Sunteți pe pagina 1din 5

Disctrete Time Feedback Linearization Method for

Dynamical Adaptive Control of Electronic Throttle


Mihaela Doina Zamfir
“Gh. Asachi”, Technical University of Iasi”
Iasi, Romania
dzamfir@tuiasi.ro

Abstract—The throttle system is an electromechanical system


represented by a nonlinear dynamical model. In construction ETM Position
some nonlinearityes exist into the throttle. The most significant sensor
nonlinearity is a dead-zone. It has its origin in a security position
of the throttle position. There are other nonlinearities e.g.
Coulomb friction and nonlinearities related to wear.
The work is focused on the control of the position of the APM ECM ETC Throttle
throttle. Discrete-time designs of electronic throttle control valve
systems are required for implementation on electric traction
borne computers. This paper presents an nonlinear, discrete-time
control system design method. These is the discrete-time
feedback linearization method. First, the system dynamics is
transformed into linear, time-invariant form. Next, the pole
placement control techniques it is applied to the transformed
control problem. In the presence of parametric variations, and
perturbation the linearized system is modified. Next, we propose Figure 1. Schematic view of the airflow control system
to use the adaptive dynamical control to minimize the
consequences of these parametric variations. The resulting
control scheme is applied on the original system and its
effectiveness is evaluated by simulation tests. Return
spring
Keywords-component; electronic throttle control, nonlinear
system, discrete-time feedback linearization, dinamic control.

DC
I. INTRODUCTION Motor
In the airflow modern control system the accelerator pedal (Ua,Ra,La)
is not directly connected to the throttle as in elder car models.
Fig. 1 shows a schematic view of the airflow control system.
The Accelerator Pedal Module (APM) sends a velocity
reference to the Engine Control Module (ECM), which
transforms the reference to a plate position and sends it to the
Electronic Throttle Module (ETM). The inner control loop Thrrotle Position
valve sensor
then moves the plate accordingly. The outer control loop is
also used for safely and monitoring of the throttle.
Electronic throttle control, ETC is an electromechanical
system [3], [4], [5], that controls the throttle valve position. Gear
Fig. 2 shows a diagram of a throttle system. This system has a
DC motor that manipulates the throttle valve, and a sensor to
measure the throttle position. The technique proposed is the
discrete-time version of the feedback linearization design
approach. Next, the pole placement strategy and dynamical
adaptive control are used for the discrete linear model. Figure 2. Throttle construction
II. DISCRETE-TIME FEEDBACK LINEARIZATION METHOD Linear control technique is used to design controllers for
The feedback linearization technique [1,], [2], [3] the transformed model. For the case of a single control input,
the transformed model can be written in the state-space form as
assumes that the system dynamics can be partitioned with
shown below.
respect to each control variable:

.
.
x1 = f1 ( x1 ) + g1 ( x1 , x2 ) z = Az + Bν (4)
.
x 2 = f 2 ( x1 , x2 ) + g 2 ( x1 , x2 , x3 ) (1) Where:
.............................................
.
x n = f n ( x1 , x2 ,..., xn ) + g n ( x1 , x2 ,..., xn )u ⎡0 1 0 ... 0⎤ ⎡0 ⎤
⎢0 0 1 ... 0⎥⎥ ⎢0 ⎥
In this form, the system dynamics appear as chains of ⎢ ⎢ ⎥
integrators with all the system nonlinearities moved to the A = ⎢. . . . .⎥ B = ⎢.⎥ (5)
input. ⎢ ⎥ ⎢ ⎥
⎢0 0 0 . 1⎥ ⎢0 ⎥
⎢⎣0 0 0 ... 0⎥⎦ ⎢⎣1⎥⎦
(n)
x1 = f ( x1 , x 2 ,..., x n ) + g ( x1 , x 2 ,.., x n )u (2) (2)

The right hand side of equation (2) is then replaced with a Next, is obtain a discrete-time, linear dynamic system :
pseudo control variable ν to yield the transformed system:
z[k + 1] = Ad z[k ] + Bdν [k ] (6)
(n)
x1 = ν = f ( x1 , x 2 ,..., x n ) + g ( x1 , x 2 ,.., x n )u (3)
In this work, the pole placement discrete-time control law
it is designed as:
ν [k ] = G[k ]z[k ] (7)

The actual control variables u[k] can then be recovered


Nominal
linear system from the pseudo-control variables ν [ k ] :

ν (z[k]) − f (x[k]) [r[k] − K T X[k] − f (x[k ])]


f(x)
u[k] = = (8)
g(x[k]) g(x[k])

x
ν III. MODELING AND CONTROL DESIGN
Throttle

+
A. FEEDBACK LINEARIZATION
r u For control design, it is of importance to know the
characteristics of controlled object . First, we build a model of
Linear x
a throttle chamber system, [3], which is controlled object (Fig.
g(x) 1). The relations between current ia and input voltage Ua in the
Control
K armature circuit is described as:

dia dθ
La + Ra ia + K e N = Ua (9)
dt dt

Where La and Ra are the inductance and resistance in the


Figure 3. The pole placement strategy circuit, respectively, Ke is the inductive voltage constant, N
is the gear ratio, and θ is the throttle angle. Now, the ⎡ . ⎤ ⎡. ⎤
electric-magnetic torque generated by the DC motor is given ⎢ x1 ⎥ = ⎢ x1 ⎥ = ⎡ x2 ⎤ ⎡ x2 ⎤ ⎡ x2 ⎤
= = (18)
⎢ . ⎥ ⎢ .. ⎥ ⎣ f ( x) + g ( x)u ⎥⎦ ⎢⎣r − K T X ⎥⎦ ⎢⎣ν ⎥⎦

T = K t ia ⎣ x2 ⎦ ⎣ x1 ⎦
(10)
The command u, for the feedback linearization discrete
Ignoring the inductance La, we have the following equation technique and pole placement method :
by substituting (10) into (9):

d ( 2 )θ dθ dθ ν ( z[k ]) − f ( x[k ]) [r[k ] − K T X [k ] − f ( x[k ])]


J +D + d k sgn + k sθ (t ) = NK t i a (11) u[k ] = = (19)
dt 2
dt dt g ( x[k ]) g ( x[k ])

where J is the equivalent moment around the throttle axis,


D is the frictional coefficient and Ks is the return spring B. DINAMICAL ADAPTIVE CONTROL
constant.
In the presence of parametric variations, and perturbation P
.. 1 N 2 Kt Ke . 1 . 1 NK ,Fig.3, the linearized system is modified.
θ = − (D + )θ − dk sgn(θ ) − Ksθ + t Ua (12)
J Ra J J Ra J
X [k + 1] = ( A + ΔA) X [k ] + ( B + ΔB)u[k ] + P[k ] =
. . (20)
We note the state variables
x1 = θ , x1 = θ (13) = f ( X [k ], P[k ]) + B p u[k ]

We propose to use the adaptive dynamical control to


⎡x ⎤
. minimize the consequences of these parametric variations,
⎢ 1⎥ = ⎡ x2 ⎤ ⎡ b1 ⎤
+ U
fig.4. Commonly, in order to introduce tracking requirements,
⎢ . ⎥ ⎢⎣ a1 x1 + a2 sgn( x21 ) + a3 x2 ⎥⎦ ⎣⎢b2 ⎦⎥ a (14)
the model-based approach (SMR) is chosen, [6], [7]. The
⎣ x2 ⎦ reference model imposes the dynamics to be followed by the
linearized system. The reference model is chosen as a copy of
y = [ x1 x2 ]
T
the theoretical linearized system in the nominal case:

where:
SMR[k + 1] = Am SMR[k ] + Bm r[k ] (21)

Kt d D N 2 Kt K s Define the error state


a1 = − ; a 2 = − k ; a 3 = −( + ( )
J J J Ra J
(15) e[k ] = X [k ] − SMR[k ] (22)
NK t
b1 = 0; b2 =
Ra J The objective of the second step of the design procedure is
to choose the control input vector du which imposes tracking
For control design the following simplified model is error stabilization, e[ nT ] → 0 in a short time. For this, in
utilized: Fig. 4, is considered un uncertain linear system and is utilized
un Reference Model (SRM). The uncertain linear system is
described now:
⎡ . ⎤ ⎡ . ⎤⎡ x2 ⎤ ⎡0⎤
⎢ x. 1 ⎥ = ⎢ x.. 1 ⎥ ⎢ ⎥ + ⎢ ⎥u (16)
⎢⎣ x 2 ⎥⎦ ⎢⎣ x 1 ⎥⎦ ⎣a 2 sgn( x 2 ) + a3 x 2 ⎦ ⎣b2 ⎦ X [k + 1] = f [ X [k ], P[k ]) + u[k ]
u[k ] = K * X [k ] (23)
Where:
As a result of the linear dependence between states and
u = U a ; f ( x) = a2 sgn( x2 ) + a3 x2 ; g ( x) = b2 (17) forces, the problem is reduced to solving a linear system.
Replacing the state X with the desired state, SMR, at any
moment in time, the system is rewritten , (24).
Then, the pole placement control techniques it is applied
to the transformed control problem. In Fig. 4, is considered un uncertain linear system .
controlled plant H. The mathematical model looks like for this
uncertain plant model:
Xm

Reference
Model (SRM)
H ( s ) = H ( s )(1 + ΔH ( s )) =
r PER e ϖ2 (26)
T = 2 (1 + W ( s )δ ( s ))
Uncertain linear s + 2ςω + ϖ 2
system
The uncertain plant model is a lightly-damped, second-
u+du
-
order system with parametric uncertainty and significant
Dynamical frequency-dependent unmodeled dynamics. The parameter
adaptive control ϖ , ξ are assumed to be about 30% uncertain, with a nominal
value of 4 respectively 2. The frequency-dependent uncertainty
at the plant input is assumed to be about 20% at low
frequency, rising to 100% at 6 radians/second. First, the
controller is tested without adaption of the parameters. The
Figure 4. Robust adaptive control. results are provided in figures (5-7). Fig. 5 provide the
disturbance rejection for the nominal and the uncertain plant
model . In Fig. 6 and Fig. 7 is presented variability position of
X[k + 2]− F(X[k +1], P[k +1]− Bpu[k] = du[k] the throttle and Bode responses obtained without and with the
reference model. Without the reference model, the parametric
X[k + 2]− F(X[k +2], P[k +2]) − Bpu[k] = du[k]+du[k +1] variations introduced in the control algorithm an important
steady-state error. With the use the reference model, (SMR),
............................................ (24) correct tracking is obtained for the components.
Z[k + h]− F(Xk +h −1], P[k +h −1]) − Bpu[k] =
(du[k]+ du[k +1]+... + du[k +h −1])
Disturbance rejection
1
Nominal
SMR[k + 2] − F ( X [k +1], P[k +1]) − Bpu[k ] = du[k ] Samples
0.8
SMR[k + 3] − F (SMR[k + 2], P[k + 2]) − Bpu[k ] =
= du[k ] + du[k +1] (25) 0.6
SMR[k + h] − F (SMR[k + h −1], P[k + h −1]) −
−Bpu[k ] = du[k ] + ... + du[k + h −1] 0.4
A m p litu d e

Because the terms in the left are determined at the moment


k , there can be computed the components of the command 0.2
vector u for I=0…h-1.
This approach allows the computation of the command 0
vector du, in terms of the restrictions imposed on the position
,at every iteration of the control loop.
This graduate research allows the determination of the -0.2
desired evolution of the position of the throttle, for tracking an
imposed state route, taking into account the exterior
perturbations. -0.4

IV SIMULATION AND RESULTS -0.6


0 0.5 1 1.5 2 2.5 3

In the presence of parametric variations, the linearized Time (sec)


system is modified . To show the advantages of the dynamical
adaptive control strategy and reference model, we introduce Figure 5. Disturbance rejection
parameter deviations and input disturbance P in the linearized
V CONCLUSION
Bode Diagram
0
There are main conclusions that can drawn. It is possible to
-50 implement fast real-time control loops in the Matlab-Simulink
M a g n itu d e ( d B )

System: Prand(:,:,3) environment. During the identification process it was


-100
Gain Margin (dB): 59.4 discovered that there were different dynamics above. The
-150
At frequency (rad/sec): 86.4 characteristics also changes with parametric variations . This
Closed Loop Stable? Yes
implies that at least two different controllers should be used.
-200 This presumption is confirmed during the linear controller
0 1 2 3 4
10 10 10 10 10 design. The discrete linear controller is designed based on the
Frequency (rad/sec) nominal model achieved from the system identification, using
pole placement method. An adaptive scheme has also been
Step Response
1
added as a part of the controller structure. The purpose is to
track and compensate the unknown parameter. In this paper,
we propose electronic throttle control using discrete
System: Prand(:,:,2)
linearization technique, pole placement and adaptive dynamic
A m p litu d e

Peak amplitude: 0.822


0.5 Overshoot (%): 0 control methods for system with unknown parameter. A
At time (sec): 3 particular control is used to obtain linearised model of the
throttle .Second feedback based on a reference model is used
0 to obtain robust properties in terms of parameter variations.
0 0.5 1 1.5 2 2.5 3 Experimental results show good performances obtained with
Time (sec) this method.

Figure 6. Step and Bode response without reference model


REFERENCES

[1] J. P. Barbot,S. Monaco, “Somme comment about linearization under


sampling”, Proc. Conf. CDC’92, Tucson, Arizona, 1992 pp.2392-2397.
[2] J. Chiasson, “Dynamic feedback linearization of the induction motor”,
Bode Diagram
IEEE Trans. Automat. Contr., Vol. 38, pp. 1588-1594, 1993.
50 [3] S.Choi, J.K. Hedric, Robust Throttle Control of Automotive Engines,
Theory and Experiment, Journal of Dynamic Systems Measurement and
Control, 1996, Vol. 118, pp.92-98
0
M a gn itu d e (d B)

[4] M .Honek, W.Slawomir, S.Peter, Control of the Electronic Throttle


System: Nominal Valve Position of SI Engine, International Conference Cybernetics and
-50 Peak gain (dB): 0.374 System: Nominal Informatics, Februarie 10-13, 2010, Vysma Boca, Slovak Republic.
At frequency (rad/sec): 10
Gain Margin (dB): 23.5
[5] K.Nakano, Modelling and Observer-based Sliding-Mode Control of
-100 At frequency (rad/sec): 575
Electronic Throttle Systems, ECTI Transactions on electronical Eng.
Closed Loop Stable? Yes Electronics, and Communications, Vol. Nr. 4, No.1, February, 2006
-150 [6] D. Pavkovic, J.Deura, M.Janszb, Adaptive control of Automative
1 2 3 4
10 10 10 10 Electronic Throttle, Control Eng Practice, 2006, Vol. 14, pp.121-136
Frequency (rad/sec) [7] M.D.Zamfir, “Dinamycal sliding mode strategy for control of series d.c.
motor”, International Conference on applied and Theoretical Electricity,
Step Response (ICATE’96), Craiova, june 6-7, Vol. 3, pp. 153-158.
1.5

1
A m p litud e

System: Nominal System: Nominal


Peak amplitude: 1.05 Final Value: 1
0.5 Overshoot (%): 4.61
At time (sec): 0.102

0
0 0.5 1 1.5 2 2.5 3
Time (sec)

Figure 7. Step and Bode response with reference model

S-ar putea să vă placă și