SIMULATION OF A VEHICLE IN LONGITUDINAL

MOTION WITH CLUTCH LOCK AND CLUTCH

RELEASE

Lars Eriksson

Vehicular Systems, Linköping University, SE-581 83 Linköping

Tel: +46-13-28 44 09, Fax:+46-13-28 20 35
E-mail: [email protected]

Abstract: A simple model for driver and vehicle in longitudinal motion is developed
and simulated. The focus is on describing and handling simulation of clutch lock and
clutch release which changes the model structure, both during start and gear shifts,
in Simulink. Special attention is given the problem of simulating start and stop of
a vehicle with rolling resistance at zero speed. Only principles for simulating the
system with variable structure is of interest and therefore the models are maintained
at lowest possible complexity. The system is successfully simulated and the validation
is performed using three scenarios: one with only clutch lock and clutch release, one
with start and stop of vehicle, and one full European drive-cycle.

Keywords: Driveline, transmission, friction, rolling resistance.

1. INTRODUCTION together and the system has only one degree of

freedom. Thus at the instant when the clutch
Simulation of complete vehicles with engine and locks the system changes structure as one state
driveline is an important tool, for example when disappears and likewise when the clutch unlocks
investigating how different driveline configura- when one state appears. Another issue is the
tions effect the fuel consumption and emissions rolling resistance at zero velocity, i.e. the torque
in a driving cycle. A vehicle contains many com- needed to start moving the vehicle from zero.
ponents and the purpose of the study dictates This is further described in Section 3. Both these
what components must be included as well as problems have been addressed by many authors
the level of detail of the components. The clutch and the solutions proposed here are not new. The
connects two masses to each other and introduces focus here is to develop simple models that isolates
intricate difficulties when simulating the clutch the simulation difficulties and are well suited for
connect and release as the structure of the system educational purposes. In addition the presentation
changes. Emphasis is on isolating and describing address both issues and solves them such that the
the simulation problems in an introductory man- model can be simulated in the readily available
ner such that students easily can understand the tool Simulink (Mat 1999).
difficulties when simulating a system with chang- The aim here is to simulate the clutch lock and
ing structure. release as well as the rolling of the vehicle without
Two issues that pose problems in simulation are losing the perspective due to a too complex total
addressed. First, when the clutch is separated the model. Therefore simple models, that capture
system has two degrees of freedom, which in the the behavior without hiding them in details, are
simulation is represented by two states. When the developed for the different vehicle subsystems
clutch is locked then the two masses are locked shown in Figure 1.
 Engine Transm. Vehicle In the expression above the kinetic friction Tc,k
and is modeled, i.e. the clutch torque when the clutch
Ie
Te Clutch wheel disks are slipping. However, when the clutch is
6 T 6 locked the clutch can transfer a higher (static)
?c torque. The quotient between the static and dy-
T̃ 6 Tw namic friction is modeled as a constant, where
?c ?
Gears Tc,s (t)
= cs,k > 1 (2)
Tc,k (t)
Fig. 1. Sketch of a vehicle consisting of engine,
The maximum static friction torque is thus
transmission, and vehicle body. The trans-
mission includes clutch, gear box, and final Tc,s (Fc (uc (t))) = Tmax,k cs,k uc (t) (3)
drive. The arrows indicate the sign conven-
The clutch and the surrounding system have two
tion: up=driving torque, down=load torque.
important system states and two important events
2. TRANSMISSION MODEL for the transition between them.

First the clutch model is described in isolation to S1 - Two mass system: The system has two
show how cluch lock and clutch release is deter- masses that rotate with speeds independent of
mined. Then the gearbox and final drive are added each other but their respective movements are
and the algebraic transformaitions over a set of connected to each other through to the transfered
gears are shown. Finally an illustrative simulation clutch torque. In this state the systems surround-
validates that the model works properly. ing the clutch have two degrees of freedom and
the equations that govern their rotations are
dωe
Ie = Te − Tc
2.1 Friction Clutch dt
dωv
Iv = Tc − Tv
Three basic assumptions are made, the friction dt
clutch is: inelastic, mass less, and there are no The torque transfered through the clutch is
friction losses (except the obvious between the
Tc = Tc,k (fc (uc )) sgn(ωe − ωv )
plates). To describe the clutch behaviour it is
necessary to connect it to two rotating masses where sgn is the signum function. The equations
with inertias Ie and Iv , corresponding to the above are valid when ωv 6= ωe .
engine side and the vehicle side respectively. A
S2 - One mass system: The clutch disc locks
torque Te is driving the system and a torque Tv is
the two rotating systems to each other and the
braking the system, see Figure 2.
two masses rotate with identical speed. Under
The clutch pedal position uc is connected to the these conditions the rotation of the surrounding
clutch discs and produces a force that presses the two masses have only one degree of freedom and
clutch discs together, Fc (uc ). Here uc (t) ∈ [0, 1] the equations that govern the rotations are
and 0 means separated clutch and 1 means full dωe
force on the clutch discs. This force influences (Ie + Iv ) = Te − Tv
dt (4)
the maximum torque that the clutch can transfer ωv = ωe
Tc,k (Fc (uc (t))). Here a simple linear model from
the input uc is used When the clutch is locked the torque transfered in
the clutch Tc,l depends on the applied torques and
Tc,k (Fc (uc (t))) = Tmax,k ∗ uc (t) (1) the inertias. The transferred torque is the driving
Iv torque minus the inertia effect of the first mass
Ie
Clutch dωe
Fc Fc Tc,l = Te − Ie
dt
Te Tv Inserting Equation 4 and manipulating gives the
clutch torque
ωe ωv
Te Iv + Tv Ie
Tc,l = (5)
Ie + Iv
Fig. 2. Sketch of the clutch that shows the sign
E1 - Clutch lock: The instant when the clutch
conventions of the applied torques and veloc-
locks the two rotating systems to each other. Prior
ities. Te is the driving torque from the engine
to the lock event the two systems rotate freely
side and Tv is the load torque on the vehicle
other and two conditions must be fulfilled for the
side. Clutch torque, Tc is not shown but the
clutch to lock:
sign convention is that it loads the engine
inertia and drives the vehicle inertia. (1) The angular velocities must match ωv = ωe .
(2) The torque that the clutch transfers under 1.5
Clutch position and Clutch locked information

locked conditions, Equation 5, must be less 1

or equal to the maximum possible static fric- 0.5

Clutch Position
tion, Equation 3. I.e the following equation 0
Clutch Locked

that must be fulfilled Tc,l ≤ Tc,s . −0.5

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5

By continuously monitoring these conditions the

0.4
instant when the clutch locks can be determined.

Input torque [Nm]

0.3

In the Simulink implementation the first condition 0.2

0.1

is checked using a hit crossing block which forces 0

−0.1

the simulation to detect the instant when the −0.2

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
velocities match. If condition 2) is also fulfilled
1.4
at that instant then the clutch is locked and the
system is set to state S2.
1.2

ωe [rad/s]
1

E2 - Clutch breakup: The condition for deter- 0.8

mining the clutch breakup is when the torque 0.6

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5

that the clutch transfers under locked condition 0.6

exceeds the maximum for the static friction torque 0.4
in the clutch
ωw [rad/s]
0.2

Tc,l > Tc,s 0

−0.2
This causes the clutch to start slipping and the 0 0.5 1 1.5 2 2.5
time [s]
3 3.5 4 4.5 5

two masses start rotating with differing speeds.

In the simulation a new state must be initialized Fig. 3. Simulation of clutch lock and unlock a
using the relation ωe = ωv . sequence of changes in the clutch control
signal, and input torque.
Cumulative energy plot
0.55
2.2 Gear box and final drive 0.5

0.45
Energy [J]

The gear box and final drive are modeled as 0.4

0.35
rigid, friction free, and mass less elements that are 0.3
Ee
Ee+Ew
Ee+Ew+Wc
lumped together into one set of gears, see Figure 1. 0.25

0.2
The cogwheels in the transmission change the 0 0.5 1 1.5 2 2.5
time [s]
3 3.5 4 4.5 5

torque and rotational speed. The gear ratio ig

connects the incoming and outgoing rotational Fig. 4. Cumulative plot of energy stored in the
velocities and torques masses and dissipated in the clutch for the
simulation shown in Figure 3.
ig = ωv /ωw , ig = Tw /Tv (6)
using (6) and (7). Wheel speed is selected as the
where subscript v corresponds to the side con-
only state in S2. A similar model of a clutch,
nected to the clutch and w the wheel side. A mass
but without gears, has previously been published
with inertia Iw connected to the wheel side of the
in Mathworks Automotive Examples (Mat 1998).
gears can be transferred to an equivalent inertia
That model is freely available at Mathworks web-
at the clutch side through
site and is well suited as a starting point. An
Iv = Iw /i2g (7) important remark is that during the simulation
it is important that the clutch is fully engaged
Using these transformations a rigid system with
uc = 0 when a change in gear ratio is made, oth-
gears can be transformed into a system shown in
erwise Simulink might have difficulties simulating
Figure 2 and these equations can be used directly.
the system.

2.3 Simulink Implementation 2.3.1. Clutch Lock and Unlock – Simulation A

simple clutch lock and unlock simulation is used
The clutch and gearbox are implemented in to validate that the clutch and gears are working
Simulink (Mat 1999) using enabled subsystems properly, Figure 3. Two rotating masses without
with resettable integrators. One subsystem imple- friction with Ie =1 and Iw =2 are connected to
ments S1 and the other S2, control logic is imple- the transmission model with gear ratio is ig =2.
mented based on E1 and E2 that determine what Initially Ie rotates at 1 rad/s and Iw does not move
subsystem to run. The wheel and engine speeds at all. The clutch signal is also zero. Between t=1 s
are selected as states in S1, which means that the and 2 s uc is ramped from 0 to 1 which result
clutch torque, inertia, and speeds are transformed in a decrease in ωe and an increase in ωw . At
t=1.25 s, ωw ∗ ie equals ωe and the clutch locks. brake pedal position Tb = Tb,max ub ub ∈ [0, 1].
At t=2 s an torque is applied on the engine side The rolling resistance is modeled as a torque that
of the clutch and the system starts to accelerate. opposes the wheel rotation, and a linear function
Between t=3 s and 4 s the clutch is ramped from 1 of the rotational speed is used Tr = m(c1 + cv ω).
to 0 and just before t=4 s the clutch releases and See for example (Gillespie 1992) for a discussion
Iw stops accelerating. of the model and the model parameters. In the
following the influence of the gravitational force
One important verification of the operation is to
is omitted.
study the energy conservation, Figure 4. In the
figure a cumulative plot of the energy, contained The vehicle mass m that has a translatory move-
in and dissipated by the system, is plotted for the ment and is connected to a rotating system
simulation shown in Figure 3. The lowest dash- through a rigid wheel with radius rw can be trans-
dotted line shows the rotational energy in Ie , formed into an equivalent rotating inertia
I ω2
Ee = e2 e when the clutch is engaged at t=1 s 2
Iw = m · rw
energy is transferred to the second inertia and
the middle line in the plot shows the sum Ee + The forces acting on the vehicle can likewise be
I ω2 I ω2 transformed to torques that act on the equiva-
Ew = e2 e + w2 w . The total rotational energy
decreases when the clutch is engaged which is lent rotating mass, T = F rw . In the simulation
due to the energy dissipated in the friction clutch implementation the state corresponding to the
while the clutch locks.RThe work produced on the vehicle speed is the equivalent rotating velocity
clutch disks is Wc = Tc,k (t) ∆ω(t) dt. The sum ωw = v/rw .
Ee +Ew +Wc is constant, 0.5 J, when no additional
energy is put into the system. At t=2 s the total 3.0.2. Lumped Model The air drag and rolling
energy starts to increase due to the added input resistance models are lumped together and param-
torque. From this validation it can be concluded eterized using a second order polynomial in ωw .
that the implemented clutch model is correct and 2
that it conserves the energy. Tw,r = c0 + c1 ωw + c2 ωw
Where c0 to c2 are fitted to a specific vehicle con-
figuration using the least squares method. Four
3. VEHICLE MODEL points have been used to determine the maximum
speed point and the three vehicle speeds 70, 90,
In this section a longitudinal vehicle model is and 110 km/h. There were no data available for
lumped to an equivalent rotational model un- velocities close to zero so the “dry friction” con-
der the assumption of a rigid wheel that rolling stant c0 is not accurately determined. However,
without slipping. The forces acting on the ve- the exact value of c0 is not important here since
hicle is modeled using the standard models de- this is only a illustrative simulation study.
scribed in for example (Kiencke and Nielsen 2000).
Forces on the vehicle that influence the motion are An important simulation aspect is how the torque
the sum of the driving force from the wheel Ff Tw is handled when the vehicle is at rest but
(which comes from friction between the wheel and there is a torque driving it. This can be handled
ground), air drag Fa , and the gravitational force analogously to standard friction models, see e.g.
which together with Newton’s second law yields (Olsson et al. 1998). When the vehicle is moving
dv the torque Tw can be subtracted directly from
m = Ff − (Fa + m g sin(α)) = Ff − Fw the driving torque to get the angular acceleration
dt
but when the vehicle is standing still the dry
where m is the vehicle mass, and α is the slope
friction and the braking torque, c0 + Tb in the
of the road, g is the gravitation. The air drag is
model for Tw , must be matched to the driving
Fa = 12 cw A ρ v 2 where cw is the air drag coeffi-
torque, T̃c . This matching is necessary since the
cient, A is the effective area, ρ the air density, and
driving torque must overcome the dry friction in
v the vehicle velocity. The wheel is considered to
order to make the vehicle move. An incorrectly
be rigid wheel slip slip is neglected and the wheel
designed friction model can produce chattering in
inertia is considered to be negligable in compar-
the vehicle speed. The implemented model is
ison to the vehicle mass. Under these conditions 
the following torque balance can be stated for the min(T̃c , c0 + Tb ) if ωw = 0
Tw = 2
wheel c 0 + c1 ω w + c2 ω w + Tb otherwise
T̃c − Tr − Tb = Ff rw The torque Tw is always applied such that it
opposes the motion.
where T̃c is a driving torque that comes from the
clutch via the gear box, Tb is the brake torque, It is important to note one detail; The necessity
rw is the wheel radius, and Tr rolling resistance. for the wheel friction model to have information
The brake torque is modeled as a function of the from the transmission about the driving torque
 300
Simple Non−Linear Driver Model
Manual
250 Gear
ue=1 1 Switch1

200 Clutch 1
Gear
From Workspace1 Gear
From Workspace
150
Torque [Nm]

1
Speed demand
2
100
Clutch
2

PI input

PI reset
Manual
Actual speed Gas Pedal
50 1
PI Driver

PI out
Switch
0
3
ue=0 Gas Pedal
max
Gas/Brake
−50 Mode Switch
0 1000 2000 3000 4000 5000 6000 MinMax Product
Engine speed [rpm] 0

min −1 4
Fig. 5. A map of the engine torque Tdemand (N, ue ). MinMax1
Brake Pedal
Brake Gain
The lines in the plot represent constant ue .
The idle speed is set to 800 RPM which is
Fig. 6. Simulink implementation of a simple
seen in the plot Tdemand (800, 0) = 0 Nm.
driver.
The maximum torque is 280 Nm and the
engine speed where T starts to decrease is Tdemand is the non-linear function, shown in Fig-
3500 RPM. ure 5, of the engine speed and the accelerator
pedal signal, ue ∈ [0, 1], from the driver.
such that the rolling resistance torque Tw can be
matched to the driving torque T̃c .
5. DRIVER MODEL

The driver is provided with the following infor-

4. ENGINE MODEL mation: reference speed, measured vehicle speed, a
clutch signal, and a gear signal. The driver is mod-
A physically justified engine model is used for the eled using a PI controller with non-linear actions
torque from spark ignited engines, see Figure 5 for governed by the clutch signal. The functionality of
the torque function. The engine torque is param- the driver is illustrated by the Simulink, Figure 6.
eterized to capture the following features: 1) The The output from the PI controller ranges from -1
maximum torque Te,max . 2) The decrease in the to 1 where 1 represents the maximum acceleration
torque curve at high speeds, which represents the and -1 represents maximum braking. Three things
choking of the engine. 3) Idling and idle speed con- that reflect a real driver are visible in the plot:
trol. 4) Engine braking at speeds over idling. The
model is derived from an ideal thermodynamic (1) The driver can not press the brake and gas
cycle that includes pumping work but neglects pedal simultaneously. This is implemented by
residual gases. The basis for the parameterization the min and max selectors connected to the
is to determine a lowest air mass flow and a output of the PI Driver.
highest air mass flow and the gas pedal position (2) When the driver pushes the clutch down he
gives a desired air mass flow in between these. also decreases the gas pedal. This is imple-
From this a desired intake manifold pressure is mented by the multiplication on gas pedal
determined using a model based on constant vol- output.
umetric effiency. The manifold pressure is then (3) When a new gear is in place and the clutch is
limited to atmospheric pressure. Finally an ideal engaged then the information from the last
Otto cycle with pumping work and friction is used gear is forgotten. This is implemented by
to determined the output torque. The resulting reseting the integral part of the PI Driver.
torque model is shown in Figure 5 as a function
of engine speed and control input ue . The focus
on air mass flow can be motivated since nowadays 5.1 Driver tuning
many engine management systems have drive-by-
wire systems that implement air mass flow control The driver was tuned to a driving cycle using a
instead of just throttle plate position control. simple manual method. When the P-part of the
PI-controller is increased the error between the
A torque developing engine dynamics is modeled desired speed and actual speed decreases but both
using a first order system with a time constant of the engine and brake torques increase. The P-part
0.1 s from the demanded torque. was tuned first and a trade off was made between
1 good following of the desired vehicle speed and
Te = Tdemand reasonable control signals. Then the integral part
0.1s + 1
 1.5 Clutch position and Clutch locked information
Clutch Position 1.5
Clutch Locked
1
1

0.5
0.5

0
0
Clutch Position
−0.5 Clutch Locked
0 5 10 15 20 25 30 −0.5
0 100 200 300 400 500 600

20
3000

15 2500
v [km/h]

2000

Ne [rpm]
10
1500

5 1000

500
0
0 5 10 15 20 25 30 0
time [s] 0 100 200 300 400 500 600

1.5
150

1
100
v [km/h]

v [km/h]
0.5
50

0
0

−0.5
10.5 10.55 10.6 10.65 10.7 10.75 10.8 10.85 10.9 10.95 11 −50
0 100 200 300 400 500 600
time [s]

250
T−Engine

Fig. 7. Start and stop of vehicle showing that there 200

150
T−Wheel
T−Brake

is no chattering in the vehicle speed at start

T [Nm]
100

and stop. 50

−50
0 100 200 300 400 500 600
was increased to yield zero stationary error within time [s]

approximately one second.

Fig. 8. Simulation results showing successful sim-
ulation of an European driving cycle.
7. CONCLUSIONS
6. SIMULATION RESULTS
A simple longitudinal vehicle model with vari-
The complete model is validated using two sce- able clutch structure is described and tested. The
narios; First, the vehicle start from zero velocity model can handle the problems of simulating
which especially validates the rolling resistance clutch lock and clutch release both when the ve-
model. Then the complete system is submitted to hicle is starting and during gear shifts. Special
a complete European driving cycle. attention is also given the problem of simulating
start and stop of a vehicle with rolling resistance
at zero speed. The system is successfully simulated
in Simulink and the validation is performed using
6.1 Vehicle Start and Stop three scenarios: one with only clutch lock and
release, one with only start and stop of vehicle,
A start and stop scenario is shown in Figure 7. and one full European driving cycle.
The top plot shows the clutch information. The
second plot shows the desired speed and the actual
speed. The third plot shows and enlargement of 8. REFERENCES
the vehicle start, note that there is no chattering
Gillespie, Thomas D. (1992). Fundamentals of
in the vehicle speed (solid line in the plot) when
Vehicle Dynamics. SAE International.
the vehicle starts to roll.
Kiencke, Uwe and Lars Nielsen (2000). Automo-
tive Control Systems For Engine, Driveline,
and Vehicle. Springer-Verlag.
Mat (1998). Simulink-Stateflow Technical Exam-
6.2 Complete Drive Cycle
ples: Using Simulink and Stateflow in Auto-
motive Applications.
A simulation of the complete European drive cycle Mat (1999). Simulink , Dynamic System Simula-
is shown in Figure 8. As it can be seen the tion for MATLAB – Using Simulink, Version
complete test cycle is successfully simulated. One 3.
thing that is interesting to note is that the engine Olsson, H, KJ Åström, C Canudas de Wit,
speed over shoots at the gear changes when the M Gäfvert and P Lischinsky (1998). Friction
vehicle is accelerating. This is due to the first order models and friction compensation. European
model of the engine dynamics in combination with Journal of Control 4(3), p.176–195.
the driver that does not have any prior knowledge
of the driving cycle it is following and does not
know of the upcoming gear change.