2015 International Conferenceon Industrial Instrumentation and Control (ICIC)

Col/ege ofEngineering Pune, India. May28-30,2015

A Scheme to Control the Speed of a DC Motor

with Time Delay using LQR-PID controller
Saurabh Srivastaval and V. S. Pandit2
Variable Energy Cyclotron Center, lIAF Bidhan Nagar, Kolkata, 700064, INDIA

I [email protected], [email protected]

Abstract- In this paper we have designed an optimal PID literature and some of the popular methods are Ziegler­
controller for the speed control of a DC motor. The optimal Nichols tuning, Cohen-Coon tuning, Internal model control,
Linear Quadratic Regulator (LQR) based PID controller is
Direct Synthesis Method, neural networks based
derived analytically for the second order transfer function of
methodologies relay based auto tuning method etc[I,2,4].
the DC motor with time delay. The state weighting matrices of
the LQR are used for finding the set of optimal PID gains. We Most of the real plants can be more closely
have compared the results with the previously developed approximated using second order plus time delay model
results of LQR-PID and found that the present approach of
compared to first order plus time delay (FOPTD) model [4].
PID tuning gives the speed control of the DC motor more close
He et al. [5] have proposed an analytical method to tune the
to the desired damping ratio and natural frequency.
PI/PID parameters in an optimal way using LQR techniques
Keywords-PID, pole-placement, dominant pole, LQR with user specified closed loop damping ratio and natural
frequency for the FOPTD model. Authors have also
extended their approach for SOPTD systems by equating the
I. INTRODUCTION
larger process pole with the derivative term of the PID
Existence of time-delay in a control loop reduces the controller and then applied the PI tuning approach using
phase margin of control systems and reduces relative LQR to obtain other two parameters.
stability. The existence of time-delay in a control loop is a
source of instability and performance deterioration [1]. Most In the present paper we will discuss that taking the
of the real time system contains time delay in their transfer derivative gain of the PID controller equal to the one of the
function and we cannot neglect them. As the high system poles does not provide optimal PID settings and
performance is always desired from the controller, many hence does not give the closed loop time response
researchers have worked on the tuning of controller for the satisfactory with user specified closed loop damping ratio
system having time-delay and few recent techniques can be and natural frequency. This technique also fails to produce
found in Ref. [2]. Proportional-Integral-Derivative (PID) satisfactory PID parameters for complex conjugate poles of
controller is the most widely used controller in the industry the systems since a single complex pole of the process
today. The popularity of PID controller is due to its cannot be eliminated with a single complex zero of the
simplicity which uses only three parameters to tune i.e controller, as they are always in conjugate pairs [6].
Proportional (Kp) term which controls the plant or system We have extended the PI/PID tuning method proposed
proportional to the input error, Integral (K,) term which by He et al. [5] for the first order plus time delay (FOPTD)
gives the change in the control input proportional to the model to develop the tuning procedure for SOPTD model
integral of the error signal and the last one is the Derivative using dominant pole placement approach [6,7]. Dominant
term (Kd) that controls the system by providing control pole placement design was first introduced by person [7]
signal proportional to the derivative of the error signal. and further explained by [4]. In the dominant pole design a
Derivative action is used in some cases to speed up the pair conjugate poles are chosen based on the requirements
response and to stabilize the system behavior [3]. In this on the closed-loop response such as rise time and percentage
paper we present the methodology of PID tuning based on overshoot.
the optimal approach of LQR and dominant pole placement
technique applicable to second order with time delay In this paper we present the simulation results of the
(SOPTD) systems with user specified closed loop damping speed control of a DC motor presented in Ref [8] subjected
to the time delay. The PID parameters are evaluated
ratio '?;c/' and natural frequency 'wc/'.
analytically and compare with the previous developed LQR­
Although the PID controller is known to be the simplest PID tuning method. The simulated results done in
and efficient controller but it requires effective and MATLAB shows remarkable closed loop time response
optimized tuning of the control parameter (Kp, K" Kd). Many with user defined closed loop performance measures up to
PID controller tuning methods have been proposed in the the certain time delay. Section 2, first briefly presents the

978-1-4799-7165-7/15/$31.00 ©2015 IEEE

Authorized licensed use limited to: UNIVERSIDADE FEDERAL DO MARANHAO. Downloaded on December 05,2023 at 16:50:20 UTC from IEEE Xplore. Restrictions apply.
 LQR solution for time delay systems and then details of the the optimum control vector u(t) subjected to the
PID tuning method for the SOPTD systems. In section 3 minimization of the cost function defined by
case study of a speed control of a DC motor is discussed
followed by the conclusion. (4)

II. MATHEMATICAL PREMINARIES FOR OBTAINING where Q is the semi positive defmite state weighting matrix
OPTIMAL LQR-PID CONTROLLER and R is the positive definite control weighting matrix.

Fig. 1 shows the step response of a typical open loop

system with time delay. The delay term can be evaluated
from the curve by drawing a tangent line on the inflection ISOPTO System TF G(S)I
point in the transient region and then calculating the
K y(t)
distance of its intersection from origin on the time axis [4]. _,
___ e->L
s' + as+ b

K
O. K

Tangent line ;It Fig. 2. Block diagram of the closed loop SOPTD system with PID
uille lion poult controller

LQR solution to the above process gives the optimal control

vector um (t) as

Um(I)= _R-1BTpX (I), t�L (5)

where P is the symmetric positive defmite Riccati

coefficient matrix which can be obtained by solving
Time s
Continuous Algebraic Riccati Equation (CARE) given as
Fig. I. Open Loop time response of the system with time delay (L)
ATp+PA +Q-PBR -IBTp=O (6)

A linear plant with time delay can be characterized by In the case of a second order process the state weighting
the state-space representation as matrix Q, and the Riccati coefficient matrix P are generally
taken as
X(/)= AX (/)+Bu(/-L), t2:0 (1)
o
l I PII PI2 PI3 l
where A, B, X and L are the corresponding state transition o I p= I I
matrix, control matrix, state matrix and the time delay term I' I PI2 P22 P23 I •
(7)
respectively. For t < L, no control signal will be effective Q3J lpl3 P33 J
and only for t 2: L the control signal comes into the picture.
Equation (1) can be written as In the optimal control it is a standard practice to design
regulator by varying Q and keeping R fixed equal to
X(/)= AX (I) , OS:t<L (2) (R=[r]=[I]) [10]. In order to fmd the optimal control for
time delay system put t = t + L in (5)
XU)= AX (I)+ Bu m(I), t2:L (3)
Um(I+ L)= u(I+ L -L) =-R-1BTPX (I+ L)
where, um(I)= u(I-L) . Fig. 2 shows the Block diagram of
For, t + L 2: L which gives
the closed loop SOPTD system with PID controller.
Representation of x (I) is given as t2: 0 (8)
X(/)=[XI(/) X 2 (/ ) X3(/)]T, The time advance feature of state variable matrix leads
where to the decomposition of control vector u(t) in two parts one
de (t) for t less than L and another for t greater than L i.e.
xl(t) = feet) dt, x2(t) = e(t), X3 (/) = -- ,
dl
Since (3) is now delay free, one can easily apply the
standard LQR approach [9] for delay free processes to fmd

Authorized licensed use limited to: UNIVERSIDADE FEDERAL DO MARANHAO. Downloaded on December 05,2023 at 16:50:20 UTC from IEEE Xplore. Restrictions apply.
 where, Ac = A - BR -I B TP . o l I0 l
I I I
From (9) and (10) knowing the values of matrices A, B o \
I ,
B =
I
0 I
(16)
(Ac)t A(L-t)
,P, e ande one can calculate the optimal control -b -aJ l-KJ
u(t). It should be noted that (9) leads to the time varying
control signal (coefficient ofx (I») while (10) has constant The matrix Ac can be determined by setting the
characteristic equation of the closed loop systems
control signal. We have not consider the control signal
given by (9) in the present paper as it leads to the time
L'. s
( ) s
= I I - A I equal to the desired closed loop equation
c

varying and larger control signal initially. Large initial ( +mSC1wc/)( s2 +2ssc/wcl +w ,).
s : Here we have utilized
control signal leads to the problem of actuator saturation and
the help of dominant pole placement technique [4, 6] where
time varying signal are generally difficult to implement in
the location of third pole is placed m times away from the
analog controller. So we have modified the control law and
real part of the dominant closed loop poles i.e. mSc/ W cl . We
consider the same control signal initially which is valid for t
greater than L. So in this paper the only control signal which call this m as relative dominance and as per the published
is valid throughout is given as literature its value should be chosen 3 or more. Matrix Ac �

can be evaluated in terms of parameters (eI"and Wei as

(11)
-\ o
A (L-t)
So we do not have to fmd the values of e Now •
-\
we will discuss how to obtain these values. From Fig. 2 the a p 13 b+ a p 23 a
s+ + a p 33
control signal in terms of the state variable is given as
= ( + mSc/W cl )( S2 + 2sSc/W c/+ w ,)
s : (17)
U(/) = K p X2 (I) + KjxI (I) +K d X3 (I) (12)
I
The transfer function (TF) of the PID controller can be Here a = r - K2 By comparing the coefficients of power of
express in s domain as
s from both sides of (17), the elements p 13 ' P 23 and p 33

u(s) Ki can be expressed as

C(s)= -- =Kp +-+K"s (13)
e(s)

In the case of unity output feedback system such as a

shown in Fig. 2, if we put the reference signal ret) = 0, we

have e(t) = - y(t). With this condition, the second order TF
with time delay can be written as a

-sL
yes) Ke - e(s)
Gs
( )= --= ---- --- (14) (18 )
U ( S) s2 + as + b U s
( ) a

The remaining three elements of the matrix P and three

in which a = 2(01 Wol and b = OJ :, ' where (01 and Wol are the elements of the matrix Q can be obtain by solving CARE
damping ratio and natural frequency of the open loop plant equation given by (6). Results are
respectively. Equation (14) can be expressed in terms of
time variable as
a

X 3 (I) = -ax 3 (/)-bx 2 (I) - Ku(/ -L).

In terms of state-space formulation the derivative of the a

state variables can be written as

o llx1(/) l 10 l P
22 a
II I I I
o \ ( ) +1
t 0 u(t-L) (15)
Ilx2 I I
-b -aJ lx3(t)J l-KJ
a

Comparing (15) with (1), it IS straightforward to obtain

matrices A and B as a

Authorized licensed use limited to: UNIVERSIDADE FEDERAL DO MARANHAO. Downloaded on December 05,2023 at 16:50:20 UTC from IEEE Xplore. Restrictions apply.
 2 2 2 2 2 2 -2 2 time delay. It can be easily seen from Fig. 3 that increasing
4(e1())e1+m (cl ())e1+2b- a ())e1
Q3 = (19) the value of time delay L the closed loop system becomes
a
more oscillatory and finally leads to instability.

Finally the value of e A cL can be evaluated as ,.

2.0
n - L =O
! \ ---- L = 0.1
i !
I .

i ! L = 0.2

r: �!\\ ,(\J\/::-:= O'

····· ···

I III (I) 112 (I) 113 (I) l

1 1
=1121(1) 122 (I) 123 (I) 1 (20)
l/31 (I) 132 (I) 133(')1�
L
[I \ f I j

Values of lij (L) are the final values of the matrix elements
U
Ei
\'ii \]
0.5 :i 'i
I;; 'i
�.
of e (A ,)'
at t = L, where i = j = 1, 2, 3. The optimal control Hi

u(t) can be obtained by using (11) as lfj

1 :1
I:"

0.0 2.0 4.0 6.0 8.0 10

Time (s)

T
Fig. 3. Closed loop time response of the speed control of a DC motor with
I P 13 l I III (L) 112 (L) 113 (L ) II x I (t) l different L
I 1 1 1 II I
U(/)= r - KI P23 I 1 / 2 1 (L) 122 ( L ) 123 ( L ) II 2 (t) I
x (21)
From the inspection of (23) the roots of the system are
lp33 J l/31 (L) 132 (L) 133 (L) J l x 3 (t) J real (-2 and -10). This implies the PID tuning for such
SOPTD system can be obtained by the previous LQR-PID
techniques as proposed in [5], by equating the derivative
Comparing the coefficients of x I (t) , X
2 (t) and x 3 (t) gain as one of the larger real system poles and then
in (21) and (12) it is straightforward to obtain the PID obtaining the Kp and K, according to the FOPTD tuning
parameters for t � 0 as procedure. The values of PID gains according to the
previous method with same (c/ and Wc/ are given in Table I.
I
K, = r- K( P13/II(L)+ P23/2 1(L)+ P33/31(L)),
I TABLE I: PID Parameters using formulation of [5]
K p = r - K (p 13 112 ( L)+ P23 I 22 ( L)+ P33 I32 ( L)) ,
L Kp Ki Kd (c/ Wc/
I
K d = r - K (p 13 113(L) + P23 123 (L) + P33 133(L)) . (22)
O.ls 1.1727 3.3418 10 0.8 3rad/s
O.2s 0.9488 2.388 10 0.8 3rad/s
III. SPEED CONTROL OF A DC MOTOR WITH TIME DELAY
VIA OPTIMAL LQR-PID CONTROLLER Fig. 4 shows the closed loop time response of the speed
control of a DC motor by using the un-optimized LQR-PID
In order to test the effectiveness of the proposed settings keeping Kd = 10 for different time delay L. It can be
methodology we have taken the standard transfer function easily seen that at L = O.ls, although the system is stable but
used for the speed control of a DC motor from [8]. Here it does not fulfill the required closed loop performance
authors proposed the evaluation of PID controller with user measures. At L = 0.2s the closed loop response becomes
specified damping ratio «(c/ = 0.8) and natural frequency (wc/ unstable. The gain margin (GM) and phase margin (PM) for
= 3rad/s) for a second order transfer function without time L =O.ls is GM = 3.15dB and PM = 28.34deg and for L =0.2s
delay. With the introduction of time delay L the TF becomes these are negative (unstable response).
Observing Fig. 4, it could be concluded that it is not
() 2 sL
= - = ----- e
_
G(s) 2 (23) always possible to obtain the PID gains for SOPTD system
V s + 12 s + 20 by equating the larger process pole with the derivative term
of the PID controller and then applying the PI tuning
Here L represents the delay due to the motor as well as from
approach using LQR to obtain other two parameters. Fig. 5
the components in the forward path. In Fig. 3 we have
shows the zoom view of Fig. 4 so that the existence of time
shown that by the introduction of time delay L with same
delay L can be seen clearly. In Fig. 6 we have plotted the
PID parameters (Kp = 27; Ki = 60; Kd =3.3318) one could
close loop time response obtained using the optimized LQR-
not get the desired closed loop time response as without

Authorized licensed use limited to: UNIVERSIDADE FEDERAL DO MARANHAO. Downloaded on December 05,2023 at 16:50:20 UTC from IEEE Xplore. Restrictions apply.
 PID settings following the present methodology for the O.ls 14.8576 33.0225 1.1287 0.8 3rad/s
SOPTD system of DC motor speed control at different O.2s 11.9241 23.9702 0.9588 0.8 3rad/s
values of time delay L. O.3s 9.l788 16.5105 0.7681 0.8 3rad/s

""

j
"""

�
1.5
-L=0.1 Kd=10
--- L=0.2 Kd=1 0

'"
�
'.
"
:�
"'"
,. I:
r:I,
t
�0.8
!!!.
1.0

:1.' (
: "" -L=O.1
50% Load
�..... ,

-S """ -S --L=0.2 Disturbance at t = 6s

"0 1.0 ,, � 0.6
Q)
Q)
�
!� : I : ! :: ········ L=0 3 .

0.. Q)
0..
:i
� � : : : : u
(f) t:
:\ : I: I:
I

I
(f) t:
0 .4 t:
0.5

,'''''
�,' �
l . f\ : \ : \: 1: :
,''t ! \: \ ! \ : ,! l
/' /\/ \, '.J � \i t: i
� :: :
H J:
V !r
I 0.2

'?
t:
t:
t:
t:
t:
r:
t:
t:
t:
t:
t:
t:

0.0 2.0 4.0 6.0 8.0 10

0.0 2.0 4.0 6.0 8.0 10
Time (s)
Time (s)
Fig. 4. Closed loop time response of the speed control of a DC motor with
Fig. 6. Closed loop time response of the speed control of a DC motor with
time delay following calculation of PID using [5].
time delay using present methodology.

14

0.04
13
•

$" : -L=0.1
(j)
0.03
E -L=0.2
o 12 .! \'
t::
"0 ········
L=0 3
Q) �
.

: r\
�O . 02
(f) � 11 U \.
o � f "�\\
0.01

0.0 0.1 0.2

o
10:J r·...
.
,\

.
.· .::.,..
-.;
_-
-
-----
--4
4·
Time (s) 0.0 2.0 4.0 6.0 ·· · · 8.0 10
Fig. 5. Zoom view of Fig. 4 for clear visibility of time delay L. Time (s)
Fig. 7. LQR-PID Controller Response using present method.
From the simulation results presented in Fig. 6, it is easy
to conclude that the present methodology gives much better
closed loop time response as compared to the previously We like to point out that above PID settings are
developed LQR-PID present system. The reason for this evaluated at a fixed value of third pole location by
improvement lies in the fact that for each value of time considering m = 4. However, one can further improve the
delay a new set of optimized PID parameters is needed in close loop time response by precisely tuning the value of m.
contrast to the previous method where Kd is fixed for any From Fig. 6 one can easily observe that for three different
value of L. The values of optimized PID settings with same values of time delay L, the percentage overshoot and settling
(c/ and Wc/ are given in Table II for comparison where for time are almost identical and are very close to the required
different values of time deay L one gets different Kd along closed loop performance measures ((c/ = 0.8) and (wc/ =
with the optimized sets of Kp and K,. 3rad/s). Fig. 7 shows the LQR-PID controller response
obtained using present method for the DC motor speed
TABLE II: PID parameters with present method control at different values of time delay L. Here we see that
as the time delay L increases we need less control effort.
This advantage can be exploited in cases where there is

Authorized licensed use limited to: UNIVERSIDADE FEDERAL DO MARANHAO. Downloaded on December 05,2023 at 16:50:20 UTC from IEEE Xplore. Restrictions apply.
 limitation in the actuator saturation. One can design a PID utilizing the pole placement techniques. To demonstrate the
controller intentionally by introducing some time delay in effectiveness of present method we have performed
the forward path. However, this way one can slightly simulations considering the example of a speed control of
increase the rise time. DC motor with time delay. It is shown that present method
gives much improved results than the previous LQR based
TABLE III: GM and PM obtained with present method.
PID tuning method where the derivative term is taken as one
L GM(dB) PM(deg) of the real pole of the system [5].
O.ls 37.51 64.18
O.2s 27.15 58.88
/\\
I
O.3s 24.35 60.13 I \
1.2

lro
Table III gives the numerical values of the robustness . ..
! ; ..-
�y � ... ..
��----------------------�
if \ ',
measures of the closed loop control system. One can easily 1.0
�

.... ... ,,'

distinguish that with the present tuning method for time en "
'i
delay L =O.ls the GM are almost 10 times more while the I O.8 I:
"
, -+50% change in Kp
PM are more than twice than the previous LQR-PID method "0 ,
,
<D ----+50% change in Ki
( for L =O.ls: GM = 3.15dB and PM = 28.34deg). � 0.6
(f) ..·....·+50% change in Kd
1.4 ,......��--.-��.,...,���.,...�-.-�--.-...,
0.4
1.2
0.2
1.0

� 0.0 2.0 4.0 6.0 8.0 10

.s0.8 --50% change in Kp Time (s)
"0
<D ---- -50% change in Ki
� 0.6 ........ -50% change in Kd
Fig. 9. Closed loop time response when the values of PID controller
(f) parameters are increased by 50%.
0.4

REFERENCES

[I]. Gang Feng, Rogelio Lozano "Adaptive Control systems" Newness

pUblic.1999.
0.0 2.0 4.0 6.0 8.0 10 [2]. Sigurd Skogestad, "Probably the best simple PID tuning rules in the
Time (s) world" Journal of process control july3,200I.
[3]. S. Srivastava, Y. Kumar, A. Misra, S. K. Thakur and V. S. Pandit
Fig. 8. Closed loop time response when the values of PID controller
"Optimum Design of Buck converter controller using LQR
parameters are decreased by 50%.
approach ", IEEE, International Conference on Advanced Computing
Technology,21-22 Sept INDIA 2013. Pp. 1-6.
In order to check the robustness of the PID controller we [4]. K. J. Astrom , T. Hagglund, PID controller, Theory Design and
perturbed the PID controller parameters (Kp , K, and Kd) by Tuning Instrument society of America: 1995 .
± 50% from the actual values and performed simulation to [5]. Jian-Bo He, Qing-Guo Wang, Tong-Heng Lee. PI/PID controller
check the behaviour of closed loop time response. Fig. 8 tuning via LQR approach. Chern Eng Sci 2000; 55(13):2429-39.

shows the closed loop time response when the values of PID [6]. G. F. Franklin, 1. David Powell, Michael Workman, "Digital control
of dynamics systems"yd edition,Pearson education.
controller parameters are decreased by 50% with same DC
[7]. P. Persson, K. J. Astrom, "Dominant pole design-a unified view of
motor speed control TF at L=0.2s. It can be seen that a PID controller tuning",in: L. Dugard, M.M. Saad, I.D. Landau (Eds.),
decrease in proportional term causes an extra larger Adaptive Systems in Control and Signal Processing 1992. France,1-3
overshoot where as a decrease in the value of Ki affects the July 1992,Pergamon Press,Oxford,1993,pp. 377-382.
settling time. Fig. 9 shows the closed loop time response [8]. S. Srivastava, A. Misra, V. S. Pandit "Auto tuned PID Controller
when the values of PID controller parameters are increased design Using Diophantine Equation ",IEEE,International Conference
on communication devices and intelligent systems Dec, 28-29 INDIA
by 50%. In this case integral term produces more overshoot
20I2. pp 492-495.
and also affects settling time. It is easy to observe that in
[9]. D. S. Naidu,Optimal Control System. CRC Press 2003.
both the cases the closed loop time response is more or less
[10). Suman Saha et.al, "A conformal mapping based fractional order
unaffected by the change in value Kd• approach for sub-optimal tuning of PID controllers with guaranteed
dominant pole placement" Commun Nonlinear Sci Numer Simul 17
IV. CONCLUSION (2012),3628-3642.

In the present paper we have developed an LQR based

tuning method of PID controller for SOPTD process

Authorized licensed use limited to: UNIVERSIDADE FEDERAL DO MARANHAO. Downloaded on December 05,2023 at 16:50:20 UTC from IEEE Xplore. Restrictions apply.