Received: March 31, 2018 241

Speed Control of DC Motor Using PID and FOPID Controllers Based on

Differential Evolution and PSO

Abdelhakim Idir1* Madjid Kidouche1 Yassine Bensafia2 Khatir Khettab3 Sid Ahmed Tadjer4

1
Applied Automation Laboratory, F.H.C., University of Boumerdes,
1 Av. de l'Independance, 35000 Boumerdes, Algeria
2
Department of Electrical Engineering, Bouira University, 10000, Algeria
3
Department of Electrical Engineering, University Mohamed Boudiaf of M'sila, 28000, Algeria
4
Electrification of Industrial Enterprises Laboratory, F.H.C.,University of Boumerdes,
1 Av. de l’Independance, 35000 Boumerdes, Algeria
* Corresponding author’s Email: [email protected]

Abstract: DC motors are widely used in industrial application for its different advantage such us high efficiency,
low costs and flexibilities. For controlling the speed of DC motor, conventional controller PI and PID were the most
widely used controllers. But due to empirically selected parameters 𝐾𝑝 , 𝐾𝑖 , 𝐾𝑑 and limitation of convention PID
controller to achieve ideal control effect for higher order systems, a Fractional order Proportional-Integral-
Derivative PID (FOPID) based on optimization techniques was proposed in this paper. The aim of this paper is to
study the tuning of a FOPID controller using intelligent soft computing techniques such as Differential Evolution
(DE) and Particle Swarm Optimization (PSO) for designing fractional order PID controller. The parameters of
FOPID controller are determined by minimizing the Integral Time Absolute Error (ITAE) between the output of
reference model and the plant. The performance of DE and PSO were compared with several simulation experiments.
The simulation results show that the DE-based FOPID controller tuning approach provides improved performance
for the setpoint tracking, error minimization, and measurement noise attenuation.
Keywords: DC motor, Fractional PID, Tuning parameters, Differential evolution, Particle swarm optimization, Cost
function.

enable the controller to provide the more flexibility

1. Introduction and stability.
It is quite difficult to optimize the parameters of
The DC motors are widely used in industrial
the FOPID controller in linear and nonlinear
application for its different advantage such us high
systems. There is a need for an effective and
efficiency, low costs and flexibilities. For
efficient global approach to optimize these
controlling the speed of DC motor different
parameters automatically.
controllers is used, most widely used controllers are
For this reason different design methods for
conventional controller PI and PID. But
FOPID controller have been reported in the
conventional PID controller has been facing lots of
literatures. In [1] the authors proposed a new
problem to achieve ideal control effect. For higher
approach for robust control by fractionalizing an
order systems, PID has not been working properly.
integer order integrator in the classical PID control
When compared with the classical three terms
scheme and they use the Sub-optimal
PID controllers, the fractional order controllers have
Approximation of fractional order transfer function
two additional control parameters defined as
to design the parameters of PID controller .In [2],
integration and differentiation orders which may
the authors designed a new tuning rules for the

International Journal of Intelligent Engineering and Systems, Vol.11, No.4, 2018 DOI: 10.22266/ijies2018.0831.24
Received: March 31, 2018 242

tuning parameters of FOPID based on Ziegler– (𝐾𝑝 , 𝐾𝑖 , 𝐾𝑑 , 𝜆, 𝜇) [19].

Nichols (Z–N) rule, A new tuning method for An evolutionary computation technique has
designing fractional order PID controllers based on become gradually popular to obtain global optimal
radial basis function (RBF) neural networks was solution in many areas. A Differential Evolution
proposed in [3]. Sharma et al [4] proposed a Optimization (DEO), particle swarm optimization
Fractional Order Fuzzy Proportional Integral (PSO) and Genetic Algorithm (GA) are stochastic
Derivative (FOFPID) controller for a two-link optimization strategy from the family of
planar rigid robotic manipulator for trajectory evolutionary computation [20, 21].
tracking problem. For tuning of parameters of all the DE has been regarded widely as a promising
controllers, Cuckoo Search Algorithm (CSA) optimization algorithm. What’s more, the optimal
optimization technique was used. Chang et al. [5] problems solved by genetic algorithms (GA) can be
proposed a novel adaptive GA for the multi-
objective optimization design of a FOPID controller obtained better solutions with PSO in comparison
and applied it to the control of an active magnetic with conventional methods. These are precisely the
bearing system. They found that the fractional PID main motivations that led us to apply DE and PSO
controllers have remarkably reduced the overshoot for FOPID controllers design.
and settling time compared with the optimized This paper proposes a new method to design a
conventional PID controller. Bingul [6] employed speed controller of a DC motor by selection of
the differential evolution (DE) algorithm to tune a FOPID parameters using DE. To show the efficiency
PID controller for unstable and integrating processes of DE, the results of this method are compared with
with time delay. The results showed that a faster PSO method. Minimization of time domain based
settling time, less or no overshoot, and higher objective function is the main focus of design
robustness were obtained with the PID tuned DE. methodology.
Cao [7] demonstrated the parameter optimization of The structure of this paper is organized as
a fractional order controller based on a modified follows: Section 2 deals with mathematical
PSO. In their paper, the improved PSO could modelling of DC motor. Sect. III introduces the
achieve faster search speed and better solution fractional order PID controller; Section IV provides
compared to the GA. Maiti et al. [8] employed PSO a brief overview of the DE and PSO algorithms.
for designing fractional order PID controllers. They Section V applies the new algorithm in this paper to
reduced significantly the percentage of overshoot, parameter setting of fractional order PID controller
rise, and settling times using FOPID controllers through a simulated calculation example; and
compared to a PID controller. Alfi and Modares [9] Section VI draws the conclusion of the whole paper.
found optimal system parameters for an unstable
nonlinear system and optimal parameters of the PID 2. Modelling of DC motor
controller using a novel Adaptive PSO (APSO).
They compared the APSO with a Linearly In order to experiment our proposed robust
Decreasing inertia Weight PSO (LDW-PSO) and the control strategy, let us apply it in numerical
GA. The APSO has a faster convergence speed than simulations to the general model of a DC motor
the GA and LDW-PSO. (DCM) as depicted in [22]. The voltage 𝑉𝑎 is applied
The controllers of the speed that are conceived to command the motor angular velocity 𝜔(𝑡). Fig.1
for goal to control the speed of DC motor are shows the schematic diagram of armature controlled
numerous: Fractional PID Controller [10, 11], DC motor.
Fractional fuzzy PID Controller [12]; Genetic
algorithm [13, 14], Particle Swarm Optimization
𝑖𝑓
[15], … etc. 𝐹𝑖𝑥𝑒𝑑
The benefit of FOPID controller is flexible to 𝑅 𝐿 𝑓𝑖𝑒𝑙𝑑
𝑖𝑎 (𝑡) 𝑇𝑚
design, more robust [16] and the most important 𝜃
advantages is the better control of dynamical
systems and less sensitivity to changes in parameters 𝑉𝑎 𝐽
𝑒𝑏
of a control system [17, 18]. In Fractional Order PID
(FOPID) besides setting the proportional, derivative 𝑓𝜃
and integral constants 𝐾𝑝 , 𝐾𝑖 , 𝐾𝑑 we have two more Figure.1 Closed loop response with PID controller
parameters λ (integral order) and μ (derivative
order). Hence, for designing FOPID controller, there
is a need of proper tuning of five parameters
International Journal of Intelligent Engineering and Systems, Vol.11, No.4, 2018 DOI: 10.22266/ijies2018.0831.24
Received: March 31, 2018 243

Step Response
The DC motors are generally used in the linear 2
range of the magnetization curve. Therefore, air gap Coventional PID
flux 𝜑 is proportional of field current:
1.5
𝜑 = 𝐾𝑓 𝑖𝑓 (1)

yout
1
Where 𝐾𝑓 is constant.
The torque 𝑇𝑚 developed by the motor is
proportional to the armature current and air gap flux: 0.5

𝑇𝑚 = 𝐾𝑚 𝑖𝑎 (2) 0
0 5 10 15 20
Times (s)
Where 𝐾𝑚 is the motor torque constant. Figure.2 Closed loop Time(s) with PID controller: 𝐾𝑝 =
response(seconds)
The motor back EMF being proportional to 10, 𝐾𝑖 = 100, 𝐾𝑑 = 0.25
speed is given as:
(6). The resulting simplified mathematical model
𝑑𝜃
𝑒𝑏 = 𝐾𝑏 𝑑𝑡 (3) form is:

Where 𝑒𝑏 is the back EMF constant. 𝜃(𝑠) 𝐾𝑚

𝐺𝑀 (𝑠) = =
The differential equation of armature circuit is: 𝑉𝛼 (𝑠) 𝑠[𝑅(𝐽𝑠 + 𝐾𝑓 ) + 𝐾𝑏 𝐾𝑚 ]
[𝐾𝑚 ⁄(𝑅𝐾𝑓 +𝐾𝑏 𝐾𝑚 )] 𝐾
𝑉𝑎 = 𝐿
𝑑𝑖𝑎
+ 𝑅𝑖𝑎 + 𝑒𝑏 (4) 𝐺𝑀 (𝑠) = = 𝑠(𝜏 𝑀+1) (7)
𝑑𝑡 𝑠(𝜏𝑠 +1) 𝑠

And the dynamic equation with moment of inertia Where:

and coefficient of friction will be: 𝜏 = 𝐾𝑚 ⁄(𝑅𝐾𝑓 + 𝐾𝑏 𝐾𝑚 ): is the time constant.
𝐾𝑀 = 𝐾𝑚 ⁄(𝑅𝐾𝑓 + 𝐾𝑏 𝐾𝑚 ): is the gain.
𝑑2 𝜃 𝑑𝜃
𝑇𝑚 = 𝐽 𝑑𝑡 2 + 𝑓 𝑑𝑡 (5) with 𝐾𝑚 = 𝐾𝑏 .
The specifications are: min. volt-age 1.5 𝑉, max.
The resulting mathematical model for controlled voltage 2.5 𝑉 , nominal voltage 2 V, max rated
DC motor is given by the following transfer function current 0.08 𝐴, no load speed 3830 𝑟/𝑚𝑖𝑛 and rated
[22]: load speed 3315 𝑟/𝑚𝑖𝑛.
For our mini DC motor the physical constants
𝐺𝑀 (𝑠) = 𝑉
𝜃(𝑠) 𝐾𝑚
= 𝑠(𝐿𝑠+𝑅)(𝐽𝑠+𝑓)+𝐾 (6) are R = 6Ω, 𝐾 𝑚 = 𝐾𝑏 = 0.1, 𝐾𝑓 = 0.2 𝑁𝑚𝑠
𝑎 (𝑠) 𝑏 𝐾𝑚 and 𝐽 = 0.01 𝑘𝑔𝑚2 /𝑠 2 .
For the considered motor parameters the transfer
Where function (7) becomes:
𝑅: Armature Resistance (𝛺).
𝜃(𝑠) 0.08
𝐿: Inductance of armature winding (𝐻). 𝐺𝑀 (𝑠) = = (8)
𝑉𝛼 (𝑠) 𝑠(0.05𝑠+1)
𝑖𝑎 : Armature current (𝐴).
𝑖𝑓 : Field current (𝐴). Fig. 2 shows the closed loop response with
𝑉𝑎 : Applied armature voltage (𝑉). conventional PID controller. Thus the system (Eq.
𝑒𝑏 : Back emf (𝑉). (8)) is marginally stable.
𝑇𝑚 : Torque developed by motor (𝑁𝑚).
𝜃: Angular displacement of motor shaft (𝑟𝑎𝑑). 3. Fractional PID Controller Design
𝜔: Angular speed of motor shaft (𝑟𝑎𝑑/𝑠𝑒𝑐).
𝐽: Equivalent moment of inertia of motor and 3.1 Fractional Calculus
load referred to motor shaft (𝑘𝑔 − 𝑚2). Fractional calculus is a generalization of the
𝑓: Equivalent friction coefficient of motor and differentiation and integration to non-integer-order
load referred to motor shaft (𝑁𝑚. 𝑠/𝑟𝑎𝑑). 𝜇
fundamental operator 𝑎𝐷𝑡 , where a and t are the
As the armature time constant for most DC bounds of the operation. The definition of the basic
Motor (𝑀) is negligible we can simplify the model operator which includes the derivative and
integration is [23]:
International Journal of Intelligent Engineering and Systems, Vol.11, No.4, 2018 DOI: 10.22266/ijies2018.0831.24
Received: March 31, 2018 244

𝑑𝜇 𝜆 1 𝑡
𝜇>0 𝑎𝐷𝑡 𝑓(𝑡) = ∫ (𝑡 − 𝜏)𝜆−1 𝑓(𝜏)𝑑𝜏 (13)
𝑑𝑡 𝜇 Γ(𝜆) 𝑎
𝜇
𝑎𝐷𝑡 = {1 𝜇=0 (9)
𝑡 The Laplace transform of the fractional derivative of
∫𝑎 (𝑑𝜏)−𝜇 𝜇<0
𝑓(𝑡) is given by:
Where 𝜇 is a fractional order of differentiation
𝐿{𝐷 𝜇 𝑓(𝑡)} = 𝑠 𝜇 𝐹(𝑠) − [𝐷 𝜇−1 𝑓(𝑡)]𝑡=0 (14)
or integration, generally 𝜇 ∈ ℝ. The negative sign of
𝜇 indicates integration while positive one means The Laplace transform of the fractional integral of
derivation [24].
𝑓(𝑡) is given as follows:
There are many mathematical definitions of
fractional derivatives [25]. One of the most
important used definitions is Grunwald-Letnikov 𝐿{𝐷 −𝜆 𝑓(𝑡)} = 𝑠 −𝜆 𝐹(𝑠) (15)
definition which is perhaps the most popular
because of its suitability for the realization of Where
discrete control algorithms. 𝐹(𝑠) is the Laplace transform of 𝑓(𝑡).
The Grünwald-Letnikov definition of fractional-
3.2 Fractional PID controller
order derivatives is expressed as [26]:
𝑡−𝑎
The Fractional Order PID (FOPID) Controller is
𝜇 1 [ ] the expansion of the generic control loop feedback
𝑎𝐷𝑡 𝑓(𝑡) = lim ∑ ℎ (−1)𝑗 (𝜇𝑗) 𝑓(𝑡 − 𝑗ℎ) (10)
ℎ→0 ℎ 𝜇 𝑗=0 mechanism (PID controller) widely used in industrial
control systems. The FOPID Controller attempts to
Γ(𝜇+1)
With (𝜇𝑗) = Γ(𝑗+1)(𝜇−𝑗+1) correct the error between a measured process
variable and a desired set point by calculating and
then outputting a corrective action that can adjust the
while the definition of fractional-order integral is
process accordingly.
expressed as:
The transfer function of the FOPID controller is
𝑡−𝜆 described as follows:
1 [
𝜆 ]
−𝜆 ℎ
𝑎𝐷𝑡 𝑓(𝑡) = lim ℎ −𝜆 ∑𝑗=0 ( 𝑗 ) 𝑓(𝑡 − 𝑗ℎ) (11)
ℎ→0
𝐺𝑐 (𝑠) = 𝐾𝑝 + 𝐾𝑖 𝑆 −𝜆 + 𝐾𝑑 𝑆𝜇 (16)
With:
Γ(𝑛−𝜆+1) The FOPID equation has five unknown parameters,
(𝜆𝑗) = 𝑗!Γ(𝜆) , Γ(1) = 1 and Γ(𝑥 + 1) = 𝑥Γ(𝑥) where 𝐾𝑝 is the proportional gain, 𝐾𝑖 is the integral
for 𝜆 ∈ ℕ, Γ(𝜆 + 1) = 𝜆! gain, 𝐾𝑑 is the derivative gain, 𝜆 is the fractional-
order integral and 𝜇 is the fractional-order derivative
where: and 𝜆, 𝜇 are positive real numbers.
𝜆! 𝜇!
(𝜆𝑗) = 𝑗!(𝜆−𝑗)! and (𝜇𝑗) = 𝑗!(𝜇−𝑗)! are the The block diagram of control system employing
Soft computing FOPID control action is shown in
binomial coefficients (𝑗 > 0).
Fig.3.
𝜆, 𝜇 : Integral and derivative Order respectively.
Γ(. ) : Gamma function 𝐹𝑖𝑡𝑛𝑒𝑠𝑠
ℎ : Step time.

Another popular definition is that of Riemann-

Liouville definition of fractional-order derivatives PSO/DE 𝑂𝑏𝑗𝑒𝑐𝑡𝑖𝑣𝑒
𝐼𝑛𝑡𝑒𝑙𝑙𝑖𝑔𝑒𝑛𝑡𝑠 𝐴𝑙𝑔𝑜𝑟𝑖𝑡ℎ𝑚𝑠
𝑓𝑢𝑛𝑐𝑡𝑖𝑜𝑛
given by:
𝐾𝑝 𝐾𝑖 𝐾𝑑 𝜆 𝜇
𝜇 1 𝑑𝑛 𝑡 𝑓(𝜏)
𝑎𝐷𝑡 𝑓(𝑡) = ∫
Γ(𝑛−𝜇) 𝑑𝑡 𝑛 𝑎 (𝑡−𝜏)𝜇−𝑛+1
𝑑𝜏 (12) 𝑉(𝑠) + 𝐸(𝑠) 𝐺𝑐 (𝑠) 𝑈(𝑠) 𝐺(𝑠) 𝜃(𝑠)
+−-
- FOPID Controller DC Motor

Where 𝑛 − 1 < 𝜇 < 𝑛

while the definition of fractional-order integral is
expressed as:
Figure.3 A block diagram of Intelligent FOPID controller

International Journal of Intelligent Engineering and Systems, Vol.11, No.4, 2018 DOI: 10.22266/ijies2018.0831.24
Received: March 31, 2018 245

where
Start
𝑉(𝑠) : Input Signal
𝐸(𝑠) : Error Signal
Initialize particles in searching space with random position
𝐺𝑐 (𝑠): Controller Transfer Function and velocity
𝐺(𝑠) : System or plant (DC Motor)
𝜃(𝑠) : Output Signal Calculate the next position and velocity of each particle

𝑈(𝑠) : Control Signal

Calculate the inertia weight and the fitness values
3.3 Cost function

To evaluate the controller performance, there are Compare and update the personal best position and speed
always several criterions of control quality like
integral of absolute error (IAE), integral of time
Compare and update the goal best position and speed
absolute error (ITAE), integral of squared error (ISE)
and integral of time squared error (ITSE) [27].
A disadvantage of the ISE and IAE criteria Yes
Early stopping
(weight all errors equally and independent of time) is
that they may result in a response with a long settling No
time and relatively small overshoot [27]. To No
overcome this drawback, an integral of time Maximum iteration
weighted absolute error (ITAE) is used in this paper Yes
as fitness function.
Therefore, the controller can be evaluated using Stop: giving gbest as optimal solution
the following performance index:
∞ End
𝐽(𝐾𝑝 , 𝐾𝑖 , 𝐾𝑑 , 𝜆, 𝜇) = ∫0 𝑡|𝑒(𝑡)|𝑑𝑡 (17)
Figure.4 Flowchart of PSO algorithm procedure
J is called as ITAE. It explains indirectly the level
that the controlled object is close to the reference The update formula of velocity and position is
model. Where 𝑡 is the time and 𝑒(𝑡) is the stated by Eqs. (18) and (19):
difference between set point and controlled variable.
𝑣𝑖𝑘+1 = 𝑤𝑖 𝑣𝑖𝑘 + 𝐶1 𝑎(𝑃𝑖 − 𝑥𝑖𝑘 ) + 𝐶2 𝑏(𝑃𝑔 − 𝑥𝑖𝑘 ) (18)
4. PSO and DE optimization methods
In this paper, the FOPID controller is optimized 𝑥𝑖𝑘+1 = 𝑥𝑖𝑘 + 𝑣𝑖𝑘+1 (19)
to achieve the optimal behaviour of the plant. The
optimizer is used to search for the optimal solution of Where:
the FOPID control gain. 𝑣𝑖𝑘 , 𝑥𝑖𝑘 : Velocity and positioning vectors of
particle 𝑖 at iteration 𝑘 respectively.
4.1 Particle swarm optimization algorithm 𝑣𝑖𝑘+1 , 𝑥𝑖𝑘+1 : Modified velocity and position of
PSO is a modern heuristic search method particle 𝑖 at the next iteration 𝑘 + 1
inspired by the social behavior of bird and fish respectively.
schooling. PSO optimization consists of designing 𝑎 ,𝑏 : Random number between 0 and 1
the optimization goal, i.e. the fitness function and 𝐶1 , 𝐶2 : Positive constants
then encoding the parameters to be searched. 𝑃𝑖 , 𝑃𝑔 : Best positions found by particle 𝑖 and 𝑔
PSO exploits a swarm of particles probing respectively
promising regions of the D-dimension search space 𝑤𝑖 : Weight function for velocity of particle 𝑖.
with adaptable velocity. It runs until the stop
condition is satisfied. The best particle’s position In order to design optimum controller, the fitness
gives the optimized parameters for the controller. function are defined in Eq. (17).
The flowchart of a typical PSO algorithm is shown
in Fig.4.

International Journal of Intelligent Engineering and Systems, Vol.11, No.4, 2018 DOI: 10.22266/ijies2018.0831.24
Received: March 31, 2018 246

In this paper a time domain quantities such as

Start maximum overshoot, rise time, setting time,
damping ratio and undamped natural frequency of
the desired dominant closed-loop poles, is used for
Problem to be optimized evaluating the FOPID controller. A set of good
and velocity control parameters P, I, D, 𝜆 and 𝜇 can yield a good
step response that will result in performance criteria
Forming initial group minimization in the time domain. To control the
Crossover plant model the following intelligent tuning methods
PSO and DE parameters are used to verify the
Compute adaptive value performance of the FOPID controller Parameters.
of individual Variation Table 1 summarizes the values of parameters
affecting the optimization. Table 2 displays the
optimization parameters for each optimization
Fulfil terminate No Choose high adaptive method used in this paper.
requirement Value of individuals
Yes 6. Simulation results and discussion
Output All optimization procedures are successful,
producing gains inside the specified bounds and
Figure.5 Flowchart of DE algorithm providing valid solutions for each case. Conventional
methods of controller tuning lead to a rise time,
4.2 Differential evolution overshoot, large settling time and steady state error
of the controlled system. Hence intelligent soft
DE was introduced by Storn and Price in 1996. computing techniques are introduces into the control
It is a stochastic, population based optimization loop. PSO and DE based FOPID tuning methods
algorithm like Genetic Algorithm. But one big have proved their excellence in giving better results
difference is that DE is developed to optimize real by improving the performance indices and the steady
parameters that are non-differentiable, non- state characteristics.
continuous, non-linear, noisy, flat, multi- Performance characteristics of process model
dimensional or have many local minima. As a result, (dynamic response characteristic of the closed loop)
the idea of mutation and crossover are substantially was indicated and compared with the intelligent
different in both the techniques. tuning methods as shown in Figs.6 and 7.
DE has better convergence to global optimum, It can be observed from the Fig.6 that, the DE
more accurate and reduced number of simulations in algorithm method gives much better time domain
comparison to other optimization techniques. performance comparatively to PSO algorithm
Minimizing the cost function generates the especially for maximum overshoot, rise time, and
controller parameters. The error criterion is settling time and also comparatively to [22] study.
considered as the cost function, and the values of
𝐾𝑝 , 𝐾𝑖 , and 𝐾𝑑 are continuously adjusted, until the Table 1. Optimization parameters
error of the closed-loop system is minimum. Eq. Optimization parameters Value
(17) shows the normally considered error criterion Number of Population (NP) 50
in control system to evaluate the performance of
D-dimensional parameter 5
controller.
The flowchart of DE algorithm is shown in Fig.5. Generation number 100

5. Tuning of the FOPID controller using Table 2. Optimization methods parameters

Optimization Process
PSO and DE optimizations
Optimization
Optimization method Value
The speed control loop of DC motor (model Parameter
number PN13KA12C) has been modeled in CR 0.8
Differential Evolution
SIMULINK PSO and DE algorithms has been F 0.9
programmed and implemented in Matlab. Particle Swarm C1 1
Optimization C2 3

International Journal of Intelligent Engineering and Systems, Vol.11, No.4, 2018 DOI: 10.22266/ijies2018.0831.24
Received: March 31, 2018 247

1.5 1.5
FOPID-PSO Setpoint
PID-DE
Output Signal

Output Signal
1 1 FOPID-DE

0.5 0.5

0 0
0 5 10 15 20 5 10 15 20
Time(s) Time(s)

(a) Figure.9 Comparison of PID-DE and FOPID-DE

1.02 FOPID-PSO 1.5 1.5

Output Signal

1.01 ZOOM
1

fitness Value
1 0.5
Best J=0.53 1 2 3 4 5
1 Numbers of generation
0.99

0.98
2.6 2.8 3 0.5
Time(s) 0 20 40 60 80 100
Numbers of generation
(b) Figure.10 Convergence of behaviours of FOPID-DE
Figure.6 Unit step response of closed-loop system: (a)
FOPID-PSO and (b) Zoom 2.6
2.8
fitness Value

2.6
1.5 2.55
FOPID-DE
Setpoint 1 2 3 4 5
Best J: 2.4509 Numbers of generation
Output Signal

1 2.5

0.5 2.45
0 20 40 60 80 100
Numbers of generation
0 Figure.11 Convergence of behaviours of the: FOPID-PSO
5 10 15 20
Time(s)
Figure.7 Unit step response of closed-loop system: As can be seen in Fig.8, the dynamic properties
FOPID-DE (overshoot and settling time) of the controlled
system response obtained from the FOPID-PSO
As can be seen in Fig.7, the dynamic properties controller are much better than those of the PID-
(overshoot and settling time) of the controlled PSO controller.
system response obtained from the FOPID-DE are Fig.9 shows the responses of the FOPID-DE and
much better than those of obtained from FOPID- PID-DE controllers with the ITAE cost function. As
PSO controller. can be seen from Fig.9, the FOPID-DE controller is
We remark also, in all figures, a sluggish initial more robust and has better trajectory tracking than
responses which is due to ITAE index (ITAE index the PID-DE. In order to compare the search
reduces the settling time and absolute error but it has performance of the different intelligent optimization
sluggish initial response). methods, PSO and DE algorithms are applied to the
Fig.8 shows the responses of the FOPID-PSO FOPID controller optimization with ITAE cost
and PID-PSO controllers with the ITAE cost function.
function. Figs.10 and 11 show the fitness values of
different algorithms and as can be seen, the fitness
1.5 value of the FOPID-DE is decreased to 0.53 after 2
Setpoint
PID-PSO generations. On the second hand, the fitness value of
Output Signal

1 FOPID-PSO the FOPID-PSO is decreased to 2.45 after 2

generations. It is clear from Fig.10 that DE
0.5 converges fast initially and requires fewer
generations to reach the optimal point. As can be
0
0 5 10 15 20 seen, through about 2 generations, the DE algorithm
Time(s) provides better convergence. Furthermore, the
Figure.8 Comparison of PID-PSO and FOPID-PSO results obtained here show that the DE algorithm

International Journal of Intelligent Engineering and Systems, Vol.11, No.4, 2018 DOI: 10.22266/ijies2018.0831.24
 Received: March 31, 2018 248

Considering all of the results from the simulation

1.5 1.1
experiments, the FOPID-DE controller can achieve
1 1.05
good performance, noises rejection and robustness,
superior to those obtained with the FOPID-PSO

yout
yout

0.5 1 controller. The FOPID-DE controller has good

tracking performance in comparison with the
0
0 5 10 15 20
0.95
0 5 10 15 20 FOPID-PSO controller. In addition, the FOPID-DE
Time(s) Time(s)
controller enhanced the flexibility and stability of the
Figure.12 FOPID controller with random output noise of PID controller. Furthermore, the implementation of
5 % of the reference signal amplitude
the controller tuning with DE is much easier than
Table 3. Parameters of controllers with the traditional methods because there is no need
𝑲𝒑 𝑲𝒊 𝑲𝒅 𝝀 𝝁 𝑭𝒊𝒕𝒏𝒆𝒔𝒔 for derivative knowledge or complex mathematical
𝑭𝑶𝑷𝑰𝑫 equations. In future studies, Bacterial Foraging (BF-
300 719 120 1.06 1.01 1.106
− 𝑫𝑬 FOPID) will be developed using these optimized
𝑭𝑶𝑷𝑰𝑫 FOPID controllers.
180.4 582 35.7 0.78 1.20 2.4509
− 𝑷𝑺𝑶

can search optimal FOPID controller parameters References

more quickly and efficiently than the PSO algorithm. [1] Y. Bensafia, K. Khettab, and A. Idir, “An
Fig.12 shows the time response characteristics for a Improved Robust Fractionalized PID Controller
step change of the system (2) with random output for a Class of Fractional-Order Systems with
noise of a magnitude equal to 5% of the reference Measurement Noise”, International Journal of
signal amplitude. Intelligent Engineering and Systems, Vol.11,
We remark the FOPID-DE controller give them No.2, pp. 200-207, 2018.
a certain diminution of the noise effect (Absolute [2] D. Valéio and J.S. da Costa, “Tuning of
error =2%). For DC Motor, the Evolution fractional PID controllers with Ziegler–Nichols-
optimization algorithms (PSO, DE) aims to find type rules”, Signal Process, Vol.86, No.10, pp.
optimal value of FOPID controller to minimize the 2771-2784, 2006.
objective function as given in Eq. (12). For ITAE [3] B. Ou, L. Song and C. Chang, “Tuning of
cost function, the parameters of the FOPID fractional PID controllers by using radial basis
controller tuned with two different algorithms and a function neural networks”, In: Proc. of the 2010
comparaison in terms of the cost function are 8th IEEE international conference on control
summarized in Tables 3. From Tables 3, the and automation, pp. 1239-1244, 2010.
parameters of the FOPID-DE controller for the cost [4] R. Sharma, K.P.S. Rana, and V. Kumar,
function ITAE are approximately close to that of the “Performance analysis of fractional-order fuzzy
FOPID-PSO controller. PID controllers applied to a robotic
manipulator”, Expert Syst. Appl. Vol.41, No.9,
7. Conclusion pp. 4274-4289, 2014.
In this work, a new design of intelligent [5] L.Y. Chang and H.C. Chen, “Tuning of
optimization-based model independent controller fractional PID controllers using adaptive genetic
tuning for DC Motor plant has been attempted. All of algorithm for active magnetic bearing system”,
the parameters related to the fractional order PID WSEAS Transactions on Systems, Vol.8, No.1,
controller were determined using PSO and DE. The pp. 158-167, 2009.
robust design of the FOPID controller is difficult to [6] Z. Bingul, “A new PID tuning technique using
compare to the PID controller, since the FOPID differential evolution for unstable and
controller includes more parameters. The parameters integrating processes with time delay”, In: Proc.
of FOPID controller were determined by minimizing of the 11th International Conference on Neural
the ITAE between the output of reference model and Information Processing, pp. 254-260, 2004.
the plant. The robustness of the FOPID-DE [7] J.Y. Cao and B.G. Cao, “Design of fractional
controller was tested in the case of presence noise at order controller based on particle swarm
the reference signal amplitude. optimization”, International Journal of Control,
Automation and Systems, Vol.4, No.6, pp. 775-
781, 2006.
International Journal of Intelligent Engineering and Systems, Vol.11, No.4, 2018 DOI: 10.22266/ijies2018.0831.24
Received: March 31, 2018 249

[8] D. Maiti, S. Biswas, and A. Konar, “Design of a [19] A. Rajasekhar, S. Das, and A. Abraham,
fractional order PID controller using particle ‘‘Fractional order PID controller design for
swarm optimization technique”, Second speed control of chopper fed dc motor drive
National Conference on Recent Trends in using artificial bee colony algorithm’’, In: Proc.
Information Systems, Vol. 30, 2008. of Nature and Biologically Inspired Computing,
[9] A. Alfi and H. Modares, “System identification World Congress, pp. 259-266, 2013.
and control using adaptive particle swarm [20] A. Jalilvand, A.A. Kimiyaghalam, and H. Kord,
optimization”, Applied Mathematical Modelling, ‘‘Optimal tuning of PID controller parameters
Vol.35, No.3, pp. 1210-1221, 2011. on a DC motor based on advanced particle
[10] D. Xue, C. Zhao, and Y.Q. Chen, “Fractional swarm optimization algorithm’’, International
Order PID Control of a DCMotor with Elastic Journal on Technical and Physical Problems of
Shaft: A Case Study”, In: Proc. of the 2006 Engineering, Vol.3, No.4, pp. 10-17, 2011.
[21] A. Biswas, S. Das, A. Abraham, and S.
American Control Conference, pp. 3182-3187,
Dasgupta, “Design of fractional-order PID
2006. controllers with an improved differential
[11] V. Mehra, S. Srivastava, and P. Varshney, evolution”, Eng. Appl. Artif. Intell., Vol.22,
“Fractional-Order PID Controller Design for No.2, pp. 343-350, 2009.
Speed Control of DC Motor”, In: Proc. of the [22] H. Tajbakhsh, and S. Balochian, “Robust
Third International Conference on Emerging Fractional Order PID Control of a DC Motor
Trends in Engineering and Technology, pp. with Parameter Uncertainty Structure”, Int. J. of
422-425, 2010. Innovative Science, Engineering & Technology,
[12] S.K. Gupta and P. Varshney, ‘‘Fractional Fuzzy Vol.1, No.6, 2014.
PID Controller for Speed Control of DC [23] A. Oustaloup, La commande CRONE:
Motor’’, In: Proc. of the Third International commande robuste d’ordre non entier (in
Conference on Advances in Computing and French), Hermes, Paris, 1991.
Communications, pp. 1-4, 2013. [24] J. Sabatier, A. Oustaloup, A.G. Iturricha, and F.
[13] D. Singh, B. Singh, and N. Singh, Levron, “CRONE control of continuous linear
“Performance Indices Based Optimal Tunining time periodic systems: Application to a testing
Criterion for Speed Control of DC Drives bench”, ISA Transactions, Vol.42, No.3, pp.
Using GA”, International Journal of Power 421-436, 2003.
Electronics and Drive System, Vol.4, No.4, pp. [25] A. Carpinteri and F. Mainardi, Fractals and
461-473, 2014. Fractional Calculus in Continuum Mechanics,
[14] A. Ahuja and B. Tandon, “Design of Fractional Springer-Verlag, Wien and New York, pp. 223-
Order PID Controller for DC Motor using 276, 1997.
Genetic Algorithm”, TELKOMNIKA [26] R. Garrappa, “A grunwald–letnikov scheme for
Indonesian Journal of Electrical Engineering, fractional operators of havriliak-negami type”,
Vol.12, No.12, pp. 8140-8151, 2014. Recent Advances in Applied, Modelling and
[15] A. Ahuja and S.K. Aggarwal, “Design of Simulation, Vol.34, pp. 70-76, 2014.
fractional order PID controller for DC motor [27] A.S. Mouayad and S.A. Bestoun, “A new
using evolutionary optimization techniques”, multiobjective performance criterion used in
Wseas Transactions on Systems and Control, PID tuning optimization algorithms”, Journal of
Vol. 9, pp. 171-182, 2014. Advanced Research, Vol.7, No.1, pp. 125-134,
[16] A. M. Concepción, Q. C. Yang, M. V. Blas, X. 2016.
Dingyü, and F. Vicente, Fractional-order
Systems and Controls: Fundamentals and
Applications, Springer-Verlag London,
Advances in Industrial Control series, 2010.
[17] X. Dingyü and Q. C.Yang, ‘‘Fractional Order
PID Control of a DC-Motor with Elastic Shaft:
a Case Study’’, In: Proc. of American Control
Conference, pp. 3182-3187, 2006.
[18] I. Podlubny, ‘‘Fractional-order systems and
PIλDμ controllers’’, IEEE Transactions on
Automatic Control, Vol.44, No.1, pp. 208-214,
1999.
International Journal of Intelligent Engineering and Systems, Vol.11, No.4, 2018 DOI: 10.22266/ijies2018.0831.24