinations of 5 vehicles and 5 cycle types, and the scatter graph
above is created (Figure 51).
Ideally, the fuel consumption of the generated test cycle should be equal to the fuel consumption of
the original cycle, which would result in every point on the graph to have same x and y coordinate,
thus all would be aligned on the x=y diagonal line. Inserting a interpolation line of the 25 points in the
scatter graph on Figure 51 and extending it to the left until it intersects one of the axis, we can see
that it is parallel to the x=y line but situated slightly below. The equation of the interpolating line is: y =
0.9787x - 0.2095, which present coefficient of 0.98 very close to 1 and has a displacement of 0.2
liters/100km downwards. In other words this means that on average, the generated test cycle will
have a fuel consumption that is 0.2 liters per hundred kilometers below the fuel consumption of the
original cycle (that is on average, but as can be seen on the Figure 51, there are points that have
higher y coordinate that x, which translates to higher fuel consumed in the generated test cycle
compared to the original one). The average underestimation of 0.2 liters on 100 kilometers represents
2% to 3% deviation from the typical values of fuel consumption (6 to 10 l/100km). The R2 value is
0.969, which represents the dispersion of the data points from the linear trend. When all the points are
aligned perfectly in a straight line, the R2 value is 1. Being the actual R
2 value, of the above
distribution, very close to 1, points are situated near the trend-line, as is observed on the chart.
66
Figure 52 - Unburned Hydrocarbons Comparison
Similar analysis can be made for the unburned hydrocarbons (HC) data of the original and generated
cycles. The points are also well aligned along the trend-line, while the later one, like in the fuel
consumption case, is almost parallel to x=y line, and slightly shifted downwards. The equation that
defines the trend-line is: y = 0.9809x - 0.0014, where y is the HC emission in grams per kilometer of
the generated test cycle and x is the HC emission of the original cycle. The slope of the trend-line is
very close to the unity, and on average the values of the generated cycle HC emission are 0.0014
grams per kilometer below the values obtained by simulating the original cycle. The deviation of
0.0014 g/km represents an error from 0.5% to 3% of underestimation of the real world original cycle.
The R2 value in this case is 0.992, which accentuates, the already observed fact, that the cluster of
data points align almost perfectly on top of the trend-line (which in its case is very close to the y=x
line).
Figure 53 - Carbon Monoxide Comparison of all 5 vehicles
y = 0.9809x - 0.0014 R² = 0.9916
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35
Gen
erat
ed
Original
HC [g/km]
y = 0.7515x + 0.0467 R² = 0.8955
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1
Gen
erat
ed
Original
CO [g/km]
67
The parameter that deviates the most between the original and generated driving cycles is carbon
monoxide (CO) emission. The discrepancy is especially evident for higher emission values (above 0.5
grams per kilometer), which is the case of “Gasoline 1 and 2” vehicles. The equation of the trend-line
in this case is significantly deviated from y=x, being y=0.75x+0.0467. That means, that on low
emission rate, the deviation between original and generated cycle in terms of CO emission are
insignificant, while on higher emission rates the deviation is increasingly greater, that is, on low rates,
the 0.75 coefficient of original emission rate (0.75x) is compensated by the +0.0467 shift upwards of
the trend-line, thus achieving similar generated (y value) and original (x value) emissions, while on the
higher emission rates (above 0.5 g/km), the trend-line upward shift does not compensate enough, and
the deviation increases. Also the R2 value decreased considerably from previous cases, which is easy
to explain judging by the spread of data points starting from 0.5g/km and higher.
Taking a look at the Table 11 we can observe that in all vehicles, except “gasoline 1”, the CO grams
per kilometer decreases when we go through the 5 driving cycles type, from urban up to high-way,
both on original as well as generated cycles. Gasoline 1 vehicle, not only does not follow the trend of
the other vehicles (Gasoline 2 included), but does increase CO emission a lot for higher average
speeds trips (it‟s important to mention that even a significant increase of CO emission per unit of time
would have been averaged by the increased mean speed, resulting in lower emission per distance
traveled), so judging by the values from the table, the “gasoline 1” vehicle is atypical and the results
should be interpreted with care. If the “gasoline 1” vehicle, which is a Euro 1 car, would be omitted,
the plot from Figure 54 will result:
Figure 54 - Carbon Monoxide comparison of 4 vehicles (all except "gasoline 1")
Now the CO emission of the generated and original cycles have much closer values, having now a
trend-line with a closer to y=x equation (y=0.9x+0.014), as well as a much higher R2, which is once
more not as close to 1 as in previous cases, because of the higher spread encountered for emissions
y = 0.903x + 0.014 R² = 0.9661
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
Gen
erat
ed
Original
CO [g/km]
68
above 0.5g/km. Trend-line equation, has a low upward shift of 0.014, and on average the generated
cycle underestimates the original cycle by 10% (y=0.9x), which is explained mainly by the not as high
acceleration rates, present in the “blocks” from which the test cycle is generated, as the ones
encountered on real world driving situations. Steep acceleration is one of the causes of higher CO
emission rate.
Figure 55 - Nitrogen monoxide and dioxide Comparison
Nitrogen Monoxide and Dioxide are mainly generated at high temperature and pressure, which are
common so high load (high acceleration) situations. High acceleration situations are not as
aggressive in the blocks used for test cycles generation by TCGT as they are in real world driving, as
they come from existent test cycles, which to assure an easy performance on rolling bench, and that
even vehicles with low power can perform them, are made with “softer” dynamics demands. This
explains why NOx is underestimated by the generated test cycle compared to the original input cycle.
The generated test cycle will have usually 12% less NOx emission than the real world counterpart
cycle. As the R2 parameter is very close to 1, the data points are aligned on a straight line, so the
deviation is constant, which helps in the later adjustment of the obtained value from the generated
cycle if the real world values is to be obtained. That is, dividing the NOx emission value obtained from
performing the test-cycle on the rolling bench by 0.88 will result in a value that is very close to real-
world case.
y = 0.8832x + 0.0125 R² = 0.9932
0
0.5
1
1.5
2
2.5
0 0.5 1 1.5 2 2.5
Gen
erat
ed
Original
NOx [g/km]
69
Figure 56 - Particulate Matter Comparison
Particulate Matter is heavily reduced in modern cars with high performance PM filters, compared to
older Diesel Vehicles. ADVISOR doesn‟t have a reliable resolution to show the PM emission rate of
the modern “Diesel 2” vehicle, leading to a chart containing only the data points of the “Diesel 1”
vehicle. Judging by the reduced sample of points, would be erroneous to take deep conclusions about
the representativeness of the generated cycles. The interpolation line equation is underestimating the
real-world data by approx. 8%, and the few points seem to be well aligned, judging by the R2 value of
0.965, relatively close to 1.
Another way of the measuring methods to tell how well the original cycle data matches the generated
test cycle data can be the correlation between the 2 datasets. This was made using the correlation
function, which in MS Excel is given by the “=correl()” command. Correlation function is given by the
Equation 4:
Equation 4
, where x and y are the values of the original and generated cycles for each compared parameter.
Correlation is a measure of how easy it is to predict the value of the corresponding point form the
second data set (of the original or generated cycle), given a point on the first data set (of the
generated or original cycle), or by another words, given x(or y) coordinate of a point, how easy is to
predict which is the y(or x) coordinate of the corresponding point. Closer the value to 1, better the
correlation. Correlation doesn‟t account for the slope or curvature of the interpolation line, thus does
y = 0.9245x + 0.004 R² = 0.9647
0
0.02
0.04
0.06
0.08
0.1
0 0.02 0.04 0.06 0.08 0.1
Gen
erat
ed
Original
PM [g/km]
70
not show if the value of the generated cycle undershoot or overshoot the target value. Judging by the
high Correlation values in the last row of Table 11, are high and close to 1, that means that by
performing the generated test cycle in lab, one can predict what will be the consumption and emission
of this same vehicle in real world driving conditions.
The overall slight underestimation of the test cycle values, while trying to represent the original cycle
values is mainly explained by the fact that the blocks used for the test cycle generation in TCGT are
mainly extractions of existent or in development test cycles, which usually represent a smooth speed
profile to ease the task of the technician who tries to perform the cycle on the test bench. Although,
should be remarked that the underestimation is much, much lower compared to the underestimation
made by the current practiced certification test cycles.
71
5.3. Influence of generated test cycle duration
Figure 57 - VSP Mode Distribution Deviation vs. cycle duration
In the Figure 57 is presented the deviation between the original and generated cycles as a function of
the duration of the generated cycle. The chart was constructed from 227 test cycles simulations,
mainly with durations of 1800 and 3000 picked from the drop-down menu (that is why we can see
higher concentrations of generated cycle‟s duration around those values). The total deviation in
percentage is calculated by the following formula (Equation 5):
( ∑(|
|)
Equation 5
As the TCGT in iterative way picks, modifies and removes blocks from database to combine them in
the best way possible to achieve the desired VSP mode distribution and average speed, the higher
duration allowed for the test cycle, more blocks it can pick, thus allowing, usually, for a closer
approach to the target. Test cycles with durations below 1200 second, using the limited blocks
database, are extremely difficult to achieve while assuring a satisfying similarity in the VSP mode
distribution. The maximum deviations found were lower than 2%.
0.0%
0.2%
0.4%
0.6%
0.8%
1.0%
1.2%
1.4%
1.6%
1.8%
2.0%
1000 1500 2000 2500 3000
Tota
l dev
iati
on
of
VSP
mo
de
curv
es
Generated Test Cycle Duration [s]
VSP mode deviation with duration
72
Using a power curve trend-line to represent the average tendency of total deviation with test cycle
duration, we can see that test cycle benefit a lot in their representativeness if their duration is allowed
to be around 2400 seconds or longer, as the reduction in deviation for cycles with higher time duration
is not as significant as the reduction achieved going from 1500 second to 2400 s.
While taking in account the energy per kilometer deviation between generated and original cycles, the
following chart (Figure 58) arises:
Figure 58 - Energy per kilometer deviation vs. cycle duration
The total deviation of energy/ton/km in percentage is calculated by the following formula:
| | Equation 6
,where:
ASEpKOC=average specific energy per kilometer of original cycle
ASEpKGC=average specific energy per kilometer of generated cycle
The chart in Figure 58 doesn‟t suggest the same evident decrease in deviation between the two
cycles with increase in the duration of the generated test cycle as it was present in the Figure 57. As
the main energy deviation comes from very high VSP modes, which are difficult to generate with the
available blocks in the database, the energy deviation does not decrease in the same way as VSP
mode distribution deviation. In any case, the reduction indeed exists, just not as evident and decisive.
Once more this suggests the user to try and generate test cycles with higher durations if possible, to
achieve lower deviations between generated and original cycles parameters.
Although it is suggested to allow a test cycle duration of about 2100 or 2400 seconds, in some cases,
even with shorter durations the achieved VSP mode distribution is as good as the one of higher
durations cycles. Actually, even though the TCGT has a dropdown menu option called “Unlimited”
0%
2%
4%
6%
8%
10%
12%
14%
16%
18%
1000 1500 2000 2500 3000
Tota
l dev
iati
on
of
kWh
/to
n/k
m
Generated Test Cycle Duration [s]
Energy per km deviation with duration
73
which allows generation of test cycles with much greater duration, the stopping criteria “if VSP mode
total deviation lower than the stopping criteria value” is met sooner rather than later, so for a specific
input data, there is a test cycle duration above which it will not generate, thus meaning that the option
for very long duration (1 hour and above) will be useful only in occasional situations where TCGT is at
difficulty to match VSP mode distribution at lower test cycle‟s durations.
5.4. Influence of the “weight vector”
The drop-down menu option presented in “2)” in chapter 4.1, allows the user to pick which region of
VSP mode distribution curve should have priority in approximation when test cycle is generated. This
is done by picking the “weight vector”, as described in Table 3. Figure 59 and Figure 60 present the
comparison of the consequences in picking a “urban” weight vector and “high-way” weight vector:
Figure 59 - VSP mode distribution with "urban" weight vector
Figure 60 - VSP mode distribution with "high-way" weight vector
74
As can be seen in the Figure 59 and Figure 60, the VSP mode distribution of the generated test cycle
is closer to the “given” original VSP mode distribution in the specific characteristic regions (low VSP
for urban case, and high VSP modes for high-way situation).
A sensitivity analysis of the impact of choosing the right “weight vector” from the drop-down menu, in
the Figure 61 is presented an average deviation of VSP mode curves for the most common “phi”
parameters and the 9 available options to pick from:
Figure 61- VSP mode deviation with "weight vector" for 4 common "phi" parameters
From the vast quantity of input data used to generate the test cycles for TCGT validation, the majority
fall into the “phi” parameter value 2, 5, 6 or 9. (as “phi” parameter was defined in Figure 38, in 4.2.1).
The values in Figure 61 are achieved from the average of 4 real-world trips each. Overall we can see
that the “Uniform” weight vector is a good pick in all cases, while for specific “phi” values, if user is not
satisfied with the VSP mode distribution of the generated test cycle, he can use another weight vector
searching for the next lowest “VSP mode deviation” in the Figure 61.
For instance, an aggressive motorway real world data (phi=9), could be well simulated if the weighting
vector chosen was the “Mixed” one. Once more it is important to mention, that due to the limited
“blocks” database, the quality of representativeness of the input data differs from case to case, and if
the user is not satisfied with initial results, he is free to adjust the variables and re-generate the test
cycle to achieve better results.
0%
2%
4%
6%
8%
10%
12%
14%
16%
18%
Un
ifo
rm
Low
-VSP
Low
-Me
d V
SP
Me
d V
SP
Me
d-H
igh
VSP
Hig
h V
SP
Hig
hw
ay
Urb
an
Mix
ed
VSP
MO
DE
DEV
IATI
ON
phi 2 phi 5 phi 6 phi 9
75
6. Applications
After going through the validation of Test Cycle Generation Tool, now it is appropriate to generate
several test cycles for specific cases.
6.1. Test Cycles based on a combination of different trips
Using the ability of TCGT to generate test cycle based on given driving cycle (speed and altitude), the
following real-world data was used (Figure 62):
The original driving cycle has a total duration of 22000 s (over 6 hours), also the altitude variation for
the whole cycle is available. As can be observed, this cycle has urban parts (low speed, around
second 5500, 11000 and 17000), mixed and highway (picks above 100 km/h). Using this data in
TCGT, the following test cycle is generated (Figure 63):
Figure 63 - Test Cycle generated through TCGT
The generated test cycle has a duration 10 times shorter than the original input cycle. As can be seen
in the Figure 64, the VSP mode distribution, average speed, energy per kilometer are very well
matched.
0
20
40
60
80
100
120
140
15
81
15
17
22
29
28
63
43
40
04
57
51
45
71
62
86
85
74
27
99
85
69
13
97
01
02
71
08
41
14
11
19
81
25
51
31
21
36
91
42
61
48
31
54
01
59
71
65
41
71
11
76
81
82
51
88
21
93
91
99
62
05
3
spee
d [
km/h
]
time [s]
0
20
40
60
80
100
120
140
15
20
10
39
15
58
20
77
25
96
31
15
36
34
41
53
46
72
51
91
57
10
62
29
67
48
72
67
77
86
83
05
88
24
93
43
98
62
10
38
11
09
00
11
41
91
19
38
12
45
71
29
76
13
49
51
40
14
14
53
31
50
52
15
57
11
60
90
16
60
91
71
28
17
64
71
81
66
18
68
51
92
04
19
72
32
02
42
20
76
12
12
80
spee
d [
km/h
]
time [s]
Figure 62 - Original Driving Cycle
76
Another way of characterizing a driving cycle is using the “acceleration vs speed” distribution. This
distribution shows where the accelerations occur with higher frequency, as at high speeds,
accelerations require high amount of power.
Figure 65 - Acceleration [m/s2] vs. Speed [km/h]; Blue - Generated, Green - Original
With exception to positive accelerations at high speed, the distributions in Figure 65 are similar. The
difference in high speed accelerations are originated once again in the “non-aggressive” nature of the
“blocks” that are used for the test cycle generation. Accelerations of 1m/s2 and more for speeds above
100 km/h are present in original cycle, but not in the generated test cycle.
Figure 64 - TCGT output for the 22000 seconds input driving cycle
77
6.2. Test Cycles based on a population (fleet)
During half a year, 50 drivers in the district of Lisbon were monitored, registering the speed and
altitude data. [20] Using the Vehicle Specific Power definition form 3.1, the VSP mode distribution for
a collection of 50 drivers was made, and is presented in Figure 66:
Figure 66 - VSP mode distribution of the 50 drivers
The “total 50” line in the Figure 66 represents the averaged (by amount of driving time) VSP mode
distribution of all 50 monitored drivers (with different vehicles and different driving conditions). The
other 3 curves were created for trips with average speeds typical for urban, rural and highway
conditions. The “mixed” or rural conditions VSP mode distribution is very close to the total VSP line,
as shown in the figure, and this trend was also observed in other data collections. This once again
suggests that the study of a greater sample of drivers of a given country, can result in the VSP mode
distribution that represents very well the average driving behavior of that nation, thus allowing the
generation of a test cycle customized specifically for that case.
Using the VSP mode distributions and mean speed from Figure 66, as input data for the TCGT, the
following results are achieved (Figure 67):
0.0%
5.0%
10.0%
15.0%
20.0%
25.0%
30.0%
35.0%
1 2 3 4 5 6 7 8 9 10 11 12 13 14
% o
f ti
me
in V
SP m
od
e
VSP mode
Total 50 Urban (<40km/h) mix(40-70km/h) Highway(>70km/h)
78
Using the VSP mode distribution of a population or a large fleet, and posterior removal of the
percentage of time from high VSP modes, and redistributing it accordingly to the low VSP modes, will
result in a VSP mode distribution with low power requirements. Using this VSP mode distribution as
input for the TCGT will result in a test cycle that is suitable for vehicles with low total specific power. It
can be observed the small difference at the VSP mode 14, which is originated from the limited high
power demand of the “blocks” form the available database for the generation process.
For example, if a test cycle is desired for a vehicle with 45 horsepower, and 1400 kg of mass, that is,
24kW/ton, the test cycle must have VSP mode distribution limited up to VSP mode 11. The procedure
to generate this test cycles, is to take the percentage of time share of VSP modes 12, 13 and 14, and
redistribute it proportionally to VSP modes 1-11, and later use this VSP distribution as input data for
TCGT, which will then generate a test cycle suitable for this specific, low power vehicle.
Both VSP mode distributions presented in Figure 64 and Figure 67 are representative of real-world
data (measurements made on typical driving behavior). Comparing it to the VSP mode distribution of
NEDC type-approval test cycle from Figure 16, it is immediate that similarities are very scarce. The
TCGT‟s main goal is to help emend this unconformity.
Figure 67 - TCGT results for the VSP mode distribution data of 50 drivers
79
7. Conclusions and Future Work
The methodology for generating test cycles, created, validated and applied in this work proved to be
accurate and representative of desired input data. It proves to be possible, using segments from
existing driving cycles (many of which are already used for vehicle certification) combine and create
new test cycles, with limited duration, whose characteristics are very similar to the real-world data
used as input.
Driving cycles with similar average speed and VSP mode distribution have a similar fuel consumption
and pollutant emission when simulated in the vehicle simulation environment (ADVISOR). Validation
of the methodology and the Test Cycle Generating Tool (TCGT) was made on two levels. Firstly, it
was verified that the TCGT can generate test cycles with very similar VSP mode distributions, while
maintaining the same average speed as the source data. Even with the reduced database of available
test cycle segments (“blocks”), the amount of possible combinations being achieved is overwhelming,
allowing from good, to “perfect overlap” VSP mode distribution matches (in some cases, it is required
to fine tune the comparison “weighting vector” or increase the test cycle duration, to achieve the
desired results). Secondly, it was simulated in ADVISOR and was revealed that both, original and
generated cycles, show very similar fuel consumption and pollutant emissions per distance of travel.
The ability to generate driving cycles, based on real-world data, with reduced duration, yet with very
similar kinematic behavior, that yields similar fuel consumption and emissions, is one of the main
achievements of the developed methodology.
On average, the deviation between the real-world driving data and the test cycle that was generated
based on it, is lower than: 1% for mean speed, 2% for specific power, 2.5% for fuel consumption and
7% for pollutant emissions. This deviations are much lower compared to the deviations between real-
world data and the, current in use, NEDC test cycle. [5] [18] [25]
As suggestions for the continuation of the work done in this thesis could be:
Improve the generation methodology: add average positive acceleration, gear shifts, number
of acceleration picks etc.
Improve the TCGT tool
o Faster computation time
o Larger “blocks” database
o Closer approximation of VSP mode curves and average speeds
o Addition of new parameters to be matched by TCGT
Collect driving data representative for specific country or region, and generate a test cycle
with type-approval characteristics
80
Bibliography
[1] "Cars produced in the world," 2012. [Online]. Available: http://www.worldometers.info/cars/.
[Accessed 2 August 2013].
[2] D. Carslaw, S. Beevers, J. Tate, E. Westmoreland and M. Williams, "Recent evidence concerning
higher NOx emissions from passenger cars and light duty vehicles," ELSEVIER - Atmospheric
Environment, no. 45, pp. 11 (7053-7063), 2011.
[3] G. N. A. Gonçalves, "Energy and Environmental Monitoring of Alternative Fuel Vehicles,"
UNIVERSIDADE TÉCNICA DE LISBOA - Instituto Superior Técnico (IST), Lisboa, 2009.
[4] Climate Action, "Reducing CO2 emissions from passenger cars," European Comission , 30 Jully
2012. [Online]. Available:
http://ec.europa.eu/clima/policies/transport/vehicles/cars/index_en.htm. [Accessed 16 June
2013].
[5] P. Mock, J. German, A. Bandivadekar and I. Riemersa, "Discrepancies between type-approval
and "real-world" fuel consumption and CO2 values," icct - the International Council on Clean
transportation, no. 2012-02, p. 13, 2012.
[6] R. Joumard, M. André, R. Vidon, P. Tassel and C. Pruvost, "Influence of driving cycles on unit
emissions from passenger cars," PERGAMON, p. 8, 2000.
[7] M. Williams and D. Carslaw, "Newdirections: science and policyeout of step on NOx and NO2?,"
Atmospheric Environment, no. 45, 2011.
[8] M. Rexeis and S. Hausberger, "Trend of vehicle emission levels until 2020- Prognosis based on
current vehicle measurements and future emission legislation.," in 6th International Conference
on Urban Air Quality, Nicosia, Cyprus, 2007.
[9] L. Pelkmans and P. Debal, "Comparison of on-road emissions with emissions measured on
chassis dynamometer test cycles," Transportation Research Part D - Elsevier, no. 11, pp. 233-
241, 2006.
[10] M. Weiss, P. Bonnel, R. Hummel and N. Steininger, "A complementary emissions test for light-
duty vehicles: Assessing the technical feasibility of candidate procedures," JRC Scientific and
Policy Reports - Report EUR 25572 EN, 2013.
[11] M. André, "Real-world driving cycles for measuring cars pollutant emissions – Part A: The
ARTEMIS European driving cycles," INRETS - Institut Nationale derecherche sur les transports et
leur securite, p. 97, 2004.
[12] "World harmonized Light vehicle Test Procedure (WLTP)," UNECE - United Nation Economic
Comission for Europe, 13 June 2012. [Online]. Available:
https://www2.unece.org/wiki/pages/viewpage.action?pageId=2523179. [Accessed 21 April
2013].
[13] Environmental Protection Agency, "Development of Emission Rates for Light-Duty Vehicles in
81
the Motor Vehicle Emissions Simulator (MOVES2010)," EPA - United States Environmental
Protection Agency, 2011.
[14] T. Lee, J. Park, S. Kwon, J. Lee and J. Kim, "Variability in operation-based NOx emission factors
with different test routes, and its effects on the real-driving emissions of light diesel vehicles,"
ELSEVIER - Science of the Total Environment, p. 9, 2013.
[15] C. Frey, K. Zhang and N. Rouphail, "Fuel Use and Emissions Comaprisions for Alternative Routes,
Time of Day, Road Grade, and Vehicles Based on In-Use Measurements," Environmental Science
& Technology, no. 42, pp. 2483-2489 (7), 2008.
[16] E. Ericsson, "Independent driving pattern factors and their influence on fuel-use and exhaust
emission factors," Pergamon, no. Transportation Research Part D, pp. 325-345, 2001.
[17] "Diesel Net - Emission Standards," ECOpoint Inc, November 2012. [Online]. Available:
www.dieselnet.com/standards. [Accessed 16 August 2013].
[18] M. Weiss, P. Bonnel, R. Hummel, A. Provenza and U. Manfredi, "On-Road Emissions of light-
Duty Vehicles in Europe," Environmental - Science & Technology, no. 45, p. 8575–8581, 2011.
[19] M. Weiss, P. Bonnel, J. Kuhlwein and others, "Will Euro 6 reduce the NOx emissions of new
diesel cars? - Insights from on-road tests with Portable Emissions Measurement Systems
(PEMS)," Atmospheric Environment - Elsevier, no. 62, pp. 657-665, 2012.
[20] N. Pereira, "Eficiência energética no sector dos transportes rodoviários: Metodologia para
quantificação do excesso de energia consumido devido ao factor comportamental na condução
de veículos automóveis ligeiros," Master Thesis in Bioenergy Engineering, FCT-UNL, Lisbon,
2011.
[21] D. Carslaw, M. Williams, J. Tate and S. Beevers, "The importance of high vehicle power for
passenger car emissions," Atmospheric Environment - ELSEVIER, vol. Atmospheric Environment,
no. 68, pp. 8-16, Nov, 2012.
[22] B. Georgi, "SAE978626," 1997.
[23] J. B. Heywood, Internal Combustion Engines Fundamentals, McGraw-Hill, Inc., 1988.
[24] "Cars, Trucks and Air Pollution - Union of Concerned Scientists," Union of Concerned Scientists,
9 March 2013. [Online]. Available: http://www.ucsusa.org/clean_vehicles/why-clean-cars/air-
pollution-and-health/cars-trucks-air-pollution.html. [Accessed 12 July 2013].
[25] L. Tiezhu, C. Xudong and Y. Zhenxing, "Comparison of fine particles emissions of light-duty
gasoline vehicles from chassis dynamometer tests and on-road measurements," Atmospheric
Evironment - ELSEVIER, no. 68, pp. 82-91, 2013.
[26] I. Caplain, F. Cazier, H. Nouali, A. Mercier, J.-C. Dechaux, V. Nollet, R. Jourmard, J.-M. Andre and
R. Vidon, "Emissions of unregulated pollutants from European gasoline and diesel passenger
cars," Atmospheric Environment - ELSEVIER, vol. 40, no. 31, pp. 5954-5966, 2006.
[27] "Acceleration Test: Debunking the Myths," efficient mileage .com, [Online]. Available:
82
http://www.efficient-mileage.com/acceleration-test.html. [Accessed 21 September 2013].
[28] P. Nash, "The Oil Drum | Drumbeat: August 3, 2011," 9 August 2011. [Online]. Available:
http://www.theoildrum.com/node/8223. [Accessed 30 September 2013].
Links to Bibliography (where available/applicable):
[1] - http://www.worldometers.info/cars/
[4] - http://ec.europa.eu/clima/policies/transport/vehicles/cars/index_en.htm
[9] - http://www.sciencedirect.com/science/article/pii/S1361920906000253
[10] - http://publications.jrc.ec.europa.eu/repository/handle/111111111/27598
[12] - https://www2.unece.org/wiki/pages/viewpage.action?pageId=2523179
[13] - http://www.epa.gov/otaq/models/moves/documents/420r11011.pdf
[14] - http://www.sciencedirect.com/science/article/pii/S0048969713005585
[15] - http://pubs.acs.org/doi/abs/10.1021/es702493v
[16] - http://www.sciencedirect.com/science/article/pii/S1361920901000037
[17] - http://www.dieselnet.com/standards
[19] - http://www.sciencedirect.com/science/article/pii/S1352231012008412
[24] - http://www.ucsusa.org/clean_vehicles/why-clean-cars/air-pollution-and-health/cars-trucks-air-
pollution.html
[25] - http://www.sciencedirect.com/science/article/pii/S135223101201093X
[26] - http://www.sciencedirect.com/science/article/pii/S1352231006000525
[18] - http://pubs.acs.org/doi/abs/10.1021/es2008424
[27] - http://www.efficient-mileage.com/acceleration-test.html
[28] - http://www.theoildrum.com/node/8223
i
ANNEX
The “blocks” not represented in 4.2.2 are presented below:
0.0
50.0
100.0
150.0
1
14
27
40
53
66
79
92
10
5
11
8
13
1
14
4
15
7
ARB02 P2
0.0
20.0
40.0
60.0
1
21
41
61
81
10
1
12
1
14
1
16
1
18
1
20
1
22
1
24
1
26
1
28
1
Advisor_Comp.
0.00
50.00
100.00
150.00
1
90
17
9
26
8
35
7
44
6
53
5
62
4
71
3
80
2
89
1
98
0
10
69
Highway
0.0
50.0
100.0
150.01
52
10
3
15
4
20
5
25
6
30
7
35
8
40
9
46
0
51
1
56
2
61
3
66
4
71
5
HWFET
0.0
50.0
100.0
150.0
15
61
11
16
62
21
27
63
31
38
64
41
49
65
51
60
66
61
71
6
HWFET
0.0
20.0
40.0
60.0
80.0
1
19
37
55
73
91
10
9
12
7
14
5
16
3
18
1
19
9
21
7
23
5
25
3
Indian Urban 2
0
50
100
150
1
20
39
58
77
96
11
5
13
4
15
3
17
2
19
1
21
0
22
9
24
8
LA92 - 1
0
50
100
1
12
23
34
45
56
67
78
89
10
0
11
1
12
2
13
3
14
4
15
5
LA92 - 2
0
50
100
150
1
14
27
40
53
66
79
92
10
5
11
8
13
1
14
4
15
7
17
0
LA92 - 3
0
50
100
1
13
25
37
49
61
73
85
97
10
9
12
1
13
3
14
5
15
7
16
9
LA92 - 4
ii
Following are some examples of results achieved with different test cycles generations:
0
10
20
30
40
50
60
70
80
90
1001
18
35
52
69
86
10
3
12
0
13
7
15
4
17
1
18
8
20
5
22
2
23
9
25
6
27
3
29
0
30
7
32
4
34
1
35
8
37
5
39
2
40
9
42
6
44
3
46
0
47
7
49
4
51
1
52
8
54
5
56
2
57
9
59
6
SC03-1
0
20
40
60
80
1 7 13 19 25 31 37 43 49
SC03-2
0
50
100
1 9 17 25 33 41 49 57 65
SC03-3
0
20
40
60
80
1
14
27
40
53
66
79
92
10
5
11
8
13
1
14
4
SC03-4
iii
iv
v
Code of the Generate1() function (with Portuguese commentary):
(the other functions, as well as the Test Cycle Generating Tool can be found and consulted on the
data disc)
% clearvars -EXCEPT eixo_ciclo eixo_vsp;
%este programa vai gerar um ciclo de Teste %usa como entrada um ficheiro que contem velocidade media e distribuicao %VSP
%no futuro й para aperfeiзoar de modo que de para pedir qq tipo ficheiro e %formatacao, com GUI e atй um executavel standalone
global nome_fich; nome_fich=uigetfile('*.xls','Escolhe o ficheiro com distribuicao VSP e vel med'); %ou
[nome_fich] dados = xlsread(nome_fich,1);
%atribui o valor da vel media e o vector distrib. VSP...
vel_med=dados(1); vsp_dado=dados(2:15);
%le o ficheiro que contem os "blocos" para construir o ciclo de conducao
if vel_med<30 blocos=xlsread('blocks.xlsx',3); elseif vel_med>=30 || vel_med<=65 blocos=xlsread('blocks.xlsx'); elseif vel_med >65 blocos=xlsread('blocks.xlsx',2); end [L,C]=size(blocos);
% cria uma celula que contem os blocos para construir o ciclo e tambem vai % conter ja os modos vsp calculados pela funcao vsp...
bloc_vsp=cell(2,C);
for i=1:C aux=blocos(:,i); aux(isnan(aux))=[]; bloc_vsp{1,i}=aux; bloc_vsp{2,i}=vsp(aux); clear aux; end clear i;
%prealocar espaco e inicializar o vector "ciclo de teste" que ira %representar o vsp desejado (de dados) ciclo=zeros(3,1);
%um ciclo que vai sempre alterar e modificar o vector "ciclo" ate que seja %suficiente prуximo do vsp desejado. As modificacoes sao: acrescentar novo %vector, amplificar, reduzir, estender ou compactar um bloco, sobrepor %blocos, tudo de modo que o vsp do "ciclo" esteja o mais prуximo do desejado
xxx=0.9; z=100;
vi
y2=105; y3=110; y5=120;
k=1; flag=0;
ciclo_novo=zeros(3,1); indice=[]; % indice dos sitios no vector "ciclo" onde se pode acrescentar tempo parado
global peso;
blo_cic_iniciado=0; blo_cic={'iniciado'}; bloco_alt={};
global duracao_ciclo;
if duracao_ciclo<3000 k_limit=0.08; elseif duracao_ciclo>=3000 k_limit=0.06; end
while flag==0 %&& size(ciclo,1)<duracao_ciclo*3
for a=1:C % percorre todos os blocos existentes em blocos.xls bloco_alt{1}=bloc_vsp{1,a}; bloco_alt{2}=bloc_vsp{1,a}*xxx; bloco_alt{3}=resample(bloc_vsp{1,a},y2,z,0); bloco_alt{4}=resample(bloc_vsp{1,a},y3,z,0); bloco_alt{5}=resample(bloc_vsp{1,a},y5,z,0);
for b=1:size(bloco_alt,2) %percorre cada alternativa do "bloco_alt" criada
sum_blo=0; dur_blo=0;
blo_cic=[blo_cic, bloco_alt(b)]; % uma celula "blo_cic" que vai armazenar
o bloco_alt mais conveniente
if size(blo_cic,2)>=2 && blo_cic_iniciado==0; %apaga a inicializacao do
blo_cic blo_cic=blo_cic(2:end); blo_cic_iniciado=1; end
for c=1:size(blo_cic,2) % calule a soma e a duracao total dos blocos ate
agora escolhidos sum_blo=sum_blo+mean(blo_cic{c})*size(blo_cic{c},1); dur_blo=dur_blo+size(blo_cic{c},1); end
mean_blo=sum_blo/dur_blo;
if mean_blo>vel_med
aux=(dur_blo*(mean_blo/vel_med)-dur_blo)/size(blo_cic,2); aux=round(aux);
vii
ciclo=[]; for d=1:size(blo_cic,2) ciclo=[ciclo;zeros(aux,1);blo_cic{d}]; end
elseif mean_blo<=vel_med
ciclo=[]; for z=1:size(blo_cic,2) ciclo=[ciclo;blo_cic{z}]; end
end
if mean(abs(vsp(ciclo)-vsp_dado).^2.*peso)< mean(abs(vsp(ciclo_novo)-
vsp_dado).^2.*peso) ciclo_novo=ciclo; elseif mean(abs(vsp(ciclo)-vsp_dado).^2.*peso)>= mean(abs(vsp(ciclo_novo)-
vsp_dado).^2.*peso) blo_cic=blo_cic(1:end-1); end
end % end do ciclo dos blocos ajustados
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% agora removemos blocos que foram escolhidos, mas
era melhor que nao%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
end % end do ciclo dos blocos da base de dados
blo_cic_aux=[]; blo_cic_aux2=[];
if size(blo_cic,2)>4
for e=size(blo_cic,2):-1:1 % verifique se retirando alguns blocos previamente
escolhidos melhora o ciclo
if e==1 blo_cic_aux=blo_cic(2:end); elseif e==size(blo_cic,2) blo_cic_aux=blo_cic(1:end-1); elseif e~=1 && e~=size(blo_cic,2) blo_cic_aux=[blo_cic(1:e-1),blo_cic(e+1:end)]; end
sum_blo=0; dur_blo=0;
for f=1:size(blo_cic_aux,2) sum_blo=sum_blo+mean(blo_cic_aux{f})*size(blo_cic_aux{f},1); dur_blo=dur_blo+size(blo_cic_aux{f},1); end
mean_blo=sum_blo/dur_blo;
if mean_blo>vel_med
viii
aux=(dur_blo*(mean_blo/vel_med)-dur_blo)/size(blo_cic_aux,2); aux=round(aux);
ciclo=[]; for g=1:size(blo_cic_aux,2) ciclo=[ciclo;zeros(aux,1);blo_cic_aux{g}]; end elseif mean_blo<=vel_med ciclo=[]; for h=1:size(blo_cic_aux,2) ciclo=[ciclo;blo_cic_aux{h}]; end end
if mean(abs(vsp(ciclo)-vsp_dado).^2.*peso)< 0.95*mean(abs(vsp(ciclo_novo)-
vsp_dado).^2.*peso) ciclo_novo=ciclo; tmnh=size(blo_cic_aux); if tmnh~=0 blo_cic_aux2=blo_cic_aux; end end end
end
if size(blo_cic_aux2,2)~=0 blo_cic=blo_cic_aux2; end
blo_cic_aux=[]; blo_cic_aux2=[];
for m=1:C % verifique se substituindo alguns blocos previamente escolhidos melhora o
ciclo
for n=size(blo_cic,2):-1:1
if n==1 blo_cic_aux=[bloc_vsp{1,m},blo_cic(2:end)]; elseif n==size(blo_cic,2) blo_cic_aux=[blo_cic(1:end-1),bloc_vsp{1,m}]; elseif n~=1 && n~=size(blo_cic,2) blo_cic_aux=[blo_cic(1:n-1),bloc_vsp{1,m},blo_cic(n+1:end)]; end
sum_blo=0; dur_blo=0;
for f=1:size(blo_cic_aux,2) sum_blo=sum_blo+mean(blo_cic_aux{f})*size(blo_cic_aux{f},1); dur_blo=dur_blo+size(blo_cic_aux{f},1); end
mean_blo=sum_blo/dur_blo;
if mean_blo>vel_med
aux=(dur_blo*(mean_blo/vel_med)-dur_blo)/size(blo_cic_aux,2); aux=round(aux);
ciclo=[]; for g=1:size(blo_cic_aux,2) ciclo=[ciclo;zeros(aux,1);blo_cic_aux{g}]; end elseif mean_blo<=vel_med ciclo=[];
ix
for h=1:size(blo_cic_aux,2) ciclo=[ciclo;blo_cic_aux{h}]; end end
if mean(abs(vsp(ciclo)-vsp_dado).^2.*peso)< 0.97*mean(abs(vsp(ciclo_novo)-
vsp_dado).^2.*peso) ciclo_novo=ciclo; tmnh=size(blo_cic_aux); if tmnh~=0 blo_cic_aux2=blo_cic_aux; end end end
if size(blo_cic_aux2,2)~=0 blo_cic=blo_cic_aux2; end end
k=k+1; if sum(abs(vsp(ciclo)-vsp_dado))<k_limit && k>1 || k>=15 flag=1; end
end %end do "ciclo repetitivo" que ira ajudar para convergir o vsp
ciclo=ciclo_novo; blo_cic_aux2=blo_cic;
while size(ciclo,1)>duracao_ciclo*1.05
blo_cic_aux=[]; ciclo_novo=ones(10,1); blo_cic=blo_cic_aux2;
for m=size(blo_cic,2):-1:1 % retire alguns blocos previamente escolhidos para satisfazer
duracao total ciclo
if size(ciclo,1)>duracao_ciclo;
if m==1 blo_cic_aux=blo_cic(2:end); elseif m==size(blo_cic,2) blo_cic_aux=blo_cic(1:end-1); elseif m~=1 && m~=size(blo_cic,2) blo_cic_aux=[blo_cic(1:m-1),blo_cic(m+1:end)]; end
sum_blo=0; dur_blo=0;
for o=1:size(blo_cic_aux,2) sum_blo=sum_blo+mean(blo_cic_aux{o})*size(blo_cic_aux{o},1); dur_blo=dur_blo+size(blo_cic_aux{o},1); end
mean_blo=sum_blo/dur_blo;
if mean_blo>vel_med
aux=(dur_blo*(mean_blo/vel_med)-dur_blo)/size(blo_cic_aux,2);
x
aux=round(aux);
ciclo=[]; for g=1:size(blo_cic_aux,2) ciclo=[ciclo;zeros(aux,1);blo_cic_aux{g}]; end
elseif mean_blo<=vel_med ciclo=[]; for h=1:size(blo_cic_aux,2) ciclo=[ciclo;blo_cic_aux{h}]; end end
if mean(abs(vsp(ciclo)-vsp_dado).^2.*peso)< mean(abs(vsp(ciclo_novo)-
vsp_dado).^2.*peso) ciclo_novo=ciclo; tmnh=size(blo_cic_aux); if tmnh~=0 blo_cic_aux2=blo_cic_aux; end end end ciclo=ciclo_novo; end
end
ciclo=ciclo_novo; global ciclo_gerado; ciclo_gerado=ciclo;
%clearvars -EXCEPT ciclo vel_med vsp_dado;
%NOTAS: nao esqueces verificar semrpre a aceleracao maxima do ciclo, para %nao ultrapassar 1.5m/s2. verifica no fim outra vez vel media, e o desvio %do vsp desejado...
assignin('base','ciclo',ciclo);
The remaining matlab functions() used by TCGT are included on the data disk, and can be thoroughly
consulted there.
All the digital content, as well as the program and more screen captures of the TCGT results can be
checked on the Data Disc Attached or on:
https://www.dropbox.com/sh/umd9146tbsialhn/yHyw0nKYPf
https://drive.google.com/folderview?id=0B0lBkanelHAQZHhNWUdkSHRoQXc&usp=sharing