Effect of other after-Treatment Systems on Particulate Emission
and Composition
Summary
A passenger car engine was equipped with a NOx-trap and the engine management was
changed in order to achieve an optimal trapping/regeneration cycle of the trap at constant
load and speed. These modifications had a big influence on the soot formation process. At
two engine mapping points the particle composition and size distribution was measured. A
first idea, if there is any indication to future problems with the measurement technique or to
future health problems due to unexpected particulate behaviour was evaluated from these
measurements. The measured particulate composition and size distribution downstream the
NOx trap are within the range of the particulate emission of engines without NOx trap even
during the fuel rich ( < 1) phases. There was no indication to problems with future particulate
emissions found.
Introduction
Beside the reduction of the particulate emission the NOx reduction will be important in order
to reach future exhaust regulations. In principle three configurations are possible depending
on the effectiveness of each component:
Reduction of the engine out
emission
Reduction with after-treatment
systems
Particulate reduction by Particulate
trap
NOx reduction by catalyst (NOX
trap, SCR…)
NOx reduction by engine
Management (EGR, Injection
timing…)
Particulate reduction by Particulate
trap
NOx reduction by catalyst and
Particulate reduction by Particulate
trap
The investigations at VKA (RWTH-Aachen) concerned two questions:
Is there any additional health effect by changing physical or chemical behaviour of the
particulate emission to be expected
Are there any problems due to the measurability of the particulate emission to be expected.
Due to these questions the NOx trap was selected for the investigations. The NOx trap
requires every few minutes a regeneration by the CO exhaust of a fuel rich combustion
( < 1). This causes a varying exhaust mass flow even at steady state conditions. (Fig. 1) The
internal combustion kinetics and the soot formation process is changing dramatically by
changing from lean to rich mode. In the moment there is very few literature, how this might
change the physical and chemical parameters of the particulate emission.
Engine, engine management and exhaust after-treatment
A prototype 4 cylinder, 1.9 l (92 kW), Common Rail Diesel engine with variable turbo charger
was used for the investigations. Full access to
the engine management system allows the
optimisation of the fuel rich combustion process
at constant torque and speed. By switching from
Sampling
lean to rich combustion most parameters
position
(injection timing, rail pressure, exhaust back
pressure, EGR-ratio, intake air flow and boost
pressure) were changed. A close coupled
oxidation catalyst was used in order to increase
the CO-concentration for the regeneration of the
NOx trap during the fuel rich phases. The NOx
trap reduces the CO-Emission efficiently by
reaction with the trapped NOx.
Fig. 1: Exhaust line
Intake air flow [l/min]
1000
Lambda
2
900
1.8
800
1.6
700
1.4
600
1.2
500
1
400
0.8
300
0.6
200
0.4
100
0.2
0
240
260
280
300
320
340
360
380
Lambda
CO [ppm]
Emissions downstream NOx Trap
CO [ppm] and HC [ppmC3]
and Intake Air Flow [slm]
Hydrocarbons [ppmC3]
0
400
Time [sec]
Fig. 2: Fuel / air ratio and exhaust mass flow at 1500 rpm / 2 bar bmep
Two engine mapping points were selected for the investigations:
1500 rpm;
2 bar bmep
(4,7 kW)
As example for High changes of the fuel/air ratio Temperature of NOx-trap
ECE - test
necessary for regeneration - fuel (210°C) at lower limit of
conditions
and air flow have to be adjusted its working temperature
3000 rpm;
8 bar bmep
(38 kW)
Outside of the Only small adjustments in the
NEDC - load
air flow are necessary
and speed map
Temperature of NOx-trap
(450°C) at upper limit of
its working temperature
Sampling System
The gaseous exhaust was measured from the undiluted exhaust between the oxidation
catalyst and the NOx trap and downstream the NOx trap. The EGR-ratio was determined by
CO2 measurement.
orifice
primary dilution tunnel
TTUN
gas
meter
Impactor
Fig. 3: Two step dilution system and sampling equipment
Exhaust-CO2 [%]
12
throttle
primary
dilution
Luft
pump
PIMP
continous
sampling (both
phases)
valve
secondary dilution air
secondary dilution tunnel
CO2 (secondary dilution)
DMA --> CPC
buffer
sample line 2 (fuel lean phase)
sample
filter
TSEK
filter
CO2
throttle
sample line 1 (fuel rich phase)
exhaust
PVBL
PNBL
PABS
throttle
A full flow dilution tunnel, controlled by a Venturi orifice was used for regulating the primary
dilution ratio. Because of the dynamic exhaust gas flow, only full flow dilution is sufficient to
measure the exhaust gas compounds. (Only the lambda- and the NOx sensor are quick
enough to measure real time signals from the undiluted exhaust.) A low primary dilution ratio
(approx.3.5) was selected in order to be comparable with future test procedures, where a low
dilution ratio seems to be necessary in order to achieve sufficient accuracy for the
measurement of the gaseous compounds. A secondary dilution (approx. 6.5) was used to
reach the maximum Temperature of 52°C at the sample position. The secondary dilution was
measured by gas meters. The total dilution ratio (approx. 23) was continually controlled by
CO2 measurement of the diluted exhaust.
The diluted exhaust was switched
to sample line 1 (fuel rich phase)
when the engine mode changed.
While the engine run 3 seconds
in rich mode, the (rich) sampling
lasts 6 seconds. Then the valve
switched back to the lean sample
line for 117 seconds.
Impactor and DMA are sampling
continually:
3.5
CO2 upstr. NOx Trap
CO2 downstr. NOx Trap
Lambda
10
3
particulate sampling
fuel rich operation
8
2.5
2
4
1.5
fuel rich
operation
2
1
lean (red) sampling line
0
120
0.5
121
122
Time [min]
Fig. 4: Lean/rich cycle and sampling time (1500 rpm; 2 bar bmep)
123
Lambda
rich (yellow) sampling line
6
It was not possible to separate the exhaust of the lean and rich combustion with the impactor.
The aerodynamic diameter was determined from the continuous sample during lean and rich
phases.
With the DMA the mobility diameter was determined. For a period of 8 minutes (4 lean/rich
cycles) the DMA voltage remained constant. One size class was sampled during this period
and counted by the CPC. During the fuel rich phases the particle number reaches maximum
or minimum values within short peaks. The particle number during the rich phases were
evaluated from these peak values and compared to the values of the lean phases. Than the
next bigger size class was measured by increasing the DMA voltage. 14 size classes were
measured from 10 to 400 nm. Than the measurement was repeated by lowering the Voltage
from step to step form 9920 V (400 nm) to 15 V (10 nm). The total test lasts about 4 hours.
Operating the NOx Trap
The engine was switched to the fuel rich regeneration mode for 3 seconds after 120 seconds
trapping. Some engine parameters for the engine mapping point 1500 rpm; 2 bar bmep are
shown for example in fig. 5.
upstr. NOx Trap;
Lambda
2.0
downstr. NOx Trap
1.5
1.0
speed
/ [1/min]
Pressure
/ [mbar]
0.5
boost pressure;
1400
exhaust back pressure
1200
1000
rich
operation
80000
70000
60000
50000
40000
30000
Turbocharger speed
Air massflow - EDC / [mg/stroke]
300
300
200
200
100
240
241
242
243
244
245
246
247
248
249
100
250
Time / [s]
VR529
Fig. 5: Air parameters at 1500 rpm, 2 bar bmep
High CO and Hydrocarbon concentrations (up to 1.5%) were reached during the rich mode
upstream the NOx trap as shown in fig. 6. The time resolution of the exhaust gas analysers is
not sufficient in order to reach the peak values within 3 seconds.
Because of the sulfate poisoning of the NOx trap a dependency of the fuel sulfur content is
not measurable. Therefore a low sulfur fuel (comparable to D4-Fuel) was used. Nevertheless
suffered the NOx trap capacity during the tests by the fuel sulfur. A desulfatisation procedure
was set up at 650 °C trap temperature. Because of the rapid aging of the trap at these high
temperature, it was not possible to desulfate the trap before each test. Therefore the trap
capacity varies from test to test.
upstr. NOx Trap;
Lambda
2.0
downstr NOx Trap
1.5
1.0
10000
1000
100
10
CO / [ppm]
O2 / [%]
HC / [C3-ppm]
0.5
10
8
6
4
2
0
10000
1000
100
10
1
200
300
400
500
600
700
800
Time / [s]
VR529
Fig. 6: Exhaust parameters at 1500 rpm, 2 bar bmep during lean/rich cycles
It might seems to be easier, to measure the exhaust of the rich phases by running the rich
phases for a longer time. Unfortunately this is not possible because of the rapid heating of
the trap during the rich phases. For example the exhaust temperatures are shown in fig. 7 for
a 30 sec rich operation compared to the used 3 sec regeneration mode.
Fig. 7: Exhaust temperature and its dependency on the regeneration duration (3000 rpm; 2 bar bmep)
For normal trap operation the trap temperatures for the selected engine mapping points are
shown in fig. 8.
Fig. 8: Temperature and NOx-concentration upstream and downstream the NOx trap
Particulate analysis
Abbildung
1
The
particulate
matter was sampled during 4 hours test time on 70mm T40 A60 Teflon
coated borosilicate filter and analysed by two
first extraction
consecutive extractions as shown in fig. 9.
Weighing of
The solubles were divided into an un-polar
the loaded
sample filter
(SOF) and a polar (WSF) fraction. The
second extraction
insoluble part (unsol) was weighted The
SOF
results of the extraction was compared with a
extract in
cyclohexane
volatility analysis by thermo gravimetry: A
WSF
extract with
sample of soot was scratched off the filters
Weighing of
20% Water in
the extracted
isopropanol
and heated on a thermo balance up to 650°C
sample filter
under Helium. Than oxygen was added and
unsol
Weighing of the
the non-volatile fraction was burnt. While the
insoluble fraction
after both extractions
weight of the sample was recorded. Some
ash remains on the thermo balance. The
Fig. 9: Particulate analysis by extraction
insoluble part as well as the non volatile
(650°C) part should mainly consist of soot. Unfortunately thermo gravimetric analysis of
particles sampled at 1500 rpm; 2 bar was not possible. Although a high amount of particles
was sampled during 4 hours sampling time it was not possible to scratch them off the filter
surface because they deposit inside the fibres.
The results of the extraction are shown in fig. 10 (1500 rpm; 2 bar bmep) and in fig. 11
(3000 rpm; 8 bar bmep). They are compared with the total mass collected by the
measurement of the aerodynamic size distribution on the impactor. A Berner low pressure
impactor was used with 10 aerodynamic size classes. The mass collected on the impactor
usually is lower, than the mass collected on the filter. First reason is, that a much lower
volatile fraction was deposit on the aluminium foils of the impactor because of the lower
pressure and the smaller un-polar surface of the foils compared with the borosilicate fibres of
the filters. For the non volatile particles the collection efficiency of the impactor is about 60%
to 70% because a part of the material is blown off the surface. For this reason the sum of the
collected material is usually little lower than the non soluble fraction collected on the filters.
Emission [mg/kWh]
500
primary dilution ratio
(lean mode):
3.5
sec. dilution ratio: 7.0
total dilution ratio: 24.5
400
Backup
SOF
WSF
Mixed Imp: lean and rich
mode was sampled together
onto the same impactor
300
Unsol
Impactor
200
100
0
Rich
Lean
Mixed Imp
Rich
1500/2; downstr. NOx trap; VR881
Lean
Mixed Imp
1500/2; upstr. NOx trap; VR887
Fig. 10: Extraction of particulate matter (1500 rpm; 2 bar bmep);
Impactor
Emission [mg/kWh]
600
Backup
SOF
500
WSF
Unsol
400
300
200
100
3000/8; downstr. NOx
trap; VR880
prim. dil = 3.5
sec. dil = 6.5
3000/8; upstr. NOx trap;
VR890
prim. dil = 3.5
sec. dil = 6.5
3000/8; upstr. NOx trap;
VR896
prim. dil = 23
sec. dil = 1.0
Mixed
Imp
Lean
Rich
Mixed
Imp
Lean
Rich
Mixed
Imp
Lean
Rich
Mixed
Imp
Lean
Rich
0
3000/8; upstr. NOx trap;
VR892
prim. dil = 23
sec. dil = 2.5
Fig. 11: Extraction of particulate matter (3000 rpm; 8 bar bmep); comparison of 1 and 2-step dilution
Conspicuous is the high soluble fraction of the particulates upstream the NOx trap at
1500 rpm; 2 bar bmep. This fraction was removed largely catalytic by the NOx trap.
Differences measured with different dilution ratios are within the statistical error of the
particulate measurement.
The thermo gravimetric analysis of the particulate samples at 3000 rpm, 8 bar shows, that the
non volatile fraction is nearly identically with the non soluble fraction (fig. 12). The volatile
fraction consists mainly of low boiling compounds like fuel and oil compounds for the samples
downstream the NOx trap and for the lean combustion. The rich combustion however
produces a significant portion (20%) of high boiling compounds, volatile between 450°C and
650°C. This can not be a part of the engine oil. It might consist out of tary compounds, build
at low combustion temperatures from the late fuel injection. This fraction is widely removed
by the NOx trap (similar to the soluble fraction, build by the rich combustion at 1500 rpm,
2 bar bmep
Fig. 12: Comparison between thermo gravimetry and extraction (3000 rpm; 8 bar bmep)
Particulate size distribution
The aerodynamic size distribution is shown in fig. 13 (1500 rpm; 2 bar bmep) and fig. 14
(3000 rpm; 8 bar bmep) In both cases the NOx trap reduces the aerodynamic diameter. It
was not possible to separate the size distribution of the fuel rich phases from the lean
phases. The measured size distribution from all particles emitted is similar to the size
distribution of state of the art vehicles. The effect, that the aerodynamic diameter was
reduced by active catalysts was observed for other catalytic systems too. Probably this effect
is mainly due to a reduction of the particle density. As shown in figure 14 the mobility
diameter is reduced a little bit by the NOx trap. The mobility diameter is comparable with the
geometric diameter. The reduction of the particle size is not only a density effect
Especially the rich combustion don’t produce a special mode of small particles.
35
1500/2; downstr. NOx trap; VR881
Particulate emission at each
aerodynamic size class [mg/kWh]
30
1500/2; upstr. NOx trap; VR887
25
20
15
10
5
0
25
50
98
194
382
774
1490
2930
5780
11400
Aerodynamic diameter (density = 1g/cm³) [nm]
Fig. 13: Aerodynamic size distribution (1500 rpm; 2 bar bmep)
120
3000/8; downstr. NOx trap; VR880
prim. dil = 3.5
sec. dil = 6.5
tot. dil = 23
Particulate emission at each
aerodynamic size class [mg/kWh]
100
80
3000/8; upstr. NOx trap; VR890
prim. dil = 3.5
sec. dil = 6.5
tot. dil = 23
60
3000/8; upstr. NOx trap; VR892
prim. dil = 23
sec. dil = 2.5
tot. dil = 57
40
3000/8; upstr. NOx trap; VR896
prim. dil = 23
sec. dil = 1.0
tot. dil = 23
20
0
25
50
98
194
382
774
1490
2930
5780
11400
Aerodynamic diameter (density = 1g/cm³) [nm]
Fig. 14: Aerodynamic size distribution (3000 rpm; 8 bar bmep)
The mobility size distribution the rich combustion differs only a little bit from the distribution of
the lean combustion. This is shown in a normalized scale (fig. 15: 3000 rpm; 8 bar bmep and
fig. 16: 1500 rpm; 2 bar bmep). A small shift to bigger diameters is within the accuracy of the
measurement. The catalytic NOx trap reduces the mobility diameter a little bit (fig. 15)
18%
lean: upstream
NOx trap VR880
Normalized Particle Number
(total = 100%)
16%
14%
riche: upstream
NOx trap VR880
12%
lean: downstr.
NOx trap VR892
10%
rich: downstr.
NOx trap VR892
8%
6%
4%
2%
0%
10
100
Mobility Diameter [nm]
Fig. 15: Mobility size distribution (3000 rpm; 8 bar bmep)
Fig. 15: Mobility size distribution (1500 rpm; 2 bar bmep)
1000