Professional Documents
Culture Documents
4th
Workshop
on
4th IFAC
IFACand
Workshop
on Control, Simulation and Modeling
Engine
Powertrain
4th
IFAC Workshop
on Control, Simulation and Modeling
Engine
Powertrain
Engine and
and
Powertrain
Control, OH,
Simulation
and online
Modeling
August
23-26,
2015. Columbus,
USA
Available
at www.sciencedirect.com
Engine and
Powertrain
Control, Simulation
and Modeling
August
23-26,
2015.
August 23-26, 2015. Columbus,
Columbus, OH,
OH, USA
USA
August 23-26, 2015. Columbus, OH, USA
ScienceDirect
A
A framework
framework for
for modeling
modeling and
and optimal
optimal
A framework
for modeling
and optimal
control
of
automatic
transmission
control of automatic transmission systems
systems
control of automatic transmission
systems
V. Nezhadali L. Eriksson
V.
Nezhadali L.
Eriksson
V.
V. Nezhadali
Nezhadali L.
L. Eriksson
Eriksson
Copyright
2015,
2015 IFAC
285 Hosting by Elsevier Ltd. All rights reserved.
2405-8963
IFAC (International Federation of Automatic Control)
Copyright
2015
IFAC
285
Copyright
under
2015 responsibility
IFAC
285Control.
Peer
review
of
International
Federation
of
Automatic
Copyright 2015 IFAC
285
10.1016/j.ifacol.2015.10.041
uf
Engine
TKi TBi
Te Tp
Tt
Torque
Converter
Tf
Transmission
Tw
Final drive Drive shaft
B1
K1
B2
K2
In
B3
B4
B5
K3
Out
K1
K2
K3
B1
B2
B3
B4
B5
400
2.4
300
1.8
200
1.2
100
0.6
0
0
0.2
0.4
0.6
0.8
F1R1
F1R1
()
MP1000 (Nm)
F1S1
F2S2
F3S3
F4S4
S
3
Tt
F 5R 5
F 4R 4
Tf C
F4S4
F3S3
S
In
F 3R 3
F5S5
F5R5
TK3
F5S5
TK3
TB5
Out
5 =
287
F 4R 4
2.3 Transmission
=
= 1
(7c)
out
4
R TB4
F2S2
F1S1
()
F2R2 F3R3
F 2R 2
0
1
287
6 =
(1 + 4 )
4
(1 + 3 )(1 + 4 )(1 + 5 )
4 + 3 4 + 3
(1 + 3 )(1 + 4 )
4 + 3 4 + 3
4 + 1
3 + 2 3
,
g1 =
4 + g1
2 + 2 3 + 3
4 + 1
3
,
g2 =
1
4 + g2
1 + 3 1+1+
2 +1 2
7 = 1
3 (1 + 4 )(1 + 5 )
8 =
(3 + 4 + 3 4 )g3 4 (1 + 1 )
8
9 =
,
g 3 = 1 + 1 + 1 2
1 + 5
(8a)
(8b)
(8c)
(8d)
(8e)
(8f)
(8g)
(8h)
Using the gear ratios, i and equation (7c)-(8h) the unknown i can be calculated. Knowing that the transmission is almost one meter long and using this as a scale, the
radii of sun gears (Si ) are read from Figure 1 and then the
radii of the ring gears (Ri ) are calculated from Ri = Si /i .
Applying the matrix method Mechanical links between
gearset components are shown in Figure 3. In order to
reduce the required number of state variables for describing all component dynamics, the linked components
of different gearsets namely sun(S)-ring(R)-carrier(C)
are considered as a single body resulting in the following
inertias:
J1 : S1
J 5 : C4 R5
J2 : R1 C2 R3
J 3 : C1 S2
J6 : S3 S4 + JT C,t J7 : C5
J4 : R2 C3 R4
J8 : S5
J1
J2
J3
J4
J5
J6
J7
J8
d1
dt
d2
dt
d3
dt
d4
dt
d5
dt
d6
dt
d7
dt
d8
dt
F1 S1 = 0
(9)
F1 R1 + F2 (S2 + R2 ) + F3 R3 = 0
(10)
+ F1 (S1 + R1 ) F2 S2 = 0
(11)
F2 R2 F3 (S3 + R3 ) + F4 R4 = TB4
(12)
F4 (S4 + R4 ) + F5 R5 = 0
(13)
+ F3 S3 + F4 S4 = Tt
(14)
F5 (S5 + R5 ) = TK3 Tf
(15)
+ F5 S5 = TK3 + TB5
(16)
I K
=
T
0
KT 0
(17)
Tw
f d
7
w )
f d
(19)
(20)
=
w
(21)
S1
0
0
0
0
dt
fd
R3
0
0
R1 R2 + S2
R + S
The wheel speed w is the last state variable in the comS2
0
0
0
1
0
plete
driveline model calculated from the vehicle longituR
R
S
R
0
2
3
3
4
,
K=
dinal
dynamics for level road condition according to:
0
0
0
R
S
R
4
4
5
1
1
dw
0
0
S3
S4
0
2 3
=
T w c d A w
rw m g cr rw (22)
2
0
0
0
0
S5 R5
dt
m rw + Jw
2
0
0
0
0
S5
where rw , m and A are wheel radius,, vehicle mass, and
0 0 0 0 0
Tt
gravity.
0 0 0 0 0
T
B4
0 1 0 0 0
T
B5
1 0 0 0 0
Using the developed state space model, optimal conTf
0 0 1 0 1
trol problems are formulated and solved for optimization
00 1 1 0
of gearshift transients. There are several properties of
where Tf is the transferred torque from the transmission gearshift which can be considered as the optimization
to the final drive and is calculated in the next section. objective, see Haj-Fraj and Pfeiffer (2002) and Haj-Fraj
Rearranging equation (17) into (18), the speed dynamics and Pfeiffer (2001) for a discussion. Here shift duration (T )
of every rotating component and the internal forces in each corresponding to the operating life of shift elements and
changes in the vehicle acceleration (jerk) corresponding
planetary gearset can be calculated from the following:
to the passenger comfort, are focused. The shift duration
as one of the objectives in the optimal control analysis is
1
I K
T
(18) represented by:
=
t
0
KT 0
T =
dt
(23)
0
The dynamic equations of the transmission components
where t denote the duration of the gearshift. The jerk, A,
i
can be simplified during stationary operations ( d
dt = 0)
is
represented by integrating the squared derivatives of the
and also when a component is fixed by means of shift
vehicle
accelerations during the gearshift as follows:
elements to another component or to the transmission
t
da
housing such that it has same dynamics as another com( )2 dt
(24)
A=
ponent or is stationary. For example during 1-2 up-shift
0 dt
equation (12) can be removed from dynamic equations where a = dw
dt .
since TB4 fixes R2 C2 R1 to the transmission housing and
4
To obtain the trade-off between min T and min A solu4 = d
dt = 0. The transmission dynamics can then be
calculated by removing J4 from I, 4 from , fourth row tions, the optimal control problem objective is often forof matrix K, second column of and TB4 from T vector. mulated as the weighted sum of the time and jerk terms
288
(25)
x(t),u(t),(t)
T or A
Implementing direct multiple shooting method, the formulated optimal control problem is transformed into a nonlinear program (NLP). The NLP is then solved by CasADi,
Andersson (2013), which is an open source optimal control
framework using Ipopt, Wachter and Biegler (2006), and
the powertrain transients are obtained. The results of this
optimization are presented and analyzed in the following
section.
A (rad/s )
(v1 T + v2 A, v1 + v2 = 1) where the points on the tradeoff are calculated by solving with different weights (v1,2 ).
However, considering the large difference in the order of
magnitude between the two objectives, when formulating
the weighted sum, A and J have to be normalized with
respect to their maximums. This makes the solution of the
weighted sum sensitive to the normalization and weight
values, which makes it difficult to calculate the trade-off
with an acceptable spread of points.
10
C12
5
0
1.32
(26)
x(t)
= f (x(t), u(t))
umin ui (t) umax
xmin xi (t) xmax
Te (e , uf (t)) Te,max
x(0) = x0
x(T ) = xT
x(0)
=0
x(T
)=0
a = 0, when solving for min T |A=0
289
1.33
1.34
s.t.
1.35
1.36
T (s)
1.37
1.38
1.39
1.4
A (rad/s3)
45 upshift
C
20
45
10
0
0.55
0.6
0.65
T (s)
0.7
0.75
0.8
Fig. 6. The trade-off between time and jerk for 1-2 and 4-5
up-shifts. At C12,45 there is a compromise between the
two objectives.
where the min and max are the upper and lower limits for
states and controls and x0 and xT correspond to the state
values at the end of the first gear and beginning of the
second gear operation, respectively. These are calculated
in similar manner as described by equations (7a)-(7c) for
rw w =6 km/h and rw w =15 km/h in 1-2 and 4-5 upshifts respectively. The requirements on x at the beginning
and end of the gearshift are applied to guaranty that the
gearshift starts and ends at stationary condition.
It should be noted that the kinematic constraint mentioned in equation (6) does not need to be added to
problem constraints as their effect is already present in
the dynamic equations of states via internal forces Fi .
289
100
0.4
0.6
0.8
1.2
1.4
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0
ub1 (Nm)
min T
min A
C
200
2000
4000
12
0.2
0.4
0.6
0.8
1.2
6000
1.4
0
200
400
600
(Nm)
2000
1000
0
u
0.2
uk3 (Nm)
400
200
mf
100
0
300
(gr/cycle)
200
b3
ub5 (Nm)
umf (gr/cycle)
300
0.2
0.4
0.6
0.8
1.2
1.4
min A
min T
C45
0
0.1
0.2
0.3
T (s)
100
50
0
1.9
0.2
0.4
0.6
0.8
1.2
1.4
1.8
1.7
0
250
200
150
0
250
0.2
0.4
0.6
0.8
1.2
1.4
0.2
0.4
0.6
0.8
1.2
1.4
0.8
1.2
1.4
0.2
0.4
0.6
0.8
1.2
1.4
0.2
0.4
0.6
0.8
1.2
1.4
0.6
(rad/s)
0.2 min0.4
A
min T
C
0.8
100
50
4.650
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.8
4.6
4.55
0
200
0.1
0.2
0.3
0.4
0.5
0.6
min A
min T
C 0.7
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.1
0.2
0.3
0.4
T (s)
0.5
0.6
0.7
0.8
45
150
3000
250
200
1000
0
100
12
0.2
0
0
0.7
150
(rad/s)
150
0
300
200
100
0
0
0.4
0.6
ice
200
0.5
T (s)
xflex (rad)
xflex (rad)
0.4
T (s)
0
100
50
0
50
0.40
0.2
0
0
Fig. 10. Optimal state transients for min T , min A and C12
cases in a 4-5 up-shift where out =7 and in =6 .
of the in-coming shift elements introduce disturbances
into the driveline such that the induced excitation is
transferred to the vehicle mass at wheels via the the
drive shaft. These excitations get partially damped by the
drive shaft flexibility, vehicle mass and wheel inertia which
increases the rate of reduction in the kinetic energy of the
transmission components and shortens the time before the
next gear can be engaged.
Subtracting the
total kinetic energy of all transmission
components (0.5 Ii i2 ) at the beginning and end of the
gearshifts, the difference is 25.17 % larger in case of the
1-2 up-shift which could be the main reason for the slower
transients compared to the 4-5 case as more reduction in
the kinetic energy is required before the next gear can be
engaged.
0.05
T (Nm)
600
400
200
dw/dt (rad/s )
0.4
0.6
T (s)
0.8
0.05
0
0.1
0.2
0.3
0.4
T (s)
0.5
0.6
1.2
1.4
min A, 45
min T, 45
min A, 12
min T, 12
C45
C0.7
0.8
1.2
1.4
12
0.05
0
d /dt (rad/s2)
0.2
0.05
0
0.2
0.4
0.6
T (s)
0.8
Fig. 11. Transmission input torque and vehicle acceleration/decceleration in min T , min A and C45 solutions
of 1-2 and 4-5 up-shifts.
5. CONCLUSION
A framework is developed for gearshift transient optimization during inertia phase via state space modeling of a
nine speed heavy duty automatic transmission. Using the
developed model and in order to analyzed the minimum
time/jerk transients and the trade-off between these, optimal control problems are formulated and solved. Two
example up-shifts are considered and in order to calculate the trade-off between time and jerk objectives while
avoiding objective function normalizations difficulties, the
minimum jerk problem is iteratively solved for various
preselected gearshift durations.
The results show that the developed framework is applicable for efficient optimization of inertia phase gearshift
transients. As future model developments, hydraulic actuation dynamics can be included and sensitivity of the
gearshift transients with respect to the parameters such
as clutch fill dynamics, components inertia and driveline
flexibilities can be analyzed.
REFERENCES
A40G (2015).
Volvo A40G hauler, http://www.
volvoce.com/SiteCollectionDocuments/VCE/
Documents%20Global/articulated%20haulers/
Brochure_A35GFS_A40GFS_SV_12_20040745_B_2015.
02.pdf, Accessed July. 2015.
Andersson, J. (2013). A General-Purpose Software Framework for Dynamic Optimization. PhD thesis, Arenberg Doctoral School, KU Leuven, Department of Electrical Engineering (ESAT/SCD) and Optimization in
Engineering Center, Kasteelpark Arenberg 10, 3001Heverlee, Belgium.
Eriksson, L. and Nielsen, L. (2014). Modeling and Control
of Engines and Drivelines. John Wiley & Sons.
Gao, B., Chen, H., Zhao, H., and Sanada, K. (2010).
A reduced-order nonlinear clutch pressure observer for
automatic transmission. Control Systems Technology,
IEEE Transactions on, 18(2), 446453.
Goetz, M., Levesley, M., and Crolla, D. (2005). Dynamics
and control of gearshifts on twin-clutch transmissions.
291
291