location of formation (engine-out or catalyst-out) are reconstructed from the .....
For more than a decade attempts have been made to store or map emission mea
...
Research Collection
Doctoral Thesis
Modal pollutant emissions model of diesel and gasoline engines Author(s): Ajtay, Delia Elisabeta Publication Date: 2005 Permanent Link: https://doi.org/10.3929/ethz-a-005163854
Rights / License: In Copyright - Non-Commercial Use Permitted
This page was generated automatically upon download from the ETH Zurich Research Collection. For more information please consult the Terms of use.
ETH Library
Diss. ETH No. 16302
Modal Pollutant Emissions Model of Diesel and Gasoline Engines
A dissertation submitted to the
SWISS FEDERAL INSTITUTE OF TECHNOLOGY ZURICH
for the
degree
of
Doctor of Technical Sciences
presented by Delia Elisabeta
Ajtay
M.Sc. Mathematics Born
11th August
Nationality
accepted
on
1975
Romanian
the recommendation of
Prof. Dr. Lino Guzella, examiner Prof. Dr. Stefan
Hausberger,
Dr. Martin Weilenmann
2005
,
co-examiner
co-examiner
Abstract
Road traffic accounts for
important part of air pollution. For roughly 30 years, emission limits have been enforced by legislation in Europe and else¬ where. The success of the stringent regulations has been monitored by the environmental agencies. Although in the early days the interest was in the ful¬ fillment of the legislation limits, real-world emissions represent today's main an
focus. Emissions models
are
used to derive
international, national and regional emis¬
sion inventories to
predict
the
using measurements performed in emissions laboratories and impact of different traffic related measures. These emission mod¬
els connect the
driving behaviour and fleet data to the test bench investigations.
The
availability
of vehicle emission models has
recent years. There are
models:
one
emission measurements from
based
on
basically
bag
two
improved significantly in the types of emissions and fuel consumption
measurements and the other based
measurements. Emission models based
fic situations similar to the
on
bag
values
give
on
instantaneous
results for the traf¬
bag. If the driving behaviour changes, new measurements with comparable driving patterns have to be per¬ formed. To account for the additional effects as load, slope or gearshift strate¬ gies, bag based models include correction functions. However, these correc¬ tion functions
based
are
which may not be
one
on a
used to fill the
small number of measurements with few vehicles
representative
for the emissions behaviour.
combination of these correction factors (i.e. when a
full
load)
can
be
Moreover, the
uphill
with
given time to speed, engine speed, torque, etc.
their
a
vehicle drives
extremely misleading.
Instantaneous emissions
modelling
maps the emissions at
a
generating "engine state", like vehicle makes it possible to integrate new, unmeasured driving patterns
over
and calculate their emission factors without further measurements.
sion factors for
a
large
number of
driving
situations
can
This
the model
Thus, emis¬
be determined from
a
small number of measurements.
l
Abstract
goal of this thesis is sumption model. Such a
The
to
develop
an
instantaneous emissions and fuel
model should be
capable
to
predict
con¬
emission factors
for any unmeasured
limited number of measurements.
Beside
road
speed diagram using a that, contributory aspects like load,
gradient, gearshift strategies be, thus, significantly more flexi¬ ble than the existing approaches and would be especially useful for the assess¬ ment of local studies (i.e. the impact of traffic management schemes, change of driving behaviour, etc). should be also included. Such
a
model shall
Due to the fact that the instantaneous emission model relates at each moment
of time the emission
signals
of the measurements is
original
emission
one
signals
to their
of the
generating engine variables,
key
measured in
a
issues for test
If these
peaks
are
flattened
by
successful result. But, the
delayed to their time of forma¬ the engine to the analysers, and
are
tion, since the exhaust gas is transported from the emission
a
the accuracy
convolution.
neglected, the emission events are correlated to the wrong second, resulting in incorrect engine status in emissions modelling. For instantaneous emissions modelling, emission values can be correlated to the correct engine state of the car only if they are at their right location on the time scale. Therefore, these delays and mixing dynamics must be compensated, i.e. the behaviour of the gas transport systems must be modelled and inverted. The modelling of the different gas transport systems is performed using linear time-varying approaches, such that emissions at their location of formation (engine-out or catalyst-out) are reconstructed from the signals recorded at the analyser. Using
dynamic aspects
of the exhaust transport
the reconstructed emissions
data,
an
are
instantaneous emissions model is
pollutants and for various categories of ve¬ hicles. The model performs reliably, emission factors from several real-world driving situations being accurately forecasted. developed
for different classes of
For the modern
gasoline cars equipped with a three-way catalyst, the approach has to be extended by modelling separately the engine-out emissions and, af¬ terwards, the catalyst-out emissions. In order to take into account the transient generation of exhaust gases, the engine-out emissions are modelled using a 4D emissions model. With the engine-out emissions as input data, a simple catalyst submodel is developed, based on the oxygen storage mechanism. The validity of the model and the parameters estimation procedure is checked by applying them to real world case studies. It is demonstrated that the model is capable of predicting the operating behaviour of the catalyst under realistic conditions and is, thus, suited for use within emissions modelling.
11
Zusammenfassung
Der Strassenverkehr ist eine der
sionsgrenzen
setzgebung
Hauptquellen der Luftverschmutzung.
Die Emis¬
Motorfahrzeugen wurden seit etwa 30 Jahren durch die Ge¬ Europa stufenweise verschärft. Der Erfolg dieser Regulierungs¬
von
in
chritte wurde seither durch Umweltbehörden kontrolliert. Während das Inter¬ esse
die
früher mehr der
Ermittlung
gesetzlichen Vorschriften galt, Emissionen im Mittelpunkt.
Erfüllung
der realen
der
stehen heute
Emissionsmodelle, welche auf Messungen an Rollenprüfständen basieren, wer¬ den benötigt um internationale, nationale und regionale Emissionsbestandsauf¬ nahmen abzuleiten. Ausserdem werden diese Modelle genutzt,
um
den Ein-
fluss unterschiedlicher
Verkehrsparameter (z.B. Geschwindigkeitslimiten) auf die Emissionen vorhersagen zu können. In diesen Emissionsmodellen werden Daten zu Fahrverhalten und Fahrzeugbeständen mit den Emissionsmessungen von Rollenprüfstandsuntersuchungen gefaltet. Seit den letzten Jahren stehen mehr und mehr
Verfügung.
Es
zwei Arten
gibt grundsätzlich
verbrauchsmodellen: der eine
Typ
basiert auf
Fahrzeugemissionsmodelle von
zeugen
nur
Emissions- und Kraftstoff¬
Sackmessungen
auf "online"
zur
Messungen. Emissionsmodelle, die auf Ergebnisse für jene Verkehrssituationen,
und der andere
Sackwerten
die
zum
basieren,
er¬
Füllen der Ab¬
gassäcke genutzt wurden. Wenn sich das Fahrverhalten ändert, müssten neue Messungen mit entsprechenden Fahrmustern durchgeführt werden. Modelle, welche auf Sackmessungen beruhen, beinhalten teilweise Korrekturfunktionen, mit denen
Zuladung, Steigung und Schaltstrate¬ gie zu berücksichtigen versucht. Allerdings basieren diese Korrekturen meist auf einer geringen Anzahl Messungen mit wenigen Fahrzeugen, welche nicht repräsentativ für das Emissionsverhalten der Flotte sein könnten. Ausserdem man
zusätzliche Effekte wie
kann die Kombination solcher einzeln ermittelter Korrekturen extrem irrefüh¬ rend sein. Wie soll z.B. eine
Steigungsfahrten Die "online"
Bergfahrt
und Fahrten mit
mit
Zuladung
Zuladung separat
Emissionsmodellierung bezieht die
berechnet
weden,
wenn
gemessen wurden?
Emissionen
zu
einem bestimm-
111
Zusammenfassung Zeitpunkt auf den aktuellen Motorzustand: Fahrzeuggeschwindigkeit, Mo¬ tordrehzahl, Drehmoment, etc. Dies ermöglicht es, neue, nicht gemessene Fahr¬ ten
muster in das Modell
Messungen
zu
zu
integrieren
und deren Emissionsfaktoren ohne weitere
berechnen. Dadurch können für eine grosse Anzahl
situationen Emissionsfaktoren
aus
einer kleinen Anzahl
von
von
Fahr¬
Messungen
be¬
stimmt werden. Das Ziel dieser Doktorarbeit ist es, ein "online" Emissions- und Kraftstoff¬
verbrauchsmodell
entwickeln. Ein solches Modell sollte
geeignet sein für für jeden beliebigen,
zu
Einzelfahrzeuge und Fahrzeugklassen Emissionsfaktoren ungemessenen Geschwindigkeitsverlauf vorherzusagen. Zudem sollten Aspek¬ te wie Ladung, Strassengefälle und Schaltsstrategien korrekt mit einbezogen werden. Ein solches Modell wird dadurch flexibler als vorhandene Ansätze und ist besonders nützlich für die
Bestimmung
von
Verkehrslenkungsmassnahmen, änderungen
von
(d.h. Einfluss Fahrverhalten, etc.).
lokalen Studien in
Aufgrund der Tatsache, dass die Emissionssignale in den "online" Modellen für jeden Zeitpunkt dem Zustand des Motors zugeordnet werden müssen, ist die zeitliche Exaktheit der Messung einer der wichtigsten Aspekte für erfolg¬ reiche Ergebnisse. Die direkten gemessenen Emissionssignale treffen jedoch auf Grund des Transports durch das Auspuffsystem und die Messleitung ver¬ spätet im Analysator ein. Zudem werden die Emissionsereignisse durch den turbulenten
Transport "verschmiert".
Wenn diese
dynamischen Aspekte des Abgastransportes vernachlässigt wer¬ den, werden die Emissionsereignisse mit zeitlich verschobenen Motorzustän¬ den korreliert. Für "online" Emissionsmodelle können Emissionswerte
nur zum
richtigen Motorenzustand korreliert werden, wenn sie sich auf den richtigen Zeitpunkt beziehen. Daher müssen diese Verschiebungen kompensiert werden, d. h. das Verhalten des den. Die nem
Modellierung
Gastransportsystems
muss
der unterschiedlichen
linearen zeit-variabeln Ansatz
modelliert und invertiert
Gastransportsysteme
durchgeführt,
so
wer¬
wird mit ei¬
dass die Emissionen
am
Ort der aus
Entstehung (Austritt aus dem Motor oder Austritt aus dem Katalysator) dem vom Analysator aufgezeichneten Signal rekonstruiert werden können.
In der
Folge
wird
aus
den zeitlich
modell für verschiedene
korrigierten Emissionsignalen ein Emissions¬ Schadstoffklassen und Fahrzeugkategorien entwickelt.
Es basiert auf der Korrelation der Emissionswerte mit Drehzahl und Drehmo¬ ment des Motors. Das Modell funktioniert für
gewisse Fahrzeugklassen
zuver¬
lässig:
die Emissionsfaktoren werden für ann"hrend alle realen Fahrsituationen
genau
vorhergesagt.
IV
Zusammenfass ung
Dieser Ansatz musste
jedoch für moderne Benzinfahrzeuge, ausgestattet mit einem Drei-Wege-Katalysator, ausgeweitet werden. Die Emissionen aus dem Motor und das Verhältnis des Katalysators werden separat modelliert. Die Emis¬ sionen liert zu
aus
dem Motor mussten mit einem vier-dimensionalen Kennfeld model¬
werden,
um
die transiente
Abgasproduktion zu berücksichtigen.
Zusätzlich
Drehzahl und Drehmoment werden die Emissionen auf die zeitliche Ab¬
leitung
des
Saugrohrdrucks bezogen.
de ein Teilmodell
Für das Verhalten des
Katalysators
wur¬
entwickelt, für welches die "engine-out" Emissionsdaten als
"Inputdaten" genutzt wurden. Es basiert auf dem Sauerstoffablagerungsmecha¬ nismus. Die Genauigkeit des Modells und die Parametrierung wurden über¬
verglichen wurden. Es wird gezeigt, dass das Model geeignet ist, die Wirkung des Katalysators unter realistischen Bedingungen vorherzusagen und das es folglich für Emissionsmodellierungen
prüft,
indem diese mit realen Fallstudien
verwendbar ist.
v
Zusammenfassung
Seite Leer / Blank leaf
Contents
1.
Introduction
1
The
1
1.3.
pollutants and the environment Legislation Real-world driving cycles
1.4.
Overview of the emission models
10
1.4.1.
Average speed
11
1.4.2.
Traffic situation models
14
1.4.3.
Instantaneous emission models
16
1.1. 1.2.
2.
1.5.
Scope
1.6.
Contribution
Modelling
8
models
of the thesis
19
21
of the exhaust gas
2.1.
Introduction
2.2.
Methodology
transport systems
23 23
of the model
27
2.2.1.
Basic model
27
2.2.2.
Evaluation of the exhaust volume flow
28
2.2.3.
Raw gas
2.2.4.
Exhaust system of the
2.2.5.
Dilution system model
38
2.2.6.
Overall validation
41
system model
30 34
car
of the inversion
43
2.3.1.
Basic inversion model
43
2.3.2.
Inversion of the
2.3.3.
Inversion of the exhaust system of the
2.3.4.
Inversion of the dilution
2.3.5.
Overall inversion
47
Static instantaneous emission model
49
2.3.
3.
5
Methodology
raw
gas
analyzer system car
analyser system
44 45 46
3.1.
Introduction
49
3.2.
Methodology
52
3.2.1.
Measurement
3.2.2.
Model
procedure
development
52 53
vii
Contents
4.
3.3.
Validation
60
3.4.
Conclusions
65
Dynamic
5.
67
4.1.
Introduction
4.2.
Dynamic engine
4.3.
Validation for different
4.4.
4.3.1.
Diesel
4.3.2.
Gasoline
68
model
loads, slopes and gear-shift strategies
.
72 75
case
77
case
78
Conclusions
81
Dynamic catalyst model 5.1.
Introduction
81
5.2.
Methodology
84
5.2.1.
Mathematical model
84
5.2.2.
Parameter estimation
87
5.2.3.
Static conversion
88
curves
5.3.
Validation
92
5.4.
Conclusions
95
6.
Conclusions and Outlook
A.
Appendix
vin
67
instantaneous emission model
97
101
A.l.
Kinematic characteristics of the real-world
A.2.
Set-up for the measurements tems modelling
driving patterns
necessary for the
.
.
101
transport sys¬ 103
1. Introduction
become ticles
decades, the environmental effect of burning fossil fuel has important issue. Smog, greenhouse effect, acid rain and toxic par¬
the past
During
an
are
increasing traffic, heating and industrial thermal stricter legislation, fuel consumption and vehicle emissions
consequences of the
processes. Under
of vehicle fleet and of average per distance have been reduced, but the increase distance travelled have counteracted these measures. Due to the
quantity
of information necessary to determine the different para¬
emissions, direct measurement becomes impractical and expensive. Therefore, models for predicting emissions, although difficult
meters
to
related
to
the traffic
develop, represent
impact
alternative to direct measurement.
pollutants
1.1. The
The
an
of traffic air
and the environment
pollution
of undesirable material in the
ence
cause
harmful
transport is
effects, both of the
one
environment and to the human health. Road
to the
major
high. Air pollution represents the pres¬ air, in quantities which are large enough to
is
sources
of air
pollution.
Since human
population
of emissions from road transport, vehicle emissions con¬ tribute to the personal exposure even more than expected from their share on is close to the
sources
total emissions
([40], [77]).
spark ignited gasoline engine. Gasoline mix¬ ture represents a blend of paraffins and aromatic hydrocarbons which combust with air at a very high efficiency. The simplified combustion reaction is: The
majority
of the vehicles
gasoline
Carbon dioxide
(CO2)
+
use a
02{in air)
—»
CO2
+
H2O
+ heat
(1.1)
(H2O) are the desired products of the fuel imperfect combustion process, the following
and water
combustion. However, due to the
1
1. Introduction
undesired
compounds
result
•
carbon monoxide
•
unburned
•
nitrogen
•
hydrogen (H2,
•
carbon dioxide
•
water
•
oxygen
as
(CO,
exhaust
at the range
hydrocarbons (HC,
oxides
(H20,
(02,
(NOx,
components1 ([24]):
at
of 0.1-6 vol. %);
the range of 500-5000
at the range of 100-4000
at the range of 0.17 vol.
(C02,
at the range of
ppm);
%);
10-13.5 vol. %);
at the range of 10-12 vol. at
ppm);
%);
the range of 0.2-2 vol. %).
gasoline vehicles, diesel cars are increasingly used due to the econ¬ In omy of operation and decrease of greenhouse gases, especially of C02. this case, the fuel is injected into a highly compressed charge of air where the temperature is high enough for the combustion to occur. Thus, diesel engines are based on a compression-ignited process. Due to the nature of this combus¬ tion process, some quantities of unburned fuel, lubricating oil emissions and Beside the
large C02
numbers of and
dry
soot
particles
result. Beside the desired components
H20, the exhaust emissions of diesel engines consist of:
particulate
•
solid exhaust: soot
•
gaseous exhaust: carbon monoxide
of
•
02,
matter
(PM);
(CO), hydrocarbons (HC) and oxides
nitrogen (NOx);
liquid exhaust: soluble organic cating oil) and liquid sulfates.
fraction
(SOF:
unburned fuel and lubri¬
Carbon dioxide
present in the fuel will be eventually transformed to carbon dioxide in the atmosphere. Even if, due to incomplete combustion, carbon All the carbon
exhaust, it is ultimately oxidized in the atmosphere to form C02. Carbon dioxide is a major source to the greenhouse effect, which monoxide may result leads
'N2
2
finally is
a
to
as
global warming. C02 production
remainder
is
an
invariable consequence
1.1.
of
as
pollutants
fossil fuel and, at least, the process of fuel
burning
efficient
The
and the environment
burning
should be
as
possible.
Carbon monoxide The main
atmosphere engines, especially of gasoline
source
combustion
for CO in the
from the exhaust of internal
vehicles
[76]. Carbon monoxide
is
poorly soluble in water. In the human body, carbon monoxide binds with haemoglobin to form carboxyhaemoglobin (COHb), causing a reduction in the oxygen carrying capacity of the blood. This determines headache, dizziness or nausea and, at high level of a
colourless, inodorous and tasteless
comes
gas that is very
COHb, becomes lethal [39]. Carbon monoxide is
produced mainly during rich combustion situations, when there is insufficient oxygen to burn completely all the hydrocarbons from the fuel into C02. Some of these rich situations appear during transient engine operations like acceleration or high torque demand. Also, when the engine is cold, it is necessary to enrich the air/fuel mixture, causing high levels of CO until the engine is warmed-up. In microenvironments in which combustion engines are used under conditions of insufficient ventilation, like underground car parks or road tunnels, the mean levels of carbon monoxide can rise to values much higher than those from the ambient outdoor air, becoming thus extremely dangerous for the human health and hence, relevant for the ventilation design. Hydrocarbons In the vehicle exhaust there is
pounds.
The most
large variety of unburned hydrocarbon com¬ important are paraffins, olefins, acetylenes and aromatics. a
As for
CO, hydrocarbons are caused by the lack of oxygen when the air/fuel mixture is rich. Beside that, other reasons for hydrocarbon emissions are: flame
quenching at the walls, filling of crevices with unburned mixture, absorption by oil layers, incomplete combustion (partial burning or misfire), bulk quenching and evaporative emissions [38]. Due to their
and
variety,
the
hydrocarbons
have different
impacts
on
human health
example, can lead to leukaemia and it is carcinogenic to humans [68]. In the troposphere, hydrocarbons react with ni¬ trogen dioxide (N02), forming ozone and photochemical smog. Ozone causes cough, throat irritations, pain on deep breath, chest tightness and, sometimes, headache and nausea [36]. Additionally, ozone determines the damaging of vegetation. on
the environment. Benzene, for
3
1. Introduction
Oxides of nitrogen
nitrogen are either direct products of the combustion in engines, like nitric oxide (NO) and nitrogen dioxide (N02), either a product of the catalytic converter like nitrous oxide (N20). The first two species are collectively de¬ noted as NOx. They are produced during combustion when oxygen reacts with nitrogen due to a high combustion temperature (> 1500°C). Nitric oxide (NO) is colorless, inodorous, tasteless and relatively non-toxic for humans. Similar to CO, nitric oxide is eventually oxidized in the atmosphere Oxides of
to
form
N02.
Nitrogen
dioxide
(N02)
is reddish-brown in
colour, extremely toxic and has a human pulmonary functions, caus¬
harsh odor.
Nitrogen dioxide has effects on ing damages of lung tissue, couching, bronchitis, etc. Beside the health effects, N02 is extremely important to monitor because: (a) it is also an absorber of visible radiation which could have a direct role on the global climate change if its concentration were to become too high; (b) it is a key factor in the formation of ozone in the troposphere and (c) it is, along with atmospheric sulfur oxides, responsible for acid rains [60]. Particulate matter Airborne
particulate
matter
(PM) represent
a
combination of
organic and inor¬ catalyst, diesel cars and
ganic substances. Gasoline vehicles with or without heavy-duty trucks, all emit particles mainly in the range of 0.1-0.2 fim in di¬ ameter. Gasoline cars equipped with three way catalytic converters emit much lower particle masses than those without, while diesel cars emit about 100 to 1000 times the particle mass of a gasoline car equipped with a catalytic con¬ verter.
Diesel
particulate matter is almost pure carbon and exists as a sub-aggregate of ultra-fine carbon spheroids with aerodynamic diameters of around 0.1 jum [69]. Apart from the presence of this unburned carbon in the exhaust, which is a con¬ sequence of incomplete combustion and implies therefore lower efficiency, the particulate matter may cause lung diseases. Significant relationships between particulate air pollution and human health have been found by epidemiological studies [64]. The particulates are usually denoted by PM2.5, which represents "particulate matter of size less than 2.5 /zm". However, the subject of particulate matter measurements and modelling will not be discussed in this paper, but excellent research on this topic can be found in
4
[53], [52], [54].
1.2.
Legislation
Sulfur oxides This sulfur
Both, gasoline and diesel fuels, contain sulfur in different
amounts.
is oxidized
(S02). Oxidation sulfuric acids, which
during
combustion and
produces
sulfur dioxide
of sulfur dioxide leads to the formation of sulfurous and
deposited to the earth by rain. This is called "acid rain" and has caused deforestation in Europe and North America and serious damages to buildings. Current regulation on sulphur oxide emissions are very strict and are presently can
be
fulfilled.
1.2.
Legislation
Motor vehicle traffic is
one
of the most
important sources for investigation in [51] indicates that
air
pollution
motor traffic throughout is the major source for air pollution in megacities, in half of them being the single most important source. Since 1950, the global vehicle fleet has grown ten times and it is estimated to double again within the next 20-30 years [56]. As cities expand, more people will drive more vehicles over greater distances and for longer time. Emissions caused by motor traffic are thus important to be
the world.
The
monitored and controlled. In
tightening of emission levels have been is¬ sued. The first passenger car emissions regulation was the directive 70/220/EEC and for heavy duty vehicle emissions the first directive was issued in 1998. The first mandatory European vehicle emission levels was set by the Euro-1 stan¬ dards introduced in the 91/441/EEC directive. Consequently, the Euro-2 stan¬ dard was set within the 94/12/EEC directive and the Euro-3 standard by the 98/69/EG directive. Presently, the Euro-4 standard is going to come in power Europe,
several directives for the
in 2006. In
Switzerland,
the ECE/UNO to
the ECE
the first emission limits
regulations.
regulations,
were
introduced in 1971,
In order to achieve the air
stricter emission limits
quality targets
conform
introduced in 1982. These
developed within the framework of the EFTA's "Stockholm group". In 1987 the next regulations, called FAV1, were enforced by setting first emission limits for diesel vehicles and by requiring three-way catalytic converter for gasoline vehicles. Stringent requirements for particles of diesel vehicles were set in 1998 within FAV2. Since 1996, the European legislative levels have been adopted in Switzerland. The evolution of Swiss regulations
for the limits
were
by adopting
were
5
1. Introduction
Class
Year
CO
HC
NOx
HC+NOx
Particles
[g/km]
[g/km]
[g/km]
[g/km]
[g/km]
0.25
0.62
Gasoline FAVl
1987
2.10
Euro 1
1991
3.16
1.13
Euro 2
1994
2.20
0.50
Euro 3
1998
2.30
0.20
0.15
Euro 4
2006
1.00
0.10
0.08
FAVl
1987
2.10
0.25
0.62
0.370
FAV2
1998
2.10
0.25
0.62
0.124
Euro 2
1994
1.00
Euro 3
1998
0.64
Euro 4
2006
0.50
Euro 5
2009
-
Diesel
0.10
Table 1.1.: Swiss and
and
European
European
are
0.080
0.50
0.56
0.050
0.25
0.30
0.025
placed
on a
standards for emissions of passenger
given
cars
in Table 1.1.
requirements, the vehicles dynamometer and driven through a specific
legislative
chassis
0.025
0.08
standards for emission levels is
To check the fulfillment of the
under test
0.70
emission
driving cycle. The
European legislative cycle (known
as
NEDC
-
New
European Driving Cy¬
cle) consists of an artificially created driving speed time series with low dynam¬ ics (see Figure 1.1). It contains a synthetic urban driving pattern (called ECE or
UDC
EUDC
-
-
Driving Cycle)
Extra Urban
Until the was
Urban
adoption
and
an
extra-urban
as
Driving Cycle).
of the Euro-3 standard
started with cold
driving pattern (known
engine
and
a
,
the
40 seconds
procedure was that the vehicle idle phase was run to warm-up
the
engine before the start of the measurements. From the introduction of Euro3, this pre-conditioning warm-up period was eliminated. In this way, emissions are measured from the beginning of a cold-start, making the fulfillment of Euro3 level much harder to acquire than the Euro-2 standard.
This
cycle, which skips the 40 seconds idle phase, is called NEDC 2000 (New European Driving Cycle 2000) or MVEG (European Motor Vehicle Emis¬ sions Group).
6
new
1.2.
200
400
600
time
Figure
1.1.:
In the United
800
1000
1200
time
[s]
Speed profiles
Legislation
of legislative NEDC
States, different emission standards
[s]
(left) and US FTP-75 (right)
are
enforced, with
even more
stringent requirements for California. The FTP-75 (Federal Test Procedure) cycle is being used in the US as a legislative cycle (Figure 1.1). The first 505 seconds of this test start when the engine is cold and represent the first part of this cycle. The test continues for another 867 seconds, at which point the vehicle is shut off. After a ten minutes interval, the first part is repeated with a warm engine. Effective model year 2000 vehicles have to be additionally tested on two Sup¬ plemental Federal Test Procedures (SFTP) designed to address short comings within the FTP-75 in the
(US06 cycle) and (2) the In
representation of: (1) aggressive, high speed driving use of air conditioning (SC03 cycle).
dynamometer driving schedule for light-duty vehi¬ California Air Resources Board. It is a more aggressive
California, the LA92 is
cles
developed by driving cycle then
the
a
the federal FTP-75. It has
higher speed, higher acceleration,
fewer stops per kilometer and less idle time. Most of the real-world emissions
strong acceleration, deceleration
are or
American
generated during
gear-shift phases.
phases like Both, European and speed and accelera¬
transient
legislative cycles have rather low maximum tion levels, which causes large discrepancies between emissions on certification tests and emissions
from real-world situations [21]. Therefore these standard
driving cycles are not representative their corresponding emission levels.
for real-world behaviour and, hence, for
7
1. Introduction
1.3. Real-world
driving cycles
agencies such as the Swiss Agency for Forest, Landscape (SAEFL), the Environmental Protection Agency
For about 20 years environmental Environment and
(EPA) in the United States, etc.,
have monitored the
success
of the emissions
regulations. The evolution in vehicle technologies (mainly in the electronic engine control systems) has caused, however, an increase in the difference be¬ tween emissions in legislation tests and those from real world driving. Thus, legislative cycles are no longer representative for real-world driving behaviour and, consequently, for the assessment of pollutants. The use of real-world driving cycles is therefore one of the key issues in emissions inventories. campaign has been conducted on Swiss roads in order to determine real-world driving behaviour [22]. Cars equipped with velocity and time logging devices were driven by special drivers, who were told to follow the flow of the traffic. During this measurement cam¬ paign 759'299 seconds of driving manner have been recorded and analysed by Within SAEFL,
statistical
an
extensive measurement
means.
Recorded data
were
characteristics. For
divided into
driving patterns
that, 14 parameters
were
based
on
the different road
defined to describe the road type:
change of the average speed during driving pat¬ tern, standard deviation of velocity, road gradient, percentage of time with con¬ stant velocity, percentage of time with velocity zero, length of the driving pat¬ mean
travel
speed, sign
of the
tern, etc.
By
means
of cluster
analysis,
tive for Swiss
12
driving patterns
which
behaviour have been selected to
driving real-world driving cycles. Each of these cycles patterns. Their corresponding speed time series
are
most
develop
representa¬
a new
contains three of these are
depicted
in
Figure
set of 4
driving 1.2.
driving situations, AE1R, AE2R and AE3R. Cycle R2 contains a motorway part A4R and two rural driving compo¬ nents, LE1R and LE2sR. Cycle R3 is formed out of rural driving LE2uR and urban driving LE3R and LE5R. Finally, R4 consists of urban driving LE6R, highway and urban stop-and-go driving, StGoHW and StGoUrb. Cycle
Rl is
composed
of three motorway
Within the frame of the
European
research program ARTEMIS
(Assessment
Reliability of Transport Emission Models and Inventory Systems), another real-world driving cycle called Common ARTEMIS Driving Cycle (CADC) has been developed [9]. The goal was to have a common cycle for all the and
8
1.3. Real-world
400
200
0
ft >m
AVl
WANVA
1.2.:
Speed profiles
,
Aa À. A
I 1500
1000
time
Figure
StGoUrb
StGoHW
500
0
1400
1200
1000
800
600R4
driving cycles
[s]
of the Swiss real-world
driving cycles Rl, R2, R3,
R4.
ARTEMIS partners, suitable for both
representative Using
for the actual
driving
bag
and instantaneous measurements and
conditions of
European
the Swiss data, the data from another multinational
cars.
project ([10])
and
European driving patterns have been identified by factorial analysis and clustering tools: congested urban, urban dense, urban with low speed, urban with free flow, urban unsteady, secondary roads unsteady, secondary rural roads, rural roads with steady speed, main-road unsteady, main-road with steady speed, motorway unsteady and motorway with additional data recorded in
Naples,
14
steady speed. aggregated road highway (Figure 1.3). Each of be attributed to a specific road
CADC is divided in three main parts which account for the
categories: urban,
rural
(i.e. extra-urban) and
sub-cycles that can "sub-category", allowing thus disaggregation of the emission levels at various driving conditions. The three main parts are independent from one another and each of them includes a pre- and a post-conditioning phase. these parts contain 4
Unlike R1-R4
cycles
or
5
which follow the NEDC's fixed,
predefined gearshift
9
1. Introduction
1
W
i
i
1
i
Rural
Urban
I WmilPlMT I
500
f
ull
i_J
II
1
2000
1500
1000
i
|J
2500
3000
I
I
3500
timersl
Figure
1.3.:
Speed
time series of the CADC
cycle
strategy, the CADC considers four strategies for the gearshift depending the technical characteristics of the vehicles
Kinematic characteristics and are
given
in
Appendix
description
(power,
mass, transmission
on
ratios).
of the Swiss and of the CADC
cycle
A.l.
cycles (MEC01) de¬ veloped by a research team from California [16], the cycle for the area of Hong Kong [74] or the Istanbul urban driving cycle [27]. Similar real-world
driving cycles
are
the modal emission
Although the real-world cycles represent a significant improvement towards representative emissions factors, they cannot take into account driving style dis¬ tribution, different loadings of the vehicle or gradients of the road. Therefore, models that are able to predict the emissions generated by these contributory aspects are of increasing interest.
1.4. Overview of the emission models
From 1995 until
2000, in Switzerland, the total vehicle-km of all road vehicles
has increased
about 7 percent
future is
by
and, by the
expected
to
year
[44]. This growth is likely
to
continue in the
2010, the increase in total vehicle-km relative
be about 19-20 percent
to 1995
[44].
Both, the increased vehicle fleet and larger distance covered yearly per vehicle have counteracted the
improvements generated by
the stricter
legislation.
Using test bench measurements, emission models are developed to obtain re¬ gional, national or international emission inventories and to predict the impact
10
1.4.
of different traffic related
Overview of the emission models
Emission models
measures.
can
be
split
into two
categories:
•
Fleet emissions models:
cycles and compute
a
they
use
weighted
emission factors of different measured
sum
that is
to fleet statistics to
multiplied
generate fleet emission models. •
Vehicle emission models:
they
allow to calculate emissions for any
pattern, for any combination of vehicle load, slope out of a limited set
of measured test
or
speed
gear-shift strategy
cycles.
depend on a large set of input parameters: traffic situation, loading, gradient of the road, driving behaviour, etc. Due to the quan¬
Vehicle emissions vehicle
tity
of information necessary to determine the different combinations related to
emissions, direct
the traffic
measurement becomes
impractical
Therefore, models for predicting emissions represent
an
and
expensive.
alternative to direct
measurement.
For
more
than
surements
a
decade attempts have been made to store
of test
on
chassis
dynamometers
way, such that emissions of other
driving
or
engine
conditions
or
map emission
mea¬
test
benches in
can
be calculated out of
a
neutral
them without additional measurements.
variety of vehicle emissions and fuel consumption models derived for different spatial and temporal scales. These models can be categorised into three main groups with increasing level of complexity: (a) average speed mod¬ els, (b) traffic situation models and (c) instantaneous (modal) models.
There is
a
1.4.1.
Average speed models
speed models relate emissions and fuel consumption to the aver¬ age speed of each driving cycle. The emission and fuel consumption rates are generated from chassis dynamometer measurements for a variety of simulated cycles at different average speed levels.
The average
example of this type of the model is the COPERT III computer program developed by the CORINAIR Working Group on behalf of the European Com¬ mission [59]. COPERT III uses linear regression to express emission factors as function of the average travelling speed (Figure 1.4).
An
11
1. Introduction
NO*-PC
Gasoline
2.0 1- EURO I
1.4
-ce
2,5
2,0
j--y«-8.5-4E=Qdx?---üJa8dx-+-QJS2il.
o-
R2 =Ü.0J772 1,5
SS
°
--
Raw Value
s
Average Valu«s
2
o
o
o
%
1,0
-p--
P.
0,5
^M:
--
8
L-l
0,0 20
40
1.4.: Calculation of
Figure
function of the average
80
60
Real
Mean
Cycles
NOx
100
140
Speed (km/h)
emissions for Euro-1
speed (source
120
COPERT
gasoline
vehicles
as
III)
prediction quality of the emission factors decreases dramatically from preECE vehicles (since emissions were high and presented good correlation with mean speed) to Euro-1 vehicles (where emissions become lower and more scat¬ tered between vehicles and operating conditions). The Pearson coefficient R2 The
for the
quality
of CO estimation
values between 0.133 and 0.159
drops
from 0.924
(for pre-ECE vehicles)
(for Euro-1 cars), depending
on
the
to
engine
capacity (Table 1.2). In
2003, about 15 European countries
were
official emission estimates, among them
using
the COPERT III model for
Belgium, Denmark, France, Greece,
Ireland, Italy and Spain. Emission factors for Euro-2, Euro-3 and Euro-4 vehi¬ cles
were
derived from the Euro-1 emission functions
An evaluation of the COPERT III model remote
sensing
measurements
in Sweden
was
[26]
favorable agreement for CO and HC emissions
using reducing
performed by
factors.
on-road
optical
with the results
showing not so and better quality for NOx emis¬
sions. In the
tion
US, the MOBILES model developed by the U.S. Environmental Protec¬
Agency [15] and the EMPAC model developed by the California Air Re¬ sources Board (CARB) [20] attempt to determine the overall emission levels,
12
Table 1.2.:
-
0.247
0.781 0.767 0.656 0.719
0.294V + 0.002478V2
0.00957V2 0.377V + 0.00283V2
0.00203V2
0.159
0.2955V + 0.0018V2
passenger
-
cars.
V represents average drivin
0.145
0.001728V2 0.245V +
0.133
0.2867V + 0.0022V2
0.23012V + -
-
0.613
0.0011639V2
0.22V +
0.1511V +
-
-
-
12.826
9.617
0.68V +
0.825
0.790
ln(V)
0.00377V2
260.788 V~u-yi
-
-
-
9.846
9.446
17.882
8.273
14.577
14.653
37.92
161.36-45.62
0.102
0.747
0.0026V2
300Y-U.797
0.158
0.406V + 0.0032V2
26.62-0.44V +
gasoline
5-130
CC>2.01 of CO emission factors for
5-130
1.4
'
400
8
measured
(solid)
and fitted
500
[s]
(dashed) step
response of
validation of the transport model for the
Right graph:
signal at the tailpipe, measured analyzer signal,
gas system in the FTP-75 test. Note: dash-dotted:
u(t),
dashed: simulated
analyzer signal, y(t),
solid:
approach is to observe how well the model sim¬ ulates the system output based on a given input.This method of simulation is a commonly applied procedure that compares the actual measured output of a model
[49]. One
system
to the
such direct
simulated output from the model.
For the validation of the was
run,
lambda
using
sensor
a
developed
vehicle mounted
(ETAS LSU)
was
raw
on
system model,
a
transient FTP-75 test
dynamometer. A continuous the end of the tailpipe and was
the chassis
mounted at
having a fast response time of 20 ms. This In addition, the raw gas sensor was used to measure the input signal u(t). signal y(t) was recorded at the analyzer.
calibrated
Using
as an
oxygen sensor,
the measured
input u(t)
of the
raw
and the identified parameters of Equation
2.7,
signal), y(t),
to the
ure
is simulated and
compared
tailpipe) corresponding output (analyser measured output, y(t). In Fig¬
system (oxygen signal
at the
the
2.8 the excellent agreement between measured and simulated output
can
be
checked.
33
2.
Modelling
of the exhaust gas transport systems
system of the
2.2.4. Exhaust
The exhaust system of the in Section 2.2.1,
as a
car can
car
be modelled,
perfect delay
and
using
the basic model described
first order system. Additional
a
com¬
plexity arises here since the volume flow of the exhaust gas may vary by a factor of fifty, depending on the engine load. Therefore, the simple approach used in Section 2.2.3 must be extended.
catalyst outlet location
The transport of the exhaust gas from the
to the
tailpipe
goes with the volume flow of the exhaust gases. This volume flow varies be¬ tween 3 1/s and
liters,
the total
1501/s for
delay
a
typical
2000
cm3
car.
For
varies between 0.1 and 6 seconds
an
exhaust system of 20
(Figure 2.2).
Therefore, the parameters of Equation 2.1 become time-varying. In conse¬ quence, the sequence of the subsystems becomes important here. The results are different if the transport time delay is considered before or after the mixing
Following the geometry of the exhaust transport system, the lay should be split ideally into two parts (corresponding to the pipes), dynamic part in between (corresponding to the silencer). chamber.
Is
Input
u(t)
Signal at the tailpipe
order
differential
aTd(t)
with the
Transport
Mixing
Transport
time de¬
(1
equation
-
a)Td(t)
m
Tm(t)
Figure
2.9.: Block
diagram
of the exhaust system model
Again it was found by least squares optimization into two equal parts (a 0.5), thus:
that the time
delay
can
be
split
=
TmWi(t ?^) ,(4 ZM)=0(t_ZkÖ>) +
This result of the
was
(2.io)
+
+
tested for two different vehicles with
completely different shapes
tailpipe systems. Equation 2.10, Td(t) and Tm(t), are now functions shall be determined according to Equations 2.2 and 2.3.
The parameters of
of the
volume flow and
Intro¬
ducing
an
additional parameter p, the coefficients of the system
can
be obtained
as:
Td(t)=pTT(t),
34
Tm(t)
=
(l-p)TT{t),
0
/
Q.
O Ü
y
/
Q.
^.---
y
f
200
n
O O
er
i
/y
.,--"
,.--
-'"
' y
/r^^'
^
„y
/
£
^-"
10
/jy
/
oo
^-
CX
y /
/ /
cted
TJ
/
/ /
60°
^
/
/
â30
*E
°
X
F jx
Validation
^
n
10
20
30
CO measured
"0
40
200
[g/km]
400
CO, 4
predicted NO
[g/km]
y
/
Oy»
"e
â
600
measured
y
/
3
Oy y
/ /
y j^y
/ / /
to
t
/
jo yd.
/
„--*
.**
*''*'
''IT /
--
^'
Q/O
/
-*
*'.
Çr ~y
^--"
y
y*-'"' o
12
Figure
3.8.:
3
4
HC measured
[g/km]
Comparison
)
between
predicted
1
2
NO
measured
with
3
4
[g/km]
map and measured
bmep-n
emission factors for the average pre Euro-1 vehicle
(E
the normalised
mean
the fractional bias:
square
FB
0.5
the fraction of predictions within
—E \ ^- ^
E
•
E
(Em+Ep) R
a
\Em
2
—
—
EmJ
•
{Ep Epj —
;——^r^-
factor two from measurements: FAC2.
Em and Ep represent the measured and predicted emission factors, Em
Ep
values and o),
the
(coxid* CRed)
the total volume of the
-
c&J
rads
TWC,
e
is
obtained:
(5.6)
rox
-
a
are
constant
representing the
volume
fraction of the gas phase, V denotes the volumetric flow of the exhaust gas, whereas OSC stands for the oxygen storage capacity of the TWC in mol/m3.
Generally,
is has been found that the
faster than the balance be
86
ones
dynamics
of the gas
of the oxygen storage and release.
species
are
Therefore, the
much mass
equations (5.4) and (5.5) can be applied as static equations. This can done by setting the left-hand side of the two equations to zero. After some
5.2.
algebra,
the
following
terms
are
Methodology
obtained for both coxid and CRed concentra¬
tions: v
V +
°Oxid
0.b-e-Vc-OSC-ki-(l-i/>o) V
.rin
v
CRed
With these
equations,
_
/< -7\
LRed
/e
~
V +
e-Vc-OSC-k2-*Po
both concentrations
be calculated from the oxygen
can
occupancy, temperature and volumetric flow.
The concentrations
are
in turn
used for the estimation of the reaction rates and,
consequently, of i/jo
conversion efficiencies of HC, CO and
further
NOx
o\
P-o)
—
are
on
•
The
characterised
as
static functions of this relative oxygen level variable. This model may appear to be
simpler
when
compared
to
other
approaches
pre¬
sented in the literature. But, this model is used for the emission factors mod¬
elling. Usually, vehicles studied for this purpose are available for a short time and only a limited number of tests are possible. Moreover, the tests are per¬ formed on chassis dynamometers and not on engine test benches (which would allow more detailed measurements). Therefore, the goal is not to have a good accuracy on an instantaneous basis, but to have a simple and efficient way of predicting the cumulated emissions over a transient cycle. In fact, the model has only five parameters to be determined (kinetic parameters and OSC) and conversion curves to be estimated. It will be shown that the tuning of these parameters is very simple and can be easily performed. Moreover, the very transparent structure of the model makes extensions or simplifications easy to implement.
5.2.2. Parameter estimation
The present model introduces
reference to vides
an
a
set
of
of parameters that has to be estimated with
set
experimental
estimation for each
computation, using
a
one
data. For each
experiment, modelling
of the measured outputs
pro¬
(coXid-> CRed). The
the model, of each output
model parameters. The
tuning
of the model
depends on the values of the requires that the tunable parame¬
ters be fitted in order to minimize the error between available measurements
and the viewed
respective as an
simulations.
Hence, the problem of model tuning
can
be
optimization problem.
87
5.
Dynamic catalyst model
For the oxygen storage
fied
submodel, the following parameters have
to
be identi¬
by tuning:
•
the activation •
pre-exponential factors Ai incorporated in the reaction rates.
Kinetic parameters: this includes the two
energies Ei
that
are
Oxygen storage capacity: the storage capacity of the catalyst determined simultaneously with the kinetic parameters.
and
has to be
performed on the chassis dynamometer with the BMW vehicle (see Table 4.1). The already measured driving cycle R3 (see Section 1.3), which covers a large area of different operating points, was The model
was
tuned to measurements
used to fit the model
The
parameters.
goodness-of-fit of the model was chosen to be the sum of squares of the sampled error between measured and computed concentrations of oxidizing and reducing species (coxid and CRed) at the catalyst outlet. For the optimization procedure, the nonlinear least-squares error algorithm "lsqnonlin" from the Matlab Optimization Toolbox [55] has been employed. performance
The results of the
simulated
vs.
which
measure
tuning
assesses
the
for the studied vehicle
are
presented
in the form of
measured coxid and CRed concentrations at the outlet of the
con¬
verter. The accuracy of the model in
predicting catalyst-out aggregated species presented Figure cycle part from 300s to 400s. The prediction of the model is remarkably good. This successful prediction indicates that the oxygen storage dynamics implemented in the model are capable of modelling the phenomenon with good accuracy. is
in
5.2 for the
5.2.3. Static conversion
curves
The behaviour of the TWC is characterised
by
the conversion efficiencies of
HC, CO and NOx. Some of the sixteen transient driving patterns have been
employed
as
experimental
data necessary to
identify
the conversion
curves as
function of the estimated relative oxygen level. The reason for using more than one transient cycle was to achieve enough points covering all possible ranges of the ROL.
Consider, for example, the instantaneous emission profile during the CADC part 1, which corresponds to an urban driving pattern (Figure 5.3). Obviously,
88
5.2.
Methodology
390
400
400
500
310
320
330
~i
500
310
340
i—
r
320
330
350
340
350
time
Figure 5.2.: ing species.
Measured
vs.
360
370
i
360
380
i
370
380
390
400
r
390
400
[s]
computed concentrations
of the
oxidising
and reduc¬
89
5.
Dynamic catalyst model >
10 —i
r~
1
~l
1
1-
5000
„
E Q.
11 il 0
..i.
100
..lit.
200
I..
300
I
400
....
500
Li
m
600
it
700
8°
A
800
900
1000 200
E a.
10000
&>
100
Juukiyiluii.-i..
t..
..
J.1
J lJ
o
5000
Figure 5.3.: Measured instantaneous NOx, HC inlet (dashed lines )and outlet (solid lines) over urban
the
and CO emissions at converter the 900 seconds of the CADC,
driving pattern.
significant overall conversion efficiency. The be, therefore, capable of matching the catalyst's breakthrough
specific catalyst
model should
attains
a
during accelerations, decelerations In order to be consisted with this
kinetic parameters and OSC
or
fuel cutoff situations.
catalyst's behaviour,
were
first the
already identified
used to generate the relative oxygen level
during the experimental data. Further on, conversion efficiencies of HC, CO and NOx were determined, but only when the values of the inlet concentrations were
above of
a
certain threshold.
NOx, for example, this threshold was set at 500 ppm. If the values of inlet NOx emissions are below this threshold it makes not a big difference for the For
cumulated emission value if the conversion range. At
points approximation.
were not
below this
a
constant
is set at 99%
conversion rate is
considered in the identification process of the conversion
efficiencies
were
or
an
at 50%
adequate
For this reason, the emissions below these threshold values
Once the events with
90
threshold,
efficiency
high
inlet concentrations
determined
on an
were
curves.
identified, the conversion
instantaneous basis. However,
we
have to
5.2.
-i
1
1
Methodology
i
r
~i~
measurements fitted function
—
.E
O
y 1
0.5-
o o
-ifrAiWtfc^-^1^» 01
0 2
0.3
0.4
atBa^èHËBBcate^iiWCTrw^iaiiiîifTOtiii
0.5
ROL
0.6
0.7
0.8
iliiiii
i,
-.
0.!
[-]
measurements
fitted function
ï
0.5
O
*!*»¥***£.
