Modeling of a Diesel Engine with VGT and EGR

including Oxygen Mass Fraction

Johan Wahlstrom and Lars Eriksson
Vehicular systems
Department of Electrical Engineering
Linkopings universitet, SE-581 83 Linkoping, Sweden
WWW: www.vehicular.isy.liu.se
E-mail: {johwa, larer}@isy.liu.se
Report: LiTH-ISY-R-2747
September 27, 2006
When citing this work, it is recommended that the citation is the
improved and extended work, published in the peer-reviewed
article Johan Wahlstrom and Lars Eriksson, Modeling diesel
engines with a variable-geometry turbocharger and exhaust gas
recirculation by optimization of model parameters for capturing
non-linear system dynamics, Proceedings of the Institution of
Mechanical Engineers, Part D, Journal of Automobile Engineering,
Volume 225, Issue 7, July 2011,
http://dx.doi.org/10.1177/0954407011398177.
Abstract
A mean value model of a diesel engine with VGT and EGR and that includes
oxygen mass fraction is developed and validated. The intended model applica-
tions are system analysis, simulation, and development of model-based control
systems. Model equations and tuning methods are described for each subsys-
tem in the model. In order to decrease the amount of tuning parameters, ows
and eciencies are modeled using physical relationships and parametric mod-
els instead of look-up tables. The static models have mean relative errors that
are equal to or lower than 6.1 %. Static and dynamic validations of the entire
model show that the mean relative errors are less than 12 %. The validations
also show that the proposed model captures the essential system properties, i.e.
a non-minimum phase behavior in the transfer function EGR-valve to intake
manifold pressure and a non-minimum phase behavior, an overshoot, and a sign
reversal in the transfer function VGT to compressor mass ow.
Contents
1 Introduction 4
1.1 Model structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.2 Measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.2.1 Stationary measurements . . . . . . . . . . . . . . . . . . 5
1.2.2 Dynamic measurements . . . . . . . . . . . . . . . . . . . 6
1.3 Parameter estimation . . . . . . . . . . . . . . . . . . . . . . . . 6
1.4 Relative error . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.5 Outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2 Manifolds 8
3 Cylinder 9
3.1 Cylinder ow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
3.2 Cylinder out temperature . . . . . . . . . . . . . . . . . . . . . . 10
3.3 Cylinder torque . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
4 EGR-valve 16
5 Turbocharger 19
5.1 Turbo inertia . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
5.2 Turbine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
5.2.1 Turbine eciency . . . . . . . . . . . . . . . . . . . . . . . 19
5.2.2 Turbine mass ow . . . . . . . . . . . . . . . . . . . . . . 21
5.3 Compressor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
5.3.1 Compressor eciency . . . . . . . . . . . . . . . . . . . . 24
5.3.2 Compressor mass ow . . . . . . . . . . . . . . . . . . . . 25
5.3.3 Compressor map . . . . . . . . . . . . . . . . . . . . . . . 27
6 Intercooler and EGR-cooler 29
7 Summary of assumptions and model equations 30
7.1 Assumptions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
7.2 Manifolds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
7.3 Cylinder . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
7.3.1 Cylinder ow . . . . . . . . . . . . . . . . . . . . . . . . . 31
7.3.2 Cylinder out temperature . . . . . . . . . . . . . . . . . . 31
7.3.3 Cylinder torque . . . . . . . . . . . . . . . . . . . . . . . . 31
7.4 EGR-valve . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
7.5 Turbo . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
7.5.1 Turbo inertia . . . . . . . . . . . . . . . . . . . . . . . . . 32
7.5.2 Turbine eciency . . . . . . . . . . . . . . . . . . . . . . . 32
7.5.3 Turbine mass ow . . . . . . . . . . . . . . . . . . . . . . 33
7.5.4 Compressor eciency . . . . . . . . . . . . . . . . . . . . 33
7.5.5 Compressor mass ow . . . . . . . . . . . . . . . . . . . . 33
8 Model tuning and validation 34
8.1 Tuning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
8.2 Validation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
2
9 Conclusions 39
A Notation 43
3
1 Introduction
Legislated emission limits for heavy duty trucks are constantly reduced. To fulll
the requirements, technologies like Exhaust Gas Recirculation (EGR) systems
and Variable Geometry Turbochargers (VGT) have been introduced. The pri-
mary emission reduction mechanisms utilized to control the emissions are that
NO
x
can be reduced by increasing the intake manifold EGR-fraction and smoke
can be reduced by increasing the air/fuel ratio (Heywood, 1988). However the
EGR fraction and air/fuel ratio depend in complicated ways on the EGR and
VGT actuation. It is therefore necessary to have coordinated control of the EGR
and VGT to reach the legislated emission limits in NO
x
and smoke. When de-
veloping and validating a controller for this system, it is desirable to have a
model that describes the system dynamics and the nonlinear eects. Therefore,
the objective of this report is to construct a mean value diesel engine model with
VGT and EGR. The model should be able to describe stationary operations and
dynamics that are important for gas ow control. The intended usage of the
model are system analysis, simulation and development of model-based control
systems. In order to decrease the amount of tuning parameters, ows and e-
ciencies are modeled based upon physical relationships and parametric models
instead of look-up tables. The model is implemented in Matlab/Simulink
using a component library.
1.1 Model structure
The structure of the model can be seen in Fig. 1. To be able to implement a
model-based controller in a control system the model must be small. Therefore
the model has only seven states: intake and exhaust manifold pressures (p
im
and p
em
), oxygen mass fraction in the intake and exhaust manifold (X
Oim
and X
Oem
), turbocharger speed (
t
), and two states describing the actuator
dynamics for the two control signals ( u
egr
and u
vgt
). These states are collected
in a state vector x
x = (p
im
p
em
X
Oim
X
Oem

t
u
egr
u
vgt
)
T
(1)
Descriptions of the nomenclature, the variables and the indices can be found in
Appendix A.
The modeling eort is focused on the gas ows, and it is important that
the model can be utilized both for dierent vehicles and for engine testing,
calibration, and certication in an engine test cell. In many of these situations
the engine operation is dened by the rotational speed n
e
, for example given
as an input from a drivecycle, and therefore it is natural to parameterize the
model using engine speed. The resulting model is thus expressed in state space
form as
x = f(x, u, n
e
) (2)
where the engine speed n
e
is considered as an input to the model, and u is the
control input vector
u = (u

u
egr
u
vgt
)
T
(3)
which contains mass of injected fuel u

, EGR-valve position u
egr
, and VGT
actuator position u
vgt
. The EGR-valve is closed when u
egr
= 0 % and open
when u
egr
= 100 %. The VGT is closed when u
vgt
= 0 % and open when
u
vgt
= 100 %.
4
EGR cooler
Exhaust
manifold
Compressor Intercooler
Cylinders
Turbine
EGR valve
Intake
manifold
W
egr
W
ei
W
eo
p
em
X
Oem
X
Oim
p
im
u

W
t
W
c
u
vgt
u
egr

t
Figure 1: A model structure of the diesel engine. It has three control inputs
and ve main states related to the engine (p
im
, p
em
, X
Oim
, X
Oem
, and
t
). In
addition, there are two states for actuator dynamics ( u
egr
and u
vgt
).
1.2 Measurements
To tune and validate the model, stationary and dynamic measurements have
been performed in an engine laboratory at Scania CV AB, and these are de-
scribed below.
1.2.1 Stationary measurements
The stationary data consists of measurements at stationary conditions in 82
operating points, that are scattered over a large operating region covering dif-
ferent loads, speeds, VGT- and EGR-positions. These 82 operating points also
include the European Stationary Cycle (ESC). The variables that were mea-
sured during stationary measurements can be seen in Tab. 1. The EGR fraction
is calculated by measuring the carbon dioxide concentration in the intake and
exhaust manifolds.
5
Table 1: Measured variables during stationary measurements.
Variable Description Unit
M
e
Engine torque Nm
n
e
Rotational engine speed rpm
n
t
Rotational turbine speed rpm
p
amb
Ambient pressure Pa
p
em
Exhaust manifold pressure Pa
p
im
Intake manifold pressure Pa
T
amb
Ambient temperature K
T
c
Temperature after compressor K
T
em
Exhaust manifold temperature K
T
im
Intake manifold temperature K
T
t
Temperature after turbine K
u
egr
EGR control signal. 0 - closed, 100 - open %
u
vgt
VGT control signal. 0 - closed, 100 - open %
u

Injected amount of fuel mg/cycle

W
c
Compressor mass ow kg/s
x
egr
EGR fraction
Table 2: Measured variables during dynamic measurements.
Variable Description Unit
M
e
Engine torque Nm
n
e
Rotational engine speed rpm
n
t
Rotational turbine speed rpm
p
em
Exhaust manifold pressure Pa
p
im
Intake manifold pressure Pa
u
egr
EGR control signal. 0 - closed, 100 - open %
u
vgt
VGT control signal. 0 - closed, 100 - open %
u

Injected amount of fuel mg/cycle

W
c
Compressor mass ow kg/s
1.2.2 Dynamic measurements
The dynamic data consists of measurements at dynamic conditions with steps in
VGT control signal, EGR control signal, and fuel injection in several dierent
operating points. The measurements are sampled with a frequency of 1 Hz,
except for the steps in fuel injection where the measurements are sampled with
a frequency of 10 Hz. These measurements are used in Sec. 8 for tuning of
dynamic models and validation of the total engine model. The variables that
were measured during dynamic measurements can be seen in Tab. 2.
1.3 Parameter estimation
Parameters in static models are estimated automatically using least squares
optimization and data from stationary measurements. Parameters in dynamic
models (volumes and an inertia) are estimated by adjusting these parameters
manually until simulations of the complete model follow the dynamic responses
in the dynamic measurements.
6
1.4 Relative error
Relative errors are calculated and used to evaluate the tuning and the validation
of the model. Relative errors for stationary measurements between a measured
variable y
meas,stat
and a modeled variable y
mod,stat
are calculated as
stationary relative error(i) =
y
meas,stat
(i) y
mod,stat
(i)
1
N

N
i=1
y
meas,stat
(i)
(4)
where i is an operating point. Relative errors for dynamic measurements be-
tween a measured variable y
meas,dyn
and a modeled variable y
mod,dyn
are cal-
culated as
dynamic relative error(j) =
y
meas,dyn
(j) y
mod,dyn
(j)
1
N

N
i=1
y
meas,stat
(i)
(5)
where j is a time sample. In order to make a fair comparison between these
relative errors, both the stationary and the dynamic relative error have the
same stationary measurement in the denominator and the mean value of this
stationary measurement is calculated in order to avoid large relative errors when
y
meas,stat
is small.
1.5 Outline
The outline of the report is as follows. Sec. 2 describes the model equations for
the intake and exhaust manifold. The cylinder ows, cylinder temperature, and
cylinder torque are modeled in Sec. 3. In Sec. 4 a model of the EGR-valve is
proposed and in Sec. 5 model equations for the turbocharger are described. The
intercooler and EGR-cooler are modeled in Sec. 6. A summary of the model
assumptions and the model equations is given in Sec. 7. Tuning and validation
of the model are performed in Sec. 8. Finally, conclusions are drawn in Sec. 9.
7
2 Manifolds
The intake and exhaust manifolds are modeled as dynamic systems with two
states each, pressure and oxygen mass fraction. The standard isothermal model
(Heywood, 1988), that is based upon mass conservation, the ideal gas law, and
that the manifold temperature is constant or varies slowly, has the dierential
equations for the manifold pressures
d
dt
p
im
=
R
a
T
im
V
im
(W
c
+ W
egr
W
ei
)
d
dt
p
em
=
R
e
T
em
V
em
(W
eo
W
t
W
egr
)
(6)
There are two sets of thermodynamic properties: air has the ideal gas constant
R
a
and the specic heat capacity ratio
a
, and exhaust gas has the ideal gas
constant R
e
and the specic heat capacity ratio
e
. The intake manifold tem-
perature T
im
is assumed to be constant and equal to the cooling temperature
in the intercooler, the exhaust manifold temperature T
em
will be described in
Sec. 3.2, and V
im
and V
em
are the manifold volumes. The mass ows W
c
, W
egr
,
W
ei
, W
eo
, and W
t
will be described in Sec. 3 to 5.
The EGR fraction in the intake manifold is calculated as
x
egr
=
W
egr
W
c
+ W
egr
(7)
Note that the EGR gas also contains oxygen that aects the oxygen fuel ratio
in the cylinder. This eect is considered by modeling the oxygen concentrations
X
Oim
and X
Oem
in the control volumes. These concentrations are dened
as (Vigild, 2001)
X
Oim
=
m
Oim
m
totim
, X
Oem
=
m
Oem
m
totem
(8)
where m
Oim
and m
Oem
are the oxygen masses, and m
totim
and m
totem
are the
total masses in the intake and exhaust manifolds. Dierentiating X
Oim
and
X
Oem
and using mass conservation (Vigild, 2001) give the following dierential
equations
d
dt
X
Oim
=
R
a
T
im
p
im
V
im
((X
Oem
X
Oim
) W
egr
+ (X
Oc
X
Oim
) W
c
)
d
dt
X
Oem
=
R
e
T
em
p
em
V
em
(X
Oe
X
Oem
) W
eo
(9)
where X
Oc
is the constant oxygen concentration in air passing the compressor,
i.e. X
Oc
= 23.14%, and X
Oe
is the oxygen concentration in the exhaust gases
out from the engine cylinders, X
Oe
will be described in Sec. 3.1.
Tuning parameters
V
im
and V
em
: manifold volumes.
Tuning method
The tuning parameters V
im
and V
em
are obtained by adjusting these parameters
manually until simulations of the complete model follow the dynamic responses
in the dynamic measurements, see Sec. 8.1.
8
3 Cylinder
Three sub-models describe the behavior of the cylinder, these are:
A mass ow model that models the ows through the cylinder, the oxygen
to fuel ratio, and the oxygen concentration out from the cylinder.
A model of the cylinder out temperature.
A cylinder torque model.
3.1 Cylinder ow
The total mass ow W
ei
into the cylinders is modeled using the volumetric
eciency
vol
(Heywood, 1988)
W
ei
=

vol
p
im
n
e
V
d
120 R
a
T
im
(10)
where p
im
and T
im
are the pressure and temperature in the intake manifold, n
e
is the engine speed and V
d
is the displaced volume. The volumetric eciency is
in its turn modeled as

vol
= c
vol1

p
im
+ c
vol2

n
e
+ c
vol3
(11)
The fuel mass ow W
f
into the cylinders is controlled by u

, which gives the

injected mass of fuel in mg per cycle and cylinder
W
f
=
10
6
120
u

n
e
n
cyl
(12)
where n
cyl
is the number of cylinders. The mass ow W
eo
out from the cylinder
is given by the mass balance as
W
eo
= W
f
+ W
ei
(13)
The oxygen to fuel ratio
O
in the cylinder is dened as

O
=
W
ei
X
Oim
W
f
(O/F)
s
(14)
where (O/F)
s
is the stoichiometric relation between oxygen and fuel masses.
During the combustion, the oxygen is burned in the presence of fuel. In
diesel engines
O
> 1 to avoid smoke. Therefore, it is assumed that
O
> 1 and
the oxygen concentration out from the cylinder can then be calculated as the
unburned oxygen fraction
X
Oe
=
W
ei
X
Oim
W
f
(O/F)
s
W
eo
(15)
Tuning parameters
c
vol1
, c
vol2
, c
vol3
: volumetric eciency constants
9
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
W
e
i

[
k
g
/
s
]
Modeled W
ei
[kg/s]


Modeled
Calculated from measurements
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
3
2
1
0
1
2
3
r
e
l

e
r
r
o
r

W
e
i

[
%
]
Modeled W
ei
[kg/s]
mean abs rel error: 0.9% max abs rel error: 2.5%
Figure 2: Top: Comparison of modeled mass ow W
ei
into the cylinders and
estimated W
ei
from measurements. Bottom: Relative errors for modeled W
ei
as function of modeled W
ei
at steady state.
Tuning method
The tuning parameters c
vol1
, c
vol2
, and c
vol3
are obtained by solving a lin-
ear least-squares problem that minimizes (W
ei
W
ei,meas
)
2
with c
vol1
, c
vol2
,
and c
vol3
as the optimization variables. The variable W
ei
is the model in
Eq. (10) and (11) and W
ei,meas
is estimated from stationary measurements as
W
ei,meas
= W
c
/(1 x
egr
). Stationary measurements are used as inputs to the
model during the tuning and the result can be seen in Fig. 2, which compares
W
ei
and W
ei,meas
.
3.2 Cylinder out temperature
The cylinder out temperature T
e
is modeled in the same way as in Skogtj arn
(2002). This approach is based upon ideal gas Seliger cycle calculations that
give the cylinder out temperature
T
e
=
sc

11/a
e
r
1a
c
x
1/a1
p
_
q
in
_
1 x
cv
c
pa
+
x
cv
c
va
_
+ T
1
r
a1
c
_
(16)
where
sc
is a compensation factor for non ideal cycles and x
cv
the ratio of fuel
consumed during constant volume combustion. The rest of the fuel (1 x
cv
) is
used during constant pressure combustion. Further, this model consists of the
pressure quotient over the cylinder

e
=
p
em
p
im
(17)
10
the pressure quotient between point 3 (after combustion) and point 2 (before
combustion) in the Seliger cycle
x
p
=
p
3
p
2
= 1 +
q
in
x
cv
c
va
T
1
r
a1
c
(18)
the specic energy contents of the charge
q
in
=
W
f
q
HV
W
ei
+ W
f
(1 x
r
) (19)
the temperature at inlet valve closing after intake stroke and mixing
T
1
= x
r
T
e
+ (1 x
r
) T
im
(20)
the residual gas fraction
x
r
=

1/a
e
x
1/a
p
r
c
x
v
(21)
and the volume quotient between point 3 (after combustion) and point 2 (before
combustion) in the Seliger cycle
x
v
=
v
3
v
2
= 1 +
q
in
(1 x
cv
)
c
pa
_
qin xcv
cva
+ T
1
r
a1
c
_ (22)
Since the equations above are non-linear and depend on each other, the cylinder
out temperature is calculated numerically using a xed point iteration which
starts with the initial values x
r,0
and T
1,0
. Then the following equations are
applied in each iteration k
q
in,k+1
=
W
f
q
HV
W
ei
+ W
f
(1 x
r,k
)
x
p,k+1
= 1 +
q
in,k+1
x
cv
c
va
T
1,k
r
a1
c
x
v,k+1
= 1 +
q
in,k+1
(1 x
cv
)
c
pa
_
q
in,k+1
xcv
cva
+ T
1,k
r
a1
c
_
x
r,k+1
=

1/a
e
x
1/a
p,k+1
r
c
x
v,k+1
T
e,k+1
=
sc

11/a
e
r
1a
c
x
1/a1
p,k+1
_
q
in,k+1
_
1 x
cv
c
pa
+
x
cv
c
va
_
+ T
1,k
r
a1
c
_
T
1,k+1
= x
r,k+1
T
e,k+1
+ (1 x
r,k+1
) T
im
(23)
In each sample during dynamic simulation, the initial values x
r,0
and T
1,0
are
set to the solutions of x
r
and T
1
from the previous sample.
Exhaust manifold temperature
The cylinder out temperature model above does not describe the exhaust man-
ifold temperature completely due to heat losses. This is illustrated in Fig. 3(a)
11
which shows a comparison between measured and modeled exhaust manifold
temperature and in this gure it is assumed that the exhaust manifold temper-
ature is equal to the cylinder out temperature, i.e. T
em
= T
e
. The relative error
between model and measurement seems to increase from a negative error to a
positive error for increasing mass ow W
eo
out from the cylinder. The exhaust
manifold temperature is measured in the exhaust manifold, thus the heat losses
to the surroundings in the exhaust pipes between the cylinder and the exhaust
manifold must be taken into consideration.
This temperature drop is modeled as a function of mass ow out from the
cylinder, see Model 1 in Eriksson (2002).
T
em
= T
amb
+ (T
e
T
amb
) e

h
tot
d
pipe
l
pipe
n
pipe
Weo cpe
(24)
where T
amb
is the ambient temperature, h
tot
the total heat transfer coecient,
d
pipe
the pipe diameter, l
pipe
the pipe length and n
pipe
the number of pipes.
Using this model, the mean and maximum absolute relative error is reduced,
see Fig. 3(b).
Approximating the solution to the cylinder out temperature
As explained above, the cylinder out temperature is calculated numerically using
the xed point iteration Eq. (23). Fig. 4 shows that it is sucient to use one
iteration in this iterative process. This is shown by comparing the solution
from one iteration with one that has a sucient number of iterations to give
a solution with 0.01 % accuracy. The maximum absolute relative error of the
solution from one iteration (compared to the solution with 0.01 % accuracy)
is 0.15 %. This error is small because the xed point iteration Eq. (23) has
initial values that are close to the solution. Consequently, it is sucient to use
one iteration in this model since the mean absolute relative error of the exhaust
manifold temperature model (compared to the measurements in Fig. 3(b)) is
1.7 %.
Tuning parameters

sc
: compensation factor for non ideal cycles
x
cv
: the ratio of fuel consumed during constant volume combustion
h
tot
: the total heat transfer coecient
Tuning method
The tuning parameters
sc
, x
cv
, and h
tot
are obtained by solving a non-linear
least-squares problem that minimizes (T
em
T
em,meas
)
2
with
sc
, x
cv
, and h
tot
as the optimization variables. The variable T
em
is the model in Eq. (23) and (24)
with stationary measurements as inputs to the model, and T
em,meas
is a sta-
tionary measurement. The result of the tuning is shown in Fig. 3(b).
12
550 600 650 700 750 800 850 900
500
600
700
800
900
T
e
m

[
K
]
Modeled T
em
[K]
Modeled
Measured
0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 0.55 0.6 0.65
10
5
0
5
10
15
r
e
l

e
r
r
o
r

T
e
m

[
%
]
W
eo
[kg/s]
mean abs rel error: 2.8% max abs rel error: 10.2%
(a) Without a model for heat losses in the exhaust pipes, i.e. Tem = Te.
550 600 650 700 750 800 850
500
600
700
800
900
T
e
m

[
K
]
Modeled T
em
[K]
Modeled
Measured
0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 0.55 0.6 0.65
6
4
2
0
2
4
6
r
e
l

e
r
r
o
r

T
e
m

[
%
]
W
eo
[kg/s]
mean abs rel error: 1.7% max abs rel error: 5.4%
(b) With model (24) for heat losses in the exhaust pipes.
Figure 3: Modeled and measured exhaust manifold temperature T
em
and rela-
tive errors for modeled T
em
at steady state.
13
510 512 514 516 518 520 522 524 526 528 530
200
400
600
800
1000
1200
1400
T
e

[
K
]
One iteration
0.01 % accuracy
510 512 514 516 518 520 522 524 526 528 530
0.2
0.15
0.1
0.05
0
0.05
0.1
0.15
r
e
l

e
r
r
o
r

[
%
]
Time [s]
Figure 4: The cylinder out temperature T
e
is calculated by simulating the total
engine model during the complete European Transient Cycle. This gure shows
the part of the European Transient Cycle that consists of the maximum relative
error. Top: The xed point iteration Eq. (23) is used in two ways: by using
one iteration and to get 0.01 % accuracy. Bottom: Relative errors between the
solutions from one iteration and 0.01 % accuracy.
3.3 Cylinder torque
The torque produced by the engine M
e
is modeled using three dierent engine
components; the gross indicated torque M
ig
, the pumping torque M
p
, and the
friction torque M
fric
(Heywood, 1988).
M
e
= M
ig
M
p
M
fric
(25)
The pumping torque is modeled using the intake and exhaust manifold pressures.
M
p
=
V
d
4
(p
em
p
im
) (26)
The gross indicated torque is coupled to the energy that comes from the fuel
M
ig
=
u

10
6
n
cyl
q
HV

ig
4
(27)
Assuming that the engine is always running at optimal injection timing, the
gross indicated eciency
ig
is modeled as

ig
=
igch
_
1
1
r

cyl
1
c
_
(28)
14
200 400 600 800 1000 1200 1400 1600 1800 2000 2200
0
500
1000
1500
2000
2500
M
e

[
N
m
]
Modeled M
e
[Nm]
Modeled
Measured
0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 0.55 0.6
8
6
4
2
0
2
4
r
e
l

e
r
r
o
r

M
e

[
%
]
W
c
[kg/s]
mean abs rel error: 1.9% max abs rel error: 7.1%
Figure 5: Comparison of measurements and model for the engine torque M
e
at steady state. Top: Modeled and measured engine torque M
e
. Bottom:
Relative errors for modeled M
e
.
where the parameter
igch
is estimated from measurements, r
c
is the compres-
sion ratio, and
cyl
is the specic heat capacity ratio for the gas in the cylinder.
The friction torque is assumed to follow a polynomial function
M
fric
=
V
d
4
10
5
_
c
fric1
n
2
eratio
+ c
fric2
n
eratio
+ c
fric3
_
(29)
where
n
eratio
=
n
e
1000
(30)
Tuning model parameters

igch
: combustion chamber eciency
c
fric1
, c
fric2
, c
fric3
: coecients in the polynomial function for the friction
torque
Tuning method
The tuning parameters
igch
, c
fric1
, c
fric2
, and c
fric3
are obtained by solving a
linear least-squares problem that minimizes (M
e
+ M
p
M
e,meas
M
p,meas
)
2
with the tuning parameters as the optimization variables. The model of M
e
+M
p
is obtained by solving M
e
+M
p
from Eq. (25) and M
e,meas
+M
p,meas
is estimated
from stationary measurements as M
e,meas
+M
p,meas
= M
e
+V
d
(p
em
p
im
)/(4).
Stationary measurements are used as inputs to the model. The result of the
tuning can be seen in Fig. 5.
15
4 EGR-valve
The mass ow through the EGR-valve is modeled as a simplication of a com-
pressible ow restriction with variable area (Heywood, 1988) and with the as-
sumption that there is no reverse ow when p
em
< p
im
. The motive for this
assumption is to construct a simple model. The model can be extended with
reverse ow, but this increases the complexity of the model since a reverse ow
model requires mixing of dierent temperatures and oxygen fractions in the ex-
haust manifold and a change of the temperature and the gas constant in the
EGR mass ow model. However, p
em
is larger than p
im
in normal operating
points, consequently the assumption above will not eect the model behavior
in these operating points. Furthermore, reverse ow is not measured and can
therefore not be validated.
The mass ow through the restriction is
W
egr
=
A
egr
p
em

egr

T
em
R
e
(31)
where

egr
=
_
2
e

e
1
_

2/e
egr

1+1/e
egr
_
(32)
Measurement data shows that Eq. (32) does not give a suciently accurate
description of the EGR ow. Pressure pulsations in the exhaust manifold or
the inuence of the EGR-cooler could be two dierent explanations for this
phenomenon. In order to maintain the density inuence (p
em
/(

T
em
R
e
)) in
Eq. (31) and the simplicity in the model, the function
egr
is instead modeled
as a parabolic function (see Fig. 6 where
egr
is plotted as function of
egr
).

egr
= 1
_
1
egr
1
egropt
1
_
2
(33)
The pressure quotient
egr
over the valve is limited when the ow is choked,
i.e. when sonic conditions are reached in the throat, and when 1 < p
im
/p
em
,
i.e. no backow can occur.

egr
=
_

egropt
if
pim
pem
<
egropt
pim
pem
if
egropt

pim
pem
1
1 if 1 <
pim
pem
(34)
For a compressible ow restriction, the standard model for
egropt
is

egropt
=
_
2

e
+ 1
_
e
e1
(35)
but the accuracy of the EGR ow model is improved by replacing the physical
value of
egropt
in Eq. (35) with a tuning parameter (Andersson, 2005).
The eective area
A
egr
= A
egrmax
f
egr
( u
egr
) (36)
16
0.7 0.75 0.8 0.85 0.9 0.95 1
0
0.2
0.4
0.6
0.8
1

e
g
r

egr
[]
0 10 20 30 40 50 60 70 80 90 100
0
0.2
0.4
0.6
0.8
1
f
e
g
r

[

]
u
egr
[%]
Figure 6: Comparison of estimated points from measurements and two sub-
models for the EGR ow W
egr
at steady state showing how dierent variables
in the sub-models depend on each other. Note that this is not a validation of
the sub-models since the estimated points for the sub-models depend on the
model tuning. Top:
egr
as function of pressure quotient
egr
. The estimated
points are calculated by solving
egr
from Eq. (31). The model is described by
Eq. (33). Bottom: Eective area ratio f
egr
as function of control signal u
egr
.
The estimated points are calculated by solving f
egr
from Eq. (31). The model
is described by Eq. (37).
is modeled as a polynomial function of the EGR valve position u
egr
(see Fig. 6
where f
egr
is plotted as function of u
egr
)
f
egr
( u
egr
) =
_
_
_
c
egr1
u
2
egr
+ c
egr2
u
egr
+ c
egr3
if u
egr

cegr2
2 cegr1
c
egr3

c
2
egr2
4 cegr1
if u
egr
>
cegr2
2 cegr1
(37)
where u
egr
describes the EGR actuator dynamic
d
dt
u
egr
=
1

egr
(u
egr
(t
degr
) u
egr
) (38)
The EGR-valve is open when u
egr
= 100% and closed when u
egr
= 0%. The
values of
egr
and
degr
have been provided by industry.
17
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
W
e
g
r

[
k
g
/
s
]
Modeled W
egr
[kg/s]


Modeled
Calculated from measurements
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14
20
10
0
10
20
30
mean abs rel error: 6.1% max abs rel error: 22.2%
r
e
l

e
r
r
o
r

W
e
g
r

[
%
]
Modeled W
egr
[kg/s]
Figure 7: Top: Comparison between modeled EGR ow W
egr
and estimated
W
egr
from measurements at steady state. Bottom: Relative errors for W
egr
at
steady state.
Tuning parameters

egropt
: optimal value of
egr
for maximum value of the function
egr
in
Eq. (33)
c
egr1
, c
egr2
, c
egr3
: coecients in the polynomial function for the eective
area
Tuning method
The tuning parameter
egropt
is obtained by solving a non-linear least-squares
problem that minimizes (W
egr
W
egr,meas
)
2
with
egropt
as the optimization
variable. In each iteration in the non-linear least-squares solver, the values
for c
egr1
, c
egr2
, and c
egr3
are set to be the solution of a linear least-squares
problem that minimizes (W
egr
W
egr,meas
)
2
for the current value of
egropt
. The
variable W
egr
is described by the model Eq. (31) and W
egr,meas
is estimated from
measurements as W
egr,meas
= W
c
x
egr
/(1x
egr
). Stationary measurements are
used as inputs to the model. The result of the tuning is shown in Fig. 7.
18
5 Turbocharger
The turbocharger consist of a turbo inertia model, a turbine model, and a
compressor model.
5.1 Turbo inertia
For the turbo speed
t
, Newtons second law gives
d
dt

t
=
P
t

m
P
c
J
t

t
(39)
where J
t
is the inertia, P
t
is the power delivered by the turbine, P
c
is the power
required to drive the compressor, and
m
is the mechanical eciency in the
turbocharger.
Tuning parameter
J
t
: turbo inertia
Tuning method
The tuning parameter J
t
is obtained by adjusting this parameter manually until
simulations of the complete model follow the dynamic responses in the dynamic
measurements, see Sec. 8.1.
5.2 Turbine
The turbine models are the total turbine eciency and the turbine mass ow.
5.2.1 Turbine eciency
One way to model the power P
t
is to use the turbine eciency
t
, which is
dened as (Heywood, 1988)

t
=
P
t
P
t,s
=
T
em
T
t
T
em
(1
11/e
t
)
(40)
where T
t
is the temperature after the turbine,
t
is the pressure ratio

t
=
p
amb
p
em
(41)
and P
t,s
is the power from the isentropic process
P
t,s
= W
t
c
pe
T
em
_
1
11/e
t
_
(42)
where W
t
is the turbine mass ow.
However, Eq. (40) is not applicable due to heat losses in the turbine which
cause temperature drops in the temperatures T
t
and T
em
. Consequently, there
will be errors for
t
if Eq. (40) is used to calculate
t
from measurements. One
way to overcome this is to model the temperature drops, but it is dicult to tune
these models since there exists no measurements of these temperature drops.
19
Another way to overcome this, that is frequently used in the literature, is to
use another eciency that are approximatively equal to
t
. This approximation
utilizes that
P
t

m
= P
c
(43)
at steady state according to Eq. (39). Consequently, P
t
P
c
at steady state.
Using this approximation in Eq. (40), another eciency
tm
is obtained

tm
=
P
c
P
t,s
=
W
c
c
pa
(T
c
T
amb
)
W
t
c
pe
T
em
_
1
11/e
t
_ (44)
where T
c
is the temperature after the compressor and W
c
is the compressor mass
ow. The temperature T
em
in Eq. (44) introduces less errors compared to the
temperature dierence T
em
T
t
in Eq. (40) due to that the absolute value of T
em
is larger than the absolute value of T
em
T
t
. Consequently, Eq. (44) introduces
less errors compared to Eq. (40) since Eq. (44) does not consist of T
em
T
t
.
The temperatures T
c
and T
amb
are low and they introduce less errors compared
to T
em
and T
t
since the heat losses in the compressor are comparatively small.
Another advantage of using Eq. (44) is that the individual variables P
t
and
m
in Eq. (39) do not have to be modeled. Instead, the product P
t

m
is modeled
using Eq. (43) and (44)
P
t

m
= P
c
=
tm
P
t,s
=
tm
W
t
c
pe
T
em
_
1
11/e
t
_
(45)
Measurements show that
tm
depends on the blade speed ratio (BSR) as a
parabolic function (Watson and Janota, 1982), see Fig. 8 where
tm
is plotted
as function of BSR.

tm
=
tm,max
c
m
(BSR BSR
opt
)
2
(46)
The blade speed ratio is the quotient of the turbine blade tip speed and the
speed which a gas reaches when expanded isentropically at the given pressure
ratio
t
BSR =
R
t

t
_
2 c
pe
T
em
_
1
11/e
t
_
(47)
where R
t
is the turbine blade radius. The parameter c
m
in the parabolic function
varies due to mechanical losses and c
m
is therefore modeled as a function of the
turbo speed
c
m
= c
m1
(
t
c
m2
)
cm3
(48)
see Fig. 8 where c
m
is plotted as function of
t
.
Tuning parameters

tm,max
: maximum turbine eciency
BSR
opt
: optimum BSR value for maximum turbine eciency
c
m1
, c
m2
, c
m3
: parameters in the model for c
m
20
0.52 0.54 0.56 0.58 0.6 0.62 0.64
0.4
0.5
0.6
0.7
0.8
0.9
2680
10640

t
m

[

]
BSR []
2000 3000 4000 5000 6000 7000 8000 9000 10000 11000
0.5
0
0.5
1
1.5
2
2.5
c
m

[

t
[rad/s]
Figure 8: Comparison of estimated points from measurements and the model
for the turbine eciency
tm
at steady state. Top:
tm
as function of blade
speed ratio BSR. The estimated points are calculated by using Eq. (44) and
(47). The model Eq. (46) is plotted at two dierent turbo speeds
t
. Bottom:
Parameter c
m
as function of turbo speed
t
. The estimated points are calculated
by solving c
m
from Eq. (46). The model is described by Eq. (48). Note that
this plot is not a validation of c
m
since the estimated points for c
m
depend on
the model tuning.
Tuning method
The tuning parameters BSR
opt
, c
m2
, and c
m3
are obtained by solving a non-
linear least-squares problem that minimizes (
tm

tm,meas
)
2
with BSR
opt
,
c
m2
, and c
m3
as the optimization variables. In each iteration in the non-linear
least-squares solver, the values for
tm,max
and c
m1
are set to be the solution of
a linear least-squares problem that minimizes (
tm

tm,meas
)
2
for the current
values of BSR
opt
, c
m2
, and c
m3
. The eciency
tm
is described by the model
Eq. (46) and
tm,meas
is estimated from measurements using Eq. (44). Station-
ary measurements are used as inputs to the model. The result of the tuning is
shown in Fig. 8 and 9.
5.2.2 Turbine mass ow
The turbine mass ow W
t
is modeled using the corrected mass ow (Heywood,
1988; Watson and Janota, 1982)
W
t

T
em
p
em
= A
vgtmax
f
t
(
t
) f
vgt
( u
vgt
) (49)
21
1 1.5 2 2.5 3 3.5 4
x 10
5
15
10
5
0
5
10
15
r
e
l

e
r
r
o
r

t
m

[
%
]
p
em
[Pa]
mean abs rel error: 4.2% max abs rel error: 13.2%
Figure 9: Relative errors for the total turbine eciency
tm
as function of
exhaust manifold pressure p
em
at steady state.
where A
vgtmax
is the maximum area in the turbine that the gas ows through.
Measurements show that the corrected mass ow depends on the pressure ratio

t
and the VGT actuator signal u
vgt
. As the pressure ratio decreases, the
corrected mass ow increases until the gas reaches the sonic condition and the
ow is choked. This behavior can be described by a choking function
f
t
(
t
) =
_
1
Kt
t
(50)
which is not based on the physics of the turbine, but it gives good agreement
with measurements using few parameters (Eriksson et al., 2002), see Fig. 10
where f
t
is plotted as function of
t
.
When the VGT control signal u
vgt
increases, the eective area increases and
hence also the ow increases. Due to the geometry in the turbine, the change in
eective area is large when the VGT control signal is large. This behavior can
be described by a part of an ellipse (see Fig. 10 where f
vgt
is plotted as function
of u
vgt
)
_
f
vgt
( u
vgt
) c
f2
c
f1
_
2
+
_
u
vgt
c
vgt2
c
vgt1
_
2
= 1 (51)
where f
vgt
is the eective area ratio function and u
vgt
describes the VGT actu-
ator dynamic
d
dt
u
vgt
=
1

vgt
(u
vgt
u
vgt
) (52)
The value of
vgt
has been provided by industry. The ow can now be modeled
by solving W
t
from Eq. (49)
W
t
=
A
vgtmax
p
em
f
t
(
t
) f
vgt
( u
vgt
)

T
em
(53)
and solving f
vgt
from Eq. (51)
f
vgt
( u
vgt
) = c
f2
+ c
f1

1
_
u
vgt
c
vgt2
c
vgt1
_
2
(54)
22
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
0.6
0.8
1
1.2
f

t

[

t
[]
20 30 40 50 60 70 80 90 100 110
0.4
0.6
0.8
1
1.2
1.4
f
v
g
t

[

]
u
vgt
[%]
Figure 10: Comparison of estimated points from measurements and two sub-
models for the turbine mass ow at steady state showing how dierent variables
in the sub-models depend on each other. Note that this is not a validation of
the sub-models since the estimated points for the sub-models depend on the
model tuning. Top: The choking function f
t
as function of the pressure ratio

t
. The estimated points are calculated by solving f
t
from Eq. (49). The
model is described by Eq. (50). Bottom: The eective area ratio function f
vgt
as function of the control signal u
vgt
. The estimated points are calculated by
solving f
vgt
from Eq. (49). The model is described by Eq. (54).
Tuning parameters
K
t
: exponent in the choking function for the turbine ow
c
f1
, c
f2
, c
vgt1
, c
vgt2
: parameters in the ellipse for the eective area ratio
function
Tuning method
The tuning parameters above are obtained by solving a non-linear least-squares
problem that minimizes (W
t
W
t,meas
)
2
with the tuning parameters as the
optimization variables. The ow W
t
is described by the model Eq. (53), (54),
and (50), and W
t,meas
is estimated from measurements as W
t,meas
= W
c
+W
f
,
where W
f
is estimated using Eq. (12). Stationary measurements are used as
inputs to the model. The result of the tuning is shown in Fig. 11.
5.3 Compressor
The compressor models the compressor eciency and the compressor mass ow.
23
20 30 40 50 60 70 80 90 100 110
6
4
2
0
2
4
6
8
r
e
l

e
r
r
o
r

W
t

[
%
]
u
vgt
[%]
mean abs rel error: 2.8% max abs rel error: 7.6%
Figure 11: Relative errors for turbine ow W
t
as function of control signal u
vgt
at steady state.
5.3.1 Compressor eciency
The compressor power P
c
is modeled using the compressor eciency
c
, which
is dened as (Heywood, 1988)

c
=
P
c,s
P
c
=
T
amb
_

11/a
c
1
_
T
c
T
amb
(55)
where T
c
is the temperature after the compressor,
c
is the pressure ratio

c
=
p
im
p
amb
(56)
and P
c,s
is the power from the isentropic process
P
c,s
= W
c
c
pa
T
amb
_

11/a
c
1
_
(57)
where W
c
is the compressor mass ow. The power P
c
is modeled by solving P
c
from Eq. (55) and using Eq. (57)
P
c
=
P
c,s

c
=
W
c
c
pa
T
amb

c
_

11/a
c
1
_
(58)
The eciency is modeled using ellipses similar to Guzzella and Amstutz
(1998), but with a non-linear transformation on the axis for the pressure ratio.
The inputs to the eciency model are
c
and W
c
(see Fig. 16). The ow W
c
is not scaled by the inlet temperature and the inlet pressure since these two
variables are constant. The ellipses can be described as

c
=
cmax

T
Q
c
(59)
is a vector which contains the inputs
=
_
W
c
W
copt

c

copt
_
(60)
where the non-linear transformation for
c
is

c
= (
c
1)
pow
(61)
24
0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 0.55 0.6
15
10
5
0
5
10
15
mean abs rel error: 3.3% max abs rel error: 14.1%
r
e
l

e
r
r
o
r

c

[
%
]
W
c
[kg/s]
Figure 12: Relative errors for
c
as function of W
c
at steady state.
and the symmetric matrix Q
c
consists of three parameters
Q
c
=
_
a
1
a
3
a
3
a
2
_
(62)
Tuning model parameters

cmax
: maximum compressor eciency
W
copt
and
copt
: optimum values of W
c
and
c
for maximum compressor
eciency
pow

: exponent in the scale function, Eq. (61)

a
1
, a
2
and a
3
: parameters in the matrix Q
c
Tuning method
The tuning parameters W
copt
,
copt
, and pow

are obtained by solving a non-

linear least-squares problem that minimizes (
c

c,meas
)
2
with W
copt
,
copt
,
and pow

as the optimization variables. In each iteration in the non-linear least-

squares solver, the values for
cmax
, a
1
, a
2
and a
3
are set to be the solution of
a linear least-squares problem that minimizes (
c

c,meas
)
2
for the current
values of W
copt
,
copt
, and pow

. The eciency
c
is described by the model
Eq. (59) to (62) and
c,meas
is estimated from measurements using Eq. (55).
Stationary measurements are used as inputs to the model. The result of the
tuning is shown in Fig. 12.
5.3.2 Compressor mass ow
The mass ow W
c
through the compressor is modeled using two dimensionless
variables. The rst variable is the energy transfer coecient (Dixon, 1998)

c
=
2 c
pa
T
amb
_

11/a
c
1
_
R
2
c

2
t
(63)
which is the quotient of the isentropic kinetic energy of the gas at the given
pressure ratio
c
and the kinetic energy of the compressor blade tip where R
c
25
0.2 0.4 0.6 0.8 1 1.2 1.4 1.6
0
0.1
0.2
0.3
0.4
2700
5300
8000
10600

c
[]

c

[

]
Figure 13: Comparison of estimated points from measurements and model for
the compressor mass ow W
c
at steady state. Volumetric ow coecient
c
as
function of energy transfer coecient
c
. The estimated points are calculated
using Eq. (63) and (64). The model (Eq. 68) is plotted at four dierent turbo
speeds
t
.
is compressor blade radius. The second variable is the volumetric ow coe-
cient (Dixon, 1998)

c
=
W
c
/
amb
R
3
c

t
=
R
a
T
amb
p
amb
R
3
c

t
W
c
(64)
which is the quotient of volume ow rate of air into the compressor and the rate
at which volume is displaced by the compressor blade where
amb
is the density
of the ambient air. The relation between
c
and
c
can be described by a part
of an ellipse (Andersson, 2005), see Fig. 13 where
c
is plotted as function of

c
.
c
1
(
t
) (
c
c
2
)
2
+ c
1
(
t
) (
c
c
2
)
2
= 1 (65)
where c
1
and c
1
varies with turbo speed
t
and are modeled as polynomial
functions.
c
1
(
t
) = c
1

2
t
+ c
2

t
+ c
3
(66)
c
1
(
t
) = c
1

2
t
+ c
2

t
+ c
3
(67)
In Fig. 14 the variables c
1
and c
1
are plotted as function of the turbo speed

t
.
The mass ow is modeled by solving
c
from Eq. (65) and solving W
c
from
Eq. (64).

c
=

1 c
1
(
c
c
2
)
2
c
1
+ c
2
(68)
W
c
=
p
amb
R
3
c

t
R
a
T
amb

c
(69)
Tuning model parameters
c
2
, c
2
: parameters in the ellipse model for the compressor mass ow
c
1
, c
2
, c
3
: coecients in the polynomial function Eq. (66)
c
1
, c
2
, c
3
: coecients in the polynomial function Eq. (67)
26
2000 3000 4000 5000 6000 7000 8000 9000 10000 11000
0.3
0.4
0.5
0.6
0.7
c

1

[

t
[rad/s]
2000 3000 4000 5000 6000 7000 8000 9000 10000 11000
10
15
20
25
30
c

1

[

t
[rad/s]
Figure 14: Comparison of estimated points from measurements and two sub-
models for the compressor mass ow at steady state showing how dierent vari-
ables in the sub-models depend on each other. Note that this is not a validation
of the sub-models since the estimated points for the sub-models depend on the
model tuning. The sub-models are the ellipse variables c
1
and c
1
as function
of turbo speed
t
. The estimated points are calculated by solving c
1
and c
1
from Eq. (65). The models are described by Eq. (66) and (67).
Tuning method
The tuning parameters c
2
and c
2
are obtained by solving a non-linear least-
squares problem that minimizes (c
1
(
t
) (
c
c
2
)
2
+c
1
(
t
) (
c
c
2
)
2
1)
2
with c
2
and c
2
as the optimization variables. In each iteration in the non-
linear least-squares solver, the values for c
1
, c
2
, c
3
, c
1
, c
2
, and
c
3
are set to be the solution of a linear least-squares problem that minimizes
(c
1
(
t
) (
c
c
2
)
2
+ c
1
(
t
) (
c
c
2
)
2
1)
2
for the current values of c
2
and c
2
. Stationary measurements are used as inputs to the model. The result
of the tuning is shown in Fig. 15.
5.3.3 Compressor map
Compressor performance is usually presented by a map with constant eciency
lines and constant turbo speed lines and with
c
and W
c
on the axes. This is
shown in Fig. 16 which has approximatively the same characteristics as Fig. 2.10
in Watson and Janota (1982). Consequently, the proposed compressor model
has the expected behavior.
27
2000 3000 4000 5000 6000 7000 8000 9000 10000 11000
15
10
5
0
5
10
r
e
l

e
r
r
o
r

W
c

[
%
]

t
[rad/s]
mean abs rel error: 3.4% max abs rel error: 13.7%
Figure 15: Relative errors for compressor ow W
c
as function of turbocharger
speed
t
at steady state.
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
1
1.5
2
2.5
3
3.5
4
W
c
[kg/s]

c

[

]
0
.
5
0
.
5
0.5
0
.5
0
.
5
0
.
5
0
.
5
5
0
.
5
5
0.55
0
.5
5
0
.
5
5
0
.
5
5
0
.
6
0
.
6
0
.
6
0.6
0
.6
0
.
6
0
.
6
0
.
6
5
0
.
6
5
0
.
6
5
0
.
6
5
0
.
6
5
0
.6
5
0.65
0
.
7
0
.
7
0
.
7
0
.
7
0
.
7
0
.7
0.7
0
.
7
3
0
.7
3
0
.
7
3
0
.7
3
4000
5000
6000
7000
8000
9000
10000
11000
12000

c
< 0.5
0.5 <
c
< 0.55
0.55 <
c
< 0.6
0.6 <
c
< 0.65
0.65 <
c
< 0.7
0.7 <
c
< 0.73
0.73 <
c
Figure 16: Compressor map with modeled eciency lines (solid line), modeled
turbo speed lines (dashed line with turbo speed in rad/s), and estimated ef-
ciency from measurements using Eq. (55). The estimated points are divided
into dierent groups. The turbo speed lines are described by the compressor
ow model.
28
6 Intercooler and EGR-cooler
To construct a simple model, that captures the important system properties,
the intercooler and the EGR-cooler are assumed to be ideal, i.e. the equations
for the coolers are
p
out
= p
in
W
out
= W
in
T
out
= T
cool
(70)
where T
cool
is the cooling temperature. The model can be extended with non-
ideal coolers, but these increase the complexity of the model since non-ideal
coolers require that there are states for the pressures both before and after the
coolers.
29
7 Summary of assumptions and model equations
A summary of the model assumptions is given in Sec. 7.1 and the proposed
model equations are given in Sec. 7.2 to 7.5.
7.1 Assumptions
To develop a simple model, that captures the dominating eects in the mass
ows, the following assumptions are made:
The intercooler and the EGR-cooler are ideal, i.e. the equations for the
coolers are
p
out
= p
in
W
out
= W
in
T
out
= T
cool
(71)
where T
cool
is the cooling temperature.
The manifolds are modeled as standard isothermal models.
All gases are considered to be ideal and there are two sets of thermody-
namic properties:
1. Air has the gas constant R
a
and the specic heat capacity ratio
a
.
2. Exhaust gas has the gas constant R
e
and the specic heat capacity
ratio
e
.
No heat transfer to or from the gas inside of the intake manifold.
No backow can occur.
The intake manifold temperature is constant.
The oxygen fuel ratio
O
is always larger than one.
7.2 Manifolds
d
dt
p
im
=
R
a
T
im
V
im
(W
c
+ W
egr
W
ei
)
d
dt
p
em
=
R
e
T
em
V
em
(W
eo
W
t
W
egr
)
(72)
x
egr
=
W
egr
W
c
+ W
egr
(73)
d
dt
X
Oim
=
R
a
T
im
p
im
V
im
((X
Oem
X
Oim
) W
egr
+ (X
Oc
X
Oim
) W
c
)
d
dt
X
Oem
=
R
e
T
em
p
em
V
em
(X
Oe
X
Oem
) W
eo
(74)
30
7.3 Cylinder
7.3.1 Cylinder ow
W
ei
=

vol
p
im
n
e
V
d
120 R
a
T
im
(75)

vol
= c
vol1

p
im
+ c
vol2

n
e
+ c
vol3
(76)
W
f
=
10
6
120
u

n
e
n
cyl
(77)
W
eo
= W
f
+ W
ei
(78)

O
=
W
ei
X
Oim
W
f
(O/F)
s
(79)
X
Oe
=
W
ei
X
Oim
W
f
(O/F)
s
W
eo
(80)
7.3.2 Cylinder out temperature
q
in,k+1
=
W
f
q
HV
W
ei
+ W
f
(1 x
r,k
)
x
p,k+1
= 1 +
q
in,k+1
x
cv
c
va
T
1,k
r
a1
c
x
v,k+1
= 1 +
q
in,k+1
(1 x
cv
)
c
pa
_
q
in,k+1
xcv
cva
+ T
1,k
r
a1
c
_
x
r,k+1
=

1/a
e
x
1/a
p,k+1
r
c
x
v,k+1
T
e,k+1
=
sc

11/a
e
r
1a
c
x
1/a1
p,k+1
_
q
in,k+1
_
1 x
cv
c
pa
+
x
cv
c
va
_
+ T
1,k
r
a1
c
_
T
1,k+1
= x
r,k+1
T
e,k+1
+ (1 x
r,k+1
) T
im
(81)
T
em
= T
amb
+ (T
e
T
amb
) e

h
tot
d
pipe
l
pipe
n
pipe
Weo cpe
(82)
7.3.3 Cylinder torque
M
e
= M
ig
M
p
M
fric
(83)
M
p
=
V
d
4
(p
em
p
im
) (84)
M
ig
=
u

10
6
n
cyl
q
HV

ig
4
(85)

ig
=
igch
_
1
1
r

cyl
1
c
_
(86)
M
fric
=
V
d
4
10
5
_
c
fric1
n
2
eratio
+ c
fric2
n
eratio
+ c
fric3
_
(87)
n
eratio
=
n
e
1000
(88)
31
7.4 EGR-valve
W
egr
=
A
egr
p
em

egr

T
em
R
e
(89)

egr
= 1
_
1
egr
1
egropt
1
_
2
(90)

egr
=
_

egropt
if
pim
pem
<
egropt
pim
pem
if
egropt

pim
pem
1
1 if 1 <
pim
pem
(91)
A
egr
= A
egrmax
f
egr
( u
egr
) (92)
f
egr
( u
egr
) =
_
_
_
c
egr1
u
2
egr
+ c
egr2
u
egr
+ c
egr3
if u
egr

cegr2
2 cegr1
c
egr3

c
2
egr2
4 cegr1
if u
egr
>
cegr2
2 cegr1
(93)
d
dt
u
egr
=
1

egr
(u
egr
(t
degr
) u
egr
) (94)
7.5 Turbo
7.5.1 Turbo inertia
d
dt

t
=
P
t

m
P
c
J
t

t
(95)
7.5.2 Turbine eciency
P
t

m
=
tm
W
t
c
pe
T
em
_
1
11/e
t
_
(96)

t
=
p
amb
p
em
(97)

tm
=
tm,max
c
m
(BSR BSR
opt
)
2
(98)
BSR =
R
t

t
_
2 c
pe
T
em
_
1
11/e
t
_
(99)
c
m
= c
m1
(
t
c
m2
)
cm3
(100)
32
7.5.3 Turbine mass ow
W
t
=
A
vgtmax
p
em
f
t
(
t
) f
vgt
( u
vgt
)

T
em
(101)
f
t
(
t
) =
_
1
Kt
t
(102)
f
vgt
( u
vgt
) = c
f2
+ c
f1

1
_
u
vgt
c
vgt2
c
vgt1
_
2
(103)
d
dt
u
vgt
=
1

vgt
(u
vgt
u
vgt
) (104)
7.5.4 Compressor eciency
P
c
=
W
c
c
pa
T
amb

c
_

11/a
c
1
_
(105)

c
=
p
im
p
amb
(106)

c
=
cmax

T
Q
c
(107)
=
_
W
c
W
copt

c

copt
_
(108)

c
= (
c
1)
pow
(109)
Q
c
=
_
a
1
a
3
a
3
a
2
_
(110)
7.5.5 Compressor mass ow
W
c
=
p
amb
R
3
c

t
R
a
T
amb

c
(111)

c
=

1 c
1
(
c
c
2
)
2
c
1
+ c
2
(112)

c
=
2 c
pa
T
amb
_

11/a
c
1
_
R
2
c

2
t
(113)
c
1
= c
1

2
t
+ c
2

t
+ c
3
(114)
c
1
= c
1

2
t
+ c
2

t
+ c
3
(115)
33
8 Model tuning and validation
To develop a model that describes the system dynamics and the nonlinear eects,
the model have to be tuned and validated. In Sec. 8.1 static and dynamic models
are tuned and in Sec. 8.2 a validation of the complete model is performed using
dynamic data. In the validation, it is important to investigate if the model
captures the essential dynamic behaviors and nonlinear eects.
8.1 Tuning
The tuning of static and dynamic models are described in the following sections.
Static models
In Tab. 3 there is a summary of the absolute relative model errors from Sec. 3 to 5
between static models and stationary measurements for each subsystem. The
stationary measurements consist of 82 operating points, that are scattered over
a large operating region with dierent loads, speeds, VGT- and EGR-positions.
These 82 operating points also include the European Stationary Cycle (ESC).
The mean absolute relative errors are equal to or lower than 6.1 %. The EGR
mass ow model has the largest mean relative error and the cylinder mass ow
model has the smallest mean relative error.
Table 3: The mean and maximum absolute relative errors between static models
and steady state measurements for each subsystem in the diesel engine model,
i.e. a summary of the mean and maximum absolute relative errors in Sec. 3
to 5.
Subsystem Mean absolute rela-
tive error [%]
Maximum absolute
relative error [%]
Cylinder mass ow 0.9 2.5
Exhaust gas temperature 1.7 5.4
Engine torque 1.9 7.1
EGR mass ow 6.1 22.2
Turbine eciency 4.2 13.2
Turbine mass ow 2.8 7.6
Compressor eciency 3.3 14.1
Compressor mass ow 3.4 13.7
Dynamic models
The tuning parameters for the dynamic models are the manifold volumes V
im
and V
em
in Sec. 2 and the turbo inertia J
t
in Sec. 5.1. These parameters
are adjusted manually until simulations of the complete model follow dynamic
responses in dynamic measurements by considering time constants. The tuning
is performed using a dynamic tuning data, the data C in Tab. 4, that consists of
77 dierent steps in VGT control signal and EGR control signal in an operating
point with 50 % load and n
e
=1500 rpm. All the data in Tab. 4 are used for
validation in Sec. 8.2. Note that the dynamic measurements are limited in
sample rate with a sample frequency of 1 Hz for data A-E and with a sample
34
Table 4: The mean absolute relative errors between diesel engine model sim-
ulation and dynamic tuning or validation data that consist of steps in VGT-
position, EGR-valve, and fuel injection. The data C and F are used for tuning
of dynamic models, the data A, B, D, E, and F are used for validation of time
constants, and all the data are used for validation of static models and essential
system properties.
VGT-EGR steps u

steps
Data name A B C D E F
Speed [rpm] 1200 1500 1900 1500
Load [%] 25 75 50 25 75 -
Number of steps 77 77 77 77 55 7
p
im
2.0 10.6 6.3 5.0 4.5 2.9
p
em
2.4 6.8 5.5 4.5 4.6 4.7
W
c
3.2 10.6 8.0 6.7 6.7 3.8
n
t
4.4 11.9 7.0 6.0 4.1 3.0
M
e
- - - - - 7.3
frequency of 10 Hz for data F. This leads to that the data does not captures the
fastest dynamics in the system.
A dynamometer is tted to the engine via an axle in order to brake or supply
torque to the engine. This dynamometer and axle lead to that the measured
engine torque has a time constant that is not modeled due to that the torque
will not be used as a feedback in the controller. However, in order to validate
the engine torque model during dynamic responses, this dynamic is modeled in
the validation as a rst order system
d
dt
M
e,meas
=
1

Me
(M
e
M
e,meas
) (116)
where M
e,meas
is the measured torque and M
e
is the output torque from the
engine. The time constant
Me
is tuned by adjusting it manually until simu-
lations of the complete model follow the measured torque during steps in fuel
injection at 1500 rpm, i.e. the data F in Tab. 4.
8.2 Validation
Due to that the stationary measurements are few, both the static and the dy-
namic models are validated by simulating the total model and comparing it with
dynamic validation data that consists of several dierent steps in VGT-position,
EGR-valve, and fuel injection. The steps in VGT-position and EGR-valve are
performed in 5 dierent operating points and the steps in fuel injection are
performed in one operating point. The result of this validation can be seen in
Tab. 4 that shows that the mean absolute relative errors are less than 12 %.
Note that the engine torque is not measured during VGT and EGR steps. The
relative errors are due to mostly steady state errors, but since the engine model
will be used in a controller the steady state accuracy is less important since a
controller will take care of steady state errors. However, in order to design a
successful controller, it is important that the model captures the essential dy-
namic behaviors and nonlinear eects. Therefore, time constants and essential
35
system properties are validated in the following sections.
Validation of time constants
In Sec. 8.1, the dynamic models are tuned by considering the time constants in
the data C in Tab. 4. These time constants are validated using the dynamic
validation data A, B, D, and E in Tab. 4. Some parts of this validation are
plotted in Fig. 17 and 18. The non-minimum phase behavior in p
im
in Fig. 17
shows that the model captures the fast dynamic in the beginning of the response
and that the model captures the slow dynamic in the end of the response. The
overshoot in the third response in Fig. 18 also shows that model captures both
fast and slow dynamics.
Validation of essential system properties
Kolmanovsky et al. (1997) and Jung (2003) show the essential system properties
for the pressures and the ows in a diesel engine with VGT and EGR. Some
of these properties are a non-minimum phase behavior in the intake manifold
pressure and a non-minimum phase behavior, an overshoot, and a sign reversal
in the compressor mass ow. These system properties are validated using the
dynamic data A-E in Tab. 4. Some parts of this validation are shown in Fig. 17
to 19. Fig. 17 shows that the model captures the non-minimum phase behavior
in the transfer function u
egr
to p
im
and the second step in Fig. 19 shows that
the model captures the non-minimum phase behavior in the transfer function
u
vgt
to W
c
. Note that the non-minimum phase behaviors in the measurements
are not obvious due to a low sample frequency. Further, the third step in Fig. 18
and the third step in Fig. 19 show that the model captures the overshoot and
the sign reversal in the transfer function u
vgt
to W
c
.
36
0 5 10 15 20
0
20
40
60
u
e
g
r

[
%
]
0 5 10 15 20
1.5
1.6
1.7
1.8
1.9
x 10
5
Time [s]
p
i
m

[
P
a
]
sim
meas
Figure 17: Comparison between diesel engine model simulation and dynamic
validation data during steps in EGR-valve position showing that the model
captures the non-minimum phase behavior in p
im
. Operating point: 25 % load,
n
e
=1900 rpm and u
vgt
=50 %.
0 5 10 15 20 25 30 35 40
20
40
60
80
100
u
v
g
t

[
%
]
0 5 10 15 20 25 30 35 40
0.15
0.2
0.25
0.3
0.35
Time [s]
W
c

[
k
g
/
s
]
sim
meas
Figure 18: Comparison between diesel engine model simulation and dynamic val-
idation data during steps in VGT position showing that the model captures the
overshoot and the sign reversal in W
c
. Operating point: 75 % load, n
e
=1200 rpm
and u
egr
=40 %.
37
0 5 10 15 20 25 30
20
40
60
80
100
u
v
g
t

[
%
]
0 5 10 15 20 25 30
0.2
0.25
0.3
0.35
Time [s]
W
c

[
k
g
/
s
]
sim
meas
Figure 19: Comparison between diesel engine model simulation and dynamic
tuning data during steps in VGT position showing that the model captures
the non-minimum phase behavior, the overshoot, and the sign reversal in W
c
.
Operating point: 50 % load, n
e
=1500 rpm and u
egr
=81.5 %.
38
9 Conclusions
A mean value model of a diesel engine with VGT and EGR including oxygen
mass fraction was developed and validated. The intended applications of the
model are system analysis, simulation, and development of model-based con-
trol systems. To be able to implement a model-based controller, the model
must be small. Therefore the model has only seven states: intake and exhaust
manifold pressures, oxygen mass fraction in the intake and exhaust manifold,
turbocharger speed, and two states describing the actuator dynamics for the
EGR-valve and the VGT-position.
Model equations and tuning methods for the model parameters was de-
scribed for each subsystem in the model. Parameters in the static models are
tuned automatically using least square optimization and stationary measure-
ments in 82 dierent operating points. The tuning shows that the mean relative
errors are equal to or lower than 6.1 %. Parameters in dynamic models are
tuned by adjusting these parameters manually until simulations of the complete
model follow the dynamic responses in the dynamic measurements. In order to
decrease the amount of tuning parameters, ows and eciencies are modeled
using physical relationships and parametric models instead of look-up tables.
Static and dynamic validations of the entire model were performed using
dynamic measurements, which consist of steps in fuel injection, EGR control
signal, and VGT control signal. The validations show mean relative errors
which are less than 12 %. The validations also show that the proposed model
captures the essential system properties, i.e. a non-minimum phase behavior
in the transfer function u
egr
to p
im
and a non-minimum phase behavior, an
overshoot, and a sign reversal in the transfer function u
vgt
to W
c
.
39
40
References
Andersson, P. (2005). Air Charge Estimation in Turbocharged Spark Ignition
Engines. PhD thesis, Linkopings Universitet.
Dixon, S. (1998). Fluid Mechanics and Thermodynamics of Turbomachinery.
Butterworth Heinemann, Woburn, 4:th edition.
Eriksson, L. (2002). Mean value models for exhaust system temperatures. SAE
2002 Transactions, Journal of Engines, 2002-01-0374, 111(3).
Eriksson, L., Nielsen, L., Brugard, J., Bergstrom, J., Pettersson, F., and An-
dersson, P. (2002). Modeling and simulation of a turbo charged SI engine.
Annual Reviews in Control, 26(1):129137.
Guzzella, L. and Amstutz, A. (1998). Control of diesel engines. IEEE Control
Systems Magazine, 18:5371.
Heywood, J. (1988). Internal Combustion Engine Fundamentals. McGraw-Hill
Book Co.
Jung, M. (2003). Mean-Value Modelling and Robust Control of the Airpath of
a Turbocharged Diesel Engine. PhD thesis, University of Cambridge.
Kolmanovsky, I., Stefanopoulou, A., Moraal, P., and van Nieuwstadt, M.
(1997). Issues in modeling and control of intake ow in variable geometry
turbocharged engines. In Proceedings of 18
th
IFIP Conference on System
Modeling and Optimization, Detroit.
Skogtj arn, P. (2002). Modelling of the exhaust gas temperature for diesel en-
gines. Masters thesis LiTH-ISY-EX-3379, Department of Electrical Engi-
neering, Linkoping University, Linkoping, Sweden.
Vigild, C. (2001). The Internal Combustion Engine Modelling, Estimation and
Control Issues. PhD thesis, Technical University of Denmark, Lyngby.
Watson, N. and Janota, M. (1982). Turbocharging the Internal Combustion
Engine. The Mechanical Press Ltd, Hong Kong.
41
42
A Notation
Table 5: Symbols used in the report
Symbol Description Unit
A Area m
2
BSR Blade speed ratio
c
p
Spec. heat capacity, constant pressure J/(kg K)
c
v
Spec. heat capacity, constant volume J/(kg K)
J Inertia kg m
2
M Torque Nm
M
e
Engine torque Nm
M
p
Pumping torque Nm
n
cyl
Number of cylinders
n
e
Rotational engine speed rpm
n
t
Rotational turbine speed rpm
(O/F)
s
Stoichiometric oxygen-fuel ratio
p Pressure Pa
P Power W
q
HV
Heating value of fuel J/kg
r
c
Compression ratio
R Gas constant J/(kg K)
R Radius m
T Temperature K
u
egr
EGR control signal. 100 - open, 0 - closed %
u
vgt
VGT control signal. 100 - open, 0 - closed %
u

Injected amount of fuel mg/cycle

V Volume m
3
W Mass ow kg/s
x
egr
EGR fraction
X
O
Oxygen mass fraction
Specic heat capacity ratio
Eciency

O
Oxygen-fuel ratio
Pressure quotient
Density kg/m
3
Time constant s

c
Volumetric ow coecient

c
Energy transfer coecient
Rotational speed rad/s
43
Table 6: Indices used in the report
Index Description
a air
amb ambient
c compressor
d displaced
e exhaust
egr EGR
ei engine cylinder in
em exhaust manifold
eo engine cylinder out
f fuel
fric friction
ig indicated gross
im intake manifold
m mechanical
t turbine
vgt VGT
vol volumetric
fuel injection
Table 7: Abbreviations used in the report
Abbreviation Description
EGR Exhaust gas recirculation
VGT Variable geometry turbocharger
44