Control of Binary Input Systems: William HOLDERBAUM
Control of Binary Input Systems: William HOLDERBAUM
|
.
|
|
|
|
|
|
|
\
|
.
|
|
|
|
|
|
|
\
|
.
|
|
|
|
|
|
|
\
|
.
|
|
|
|
|
|
|
\
|
.
|
|
|
|
|
|
|
\
|
.
|
|
|
|
|
|
+ + +
1 2
0
0
0
0
1
0
0
0
0
1
0
0
0
0
0
1
1
1
0
0
1
0
1
0
1
0
1 2 3 1 2 3
0
1
1
1
1
1
2
|
\
|
.
|
|
|
|
|
|
|
\
|
.
|
|
|
|
|
|
i
m
Between two commutations of the system, the input u is a constant vector denoted ( ) Config u
i
. For example if
the number input m= 2 then the ( ) Config u
i
set becomes :
( )
{ } Config u i
i
/ = =
|
\
|
.
|
|
\
|
.
|
|
\
|
.
|
|
\
|
.
|
1 4
0
0
1
0
0
1
1
1
1 2 3 4
III. BOOLEAN CONTROL
The aim of the study is to determine the sequence of ( ) Config u
i
for vector u for the state vector x
to reach a desired fixed state (10), denoted by ( ) x
d
.
Let us consider a vector c which represents a position error vector associated with :
( ) ( ) c = x x
d p
(2)
where
( ) x
p
is the current position of the model in the hyperplane associated with the state space and ( ) x
d
is
the target. We calculate the vector
V
i
associated with ( ) x
i
such as ( )
V x
i i
=
def
, for each configuration
( ) Config u
i
associated with the current position
( ) x
p
( ) ( ) ( )
( )
{ }
V x f x Config
i
u
i
i i p
m
= =
=
def
1 2
,
(3)
Our aim is to determine the configuration ( ) Config u
i
minimizing the angle o between c and
V
i
. We can
illustrate this procedure with the example shown figure 1, where x e9
2
and { } u e 01 ,
2
.
0
V
4
V
1
x
1
c
x
2
x
1
( ) x
p
V
2
o
V
3
( ) x
d
Fig.1 : Control in the state space
In that case, the vector
V
3
has the smallest angle with the vector c .
V
3
corresponds to the control
( ) Config u
3
=
(
1
0
applied to the model.
Interpretation : If the target ( ) x
d
is fixed and constant, then ( ) x
i
can be defined by the equation (eq.
4) in relation to the derivative of c (eq. 2) when the configuration u
i
is applied.
Control of Binary Input Systems
www.iosrjen.org 3 | P a g e
( ) x
d
dt
i
i
=
|
\
|
.
|
c
(4)
We calculate the inner product between c and
V
i
( ( )
V x
i i
=
def
) by using :
( ) c c o , . .cos
V V
i i
= (5)
then c c
c
, .
V
T
i
i
d
dt
=
|
\
|
.
| (6)
or ( )
( ) ( )
d
dt
d
dt
d
dt T
T
T
1
2 2
c c
c
c c
c
.
. .
`
)
=
+
(7)
consequently
( )
|
\
|
.
| =
|
\
|
.
|
|
c
c
c
T
1
2
2
.
d
dt
d
dt
i
i
(8)
The cosine of the angle is given by : ( ) cos
.
,
o
c
c
=
V
V
i
i
(9)
The configuration ( ) Config u
i
is chosen by minimizing angle o . The minimization of angle o is deduced by
the cosine maximum :
( ) ( )
{ }
max cos
i
m
o
= 1 2 ,
(10)
which can be written as :
( )
max
d
dt
d
dt
i
i
m
i
=
|
\
|
.
|
|
|
\
|
.
|
1 2 ; ;
1
2
2
c
c
c
(11)
So the control criterion is expressed by equation 11. By now we can show that the produce error due to this
developed strategy control is bounded.
Proof : The operation principle is such that the deviation with respect to the next point, is less than smallest of
two following terms : the initial deviation and the maximum allowable variations.
The proof is explained with the figure 2.
0
V
4
V
1
x
1
c
x
2
x
1
( ) ( ) x t
p
V
2
o
V
3
( ) ( ) x t
d
( ) ( ) x t
p
+1
A
3
Fig.2 : Control in the state space
In fact the deviation with respect to the next point is stated by A
3
, consequently this deviation is less
than smallest of these two terms : the initial deviation c and the maximum allowable variations. This maximum
Control of Binary Input Systems
www.iosrjen.org 4 | P a g e
allowable variations is calculated for each configuration i { } ( ) where i
m
= 1 2 , and represented by a set of
V
i
which obviously depend of the sampling time. That's why we can say that the error due to the strategy
control is bounded.
IV. TRACKING PREDEFINED TRAJECTORY
Our main objective here is to allow the system to track a desired trajectory. The desired trajectory is
predefined (10) in the state space (figure 3), and it is defined between a starting point M
1
to an end point M
E
.
Fig.3 : Predefined trajectory in the state space
This trajectory is not supposed to be defined in real time, and consequently this definition form will be
settled in order to respect specification sheets. In general case the equation form trajectory is unknown that's why
it is necessary to discrete. So the defined continuous trajectory is changed into file data points. Hence we use the
parametric equation to obtain the expected state trajectory, and which can be represented by a set of point
( )
( )
( )
( )
( )
M x k x k
k
d d
=
1 2
, . This description is interpreted as follows :
( ) ( )
( )
( ) ( )
( )
x x k
x x k
d d
n
d
n
d
1 1
=
=
(12)
where
( )
( ) x k
j
d
( j n = 1, ) represents the k
th
discrete point define on the state variable trajectory by the
parameter | | k E N e c 1, . The sampling period is chosen according to the dynamic system. The whole points
number E depends from sampling period.
In order to minimize the number of switching, we define a domain of dimensional n all around this
parameter curve whose boundary will be denoted the enveloping curve of the desired trajectory. This region
contains the evolution trajectory of the system. We calculate the control u in closed-loop at each instant of
intersection of the state trajectory with the enveloping curve of the desired trajectory (figure 3).
Methodology
The aim of this procedure is to minimize the commutation, and to control system in a bandwidth. For
that we have to determine when the current position (x)
p
is inside or outside of the boundary domain previously
defined by enveloping curve. The location of the current position (x)
p
is necessary to know when a new Boolean
control has to be computed. A new configuration ( ) Config u
i
is calculated when the current position (x)
p
is
outside so that the system converges inside the bandwidth. More precisely the idea is to select the velocity
vectors which cross the envelope just when the current position (x)
p
goes outside it. So we choose between them
the smallest angle with the vector c .
Control of Binary Input Systems
www.iosrjen.org 5 | P a g e
Conversely if the current position (x)
p
is inside any control is computed, but the same control previously
defined (when the current position (x)
p
goes outside the boundary domain) has to be applied. This last
consideration has for objective to minimize the commutation frequency since the same control is kept in the
bandwidth.
The method is summarized here, and presented in several steps in the next section. As to explain to the
beginning of methodology we have to know at all moment the position (x)
p
, in comparison of the discrete desired
trajectory. To realize that, the minimal distance is computed with the discrete state trajectory with respect to
whole points in the first step, but the major drawback is the compute time is too long. To overcome this problem,
an observation window is defined around the predefined state trajectory, and by now the minimal distance have
to be computed by this observation window. If the minimal distance corresponds to window 's boundary then the
window is moved until the minimal distance does not belong to the limit of the window. So the compute time can
be reduced. With the computed minimal distance, a vector orthogonal can be formed, this allows to discern
which vectors cross the domain previously defined by using the inner product. Hence we can choose the
configuration between them and the vector c by the method previously exposed in section III.
1
In the first step : Initialization :
To initialize we search for the M
k
on the discrete state trajectory with a minimal distance from
( ) x
p
. The
distance in 9
n
is calculated by using :
( )
( )
( )
( )
( )
( )
( )
( )
d x k x x k x
k to E
d p
n
d
n
p
min
min =
=1
1 1
2 2
- + + - (13)
k
d
min
: it is the parameter on this curve that corresponds to the minimal distance.
The consideration where k
d
min
is not unique has to be taken account here. That means there are several
minimal distance d
min
due to geometrical form of the desired trajectory. In that case, the choice is performed by
taking the nearest value parameter k
d
min
with the whole number parameter E. This selection is realized in order
to reach an end point M
E
by the predefined trajectory as rapidly as possible, thats why this procedure of choice
has been defined.
An observation window is defined (figure 4), called win around this discrete state trajectory. The size of win
is defined by five points. The point M
k
d
min
which has been found corresponding to k
d
min
is used to form this
window 's boundary as follows :
-center of window M
k
d
min
-lower bound of the window
=
M
k k
k
win d
win min
min
min
with 2
-upper bound of the window
= +
M
k k
k
win d
win max
min
max
with 2
( ) x
p
Observer window
(win)
x
2
x
1
d
min
M
k
win min
M
k
win max
M
k
dmin
Fig.4 : Observation window around state trajectory in state space
Control of Binary Input Systems
www.iosrjen.org 6 | P a g e
For the following sampling time we go to the second step.
2
In the next iteration we repeat the calculation of the minimal distance between the current position and the
observation window. This minimal distance calculated on the observation window allow to reduce the compute
time, because the calculation is only performed on five points and not whole points on this discrete desired state
trajectory.
( ) ( )
( )
( ) ( )
( )
d x k x x k x
k
d p
n
d
n
p
win
min
min =
1 1
2 2
( ) - + + ( ) - (14)
with k k
win win
=
min
to k
win max
( ) x
d
represents the target such that ( ) ( ) x k
d
is the set of states which belongs to the window.
( ) ( ) ( )
{ }
x k x win
d d
e
If the new computed point
min
d
k
M (resp. d
min
with this minimal distance) corresponds to the window's
boundary ( k k
d win
min
min
= or k
win max
); then the window on this curve is moved with respect to previous value
of k
win min
and k
win max
:
k k
k k
win win
win win
min min
max max
= +
= +
1
1
This step by using distance minimal (equ. 14) and shifting window is repeated until the position
k
d
min
(resp. d
min
) on parametrized curve does not belong anymore to the window limits.
( ) x
p
x
2
x
1
d
min
M
k
win min
M
k
win max
M
k
dmin
M
k
dmin
New
Fig.5 : Observation window moving around desired trajectory in state space
In this example (figure 5) d
min
corresponds to the point
wi n
k
M
max
belonging to window's boundary, then
the window is moved as previously defined.
If the new computed point M
k
d
min
(resp. d
min
with this minimal distance) does not correspond to the
window's boundary; then the new boundary limits ( k
win min
and k
win max
) of the observation window are defined
and the algorithm can be go on to the third step.
The new definition of the observation window which is occurred in this stage, is expressed by taking M
k
d
min
as
the center of the window as follows :
-center of window M
k
d
min
-lower bound of the window
=
M
k k
k
win d
win min
min
min
with 2
-upper bound of the window
= +
M
k k
k
win d
win max
min
max
with 2
Control of Binary Input Systems
www.iosrjen.org 7 | P a g e
3
The width between desired state trajectory and the envelope is defined by the distance d
c
. In other words d
c
represents the envelop around the desired state trajectory. The value of d
c
is chosen by the user such that
efficiently about the reduction of switching frequency and accuracy in order to follow desired state trajectory
nearest possible have to be respected.
The calculated distance d
min
is compared with this distance d
c
.
-If d d
c min
( thus we are inside the bandwidth and hence we apply the same control as for the previous sampling
time. Then we return to the second step at the next sampling time.
-If d d
c min
) then we go to the forth step.
4
The vector
V
orth
is formed, defined below, using the minimal distance
( )
min
orth
V
d
k p
M x =
(15)
By using the inner product between
V
orth
and
V
i
, the vectors which cross the envelope are determined.
( ) x
p
x
2
x
1
d
min
M
k
win min
M
k
win max
M
k
dmin
d
c
V
1
V
2
V
4
V
3
V
orth
Fig.6 : Selection vectors which cross the envelope in state space
For this example (figure 6), the vector
V
3
and
V
4
which cross the tube will be selected to go in the fifth step.
5
The configuration u
i
of input vector u is chosen among the remaining vectors which cross the envelope
and we select the best vector by the method described in 3 (Boolean Control).
The algorithm is restarted after the new sampling time in the second step. Repeating the above process we
converge to the regulation domain.
This procedure is illustrated by the following scheme (figure 7):
( ) x
p
x
2
x
1
V
4
V
3
M
E
o
3
o
4
c
d
min
M
k
win min
M
k
dmin
M
k
win max
Fig.7 : Tracking in state space
Here vectors
V
3
and
V
4
cross the tube. The vector
V
4
has the smallest angleo
4
in comparison of o
3
with
the vector c . Therefore the vector
V
4
is chosen for the control in order to converge to the regulation region.
The center of this regulation region is represented by the end point M
E
as described in the next section.
Control of Binary Input Systems
www.iosrjen.org 8 | P a g e
Regulation problem
A domain D
r
is defined around of the end point M
E
(cf. figure 8). The goal of this domain is to
minimize the commutation as previously described method.
v
3
v
4
v
2
v
1
c
( ) x
p
M
E
Regulation region
x
1
x
2
o
3
Fig.8 : Regulation in state space
If
( ) x D
p r
e then the determination of the sequence u is made by using the method presented in 3 (Boolean
Control). For example in figure 8 the vector
V
3
will be applied for the control.
If
( ) x D
p r
e then the same previous computed control, corresponding to
( ) x
p
when it was outside of the
domain D
r
is applied.
Application
This method is applied on a power electronic circuit illustrated in figure 9 modeled by a second order
composed of inductor L1 and capacitor C1, resistances and switching components. This electronic circuit is
supplied by two continuous voltages E1 and E2. The aim is to control the magnetic flux |
L1
in the inductor and
the electrical charge q
C1
in the capacitor in order to get a desired magnetic flux and electrical charge.
E1
E2
S2
S1
R3
R1
R2 L1
C1
Fig.9 : scheme of electronics circuit
Modeling
The classical approach to study a switching electronic circuit is based on the determination of all
possible topologies of the circuit with respect of each possible configurations. In fact there are 2
m
topologies for
the electronic circuit. This approach become very complicated when the number of switching elements increase.
The proposed method, to obtain the model, is to use Bond-Graph methodology (11)(12). This approach allows to
determine an unique state space equation of the electronic circuit under the form defined in (eq. 1). This unique
model represent the association between switching elements and the circuit. The state equation so obtained
includes Boolean terms. The control vector u is consequently straightforward since it contains the Boolean terms
The Bond-Graph is given figure 10 associated with the physical system :
Control of Binary Input Systems
www.iosrjen.org 9 | P a g e
Se:E1
1
1
1
1
0
1
0
R:Ron
R:R1
C:C1
MTF:1/m1
Se:E2
1
R:Ron
MTF:1/m2
I:L1
1
1
R:R3
R:R2
Fig.10 : Bond-Graph model
Ron is the internal resistance of the switches when they are ON, and m
i
( ) i = 1 2 , are Boolean
variables representing the switching logical rules.
m
i
for i =1,2 are the Boolean control input which represents the state of two switches. m
i
= 0 if the
i
th
transistor is in OFF state and m
i
= 1 if it is in ON state.
The mathematical model of the circuit can be represented in this form :
( ) { ( ) x A u x B u u = + with x =
|
\
|
.
|
|
|
L1
C1
The input vector u is composed of the Boolean parameters as
1
2
u
u
u
m
m
=
(
=
(
1
2
as the input vector . The
matrices A and B are expressed as: A
a a
a a
r
=
(
1
11 12
21 22
A
; B
b b
b b
r
=
(
1
11 12
21 22
A
with :
A
r
u u u = + + + + + 1
1 2 1
R3
Ron
R3
R1
R2
Ron
R2
Ron
R2
R1
a u
11 1
1 = + +
|
\
|
.
|
R2
L1
R3
Ron
R3
R1
a
12
=
R2
C1R1
a
21
=
R2
L1R1
a u u u
22 1 2 1
1 = + + +
|
\
|
.
|
1
C1R1
R3
Ron
R2
Ron
R2
Ron
b
11
= E1
R2
Ron
b u
12 1
1 = + +
|
\
|
.
| E2
R2
Ron
R3
Ron
R3
R1
b u
21 2
1 = +
|
\
|
.
| E1
R3
RonR1
R2
Ron
+ E1
R2
RonR1
Control of Binary Input Systems
www.iosrjen.org 10 | P a g e
b
22
= E
R2
RonR1
2
Consider the electrical parameters defined as :
E1 = 12 volts
E2 = 4 volts
Ron = 0.1
R1 = 12
R2 = 2
R3 = 40
L1 = 100mH
C1 = 4.7mF
O
O
O
O
Then the state equation becomes :
( )
x
u
u u
x
u
u
u
r r
+ +
(
+
+
+
(
=
1
866 8000
17 73 1 420 20
1
3464 32000
420 8000 666
1
1 2
1
2
A A
.
.
.
.
35.46
-1.66
240
with
1 1
4 5 420 20
1 2
A
r
u u
=
+ + .
The simulation software MATLAB-SIMULINK has been used for studying the response of the
electrical system.
The sampling time period used for simulation is t
sim
=0.001 sec. This system in closed loop can be represented
by :
c
x
d
u
x
q
=
|
\
|
.
|
|
L1
C1
System +
-
Control
Fig.11 : Control scheme of electrical system
Simulation:
The target ( ) x
d
or M
E
, and the predefined trajectory (as a linear trajectory) have been chosen inside the
reachable domain, with :
( ) x
q
d
d
=
=
=
|
\
|
.
|
|
L1
C1
3
0.4
Wb
C
Figure 12 represents the trajectory in the state space.
0 0.5 1 1.5 2 2.5 3 3.5
-0.05
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
|
L1
q
C1
Fig .12 : Trajectory in the state space
Control of Binary Input Systems
www.iosrjen.org 11 | P a g e
The time responses of electrical charge q
C1
in the capacitor and the magnetic linkage |
L1
in the inductor are
represented figure 13 :
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35
0
1
2
3
4
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35
0
0.2
0.4
0.6
t
t
|
L1
q
C1
Fig.13 :Time variations of the magnetic linkage |
L1
and the electrical charge q
C1
.
Figure 14 shows the variation of input variables :
0 0.05 0.1 0.15 0.2 0.25 0.3
0
0.5
1
0 0.05 0.1 0.15 0.2 0.25 0.3
0
0.5
1
t
t
u
1
u
2
Fig.14 : Time variation of the input Boolean controls
The simulation results show that the predefined tracking trajectory, obtained with the Boolean control strategy.
V. A REAL TIME TRACKING TRAJECTORY
In this section we present a control law allowing the system to track a real time trajectory (13) in the state space.
We use the same principle for the control law design as defined previously in III (Boolean Control).
We define a n -dimensional space all around the desired trajectory ( ) ( ) x
d
t so as to minimize the switching.
The control strategy in closed-loop consists in determining the configuration u at each time of intersections
with the domain according to the point ( ) ( ) x
d
t . The minimization of the commutation frequency is achieved
because the same control is kept in the domain. The details of this procedure is outlined below.
Procedure
The time of switching is determined by suitable error boundaries. When the current position
( ) ( ) x t
p
goes
outside the domain, the next configuration ( ) Config
i
u of input vector u is determined by the methodology
discussed above in III (Boolean Control). The algorithm is explained as follows :
+ In the first case, a domain D
s
is defined around the point ( ) ( ) x
d
t .
+ We study if the point
( ) ( ) x t
p
is inside the domain D
s
:
Control of Binary Input Systems
www.iosrjen.org 12 | P a g e
- if
( ) x D
p s
e , then the determination of the sequence u is made by using the method presented in section III
(Boolean Control).
- if
( ) x D
p s
e , then the same previous step control corresponding to
( ) x
p
when it was outside of the domain
D
r
is applied.
The same algorithm is restarted at each sampling time. This procedure is illustrated by the following example
(figure 15):
( )
V
4
t
x
2
x
1
( ) ( ) x t
p
( )
V
1
t ( )
V
2
t
( )
V
3
t
( ) ( ) x t
d
( ) D t
s
( ) ( ) 1 x t
p
+
( )
V 1
1
t +
( )
V 1
2
t +
( )
V 1
3
t +
( ) ( ) 1 x t
d
+
( ) D t
s
+1
( )
V 1
4
t +
Real time
Trajectory
Fig.15: Tracking in state space
In this example at the instant t the vector
( )
V
1
t is chosen and at t+1 the vector
( )
V
2
t +1 will be
selected.
Application
The proposed control law is illustrated on the same electronic circuit as described previously (figure 9).
Simulation:
The desired vector ( ) x
d
has been simulated by two sinusoidal waveforms such as :
( )
( )
x
x t
x t
d
d
d
=
= +
= +
|
\
|
.
|
|
|
1
2
3
1
2
6
04
1
5
3
sin .
. sin .
These equations simulate the vector ( ) x
d
in real time. Note that frequency of the waveform must respect the
bandwidth of the system.
Figure 16 shows the trajectory in the state space
2.2 2.4 2.6 2.8 3 3.2 3.4 3.6
0.15
0.2
0.25
0.3
0.35
0.4
0.45
0.5
0.55
0.6
0.65
|
L1
q
C1
Fig .16 : Trajectory in the state space
Control of Binary Input Systems
www.iosrjen.org 13 | P a g e
The magnetic linkage |
L1
in the inductor and the electrical charge q
C1
in the capacitor, with the
corresponding desired trajectory are represented figure 17 and figure 18.
0 0.5 1 1.5 2 2.5
2
2.5
3
3.5
4
0 0.5 1 1.5 2 2.5
2
2.5
3
3.5
4
x
d1
|
L1
t
t
Fig .17 : Time variation of the magnetic linkage |
L1
compared to the desired trajectory
0 0.5 1 1.5 2 2.5
0
0.2
0.4
0.6
0.8
0 0.5 1 1.5 2 2.5
0
0.2
0.4
0.6
0.8
t
t
x
d 2
q
C1
Fig .18 : Time variation of the electric charge q
C1
compared to the desired trajectory
Figure 19 represents the Boolean control :
Control of Binary Input Systems
www.iosrjen.org 14 | P a g e
0 0.1 0.2 0.3 0.4 0.5
0
0.5
1
0 0.1 0.2 0.3 0.4 0.5
0
0.5
1
t
t
u
2
u
1
Fig .19 : Boolean controls
VI. CONCLUSION
In this paper a method is proposed for the control design of systems with Boolean inputs. The problem
of regulation and tracking has been discussed. The proposed control law has been applied in the initially time to
follow a predefined trajectory from the initial state and the second time to track a real time trajectory.
In both cases some remarks have been presented about the domains. If these domains are very large
then we reduce the commutation numbers but we have low precision. Conversely if these domains are very
small, the commutations are enhanced and we have high precision. The domains can be defined by the
specification sheets.
The Bond-Graph approach has been used in this paper. One of this major advantage for Boolean control
is to represent the model by a unique state equation including control vector. Indeed this control vector is
characterized by the Boolean value. The other technique has drawback to obtain a multi-model representation,
and it is less practicable when the number of switches items rises.
The improvements of this approach presented in this paper are :
To consider whole model which give a direct link for Boolean control system.
To reduce the switching frequency in order to have energy low in the commutation.
The same example has been used in the two cases to illustrate the control scheme. Simulation curves
have shown that the vector
( ) x
p
reaches the desired value ( ) x
d
.
The proposed strategy seems a few similar with the Direct Self Control (DSC). However the DSC
technique is applied to power converter in order to control the electrical motor, while our method is considered
as a whole method for Boolean input system as well as hydraulic system (valve control), chemical processes and
so on. In addition a control criterion has been developed which minimize angle, and the news technique for
tracking trajectory (predefined and real time trajectory) has been presented.
Currently new studies are performed in the laboratories. Their goal are to define a minimization
frequency switching criterion and other part are to compare different strategies of Boolean control presented in
introduction.
REFERENCES
1). Seguier G., Labrique F. (1989).
"Les convertisseur de l'lectronique de puissance", Editions technique et documentation Lavoisier.
2). S Nonaka et Y Neba (1987).
"Analysis of a P.W.M GTO current source inverter fed induction motor drive system", IEEE
Transaction on Industry Applications, Vol. 23, No 2.
3). H Sira-Ramirez(1989).
"Sliding regimes in general non linear systems : a relative degree approach", Int. J. Control, Vol. 50, No
4, pp.1487-1506.
4). J E.Slotine(1984).
"Sliding controller for non-linear systems", Int. J. Control, Vol. 40, No 2, pp.421-434.
5). Abadie V (1994).
Control of Binary Input Systems
www.iosrjen.org 15 | P a g e
"Commande des systmes continus a entres binaire. Application aux machines electriques", Thesis of
UNIVERSITE DES SCIENCES ET TECHNIQUE DE LILLE No 1310. In french.
6). Habetler T.G., Profumo F., Pastorelli M., Tolbert L.M.(1992).
"Direct torque control of induction machines using space vector modulation", IEEE Transaction on
Industry Applications, Vol. 28, No 5.
7). W .Leonhard, (1985)
"Control of electrical drives", Springer, Berlin
8). Depenbrock M. (1988).
"Direct self control of inverter fed induction machines", IEEE Transaction on Power Electronics, Vol.
3, No 4.
9). Abadie V. Dauphin-Tanguy G (1993).
"Opened Loop control of switching linear system", Journal of the Franklin Institute, Vol. 330, No 5,
pp.799-813.
10). Holderbaum W., Dauphin-Tanguy G ., Borne P (1998).
"Boolean control for linear system", I.S.I.A.C International Symposium on Intelligent Automation and
control, Wac'98 Anchorage , USA .
11). Borne P., Dauphin-Tanguy G., Richard J.P., Rotella F., Zambettakis I.(1992).
"Modelisation et identification des processus" Tome 1. Collection : mthode et pratique de l'ingnieur.
Edition Technip.
12). Ducreux J.P., Castelain A., Dauphin-Tanguy G. and Rombaut C. (1992),
"Power electronics and electrical machines modelling using Bond-Graph ", IMACS Transactions on
"Bond-Graph for Engineers "( eds. Dauphin-Tanguy G. and Breedveld P. ) Elsevier, NewYork
13). Holderbaum W., Dauphin-Tanguy G ., Borne P (1998).
"Tracking control problem for switching linear system", CESA'98 IEEE/SMC Conference Hammamet
(Tunisia).