International Journal of Automation and Computing 6(3), August 2009, 267-276

DOI: 10.1007/s11633-009-0267-4
A New Energy Optimal Control Scheme for a Separately
Excited DC Motor Based Incremental Motion Drive
Milan A. Sheta
1
Vivek Agarwal
2,
Paluri S. V. Nataraj
1
1
Systems and Control Engineering, Indian Institute of Technology Bombay, India
2
Department of Electrical Engineering, Indian Institute of Technology Bombay, India
Abstract: This paper considers minimization of resistive and frictional power dissipation in a separately excited DC motor based
incremental motion drive (IMD). The drive is required to displace a given, xed load through a denite angle in specied time, with
minimum energy dissipation in the motor windings and minimum frictional losses. Accordingly, an energy optimal (EO) control
strategy is proposed in which the motor is rst accelerated to track a specic speed prole for a pre-determined optimal time period.
Thereafter, both armature and eld power supplies are disconnected, and the motor decelerates and comes to a halt at the desired
displacement point in the desired total displacement time. The optimal time period for the initial acceleration phase is computed so
that the motor stores just enough energy to decelerate to the nal position at the specied displacement time. The parameters, such
as the moment of inertia and coecient of friction, which depend on the load and other external conditions, have been obtained using
system identication method. Comparison with earlier control techniques is included. The results show that the proposed EO control
strategy results in signicant reduction of energy losses compared to the existing ones.
Keywords: Energy optimal control, speed prole, incremental motion drive (IMD).
1 Introduction
Certain applications such as robots, automatic machine
tools, and electric cranes require that a load is moved
from one location to another, where it is held for a cer-
tain time duration, before the next motion command is is-
sued. An electric drive that facilitates the repeated step-
ping of the load through incremental steps at specied fre-
quency or time intervals is called an incremental motion
drive (IMD)
[1,2]
.
Any motor can be used in an IMD application. How-
ever, the performance of the IMD will be determined by
the type of motor used. DC motors are a strong candi-
date for an IMD due to their simple control requirements,
precise control over a wide bandwidth, and fast response
features. Among the DC motors, a permanent magnet DC
(PMDC) motor may be considered for an IMD application
due to its high power density, high eciency, and better
controllability
[3]
. However, the magnetic eld in a PMDC
motor cannot be varied, ruling out the operation in the
eld weakening mode. This puts a serious limitation on the
use of a PMDC motor in an IMD. A series or shunt con-
nected DC motor can be used, but has limitations because
independent control of the armature and eld circuit is not
possible. Therefore, a separately excited DC motor is by
far the most eligible candidate for an IMD application.
In the last few decades, AC motors have also become pop-
ular for use in electric drives due to the availability of high
power inverters and powerful processors (e.g., digital signal
processors), the latter being a requirement for the usually
complex and non-linear control of the AC drives. In fact, it
is due to this control complexity that the popular AC mo-
tor control methods like the vector control try to emulate
the DC motor control characteristics. Among the AC mo-
Manuscript received February 18, 2008; revised December 24, 2008
*Corresponding author. E-mail address: [email protected]
tors, permanent magnet synchronous motor (PMSM) drives
are popular due to their high power to weight ratio, large
torque to inertia ratio and high eciency. However, one
of the major disadvantages of a PMSM is the presence of
torque ripple
[4]
. This ripple may particularly be a matter
of concern in applications requiring precise control. Fur-
ther, a PMSM with damper windings consumes more en-
ergy (i.e., more losses) during the transient, start-up phase
as the permanent magnets exert a braking torque
[5]
. An-
other problem with a PMSM is the possibility of its per-
manent magnets getting de-magnetized during an electric
fault or in due course of their life. This will hamper the
performance of the PMSM.
In view of the drawbacks discussed above in addition to
the higher cost and losses in the power stage and higher
control complexity, the AC drives are usually preferred for
high power applications only. For low power rating applica-
tions, including the battery operated autonomous systems
(e.g., robots), a DC motor drive is still considered a better
alternative
[6,7]
.
This paper is concerned with a DC motor based IMD.
As stated previously, among the DC motors, a separately
excited DC motor is a highly suitable conguration for an
IMD, in which the eciency can be enhanced and better
controllability can be achieved with the help of variable
eld ux, which can be independently adjusted. If the mo-
tor is connected to a constant load, then for achieving an
overall high eciency, an appropriate combination of the
armature and eld power can be used. Thus, the motor
can be operated with an optimized reduced power
[6,8,9]
.
Minimization of the energy losses not only improves the
eciency of the drive but also the performance of the drive.
For example, it causes an indirect reduction in the arma-
ture current. From a practical point of view, this has several
advantages. The reduced armature current eliminates the
possibility of saturation in the power ampliers and convert-
268 International Journal of Automation and Computing 6(3), August 2009
ers, thereby inhibiting nonlinear eects in the systems
[10]
.
A reduced armature current also leads to reduced losses
in the control circuit. Further, a reduction in the arma-
ture current also yields a reduction in the I
2
R losses and
magnetic ux to the necessary minimum levels
[11]
. This in
turn results in a reduction of the core losses and stray-load
losses
[12]
. Therefore, a motor with lower power rating can
be used for a given application.
In incremental motion control, the driving motor mostly
remains in the dynamic state thus consuming a large
amount of energy
[1214]
. Therefore, the resistive and fric-
tional losses greatly depend on the speed prole through
which the motor is rotated. An elegant energy optimal (EO)
control scheme was proposed by Trzynadlowski
[12]
to min-
imize the energy losses in the IMD. In this scheme, using
the calculus of variations, an optimal parabolic or trape-
zoidal speed prole is generated, which minimizes the en-
ergy losses. The energy losses corresponding to this opti-
mized speed prole are further minimized with respect to
the eld current and displacement time. One of the draw-
backs of this scheme is that if the obtained optimal value of
eld current is higher than the specied allowable maximum
value, then it is set at its maximum value, i.e., it is assumed
that the energy losses are minimum at the maximum eld
current value. Actually, it may not be true because any
value in the given range of eld current may correspond to
minimum losses. While optimizing the losses with respect
to the displacement time, if the obtained displacement time
value is less than the specied maximum value, then the ob-
tained value itself is used. In applications where the load
must stop exactly at the specied displacement time, this
may not be applicable. Another drawback of this scheme
is that only resistive losses are minimized with respect to
the speed prole, while frictional losses are not considered.
It is shown in a subsequent section of this paper that if
frictional losses are also considered, better performance in
terms of losses can be achieved. The disadvantages, dis-
cussed above, limit the application of the scheme proposed
by Trzynadlowski
[12]
.
This paper presents a new EO control scheme by tak-
ing into account the frictional losses for the minimization
of overall energy losses and optimal use of kinetic energy
stored in the motor. The considerations are limited to con-
stant load-torque and possible angular displacement with
respect to specied displacement time and maximum al-
lowable speed. Both the resistive and frictional losses are
considered. In the proposed scheme, the motor is rst accel-
erated as per a specic speed prole up to a pre-computed
maximum speed for a pre-computed intermediate displace-
ment time. Subsequently, both armature and eld power
supplies are disconnected. The motor decelerates and comes
to a halt exactly at the specied displacement time. The
specic EO speed prole through which the motor must be
rotated during the initial acceleration phase is obtained us-
ing the theory of calculus of variations. The parameters
such as the coecient of friction, moment of inertia, etc.,
which depend on the type of load and the external condi-
tions in which the drive is required to work, are obtained
using system identication method, which is both easy and
accurate. It is assumed that the load and external environ-
mental conditions, for which the system identication has
been done, do not vary signicantly during an operation
cycle.
A modication to Trzynadlowski

s scheme
[12]
is also in-
cluded, which takes into account the frictional losses. It is
shown that this modication improves the performance of
the drive as compared to the original scheme.
The rest of this paper is organized as follows. The ba-
sic concept behind the proposed scheme and the determi-
nation of the EO speed prole are discussed in Section 2.
Section 3 discusses the implementation of the proposed
control scheme. Section 4 presents the details of the ear-
lier EO control
[12]
, modied to take into account the fric-
tional losses. Discussion and comparison of various control
schemes is included in Section 5. Experimental results cor-
responding to the proposed scheme and comparison with
the simulation results are included in Section 6. Finally,
the main conclusions of this work are given in Section 7.
2 The proposed energy optimal control
scheme
In the proposed scheme, the motor is accelerated up to a
speed (t
d
) for a pre-computed optimal time period [0, t
d
]
with a pre-computed EO speed prole (

(t)) to minimize
the total losses (see Fig. 1 (a)).

(t) can be determined

by minimizing the total losses with respect to the speed
(t) by using the theory of calculus of variation along with
the given constraints
[15]
. However, the eld current is still
an unknown quantity. To determine an optimal value of
the eld current, the total losses corresponding to

(t) are
minimized with respect to the eld current.

(t) is implemented by varying the armature voltage

E
a
(t), while eld current I
f
is set at its optimal value that
depends on the pre-computed disconnecting speed (t
d
),
optimal intermediate displacement time t
d
, electrical and
mechanical parameters of the motor like armature and eld
resistances, inductance of the armature circuit, moment of
inertia, load torque, viscous friction, etc. The power sup-
plies are disconnected at time instant t
d
, thereby forcing
armature current i
a
(t) and I
f
to zero. Therefore, after t
d
,
the electrical losses are zero.
The kinetic energy stored in the motor up to instant t
d

takes the motor to its nal position. As the stored energy

is spent, the motor decelerates and comes to a halt at the
desired displacement
d
, in the desired displacement time
t
d
as shown in Fig. 1 (a). The time interval [t
d
, t
d
] depends
on the motor speed (t
d
), at which the supplies are dis-
connected. It also depends on the motor time constant,
connected load torque, and displacement
d

d
.
The analysis to get the overall EO speed prole (

(t)
and
dec
(t)) is done in two steps.
The rst step includes determination of
d

d
and t
d
for
(t
d
) [
min
,
max
]. The second step includes the deter-
mination of

(t) for the disconnecting speed (t

d
).

(t)
is determined such that it minimizes the total losses by
considering it as an isoperimetric problem of calculus of
variation with integral constraint (
od
), initial condition
(

(0) = 0), and nal condition (

(t
d
) = (t
d
)). These
two steps are described in details next.
M. A. Sheta et al. / A New Energy Optimal Control Scheme for a Separately Excited DC Motor 269
(a)
(b)
Fig. 1 (a) EO speed prole of an IMD using proposed technique.
t
d
is the time instant at which power supplies are disconnected,
after which the motor decelerates and comes to a halt exactly at
time instant t
d
; (b) EO speed proles at dierent disconnecting
speeds
Step 1. The motor torque equation is
J
.
+f +T
l
= k
m
i
f
i
a
(1)
where k
m
is the motor constant, i
a
, i
f
are the armature and
the eld currents, J is the equivalent mass moment of iner-
tia at the motor shaft, f is the coecient of viscous friction,
and T
l
is the constant load torque. After t
d
[0, t
d
], i
a
and
i
f
are zero because the power supplies are disconnected at
t
d
. Using this,
dec
(t) for (t
d
) [
min
,
max
] and t t
d

can be determined by solving (1) as

dec
(t) =
T
l
f
+
_
(t
d
) +
T
l
f
_
e

(tt
d

(2)
where = J/f. Substituting
dec
(t) = 0 at t = t
d
in (2)
and solving for t
d
yields
t
d
= t
d
+ ln
_
T
l
f
(t
d
) +
T
l
f
_
. (3)
The corresponding displacement required during the inter-
val 0 to t
d
is

od
=
d

d

d
(4)
which is the area covered by the speed versus time curve,
from 0 to t
d
as shown in Fig. 1 (a). In (4),
d

d
can be
determined by integrating (2) over the time interval [t
d
, t
d
]
and is given by

d
=

f
_
_
T
l
+f (t
d
)
_
_
1 e

t
d

_
T
l
t
d

_
. (5)
The constants (t
d
),
d

d
, and t
d
determined from the
above analysis are used as the nal condition, integral con-
straint, and integral limit, respectively, in that order.
Step 2. To determine

(t) and the optimal value of

eld current, I

f
for t [0, t
d
], consider the energy losses
supplied by the source
W(t, ,
.
, i
f
) =
_
t
d

0
(
2
f +i
2
a
R
a
+i
2
f
R
f
)dt (6)
where R
a
and R
f
represent the armature and the eld cir-
cuit resistances, respectively. Using (1) and (6), we get
W(t, ,
.
, i
f
) =
_
t
d

0
_

2
f +
_
J
.
+f +T
l
k
m
i
f
_
2
R
a
+i
2
f
R
f
_
dt. (7)
The EO speed prole can now be determined by minimiz-
ing (7) within the constraints
od
, (0), and (t
d
). This
can be posed as an isoperimetric problem as shown below.
Isoperimetric problem formulation. Consider the
problem of determining a curve that minimizes a given func-
tion W(t, ,
.
, i
f
) with respect to within an integral con-
straint and the given initial and nal values
[15]
:
minimize W(t, ,
.
, i
f
) =
_
t
d

0
f(t, ,
.
, i
f
)dt (8)
with respect to (t), with initial and nal conditions (0) =
0, (t
d
) [
min
,
max
] subject to the constraint

od
=
_
t
d

0
g(t,
.
, i
f
)dt (9)
where
f(t, ,
.
, i
f
) =
2
f +
_
J
.
+f +T
l
k
m
i
f
_
2
R
a
+i
2
f
R
f
(10)
g(t, ,
.
, i
f
) =

. (11)
In order that

(t) be the solution of (7), it is necessary

that it should be an extremal of the following integral
_
t
d

0
(f(t, ,
.
, i
f
) +g(t, ,
.
, i
f
))dt (12)
for a certain constant , called the Lagrange multiplier.
To determine

(t) for the extremal solution of (11), the

necessary condition is
[3]

(f +g) +
d
dt
_

(f +g)
_
= 0. (13)
Equation (13) can be simplied into the following form:
m
2
..

m
1

1
= 0 (14)
270 International Journal of Automation and Computing 6(3), August 2009
where m
2
= R

a
J
2
, m
1
= f + R

a
f
2
,
1
= (2R

a
fT
l
+ )/2,
and R

a
= R
a
/(k
2
m
i
2
f
). In time-domain, the solution of (14)
is obtained as

(t) =

1
(cosh
n
t 1)+
.
(0)
n
m
2
sinh
n
t
m
2

2
n
(15)
where
1
and
.
(0) are obtained by simultaneous solution
of (16) and (17), given below:
.
(0) =
m
2

2
n
(t
d
) +
1
(1 cosh
n
t
d
)

n
m
2
sinh
n
t
d

(16)

1
=

n
_

od
m
2

2
n
+
.
(0)m
2
(1 cosh
n
t
d
)
_
sinh
n
t
d

n
t
d

(17)
where
n
=
_
m
1
/m
2
.
By taking the Laplace transform of (15), the transfer
function for a step input V
i
is given by

(s)s
V
i
=
.
(0)
V
i
m
2
s +

1
V
i
m
2
(s
2

2
n
)
. (18)
Several values are possible for (t
d
) [
min
,
max
] as
shown in Fig. 1 (b). Therefore, interval [
min
,
max
] is di-
vided into a set of equidistant points. EO speed proles
(

i
(t)) are determined at each point, and the correspond-
ing losses are compared. This comparison gives the overall
EO speed prole

(t), the optimal supply disconnecting

speed

ods
(t
d
) and the corresponding overall optimal value
of the eld current, I
f

. A systematic way to determine

the above unknowns is given below.
Algorithm 1. The procedure to determine
ods

(t
d
)
[
max
,
min
], I

f
[I
f min
, I
f max
] and

(t) is as follows:
Step 1. Divide the interval [
min
,
max
] into a set of
p equidistant points given by
mi
, i = 1, 2, , p and set
i = 1.
Step 2. If i > p, then go to Step 6, else determine t
d

and
od
from (3) and (4) for (t
d
) =
mi
.
Step 3. Determine

i
(t) from (15) for the initial and
nal conditions (0) = 0, (t
d
) =
mi
and integral con-
straint
od
.
Set the motor

s eld current as an unknown variable,

which is required to be optimized.
Step 4. Optimal value of the eld current I

f
for the
corresponding
mi
and

i
(t) can be determined by using
Euler

s explicit numerical integration method. For this pur-

pose, let T be the sampling time. Then, (8) can be written
in the following discrete form
W(nT) =T
_
f(nT, (nT),
.
(nT), i
f
, )

+
+W((n 1)T) (19)
where W(0) = 0, nT {0, T, 2T, , mT} and
mT = t
d
. The objective is to determine I

f

[I
f min
, I
f max
], such that W(mT) is minimum. This can
be done as follows:
Step 4.1 Divide the interval [I
f min
, I
f max
] into a set of
q equidistant points given by I
fr
, r = 1, 2, , q and set
r = 1.
Step 4.2 If r > q, then go to Step 4.4, else determine
W
r
= W(mT) for I
fr
and
mi
.
Step 4.3 Set r = r + 1 and go to Step 4.2.
Step 4.4 Set I

fi
{I
f1
, I
f2
, , I
fq
} corresponding to
W

i
= min{W
1
, W
2
, , W
q
}.
Step 5. Set i = i + 1 and go to Step 2.
Step 6. Determine the overall optimal value of
eld current I

f
{I

f1,
I

f2,
, , I

fp,
},

ods
(t
d
)
{
m1
,
m2
,
m3
, ,
mp
} and t
d
corresponding to
W

= min{W

1
, W

2
, W

3
, , W

p
}.
Step 7. From (15), determine

(t) corresponding to
I

f
and

ods
(t
d
).
3 Implementation of the proposed con-
trol scheme
After determination of the overall EO speed prole

(t), the motor needs to track this prole up to the op-

timal intermediate displacement time t
d
. Because of the
moment of inertia, viscous friction, load torque, armature
resistance, and inductance, the motor has a certain set-
tling time and steady-state error that causes deviation from

(t), if the motor is fed with the exact voltage equivalent

of

(t) as shown in Fig. 2, where V

EOS
(t) corresponds to
voltage equivalent of

(t), and T
M
(s) is the motor trans-
fer function. Therefore, it is necessary to improve the set-
tling time and to make the steady-state error zero, which
is done with the help of a proportional-integral-derivative
(PID) controller as discussed later in this section.
Fig. 2 Deviation from the desired EO speed prole due to set-
tling time and steady-state error of the separately excited DC
motor
After improving the motor transient and steady-state re-
sponse, the motor is fed with the armature voltage equiv-
alent to

(t). This variation is achieved at the output

of the tracking controller from the constant voltage sup-
ply as shown in Fig. 3. Note that both the armature and
eld supplies are disconnected at time instant t
d
. Deter-
mination of the PID controller C
PID
(s) and the tracking
controller C
T
(s) is shown next. It must be pointed out
that any other controller that makes the steady-state error
zero and improves the transient response can also be used
instead of the PID controller.
Fig. 3 Block diagram of the control system used for implement-
ing and testing the proposed scheme
M. A. Sheta et al. / A New Energy Optimal Control Scheme for a Separately Excited DC Motor 271
3.1 Determination of PID controller
CCC
PID PID PID
(sss)
To track the overall EO speed prole, the drive must have
a lower settling time than the given displacement time and
zero steady-state error. This subsection discusses the pole
cancelation technique to determine the PID parameters.
Let the transfer function of the PID controller be given
by
C
PID
(s) = k
p
(1 +
1

i
s
+
d
s) (20)
where k
p
is the proportional gain, and
i
,
d
are the integral
and derivative time constants, respectively.
The PID constants are determined such that the actual
poles of the motor transfer function T
M
(s) are canceled out
by proper selection of
i
and
d
, i.e., the second-order sys-
tem is reduced to a rst-order system. The proportional
gain k
p
is chosen to improve the settling time.
Let the motor transfer function be given by
T
M
(s) =
(s)
E
a
(s)
=
K
M
JL
a
s
2
+ (R
a
J +L
a
f)s + (R
a
f +K
M
K
B
)
(21)
where R
a
and L
a
are the resistance and inductance respec-
tively of the armature circuit, and K
B
= K
M
= k
m
I
f
. To
cancel the poles of T
M
(s),
i
and
d
should be chosen as
below:

i
=
R
a
J +L
a
f
R
a
f +K
M
K
B
,
d
=
JL
a
R
a
J +L
a
f
. (22)
3.2 Determination of tracking controller
CCC
TTT
(sss)
After
i
and
d
are determined, T
MPID
(s) (see Fig. 3) is
given by
T
MPID
(s) =
(s)
V
EOS
(s)
=
K
M
k
p

i
(R
a
f +K
M
K
B
)s +K
M
k
p
. (23)
In (23), k
p
is an adjustable parameter that can be used
to track

(t) satisfactorily.
The overall transfer function of the control system shown
in Fig. 3, is given by
(s)
V
i
(s)
= C
T
(s)T
MPID
(s). (24)
Since the eect of the poles of T
MPID
(s) on C
T
(s) is
insignicant, T
MPID
(s) can be taken as unity. By taking
Laplace transform of

(t), C
T
(s) =

(s)/V
i
(s) is deter-
mined. Therefore, from (24), we have
C
T
(s)

=
(s)
V
i
(s)

(s)
V
i
(s)
. (25)
Equation (25) indicates that the speed of the motor, us-
ing this control scheme, can be approximately tracked as
per the overall EO speed prole

(t). It must be pointed

out that if switch S is kept ON continuously (see Fig. 3),
even beyond t
d
, the system will get unstable. Both the
controllers (C
T
(s) and C
PID
(s)) must be reset during the
interval [Nt
d
, Nt
d
] (N = 1, 2, ), before the commence-
ment of the next cycle.
4 EO control
[12]
, modied to account for
frictional losses
In this section, the analysis of Trzynadlowski

s scheme
[12]
is expanded to take into account the frictional losses, which
are not considered in the original scheme. Accordingly, the
EO speed prole of the drive is obtained by considering the
frictional plus armature and eld resistive losses.
The EO speed prole is determined by minimizing (7)
with respect to (t) for the time period 0 to t
d
. The de-
sired EO speed prole

FL
(t) should satisfy the integral
constraint

d
=
_
t
d
0

FL
(t)dt (26)
which is subject to the initial and nal conditions

FL
(0) =
0 and

FL
(t
d
) = 0, respectively.

FL
(t) can be determined
by substituting

(t
d
) = 0,
od
=
d
, and t
d
= t
d
in (15),
(16), and (17).
After determining

FL
(t), the optimal value of the eld
current I

FL
is calculated by minimizing the total energy
losses for

FL
(t) with respect to the eld current.
For implementation of this modied control scheme, the
same control system, as shown in Fig. 3, is used; the only
dierence being that now both the controllers remain active
from 0 to t
d
. C
PID
(s) is determined for I

FL
in a similar
manner as discussed in Section 3, while C
T
(s) is determined
by taking the Laplace transform of

FL
(t) for a step input
V
i
(s). C
T
(s) is given by
C
T
(s) =

FL
(s)
V
i
(s)
. (27)
5 Discussion and comparison of various
control schemes
To highlight and study the improvements obtained with
the proposed technique over earlier techniques, the follow-
ing control schemes were implemented:
1) EO control
[12]
;
2) EO control
[12]
, modied to take into account frictional
losses;
3) Proposed EO control scheme.
In addition to these, a conventional control scheme, the
state feedback technique, is also implemented, which does
not involve any optimization. The idea is to show the im-
portance of optimization, without which the drive incurs
huge losses.
A separately excited DC motor with the following speci-
cations was considered: P
rated
= 250 W, E
a
(rated) = 250 V,
R
a
= 46 , R
f
= 1000 , I
f min
= 0.07 A, I
f max
= 0.15 A,

min
= 180 rad/s,
max
= 320 rad/s. A given load of
0.1239 Nm (T
l
) is required to be displaced by 565.48 rad
(
d
), i.e., 90 rotations in 3 s (t
d
).
Complex and time-consuming computations can be done
for the determination of the other unknowns like equiv-
alent moment of inertia, frictional coecient, etc. How-
ever, this may not give any additional advantage since the
272 International Journal of Automation and Computing 6(3), August 2009
motor performance signicantly depends on the environ-
ment in which the drive is required to work. The perfor-
mance also depends on the type of load connected to the
drive. The easiest and accurate way is to identify the sys-
tem (system identication) in the same environment and
determine the unknowns. For the given system, a second-
order system was identied, and the following parameters
were obtained: J = 0.00081 kgm
2
, f = 0.00021734 Nms.,
K
B
= K
M
= 5.8104 I
f
Nm/A
2
, L
a
is neglected, where K
B
and K
M
are the back electromotive force (emf) and torque
constants, respectively, of the motor.
The comparison of losses, armature current variation, po-
sition, and speed proles corresponding to these schemes,
are included later in this section.
5.1 EO control
[12]
The earlier, EO control scheme
[12]
is discussed in this
subsection. The control system shown in Fig. 3 remains
the same. The only dierence is that both the PID and the
tracking controllers are operated from 0 to t
d
. Similarly, the
motor armature and eld supplies also remain ON during
[0, t
d
], i.e., the power supplies are no longer switched OFF
at t
d
, unlike the case with the proposed scheme.
The procedure to determine the EO speed prole and
tracking controller C
T
(s) for the given motor specications
is given below:
Step 1. For a given
d
,
max
, and t
d
, check the inequality
given by

d
>
2
3

max
t
d
. (28)
For the drive specications given, (28) is not satised.
Therefore, parabolic EO prole is applicable and can be
determined.
Step 2. Determine optimal values of the eld current
I

f(p)
and displacement time t

d(p)
from the equations given
below
[12]
:
I

f(p)
=
_
2
R
a
R
f
_
T
l
k
m
_
2
_1
4
(29)
t

d(p)
=
_
12
_
J
d
T
l
_
2
_1
4
. (30)
Step 3. If I

f(p)
i
f max
and/or t

d(p)
t
d
, then set their
corresponding values to maximum.
In the present case, I

f(p)
= 0.0804 A and is less than
i
f max
(0.15 A). Therefore, the inequality is not satised.
However, t

d(p)
is obtained as 3.5786 s, and it satises the
inequality, so t

d(p)
is taken as 3 s.
Step 4. Using the information obtained in Step 3, deter-
mine the parabolic EO speed prole

(p)
(t) from the equa-
tion given below:

(p)
(t) =
6
d
(t

d(p)
)
3
(t

d(p)
t t
2
). (31)
Equation (31) gives

(p)
(t) = 125.637(3t t
2
). (32)
Step 5. By taking Laplace transform of (32) for a given
step input V
i
(s) (0100 V), C
T
(s) is given by
C
T
(s) =

(s)
V
i
(s)
=
3.77s
2
2.5123s
s
3
. (33)
It may be noted that the EO speed prole given by (32)
is the prole that minimizes only the resistive losses in the
armature and eld windings. The frictional losses are not
minimized.
To determine the PID controller parameters, consider the
motor transfer function calculated for I

f(p)
= 0.0804 A.
(s)
E
a
(s)
=
12.5
s + 6.133
. (34)
Using (22),
d
and
i
are obtained as 0 and 0.1631, respec-
tively. The proportional gain, k
p
, is an adjustable param-
eter that determines the deviation from the desired speed
prole. For satisfactory tracking of the actual speed prole,
k
p
, is taken as 12 so that the settling time turns out equal
to the actual settling time of the motor as it is very low
compared to the given displacement time. Now, the system
is able to track the desired prole, and the motor is fed with
the output of the tracking controller as shown in Fig. 3.
5.2 EO control
[12]
modied to take into ac-
count frictional losses
From Section 4,

FL
(t) is obtained as

FL
(t) = 332.7083(cosh(1.5047t 1))+
325.4990(sinh(1.5047t)). (35)
C
T
(s) is calculated from (35) for a given V
i
(s) and is given
by
C
T
(s) =

FL
(s)
V
i
(s)
=
4.889s 7.492
s
2
2.245
. (36)
Further, the total losses given by (7) are minimized with
respect to the eld current for obtained

FL
(t). The vari-
ation of the energy losses with respect to the eld current
is shown in Fig. 4 using dotted line. I

FL
= 0.0945 A gives
minimum losses and is taken as the optimal eld current.
Using the value of I

FL
,
d
, and
i
, for the PID controller, are
calculated from (22) as 0 and 0.1185, respectively. k
p
= 12
is valid in this case also.
Fig. 4 Energy losses versus eld current for the IMD
M. A. Sheta et al. / A New Energy Optimal Control Scheme for a Separately Excited DC Motor 273
5.3 Proposed EO control scheme
This subsection presents the obtained parameters of the
tracking and PID controllers for the proposed scheme. From
the proposed algorithm,

(t),

ods
(t
d
), I

f
, and t
d
are
determined for the given drive specications as follows:
The interval [
min
,
max
] is divided into a set of eight
equidistant points. For each point, intermediate displace-
ment time t
d
and EO speed prole are calculated. Further,
the total losses are minimized with respect to the eld cur-
rent for every EO speed prole.
Fig. 4 shows the variation of the energy losses with eld
current for each disconnecting speed. From the gure, it
is obvious that minimum losses occur at the disconnect-
ing speed,

ods
(t
d
) = 320 rad/s, optimal eld current
I

f
= 0.1205 A, and optimal, intermediate disconnecting
time instant, t
d
= 1.3395 s. The overall EO speed prole,

(t) is given by

(t) = 399.0148(cosh(1.8995t 1))+

391.4802(sinh(1.8995t)). (37)
By taking the Laplace transform of (37), C
T
(s) is obtained
as
C
T
(s) =

(s)
V
i
(s)
=
7.436s 14.4
s
2
3.608
. (38)
Likewise,
d
and
i
, for PID controller, are calculated
from (22) as 0 and 0.0731, respectively, and k
p
is set at 12
as before.
5.4 State feed-back control
As an example of a conventional control technique used
for applications such as electric drives, state feedback or
the pole placement method
[16]
is considered in this section,
and disadvantages of arbitrary design of this control scheme
are presented. No optimization is involved here. Fig. 5
shows the general block diagram of the state feedback con-
trol scheme for the DC motor. K
M
and K
B
are functions
of the eld current.
Fig. 5 General block diagram of the state feedback control sys-
tem
To design the state feedback controller for position con-
trol, the eld current I
f
is arbitrarily taken as 0.1 A lying
between I
f min
and I
f max
. From the chosen value of I
f
,
motor system

s state-space representation is given by

.
X= AX +Bu
and
Y = CX (39)
where
X =
_

d

_
, u = E
a
, Y =
d
and the matrices A, B, and C are computed by using the
given parameters
A =
_
0 1
0 9.34
_
, B =
_
0
15.6
_
, C =
_
1 0
_
.
(40)
Using (39) and (40), the corresponding transfer function is
given by
T
SFB
(s) =

d
(s)
E
a
(s)
=
15.6
s
2
+ 9.34s
. (41)
From the transfer function (41), it is obvious that the sys-
tem is unstable. The objective is to determine the state
feedback gain matrix K = [k
1
k
2
], which stabilizes the sys-
tem and gives the desired transient response. Since the
given displacement time is 3 s, the system must have a set-
tling time equal to the given displacement time t
d
. In the
given case, the inductance of the armature circuit is ne-
glected as it is very less. Therefore, the armature current
cannot be considered as a state vector. However, in gen-
eral, if L
a
is not negligible, then armature current must be
considered as a state vector. In this case, DC motor sys-
tem becomes a third-order system, and it needs to stabilize
three poles. For that case, let the new characteristic equa-
tion, which gives the desired response with state feedback
control be
(s
2
+ 2
nt
s +
2
nt
)(s +P) = 0. (42)
In (42), and
nt
are so determined that the characteris-
tic equation gives the desired transient response, provided
that P is such that it does not aect the transient response
obtained with the help of and
nt
. P is generally taken
as 10
nt
.
For the control law implementation, is assumed to be
0.8 and for the desired settling time,
nt
is obtained as
1.6667 (as the armature current state vector is not present
in the given case, the second term in the characteristic equa-
tion, i.e., s +P is ignored). Therefore, the poles (0, 9.34)
of the original characteristic equation (denominator of (41))
must be placed at 1.333j, i.e., the original characteristic
equation must be modied as s
2
+ 2.667s + 2.778 = 0. Af-
ter inserting the state feedback gain matrix in the original
system (represented by (39)), the new state matrix, A
FB
is given by A
FB
= A BK. For the desired transient re-
sponse, K must be determined such that A
FB
has the above
desired characteristic equation. Accordingly, K is obtained
as K = [0.1780 0.4277].
For zero steady-state error, a gain constant, k
g
= 0.1927
is inserted after the desired reference input,
d
, as shown in
Fig. 5.
274 International Journal of Automation and Computing 6(3), August 2009
5.5 Summary of comparison of energy
losses, armature current variation, po-
sition and speed proles for various
control schemes
Fig. 6 shows the position proles of the drive obtained
with various control schemes discussed in this section. It
is seen that all these proles have the same settling time
(3 s) and steady-state value (565.48 rad), even though the
variation of position of the motor with respect to time is
dierent.
Fig. 6 Variation of position with respect to time for various con-
trol schemes (Note that the initial starting time, the desired dis-
placement and the total displacement time t
d
are the same for
all the proles, irrespective of the control scheme used.)
Fig. 7 compares the speed proles obtained with various
techniques. It is seen that in state feedback, the speed pro-
le exceeds
max
, which may be unacceptable. While in
the case of original EO control scheme
[12]
and modied EO
control scheme
[12]
, to maintain the symmetry of the speed
prole, armature current in the reverse direction is required.
This kind of braking is not required in the proposed tech-
nique as shown in Fig. 8.
Fig. 7 Variation of speed with respect to time for various con-
trol schemes (Note that the initial starting time instant and the
nal specied displacement time t
d
are the same for all control
schemes.)
Fig. 9. shows that the conventional state feedback con-
trol, EO control
[12]
, and EO control
[12]
(modied), incur en-
ergy losses that are approximately 2.23, 1.61, and 1.53 times
higher than in the proposed scheme. Because of very high
losses in the case of state feedback control, as compared to
other control schemes, the maximum value of armature cur-
rent is approximately 2.5 times, i.e., high armature current
rated motor is required if state feedback control is applied
for the same displacement angle and time.
Fig. 8 Variation of the armature current with respect to time
for various control schemes
Fig. 9 Comparison of losses incurred by the IMD
In the proposed control scheme, the supplies are discon-
nected after time instant 1.3395 s. Therefore, the losses
remain constant after time instant 1.3395 s. After discon-
necting both power supplies, the stored kinetic energy is
dissipated as frictional losses, and the motor comes to a
halt at the desired displacement time t
d
and displacement
angle
d
.
6 Experimental results
To validate the theory and simulation results, a labo-
ratory prototype of the incremental motion DC drive was
built. An MOSFET (IRFP450) based buck-boost type DC-
to-DC converter was used to power the drive. Intel

s micro-
controller 8052 was used for controlling the power converter.
M. A. Sheta et al. / A New Energy Optimal Control Scheme for a Separately Excited DC Motor 275
The microcontroller generates variable duty cycle pulses
corresponding to the EO speed prole, which are fed into
the gate of the MOSFET through an isolation cum driver
IC, Agilent

s HPCL3120. This makes the power converter

s
output voltage (which supplies the armature of the DC mo-
tor) vary in accordance with the desired EO speed prole.
Fig. 10 shows the photograph of the experimental setup.
Fig. 10 Photograph of the experimental set up
The speed sensor used for the experiment is a perma-
nent magnet tacho generator that generates 1 V for a motor
speed of 17.16 rad/s. In the experimental system, overall 87
rotations were achieved with a total energy loss of 56 Ws.
Fig. 11 shows a comparison of the simulation results and
the experimental results. Figs. 11 (a) and (b) show the EO
armature voltage proles obtained with Matlab simulation
and with experimental prototype respectively. Figs. 11 (c)
and (d) show the resulting speed proles. It is important to
note that in the armature voltage prole obtained experi-
mentally (see Fig. 11 (b)), the voltage is not zero even after
disconnecting the supplies at the disconnecting speed. This
is because the motor is still in rotation, and the presence of
residual magnet eld causes the motor to work as a gener-
ator until the machine comes to a halt. However, this does
not aect the operation of the IMD since the armature ter-
minals are open (supply is disconnected). Experimental re-
sults show good agreement with the theoretical/simulation
results.
(a)
(b)
(c)
(d)
Fig. 11 Comparison of experimental results with simulation re-
sults. (a) EO armature voltage prole obtained with Matlab;
(b) EO armature voltage prole obtained with the experimental
prototype; (c) EO speed prole obtained with Matlab (A scaling
of 1:17.16 has been used for comparison with experimental out-
put); (d) EO speed prole obtained with experimental prototype
(The speed sensor used, generates 1 V for 17.16 rad/s.)
276 International Journal of Automation and Computing 6(3), August 2009
7 Conclusions
A new EO control scheme for a separately excited DC
motor based IMD has been proposed and veried through
computer simulations and experiments. The scheme mini-
mizes not only the resistive losses in the armature and eld
windings, but also the frictional losses. The minimization of
the energy dissipation in the drive is achieved by minimizing
the losses with respect to the speed prole and the eld cur-
rent and by using the kinetic energy stored in the system.
The performance of the proposed scheme, with respect to
the total energy losses, has been compared with other EO
control schemes, and its superiority has been established.
Also, the importance of optimization in the control of elec-
tric drives has been highlighted.
The merits of the proposed scheme can be summarized
as below:
1) The energy losses being low, the armature power re-
quirement decreases, which implies that a lower rating mo-
tor can be used for the same application.
2) On account of low energy losses, the thermal dissipa-
tion requirements are less stringent.
3) Braking current is not required because the motor
naturally comes to a halt (zero speed), unlike most other
schemes.
One of the drawbacks of the proposed scheme is the as-
sumption that the load and the surrounding environmental
conditions do not change. This restricts its applications.
However, in this paper, the major objective was to demon-
strate a new scheme. In the future, it is proposed to con-
sider conditions involving varying load and environmental
conditions. This would require real-time system identica-
tion, for which a digital signal processor (DSP) or a eld
programmable gate array (FPGA) will have to be employed.
This work would be reported in a future paper.
References
[1] C. K. Lai, K. K. Shyu. A Novel Motor Drive Design
for Incremental Motion System via Sliding-mode Control
Method. IEEE Transactions on Industrial Electronics, vol.
52, no. 2, pp. 499507, 2005.
[2] J. Tal, S. Kahne. Control and Component Selection for In-
cremental Motion Systems. Automatica, vol. 9, pp. 501
507, 1973.
[3] C. C. Chan, R. Zhang, K. T. Chau, J. Z. Jiang. Optimal
Eciency Control of PM Hybrid Motor Drives for Elec-
trical Vehicles. In Proceedings of the 28th Annual IEEE
Power Electronics Specialists Conference, vol. 1, pp. 363
368, 1997.
[4] J. Holtz, L. Springob. Identication and Compensation of
Torque Ripple in High Precision Permanent Magnet Motor
Drives. IEEE Transactions on Industrial Electronics, vol.
43, no. 2, pp. 309320, 1996.
[5] P. Pillay, R. Krishnan. Modeling, Simulation, and Anal-
ysis of Permanent-magnet Motor Drives Part 1: The
Permanent-magnet Synchronous Motor Drive. IEEE Trans-
actions on Industry Applications, vol. 25, no. 2, pp. 274
279, 1989.
[6] G. Zhang, A. Schmidhofer, A. Schmid. Eciency Optimisa-
tion at DC Drives for Small Electrical Vehicles. In Proceed-
ings of IEEE International Conference on Industrial Tech-
nology, IEEE Press, vol. 2, pp. 11501155, 2003.
[7] E. E. EI-kholy, S. S. Shokralla, A. H. Morsi, S. A. EI-
Absawy. Improved Performance of Rolling Mill Drives Us-
ing Hybrid Fuzzy-PI Controller. In Proceedings of the 5th
International Conference on Power Electronics and Drive
Systems, IEEE Press, vol. 2, pp. 10101015, 2003.
[8] T. Egami, J. Wang, T. Tsuchiya. Eciency Optimized
Speed Control System Synthesis Method Based on Im-
proved Optimal Regulator Theory Application to Sep-
arately Excited DC Motor System. IEEE Transactions on
Industrial Electronics, vol. IE-32, no. 4, pp. 372380, 1985.
[9] A. Kusko, D. Galler. Control Means for Minimization of
Losses in AC and DC Motor Drives. IEEE Transactions on
Industry Applications, vol. IA-19, no. 4, pp. 561570, 1983.
[10] S. Murtuza, N. Narasimhamurthi, S. T. Dilodovico. DC
Servo Control-academic and Real. In Proceedings of the
American Control Conference, IEEE Press, vol. 3, pp. 2036
2037, 1995.
[11] R. N. Danbury. Servomechanisms for Incremental Motion:
Power Dissipation Considerations. Mechatronics, vol. 4, no.
1, pp. 2536, 1994.
[12] A. M. Trzynadlowski. Energy Optimization of a Certain
Class of Incremental Motion DC Drives. IEEE Transactions
on Industrial Electronics, vol. 35, no. 1, pp. 6066, 1988.
[13] F. J. Zheng, G. Cook. Energy Optimal Control for Steel
Rolling. IEEE Transactions on Industrial Electronics, vol.
IE-32, no. 4, pp. 388392, 1985.
[14] J. C. Brierley, R. E. Colyer, A. M. Trzynadlowski. The
SOAR Method for Computer Aided Design of Energy-
optimal Positioning DC Drive Systems. In Proceedings of
Industry Applications Society Annual Meeting, IEEE Press,
vol. 1, pp. 464467, 1989.
[15] E. R. Pinch. Optimal Control and the Calculus of Varia-
tions, Oxford University Press Inc., New York, USA, 1993.
[16] I. J. Nagrath, M. Gopal. Control Systems Engineering, New
Age International Ltd., New Delhi, India, 1989.
Milan A. Sheta received the bachelor degree in electrical
engineering from Vyavasayi Vidya Pratishthan Engineering Col-
lege, Rajkot, Gujarat, India in 2002, and the M. Tech. degree in
systems and control engineering at the Indian Institute of Tech-
nology Bombay, India, in 2005.
Vivek Agarwal received the bachelor
degree in physics from St. Stephen

s col-
lege, Delhi University, USA, the Master de-
gree in electrical Engineering from the In-
dian Institute of Science, India, and the
Ph. D. degree in the Depterment of Electri-
cal and Computer Engineering, University
of Victoria, Canada in 1994. He worked for
Statpower Technologies, Burnaby, Canada,
as a research engineer. In 1995, he joined
the Department of Electrical Engineering, Indian Institute of
Technology, Bombay, India, where he is currently a professor.
He is a senior member of IEEE, fellow of IETE, and a life mem-
ber of ISTE.
His research interests include power electronics, modeling and
simulation of new power converter congurations, intelligent
and hybrid control of power electronic systems, power quality,
EMI/EMC, and conditioning of energy from non-conventional
energy sources.
Paluri S. V. Nataraj received the
Ph. D. degree in process dynamics and con-
trol from Indian Institute of Technology
Madras, India in 1987. He is a faculty of the
systems and control engineering at Indian
Institute of Technology, Bombay, India.
His research interests include robust con-
trol, process automation, nonlinear system
analysis and control, and reliable comput-
ing.