Cylinder pressure-based control in heavy-duty EGR diesel ... · Page 2 of 17 combustion phasing, the stable operation range can be expanded [1, 2]. Furthermore, the combustion rate

Page 1 of 17

2010-01-0564

Cylinder pressure-based control in heavy-duty EGR diesel engines

using a virtual heat release and emission sensor

Frank Willems, Erik Doosje, Frank Engels, Xander Seykens TNO Automotive

Copyright © 2010 SAE International

ABSTRACT

This paper presents a cylinder pressure-based control (CPBC) system for conventional diesel combustion with high EGR levels. Besides the commonly applied heat release estimation, the CPBC system is extended with a new virtual NOx and PM sensor. Using available cylinder pressure information, these emissions are estimated using a physically-based combustion model. This opens the route to advanced On-Board Diagnostics and to optimized fuel consumption and emissions during all operating conditions.

The potential of closed-loop CA50 and IMEP control is demonstrated on a multi-cylinder heavy-duty EGR engine. For uncalibrated injectors and fuel variations, the combustion control system makes the engine performance robust for the applied variations and reduces the need for a time consuming calibration process. Cylinder balancing is shown to enable auto-calibration of fuel injectors and to enhance fuel flexibility. For both Biodiesel and US diesel, the effects on NOx and PM emissions are partly compensated for by combined CA50 and IMEP control. This can be further improved by application of (virtual) emission sensors. Furthermore, it is shown that this combustion controller shows good transient performance during load changes.

The virtual emission sensor is successfully implemented for real-time control. For operating conditions with high EGR rates and varying injection timing, the predictions of the virtual NOx and PM sensor are compared with measurements. NOx emission prediction inaccuracy is typically on the order of 12%, which is comparable to commercially available sensors. The predicted PM emissions show good qualitative agreement, but need further improvement for application in DPF regeneration and PM emission control strategies. Robust emission control is essential to meet future requirements for On-Board Diagnostics and In-Use Compliance.

INTRODUCTION

Closed-loop combustion control is essential for Homogeneous Charge Compression Ignition (HCCI) and Pre-mixed Charge Compression Ignition (PCCI). Figure 1 shows a frequently applied classification of different combustion concepts. HCCI and PCCI are characterized by ultra low nitrogen oxides (NOx) and Particulate Matter (PM) emissions. Both concepts rely on auto ignition of a mixture, which is typically created by fuel injection that is (nearly) completed prior to the start of combustion. These advanced concepts are sensitive to operating conditions, since ignition delay is determined by the chemical reaction kinetics. By controlling the

Page 2 of 17

combustion phasing, the stable operation range can be expanded [1, 2]. Furthermore, the combustion rate has to be controlled to avoid noise and engine damage, especially at high loads. An overview on HCCI control approaches can be found in e.g., [2, 3].

Figure 1: Possible classification of different combustion concepts [4].

More conventional combustion concepts also benefit from closed-loop combustion control [5-8]. In 2008, Volkswagen and GM introduced these control systems on the market for their new generation diesel engines [9, 10]. By using cylinder pressure information, combustion phasing and indicated mean effective pressure (IMEP) are controlled. This cylinder pressure-based control (CPBC) system makes the engine robust for variations in fuel quality and component tolerances. Furthermore, fuel rate variations between cylinders are reduced up to a factor 4. As shown in [11, 12], CPBC systems significantly reduces variations in engine out NOx and PM emissions, such that the most stringent emission legislation can be met. It also reduces noise and vibrations [13]. Nakayama et al. [14] demonstrate that combustion phasing control prevents the engine from misfires, improves torque response, and reduces emissions during transient driving conditions.

For heavy-duty applications, closed-loop combustion control results are reported in the literature. However, these studies mainly focus on HCCI and PCCI concepts, see e.g. [14 - 19]. Conventional diesel combustion concepts can also benefit from CPBC. Besides the robustness for fuel variations and production tolerances, the auto-calibration capabilities of these closed-loop control systems are of interest. This can dramatically reduce the time consuming calibration of today's engine control systems. Engine performance can also benefit from these adaptation capabilities during real-world driving conditions: closed-loop combustion control makes the engine robust for unanticipated variations in ambient conditions, ageing and wear. This greatly improves real-world performance, such that future requirements for On-Board Diagnostics (OBD) and In-Use Compliance (IUC) can be met.

In this paper, we focus on cylinder pressure-based control for late DI combustion in a heavy-duty diesel engine with high Exhaust Gas Recirculation (EGR) rates. The main contribution of this work is the extension of the standard CPBC system with a virtual NOx and PM sensor. This innovative sensor is a further development of the virtual NO and soot sensor introduced in earlier work [20]. It uses cylinder pressure information as main input. With the availability of this sensor, engine performance can be optimized during all operating conditions, realizing the requested power output with minimal fuel consumption within the limits set by emission legislation. Furthermore, it opens the route to advanced diagnostics. In the first part of this paper, the potential

Page 3 of 17

of the CPBC system is examined for three cases: auto-calibration of fuel injectors, fuel flexibility, and transient performance. In the second part, the real-time prediction capability of the virtual emission sensor is demonstrated from experiments. Furthermore, an outlook is given on future developments in model-based combustion control systems.

EXPERIMENTAL SET-UP

ENGINE TEST PLATFORM

For advanced combustion research, a flexible heavy-duty engine test platform is available at TNO [21], see Figure 2. This platform is based on a 12 ℓ, 390 kW heavy-duty diesel engine. This engine is equipped with two-stage turbo charging and both a low-pressure (LP) and high-pressure (HP) EGR system. More specifications of the engine are given in Table 1. The high-pressure EGR valve and variable turbine geometry (VTG) turbocharger are controlling the high-pressure EGR flow. A low-pressure EGR valve and back pressure valve are controlling the low-pressure EGR flow. TNO’s in-house developed FLexible Engine Control System (FLECS) is applied to have maximal flexibility in fuel control. All the measurements presented in this study are performed with one main fuel injection event and fuel injection pressure of 1800 bar. During a fuel injection event, the fuel line pressure is not constant, due to the unit pump operating principle.

HP-EGR valve

HP-EGR

cooler

Inter

cooler

DPF

LP-EGR cooler

Backpressure

Valve (BPV)

LP-EGR valve

GT40V – VTGGT47

turbo

Airflow

Exhaust

out

P

T

P

T

Pcyl

6x

P

T

T

P

T

P

T

T

P T

T

P

T

P

TP

TT

λλλλsmoke

HP-EGR

1 2 3 4 5 6

Bypass

Valve

PCO2

emissions

CO2

smoke

emissions : Mexa 9500

: AVL 415

P

T

: Pressure

: Temperature

λλλλ : UEGO

Pcyl : 6x Kistler pressure sensor

DPF:Diesel particulate filter

emissions

HP : High PressureEGR system

LP : Low PressureEGR system

HP-EGR valve

HP-EGR

cooler

Inter

cooler

DPF

LP-EGR cooler

Backpressure

Valve (BPV)

LP-EGR valve

GT40V – VTGGT47

turbo

Airflow

Exhaust

out

PP

TT

PP

TT

PcylPcyl

6x

PP

TT

TT

PP

TT

PP

TT

TT

PP TT

TT

PP

TT

PP

TTPP

TTTT

λλλλλλλλsmoke

HP-EGR

1 2 3 4 5 6

Bypass

Valve

PPCO2CO2

emissions

CO2CO2

smoke

emissions : Mexa 9500

: AVL 415

PP

TT

: Pressure

: Temperature

λλλλλλλλ : UEGO

PcylPcyl : 6x Kistler pressure sensor

DPF:Diesel particulate filter

emissions

HP : High PressureEGR system

LP : Low PressureEGR system

Figure 2: Schematic view of the experimental set-up

Page 4 of 17

Table 1: Specification of the experimental set-up

Specifications

Diesel engine 12 ℓ, 6 cylinder in-line

Fuel injection equipment (FIE) 2400 bar unit pump system

Fuel EN590

Diesel Particulate Filter (DPF) 32 ℓ catalyzed sintered metal type

Emission measurement system Horiba MEXA 9500:

NOx, NO, THC, CO, CO2 and O2

The emission measurement system consists of two Horiba MEXA9500 units. The overall EGR ratio was determined using the measured CO2 concentrations in the engine exhaust and intake manifold. An additional analyzer measures the CO2 concentration between the compressor stages; this was used to determine the distribution of EGR over the high- and low-pressure route. For the studied operating points, approximately 20% of the total EGR flow was routed through the low-pressure circuit. Smoke numbers were measured with an AVL 415S smoke meter. For combustion analysis, Kistler pressure transducers were installed in all cylinders. Both the pressure signals and the current traces of the fuel injectors are recorded using an AVL 621 Indimodul.

CLOSED-LOOP COMBUSTION CONTROL SYSTEM

Figure 3 shows the closed-loop combustion control system that is installed on the presented engine. This control system uses the available cylinder pressure information from the installed Kistler sensors. Data acquisition is automatically triggered by the pulses from the crank shaft encoder and cylinder pressure is sampled with a resolution of 1 oCA. A 40 MHz, 2Mgate Field Programmable Gate Array (FPGA) is used for pressure signal analysis and real-time heat release computation. This virtual heat release sensor is based on the method presented in [22].

For the individual cylinders, the crank angle corresponding to 50% of the maximum apparent heat release (CA50) and the indicated mean effective pressure (IMEP) are determined every cycle. Based on the detected error, the PI controllers adjust the injection timing and fuel quantity in the next cycle to achieve the desired CA50 and IMEP values, respectively. This is called cycle-to-cycle control or individual cylinder control. Cylinder balancing is achieved by simultaneously running the six individual cylinder controllers with identical set points. For the presented system, real-time cylinder balancing is demonstrated up to 2000 rpm. Both the PI controllers and actual fuelling control are modeled in Matlab/Simulink and coded using the Real Time Workshop. Calibration of the controllers is done for several step responses. Goal was a response time of 5 cycles to meet 90% of the set-point value with minimal cylinder-to-cylinder variation.

To avoid unacceptable noise levels and structural damage, the cylinder pressure rise (dp/dα) and cylinder peak pressure (pmax) are monitored. In the CPBC system, the following limits are applied: 15 bar/CA and maximal pressure of 200 bar. This is especially of importance for pre-mixed type of combustion, like PCCI.

Page 5 of 17

dp/dα

pmaxpmax

IMEP

PI-Control

Cylinder

Pressure

sensor

CA50

PI-Control

Engine

speed

Torque

Feedforward

control

CA50, desired

IMEP, desired

CA50

IMEP+

-

+

-

Fuel

Timing

Safety

limits

pmax

Safety

limits

Virtual heat

release sensor

IMEP

PI-Control

Cylinder

Pressure

sensor

CA50

PI-Control

Torque

Feedforward

control

CA50, desired

IMEP, desired

CA50

IMEP+

-

+

-

quantity

Timing

Safety

limits

pmax

Safety

limits

FPGA

dp/dα

Figure 3: Scheme of the closed-loop combustion control system.

EXPERIMENTAL RESULTS

To demonstrate the potential of the applied closed-loop combustion control strategy, both robustness and transient performance are examined. Therefore, three cases are studied:

• Auto-calibration of fuel injectors;

• Fuel composition variations (fuel flexibility);

• Fuel quantity steps.

For these cases, the effect of the control system on CA50, IMEP and emissions is examined.

AUTO-CALIBRATION OF FUEL INJECTORS

First, tests are done with un-calibrated fuel injectors. This was realized by disturbing the trim factors of the individual injectors that are used in the engine management system. As shown in Figure 4, this results in a significant IMEP variation up to 1 bar between the different cylinders. Cylinder balancing based on closed-loop IMEP control compensates for this and realizes the desired IMEP value (4.7 bar) for all cylinders. Figure 4 shows the results of an experiment in which the closed-loop IMEP controller is sequentially activated for the individual cylinders. Note that the IMEP controllers for cylinder 1 and 2 were already activated for the shown time traces.

With this control system, the calibration effort of the injectors can be drastically reduced. By applying this control system in vehicles, also on-road injector performance variations can be compensated for and injector fouling and clogging can be detected over life time. According to [10], closed-loop IMEP control can also minimize variations in injected quantities, especially in multiple injection strategies.

Page 6 of 17

0 20 40 60 80 100 120 140 160 180 2003.8

4

4.2

4.4

4.6

4.8

5

CycleNo

IME

P [

bar]

cyl1

cyl2

cyl3

cyl4

cyl5

cyl6

Figure 4: Effect of IMEP control (IMEPdesired=4.7 bar, 1200 rpm, SOI=-10° ATDC)

FUEL FLEXIBILITY

In the second case, the robustness of the engine is examined for varying fuel conditions. More precisely, the effect of three different fuels on combustion phasing, power output and NOx and smoke emissions is determined. The studied fuels are listed in Table 2.

Table 2: Overview of studied fuels

Case Fuel type

A (ref) Standard diesel (EN590)

B Biodiesel (EN14214)

C US diesel

For the selected operating point, the following experiments are done:

1. Open-loop CA50 and IMEP control: for all fuels, injection timing and fuel quantity values corresponding to standard diesel are applied. This represents the case of drop-in replacement;

2. Closed-loop IMEP control: by adjusting fuel quantity, IMEP is restored to the value that is found using standard diesel;

3. Closed-loop CA50 and IMEP control: by adjusting fuel quantity and injection timing, both IMEP and CA50 are controlled to the values that are found for standard diesel;

Page 7 of 17

Figure 5 illustrates the effect on combustion phasing when standard diesel is replaced by Biodiesel and US diesel. For the open-loop controlled case, Biodiesel shows an advance in CA50 of more than 1 oCA, whereas the advance is negligible for US diesel fuel. The closed-loop control system compensates fully for this. In both cases, the resulting CA50 is controlled to the set point: CA50desired= 5° CA ATDC (right-hand bar for Biodiesel and US Diesel). This figure also illustrates the coupling between both control loops; combustion phasing changes when only closed-loop IMEP control is enabled (middle bar). This is most visible for Biodiesel: CA50 is retarded due to the increased injection duration. For US Diesel, the opposite effect is visible. Here, the injection duration decreases and CA50 is advanced.

CA50 open-loopIMEP open-loop

CA50 open-loop

IMEP closed-loopCA50 closed-loop

IMEP closed-loop

EN590 Biodiesel US Diesel

0.0

1.0

2.0

3.0

4.0

5.0

6.0

1

CA

50[°

CA

AT

DC

]

CA50 open-loopIMEP open-loop

CA50 open-loop

IMEP closed-loopCA50 closed-loop

IMEP closed-loop

EN590 Biodiesel US Diesel

0.0

1.0

2.0

3.0

4.0

5.0

6.0

1

CA

50[°

CA

AT

DC

]

Figure 5: Effect of closed-loop CA50 and IMEP control on CA50 (1200 RPM, λ=1.5, 46% EGR), CA50desired=5°CA ATDC, IMEPdesired=8.2bar.

Figure 6 shows the effect of the different fuels and controllers on IMEP. As the heating value of the tested fuels is different from standard diesel, IMEP changes for unchanged injection duration. Corresponding to their heating values, Biodiesel shows a reduction in IMEP, while US Diesel shows an increase. The closed-loop IMEP controller compensates fully for this. This figure also illustrates that the coupling between closed-loop CA50 control and resulting IMEP is small in the studied operating point (middle bar for Biodiesel and US Diesel).

Page 8 of 17

CA50 open-loop

IMEP open-loop

CA50 closed-loopIMEP open-loop

CA50 closed-loop

IMEP closed-loop

EN590 Biodiesel

US Diesel

5.0

5.5

6.0

6.5

7.0

7.5

8.0

8.5

9.0

1

IME

P [

bar]

CA50 open-loop

IMEP open-loop

CA50 closed-loopIMEP open-loop

CA50 closed-loop

IMEP closed-loop

EN590 Biodiesel

US Diesel

5.0

5.5

6.0

6.5

7.0

7.5

8.0

8.5

9.0

1

IME

P [

bar]

Figure 6: Effect of closed-loop IMEP and CA50 control on IMEP (1200 RPM, λ=1.5, 46% EGR), CA50desired=5°CA ATDC, IMEPdesired=8.2bar.

NOx emission

The changes in combustion phasing are reflected in the engine-out emission levels. In Figure 7, NOx emissions are shown that correspond to the results shown in Figure 6. For Biodiesel with closed-loop IMEP control, the injection duration is increased and CA50 is retarded. This leads to reduced NOx emissions. In case of US Diesel, the injection duration is decreased; CA50 is advanced, which results in higher NOx emissions. The right-hand case shows the effect of combined CA50 and IMEP control, which restores both combustion phasing and IMEP to the values corresponding to standard diesel. This results in a 26% reduction of NOx emissions for Biodiesel and 5% reduction for US Diesel, compared to the open-loop controlled case.

Page 9 of 17

CA50 open-loopIMEP open-loop

CA50 open-loop

IMEP closed-loopCA50 closed-loopIMEP closed-loop

EN590

Biodiesel

US Diesel

0.50

0.60

0.70

0.80

0.90

1.00

1.10

1.20

1

NO

x [

g/k

Wh

]

CA50 open-loopIMEP open-loop

CA50 open-loop

IMEP closed-loopCA50 closed-loopIMEP closed-loop

EN590

Biodiesel

US Diesel

0.50

0.60

0.70

0.80

0.90

1.00

1.10

1.20

1

NO

x [

g/k

Wh

]

Figure 7: Effect of closed-loop CA50 and IMEP control on brake-specific engine-out NOx emissions for

Biodiesel and US Diesel (1200 RPM, λ=1.5, 46% EGR), CA50desired=5°CA ATDC, IMEPdesired=8.2bar.

CA50 open-loop

IMEP open-loop

CA50 open-loop

IMEP closed-loopCA50 closed-loop

IMEP closed-loop

EN590

Biodiesel

US Diesel

0.00

0.50

1.00

1.50

2.00

2.50

1

Sm

oke [

FS

N]

CA50 open-loop

IMEP open-loop

CA50 open-loop

IMEP closed-loopCA50 closed-loop

IMEP closed-loop

EN590

Biodiesel

US Diesel

0.00

0.50

1.00

1.50

2.00

2.50

1

Sm

oke [

FS

N]

Figure 8: Effect of CA50 and IMEP control on smoke numbers for Biodiesel and US Diesel (1200 RPM, λ=1.5, 46% EGR), CA50desired=5°CA ATDC, IMEPdesired=8.2bar.

Page 10 of 17

PM emission

In Figure 8, the corresponding smoke numbers are shown. Due to its high oxygen content, Biodiesel has a much lower smoke number, while US Diesel performs about the same as standard diesel. For Biodiesel with only closed-loop IMEP control (middle bar), smoke numbers slightly increase, because of the retarded combustion phasing. In that case, smoke numbers decrease significantly for US Diesel, since combustion phasing is advanced. By applying both CA50 and IMEP control (right-hand case), the smoke number returns to a value similar to the open-loop situation for both fuels.

In summary, the main benefit of the CPBC system for varying fuels is that the combustion phasing is corrected for a condition that is unknown to the (open-loop) engine management system. As a result, engine performance will greatly be improved, since emissions and fuel consumption are seen to be sensitive for combustion phasing variations, see also [7, 10, 13]. Furthermore, the calibration effort can be dramatically reduced due to the auto-calibration capabilities of the closed-loop control system. Multiple calibrations for various fuels can thus be omitted and combustion control is possible even for continuously varying fuel blends. However, it is noted that the closed-loop combustion control system cannot guarantee a compliance with emission law; effects originating from the fuel's chemical composition are unknown to the system and cannot be accounted for. Typically, information on the NOx level is needed, which can then be used to correct the combustion phasing by generating a new target CA50 value. This NOx information can be provided by a NOx sensor. An alternative solution is the use of a virtual sensor, which predicts NOx emissions based on the measured cylinder pressure. This virtual sensor is presented in the second part of this paper.

FUEL QUANTITY STEPS

The transient performance of the closed-loop combustion controller was examined for a step wise change in fuel quantity from 50 to 90 mg/cycle. Figure 9 shows the results for the open-loop controlled case, where injection timing is kept constant: start of injection (SOI) at -10°CA ATDC. By increasing the injection duration to realize the desired fuel quantities, CA50 shifts from approximately 1°CA ATDC towards 5°CA ATDC. Figure 10 shows the results for cylinder balancing based on closed-loop CA50 control. The CA50 set point was 5°CA ATDC for all cylinders. For the low fuel quantity, the controller adapts the injection timing, such that the desired CA50 value is reached. During the step change (around cycle number 95), CA50 is retarded for a few engine cycles. However, within a few cycles it is controlled to the desired value. This demonstrates the fast response of the CA50 controller, which is a pre-requisite for accurate transient control. Current research concentrates on tracking control; realize a CA50 trajectory as a function of varying engine operating conditions.

Page 11 of 17

0 20 40 60 80 100 120 140 160 180 200-2

0

2

4

6

8

10

CycleNo

CA

50 [

°CA

AT

DC

]

cyl1

cyl2

cyl3

cyl4

cyl5

cyl6

Figure 9: CA50 response for a fuel quantity step with open-loop control (1200 RPM, IMEP=8.2 Bar (after step),

λ=1.35, 47% EGR, SOI=-10°CA ATDC).

0 20 40 60 80 100 120 140 160 180 200-2

0

2

4

6

8

10

CycleNo

CA

50 [

°CA

AT

DC

]

cyl1

cyl2

cyl3

cyl4

cyl5

cyl6

Figure 10: Effect of closed-loop CA50 control on combustion phasing for a fuel quantity step (1200 RPM,

IMEP=8.2 bar (after step), λ=1.35, 47% EGR)

VIRTUAL HEAT RELEASE AND EMISSION SENSOR

Current state-of-the art is to use measured in-cylinder pressure signals to derive variables like e.g., CA50, IMEP and maximum in-cylinder pressure. However, more information can be obtained from the available pressure

Page 12 of 17

signal: the formation of in-cylinder NOx and Particulate Matter (PM) emissions. Currently, PM sensors are not commercially available. Therefore, a virtual PM emission sensor has great added value, especially for DPF control. Furthermore, the virtual NOx sensor has a cost advantage: NOx sensors can be omitted or less expensive versions can be used which are less accurate and/or have slower response times.

In [20], a virtual emission sensor has already been introduced. Based on a physically-based combustion model, it allows a prediction of both NO and soot emission using measured in-cylinder pressure and manifold conditions (e.g. humidity and EGR flow). This sensor is further developed to improve accuracy, especially for high EGR operating conditions, and to reduce the computational effort.

Heat release

model

Emission

model

NO/NO2

Correlation

PM

correlation

Diagnostics

pcyl

CA

IMEP

pmax

dp/dα

ROHR

NO

soot

NOx

PM

Fuelling Intake

manifold

conditions

HC

sulphur

Figure 11: Block scheme of virtual heat release and emission sensor

The emission sensor is coupled to the virtual heat release sensor, as illustrated in Figure 11. The original emission model is extended, such that NOx and PM emissions are predicted. In this work, empirically-based corrections are used to compute the corresponding NOx (i.e. including NO2) and PM emissions. The NO2 part is determined from mapped NO/NO2 ratios over the engine operating range. For PM emissions, an empirical correlation is used, which is especially designed for the studied engine. This commonly applied correlation determines PM emission from measured smoke values, HC emissions, and sulfur content in the fuel. In this emission sensor, the smoke values are assumed to describe the carbon part (i.e. soot mass) in the PM correlation. Therefore, this part is replaced by the predicted soot mass. The soluble organic fraction is computed from mapped HC emissions. The emission model uses the estimated heat release profile as main input. At present, the virtual emission sensor is successfully implemented for real-time control on a desktop PC.

To meet future emission targets, it becomes increasingly important to optimize the synergy between engine and aftertreatment: Integrated Emission Management (IEM). Main focus is to realize minimal fuel consumption within the limits set by emission legislation at any instant. For this concept, the on-line availability of NOx and PM information is crucial. Examples of IEM are thermal management for DPF regeneration and integrated NOx control by EGR/SCR-balancing. Furthermore, virtual sensing enables more direct emission control. Instead of controlling mass flow rate or boost pressure to control NOx emissions, as is the case in conventional EGR/VTG control, the control strategy can be directly based on NOx emission data. This will result in reduced response time and will therefore enhance transient emission control. Running the virtual emission sensor in parallel to existing observers on the engine will enable to monitor ageing and malfunction of engine components more closely, as model parameters are seen to be drifting. This will

Page 13 of 17

enhance the OBD system performance and will better ensure In-Use Compliance to emission and OBD legislation under varying conditions (ambient conditions, fuel quality) and engine state (aging, wear and production tolerances).

PREDICTION ACCURACY OF VIRTUAL EMISSION SENSOR

For the studied engine, total engine-out NOx and PM emissions are predicted as the sum of the contributions of each individual cylinder. In this work, no distinction in manifold and fueling conditions is made for the individual cylinders. Figure 12 shows the measured and predicted NOx and PM emissions for injection timing sweeps in two different operating points with fixed VTG and EGR valve positions. The error bands indicate the deviation in emissions that would occur if only one cylinder would be used to predict the total engine-out emission (i.e. cylinder emission value times 6). Although the NOx emissions are very low, the virtual emission sensor is well capable of predicting the absolute NOx emission level for both EGR levels at the varying timings. Predicted NOx emissions are typically within 12% of measured values, which is close to the measurement accuracy of commercially available NOx sensors. This accuracy is sufficient to allow the application of the virtual emission sensor in model-based engine/aftertreatment control algorithms, e.g. for SCR catalyst control.

-5 0 5 10 15 200

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

CA50 [o ca aTDC]

NO

x em

issio

n [g

/kW

h]

51%50%

50% 50% 48%

42.5%

43%

47%46%

42%

Simulated IMEP 6.8 bar @ 1200 rpm

Simulated IMEP 6.1 bar @ 1830 rpm

Measured IMEP 6.8 bar @ 1200 rpm

Measured IMEP 6.1 bar @ 1830 rpm

-2 0 2 4 6 8 10 12 14 16 180

0.2

0.4

0.6

0.8

1

CA50 [o ca aTDC]

PM

em

issio

n [g

/kW

h]

Simulated IMEP 6.8 bar @ 1200 rpm

Simulated IMEP 6.1 bar @ 1830 rpm

Measured IMEP 6.8 bar @ 1200 rpm

Measured IMEP 6.1 bar @ 1830 rpm

Figure 12: Measured and predicted NOx and PM emissions. Error bands indicate deviations between individual cylinders. Depicted percentages denote the actual EGR percentage.

The predicted PM emissions show good qualitative agreement with measurements. This indicates that the virtual emission sensor can capture the sensitivity of PM emission to important operating variables, such as injection timing and EGR rate. Although this is an important property for control, application of the virtual PM sensor in model-based control strategies is not justified at present. With absolute values deviating between 2% up to a factor of 2 from measured values, accurate DPF soot loading and emission control is not possible yet.

Page 14 of 17

The predicted instantaneous in-cylinder soot mass is very sensitive to the balance between the mass of fuel injected and the mass of fuel already burned. The latter is computed for each individual cylinder from the heat release rate derived from the measured in-cylinder pressure. For the fueling rate, however, all cylinders use an injection profile that is based on the total fuel flow to the engine. This will result in an inaccuracy in the aforementioned balance. Part of the observed deviation between predicted and measured PM emission is therefore believed to originate from the fact that no distinction is made in fueling for the individual cylinders. Using cylinder specific fueling data will increase the absolute accuracy, but comes at the cost of additional sensors.

OUTLOOK

For future emission legislation, it becomes more and more important to meet the further reducing emission targets under all operating conditions. Therefore, growing attention is paid to performance optimization and robustness enhancement. At TNO, an advanced combustion model is developed that can predict engine out emissions as well as in-cylinder pressure. This model is based on [23] and requires the following input data: fueling parameters (quantity, timing, injection pressure) and manifold conditions (e.g. EGR rate, pressure, temperature). The manifold conditions are used to determine the initial in-cylinder conditions of pressure, temperature and composition.

Virtual sensor

Engine control

Model-

based

set-point

generator

Torque control

EGR/VTG

control

CA50/IMEP

controlDesired

powerFuelling

VVA

EGR

VTG

Cylinder

pressure

Air flow

Pressure

Temperature

Advanced

combustion

model

Power output

Fuel consumption

Emissions

CA50, IMEP, ISFC

dp/dα, pmax

NOx, NO2, PM

Diagnostics

Figure 13: Illustration of a model-based engine control system

VIRTUAL SENSORS

Future engine control systems can greatly benefit from a model-based control approach. This is illustrated in Figure 13. By running the advanced combustion model in the virtual sensor, more detailed information about

Page 15 of 17

the actual engine state can be on-line determined, including in-cylinder pressure. Note that closed-loop combustion control can not fully rely on virtual sensors. For example, the influence of fuel quality can only be determined when in-cylinder pressure and emission data are available, as reference to recalibrate model parameters. The availability of a virtual in-cylinder pressure sensor will, however, have significant cost advantages: it will reduce the need to equip all individual cylinders with an in-cylinder pressure sensor.

Preliminary studies have shown that the virtual in-cylinder pressure sensor (with heat release computation and virtual emission sensor) is capable of predicting the CA50 with an absolute accuracy of 0.5 oCA. This accuracy is of the same order as obtained for CA50 values derived from measured in-cylinder pressure data and is therefore, sufficiently accurate for combustion phasing control. Using this sensor, important engine and aftertreatment control variables, such as the exhaust gas temperature, maximum in-cylinder pressure (rise) and

specific fuel consumption are expected to be predicted with a relative accuracy of ≤ 4%. This level of accuracy validates the use of the sensor in model-based control applications. The virtual in-cylinder pressure sensor will be presented in more detail in separate publication.

The use of the virtual in-cylinder pressure sensor will further enhance the OBD system performance. Now that the model is able to describe the interaction between fueling and engine performance (torque) as well as emissions (NOx, PM), the source of an observed deviation from reference values can be located more accurately

OFF-LINE OPTIMIZATION

The advanced combustion model can also be used to perform concept studies on engine and aftertreatment configuration and control design. Furthermore, the developed model also has added value as enabler for off-line controller calibration. This model can be used to set-up a rough calibration of the required engine maps that are implemented in the engine control algorithm. With this model-aided calibration approach, engine test bed time can be limited to fine-tuning of the rough calibration, reducing calibration effort, time and hence costs. Besides this, the area of interest for testing can be determined a priori: Design-of-Experiments (DOE). This will also reduce the number of costly engine measurements.

ON-LINE OPTIMIZATION

The predictive capability of the advanced combustion model makes it possible to on-line determine and optimize the control action before execution, see also Hadler et al. [10]. Instead of using fixed open-loop control target values, the most optimal value for the control actions can be generated based on available information. During optimization of the control point, also the preferred actuator for the required control action can be evaluated. This will further enhance transient torque and emission control, since different actuators are preferred for fast and slow transients. For example, changes in injection timing (fast response) and EGR rate (slow response) can both be used to reduce NOx emissions.

CONCLUSIONS

The potential of cylinder pressure-based control is examined in a heavy-duty engine for late DI diesel combustion with high EGR rates. To demonstrate transient performance and robustness of the controller, fuel steps and experiments with uncalibrated injectors and different fuels are done. Furthermore, a new virtual NOx and PM emission sensor is introduced, which is suitable for use in current state-of-the-art closed-loop combustion control systems.

From this study, the following is concluded:

Page 16 of 17

• Closed-loop combustion control system is successfully implemented on a six cylinder heavy-duty diesel engine. This real-time control system is able to simultaneously control CA50 and IMEP for both individual cylinders (cycle-to-cycle-control) and all cylinders in parallel (cylinder balancing);

• Fuel step responses indicate that the control response is sufficiently fast to follow transient conditions;

• Auto-calibration features of the CPBC system are demonstrated for varying fuel injector performance and fuel composition. This dramatically reduces the calibration time and costs of engine control systems. In addition, the closed-loop control system can deal with varying ambient conditions, ageing and wear during life time. This will better ensure In-Use Compliance;

• The applied combustion control system is capable to realize the desired combustion phasing and power output. However, it cannot fully compensate for changes in emissions (NOx and PM). In order to do so, emission data has to be available on-line;

• A new virtual NOx and PM sensor is successfully implemented for real-time control of engine-out emissions. For two timing sweeps and high EGR rates, the NOx prediction accuracy is shown to be comparable with commercially available NOx sensors. For PM emissions, only the qualitative accuracy is acceptable. Absolute accuracy is expected to increase when more detailed information on individual cylinder fueling rates is available.

REFERENCES

1. Kathi Epping, Salvador Aceves, Richard Bechthold, John Dec, The potential of HCCI combustion for high efficiency and low emissions, SAE 2002-01-1923

2. Homogeneous Charge Compression Ignition (HCCI) engines: Key research and development issues, SAE publication PT-94, edited by Frank Zhao et al., 2003

3. Bengtsson, P. Strandh, R. Johansson, P. Tunestål, B. Johansson, Hybrid Control of Homogeneous Charge Compression Ignition (HCCI) Engine Dynamics, International Journal of Control, Vol. 79, No. 5, 2006, pp 422-448.

4. T. Ryan, Engine controls; future requirements and challenges, presentation at IEEE/ ASME/ SAE Workshop on Open Problems and Challenges in Automotive Control, 2008.

5. Maru Yoon, Kangyoon Lee, Myoungho Sunwoo, Byuounggul Oh, Cylinder pressure based combustion phasing control of a CRDI diesel engine, SAE 2007-01-0772

6. Shigeki Nakayama, Taku Ibuki, Hitoshi Hosaki,, Hiroyuki Tominaga, An application of model based combustion control to transient cycle-to-cycle diesel combustion, SAE 2008-01-1311

7. Matthias Lamping, Thomas Körfer, Thorsten Schnorbus, Stefan Pischinger, Yunji Chen, Tomorrows diesel fuel diversity – challenges and solutions, SAE 2008-01-1731

8. Ming Zeng, Yuyu Tan, Graham Reader, Usman Asad, Xiaoye Han, Meiping Wang, Prompt heat release analysis to improve diesel low temperature combustion, SAE 2009-01-1883

9. New GM V6 diesel has cylinder-pressure monitoring, SAE tech briefs, 2007, www.sae.org/automag/techbriefs/04-2007/1-115-4-22.pdf

10. Hadler et al., Volkswagen's new 2.0l TDI engine fulfils the most stringent emission standards, 29th International Vienna Motor Symposium, 2008

11. Jens Jeschke, Thomas Lang, Jürgen Wendt, Dieter Mannigel, Michael Henn, Hans-Georg Nitzke, Combustion Control for Diesel Engines with Direct Injection, 16th Aachen Colloquium Automobile and Engine Technology, 2007

Page 17 of 17

12. John Pinson, Bringing the low NOx diesel under control, presentation at the Diesel Engine Emission Reduction (DEER) conference, Detroit, 2006

13. M. Hillion, H. Buhlbuck, J. Chauvin, N. Petit, Combustion control of diesel engines using injection timing, SAE 2009-01-0367

14. Jason Souder, Closed-loop control of a multi-cylinder HCCI engine, SAE HCCI symposium, 2004 15. Kevin Duffy, Eric Fluga, Steve Faulkner, David Heaton, Charles Schleyer, Rafal Sobotowski; Latest

developments in heavy duty Diesel HCCI, in proceeding of IFP international conference Which Fuels For Low CO2 Engines?; P. Duret (Editor), Editions TECHNIP, Paris, pp. 1-10, 2004

16. Johan Bengtsson, Petter Strandh, Rolf Johansson, Per Tunestål and Bengt Johansson, Multi-Output Control of a Heavy Duty HCCI Engine Using Variable Valve Actuation and Model Predictive Control, SAE 2006-01-0873

17. R. Reitz, J. von der Ehe, Use of in-cylinder pressure measurement and the response surface method for combustion feedback control in a diesel engine, Proc. IMechE Vol. 220, Part D: J. Automobile Engineering, pp. 1657-1666, 2006

18. H. Husted, D. Kruger, G. Fattic, G. Ripley and E. Kelly, Cylinder Pressure-Based Control of Pre-Mixed Diesel Combustion, SAE 2007-01-0773

19. Magnus Lewander, Bengt Johansson, Per Tunestål, Nathan Keeler, Simon Tullis, Nebosja Milovanovic, Pär Bergstrand, Evaluation of the operating range of partially premixed combustion in a multi cylinder heavy duty engine with extensive EGR, SAE 2009-01-1127

20. X.L.J. Seykens, R.S.G. Baert, L.M.T. Somers, F.P.T. Willems, Experimental Validation of Extended NO and Soot Model for Advanced HD Diesel Engine Combustion, SAE 2009-01-0683

21. Mark van Aken, Frank Willems, Dirk-Jan de Jong, Appliance of high EGR rates with a short and long route EGR system on a heavy-duty diesel engine, SAE 2007-01-0906

22. Carl Wilhelmsson, Per Tunestål, Bengt Johansson, Model Based Engine Control Using ASICs: A Virtual Heat Release Sensor, published in: Les Rencontres Scientifiques de l'IFP: "New Trends in Engine Control, Simulation and Modelling", Paris 2006

23. X.L.J. Seykens, Development and validation of a phenomenological diesel engine combustion model, Ph.D. thesis, Eindhoven University of Technology, The Netherlands, February 2010

CONTACT INFORMATION

Frank Willems, Ph.D. TNO Automotive P.O. Box 756 5700 AT Helmond The Netherlands Tel: +31 15 2697471 Email: [email protected]

ACKNOWLEDGMENTS

This research is done in the framework of the PREDUCE research program (Program for Reduction of Emissions from Diesel towards Ultra Clean Engines). This program is partially funded by the Dutch Ministry of Economical affairs.