Model Validation
Validation of the three-way catalytic converter model is performed by comparing measured
vehicle dynamometer emissions with simulated emissions during the cold transient
phase of an FTP75 certification driving cycle. The vehicle chosen for this purpose
is a Volkswagen Jetta equipped with the low-end torque Volkswagen 2.0L-2V-85 kW
engine (165 Nm at 2600 rpm). A thorough revamp of the engine, its control, the gasoline
supply and the exhaust system made this vehicle the first Partial Zero Emission
Vehicle (PZEV) of Volkswagen [178]. To ensure ultra-fast catalyst light-off the engine
is started in rich mode (
[FONT=StandardSymL-Slant_167]l [/FONT]= 0.6 – 0.75) with secondary air being added into the
manifold targeting for exhaust compositions of
[FONT=StandardSymL-Slant_167]l [/FONT]= 1.05 – 1.25.
air pump
oxygen sensor (HEGO)
oxygen sensor (HEGO)
2nd catalytic brick
1st catalytic brick
temperature sensor (Pt200)
oxygen sensor (UEGO)
secondary air injection
Figure 7.1: Schematic sketch of Bosch Motronic ME7.1.1 system architecture (source:
Robert Bosch GmbH, GS/ESK) - incomplete.
92
UEGO (mounted to the exhaust manifold) HEGO (downstream 2nd brick)
3” x 4.66”, 900 cpsi, 2.5 mil 4.5” x 4.66”, 900 cpsi, 2.5 mil
200 g/ft
3, Pd:Rh = 14:1 100 g/ft3, Pd:Rh = 14:1
HEGO (mid brick)
TC2
temperature sensor
TC1
TC3 TC4
1st brick 2nd brick
Figure 7.2: Volkswagen’s Jetta 2.0l-2V-85 kW SULEV compliant exhaust system
equipped with thermocouples (TC1 - TC4).
The air-gap insulated manifold acts during this phase as a so-called thermoreactor,
completing the combustion process in a hot flame [178]. The temperature sensor located
downstream in the takedown pipe checks for proper operation of thermoreactor
and secondary air pump (figure 7.1). Since the SULEV emission thresholds (figure 1.1)
have to be fulfilled for mileages of 150 thousand miles without further facilitation, the
catalytic converter system is located in the underfloor position to avoid severe aging
conditions under full load operation. The system contains two bricks of the same
diameter. The first brick is shorter and has twice the precious metal loading of the
second brick (figure 7.2). There is one oxygen sensor with step characteristic mounted
between the two bricks and another one behind the second brick.
Modal measurements of the raw gas entering the catalytic converter system and of the
gas behind the first brick are performed at 1 Hz by means of Fourier-Transformation-
Infrared-Spectrometry (FTIR) with Volkswagen’s System for Emission Sampling And
Measurement (SESAM) [8].
Accumulated mass of carbon monoxide, hydrocarbons and nitrogen oxides behind the
first brick during the cold transient phase (
t = 0. . .507 s) of the FTP75 test procedure
are shown in figure 7.3. During cold start – which only needs a couple of seconds
for this application – a major fraction of total carbon monoxide and hydrocarbons is
emitted. After light-off, the temperatures are high enough for complete combustion
under lean conditions. Hydrocarbons are even converted under rich conditions: only
a slight and almost constant increase in hydrocarbon mass is registered after the first
brick – emphasizing the importance of converter light-off on hydrocarbon emissions.
The emissions of carbon monoxide and nitrogen monoxide after the first brick are substantial.
Especially nitrogen oxide emissions are significant between
t = 180. . .300 s
and thus due to warmed-up catalyst performance. In general, NO
x control strategies
focus on closed-loop, stabilized engine operation [174].
93
0 60 120 180 240 300 360 420 480
t
1
2
3
4
m
CO, mNOx
[g]
0.00
0.01
0.02
0.03
0.04
0.05
0.06
0.07
0.08
m
C3Hx
[g]
20
40
60
80
100
v [km/h]
C
3Hx
CO
NO
x
Figure 7.3: Measured accumulated carbon monoxide, nitrogen oxide and hydrocarbon
mass behind first brick during the cold transient phase of the FTP75 driving cycle.
The stoichiometry of the catalyst inlet emissions are crucial for three-way catalytic
emission control and thus for a sound description of the catalyst behavior modeled
in this work (see chapter 4). It is evident from figures 2.2 and 2.3 that the dynamic
response of the oxygen concentration signal is significantly slower than the dynamic
response of the polar species measured by FTIR technique. The dynamics of all analyzed
emission components are damped compared to the relative air to fuel ratio signal
provided by the linear oxygen sensor. Thus the oxygen concentration in the feed
is calculated from equation 2.17 based on the linear oxygen sensor (UEGO) signal
which exhibits a much better dynamic response to changes in gas phase composition
(figure 7.4)
1. The hydrogen concentration is assumed to be one third of the carbon
monoxide concentration (see chapter 3).
With the calculated oxygen concentrations the resulting set of engine out emissions still
underestimates the dynamics in engine out hydrogen, carbon monoxide, hydrocarbon
and nitrogen oxide emissions but obeys the fast dynamics of the measured relative air
to fuel ratio.
1
The relative air to fuel ratio signal is obtained from an extra Bosch LSU sensor not connected to the
engine control unit. This sensor is powered up well in advance to cranking the engine – thus providing
sound readings from the very first second. The obvious offset of the measured oxygen concentration
(not used in this work) is probably caused by a leakage in the vehicle roll exhaust sample piping.
94
0 60 120 180 240 300 360 420 480
t
0.0
1.0
2.0
3.0
4.0
5.0
y
O2
[%]
calculated
measured
20
40
60
80
100
v [km/h]
Figure 7.4: Converter inlet oxygen concentrations calculated from equation 2.17 based
on the linear oxygen sensor (UEGO) signal compared with the respective oxygen concentration
measured by means of a paramagnetic oxygen analyzer.
7.1 Thermal Behavior
Feed temperature estimates
Even though temperature readings of thermocouples
placed within the tube in front of the first brick are available (thermocouple TC2 in
figure 7.2), simulations based on these values result in way too low catalyst bed temperatures:
due to radiative heat losses temperature sensors positioned within a hot gas
phase surrounded by colder walls do have lower body temperatures (and thus provide
lower readings) compared to the gas phase they are positioned in [112, 121]. Thus
a feed gas temperature estimate believed to be more realistic is obtained by ”interpolation”
of the erroneous reading in front of the catalyst (TC2) and a hotter – also
erroneous – reading even further upstream (TC1 in figure 7.2). The empirical formula
T
g,in 0.3TTC1+0.7TTC2 yields reasonable feed temperature values for the first brick.
This can be seen in the following, where start-up measurements of the FTP75 cycle of
figure 2.2 are compared to simulations (figure 7.5).
Heat of reaction for oxygen storage and release
In order to check the influence of
the heat of reaction of oxygen storage and release, several simulation results are compared
with temperatures measured at two positions within the first brick. A quite good
description of the temperature progression with time at different positions within the
catalyst is obtained with the model described in chapter 4 and the parameters obtained
95
in chapter 5 and 6
2. The heat of oxygen storage is varied between a value of -760
(equivalent to the heat of reaction 3.14), a value of -200 kJ/(mol O
2) and zero.
25
50
75
100
v [km/h]
200
400
600
800
T [
oC]
D
hR,os = -760 kJ/(mol O2)
200
400
600
800
T [
oC]
D
hR,os = -200 kJ/(mol O2)
0 60 120 180 240 300 360 420 480
t
200
400
600
800
T [
oC]
D
hR,os = 0 kJ/(mol O2)
x = 15mm (measured)
x = 15mm (simulated)
x = 60mm (measured)
x = 60mm (simulated)
Figure 7.5: Measured and simulated temperatures within the first brick, assuming different
values of the enthalpy
DhR,os of oxygen storage reaction.
The value of -200 (close to the value of -220 kJ/(mol O
2) associated with palladium
oxidation) has been reported for calorimetric properties of ceria and ceria-zirconia
based palladium three-way catalysts [180]. The literature value of -200 kJ/(mol O
2)
provides the best description of the thermal behavior under warmed-up conditions.
2
Since the precious metal loading in the first brick is twice the loading of the catalyst the kinetic
parameters are obtained for, all kinetic rate constants are assumed to be twice as high.
96
With both a higher and a lower heat of oxygen storage and release significant deviations
between measured and simulated temperatures arise – especially during deceleration
associated with fuel cut and subsequent rich purges (figure 7.5, 120 s
< t < 360 s).
7.2 Cold Start Conversion
Due to the thermoreactor cold start strategy the first brick experiences temperatures
in excess of 700
C within a couple of seconds after cranking the engine (figure 7.5).
The critical cold start phase is passed within less than 20 s just before the first road
load is put on the engine. Figure 7.6 allows a comparison of measured and simulated
combustible species concentrations during this phase.
0 5 10 15
t
C
3H8
CO
0 5 10 15 20
t
100
200
300
400
500
y
C3Hx
[ppm]
C
3H6
engine out measured
mid brick measured
mid brick simulated
0.0
0.5
1.0
1.5
2.0
2.5
y
H2
, y
CO [%]
H
2
Figure 7.6: Measured and simulated hydrogen, carbon monoxide and hydrocarbon
emissions after the first brick (”mid brick”) during cold start.
Hydrocarbons are mainly emitted during the very first seconds of the driving cycle and
to a lesser degree during the corresponding seconds of the hot transient phase. More
than 10 different hydrocarbon species are measured by SESAM. They are grouped into
a fraction with relatively low catalytic ignition temperature represented by propene
and another fraction with relatively high catalytic ignition temperature represented by
propane.
97
During light-off propene shows a similar behavior as carbon monoxide with full conversion
after approximately 8 s. Propane light-off is somewhat delayed with full conversion
after 10 s. The evident time shift between simulated and measured peak hydrocarbon
emissions is due to the catalyst’s ability to store hydrocarbons in an intermediate
temperature range – a phenomenon not covered by the mathematical model. As can
be seen, hydrocarbon storage has no substantial influence on total hydrocarbon emissions
as long as hydrocarbon desorption occurs well in advance of catalyst light-off.
7.3 Transient Phase Conversion
During light-off the rates of catalytic combustion are limited by the low temperature.
In the hot catalyst under rich feed conditions after removal of residual oxygen, combustion
rates are limited by the rate of oxygen release from the oxygen storage. Under
lean feed conditions the conversion of nitrogen oxide to nitrogen is limited by the rate
of oxygen uptake.
0 60 120 180 240 300 360 420 480
t
0.950
0.975
1.000
1.025
1.050
lambda [-]
post 1st brick (simulation)
25
50
75
100
v [km/h]
0.900
0.925
0.950
0.975
1.000
1.025
1.050
lambda [-]
engine out (measurement)
Figure 7.7: Relative Air to Fuel ratio across the first brick.
If oxygen uptake and release were high enough all deviations from stoichiometric composition
would be buffered, resulting in a steady
[FONT=StandardSymL-Slant_167]l [/FONT]= 1 signal behind the catalyst. The
simulated exhaust composition after the first brick shows this behavior during idling
98
with low load after the first cruising period (around
t 150 s). Higher load operation
is accompanied with lean and rich breakthroughs (figure 7.7).
Significant nitrogen oxide breakthrough occurs during the following high speed operation
between second 200 and 300 (figure 7.3). It is evident from the time resolved
traces in figure 7.8 that these emissions are due to isolated peaks. While overall conversion
is well in excess of 80%, oxygen storage during lean engine operation seems
to be not sufficiently high to buffer all the oxygen – hindering complete nitrogen oxide
reduction.
During acceleration with high engine load the first brick emits significant amounts of
carbon monoxide under rich conditions. The accumulated mass of carbon monoxide
shows a pronounced increase around
t 350 s (figure 7.3). These few seconds add
approximately 20% to the mass accumulated during the cold transient phase. An explanation
will be given by considering the (simulated) change of stored oxygen.
It can be seen from the measured pre-first brick gas composition and the calculated
post-first brick gas composition (figure 7.7) that the average engine exhaust is slightly
rich. The exhaust composition is controlled such that the oxygen sensor’s voltage
behind the second brick is always 600-650 mV to ensure further high conversion of
nitrogen monoxide in the second brick [178].
With the catalyst storage being initially completely oxidized due to lean cold start with
secondary air injection, the slightly rich operation afterwards results in an average
decrease of the stored oxygen (figure 7.10). During the second hill (
t 150. . .300 s) an
average oxidation extend of roughly 35% is maintained providing high oxygen uptake
rates at still good oxygen release rates. During deceleration with fuel cut (
t 320 s) the
storage material is rapidly oxidized. During the following rich purge (
t 320. . .350 s)
oxygen is released from the storage until the storage is virtually empty. At
t = 350 s
the engine accelerates the vehicle with rich combustion and relatively high load. The
carbon monoxide conversion now suffers from a completely reduced storage material,
unable to compensate oxygen shortage in the feed. This explains the strong carbon
monoxide increase at about 350 s (figures 7.3 and 7.9).
After coldstart the measured hydrocarbon conversion is almost complete (figures 7.3
and 7.11). Using the kinetic data for steam reforming and water gas shift reaction determined
separately in section 5.4.2 and given in table 5.1, the simulated propane conversion
would have been too low. In other words: measured hydrocarbon conversion
under transient rich conditions with the Volkswagen Jetta 2.0l SULEV development
application is much better than the one predicted by simulation based on kinetic data
retrieved from steady state experiments. It is evident from data discussed in chapter 8
and appendix C that these differences most likely arise from sulfur poisoning: Kinetic
measurements discussed in chapter 5 are performed with an exhaust sulfur content of
y
SO2 = 14 ppm, while the certification driving cycle measurements discussed here are
performed with a much lower sulfur level. Furthermore in the presence of sulfur the
catalyst transforms into a less active state during prolonged steady state rich operation.
During real engine operation the catalyst is subjected to rich conditions no longer than
40 sec (see figure 7.7). Sulfur blocking the catalyst’s active sites as e.g. hydrogen
99
500
1000
1500
y
NO [ppm]
post 1st brick (measurement)
25
50
75
100
v [km/h]
500
1000
1500
2000
2500
3000
3500
y
NO
+
[ppm]
engine out (measurement)
180 240 300 360 420 480
t
500
1000
1500
y
NO [ppm]
post 1st brick (simulation)
Figure 7.8: Comparison of measured nitrogen monoxide emissions in front and behind
the first brick with simulated emissions behind the first brick.
100
0.5
1.0
y
CO [ppm]
25
50
75
100
v [km/h]
0.5
1.0
1.5
2.0
y
CO
+
[ppm]
180 240 300 360 420 480
t
0.5
1.0
y
CO [ppm]
engine out (measurement)
post 1st brick (measurement)
post 1st brick (simulation)
Figure 7.9: Comparison of measured carbon monoxide emissions in front and behind
the first brick with simulated emissions behind the first brick.
101
0 60 120 180 240 300 360 420 480
t
0.00
0.25
0.50
0.75
1.00
X
OSC [-]
Figure 7.10: Simulated averaged oxidation extent of the storage material.
sulfide are swept away in the subsequent lean phase. Therefore propane conversion
was calculated with the same rate parameters as propene in figure 7.11. Nevertheless
some break-through of hydrocarbons is simulated during the rich acceleration at about
350 s.
7.4 Discussion
With model parameters and properties obtained from the suppliers (e.g. substrate cell
density and wall thickness, washcoat loading, precious metal loading), from literature
(e.g. heat capacities, transfer coefficients, oxygen diffusivity within oxygen storage)
and kinetic and capacity data from isothermal measurements performed on a sample
with half the precious metal loading and a 25 % lower specific geometric surface area
the behavior of a full size catalytic converter operated under highly transient thermal
and stoichiometric conditions can be described reasonably well. This is in particular
remarkable since one single set of species conversion kinetic parameters obtained
from both fuel lean (oxygen saturated catalyst) and fuel rich (oxygen depleted catalyst)
conditions (table 5.1) is applied. During Volkswagen Jetta 2.0l SULEV FTP
drive cycle simulation however the catalyst does change between periods of oxygen
release and oxygen uptake. Hence the oxygen storage is usually not completely filled
or completely empty but somewhere inbetween (0
x 1, see figure 7.10).
While the independently parameterized model provides a good description of the cold
start behavior and the thermal behavior throughout the cold transient FTP cycle quantitative
differences between measurement and simulation are evident from a comparison
of transient carbon monoxide and nitrogen oxide emissions on a peak to peak basis
(figure 7.9 and 7.8). With the significantly higher sulfur content applied throughout
all kinetic measurements, the scale up by more than two orders of magnitude, poor
time resolved emission analysis, lack of hydrogen concentration analysis and a questionable
flow distribution within the first brick (see inlet cone geometry figure 7.2) the
degree of carbon monoxide and nitrogen oxide conversion prediction is believed to be
remarkably good. Drive cycle simulations with similar models reported in literature
provide roughly the same accuracy [126, 146, 4].
102
500
1000
y
C3Hx
[ppm]
25
50
75
100
v [km/h]
500
1000
1500
2000
2500
3000
y
C3Hx
+
[ppm]
180 240 300 360 420 480
t
500
1000
y
C3Hx
[ppm]
engine out (measurement)
post 1st brick (measurement)
post 1st brick (simulation)
Figure 7.11: Comparison of measured hydrocarbon emissions in front and behind the
first brick with simulated emissions behind the first brick. After coldstart engine out
hydrocarbons are assumed to be propene only.
103
The simulations underestimate the warmed-up propane conversion under rich conditions
compared to the measured – almost complete – alkane conversion. Similar shortcomings
in predicting transient improved hydrocarbon conversion under rich conditions
are reported in literature [166]. Steady state experiments with a rich bias to mean
air to fuel ratio (e.g.
[FONT=StandardSymL-Slant_167]l [/FONT]=0.98 ±0.01, f =1 Hz) carried out in this work (section 5.4.2)
maintain the catalyst in an oxygen depleted state over its full length. With the Volkswagen
Jetta 2.0l SULEV application described before the oxygen storage is controlled at
an intermediate oxidation state during transient operation by applying low frequency
oxidation and reduction cycles with reasonable amplitude. It has been reported that the
overall conversion of a three-way catalytic converter system can be enhanced by prolonged
rich/lean phases as long as full system break through is avoided (e.g. [90, 111]).
Further experiments towards the understanding of this transient improved rich conversion
phenomenon are discussed in chapter 8 and appendix C.