This action might not be possible to undo. Are you sure you want to continue?
Automatic Generation of
Combustion Engine Models using MatLab
&
Idle Drive Train Model in MatLab / Simulink
L.H.J. Römers
Tutors:
Dipl. Ing. B. Pennec LuK GmbH & Co.
Dr. S. Jones LuK GmbH & Co.
Dr. P.A. Veenhuizen Technical University of Eindhoven
LuK GmbH & Co. Technical University of Eindhoven
Luc Römers / XB department Traineeship report
 1 
Preface
I would like to thank all employees of the dynamic simulation department at LuK Bühl, for
giving me the chance to do an interesting and varied traineeship. Besides gaining a lot of
technical knowledge, I have experienced what it is like to be part of a professional
development team.
Special thanks go out to Bertrand Pennec and Dr. Stephen Jones for their great support during
the last months. I would also like to thank Dr. Ad Kooy, as my contact at LuK, for giving me
the chance to gain these experiences.
Luc Römers
LuK GmbH & Co. Technical University of Eindhoven
Luc Römers / XB department Traineeship report
 2 
Table of contents
1. Introduction 4
1.1 History of the company LuK GmbH & Co. 4
1.2 Goal of the traineeship 6
Automatic generation of combustion engine models using MatLab 7
2. Generation of pressurebased engine models 8
2.1 Geometry and kinematics of the crank mechanism 10
2.2 Dynamics of the crank mechanism 11
2.3 Mass torque 12
2.4 Compression / expansion torque 13
2.5 Combustion torque 17
2.5.1 Isermann method 18
2.5.2 Isermann/Pennec method 20
2.6 Total torque resulting from gas forces 22
2.7 Parameter maps 23
2.8 Reducing the parameter maps 31
2.9 Peaktopeak comparison 33
2.10 Output for DyFaSim software 40
3. Generation of standard engine models 42
3.1 Combustion engine data sheet 42
3.2 Replacing missing data by estimates 43
3.2.1 Geometry data 43
3.2.2 Net torque curve at full load 43
3.2.3 Drag torque curve 44
3.2.4 Intake pressure curve at full load 45
4. Graphical User Interface in MatLab 46
4.1 Graphical User Interface, pressurebased engine model 46
4.2 Graphical User Interface, standard engine model 54
5. Comparison between pressurebased and standard models 59
Idle drive train model in Matlab / Simulink 61
6. Idle drive train model in Matlab / Simulink 61
6.1 Virtual test bench for Simulink representation of LuK arc springs 61
6.2 Drive train model for idle simulation 64
6.3 Comparison with DyFaSim counterpart 67
LuK GmbH & Co. Technical University of Eindhoven
Luc Römers / XB department Traineeship report
 3 
7. Conclusions 68
List of symbols 69
References 71
LuK GmbH & Co. Technical University of Eindhoven
Luc Römers / XB department Traineeship report
 4 
1. Introduction
1.1 History of the company LuK GmbH & Co.
In Bühl, at the edge of the Black Forest, the headquarters of the international company LuK is
situated. At the company LuK, 9000 employees working at 20 production locations in
Germany, Brazil, GreatBritain, India, Mexico, SouthAfrica, Hungarian, France, Korea and
the United States yearly produce more than 14 million clutches for passenger cars and tractors,
over 3 million lockup clutches and over 5 million dualmass flywheels.
One in every four cars produced worldwide contains a LuK clutch. The LuK dualmass
flywheel (DMF) increases comfort and decreases fuel consumption in millions of cars
worldwide.
History of LuK GmbH & Co.
1965 Foundation of LuK in Bühl Germany
1974 Foundation of LuK Brazil and LuK Mexico
1976 Foundation of AS (AutoteileService GmbH)
Development of torsion damped clutch plates
1977 Foundation of LuK Inc. USA
1982 New LuK location in Unna
1985 LuK dualmass flywheel (DMF): new standard for noise isolation in the drive train
1986 Takeover of AFT in Werdohl
1987 Takeover of Repco S.A. (today LuK SouthAfrica)
Takeover Laycock in England (today LuK UK)
1989 Opening LuK (UK) Ldt.
1990 Takeover of TCM Ldt. in Hereford/England (today Hereford Ltd.)
1991 Jointventure between LuK and Barmag (future BarLuK)
Takeover of automotivehydraulic division Vickers (today LuK FahrzeugHydraulik)
First electronic clutch management for BMW Alpina
LuK GmbH & Co. Technical University of Eindhoven
Luc Römers / XB department Traineeship report
 5 
1992 Foundation of LuK GetriebeSysteme in Bühl
1993 Complete takeover BarLuK (today LuK Automobiltechnik)
1994 In corporation with a car manufacturer LuK develops parts for a CVT transmission with
a maximum torque of about 300 Nm
1995 Foundation LuK Korea
1996 Foundation Beijing (LuK) Liaison Office in China
Jointventure between Rane und LuK (Rane LuK Clutch Ltd.) in India
Foundation LuK Savaria in Hungarian (start of production, late 1997)
SACclutch (Self Adjusting Clutch)
1997 The LuKEKM is produced in series for the Mercedes Aclass
The production location in Wooster/Ohio starts to produce the first torqueconverter for
light trucks in series
1998 Takeover of Kongsberg TechMatic in England and Norway
Foundation of AS divisions in Spane, UK, USA, Brazil, Mexico, SouthAfrica, Poland,
Russia und Czechia
1999 Takeover of Holzhäuser, Germany
Foundation AS Argentina
Opening LuK Leamington, UK
New foundry LuK Brazil
LuK presents CVT parts for high torques. Audi Multitronic®: the first efficient CVT
transmission for more than 300 Nm is produced in series
Valeo sells LuK shares to INA Holding
2000 Opening of LuK production location in Bußmatten, Germany
LuK takes over the production of clutch release systems form INA
Opening of LuK tool shop in Kappelrodeck, Germany
Automated manual transmission, Easytronic®, is produced in series for the Opel Corsa
LuK GmbH & Co. Technical University of Eindhoven
Luc Römers / XB department Traineeship report
 6 
2001 LuK sells LuK Norway
LuK fully takes over Rane LuK (future LuK India Private Ltd.)
2002 The 20 millionth dualmass flywheel is produced in Bühl
2003 The LuK integrated crankshaft damper is produced in series for the VW Touareg
2004 In Bühl the production of torqueconverters is started
2005 Takover of APTEC (future LuK Friction)
Takeover of AP France (future LuK France)
Opening LuK Korea
Research and development are of great importance at LuK. Therefore about one of every six
employees works in this field. The inventiveness of the product developer creates innovations
for the automotive world of tomorrow.
1.2 Goal of the traineeship
The main assignment is creating an automated and simple way of generating combustion
engine models, for use in the LuKDyFaSim simulation software. For this, the mathematical
tool MatLab will be used.
Because the schedule allowed it, there was the possibility to do a second assignment. This
assignment consists of the buildup of an idle drive train model in MatLab / Simulink. This
model will be used to investigate the unwanted phenomenon of subharmonic vibrations at idle
operation.
LuK GmbH & Co. Technical University of Eindhoven
Luc Römers / XB department Traineeship report
 7 
Automatic generation of combustion engine models using MatLab
The goal of this assignment is to create a method, which allows automatic generation of
combustion engine models for the LuK DyFaSim simulation software. DyFaSim is a software
package developed by LuK and is used to simulate the dynamics of the vehicle drive train. The
program is based on a modular structure, allowing single elements like the combustion engine,
dualmass flywheel, clutch and transmission to be joined to form a complete drive train.
The combustion engine module of DyFaSim uses an empirical method to simulate the dynamic
engine torque in all operating points. To create a DyFaSim engine model, a large amount of
data from the car manufacturer needs to be processed. This has been a semimanual process
until now, carried out by only a few engineers due to its complex nature. Automating the
complete modelling process would reduce the manual effort, thereby increasing the capacity of
the involved engineers. For this cause, the mathematical software package MatLab will be used
to create an automatic modelling process, based on the semimanual process used until now.
The automatic engine model generation will handle two different situations:
 creating an accurate combustion engine model based on cylinder pressure data
 creating a standard combustion engine model, in case very little data is available
These two situations will be referred to as pressurebased model generation and standard model
generation. A full description of the pressurebased model will be given in chapter 2, the
standard model will be treated in chapter 3. In chapter 4 an explanation of the graphical user
interface in MatLab will be presented.
Engine geometry
Cylinder pressure data
Possibly static torque curve
Possibly intake pressure
Complete & optimized
parameter maps
Simplified maps based on
values of experience
Automatic generation
of a script for
DyFaSim engine module
''Motor_pressurebased.dvm'' ''Motor_standard.dvm''
Possibly engine geometry
START
Figure 1.1: overview of the MatLab program
LuK GmbH & Co. Technical University of Eindhoven
Luc Römers / XB department Traineeship report
 8 
2. Generation of pressurebased engine models
The purpose of this model is to simulate the dynamic torque output in all operating points. The
model is based on cylinder geometry data and cylinder pressure data, these are delivered by the
car manufacturer. First, the torque resulting from one single cylinder will be modelled, which
will finally be cascaded into a whole engine. The basic operating method of a fourstroke cycle
is shown in the next figure.
Compression and combustion cause a pressure rise inside the cylinder. An example is given in
the next diagram.
The pressure rise results in a force on the piston, which in turn results in a crankshaft torque.
This torque varies strongly with crank angle, therefore it is called the dynamic cylinder torque.
Exhaust
Combustion &
Expansion
Compression Intake
0°<ϕ<180° 180°<ϕ<360° 360°<ϕ<540° 540°<ϕ<720°
Figure 2.1: fourstroke cycle [ENCA98]
0
15
30
45
60
0 100 200 300 400 500 600
Volume [cm³]
P
r
e
s
s
u
r
e
[
b
a
r
]
Figure 2.2: cylinder pressure over full cycle
LuK GmbH & Co. Technical University of Eindhoven
Luc Römers / XB department Traineeship report
 9 
If we look at the dynamic torque resulting from one cylinder, it is possible to subdivide it into
three sources:
torque due to acceleration forces working on the oscillating piston mass
torque due to pressure buildup resulting from compression/expansion
torque due to pressure buildup resulting from the combustion process
The last two components form the torque due to gas forces. Examples of all three components
are given in the figure below.
These three components will be described using cylinder geometry data, cylinder pressure data
and an empirical part. Once all components are simulated, they can be summed to form the
total dynamic torque resulting from one cylinder. It is important to realise that this is the gross
engine torque. The net torque can be found by subtracting the engine friction. However, in the
LuK DyFaSim software, friction is taken into account outside the engine module.
Before starting the modelling process of these torque components, a description of the
geometry, kinematics and dynamics of the crank mechanism will be given.
Figure 2.3: components of the dynamic torque [PENN02]
LuK GmbH & Co. Technical University of Eindhoven
Luc Römers / XB department Traineeship report
 10 
2.1 Geometry and kinematics of the crank mechanism
The following schematic drawing shows the crank mechanism. It consists of the piston,
connecting rod and the crankshaft.
d = bore diameter
l = connecting rod length
r = crank radius
λ = r/l
TDC = top dead center
BDC = bottom dead center
x = piston displacement, relative to TDC
φ = crank angle, clockwise relative to TDC
4
2
2
d
r V
d
⋅
⋅ ⋅ =
π
displacement volume
c
V
= clearance volume
c
h c
V
V V +
= ε
compression ratio
Using these geometry parameters, the displacement of the piston (relative to TDC) can be
described. From this expression the piston speed and acceleration can be derived [MOOG00]:
(2.1)
(2.2)
(2.3)
Here ω represents the rotational speed of the crankshaft, also known as the engine speed.
Figure 2.4: crank mechanism
( )
(
¸
(
¸
− − + − = ) sin 1 1 ( *
1
cos 1 * ) (
2 2
ϕ λ
λ
ϕ ϕ r x
(
(
¸
(
¸
− ⋅
⋅
+ ⋅ =
•
ϕ λ
ϕ λ
ϕ ω ϕ
2 2
sin 1 2
) 2 sin(
sin * ) ( r x
( )
(
(
¸
(
¸
− ⋅
⋅
+
−
⋅
+ ⋅ =
• •
3
2 2
2 3
2 2
2
sin 1 4
) 2 ( sin
sin 1
) 2 cos(
cos * ) (
ϕ λ
ϕ λ
ϕ λ
ϕ λ
ϕ ω ϕ r x
LuK GmbH & Co. Technical University of Eindhoven
Luc Römers / XB department Traineeship report
 11 
2.2 Dynamics of the crank mechanism
Knowing the kinematics, a dynamical analysis of the piston/crank mechanism can be made.
The next figure shows the forces acting on the various components, as a result of an applied
force on the piston. The resulting force on the crankshaft can be divided into a radial part F
c,r
and a tangential part F
c,t
. Of course only the tangential part will result in a crankshaft torque.
The relation between the applied force at the piston and the tangential force at the crankshaft
can be described as follows [MOOG00]:


¹

\

− ⋅
⋅
+ =
ϕ λ
ϕ λ
ϕ
2 2
,
sin 1 2
) 2 sin(
sin *
applied t c
F F (2.4)
If this expression is multiplied with the crank radius, the crankshaft torque results:


¹

\

− ⋅
⋅
+ =
ϕ λ
ϕ λ
ϕ
2 2
sin 1 2
) 2 sin(
sin * *
applied crank
F r T (2.5)
Figure 2.5: dynamics of the crank mechanism
LuK GmbH & Co. Technical University of Eindhoven
Luc Römers / XB department Traineeship report
 12 
2.3 Mass torque
From the kinematical analysis, the acceleration of the piston as a function of the crank angle is
known. By using Newton’s second law, the forces involved can be determined:
) (ϕ
• •
⋅ − = x m F
osc
(2.6)
Where m
osc
stands for the oscillating mass of one cylinder. This force can be converted to a
torque working on the crankshaft [SEGL00]:


¹

\

− ⋅
⋅
+ =
ϕ λ
ϕ λ
ϕ
2 2
sin 1 2
) 2 sin(
sin * * F r T
mass
(2.7)
( )



¹

\

− ⋅
⋅
+
−
⋅
+


¹

\

− ⋅
⋅
+ ⋅ ⋅ − =
3
2 2
2 3
2 2 2 2
2 2
sin 1 4
) 2 ( sin
sin 1
) 2 cos(
cos *
sin 1 2
) 2 sin(
sin *
ϕ λ
ϕ λ
ϕ λ
ϕ λ
ϕ
ϕ λ
ϕ λ
ϕ ω r m T
osc mass
(2.8)
The mass torque depends quadratically on the engine speed. The next figure displays the mass
torque over one complete cycle. It can be seen that the amplitude increases by a factor of four
when the engine speed is doubled.
The derived formula reveals that the mass torque depends only on cylinder geometry
parameters and engine speed.
0 180 360 540 720
0
Crank angle [degrees]
2000 RPM
4000 RPM
M
a
s
s
t
o
r
q
u
e
[
N
m
]
Figure 2.6: mass torque over one full cycle
LuK GmbH & Co. Technical University of Eindhoven
Luc Römers / XB department Traineeship report
 13 
2.4 Compression / expansion torque
Compression of the gas inside the cylinder results in a pressure rise and consequently in a force
on the piston. This force generates a torque due to compression/expansion at the crankshaft.
Even in case no combustion process is taking place, the compression/expansion torque will still
be present.
To simulate this torque component, an analysis of the cylinder pressure is needed. Cylinder
pressure measurements, in various engine operating points, are delivered by the car
manufacturer. For this pressure analysis, the following assumptions are made:
the gas inside the cylinder is ideal
the compression/expansion is adiabatic reversible, or isentropic
the isentropic coefficient κ is constant, i.e. does not change with temperature
Under these conditions, one can state:
constant V P
κ
= ⋅ (2.9)
( ) ( )
κ
d c intake
κ
comp/exp
V V P ) V( ) ( P + ⋅ = ⋅ ϕ ϕ (2.10)
κ
d c
intake comp/exp
) V(
V V
P ) ( P


¹

\
 +
⋅ =
ϕ
ϕ
(2.11)
Where the angle dependent cylinder volume can be written as [PISC89]:
) ( *
4
*
) (
2
ϕ
π
ϕ x
d
V V
c
+ = (2.12)
( )
(
¸
(
¸
− − + − +
−
⋅ ⋅
= ) sin 1 1 ( *
1
cos 1
1
2
*
4
) (
2 2
2
ϕ λ
λ
ϕ
ε
π
ϕ
d r
V (2.13)
This expression for the volume contains only cylinder geometry parameters, which are known.
The only remaining parameter needed to describe the compression/expansion pressure is the
intake pressure. At this point the pressure measurements are used. The following graph shows
an example of a cylinder pressure measurement. The intake pressure is found by taking an
average between 175° and 185° crank angle, where the piston reaches bottom dead center. The
isentropic equation can now be used to calculate the pressure due to compression/expansion
over the next 360°. The next graph shows the total cylinder pressure and the component
resulting from compression/expansion.
LuK GmbH & Co. Technical University of Eindhoven
Luc Römers / XB department Traineeship report
 14 
0 100 200 300 400 500 600 700
0
10
20
30
40
50
60
70
80
90
100
Crank Angle [rpm]
C
y
l
i
n
d
e
r
p
r
e
s
s
u
r
e
[
b
a
r
]
Intake pressure: average
between 175°& 185°
0 100 200 300 400 500 600 700
0
10
20
30
40
50
60
70
80
90
100
Crank Angle [rpm]
C
y
l
i
n
d
e
r
p
r
e
s
s
u
r
e
[
b
a
r
]
Intake pressure: average
between 175°& 185°
Crank angle [degrees] Crank angle [degrees]
0 100 200 300 400 500 600 700
0
10
20
30
40
50
60
70
80
90
100
Crank Angle [rpm]
C
y
l
i
n
d
e
r
p
r
e
s
s
u
r
e
[
b
a
r
]
measured cylinder pressure
simulated comp./exp. pressure
Crank angle [degrees]
0 100 200 300 400 500 600 700
0
10
20
30
40
50
60
70
80
90
100
Crank Angle [rpm]
C
y
l
i
n
d
e
r
p
r
e
s
s
u
r
e
[
b
a
r
]
measured cylinder pressure
simulated comp./exp. pressure
Crank angle [degrees] Crank angle [degrees]
The compression/expansion pressure can be converted into a crankshaft torque:
( )
g surroundin comp/exp
2
piston
P ) ( P
4
d π
) ( F −
⋅
= ϕ ϕ
(2.14)
( )


¹

\

− ⋅
⋅
+ ⋅ ⋅ −
⋅
=
ϕ
ϕ
ϕ ϕ ϕ
2 2
g surroundin comp/exp
2
comp/exp
sin λ 1 2
) sin(2 λ
sin r P ) ( P
4
d π
) ( T (2.15)
Figure 2.7: cylinder pressure measurement
Figure 2.8: simulation of compression/expansion pressure
LuK GmbH & Co. Technical University of Eindhoven
Luc Römers / XB department Traineeship report
 15 
The same conversion is made for the measured total cylinder pressure. The resulting torque
curves are displayed in the next graph.
0 100 200 300 400 500 600 700
400
200
0
200
400
600
800
1000
Crank Angle [rpm]
T
o
r
q
u
e
[
N
m
]
torque from measurements
simulated comp./exp. torque
Crank angle [degrees]
0 100 200 300 400 500 600 700
400
200
0
200
400
600
800
1000
Crank Angle [rpm]
T
o
r
q
u
e
[
N
m
]
torque from measurements
simulated comp./exp. torque
0 100 200 300 400 500 600 700
400
200
0
200
400
600
800
1000
Crank Angle [rpm]
T
o
r
q
u
e
[
N
m
]
torque from measurements
simulated comp./exp. torque
Crank angle [degrees] Crank angle [degrees]
Before top dead center both curves are basically the same, since pressure rise due to
combustion is not present in this region. There are differences however, since the simulated
compression/expansion torque is based on ideal conditions. This is accounted for by
introducing a correction factor, which is defined as follows:
) Min(T
) Min(T
cf
im comp/exp,s
t measuremen
=
(2.16)
The simulated compression/expansion torque is multiplied with this factor, resulting in an
equal minimum.
0 100 200 300 400 500 600 700
400
200
0
200
400
600
800
1000
Crank Angle [rpm]
T
o
r
q
u
e
[
N
m
]
torque from measurements
simulated comp./exp. torque
Crank angle [degrees]
0 100 200 300 400 500 600 700
400
200
0
200
400
600
800
1000
Crank Angle [rpm]
T
o
r
q
u
e
[
N
m
]
torque from measurements
simulated comp./exp. torque
0 100 200 300 400 500 600 700
400
200
0
200
400
600
800
1000
Crank Angle [rpm]
T
o
r
q
u
e
[
N
m
]
torque from measurements
simulated comp./exp. torque
Crank angle [degrees] Crank angle [degrees]
Figure 2.9: compression / expansion torque
Figure 2.10: corrected compression / expansion torque
LuK GmbH & Co. Technical University of Eindhoven
Luc Römers / XB department Traineeship report
 16 
In the region between 540° and 720° there are still differences however, especially at high
engine speeds. This occurs because the cylinder pressure does not completely drop to the
intake pressure after the combustion process. There is a distinct difference between intake
pressure and exhaust pressure.
0 100 200 300 400 500 600 700
400
200
0
200
400
600
800
Crank Angle [rpm]
T
o
r
q
u
e
[
N
m
]
torque from measurement
simulated comp./exp. torque
Crank angle [degrees]
0 100 200 300 400 500 600 700
400
200
0
200
400
600
800
Crank Angle [rpm]
T
o
r
q
u
e
[
N
m
]
torque from measurement
simulated comp./exp. torque
0 100 200 300 400 500 600 700
400
200
0
200
400
600
800
Crank Angle [rpm]
T
o
r
q
u
e
[
N
m
]
torque from measurement
simulated comp./exp. torque
Crank angle [degrees] Crank angle [degrees]
To take this into account, the simulated compression/expansion pressure will be defined on
three different regions:
¦
¦
¦
¦
¦
¦
¹
¦
¦
¦
¦
¦
¦
´
¦
° > =
° ≤ ≤ °


¹

\
 +
⋅ =
° < =
540 for P ) ( P
540 180 for
) V(
V V
P ) ( P
180 for P ) ( P
exhaust comp/exp
κ
d c
intake comp/exp
intake comp/exp
ϕ ϕ
ϕ
ϕ
ϕ
ϕ ϕ
(2.17)
P
exhaust
is an average pressure taken from the measurement in the region between 625° and 635°
crank angle. The simulated compression/expansion torque, based on this pressure definition, is
displayed in the next graph. It can be seen that the differences in the exhaust region are much
smaller now.
Figure 2.11: comp./exp. torque, without adapted exhaust pressure
LuK GmbH & Co. Technical University of Eindhoven
Luc Römers / XB department Traineeship report
 17 
0 100 200 300 400 500 600 700
400
200
0
200
400
600
800
Crank Angle [rpm]
T
o
r
q
u
e
[
N
m
]
torque from measurement
simulated comp./exp. torque
Crank angle [degrees]
0 100 200 300 400 500 600 700
400
200
0
200
400
600
800
Crank Angle [rpm]
T
o
r
q
u
e
[
N
m
]
torque from measurement
simulated comp./exp. torque
0 100 200 300 400 500 600 700
400
200
0
200
400
600
800
Crank Angle [rpm]
T
o
r
q
u
e
[
N
m
]
torque from measurement
simulated comp./exp. torque
Crank angle [degrees] Crank angle [degrees]
This new compression/expansion torque has an average value different from zero. The average
over one complete cycle can be expressed as follows [PENN02]:
ϕ
ϕ λ
ϕ λ
ϕ
π
π
π
π
d ) P P
d
r T
intake exhaust verage comp/exp,a


¹

\

−
⋅
+ ⋅ −
⋅
⋅ =
∫
) ( sin 1
) 2 sin(
) sin( (
4 4
1
2 2
4
3
2
(2.18)
The purpose of this average value will become clear later on, when the combustion process is
taken into account.
Hereby the analysis of the compression/expansion torque is completed. It is now possible to
simulate this torque using geometry data and three parameters: correction factor, intake
pressure and exhaust pressure.
( ) ( )
exhaust intake comp/exp comp/exp
P P cf geometry T T , , , , ϕ ϕ = (2.19)
2.5 Combustion torque
Describing a combustion process using thermodynamics is difficult, since it depends on a large
number of parameters. Therefore an empirical method, based on mathematics, is used. The
original empirical model is known as the Isermann method. LuK has improved this method.
The improved empirical model is called the Isermann/Pennec method. A short description of
both methods will be given.
The crankshaft torque resulting from the combustion process can found by calculating the
difference between total gas torque and compression/expansion torque. The total gas torque is
known from the cylinder pressure measurements; furthermore the compression/expansion
Figure 2.12: comp./exp. torque, with adapted exhaust pressure
LuK GmbH & Co. Technical University of Eindhoven
Luc Römers / XB department Traineeship report
 18 
torque has been simulated. After subtracting both curves, the combustion torque remains. This
is shown in the next figures.
0 100 200 300 400 500 600 700
100
0
100
200
300
400
500
600
Crank Angle [rpm]
T
o
r
q
u
e
[
N
m
]
(T
measured
T
comp/exp,sim
)
Crank angle [degrees]
0 100 200 300 400 500 600 700
100
0
100
200
300
400
500
600
Crank Angle [rpm]
T
o
r
q
u
e
[
N
m
]
(T
measured
T
comp/exp,sim
)
Crank angle [degrees] Crank angle [degrees]
2.5.1 Isermann method
The Isermann method [ISER99] uses the following formula to describe the combustion torque
shown in figure 2.14.
( )
° < < ° ⋅
⋅ ⋅
=
⋅ −
540 360 for e
T π 16
T
max
2
3
max
static
combustion
ϕ ϕ
ϕ
ϕ
ϕ
ϕ
) (
2
(2.20)
Figure 2.13: combustion torque is the difference between both curves
Figure 2.14: torque due to combustion
LuK GmbH & Co. Technical University of Eindhoven
Luc Römers / XB department Traineeship report
 19 
It contains two parameters, T
static
and ϕ
max
, which can be determined from the measured data.
The static torque is the average value over one full cycle:
∫ ∫
= =
π π
ϕ
ϕ
ϕ ϕ
4
0
4
0
d
d
dV
P
4π
1
)d ( T
4π
1
T
measured static (2.21)
The angle of maximum combustion torque (ϕ
max
) corresponds to the maximum in figure 2.14.
The effect of both parameters on the combustion torque is shown in the next graphs [PENN02].
0
200
400
600
800
1000
0 180 360 540 720
Crank angle[°]
T
o
r
q
u
e
[
N
m
]
Tstat=20Nm
Tstat=40Nm
Tstat=60Nm
0
200
400
600
800
1000
1200
0 180 360 540 720
Crank angle [°]
T
o
r
q
u
e
[
N
m
]
phi_max =
15°
phi_max =
25°
phi_max =
35°
Experience has shown that the accuracy of the Isermann model is limited. Therefore the
improved Isermann/Pennec model has been developed.
Figure 2.15: influence of T
stat
on the combustion torque
Figure 2.16: influence of φ
max
on the combustion torque
LuK GmbH & Co. Technical University of Eindhoven
Luc Römers / XB department Traineeship report
 20 
2.5.2 Isermann/Pennec method
This method uses two additional shape parameters to simulate the combustion torque more
accurately. The derivation of this empirical model is beyond the scope of this report. The final
Isermann/Pennec combustion torque is a function of the crank angle and four parameters:
( ) β α ϕ ϕ , , , ,
max static combustion combustion
T T T = (2.22)
The effect of the additional shape parameters α and β on the combustion torque is shown in the
graphs below [PENN02].
0
100
200
300
400
500
0 180 360 540 720
Crank angle [°]
T
o
r
q
u
e
[
N
m
]
beta = 0.3
beta = 0.5
beta = 0.7
0
100
200
300
400
500
600
0 180 360 540 720
Crank angle [°]
T
o
r
q
u
e
[
N
m
]
alpha = 1
alpha = 3
alpha = 6
Figure 2.17: influence of β on the combustion torque
Figure 2.18: influence of α on the combustion torque
LuK GmbH & Co. Technical University of Eindhoven
Luc Römers / XB department Traineeship report
 21 
Having found the values for T
static
and ϕ
max
, the values for α and β can be optimized. To
compare different combinations, the standard deviation between measurement and simulation
is calculated:
( ) ϕ ϕ ϕ
π
β α σ
π
d T T
simulation measured
∫
− =
4
0
2 2
) ( ) (
4
1
) , (
(2.23)
Experience has shown results similar to figure 2.19. There is no unique combination of α and β
where an absolute minimum standard deviation is reached. There is a line on which the results
are comparably good. This simplifies the optimization process considerably. One shape
parameter can have a fixed value, after which the other shape parameter is optimized to find
the minimum standard deviation.
0.1
0.2
0.3
0.4
0.5
1
2
3
4
5
6
7
8
9
10
0
10
20
30
40
50
60
70
80
90
100
110
s
i
g
m
a
[
N
m
]
beta [/]
alpha [/]
100110
90100
8090
7080
6070
5060
4050
3040
2030
1020
010
0 100 200 300 400 500 600 700
100
0
100
200
300
400
500
Crank Angle [rpm]
T
o
r
q
u
e
[
N
m
]
(T
measured
T
comp/exp,simulated
)
T
combustion,simulated
Crank angle [degrees]
0 100 200 300 400 500 600 700
100
0
100
200
300
400
500
Crank Angle [rpm]
T
o
r
q
u
e
[
N
m
]
(T
measured
T
comp/exp,simulated
)
T
combustion,simulated
Crank angle [degrees] Crank angle [degrees]
Figure 2.19: standard deviation for different combinations of α and β [PENN02]
Figure 2.20: simulated combustion torque
LuK GmbH & Co. Technical University of Eindhoven
Luc Römers / XB department Traineeship report
 22 
Figure 2.20 gives an example of a combustion torque simulated with the Isermann/Pennec
method. Here α was fixed, after which β was optimized to find the best fit.
The Isermann/Pennec empirical model makes it possible to accurately describe the torque
resulting from combustion, using only three parameters: static torque, angle of maximum
combustion torque and beta.
( ) ( ) β ϕ ϕ ϕ , , ,
max static combustion combustion
T T T = (2.24)
2.6 Total torque resulting from gas forces
The total crankshaft torque resulting from gas forces is the sum of compression/expansion
torque and combustion torque. This dynamic torque can now be accurately simulated using
only geometry data and six parameter values:
( )
exhaust intake comp/exp comp/exp
P P cf geometry T T , , , = (2.25)
( ) β ϕ , ,
max static combustion combustion
T T T = (2.26)
( ) β ϕ , , , , , ,
, , max exhaust intake static sim gas sim gas
cf P P T geometry T T = (2.27)
A result of this method is displayed in figure 2.21. A comparison between measurement and
simulation is shown.
0 100 200 300 400 500 600 700
200
100
0
100
200
300
400
500
600
Crank Angle [rpm]
T
o
r
q
u
e
[
N
m
]
dynamic gas torque, measured
dynamic gas torque, simulated
Figure 2.21: dynamic torque form measurement and simulation
LuK GmbH & Co. Technical University of Eindhoven
Luc Römers / XB department Traineeship report
 23 
2.7 Parameter maps
The previous analysis is based on only one single operating point. The car manufacturer
delivers pressure measurements in many operating points. For each pressure measurement, the
torque resulting from gas forces is simulated using the Isermann/Pennec method.
The engine speed at which a cylinder pressure measurement is conducted is known. A value
for the static torque results from the pressure analysis. Note that this is the static torque of only
one cylinder. Multiplying this value with the number of cylinders delivers the total static
engine torque.
cyl cylinder static engine static
N T T ⋅ =
, ,
(2.29)
The position in the engine map, at which the pressure measurement is conducted, is now
known. Five additional parameters belong to this particular operating point. Because pressure
measurements at many different operating points are available, it is possible to create maps for
each of these parameters. These maps describe parameter variation with engine speed and
static engine torque:
( )
static intake intake
,T P P ω =
( )
static exhaust exhaust
,T P P ω =
( )
static
,T cf cf ω = (2.30)
( )
static max max
,T ω ϕ ϕ =
( )
static
,T ω β β =
With these five parameter maps (and geometry data), it is possible to simulate the dynamic
engine torque in every operating point of the engine. An example will be used to show how the
maps are constructed.
LuK GmbH & Co. Technical University of Eindhoven
Luc Römers / XB department Traineeship report
 24 
The next figure shows the engine map of a 2.0L gasoline engine. In this particular case the car
manufacturer has delivered cylinder pressure data in over 220 operating points. For each of
these measurements the dynamic torque is simulated, using the method described before. This
results in the following engine map:
0 1000 2000 3000 4000 5000 6000 7000 8000
0
50
100
150
200
250
Speed [rpm]
T
s
t
a
t
e
n
g
i
n
e
[
N
m
]
Engine map
It can be seen that pressure measurements have been conducted at idle, coast, partload and
fullload operation. Five parameters, needed to reconstruct the dynamic engine torque, are
found for each of these operating points. These parameters can be displayed as a function of
engine speed and static engine torque. For this particular engine the following results are
found.
0
1000
2000
3000
4000
5000
6000
7000
0
50
100
150
200
0
0.5
1
1.5
2
2.5
3
3.5
Speed [rpm]
Intake pressure
T
stat
engine
[Nm]
P
in
t
a
k
e
[
b
a
r
]
0
1000
2000
3000
4000
5000
6000
7000
0
50
100
150
200
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
Speed [rpm]
Exhaust pressure
T
stat
engine
[Nm]
P
e
x
h
a
u
s
t
[
b
a
r
]
Figure 2.22: engine map, operating points in which pressure data is available
Figure 2.23: intake pressure Figure 2.24: exhaust pressure
LuK GmbH & Co. Technical University of Eindhoven
Luc Römers / XB department Traineeship report
 25 
These loose values need to be converted to maps, covering all possible engine operating points.
Before these maps can be created, possible “holes” in the dataset need to be taken into account.
The operating points at which pressure data is delivered are different for each engine. The
example used above shows an ideal case, with over 220 operating points covering the engine
map. Very often however, this is not the case. Some common problems, caused by a limited
amount of measurements, are as follows:
no cylinder pressure measurements at the start/idle region (300  750 rpm)
missing cylinder pressure data at coast operation
no, or little, cylinder pressure data at partload operation
missing cylinder pressure data at high engine speeds
Each of these cases has to be taken into account. This will be explained later on by using a
second (worstcase) example.
First the example of figure 2.22 is treated. Here only data at the start/idle region is missing. In
this region a number of points are added, based on values of experience. Secondly, a line of
0
1000
2000
3000
4000
5000
6000
7000
0
50
100
150
200
0.2
0.3
0.4
0.5
0.6
0.7
Speed [rpm]
Shape factor beta
T
stat
engine
[Nm]
B
e
t
a
[

]
0
1000
2000
3000
4000
5000
6000
7000
0
50
100
150
200
20
25
30
35
40
45
50
Speed [rpm]
Combustion angle
T
stat
engine
[Nm]
P
h
i
m
a
x
[
d
e
g
r
e
e
s
]
0
1000
2000
3000
4000
5000
6000
7000
0
50
100
150
200
0.85
0.9
0.95
1
1.05
1.1
1.15
Speed [rpm]
Correction factor
T
stat
engine
[Nm]
k
f
[

]
Figure 2.25: combustion angle Figure 2.26: correction factor
Figure 2.27: shape factor beta
LuK GmbH & Co. Technical University of Eindhoven
Luc Römers / XB department Traineeship report
 26 
points at maximum engine torque is added. This is done to create a square engine map. The
reason for this has to do with the DyFaSim software (the DyFaSim combustion engine module
requires a square mesh for each parameter). The parameter values stay constant at torques
higher than the fullload torque curve. After these additions, a linear interpolation along every
load line is carried out. Interpolated points are added with a 50rpm spacing. The results are
displayed in the following engine map.
There are now enough operating points to create the full parameter maps. A number of
interpolation techniques have been considered. Experience has shown that basic linear
interpolation between the operating points delivers the best results. Higher order techniques
(cubic, spline) cause problems when the number of operating points is limited. The five
parameter maps are displayed in the following figures.
0 1000 2000 3000 4000 5000 6000 7000
0
50
100
150
200
250
Engine map
Speed [rpm]
T
s
t
a
t
e
n
g
i
n
e
[
N
m
]
Figure 2.28: engine map, after adding points and interpolation
Figure 2.29: intake pressure map Figure 2.30: exhaust pressure map
LuK GmbH & Co. Technical University of Eindhoven
Luc Römers / XB department Traineeship report
 27 
These maps are generated using an engine speed spacing of 50 rpm and a static torque spacing
of 2,5 Nm.
Figure 2.31: combustion engine map Figure 2.32: correction factor map
Figure 2.33: beta map
LuK GmbH & Co. Technical University of Eindhoven
Luc Römers / XB department Traineeship report
 28 
Now a second example is treated. It is used to illustrate some other techniques to “fill the
map”. The previous example was a bestcase scenario, with over 220 operating points
available. Now a worstcase scenario is considered. Another 2.0L gasoline engine is
considered. For this engine very little pressure data is delivered. Only fullload pressure
measurements have been conducted, as can be seen in the next figures.
1000 2000 3000 4000 5000 6000 7000
0
50
100
150
200
250
300
Speed [rpm]
T
s
t
a
t
e
n
g
i
n
e
[
N
m
]
Engine map
0
1000
2000
3000
4000
5000
6000
0
50
100
150
200
250
300
0
0.5
1
1.5
2
2.5
3
3.5
Speed [rpm]
Intake pressure
T
stat
engine
[Nm]
P
in
t
a
k
e
[
b
a
r
]
0
1000
2000
3000
4000
5000
6000
0
50
100
150
200
250
300
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
Speed [rpm]
Exhaust pressure
T
stat
engine
[Nm]
P
e
x
h
a
u
s
t
[
b
a
r
]
0
1000
2000
3000
4000
5000
6000
0
50
100
150
200
250
300
0.85
0.9
0.95
1
1.05
1.1
1.15
Speed [rpm]
Correction factor
T
stat
engine
[Nm]
c
f
[

]
Figure 2.34: engine map
Figure 2.35: intake pressure Figure 2.36: exhaust pressure
Figure 2.37: combustion angle Figure 2.38: correction factor
0
1000
2000
3000
4000
5000
6000
0
50
100
150
200
250
300
20
25
30
35
40
45
50
Speed [rpm]
Combustion angle
T
stat
engine
[Nm]
P
h
i
m
a
x
[
d
e
g
r
e
e
s
]
LuK GmbH & Co. Technical University of Eindhoven
Luc Römers / XB department Traineeship report
 29 
In this case more actions are needed to fill the engine map. Again operating points in the
start/idle region are added, as well as the line at maximum torque. Furthermore additional
points at coast operation are added. Parameter values at these added points are again based on
values of experience. Next a linear interpolation along the load lines is made, after which the
parameter maps can be created. The results are shown in the next figures.
0 1000 2000 3000 4000 5000 6000 7000
0
50
100
150
200
250
300
Engine map
Speed [rpm]
T
s
t
a
t
e
n
g
i
n
e
[
N
m
]
0
1000
2000
3000
4000
5000
6000
0
50
100
150
200
250
300
0.2
0.3
0.4
0.5
0.6
0.7
Speed [rpm]
Shape factor beta
T
stat
engine
[Nm]
B
e
t
a
[

]
Figure 2.39: shape factor beta
Figure 2.40: engine map, after adding points and interpolation
LuK GmbH & Co. Technical University of Eindhoven
Luc Römers / XB department Traineeship report
 30 
Figure 2.41: intake pressure map Figure 2.42: exhaust pressure map
Figure 2.43: combustion engine map Figure 2.44: correction factor map
Figure 2.45: beta map
LuK GmbH & Co. Technical University of Eindhoven
Luc Römers / XB department Traineeship report
 31 
These results show that, although only limited data is available, parameter values are found for
all engine operating points. This way a good estimate for the dynamic engine torque, based on
these interpolated parameter values can be made.
2.8 Reducing the parameter maps
The current parameter maps are very accurate, since they are created with a fine mesh:
speed spacing = 50 rpm
static torque spacing = 2,5 Nm
This results in a very large matrix for each parameter. Since the DyFaSim combustion engine
module is not designed for matrices of this size, the amount of data will have to be reduced. A
new spacing for engine speed and static torque can be defined, after which the matrices are
reduced in size using this new spacing. In case a large spacing is chosen, accuracy of the
engine model will be lost. A suitable compromise between data reduction and loss of accuracy
will have to be made. Experience has shown that the following spacing delivers satisfying
results:
speed spacing = 250 rpm
static torque spacing = 25 Nm
Original points, lying in between, are erased from the matrix. By implementing this new
spacing, only 2% of the original amount of data remains.
To visually check the quality of this reduction process, both the full parameter maps and the
reduced parameter maps are displayed. An example is given in the following figures. This
reduction process belongs to the first example of the previous paragraph. In the MatLab
program the axes of both figures are linked. This makes it possible to rotate the figures
simultaneously.
Figure 2.46: reduction of the intake pressure map
LuK GmbH & Co. Technical University of Eindhoven
Luc Römers / XB department Traineeship report
 32 
Figure 2.47: reduction of the exhaust pressure map
Figure 2.48: reduction of the combustion angle map
Figure 2.49: reduction of the correction factor map
LuK GmbH & Co. Technical University of Eindhoven
Luc Römers / XB department Traineeship report
 33 
Experience has shown that a good visual quality check is possible this way. In case the reduced
maps are unsatisfying, resulting from an excessively large spacing, a new (smaller) spacing can
be defined.
2.9 Peaktopeak comparison
Although the visual quality check can be very useful, a more quantitative method is developed.
A comparison will be made between the measured dynamic torque (resulting from pressure
measurements) and the simulated dynamic torque. For each operating point in which
measurement data is available, the dynamic torque is simulated using the full as well as the
reduced parameter maps.
Now a comparison between the peaktopeak torques is made. The peaktopeak torque is the
difference between the minimum and maximum of the dynamic torque curve over two
crankshaft revolutions. Here the dynamic torque of the complete engine is considered (figure
2.52). The total engine torque can be found by cascading the onecylinder torque into a full
engine. For an ncylinder inline engine the total engine torque is given by:
∑
−
=
+ =
1
0
) * (
cyl
n
i
cyl
cyl engine
n
4π
i T T ϕ
(2.31)
The next figures show the cylinder cascading for a 4cylinder engine. The dynamic torques
from measurements and simulation are displayed in one graph. Both simulation with the full
maps and simulation with the reduced maps are displayed.
Figure 2.50: reduction of the beta map
LuK GmbH & Co. Technical University of Eindhoven
Luc Römers / XB department Traineeship report
 34 
The peaktopeak torque for these three curves is given by:
) min(T ) max(T T
ed eng,measur ed eng,measur d pp,measure
− =
) min(T ) max(T T
full eng,sim full eng,sim full pp,sim , , ,
− =
(2.32)
) min(T ) max(T T
reduced eng,sim reduced eng,sim reduced pp,sim , , ,
− =
T
peaktopeak
Figure 2.51: dynamic torque of one cylinder
Figure 2.52: dynamic torque of full engine
LuK GmbH & Co. Technical University of Eindhoven
Luc Römers / XB department Traineeship report
 35 
For the 220 operating points of the previous example, the 1cylinder results are displayed in
figure 2.53.
The full (4cylinder) engine torque is displayed in figure 2.54. From these results, the peakto
peak torques can be found.
Figure 2.53: dynamic torque of one cylinder, all available operating points
Figure 2.54: dynamic torque of full engine, all available operating points
LuK GmbH & Co. Technical University of Eindhoven
Luc Römers / XB department Traineeship report
 36 
The peaktopeak values are determined and plotted as a function of engine speed and static
engine torque. The graphs below show a threedimensional view of the results, along with the
twodimensional side views.
Figure 2.57 displays the peaktopeak torque as a function of static engine torque. A clear
linear relation can be seen. This result is expected for all naturally aspirated engines.
Figure 2.55: absolute peaktopeak values
Figure 2.56: absolute peaktopeak values, side view 1 Figure 2.57: absolute peaktopeak values, side view 2
LuK GmbH & Co. Technical University of Eindhoven
Luc Römers / XB department Traineeship report
 37 
Based on these absolute values, it is difficult to evaluate the quality of the simulated curves.
Therefore the relative peaktopeak differences between measurement and both simulations are
calculated:
( )
100%
T
T T abs
diff
measured pp
measured pp full sim pp
full im relative,s
⋅


¹

\
 −
=
,
, , ,
, (2.33)
( )
100%
T
T T abs
diff
measured pp
measured pp reduced sim pp
reduced im relative,s
⋅


¹

\
 −
=
,
, , ,
,
(2.34)
The results are displayed in the following graphs. The red points represent the relative
difference between measurement and simulation with the accurate parameter maps. The green
points represent the relative difference between measurement and simulation with the reduced
maps.
Figure 2.58: relative peaktopeak difference
Figure 2.59: relative peaktopeak difference, side view 1 Figure 2.60: relative peaktopeak difference, side view 2
LuK GmbH & Co. Technical University of Eindhoven
Luc Römers / XB department Traineeship report
 38 
The simulation with the accurate maps is clearly better than the simulation with the reduced
maps. At high engine torques, this loss of accuracy is less obvious. Furthermore it can be seen
that the results at coast are very bad. There is no combustion torque at coast operation, only
compression/expansion torque. The cylinder pressure at coast cannot be described accurately
by the method explained in chapter 2.4.
These relative differences are based on dynamic torque resulting from gas forces only. If the
mass torque is also taken into account the absolute peaktopeak values change, which has an
effect on the relative error size. The next graphs show the results in case mass torque is
included.
Figure 2.61: absolute peaktopeak values, mass torque taken into account
Figure 2.62: absolute peaktopeak values, side view 1 Figure 2.63: absolute peaktopeak values, side view 2
LuK GmbH & Co. Technical University of Eindhoven
Luc Römers / XB department Traineeship report
 39 
In figure 2.62 the effect of the mass torque can be seen. At low engine speeds the combustion
torque is dominant and at high engine speeds the mass torque is dominant. This occurs because
combustion torque and mass torque are phase shifted and the mass torque increases
quadratically with engine speed. Comparing figure 2.62 and figure 2.56 clearly shows the
effect of the mass torque.
Again the relative differences are calculated, the results are displayed below.
Now all relative differences are smaller than 10%. The large error at coast operation has also
disappeared. Since mass torque is very dominant at coast, the bad accuracy of the gas torque
simulation is of minimum importance.
Figure 2.64: relative peaktopeak difference, mass torque taken into account
Figure 2.65: relative peaktopeak difference, side view 1 Figure 2.66: relative peaktopeak difference, side view 2
LuK GmbH & Co. Technical University of Eindhoven
Luc Römers / XB department Traineeship report
 40 
2.10 Output for DyFaSim software
If the peaktopeak analysis is satisfactory, the created parameter maps need to be stored in a
format suitable for the DyFaSim software. Therefore an outputfile is created in ASCIIformat,
containing the reduced parameter maps along with the geometry data of the engine. This file
can be uploaded into the DyFaSim combustion engine module, after which the engine module
is ready for use.
The figure below shows the main window of the DyFaSim combustion engine module. In this
case a generated ASCII file of a 6cylinder engine has been uploaded.
For each cylinder, the geometry data and parameter maps are defined (inside the red square).
With these values, DyFaSim is able to simulate the dynamic engine torque in every operating
point.
Figure 2.67: DyFaSim combustion engine module
LuK GmbH & Co. Technical University of Eindhoven
Luc Römers / XB department Traineeship report
 41 
The figures below show the windows containing the engine data. Geometry data is collected in
figure 2.68. The five parameter maps are defined in figure 2.69. Each “edit” button opens a
parameter map. An example is shown in figure 2.70.
Figure 2.70 shows an intake pressure map in DyFaSim. In this case it is a map with a 250rpm
speed spacing and a 50Nm torque spacing. In the bottom left corner the intake pressure matrix
is shown. On the right side a twodimensional graph of this matrix is displayed.
Figure 2.68: geometry data
Figure 2.70: intake pressure map
Figure 2.69: parameter maps
LuK GmbH & Co. Technical University of Eindhoven
Luc Römers / XB department Traineeship report
 42 
3. Generation of standard engine models
Although pressurebased engine models deliver very accurate results, cylinder pressure data is
not always available. In many cases only basic engine data is known. For these situations a
second operating mode has been developed: standard engine model generation.
This operating mode needs only basic engine data to create a full model. The modelinput
consists of:
• engine geometry data
• net torque curve at full load
• drag torque curve
• intake pressure curve at full load
3.1 Combustion engine data sheet
LuK has developed a standardised combustion engine data sheet. This Excel data sheet is filled
out by the car manufacturer. An example of the data sheet is given in figure 3.1.
Luk GmbH & Co.
77815 Bühl
Version 2.2.1, 07.12.05
Date
There is 0 essential data missing essential optional
Engine geometry
Value Unit
Nominal displacement 2000 [cm^3]
Number of cylinders 4 [/]
Bore 81 [mm]
Stroke 95.5 [mm]
Compression ratio 18 [/]
Connecting rod length 144 [mm]
Oscillating mass 1029 [kg]
Crankshaft inertia 0.04 [kgm^2]
Calculated Displacement 1968 [cm^3]
Power at full load
Speed [rpm] Net torque [Nm] Drag torque [Nm] Gross torque [Nm] Engine Power [kW]
250 70 30 100 1.8
500 100 30 130 5.2
750 120 30 150 9.4
1000 170 30 200 17.8
1250 200 32 232 26.2
1500 240 34 274 37.7
1750 300 36 336 55.0
2000 320 38 358 67.0
2500 320 42 362 83.8
3000 320 46 366 100.5
3500 280 54 334 102.6
4000 240 60 300 100.5
4500 150 66 216 70.7
5000 0 0.0
5500 0 0.0
6000 0 0.0
6500 0 0.0
7000 0 0.0
Speed [rpm] Pressure [bar]
250 0.5
500 0.8
750 1
1000 1.2
1250 1.4
1500 1.6
1750 2
2000 2.2
2500 2.3
3000 2.4
3500 2.5
4000 2.6
4500 2.6
5000 2.6
5500
6000
6500
7000
Intake pressure
Torque curve at full load
Comments
at speed [rpm]
20003000
Basic Data Sheet for Combustion Engine
Manufacturer
Company name
Engine name & calibration version
2L Diesel
Engineer responsible for data
Engineer (Company)
Tel. number
5000 103 3500 320
Max engine speed [rpm] Max power [kW] at speed [rpm] Max net torque [Nm]
+00 00 000 000
Email
engineer@company.com DD.MM.YY
Fuel type Fuel injection system
Unit injector or Unit pump system Turbocharger (1X)
Air charging system
Diesel
100
0
100
200
300
400
0 1000 2000 3000 4000 5000 6000 7000
Speed [rpm]
T
o
r
q
u
e
[
N
m
]
0.0
25.0
50.0
75.0
100.0
125.0
P
o
w
e
r
[
k
W
]
Net Trq
Drag Trq
Gross Trq
Power
0
0.5
1
1.5
2
2.5
3
0 1000 2000 3000 4000 5000 6000 7000
Speed [rpm]
I
n
t
a
k
e
p
r
e
s
s
u
r
e
[
b
a
r
]
Intake Pressure
Figure 3.1: combustion engine data sheet
LuK GmbH & Co. Technical University of Eindhoven
Luc Römers / XB department Traineeship report
 43 
Since the data sheet has a standard layout, it is possible to extract information in an automated
way. MatLab creates an interface with the Excel data sheet and imports the required data. The
car manufacturer often returns a data sheet in which data is missing. In this case the missing
data is replaced with experience values based on a similar engine. This process will be
explained in the next paragraph.
3.2 Replacing missing data by estimates
As stated in the beginning of this chapter, there are four essential inputs. In case any of these
inputs are unavailable, they need to be replaced by a good estimate. How these estimates are
determined will be explained in this paragraph.
3.2.1 Geometry data
In case geometry data is unavailable, the user has to define:
• engine type (gasoline or diesel)
• displacement volume
• number of cylinders
All other geometry data is replaced by values of experience. These experience values are
different for gasoline and diesel engines, as can be seen in the tables below.
Gasoline Value Unit
Nominal displacement User defined [liter]
Number of cylinders User defined []
Bore 0,080 [m]
Stroke 0,100 [m]
Compression ratio 10,5 []
Connecting rod length 0,150 [m]
Oscillating mass 0,5 [kg]
3.2.2 Net torque curve at full load
In case the net torque curve at full load is unavailable, the user needs to define a maximum
value for the net torque. A predefined torque curve is first scaled up or down, depending on the
displacement volume. Secondly this scaled curve will be adapted to match the userdefined
maximum.
Net torque curves are predefined for a four different situations:
• gasoline, naturally aspirated
• gasoline, turbocharged
• diesel, naturally aspirated
• diesel, turbocharged
Diesel Value Unit
Nominal displacement User defined [liter]
Number of cylinders User defined []
Bore 0,080 [m]
Stroke 0,100 [m]
Compression ratio 18 []
Connecting rod length 0,150 [m]
Oscillating mass 1,0 [kg]
LuK GmbH & Co. Technical University of Eindhoven
Luc Römers / XB department Traineeship report
 44 
The four predefined curves are displayed in figure 3.2.
These are experience curves for 2.0L engines of different configurations. Depending on the
chosen displacement volume these curves are scaled up or down:
2
V
) ( T ) ( T
d
predefined net scaled net
⋅ = ω ω
, ,
(3.1)
This scaled curve is then compared to the userdefined maximum torque. In case the scaled
torque curve exceeds the userdefined maximum, all values above this maximum are cut off:
( )
max scaled net
T ) ( T max ≡ ω
,
(3.2)
In case the scaled torque curve lies below the userdefined maximum it is scaled up once again:
( ) ) ( T max
T
) ( T ) ( T
scaled net
max
scaled net scaled net
ω
ω ω
,
, ,
⋅ = (3.3)
This method has proven to deliver a good estimate for the net torque curve.
3.2.3 Drag torque curve
The car manufacturer delivers net torque data. However the combustion engine model is based
on gross torque. Therefore drag torque has to be taken into consideration.
In case no drag torque data is available a similar process as for the net torque curve is used.
Again, predefined curves form the basis. For the drag torque there is one curve for a 2.0L
diesel engine and one curve for a 2.0L gasoline engine (figure 3.3).
Figure 3.2: standard full load torque curves
LuK GmbH & Co. Technical University of Eindhoven
Luc Römers / XB department Traineeship report
 45 
These curves are also scaled using the displacement volume:
2
V
) ( T ) ( T
d
predefined drag scaled drag
⋅ = ω ω
, ,
(3.4)
3.2.4 Intake pressure curve at full load
The intake pressure at full load is also predefined for four different situations. Again the
division is made between diesel and gasoline, turbocharged and naturally aspirated. In case the
engine is naturally aspirated, the intake pressure will be equal to the surrounding atmospheric
pressure, approximately 1 bar. For turbocharged engines, predefined curves are used. The
curves for turbocharged gasoline and diesel engines are shown in figure 3.4.
Figure 3.3: standard drag torque curves
Figure 3.4: standard intake pressure curves
LuK GmbH & Co. Technical University of Eindhoven
Luc Römers / XB department Traineeship report
 46 
4. Graphical User Interface in MatLab
Both methods described in the previous chapters have been implemented in MatLab. The main
goal of the program is easy operation. This has been accomplished by using the Graphical User
Interface tools available in MatLab. For both the pressurebased and the standard engine model
generation an example will be given.
A compiled version of the program is available. This version can be installed and used on
every computer within LuK, without the need for a MatLab license. Because MatLab licenses
are expensive and scarce, a standalone version created with MatLab compiler has many
advantages.
4.1 Graphical User Interface, pressurebased engine model
To explain the graphical user interface for the pressurebased operating mode, an example is
given. A DyFaSim engine model is needed for a new engine. The manufacturer has delivered
the geometry data and has conducted cylinder pressure measurements.
When the program is started, the user has the possibility to select a language:
After selecting a language, the user can choose between the two operating modes. Since
cylinder pressure data is available, the pressurebased model generation is chosen.
Figure 4.1: language selection
Figure 4.2: selecting an operating mode
LuK GmbH & Co. Technical University of Eindhoven
Luc Römers / XB department Traineeship report
 47 
MatLab now shows the database in which the geometry data of all engines is stored. To protect
customer data, a nameless exampleengine is chosen.
The engine named “example.txt” is selected. A new window opens in which all available
cylinder pressure data for this engine is shown. There is cylinder pressure available at idle, 6
partloads and fullload.
In this case all pressure data is selected. Of course it is also possible to select less data.
Figure 4.3: manufacturer / engine selection
Figure 4.4: pressure data selection
LuK GmbH & Co. Technical University of Eindhoven
Luc Römers / XB department Traineeship report
 48 
An overview of the selected engine and pressure data is displayed in two info boxes.
The program asks the user whether pressure data at coast is available. If coast data is not
available, which is the case for this example, a dialog box appears. In it are predefined
experience values for the intake pressure and exhaust pressure at coast. The user has the
possibility to adapt these values, in case this is necessary.
At this point the program is ready to start the calculation of the parameter values for all
selected cylinder pressure measurements. The user has the possibility to view the results of the
calculation process for every selected pressure measurement. Because the plot actions slow the
program down, one can also choose not to view these initial results.
Figure 4.6: info box showing selected pressure data Figure 4.5: info box showing selected engine
Figure 4.7: question dialog, coast data Figure 4.8: dialog box for coast values
Figure 4.9: question dialog, plot actions
LuK GmbH & Co. Technical University of Eindhoven
Luc Römers / XB department Traineeship report
 49 
If “Yes” is chosen, the following figure appears. On the left side the dynamic torque from
measurement and simulation is shown. These curves are continuously updated during the
calculation process. On the right side, the engine map shows the (black) operating points that
have already been calculated. The purple point corresponds with the current dynamic torque
shown on the left.
When the calculation process is completed, the user can adapt the simulated dynamic idle
torque. Pressure measurements in this operating point are often unreliable. Therefore the
simulated static torque at idle and a value known from experience are shown. The user can
decide which value should be used.
The parameter maps can now be calculated. A dialog box appears in which the user can insert
values for the speed and torque spacing. These values will be used to generate the reduced
maps.
Figure 4.10: dynamic torque plot, along with position in engine map
Figure 4.11: possibility to adapt idle simulation
LuK GmbH & Co. Technical University of Eindhoven
Luc Römers / XB department Traineeship report
 50 
The final results are displayed using six figures, i.e. one for the engine map and five for the
parameter maps. The user now has the ability to visually evaluate the results. The parameter
graphs show the accurate and the reduced maps. The axes of both graphs are linked, so they
can be turned simultaneously.
Figure 4.12: dialog box for spacing values
Figure 4.13: graphs showing engine map and parameter maps
LuK GmbH & Co. Technical University of Eindhoven
Luc Römers / XB department Traineeship report
 51 
The program continues after pressing a random key on the keyboard. If the reduction process is
unsatisfying, due to large spacing values, new spacing values can be entered.
If the user is satisfied with the results, the peaktopeak analysis is made. The results are
displayed in six figures.
Figure 4.14: possibility to redefine spacing values
Figure 4.15: single cylinder dynamic torque Figure 4.16: dynamic torque of full engine
Figure 4.18: relative difference, only gas torque Figure 4.17: absolute peaktopeak values, only gas torque
LuK GmbH & Co. Technical University of Eindhoven
Luc Römers / XB department Traineeship report
 52 
After viewing the results of the peaktopeak analysis, a DyFaSim engine model is created. The
user can select a location to save the final model.
Finally there is a possibility to save a 1cylinder representation of the engine model. These
single cylinder models are used in case crankshaft dynamics are taken into account. In this case
each piston is a separate mass, on which a 1cylinder combustion engine is mounted.
Figure 4.19: absolute peaktopeak values, gas & mass torque Figure 4.20: relative difference, gas & mass torque
Figure 4.21: select a location to save the DyFaSim engine model
Figure 4.22: possibility to create a 1cylinder engine model
LuK GmbH & Co. Technical University of Eindhoven
Luc Römers / XB department Traineeship report
 53 
After saving the DyFaSim model(s), the operation is completed. The following window
appears:
There is a possibility to make a printout of the final results. For this purpose, all of the
generated maps are displayed in one figure.
Figure 4.23: info box, informing user that the operation is completed
Figure 4.24: overview of results, possibility to make a printout
LuK GmbH & Co. Technical University of Eindhoven
Luc Römers / XB department Traineeship report
 54 
4.2 Graphical User Interface, standard engine model
To show the graphical user interface for the standard operating mode, the following example is
used: an engine model is needed for a 2.0L 4cylinder diesel engine. For this engine no
cylinder pressure data is available. The only known data is:
geometry data
torque curve at full load
drag torque curve
intake pressure at full load
The data is collected in an Excel data sheet (this standard data sheet is filled out by the car
manufacturer). When the program is started, the user has the possibility to select a language:
Now the user can choose between the two operating modes. Since no cylinder pressure data is
available, the standard model generation is needed.
Now the user can define whether geometry data is available for this engine. In case no
geometry data is available, estimated values are used.
Figure 4.25: language selection
Figure 4.26: selecting an operating mode
Figure 4.27: dialog box, geometry data
LuK GmbH & Co. Technical University of Eindhoven
Luc Römers / XB department Traineeship report
 55 
For this example geometry data is available. The user can now select the data sheet in which
the geometry data is collected. MatLab opens the data sheet, reads the data and displays it in a
dialog box. The user now has the ability to check and/or change the displayed data.
The next question is related to the air supply. In this example a turbocharger is present.
Figure 4.28: selection of the Excel data sheet
Figure 4.29: dialog box, engine data
Figure 4.30: question dialog, air supply
LuK GmbH & Co. Technical University of Eindhoven
Luc Römers / XB department Traineeship report
 56 
The program asks the user whether full load data is available. This is the case, so the data can
be extracted from the data sheet. In case no full load data is available, a predefined full load
curve is scaled up or down.
The drag torque is treated in a similar way. In case no data is available, a predefined drag
torque curve is scaled up or down, depending on the displacement volume. For this example
drag torque data is available in the data sheet.
When no intake pressure data (at full load) is available, predefined values of a similar engine
are used. In this case the intake pressure characteristic is filled out in the data sheet.
The last question is related to the behavior of the throttle valve at engine shutdown (stop). This
information is needed to estimate the parameter maps in the start/stop region.
Figure 4.31: question dialog, full load characteristic
Figure 4.32: question dialog, drag torque
Figure 4.33: question dialog, intake pressure
Figure 4.34: question dialog, throttle valve
LuK GmbH & Co. Technical University of Eindhoven
Luc Römers / XB department Traineeship report
 57 
Before creating the script file for DyFaSim, the resulting engine map and intake pressure map
are shown (figure 4.35). On the left side the net and gross (fullload) torque curves are shown,
along with the drag torque curve. Of course the drag torque is the difference between the first
two. The power output is also shown, scaled on the right Yaxis.
After pressing a random key on the keyboard the program continues. The DyFaSim script file
is generated, after which the user can select a location to store it.
Figure 4.35: resulting engine map and intake pressure map
Figure 4.36: select a location to save the DyFaSim engine
LuK GmbH & Co. Technical University of Eindhoven
Luc Römers / XB department Traineeship report
 58 
Again there is a possibility to create a 1cylinder representation of the engine.
After saving the DyFaSim model(s), the operation is completed. The next window appears,
after which the program automatically shuts down.
Figure 4.37: possibility to create a 1cylinder engine model
Figure 4.38: info box, informing user that the operation is completed
LuK GmbH & Co. Technical University of Eindhoven
Luc Römers / XB department Traineeship report
 59 
5. Comparison between pressurebased and standard models
To investigate the differences between pressurebased and standard engine models, a harmonic
analysis will be made. The harmonic analysis will be based on the fullload torque, since this is
the most critical operating point.
Each engine model is loaded into DyFaSim and mounted on a large rotational mass. A
simulation is performed, during which the mass is accelerated with a large constant torque.
Upon this large constant torque, the dynamic fullload engine torque is superimposed. The
applied torque accelerates the mass from 1000 rpm until maximum engine speed. A harmonic
analysis of the torque signal is made. The torque amplitude of the main excitation order is
compared. For 4cylinder engine this is the second order (two excitations per revolution). For a
6cylinder engine this is the third order (three excitations per revolution).
The comparison is made for two very different engines:
• 2.0 L, 4cylinder diesel, turbocharged
• 3.0 L, 6cylinder gasoline, naturally aspirated
For both engines a pressurebased model and a standard model has been generated. The
resulting DyFaSim engine models are used to perform the harmonic analysis.
Figure 5.1 shows the result for the 4cylinder (turbocharged) diesel engine. Here the second
order torque amplitude is compared.
pressurebased model
standard model
(second order)
Figure 5.1: harmonic analysis, 2.0L diesel 4cylinder (turbocharged)
LuK GmbH & Co. Technical University of Eindhoven
Luc Römers / XB department Traineeship report
 60 
Figure 5.2 shows the result for the 6cylinder (naturally aspirated) gasoline engine. Here the
third order torque amplitude is compared.
In both graphs the effect of the mass torque can clearly be recognised. At low engine speeds
the combustion torque is dominant. Since mass torque and combustion torque are phase shifted
and mass torque increases quadratically with engine speed, both components cancel each other
out at a certain speed. At higher engine speeds the mass torque is dominant, causing the
combined amplitude to increase again.
These results show a clear similarity between the pressurebased and standard engine models.
At partload the differences will probably be larger, since the standard model is largely an
estimate there. For the most critical (and therefore the most important) fullload operation, the
results of the standard model are acceptable.
pressurebased model
standard model
(third order)
Figure 5.2: harmonic analysis, 3.0L gasoline 6cylinder (naturally aspirated)
LuK GmbH & Co. Technical University of Eindhoven
Luc Römers / XB department Traineeship report
 61 
6. Idle drive train model in MatLab / Simulink
The goal of this second assignment is to build an idle drive train model in Simulink. This
model will be used to make special simulations. In case complicated engine management
components need to be simulated, this cannot be done immediately in the DyFaSim software.
In Simulink it is easier and faster to build and test new components. Furthermore Simulink is
the industry standard, making it possible to exchange models with the customer.
The main goal is to construct an idle model that delivers the same results as its DyFaSim
counterpart. This model will then be used to investigate the unwanted phenomenon of sub
harmonic vibrations at idle operation. Experience has shown that cylinderbalancing controllers
can contribute to subharmonic vibrations of the dualmass flywheel. It is much more practical
to build and test a cylinderbalancing controller in Simulink, than it is in DyFaSim.
6.1 Virtual test bench for Simulink representation of LuK arc springs
Before considering the complete drive train model, a new Simulink representation of the LuK
arc springs needs to be tested. A virtual test bench has been built in Simulink (figure 6.1). The
test bench contains two masses (primary and secondary) with an arc spring set in between.
phi _pri m_Rad
phi _sec_Rad
omega_pri m_Rad
omega_sec_Rad
Trq_Spri ngs_Nm
LoadTrq_Nm
Engi neTrq_Nm
omega_sec_Rad
Warning model contains no basic friction, no friction control plate,
no flange inertia & no inner damper
Created by Dr. Stephen Jones, Dr. Mathias Müller & Luc Römers,
XB Department, LuK GmbH
Version 2.2  Beta
MatLab Simulink Sfunction Representation of LuK DMF Arc Springs
LuK GmbH &Co owns all the intellectual propertyrights, including copyright and rights of confidence,in these
computer programs. Such intellectual propertyrights extend to the structure, sequence,organisation and
presentation of the programs and theymaynot be, in whole or part, copied, used or transmitted in anyform
or byanymeans, including electronically, without prior written permission of LuKGmbH &Co.
Si m_LoadTrq_Nm
Si m_Engi neTrq_Nm
Si m_omega_sec_Rad
Si m_omega_pri m_Rad
1
s
1
s
[ti me, TrqLoad]
[ti me, TrqEng]
TrqLef t
TrqRight
omega
Equati on of Moti on
of Secondary Mass
TrqLef t
TrqRight
omega
Equati on of Moti on
of Pri mary Mass
DMF Arc Spri ng Model s
omega_prim_Rad
Figure 6.1: virtual test bench containing Simulink arc spring representation
LuK GmbH & Co. Technical University of Eindhoven
Luc Römers / XB department Traineeship report
 62 
The subsystem containing the arc springs is shown in figure 6.2. This particular dualmass
flywheel contains two inner arc springs and two outer arc springs.
TrqO1_Nm
TrqO2_Nm
TrqI1_Nm
TrqI2_Nm
phi _pri m_Rad
phi _sec_Rad
omega_pri m_Rad
omega_sec_Rad
1
Trq_Spri ngs_Nm
Si m_phi _rel ati ve_Rad
Si m_phi _sec_Rad
Si m_TrqO1_Nm
Si m_phi _pri m_Rad
Si m_Trq_Spri ngs_Nm
Si m_TrqI2_Nm
Si m_TrqI1_Nm
Si m_TrqO2_Nm
SFuncti on
Outer Arc Spri ng 2
SFuncti on
Outer Arc Spri ng 1
SFuncti on
Inner Arc Spri ng 2
SFuncti on
Inner Arc Spri ng 1
4
omega2
3
omega1
2
phi 2
1
phi 1
An increasing positive torque is applied on the primary mass (see figure 6.1), representing the
engine torque. Simultaneously an increasing negative torque is applied on the secondary mass,
representing load torque. This way the dualmass flywheel is twisted, thereby compressing the
arc springs. Having reached a maximum value, the torque on both sides is decreased again.
After simulation the overall arc spring reaction torque can be plotted against the relative angle.
This results in the combined arc spring characteristic. One static simulation and multiple
dynamic simulations are performed. This is done because the characteristic is speed dependent.
The same test bench has been built in DyFaSim, making it possible to compare results. Figure
6.3 and 6.4 show the Simulink and DyFaSim results. It can be seen that the results of both
models are 99% the same. Having validated this new Simulink arc spring model, it can be
implemented into the model of the complete drive train.
Figure 6.2: Sfunctions describing arc springs
LuK GmbH & Co. Technical University of Eindhoven
Luc Römers / XB department Traineeship report
 63 
Figure 6.3: overall arc spring characteristic, Simulink (left) & DyFaSim (right)
Figure 6.4: overall arc spring characteristic, DyFaSim & Simulink
LuK GmbH & Co. Technical University of Eindhoven
Luc Römers / XB department Traineeship report
 64 
6.2 Drive train model for idle simulation
The drive train model described in this paragraph consists of a combustion engine, idle drive
train and an engine management module. Figure 6.5 shows the top layer of the model. The
following Simulink modules were already developed by LuK:
• combustion engine module (idle only)
• arc spring module
Additional elements have been built to complete the idle drive train:
• flange in dualmass flywheel
• inner damper in dualmass flywheel
• basic friction in dualmass flywheel
• idle speed controller (including speed filter)
• oscillating inertia of the combustion engine
Furthermore the engine module is improved by correcting some small errors.
kinematik_primär_RAD
Mkorrektur_regler_Nm
Mzy linder_Nm (ohne Reibung)
Mmotor_Nm
Mbestandteile_Nm (ohne Reibung)
Verbrennungsmotor
kinematik_primär_RAD Mkorrektur_regler_Nm
Motormanagement
Model l steuerung
Mmotor_Nm
kinematik_primär_RAD
Mbogenf eder_grundreibung_Nm
kinematik_sekundär_RAD
Ei nfache Leerl auf Antri ebsstrang i ncl . ZMS
kinematik_primär_RAD
Mmotor_Nm
Mkorrektur_regler_Nm
Figure 6.5: idle drive train model, top layer
LuK GmbH & Co. Technical University of Eindhoven
Luc Römers / XB department Traineeship report
 65 
Figure 6.6 shows the combustion engine model (orange subsystem in figure 6.5). Four
subsystems produce the different torque components. From top to bottom, the components are:
• mass torque
• compression / expansion torque
• combustion torque
• friction torque
For the combustion torque, the Isermann empirical model is used. All torque components are
summed, after which the total torque is sent to the drive train (primary mass).
The engine is controlled by the idle speed controller (inside the engine management
subsystem). The oscillating primary speed is filtered, resulting in an average value. This
average value is fed into a PIcontroller, which controls the static value of the combustion
torque. For this particular model the P and Igains are shaped (a function of the errorsize).
The total torque consists of:
Mass torque
Compression/Expansion torque
Combustion (Isermann) torque
Friction torque
3
Mbestandtei le_Nm
(ohne Rei bung)
2
Mmotor_Nm
1
Mzyli nder_Nm
(ohne Rei bung)
(25 40/4)
Verbrennungsparameter
(aus workspace)
kinematik_primär_RAD
winkel_max_degrees
Mstat_Nm
Mkorrektur_regler_Nm
M_Isermann_Nm
M_Isermann_indiv _Nm
Verbrennungsmoment
Mreibung_Nm
Motor Rei bung
mosz
Lamda
Kurbelradius
Kurbelwellen kinematik
Mmasse_Nm
M_masse_indiv _Nm
Massenmoment
kinematik_primär_RAD
Kurbelradius
Lamda
Verdichtungsv er.
Durchmesser
Mk/e_Nm
Mk/e_indiv _Nm
K/EMoment
[Motparam]
From
M_masse_indiv _Nm
Mk/e_indiv _Nm
M_Isermann_indiv _Nm
Mzy linder_Nm
Berechnung der
Zyli ndermomente
CalculateIndi vCyli
Acti vation
2
Mkorrektur_regl er_Nm
1
kinemati k_primär_RAD
<kinematik_primär_RAD>
Figure 6.6: combustion engine module
LuK GmbH & Co. Technical University of Eindhoven
Luc Römers / XB department Traineeship report
 66 
Figure 6.7 gives an overview of the idle drive train model (yellow subsystem in figure 6.5). For
idle operation the drive train can be modeled using three masses:
• crankshaft & primary mass DMF
• flange
• secondary mass DMF
The equation of motion of these three masses is represented by the rectangular subsystems in
the figure below. Between the primary mass and the flange, the arc springs are situated
(subsystem with DMF picture). The inner damper is mounted between the flange and the
secondary mass. The final component is the basic friction working between the primary and
secondary mass.
beschl _sekundär_RAD
geschw_fl ansch_RAD
7
beschl _sekundär_RAD
6
geschw_sekundär_RAD
5
wi nkel _sekundär_RAD
4
Mbogenfeder_Nm
3
beschl _pri mär_RAD
2
gesch_pri mär_RAD
1
wi nkel _pri mär_RAD
grundrei bung
Si mTrqEng
Si mTrqLoad
n_fl ansch
1
s
1
s
1
s
1
s
delta_omega ID_Rueckwirk moment
Innendämpfer
DMF Arc Spri ng Model s
ZMS_Rueckwirk moment
ID_Rueckwirk moment
Flanschmassen geschwindigkeit
Flanschmassen beschleunigung
Bewegungsgl ei chung des
Fl ansches
ID_Rueckwirk moment
Mgrundreibung
Mgetriebe_schlepp_Nm
geschw_sekundär_RAD
beschl_sekundär_RAD
Bewegungsglei chung der Sekundärschwungschei be
Mmotor_Nm
Mgrundreibung
Mzms_Nm
geschw_primär_RAD
beschl_primär_RAD
Bewegungsgl ei chung der
KW+ Pri märschwungschei be
2
Mgetri ebe_schl epp_Nm
1
Mmotor_Nm
geschw_sekundär_RAD
geschw_primär_RAD
geschw_primär_RAD
beschl_primär_RAD
The inertia of the first mass is a function of the crank angle. It consists of a static inertia
(belonging to primary DMF mass and crankshaft) and a dynamic inertia (belonging to the
pistons).
Figure 6.7: drive train module
LuK GmbH & Co. Technical University of Eindhoven
Luc Römers / XB department Traineeship report
 67 
6.3 Comparison with DyFaSim counterpart
The complete idle drive train model is compared to its DyFaSim counterpart. Both models are
given the same initial conditions and are simulated for five seconds. A RungeKutta (ode4)
fixedstep solver is used.
The speedsignals of the primary and secondary mass are compared. The results are displayed
in figure 6.8 until 6.11.
These results show that both models are 99% identical. The remaining small differences are
probably caused by the solver methods. Even though RungeKutta is used in both programs,
the solver techniques are not completely the same.
Figure 6.8: primary speed (0 < t < 5) Figure 6.9: primary speed (0 < t < 0.5)
Figure 6.10: secondary speed (0 < t < 5) Figure 6.11: secondary speed (0 < t < 0.5)
LuK GmbH & Co. Technical University of Eindhoven
Luc Römers / XB department Traineeship report
 68 
7. Conclusions
Automatic generation of combustion engine models using MatLab
A tool has been created that enables all engineers at the dynamic simulation department of LuK
to generate engine models for the DyFaSim simulation software. It forms an automatic
interface between manufacturer engine data and the DyFaSim software.
Accurate engine models can be generated based on cylinder pressure data. In case no cylinder
pressure data is available, an engine model based on standard engine data can be generated.
Both methods have been compared by means of a harmonic analysis. Based on this analysis it
can be concluded that the standard model generation also delivers satisfying results.
The MatLabbased tool is available as a standalone package, without the need for a MatLab
license. The graphical user interface provides easy operation. The current version makes it
possible to choose between two languages. Further expansion to three or more languages is a
possibility.
Furthermore a guideline has been written, describing the use of the DyFaSim combustion
engine module and the new MatLabtool. A training course will be organized to familiarize all
engineers with this new tool.
Idle drive train model in MatLab / Simulink
An idle drive train model has been built in MatLab / Simulink. The model contains a
combustion engine module, drive train module (including dualmass flywheel) and an engine
management module.
The main goal was to create a model that is similar to its DyFaSim counterpart. Comparison of
output signals has shown that this goal has been accomplished. The Simulink model will be
used for special simulations, involving complex engine management components. A starter
motor could be added to the model, making it possible to do start / stop simulations.
LuK GmbH & Co. Technical University of Eindhoven
Luc Römers / XB department Traineeship report
 69 
List of symbols
Symbol Unit Definition
BDC [] Bottom dead centre
cf [] Correcting factor
d [m] Bore diameter
diff
relative,sim,full
[%] Relative difference, measurement & full map simulation
diff
relative,sim,reduced
[%] Relative difference, measurement & reduced map simulation
F
applied
[N] Force applied on piston
F
c,r
[N] Radial force on crank
F
c,t
[N] Tangential force on crank
F
piston
[N] Force acting on piston
l [m] Connecting rod length
m
osc
[kg] Oscillating mass
n
cyl
[] Number of cylinders
P [Pa] Cylinder pressure
P
comp/exp
[Pa] Compression / expansion pressure
P
exhaust
[Pa] Exhaust pressure
P
intake
[Pa] Intake pressure
P
surrounding
[Pa] Surrounding pressure
r [m] Crank radius
T
combustion
[Nm] Torque due to combustion
T
comp/exp
[Nm] Torque due to compression / expansion
T
crank
[Nm] Crankshaft torque
T
drag,scaled
[Nm] Scaled drag torque curve
T
drag,predefined
[Nm] Predefined drag torque curve
T
engine
[Nm] Engine torque
T
gas,sim
[Nm] Simulated gas torque
T
mass
[Nm] Torque due to mass forces
T
max
[Nm] Maximum torque
T
measured
[Nm] Torque from measurement
T
net,scaled
[Nm] Scaled net torque curve
T
net,predefined
[Nm] Predefined net torque curve
T
pp, measured
[Nm] Peaktopeak torque, measurement
T
pp,sim,full
[Nm] Peaktopeak torque, full map simulation
T
pp,sim,reduced
[Nm] Peaktopeak torque, reduced map simulation
T
simulation
[Nm] Simulated torque
T
static
[Nm] Static engine torque
T
static,cylinder
[Nm] Static cylinder torque
T
static,engine
[Nm] Static engine torque
TDC [] Top dead centre
V [m
3
] Cylinder volume
V
c
[m
3
] Clearance volume
V
d
[m
3
] Displacement volume
x [m] Piston position from top dead centre
α
[] Shape factor
β
[] Shape factor
LuK GmbH & Co. Technical University of Eindhoven
Luc Römers / XB department Traineeship report
 70 
ε
[] Compression ratio
ϕ
[rad] Crank angle
ϕ
max
[rad] Angle of maximum combustion torque
κ
[] Isentropic coefficient
λ
[] Crank/rod ratio
ω
[rad/s] Engine speed
LuK GmbH & Co. Technical University of Eindhoven
Luc Römers / XB department Traineeship report
 71 
References
[ENCA98] Microsoft Encarta 98, Multimedia Encyclopaedia
[ISER99] Hardwareintheloop simulation for the design and testing of enginecontrol
systems, R. Isermann, J. Schaffnit, S. Sinsel, Technical University of Darmstadt
(1999)
[MOOG00] Modellierung eines Verbrennungsmotors, O. Moog, Diplomarbeitsbericht LuK
(2000)
[PENN02] Dynamische Modellierung eines Verbrennungsmotors, B. Pennec,
Praktikumbericht LuK (2002)
[PISC89] Thermodynamik der Verbrennungsmaschine, R. Pischinger, G. Kraßnig,
G. Taucar; Th. Sams, SpringerVerlag, Wien (1989)
[PULK97] Engineering Fundamentals of the Internal Combustion Engine, W. Pulkrabek,
Pearson PrenticeHall, New Jersey (1997)
[SEGL00] Berechnung der Massenkräfte am Verbrennungsmotor, R. Segler, experiment
report LuK (2000)
[SOME00] Dictaat Verbrandingsmotoren 2, B. Somers, R. Baert, Technical University of
Eindhoven (2000)
[URLA94] Verbrennungsmotoren, A. Urlaub, SpringerVerlag, Berlin (1994)
LuK GmbH & Co. Luc Römers / XB department
Technical University of Eindhoven Traineeship report
Preface
I would like to thank all employees of the dynamic simulation department at LuK Bühl, for giving me the chance to do an interesting and varied traineeship. Besides gaining a lot of technical knowledge, I have experienced what it is like to be part of a professional development team. Special thanks go out to Bertrand Pennec and Dr. Stephen Jones for their great support during the last months. I would also like to thank Dr. Ad Kooy, as my contact at LuK, for giving me the chance to gain these experiences.
Luc Römers
1
LuK GmbH & Co. Luc Römers / XB department
Technical University of Eindhoven Traineeship report
Table of contents
1. Introduction 1.1 History of the company LuK GmbH & Co. 1.2 Goal of the traineeship 4 4 6
Automatic generation of combustion engine models using MatLab 2. Generation of pressurebased engine models 2.1 Geometry and kinematics of the crank mechanism 2.2 Dynamics of the crank mechanism 2.3 Mass torque 2.4 Compression / expansion torque 2.5 Combustion torque 2.5.1 Isermann method 2.5.2 Isermann/Pennec method 2.6 Total torque resulting from gas forces 2.7 Parameter maps 2.8 Reducing the parameter maps 2.9 Peaktopeak comparison 2.10 Output for DyFaSim software Generation of standard engine models 3.1 Combustion engine data sheet 3.2 Replacing missing data by estimates 3.2.1 Geometry data 3.2.2 Net torque curve at full load 3.2.3 Drag torque curve 3.2.4 Intake pressure curve at full load Graphical User Interface in MatLab 4.1 Graphical User Interface, pressurebased engine model 4.2 Graphical User Interface, standard engine model Comparison between pressurebased and standard models
7 8 10 11 12 13 17 18 20 22 23 31 33 40 42 42 43 43 43 44 45 46 46 54 59
3.
4.
5.
Idle drive train model in Matlab / Simulink 6. Idle drive train model in Matlab / Simulink 6.1 Virtual test bench for Simulink representation of LuK arc springs 6.2 Drive train model for idle simulation 6.3 Comparison with DyFaSim counterpart
61 61 61 64 67
2
LuK GmbH & Co. Conclusions 68 69 71 List of symbols References 3 . Luc Römers / XB department Technical University of Eindhoven Traineeship report 7.
In Bühl. the headquarters of the international company LuK is situated. in Hereford/England (today Hereford Ltd. At the company LuK. Mexico. (today LuK SouthAfrica) Takeover Laycock in England (today LuK UK) 1989 1990 1991 Opening LuK (UK) Ldt. France. The LuK dualmass flywheel (DMF) increases comfort and decreases fuel consumption in millions of cars worldwide. One in every four cars produced worldwide contains a LuK clutch. Takeover of TCM Ldt. India. GreatBritain. Introduction 1. Hungarian. 1965 1974 1976 Foundation of LuK in Bühl Germany Foundation of LuK Brazil and LuK Mexico Foundation of AS (AutoteileService GmbH) Development of torsion damped clutch plates 1977 1982 1985 1986 1987 Foundation of LuK Inc.1 History of the company LuK GmbH & Co. at the edge of the Black Forest. 9000 employees working at 20 production locations in Germany. Brazil. History of LuK GmbH & Co. Luc Römers / XB department Technical University of Eindhoven Traineeship report 1.) Jointventure between LuK and Barmag (future BarLuK) Takeover of automotivehydraulic division Vickers (today LuK FahrzeugHydraulik) First electronic clutch management for BMW Alpina 4 . USA New LuK location in Unna LuK dualmass flywheel (DMF): new standard for noise isolation in the drive train Takeover of AFT in Werdohl Takeover of Repco S. Korea and the United States yearly produce more than 14 million clutches for passenger cars and tractors.A. SouthAfrica.LuK GmbH & Co. over 3 million lockup clutches and over 5 million dualmass flywheels.
Russia und Czechia 1999 Takeover of Holzhäuser. Brazil. UK New foundry LuK Brazil LuK presents CVT parts for high torques. late 1997) SACclutch (Self Adjusting Clutch) 1995 1996 1997 The LuKEKM is produced in series for the Mercedes Aclass The production location in Wooster/Ohio starts to produce the first torqueconverter for light trucks in series 1998 Takeover of Kongsberg TechMatic in England and Norway Foundation of AS divisions in Spane. Germany Foundation AS Argentina Opening LuK Leamington. USA.) in India Foundation LuK Savaria in Hungarian (start of production. Luc Römers / XB department Technical University of Eindhoven Traineeship report 1992 1993 1994 Foundation of LuK GetriebeSysteme in Bühl Complete takeover BarLuK (today LuK Automobiltechnik) In corporation with a car manufacturer LuK develops parts for a CVT transmission with a maximum torque of about 300 Nm Foundation LuK Korea Foundation Beijing (LuK) Liaison Office in China Jointventure between Rane und LuK (Rane LuK Clutch Ltd. Mexico. Easytronic®. SouthAfrica. Audi Multitronic®: the first efficient CVT transmission for more than 300 Nm is produced in series Valeo sells LuK shares to INA Holding 2000 Opening of LuK production location in Bußmatten. Poland. is produced in series for the Opel Corsa 5 .LuK GmbH & Co. UK. Germany Automated manual transmission. Germany LuK takes over the production of clutch release systems form INA Opening of LuK tool shop in Kappelrodeck.
for use in the LuKDyFaSim simulation software. This model will be used to investigate the unwanted phenomenon of subharmonic vibrations at idle operation. This assignment consists of the buildup of an idle drive train model in MatLab / Simulink. Therefore about one of every six employees works in this field. For this. 6 . there was the possibility to do a second assignment.) 2002 2003 2004 2005 The 20 millionth dualmass flywheel is produced in Bühl The LuK integrated crankshaft damper is produced in series for the VW Touareg In Bühl the production of torqueconverters is started Takover of APTEC (future LuK Friction) Takeover of AP France (future LuK France) Opening LuK Korea Research and development are of great importance at LuK. the mathematical tool MatLab will be used.LuK GmbH & Co.2 Goal of the traineeship The main assignment is creating an automated and simple way of generating combustion engine models. 1. Luc Römers / XB department Technical University of Eindhoven Traineeship report 2001 LuK sells LuK Norway LuK fully takes over Rane LuK (future LuK India Private Ltd. The inventiveness of the product developer creates innovations for the automotive world of tomorrow. Because the schedule allowed it.
carried out by only a few engineers due to its complex nature. in case very little data is available These two situations will be referred to as pressurebased model generation and standard model generation. For this cause.1: overview of the MatLab program 7 .LuK GmbH & Co. In chapter 4 an explanation of the graphical user interface in MatLab will be presented.creating an accurate combustion engine model based on cylinder pressure data . allowing single elements like the combustion engine. Luc Römers / XB department Technical University of Eindhoven Traineeship report Automatic generation of combustion engine models using MatLab The goal of this assignment is to create a method. dualmass flywheel. clutch and transmission to be joined to form a complete drive train. which allows automatic generation of combustion engine models for the LuK DyFaSim simulation software. A full description of the pressurebased model will be given in chapter 2. The program is based on a modular structure.dvm'' Figure 1. the mathematical software package MatLab will be used to create an automatic modelling process. START Possibly engine geometry Engine geometry Possibly static torque curve Possibly intake pressure Cylinder pressure data Simplified maps based on values of experience Complete & optimized parameter maps Automatic generation of a script for DyFaSim engine module ''Motor_standard. the standard model will be treated in chapter 3. based on the semimanual process used until now. a large amount of data from the car manufacturer needs to be processed. Automating the complete modelling process would reduce the manual effort.creating a standard combustion engine model. To create a DyFaSim engine model. The automatic engine model generation will handle two different situations: . thereby increasing the capacity of the involved engineers. The combustion engine module of DyFaSim uses an empirical method to simulate the dynamic engine torque in all operating points.dvm'' ''Motor_pressurebased. This has been a semimanual process until now. DyFaSim is a software package developed by LuK and is used to simulate the dynamics of the vehicle drive train.
The basic operating method of a fourstroke cycle is shown in the next figure. This torque varies strongly with crank angle. which will finally be cascaded into a whole engine.1: fourstroke cycle [ENCA98] Compression and combustion cause a pressure rise inside the cylinder. these are delivered by the car manufacturer. therefore it is called the dynamic cylinder torque. First. 60 45 Pressure [bar] 30 15 0 0 100 200 300 400 500 600 V olum e [cm ³] Figure 2. Intake Compression Combustion & Expansion Exhaust 0°<ϕ<180° 180°<ϕ<360° 360°<ϕ<540° 540°<ϕ<720° Figure 2.2: cylinder pressure over full cycle The pressure rise results in a force on the piston. which in turn results in a crankshaft torque. The model is based on cylinder geometry data and cylinder pressure data.LuK GmbH & Co. 8 . the torque resulting from one single cylinder will be modelled. Luc Römers / XB department Technical University of Eindhoven Traineeship report 2. Generation of pressurebased engine models The purpose of this model is to simulate the dynamic torque output in all operating points. An example is given in the next diagram.
However. cylinder pressure data and an empirical part. Figure 2. friction is taken into account outside the engine module. a description of the geometry. 9 .LuK GmbH & Co.3: components of the dynamic torque [PENN02] These three components will be described using cylinder geometry data. they can be summed to form the total dynamic torque resulting from one cylinder. kinematics and dynamics of the crank mechanism will be given. Once all components are simulated. Luc Römers / XB department Technical University of Eindhoven Traineeship report If we look at the dynamic torque resulting from one cylinder. The net torque can be found by subtracting the engine friction. it is possible to subdivide it into three sources: torque due to acceleration forces working on the oscillating piston mass torque due to pressure buildup resulting from compression/expansion torque due to pressure buildup resulting from the combustion process The last two components form the torque due to gas forces. in the LuK DyFaSim software. It is important to realise that this is the gross engine torque. Examples of all three components are given in the figure below. Before starting the modelling process of these torque components.
Luc Römers / XB department Technical University of Eindhoven Traineeship report 2. d l r λ = = = = bore diameter connecting rod length crank radius r/l TDC = top dead center BDC = bottom dead center x = piston displacement. also known as the engine speed.2) λ ⋅ cos(2ϕ ) λ3 ⋅ sin 2 (2ϕ ) x (ϕ ) = r ⋅ ω * cos ϕ + + 1 − λ2 sin 2 ϕ 4 ⋅ 1 − λ2 sin 2 ϕ 2 ( ) 3 (2. From this expression the piston speed and acceleration can be derived [MOOG00]: 1 x(ϕ ) = r * (1 − cos ϕ ) + * (1 − 1 − λ2 sin 2 ϕ ) λ (2.10  . .3) Here ω represents the rotational speed of the crankshaft.4: crank mechanism Vc + Vh Vc compression ratio Using these geometry parameters. clockwise relative to TDC displacement volume 4 Vc = clearance volume Vd = 2 ⋅ r ⋅ π ⋅d2 ε= Figure 2. relative to TDC φ = crank angle.1 Geometry and kinematics of the crank mechanism The following schematic drawing shows the crank mechanism. connecting rod and the crankshaft.LuK GmbH & Co.1) • λ ⋅ sin(2ϕ ) x(ϕ ) = r ⋅ ω * sin ϕ + 2 ⋅ 1 − λ2 sin 2 ϕ •• (2. It consists of the piston. the displacement of the piston (relative to TDC) can be described.
r and a tangential part Fc.5) .2 Dynamics of the crank mechanism Knowing the kinematics. Of course only the tangential part will result in a crankshaft torque. The next figure shows the forces acting on the various components. as a result of an applied force on the piston. The resulting force on the crankshaft can be divided into a radial part Fc.t = Fapplied * sin ϕ + 2 ⋅ 1 − λ2 sin 2 ϕ (2.5: dynamics of the crank mechanism The relation between the applied force at the piston and the tangential force at the crankshaft can be described as follows [MOOG00]: λ ⋅ sin(2ϕ ) Fc.LuK GmbH & Co. Luc Römers / XB department Technical University of Eindhoven Traineeship report 2. a dynamical analysis of the piston/crank mechanism can be made. the crankshaft torque results: λ ⋅ sin(2ϕ ) Tcrank = r * Fapplied * sin ϕ + 2 ⋅ 1 − λ2 sin 2 ϕ (2.t.4) If this expression is multiplied with the crank radius.11  . Figure 2.
8) ( ) 3 The mass torque depends quadratically on the engine speed. The next figure displays the mass torque over one complete cycle.LuK GmbH & Co.6: mass torque over one full cycle The derived formula reveals that the mass torque depends only on cylinder geometry parameters and engine speed. .12  . By using Newton’s second law. Luc Römers / XB department Technical University of Eindhoven Traineeship report 2.3 Mass torque From the kinematical analysis. the acceleration of the piston as a function of the crank angle is known. the forces involved can be determined: •• F = − mosc ⋅ x (ϕ ) (2. This force can be converted to a torque working on the crankshaft [SEGL00]: λ ⋅ sin(2ϕ ) Tmass = r * F * sin ϕ + 2 ⋅ 1 − λ2 sin 2 ϕ (2. It can be seen that the amplitude increases by a factor of four when the engine speed is doubled. Mass torque [Nm] 0 2000 RPM 4000 RPM 0 180 360 540 720 Crank angle [degrees] Figure 2.6) Where mosc stands for the oscillating mass of one cylinder.7) λ ⋅ sin(2ϕ ) Tmass = −mosc ⋅ ω 2 ⋅ r 2 * sin ϕ + 2 ⋅ 1 − λ2 sin 2 ϕ λ3 ⋅ sin 2 (2ϕ ) * cos ϕ + λ ⋅ cos(2ϕ ) + 1 − λ2 sin 2 ϕ 4 ⋅ 1 − λ2 sin 2 ϕ (2.
e. At this point the pressure measurements are used. Luc Römers / XB department Technical University of Eindhoven Traineeship report 2.10) V + Vd Pcomp/exp (ϕ) = Pintake ⋅ c V(ϕ) κ (2.11) Where the angle dependent cylinder volume can be written as [PISC89]: V (ϕ ) = Vc + π *d 2 4 * x(ϕ ) (2. Even in case no combustion process is taking place. one can state: P ⋅ V κ = constant Pcomp/exp (ϕ ) ⋅ (V( ϕ )) = Pintake ⋅ (Vc + Vd ) κ κ (2. . Cylinder pressure measurements. in various engine operating points. To simulate this torque component. the compression/expansion torque will still be present.13) This expression for the volume contains only cylinder geometry parameters. This force generates a torque due to compression/expansion at the crankshaft. The next graph shows the total cylinder pressure and the component resulting from compression/expansion. The isentropic equation can now be used to calculate the pressure due to compression/expansion over the next 360°. The only remaining parameter needed to describe the compression/expansion pressure is the intake pressure. an analysis of the cylinder pressure is needed. For this pressure analysis. the following assumptions are made: the gas inside the cylinder is ideal the compression/expansion is adiabatic reversible. or isentropic the isentropic coefficient κ is constant. i. The following graph shows an example of a cylinder pressure measurement.4 Compression / expansion torque Compression of the gas inside the cylinder results in a pressure rise and consequently in a force on the piston.9) (2. which are known. The intake pressure is found by taking an average between 175° and 185° crank angle.13  . does not change with temperature Under these conditions. where the piston reaches bottom dead center.LuK GmbH & Co. are delivered by the car manufacturer.12) V (ϕ ) = π ⋅r ⋅d2 2 1 * + (1 − cos ϕ ) + * (1 − 1 − λ2 sin 2 ϕ ) ε −1 λ 4 (2.
Luc Römers / XB department Technical University of Eindhoven Traineeship report 100 90 80 70 Cylinder pressure [bar] 60 50 40 30 Intake pressure: average between 175°& 185° 20 10 0 0 100 200 300 400 500 600 700 Crank Angle [degrees] Crank angle [rpm] Figure 2.15) .LuK GmbH & Co.7: cylinder pressure measurement 100 90 measured cylinder pressure simulated comp.14  .8: simulation of compression/expansion pressure The compression/expansion pressure can be converted into a crankshaft torque: Fpiston(ϕ) = π⋅d2 (Pcomp/exp(ϕ) − Psurrounding ) 4 (2. pressure 80 70 Cylinder pressure [bar] 60 50 40 30 20 10 0 0 100 200 300 400 500 600 700 Crank angle [degrees] Crank Angle [rpm] Figure 2.14) π ⋅d2 (Pcomp/exp(ϕ) − Psurrounding )⋅ r ⋅ sinϕ + λ ⋅ sin(2ϕ) 2 Tcomp/exp(ϕ) = 4 2 ⋅ 1 − λ 2 sin ϕ (2./exp.
sim ) (2. since the simulated compression/expansion torque is based on ideal conditions. There are differences however.16) The simulated compression/expansion torque is multiplied with this factor. This is accounted for by introducing a correction factor. The resulting torque curves are displayed in the next graph.LuK GmbH & Co. torque 800 600 Torque [Nm] 400 200 0 200 400 0 100 200 300 400 500 600 700 Crank Angle [rpm] Crank angle [degrees] Figure 2. 1000 torque from measurements simulated comp. which is defined as follows: cf = Min(Tmeasurement ) Min(Tcomp/exp.9: compression / expansion torque Before top dead center both curves are basically the same./exp. Luc Römers / XB department Technical University of Eindhoven Traineeship report The same conversion is made for the measured total cylinder pressure. 1000 torque from measurements simulated comp.15  ./exp. resulting in an equal minimum.10: corrected compression / expansion torque . since pressure rise due to combustion is not present in this region. torque 800 600 Torque [Nm] 400 200 0 200 400 0 100 200 300 400 500 600 700 Crank Angle [rpm] Crank angle [degrees] Figure 2.
The simulated compression/expansion torque.16  .17) for ϕ > 540° Pexhaust is an average pressure taken from the measurement in the region between 625° and 635° crank angle. the simulated compression/expansion pressure will be defined on three different regions: P comp/exp(ϕ) = Pintake Vc + V d Pcomp/exp(ϕ) = Pintake ⋅ V(ϕ) Pcomp/exp(ϕ) = Pexhaust for ϕ < 180° κ for 180° ≤ ϕ ≤ 540° (2.LuK GmbH & Co. This occurs because the cylinder pressure does not completely drop to the intake pressure after the combustion process. especially at high engine speeds. without adapted exhaust pressure To take this into account. Luc Römers / XB department Technical University of Eindhoven Traineeship report In the region between 540° and 720° there are still differences however./exp. 800 torque from measurement simulated comp.11: comp. torque 600 400 Torque [Nm] 200 0 200 400 0 100 200 300 400 500 600 700 Crank Angle [rpm] Crank angle [degrees] Figure 2. torque. based on this pressure definition. is displayed in the next graph./exp. . There is a distinct difference between intake pressure and exhaust pressure. It can be seen that the differences in the exhaust region are much smaller now.
18) The purpose of this average value will become clear later on.average = 1 4π 4π ∫r⋅ π ⋅d 2 3π λ ⋅ sin( 2ϕ ) dϕ ( Pexhaust − Pintake ) ⋅ sin(ϕ ) + 2 2 4 1 − λ sin (ϕ ) (2. based on mathematics. The improved empirical model is called the Isermann/Pennec method. since it depends on a large number of parameters. Luc Römers / XB department Technical University of Eindhoven Traineeship report 800 torque from measurement simulated comp./exp./exp. cf . geometry. torque 600 400 Torque [Nm] 200 0 200 400 0 100 200 300 400 500 600 700 Crank Angle [rpm] Crank angle [degrees] Figure 2. The average over one complete cycle can be expressed as follows [PENN02]: Tcomp/exp.5 Combustion torque Describing a combustion process using thermodynamics is difficult.12: comp. intake pressure and exhaust pressure. The total gas torque is known from the cylinder pressure measurements. torque. It is now possible to simulate this torque using geometry data and three parameters: correction factor.LuK GmbH & Co. Pintake . when the combustion process is taken into account. The crankshaft torque resulting from the combustion process can found by calculating the difference between total gas torque and compression/expansion torque.19) 2. Therefore an empirical method. The original empirical model is known as the Isermann method. Hereby the analysis of the compression/expansion torque is completed. LuK has improved this method. Tcomp/exp (ϕ ) = Tcomp/exp (ϕ . is used. with adapted exhaust pressure This new compression/expansion torque has an average value different from zero. A short description of both methods will be given. Pexhaust ) (2.17  . furthermore the compression/expansion .
14: torque due to combustion 2.5.13: combustion torque is the difference between both curves 600 (T 500 measured T comp/exp.LuK GmbH & Co. Figure 2. Luc Römers / XB department Technical University of Eindhoven Traineeship report torque has been simulated.18  .sim ) 400 Torque [Nm] 300 200 100 0 100 0 100 200 300 400 500 600 700 Crank angle [degrees] Angle [rpm] Figure 2. Tcombustion (ϕ ) = 16 ⋅ π ⋅ Tstatic (ϕ max )3 ϕ ⋅e 2 − 2⋅ ϕ ϕ max for 360 ° < ϕ < 540 ° (2. After subtracting both curves. This is shown in the next figures.14.20) .1 Isermann method The Isermann method [ISER99] uses the following formula to describe the combustion torque shown in figure 2. the combustion torque remains.
which can be determined from the measured data.LuK GmbH & Co. Luc Römers / XB department Technical University of Eindhoven Traineeship report It contains two parameters. . 1000 800 Torque [Nm] 600 Tstat=20Nm Tstat=40Nm Tstat=60Nm 400 200 0 0 180 360 540 720 Crank angle[° ] Figure 2.16: influence of φmax on the combustion torque Experience has shown that the accuracy of the Isermann model is limited.14. The effect of both parameters on the combustion torque is shown in the next graphs [PENN02].15: influence of Tstat on the combustion torque 1200 1000 800 Torque [Nm] 600 phi_max = 15° phi_max = 25° phi_max = 35° 400 200 0 0 180 360 540 720 Crank angle [°] Figure 2. The static torque is the average value over one full cycle: T static = 1 4π 4π ∫ Tmeasured (ϕ )d ϕ = 0 1 4π 4π ∫P 0 dV dϕ dϕ (2.21) The angle of maximum combustion torque (ϕmax) corresponds to the maximum in figure 2. Tstatic and ϕmax.19  . Therefore the improved Isermann/Pennec model has been developed.
LuK GmbH & Co.α .20  . 500 400 Torque [Nm] 300 beta = 0. Tstatic .7 200 100 0 0 180 360 Crank angle [° ] 540 720 Figure 2.17: influence of β on the combustion torque 600 500 400 Torque [Nm] 300 alpha = 1 alpha = 3 alpha = 6 200 100 0 0 180 360 Crank angle [° ] 540 720 Figure 2.2 Isermann/Pennec method This method uses two additional shape parameters to simulate the combustion torque more accurately. β ) (2. The final Isermann/Pennec combustion torque is a function of the crank angle and four parameters: Tcombustion = Tcombustion (ϕ .3 beta = 0.5.18: influence of α on the combustion torque .5 beta = 0.22) The effect of the additional shape parameters α and β on the combustion torque is shown in the graphs below [PENN02]. Luc Römers / XB department Technical University of Eindhoven Traineeship report 2. The derivation of this empirical model is beyond the scope of this report. ϕ max .
2 0.simulated ) combustion. β ) = 1 4π 4π ∫ (T 0 measured (ϕ ) − Tsimulation (ϕ ) ) d ϕ 2 (2. There is no unique combination of α and β where an absolute minimum standard deviation is reached.1 0.LuK GmbH & Co. To compare different combinations.5 1 2 4 3 5 alpha [/] 6 7 8 9 10 100110 90100 8090 7080 6070 5060 4050 3040 2030 1020 010 Figure 2. 110 100 90 sigma [Nm] 80 70 60 50 40 30 20 10 0 0.23) Experience has shown results similar to figure 2. There is a line on which the results are comparably good.19: standard deviation for different combinations of α and β [PENN02] 500 (T T 400 measured T comp/exp. One shape parameter can have a fixed value.simulated 300 Torque [Nm] 200 100 0 100 0 100 200 300 400 500 600 700 Crank Angle [rpm] Crank angle [degrees] Figure 2.3 beta [/] 0.19.4 0. This simplifies the optimization process considerably. Luc Römers / XB department Technical University of Eindhoven Traineeship report Having found the values for Tstatic and ϕmax. the standard deviation between measurement and simulation is calculated: σ 2 (α .21  . the values for α and β can be optimized. after which the other shape parameter is optimized to find the minimum standard deviation.20: simulated combustion torque .
LuK GmbH & Co. Luc Römers / XB department
Technical University of Eindhoven Traineeship report
Figure 2.20 gives an example of a combustion torque simulated with the Isermann/Pennec method. Here α was fixed, after which β was optimized to find the best fit. The Isermann/Pennec empirical model makes it possible to accurately describe the torque resulting from combustion, using only three parameters: static torque, angle of maximum combustion torque and beta.
Tcombustion (ϕ ) = Tcombustion (ϕ , Tstatic , ϕ max , β )
(2.24)
2.6
Total torque resulting from gas forces
The total crankshaft torque resulting from gas forces is the sum of compression/expansion torque and combustion torque. This dynamic torque can now be accurately simulated using only geometry data and six parameter values:
Tcomp/exp = Tcomp/exp ( geometry, cf , Pintake , Pexhaust )
(2.25) (2.26) (2.27)
Tcombustion = Tcombustion (Tstatic , ϕ max , β )
Tgas ,sim = Tgas ,sim ( geometry, Tstatic , Pintake , Pexhaust , cf , ϕ max , β )
A result of this method is displayed in figure 2.21. A comparison between measurement and simulation is shown.
600
500
dynamic gas torque, measured dynamic gas torque, simulated
400
300
Torque [Nm]
200
100
0
100
200
0
100
200
300
400
500
600
700
Crank Angle [rpm]
Figure 2.21: dynamic torque form measurement and simulation
 22 
LuK GmbH & Co. Luc Römers / XB department
Technical University of Eindhoven Traineeship report
2.7
Parameter maps
The previous analysis is based on only one single operating point. The car manufacturer delivers pressure measurements in many operating points. For each pressure measurement, the torque resulting from gas forces is simulated using the Isermann/Pennec method. The engine speed at which a cylinder pressure measurement is conducted is known. A value for the static torque results from the pressure analysis. Note that this is the static torque of only one cylinder. Multiplying this value with the number of cylinders delivers the total static engine torque.
Tstatic,engine = Tstatic ,cylinder ⋅ N cyl
(2.29)
The position in the engine map, at which the pressure measurement is conducted, is now known. Five additional parameters belong to this particular operating point. Because pressure measurements at many different operating points are available, it is possible to create maps for each of these parameters. These maps describe parameter variation with engine speed and static engine torque:
Pintake = Pintake (ω,Tstatic )
Pexhaust = Pexhaust (ω,Tstatic ) cf = cf (ω,Tstatic )
(2.30)
ϕ max = ϕ max (ω,Tstatic )
β = β (ω,Tstatic )
With these five parameter maps (and geometry data), it is possible to simulate the dynamic engine torque in every operating point of the engine. An example will be used to show how the maps are constructed.
 23 
LuK GmbH & Co. Luc Römers / XB department
Technical University of Eindhoven Traineeship report
The next figure shows the engine map of a 2.0L gasoline engine. In this particular case the car manufacturer has delivered cylinder pressure data in over 220 operating points. For each of these measurements the dynamic torque is simulated, using the method described before. This results in the following engine map:
Engine map
250
200
T stat engine [Nm]
150
100
50
0
0
1000
2000
3000
4000
5000
6000
7000
8000
Speed [rpm]
Figure 2.22: engine map, operating points in which pressure data is available
It can be seen that pressure measurements have been conducted at idle, coast, partload and fullload operation. Five parameters, needed to reconstruct the dynamic engine torque, are found for each of these operating points. These parameters can be displayed as a function of engine speed and static engine torque. For this particular engine the following results are found.
Intake pressure
Exhaust pressure
3.5 3 2.5 2 1.5 1 0.5 0 200 150 100 50 2000 0 1000 0 4000 3000 6000 5000
4.5 4 3.5
Pexhaust [bar]
3 2.5 2 1.5 1 0.5 0 200 150 100 50 2000 0 1000 0 4000 3000 6000 5000
Pintake [bar]
7000
7000
Tstat engine [Nm]
Speed [rpm]
Tstat engine [Nm]
Speed [rpm]
Figure 2.23: intake pressure
Figure 2.24: exhaust pressure
 24 
or little. Luc Römers / XB department Combustion angle Technical University of Eindhoven Traineeship report Correction factor 50 45 1. This will be explained later on by using a second (worstcase) example.85 7000 200 150 100 50 2000 0 1000 0 4000 3000 6000 5000 7000 Tstat engine [Nm] Speed [rpm] Tstat engine [Nm] Speed [rpm] Figure 2. are as follows: no cylinder pressure measurements at the start/idle region (300 .26: correction factor 0. The operating points at which pressure data is delivered are different for each engine.3 0. First the example of figure 2. this is not the case.25  . Before these maps can be created. Some common problems. The example used above shows an ideal case.15 1.2 200 150 100 50 2000 0 1000 0 4000 3000 6000 5000 7000 Tstat engine [Nm] Speed [rpm] Figure 2. covering all possible engine operating points. In this region a number of points are added.27: shape factor beta These loose values need to be converted to maps. Very often however.6 Beta [] 0. a line of .05 Phimax [degrees] 40 35 30 25 20 200 150 100 50 2000 0 1000 0 4000 3000 6000 5000 kf [] 1 0.22 is treated. Here only data at the start/idle region is missing. caused by a limited amount of measurements. Secondly.5 0.9 0.750 rpm) missing cylinder pressure data at coast operation no. with over 220 operating points covering the engine map.7 0. cylinder pressure data at partload operation missing cylinder pressure data at high engine speeds Each of these cases has to be taken into account.LuK GmbH & Co.4 0. based on values of experience.95 0.1 1. possible “holes” in the dataset need to be taken into account.25: combustion angle Shape factor beta Figure 2.
The reason for this has to do with the DyFaSim software (the DyFaSim combustion engine module requires a square mesh for each parameter). Luc Römers / XB department Technical University of Eindhoven Traineeship report points at maximum engine torque is added. Higher order techniques (cubic. After these additions. after adding points and interpolation There are now enough operating points to create the full parameter maps. The five parameter maps are displayed in the following figures.LuK GmbH & Co. Interpolated points are added with a 50rpm spacing.29: intake pressure map Figure 2. Experience has shown that basic linear interpolation between the operating points delivers the best results.26  . a linear interpolation along every load line is carried out. Engine map 250 200 Tstat engine [Nm] 150 100 50 0 0 1000 2000 3000 4000 5000 6000 7000 Speed [rpm] Figure 2. A number of interpolation techniques have been considered. Figure 2. spline) cause problems when the number of operating points is limited.28: engine map. The parameter values stay constant at torques higher than the fullload torque curve. The results are displayed in the following engine map. This is done to create a square engine map.30: exhaust pressure map .
31: combustion engine map Figure 2.LuK GmbH & Co. Luc Römers / XB department Technical University of Eindhoven Traineeship report Figure 2.32: correction factor map Figure 2.33: beta map These maps are generated using an engine speed spacing of 50 rpm and a static torque spacing of 2.5 Nm. .27  .
For this engine very little pressure data is delivered.35: intake pressure Combustion angle Figure 2.5 3 2.5 2 1. It is used to illustrate some other techniques to “fill the map”.5 0 300 250 200 150 100 50 0 1000 0 3000 2000 4000 6000 5000 4. with over 220 operating points available.28  . Engine map 300 250 Tstat engine [Nm] 200 150 100 50 0 1000 2000 3000 4000 5000 6000 7000 Speed [rpm] Figure 2.37: combustion angle Figure 2.5 0 300 250 200 150 100 50 0 1000 0 3000 2000 4000 6000 5000 Pintake [bar] Tstat engine [Nm] Speed [rpm] Tstat engine [Nm] Speed [rpm] Figure 2.5 1 0.05 Phimax [degrees] 40 35 30 25 20 300 250 200 150 100 50 0 1000 0 3000 2000 4000 6000 5000 cf [] 1 0. as can be seen in the next figures.38: correction factor .1 1. Another 2.15 1.36: exhaust pressure Correction factor 50 45 1.5 2 1. The previous example was a bestcase scenario.5 P exhaust [bar] 3 2.0L gasoline engine is considered.95 0.LuK GmbH & Co.85 300 250 200 150 100 50 0 1000 0 3000 2000 4000 6000 5000 Tstat engine [Nm] Speed [rpm] Tstat engine [Nm] Speed [rpm] Figure 2.5 4 3.9 0.34: engine map Intake pressure Exhaust pressure 3. Luc Römers / XB department Technical University of Eindhoven Traineeship report Now a second example is treated. Now a worstcase scenario is considered.5 1 0. Only fullload pressure measurements have been conducted.
Luc Römers / XB department Technical University of Eindhoven Traineeship report Shape factor beta 0.3 0. Again operating points in the start/idle region are added.2 300 250 200 150 100 50 0 1000 0 3000 2000 4000 6000 5000 Tstat engine [Nm] Speed [rpm] Figure 2.6 Beta [] 0.29  . Engine map 300 250 Tstat engine [Nm] 200 150 100 50 0 0 1000 2000 3000 4000 5000 6000 7000 Speed [rpm] Figure 2. The results are shown in the next figures.40: engine map. after which the parameter maps can be created. as well as the line at maximum torque. Furthermore additional points at coast operation are added. Parameter values at these added points are again based on values of experience.4 0.5 0.39: shape factor beta In this case more actions are needed to fill the engine map.LuK GmbH & Co. after adding points and interpolation . Next a linear interpolation along the load lines is made.7 0.
41: intake pressure map Figure 2.45: beta map .LuK GmbH & Co.30  .43: combustion engine map Figure 2.44: correction factor map Figure 2.42: exhaust pressure map Figure 2. Luc Römers / XB department Technical University of Eindhoven Traineeship report Figure 2.
Since the DyFaSim combustion engine module is not designed for matrices of this size. In case a large spacing is chosen. Experience has shown that the following spacing delivers satisfying results: speed spacing = 250 rpm static torque spacing = 25 Nm Original points. This reduction process belongs to the first example of the previous paragraph. after which the matrices are reduced in size using this new spacing. An example is given in the following figures. only 2% of the original amount of data remains. 2. A new spacing for engine speed and static torque can be defined. accuracy of the engine model will be lost. To visually check the quality of this reduction process.8 Reducing the parameter maps The current parameter maps are very accurate. the amount of data will have to be reduced. By implementing this new spacing. both the full parameter maps and the reduced parameter maps are displayed. Luc Römers / XB department Technical University of Eindhoven Traineeship report These results show that.LuK GmbH & Co. In the MatLab program the axes of both figures are linked. based on these interpolated parameter values can be made.31  . Figure 2. This way a good estimate for the dynamic engine torque. This makes it possible to rotate the figures simultaneously. lying in between. although only limited data is available. parameter values are found for all engine operating points.46: reduction of the intake pressure map . are erased from the matrix. since they are created with a fine mesh: speed spacing = 50 rpm static torque spacing = 2. A suitable compromise between data reduction and loss of accuracy will have to be made.5 Nm This results in a very large matrix for each parameter.
49: reduction of the correction factor map .32  .48: reduction of the combustion angle map Figure 2. Luc Römers / XB department Technical University of Eindhoven Traineeship report Figure 2.47: reduction of the exhaust pressure map Figure 2.LuK GmbH & Co.
52). 2. The total engine torque can be found by cascading the onecylinder torque into a full engine. A comparison will be made between the measured dynamic torque (resulting from pressure measurements) and the simulated dynamic torque. . a new (smaller) spacing can be defined. For an ncylinder inline engine the total engine torque is given by: ncyl −1 Tengine = ∑T i =0 cyl (ϕ + i * 4π ) ncyl (2. resulting from an excessively large spacing. The peaktopeak torque is the difference between the minimum and maximum of the dynamic torque curve over two crankshaft revolutions. the dynamic torque is simulated using the full as well as the reduced parameter maps. Here the dynamic torque of the complete engine is considered (figure 2.31) The next figures show the cylinder cascading for a 4cylinder engine. Luc Römers / XB department Technical University of Eindhoven Traineeship report Figure 2. The dynamic torques from measurements and simulation are displayed in one graph.33  .9 Peaktopeak comparison Although the visual quality check can be very useful. Now a comparison between the peaktopeak torques is made.50: reduction of the beta map Experience has shown that a good visual quality check is possible this way. In case the reduced maps are unsatisfying. For each operating point in which measurement data is available.LuK GmbH & Co. Both simulation with the full maps and simulation with the reduced maps are displayed. a more quantitative method is developed.
reduced = max(Teng.measure d = max(Teng.reduced ) − min(Teng.reduced ) (2.52: dynamic torque of full engine The peaktopeak torque for these three curves is given by: T pp.sim .sim .sim . full ) − min(Teng.sim .51: dynamic torque of one cylinder Tpeaktopeak Figure 2.sim .LuK GmbH & Co. full = max(Teng. full ) T pp.measur ed ) − min(Teng.32) .34  .measur ed ) T pp.sim . Luc Römers / XB department Technical University of Eindhoven Traineeship report Figure 2.
35  . the peaktopeak torques can be found. From these results.54: dynamic torque of full engine.LuK GmbH & Co.54. Figure 2. all available operating points . Luc Römers / XB department Technical University of Eindhoven Traineeship report For the 220 operating points of the previous example. the 1cylinder results are displayed in figure 2. Figure 2.53: dynamic torque of one cylinder. all available operating points The full (4cylinder) engine torque is displayed in figure 2.53.
LuK GmbH & Co. This result is expected for all naturally aspirated engines.57: absolute peaktopeak values. . side view 2 Figure 2.56: absolute peaktopeak values. Figure 2.36  .57 displays the peaktopeak torque as a function of static engine torque.55: absolute peaktopeak values Figure 2. A clear linear relation can be seen. along with the twodimensional side views. Luc Römers / XB department Technical University of Eindhoven Traineeship report The peaktopeak values are determined and plotted as a function of engine speed and static engine torque. The graphs below show a threedimensional view of the results. side view 1 Figure 2.
sim . full = T pp . The green points represent the relative difference between measurement and simulation with the reduced maps.60: relative peaktopeak difference. measured ) ⋅ 100% diff relative.reduced − T pp . sim . Therefore the relative peaktopeak differences between measurement and both simulations are calculated: abs (T pp . side view 1 Figure 2.reduced = T pp .sim . Figure 2. The red points represent the relative difference between measurement and simulation with the accurate parameter maps.33) (2.measured (2.59: relative peaktopeak difference.37  .58: relative peaktopeak difference Figure 2.34) The results are displayed in the following graphs. it is difficult to evaluate the quality of the simulated curves. Luc Römers / XB department Technical University of Eindhoven Traineeship report Based on these absolute values.LuK GmbH & Co.sim . side view 2 . full − T pp . measured abs (T pp .measured ) ⋅ 100% diff relative.
38  .4. Furthermore it can be seen that the results at coast are very bad.62: absolute peaktopeak values.LuK GmbH & Co. The next graphs show the results in case mass torque is included. mass torque taken into account Figure 2.63: absolute peaktopeak values. this loss of accuracy is less obvious.61: absolute peaktopeak values. At high engine torques. side view 1 Figure 2. only compression/expansion torque. Figure 2. There is no combustion torque at coast operation. which has an effect on the relative error size. If the mass torque is also taken into account the absolute peaktopeak values change. Luc Römers / XB department Technical University of Eindhoven Traineeship report The simulation with the accurate maps is clearly better than the simulation with the reduced maps. The cylinder pressure at coast cannot be described accurately by the method explained in chapter 2. These relative differences are based on dynamic torque resulting from gas forces only. side view 2 .
56 clearly shows the effect of the mass torque.62 and figure 2. This occurs because combustion torque and mass torque are phase shifted and the mass torque increases quadratically with engine speed. The large error at coast operation has also disappeared. side view 1 Figure 2.66: relative peaktopeak difference.65: relative peaktopeak difference.62 the effect of the mass torque can be seen. . Comparing figure 2. the bad accuracy of the gas torque simulation is of minimum importance. side view 2 Now all relative differences are smaller than 10%.LuK GmbH & Co. Luc Römers / XB department Technical University of Eindhoven Traineeship report In figure 2. Since mass torque is very dominant at coast. Again the relative differences are calculated. At low engine speeds the combustion torque is dominant and at high engine speeds the mass torque is dominant.39  . the results are displayed below.64: relative peaktopeak difference. Figure 2. mass torque taken into account Figure 2.
10 Output for DyFaSim software If the peaktopeak analysis is satisfactory. In this case a generated ASCII file of a 6cylinder engine has been uploaded. the geometry data and parameter maps are defined (inside the red square). DyFaSim is able to simulate the dynamic engine torque in every operating point. With these values.67: DyFaSim combustion engine module For each cylinder. Luc Römers / XB department Technical University of Eindhoven Traineeship report 2. containing the reduced parameter maps along with the geometry data of the engine. Figure 2.40  .LuK GmbH & Co. This file can be uploaded into the DyFaSim combustion engine module. . after which the engine module is ready for use. The figure below shows the main window of the DyFaSim combustion engine module. Therefore an outputfile is created in ASCIIformat. the created parameter maps need to be stored in a format suitable for the DyFaSim software.
Each “edit” button opens a parameter map. In this case it is a map with a 250rpm speed spacing and a 50Nm torque spacing.69.68.68: geometry data Figure 2. Figure 2. On the right side a twodimensional graph of this matrix is displayed. In the bottom left corner the intake pressure matrix is shown.70. The five parameter maps are defined in figure 2.70 shows an intake pressure map in DyFaSim.70: intake pressure map .41  . Geometry data is collected in figure 2. An example is shown in figure 2. Figure 2. Luc Römers / XB department Technical University of Eindhoven Traineeship report The figures below show the windows containing the engine data.LuK GmbH & Co.69: parameter maps Figure 2.
cylinder pressure data is not always available.7 0.0 7000 Speed [rpm] Intake pressure Speed [rpm] 250 500 750 1000 1250 1500 1750 2000 2500 3000 3500 4000 4500 5000 5500 6000 6500 7000 Pressure [bar] 0.3 2.0 Net Trq Drag Trq 300 100.0 Gross Trq Power Torque [Nm] 200 75.MM.2.com Date DD.0 Power [kW] Torque curve at full load Speed [rpm] 250 500 750 1000 1250 1500 1750 2000 2500 3000 3500 4000 4500 5000 5500 6000 6500 7000 Net torque [Nm] 70 100 120 170 200 240 300 320 320 320 280 240 150 Drag torque [Nm] 30 30 30 30 32 34 36 38 42 46 54 60 66 Gross torque [Nm] 100 130 150 200 232 274 336 358 362 366 334 300 216 0 0 0 0 0 Power at full load Engine Power [kW] 1.5 0 0 1000 2000 3000 4000 5000 6000 7000 Intake Pressure Comments Speed [rpm] Figure 3.1.6 100.8 5. An example of the data sheet is given in figure 3.42  .6 2.0 0 25.0 83.6 3 Intake pressure [bar] 2.1.5 1 0.4 2.0 0.5 102. Luc Römers / XB department Technical University of Eindhoven Traineeship report 3.12. This Excel data sheet is filled out by the car manufacturer.5 2.0 0.0 0.2 37.2 9.1 Combustion engine data sheet LuK has developed a standardised combustion engine data sheet.LuK GmbH & Co.5 0.5 2 1.6 2.0 100 50.0 100 0 1000 2000 3000 4000 5000 6000 0.4 1.8 26.8 1 1.5 18 144 1029 0.04 1968 Unit [cm^3] [/] [mm] [mm] [/] [mm] [kg] [kgm^2] [cm^3] 400 125.7 55. The modelinput consists of: • • • • engine geometry data net torque curve at full load drag torque curve intake pressure curve at full load 3.5 70. For these situations a second operating mode has been developed: standard engine model generation.0 67.05 Email engineer@company.4 17.YY Max net torque [Nm] 320 at speed [rpm] 20003000 There is 0 essential data missing essential optional Engine geometry Nominal displacement Number of cylinders Bore Stroke Compression ratio Connecting rod length Oscillating mass Crankshaft inertia Calculated Displacement Value 2000 4 81 95.1: combustion engine data sheet . Basic Data Sheet for Combustion Engine Manufacturer Company name Fuel type Diesel Fuel injection system Unit injector or Unit pump system Engine name & calibration version 2L Diesel Air charging system Turbocharger (1X) Max engine speed [rpm] 5000 Engineer responsible for data Engineer (Company) Max power [kW] 103 at speed [rpm] 3500 Tel.2 2. Generation of standard engine models Although pressurebased engine models deliver very accurate results.0 0. In many cases only basic engine data is known. 07.6 2 2. This operating mode needs only basic engine data to create a full model. Luk GmbH & Co. number +00 00 000 000 77815 Bühl Version 2.2 1.8 100.
The car manufacturer often returns a data sheet in which data is missing.5 Unit [liter] [] [m] [m] [] [m] [kg] Diesel Nominal displacement Number of cylinders Bore Stroke Compression ratio Connecting rod length Oscillating mass Value User defined User defined 0.150 0. they need to be replaced by a good estimate. 3.080 0. naturally aspirated • gasoline. Secondly this scaled curve will be adapted to match the userdefined maximum. This process will be explained in the next paragraph. A predefined torque curve is first scaled up or down.LuK GmbH & Co. the user needs to define a maximum value for the net torque.2.100 18 0. These experience values are different for gasoline and diesel engines.1 Geometry data In case geometry data is unavailable.2 Replacing missing data by estimates As stated in the beginning of this chapter.0 Unit [liter] [] [m] [m] [] [m] [kg] 3. 3. In this case the missing data is replaced with experience values based on a similar engine. there are four essential inputs.2 Net torque curve at full load In case the net torque curve at full load is unavailable. How these estimates are determined will be explained in this paragraph.080 0.2.100 10. Net torque curves are predefined for a four different situations: • gasoline. depending on the displacement volume. as can be seen in the tables below. turbocharged • diesel. MatLab creates an interface with the Excel data sheet and imports the required data.5 0.43  . the user has to define: • engine type (gasoline or diesel) • displacement volume • number of cylinders All other geometry data is replaced by values of experience. it is possible to extract information in an automated way. naturally aspirated • diesel.150 1. In case any of these inputs are unavailable. Gasoline Nominal displacement Number of cylinders Bore Stroke Compression ratio Connecting rod length Oscillating mass Value User defined User defined 0. turbocharged . Luc Römers / XB department Technical University of Eindhoven Traineeship report Since the data sheet has a standard layout.
2) In case the scaled torque curve lies below the userdefined maximum it is scaled up once again: Tnet . scaled (ω) = Tnet . 3.3). scaled (ω)) (3. predefined curves form the basis.2. scaled (ω) ⋅ Tmax max(Tnet .0L engines of different configurations.LuK GmbH & Co. scaled (ω) = Tnet . In case the scaled torque curve exceeds the userdefined maximum. Luc Römers / XB department Technical University of Eindhoven Traineeship report The four predefined curves are displayed in figure 3. In case no drag torque data is available a similar process as for the net torque curve is used. Figure 3.3 Drag torque curve The car manufacturer delivers net torque data. However the combustion engine model is based on gross torque.2: standard full load torque curves These are experience curves for 2. . predefined (ω) ⋅ Vd 2 (3.1) This scaled curve is then compared to the userdefined maximum torque.2.3) This method has proven to deliver a good estimate for the net torque curve.0L diesel engine and one curve for a 2. For the drag torque there is one curve for a 2.44  . all values above this maximum are cut off: max (Tnet .scaled (ω)) ≡ Tmax (3. Again.0L gasoline engine (figure 3. Depending on the chosen displacement volume these curves are scaled up or down: Tnet . Therefore drag torque has to be taken into consideration.
Again the division is made between diesel and gasoline. Luc Römers / XB department Technical University of Eindhoven Traineeship report Figure 3. the intake pressure will be equal to the surrounding atmospheric pressure. approximately 1 bar. The curves for turbocharged gasoline and diesel engines are shown in figure 3.4.4 Intake pressure curve at full load The intake pressure at full load is also predefined for four different situations.4: standard intake pressure curves .3: standard drag torque curves These curves are also scaled using the displacement volume: V Tdrag .2. predefined (ω) ⋅ d 2 (3. For turbocharged engines. predefined curves are used. Figure 3. turbocharged and naturally aspirated. scaled (ω) = Tdrag .4) 3.45  . In case the engine is naturally aspirated.LuK GmbH & Co.
2: selecting an operating mode . A compiled version of the program is available.46  . The main goal of the program is easy operation. A DyFaSim engine model is needed for a new engine. Since cylinder pressure data is available. the user has the possibility to select a language: Figure 4. The manufacturer has delivered the geometry data and has conducted cylinder pressure measurements. Graphical User Interface in MatLab Both methods described in the previous chapters have been implemented in MatLab. the pressurebased model generation is chosen. For both the pressurebased and the standard engine model generation an example will be given.LuK GmbH & Co. Because MatLab licenses are expensive and scarce.1 Graphical User Interface. an example is given. pressurebased engine model To explain the graphical user interface for the pressurebased operating mode. Luc Römers / XB department Technical University of Eindhoven Traineeship report 4. 4. Figure 4. the user can choose between the two operating modes. without the need for a MatLab license.1: language selection After selecting a language. When the program is started. This version can be installed and used on every computer within LuK. a standalone version created with MatLab compiler has many advantages. This has been accomplished by using the Graphical User Interface tools available in MatLab.
47  . There is cylinder pressure available at idle. Luc Römers / XB department Technical University of Eindhoven Traineeship report MatLab now shows the database in which the geometry data of all engines is stored.3: manufacturer / engine selection The engine named “example. Figure 4. . a nameless exampleengine is chosen. To protect customer data. Of course it is also possible to select less data.txt” is selected. A new window opens in which all available cylinder pressure data for this engine is shown. Figure 4.4: pressure data selection In this case all pressure data is selected.LuK GmbH & Co. 6 partloads and fullload.
8: dialog box for coast values At this point the program is ready to start the calculation of the parameter values for all selected cylinder pressure measurements. Figure 4. coast data Figure 4. plot actions . The user has the possibility to adapt these values. in case this is necessary. Figure 4. If coast data is not available.5: info box showing selected engine Figure 4. Figure 4.48  .LuK GmbH & Co. a dialog box appears. which is the case for this example. The user has the possibility to view the results of the calculation process for every selected pressure measurement. one can also choose not to view these initial results. Luc Römers / XB department Technical University of Eindhoven Traineeship report An overview of the selected engine and pressure data is displayed in two info boxes. Because the plot actions slow the program down. In it are predefined experience values for the intake pressure and exhaust pressure at coast.9: question dialog.6: info box showing selected pressure data The program asks the user whether pressure data at coast is available.7: question dialog.
Figure 4. These values will be used to generate the reduced maps. . A dialog box appears in which the user can insert values for the speed and torque spacing. Pressure measurements in this operating point are often unreliable. the user can adapt the simulated dynamic idle torque. Figure 4. These curves are continuously updated during the calculation process. Therefore the simulated static torque at idle and a value known from experience are shown. the engine map shows the (black) operating points that have already been calculated. The purple point corresponds with the current dynamic torque shown on the left.49  . The user can decide which value should be used.10: dynamic torque plot. Luc Römers / XB department Technical University of Eindhoven Traineeship report If “Yes” is chosen. On the right side. along with position in engine map When the calculation process is completed. On the left side the dynamic torque from measurement and simulation is shown.11: possibility to adapt idle simulation The parameter maps can now be calculated. the following figure appears.LuK GmbH & Co.
The user now has the ability to visually evaluate the results. i. Figure 4.e.LuK GmbH & Co. The axes of both graphs are linked.50  . Luc Römers / XB department Technical University of Eindhoven Traineeship report Figure 4. The parameter graphs show the accurate and the reduced maps. one for the engine map and five for the parameter maps.13: graphs showing engine map and parameter maps . so they can be turned simultaneously.12: dialog box for spacing values The final results are displayed using six figures.
15: single cylinder dynamic torque Figure 4. Luc Römers / XB department Technical University of Eindhoven Traineeship report The program continues after pressing a random key on the keyboard.16: dynamic torque of full engine Figure 4.LuK GmbH & Co.51  .17: absolute peaktopeak values.14: possibility to redefine spacing values If the user is satisfied with the results. due to large spacing values.18: relative difference. only gas torque Figure 4. If the reduction process is unsatisfying. the peaktopeak analysis is made. new spacing values can be entered. Figure 4. The results are displayed in six figures. only gas torque . Figure 4.
Figure 4. a DyFaSim engine model is created. gas & mass torque Figure 4. The user can select a location to save the final model.LuK GmbH & Co. Luc Römers / XB department Technical University of Eindhoven Traineeship report Figure 4.21: select a location to save the DyFaSim engine model Finally there is a possibility to save a 1cylinder representation of the engine model. Figure 4.20: relative difference. gas & mass torque After viewing the results of the peaktopeak analysis.19: absolute peaktopeak values. These single cylinder models are used in case crankshaft dynamics are taken into account. In this case each piston is a separate mass.22: possibility to create a 1cylinder engine model .52  . on which a 1cylinder combustion engine is mounted.
53  . Luc Römers / XB department Technical University of Eindhoven Traineeship report After saving the DyFaSim model(s). The following window appears: Figure 4. the operation is completed. Figure 4. informing user that the operation is completed There is a possibility to make a printout of the final results.LuK GmbH & Co. all of the generated maps are displayed in one figure. For this purpose. possibility to make a printout .24: overview of results.23: info box.
Figure 4.54  . Since no cylinder pressure data is available. In case no geometry data is available.2 Graphical User Interface. When the program is started. Figure 4.26: selecting an operating mode Now the user can define whether geometry data is available for this engine.0L 4cylinder diesel engine. standard engine model To show the graphical user interface for the standard operating mode. The only known data is: geometry data torque curve at full load drag torque curve intake pressure at full load The data is collected in an Excel data sheet (this standard data sheet is filled out by the car manufacturer).25: language selection Now the user can choose between the two operating modes. the user has the possibility to select a language: Figure 4. For this engine no cylinder pressure data is available. the following example is used: an engine model is needed for a 2.LuK GmbH & Co. estimated values are used. geometry data .27: dialog box. the standard model generation is needed. Luc Römers / XB department Technical University of Eindhoven Traineeship report 4.
29: dialog box. air supply . Figure 4. In this example a turbocharger is present. Figure 4.28: selection of the Excel data sheet Figure 4.30: question dialog.LuK GmbH & Co. The user can now select the data sheet in which the geometry data is collected. MatLab opens the data sheet. reads the data and displays it in a dialog box.55  . The user now has the ability to check and/or change the displayed data. Luc Römers / XB department Technical University of Eindhoven Traineeship report For this example geometry data is available. engine data The next question is related to the air supply.
Figure 4. In case no data is available. This information is needed to estimate the parameter maps in the start/stop region. Figure 4. depending on the displacement volume. Figure 4. Luc Römers / XB department Technical University of Eindhoven Traineeship report The program asks the user whether full load data is available. throttle valve . In this case the intake pressure characteristic is filled out in the data sheet.31: question dialog. predefined values of a similar engine are used. In case no full load data is available. For this example drag torque data is available in the data sheet. This is the case. intake pressure The last question is related to the behavior of the throttle valve at engine shutdown (stop). drag torque When no intake pressure data (at full load) is available.LuK GmbH & Co.32: question dialog. a predefined full load curve is scaled up or down.56  .34: question dialog.33: question dialog. Figure 4. a predefined drag torque curve is scaled up or down. so the data can be extracted from the data sheet. full load characteristic The drag torque is treated in a similar way.
along with the drag torque curve. The DyFaSim script file is generated. Figure 4. The power output is also shown. Luc Römers / XB department Technical University of Eindhoven Traineeship report Before creating the script file for DyFaSim. On the left side the net and gross (fullload) torque curves are shown. scaled on the right Yaxis. the resulting engine map and intake pressure map are shown (figure 4.35: resulting engine map and intake pressure map After pressing a random key on the keyboard the program continues. after which the user can select a location to store it.36: select a location to save the DyFaSim engine . Figure 4.LuK GmbH & Co.35). Of course the drag torque is the difference between the first two.57  .
the operation is completed. Figure 4.58  . informing user that the operation is completed .37: possibility to create a 1cylinder engine model After saving the DyFaSim model(s). after which the program automatically shuts down.38: info box. Luc Römers / XB department Technical University of Eindhoven Traineeship report Again there is a possibility to create a 1cylinder representation of the engine.LuK GmbH & Co. The next window appears. Figure 4.
2. A simulation is performed. The resulting DyFaSim engine models are used to perform the harmonic analysis. Each engine model is loaded into DyFaSim and mounted on a large rotational mass. Upon this large constant torque.0L diesel 4cylinder (turbocharged) . Here the second order torque amplitude is compared.0 L. The applied torque accelerates the mass from 1000 rpm until maximum engine speed. The torque amplitude of the main excitation order is compared.1: harmonic analysis. The comparison is made for two very different engines: • 2. naturally aspirated For both engines a pressurebased model and a standard model has been generated. pressurebased model standard model (second order) Figure 5.1 shows the result for the 4cylinder (turbocharged) diesel engine. The harmonic analysis will be based on the fullload torque. a harmonic analysis will be made.LuK GmbH & Co. For 4cylinder engine this is the second order (two excitations per revolution).59  . For a 6cylinder engine this is the third order (three excitations per revolution). 6cylinder gasoline. Figure 5. Comparison between pressurebased and standard models To investigate the differences between pressurebased and standard engine models. the dynamic fullload engine torque is superimposed. Luc Römers / XB department Technical University of Eindhoven Traineeship report 5.0 L. since this is the most critical operating point. 4cylinder diesel. A harmonic analysis of the torque signal is made. during which the mass is accelerated with a large constant torque. turbocharged • 3.
Since mass torque and combustion torque are phase shifted and mass torque increases quadratically with engine speed. since the standard model is largely an estimate there. 3.0L gasoline 6cylinder (naturally aspirated) In both graphs the effect of the mass torque can clearly be recognised. pressurebased model standard model (third order) Figure 5.60  .LuK GmbH & Co. causing the combined amplitude to increase again. For the most critical (and therefore the most important) fullload operation. At higher engine speeds the mass torque is dominant. At partload the differences will probably be larger. Luc Römers / XB department Technical University of Eindhoven Traineeship report Figure 5. Here the third order torque amplitude is compared. . the results of the standard model are acceptable. At low engine speeds the combustion torque is dominant.2: harmonic analysis. These results show a clear similarity between the pressurebased and standard engine models. both components cancel each other out at a certain speed.2 shows the result for the 6cylinder (naturally aspirated) gasoline engine.
be. It is much more practical to build and test a cylinderbalancing controller in Simulink. Such intellectual property rights extend to the structure. this cannot be done immediately in the DyFaSim software. In Simulink it is easier and faster to build and test new components. T rqEng] EngineTrq_Nm TrqLef t omega TrqRight omega_prim_Rad Sim_omega_prim_Rad 1 s phi_prim_Rad Equation of Motion of Primary Mass phi_sec_Rad Trq_Springs_Nm omega_prim_Rad omega_sec_Rad DMF Arc Spring Models 1 s TrqLef t omega Sim_omega_sec_Rad omega_sec_Rad [time. no friction control plate. Luc Römers / XB department Technical University of Eindhoven Traineeship report 6. including electronically. This model will be used to make special simulations.1: virtual by any means. 6.2 .1 Virtual test bench for Simulink representation of LuK arc springs Before considering the complete drive train model. This model will then be used to investigate the unwanted phenomenon of subharmonic vibrations at idle operation. The test bench contains two masses (primary and secondary) with an arc spring set in between. XB Department. in whole orwrittencopied. The main goal is to construct an idle model that delivers the same results as its DyFaSim counterpart. including copyright and rights of confidence. used of LuK GmbH &inCo. In case complicated engine management components need to be simulated. Mathias Müller & Luc Römers. A virtual test bench has been built in Simulink (figure 6. Idle drive train model in MatLab / Simulink The goal of this second assignment is to build an idle drive train model in Simulink. a new Simulink representation of the LuK arc springs needs to be tested. Experience has shown that cylinderbalancing controllers can contribute to subharmonic vibrations of the dualmass flywheel. permission or transmitted spring Simulink arc any or without prior .Beta LuK GmbH & Co owns all the intellectual property rights.61  . form representation test bench containingpart. than it is in DyFaSim. sequence.in these computer programs. Stephen Jones. LuK GmbH Version 2. Furthermore Simulink is the industry standard. T rqLoad] LoadT rq_Nm TrqRight Equation of Motion of Secondary Mass Sim_LoadT rq_Nm MatLab Simulink Sfunction Representation of LuK DMF Arc Springs Warning model contains no basic friction.LuK GmbH & Co. Dr.organisation and presentation of the programs and they may not Figure 6.1). making it possible to exchange models with the customer. Sim_EngineT rq_Nm [time. no flange inertia & no inner damper Created by Dr.
LuK GmbH & Co.1).62  . This particular dualmass flywheel contains two inner arc springs and two outer arc springs. it can be implemented into the model of the complete drive train. Figure 6. thereby compressing the arc springs.4 show the Simulink and DyFaSim results. It can be seen that the results of both models are 99% the same. making it possible to compare results. . The same test bench has been built in DyFaSim. the torque on both sides is decreased again. Having reached a maximum value.2: Sfunctions describing arc springs An increasing positive torque is applied on the primary mass (see figure 6. This results in the combined arc spring characteristic. representing load torque.2. Having validated this new Simulink arc spring model.3 and 6. This is done because the characteristic is speed dependent. representing the engine torque. After simulation the overall arc spring reaction torque can be plotted against the relative angle. This way the dualmass flywheel is twisted. Luc Römers / XB department Technical University of Eindhoven Traineeship report The subsystem containing the arc springs is shown in figure 6. Sim_phi_prim_Rad Si m_phi_relative_Rad Sim_phi_sec_Rad Sim_T rqO1_Nm 1 phi1 2 phi2 3 omega1 4 omega2 phi_prim_Rad phi_sec_Rad omega_prim_Rad omega_sec_Rad SFunction Outer Arc Spring 1 TrqO1_Nm Sim_T rqO2_Nm TrqO2_Nm Sim_Trq_Springs_Nm SFunction Outer Arc Spring 2 1 Trq_Springs_Nm Sim_TrqI1_Nm T rqI1_Nm SFunction Inner Arc Spring 1 Sim_TrqI2_Nm T rqI2_Nm SFunction Inner Arc Spring 2 Figure 6. One static simulation and multiple dynamic simulations are performed. Simultaneously an increasing negative torque is applied on the secondary mass.
Luc Römers / XB department Technical University of Eindhoven Traineeship report Figure 6.63  .4: overall arc spring characteristic. DyFaSim & Simulink .LuK GmbH & Co.3: overall arc spring characteristic. Simulink (left) & DyFaSim (right) Figure 6.
Luc Römers / XB department Technical University of Eindhoven Traineeship report 6.5 shows the top layer of the model. top layer . ZMS Mkorrektur_regler_Nm Mkorrektur_regler_Nm kinematik_primär_RAD kinematik_primär_RAD Motormanagement Figure 6. Figure 6.5: idle drive train model.64  . Modellsteuerung Mzy linder_Nm (ohne Reibung) kinematik_primär_RAD kinematik_primär_RAD Mmotor_Nm Mmotor_Nm Mmotor_Nm Mbogenf eder_grundreibung_Nm Mkorrektur_regler_Nm Mbestandteile_Nm (ohne Reibung) kinematik_sekundär_RAD Verbrennungsmotor Einfache Leerlauf Antriebsstrang incl.2 Drive train model for idle simulation The drive train model described in this paragraph consists of a combustion engine. The following Simulink modules were already developed by LuK: • • combustion engine module (idle only) arc spring module Additional elements have been built to complete the idle drive train: • • • • • flange in dualmass flywheel inner damper in dualmass flywheel basic friction in dualmass flywheel idle speed controller (including speed filter) oscillating inertia of the combustion engine Furthermore the engine module is improved by correcting some small errors. idle drive train and an engine management module.LuK GmbH & Co.
From top to bottom. resulting in an average value. the components are: • • • • mass torque compression / expansion torque combustion torque friction torque For the combustion torque.LuK GmbH & Co. Mk/e_indiv _Nm Durchmesser Mbestandteile_Nm (ohne Reibung) 1 kinematik_primär_RAD <kinematik_primär_RAD> K/EMoment kinematik_primär_RAD M_Isermann_Nm winkel_max_degrees 2 Mmotor_Nm (25 40/4) Mstat_Nm M_Isermann_indiv _Nm Verbrennungsparameter (aus workspace) 2 Mkorrektur_regler_Nm Mkorrektur_regler_Nm Verbrennungsmoment The total torque consists of: Mass torque Compression/Expansion torque Combustion (Isermann) torque Friction torque Mreibung_Nm Motor Reibung Figure 6. The engine is controlled by the idle speed controller (inside the engine management subsystem). The oscillating primary speed is filtered.6 shows the combustion engine model (orange subsystem in figure 6.65  . This average value is fed into a PIcontroller.kinematik Mk/e_indiv _Nm Mzy linder_Nm Massenmoment 1 Mzylinder_Nm (ohne Reibung) M_Isermann_indiv _Nm kinematik_primär_RAD Mk/e_Nm Kurbelradius Berechnung der Zylindermomente Lamda 3 Verdichtungsv er. Luc Römers / XB department Technical University of Eindhoven Traineeship report Figure 6. All torque components are summed.6: combustion engine module . after which the total torque is sent to the drive train (primary mass). which controls the static value of the combustion torque. Four subsystems produce the different torque components.5). the Isermann empirical model is used. mosz Mmasse_Nm CalculateIndivCyli Lamda Activation Kurbelradius M_masse_indiv _Nm M_masse_indiv _Nm [Motparam] From Kurbelwellen. For this particular model the P and Igains are shaped (a function of the errorsize).
5). Between the primary mass and the flange.7: drive train module The inertia of the first mass is a function of the crank angle. The inner damper is mounted between the flange and the secondary mass.66  . It consists of a static inertia (belonging to primary DMF mass and crankshaft) and a dynamic inertia (belonging to the pistons). SimT rqEng 1 s 1 winkel_primär_RAD 1 Mmotor_Nm Mmotor_Nm geschw_primär_RAD Mgrundreibung beschl_primär_RAD Mzms_Nm geschw_primär_RAD 2 gesch_primär_RAD 3 beschl_primär_RAD beschl_primär_RAD Bewegungsgleichung der KW + Primärschwungscheibe 1 s 1 s geschw_primär_RAD 4 Mbogenfeder_Nm geschw_flansch_RAD DMF Arc Spring Models ZMS_Rueckwirk moment Flanschmassen. Luc Römers / XB department Technical University of Eindhoven Traineeship report Figure 6.LuK GmbH & Co. For idle operation the drive train can be modeled using three masses: • • • crankshaft & primary mass DMF flange secondary mass DMF The equation of motion of these three masses is represented by the rectangular subsystems in the figure below. .7 gives an overview of the idle drive train model (yellow subsystem in figure 6. The final component is the basic friction working between the primary and secondary mass.geschwindigkeit n_flansch grundreibung ID_Rueckwirk moment Flanschmassen. the arc springs are situated (subsystem with DMF picture).beschleunigung Bewegungsgleichung des Flansches delta_omega ID_Rueckwirk moment Innendämpfer 1 s 5 winkel_sekundär_RAD ID_Rueckwirk moment geschw_sekundär_RAD Mgrundreibung beschl_sekundär_RAD geschw_sekundär_RAD 6 geschw_sekundär_RAD 7 beschl_sekundär_RAD 2 Mgetriebe_schlepp_Nm Mgetriebe_schlepp_Nm beschl_sekundär_RAD Bewegungsgleichung der Sekundärschwungscheibe SimT rqLoad Figure 6.
The speedsignals of the primary and secondary mass are compared.10: secondary speed (0 < t < 5) Figure 6. The results are displayed in figure 6.LuK GmbH & Co. . Even though RungeKutta is used in both programs. Luc Römers / XB department Technical University of Eindhoven Traineeship report 6.8 until 6.67  .11: secondary speed (0 < t < 0.9: primary speed (0 < t < 0.3 Comparison with DyFaSim counterpart The complete idle drive train model is compared to its DyFaSim counterpart.5) Figure 6. Both models are given the same initial conditions and are simulated for five seconds.8: primary speed (0 < t < 5) Figure 6. the solver techniques are not completely the same.5) These results show that both models are 99% identical.11. The remaining small differences are probably caused by the solver methods. Figure 6. A RungeKutta (ode4) fixedstep solver is used.
drive train module (including dualmass flywheel) and an engine management module. involving complex engine management components. A training course will be organized to familiarize all engineers with this new tool. The model contains a combustion engine module. The main goal was to create a model that is similar to its DyFaSim counterpart. Based on this analysis it can be concluded that the standard model generation also delivers satisfying results. The MatLabbased tool is available as a standalone package. A starter motor could be added to the model. It forms an automatic interface between manufacturer engine data and the DyFaSim software. Conclusions Automatic generation of combustion engine models using MatLab A tool has been created that enables all engineers at the dynamic simulation department of LuK to generate engine models for the DyFaSim simulation software. Accurate engine models can be generated based on cylinder pressure data.68  . an engine model based on standard engine data can be generated. making it possible to do start / stop simulations. The graphical user interface provides easy operation. Luc Römers / XB department Technical University of Eindhoven Traineeship report 7. Furthermore a guideline has been written. describing the use of the DyFaSim combustion engine module and the new MatLabtool. In case no cylinder pressure data is available. Both methods have been compared by means of a harmonic analysis.LuK GmbH & Co. The current version makes it possible to choose between two languages. Idle drive train model in MatLab / Simulink An idle drive train model has been built in MatLab / Simulink. without the need for a MatLab license. The Simulink model will be used for special simulations. Further expansion to three or more languages is a possibility. . Comparison of output signals has shown that this goal has been accomplished.
sim. measurement & full map simulation Relative difference.sim.sim.cylinder Tstatic.scaled Tdrag. full map simulation Peaktopeak torque.69  . Luc Römers / XB department Technical University of Eindhoven Traineeship report List of symbols Symbol BDC cf d diffrelative. reduced map simulation Simulated torque Static engine torque Static cylinder torque Static engine torque Top dead centre Cylinder volume Clearance volume Displacement volume Piston position from top dead centre Shape factor Shape factor . measurement Peaktopeak torque.engine TDC V Vc Vd x α β Unit [] [] [m] [%] [%] [N] [N] [N] [N] [m] [kg] [] [Pa] [Pa] [Pa] [Pa] [Pa] [m] [Nm] [Nm] [Nm] [Nm] [Nm] [Nm] [Nm] [Nm] [Nm] [Nm] [Nm] [Nm] [Nm] [Nm] [Nm] [Nm] [Nm] [Nm] [Nm] [] [m3] [m3] [m3] [m] [] [] Definition Bottom dead centre Correcting factor Bore diameter Relative difference.r Fc. measurement & reduced map simulation Force applied on piston Radial force on crank Tangential force on crank Force acting on piston Connecting rod length Oscillating mass Number of cylinders Cylinder pressure Compression / expansion pressure Exhaust pressure Intake pressure Surrounding pressure Crank radius Torque due to combustion Torque due to compression / expansion Crankshaft torque Scaled drag torque curve Predefined drag torque curve Engine torque Simulated gas torque Torque due to mass forces Maximum torque Torque from measurement Scaled net torque curve Predefined net torque curve Peaktopeak torque.sim.sim Tmass Tmax Tmeasured Tnet.LuK GmbH & Co.reduced Fapplied Fc.t Fpiston l mosc ncyl P Pcomp/exp Pexhaust Pintake Psurrounding r Tcombustion Tcomp/exp Tcrank Tdrag.predefined Tengine Tgas.full diffrelative. measured Tpp.reduced Tsimulation Tstatic Tstatic.scaled Tnet.full Tpp.predefined Tpp.
LuK GmbH & Co. Luc Römers / XB department Technical University of Eindhoven Traineeship report ε ϕ ϕmax κ λ ω [] [rad] [rad] [] [] [rad/s] Compression ratio Crank angle Angle of maximum combustion torque Isentropic coefficient Crank/rod ratio Engine speed .70  .
Pischinger. Wien (1989) Engineering Fundamentals of the Internal Combustion Engine. Sinsel. Technical University of Eindhoven (2000) Verbrennungsmotoren. Taucar. Somers. R. A. G. Pennec. Th. J. Kraßnig. G.LuK GmbH & Co. Baert. R. R. Sams. New Jersey (1997) Berechnung der Massenkräfte am Verbrennungsmotor. Pearson PrenticeHall. Luc Römers / XB department Technical University of Eindhoven Traineeship report References [ENCA98] [ISER99] Microsoft Encarta 98. SpringerVerlag. Diplomarbeitsbericht LuK (2000) Dynamische Modellierung eines Verbrennungsmotors. Segler. S. O. Multimedia Encyclopaedia Hardwareintheloop simulation for the design and testing of enginecontrol systems. R. Moog. experiment report LuK (2000) Dictaat Verbrandingsmotoren 2.71  . Isermann. Berlin (1994) [MOOG00] [PENN02] [PISC89] [PULK97] [SEGL00] [SOME00] [URLA94] . Pulkrabek. Praktikumbericht LuK (2002) Thermodynamik der Verbrennungsmaschine. B. SpringerVerlag. Urlaub. Technical University of Darmstadt (1999) Modellierung eines Verbrennungsmotors. Schaffnit. B. W.
This action might not be possible to undo. Are you sure you want to continue?
We've moved you to where you read on your other device.
Get the full title to continue listening from where you left off, or restart the preview.