You are on page 1of 6

DESIGN AND ANALYSIS OF FO PID

CONTTROLLERS IN NON-LINEAR PROCESS


V Ajantha1, M Poornima1, G.Balasubramanian1,K.Ramkumar1,R.Kalpana2
1
School Electrical and Electronics Engineering, SASTRA University, Thanjavur, India
Srinivasa Ramanujan Centre, Kumbakonam, India
ajantharaman@yahoo.com
m.poornima05@gmail.com
balu_eie@eie.sastra.edu

Abstract This paper presents the analysis of Fractional Order II. FRACTIONAL CALCULUS BASICS
Proportional Integral and Derivative (FO PID) controller for Fractional Calculus is a field of mathematic study
Non-Linear System. It is inferred that conventional tuning that deals about definitions of the calculus integral and
methods gives unsatisfactory response when used for processes derivative operators which are real numbers [3]. Since Integral
experiencing negative destabilizing effects due to strong
Order Calculus (IOC) is always had been self-sufficient and
nonlinearities. Henceforth the goal of this research is to obtain
optimal parameters using FO PID controllers .It shows better
due to complexity involved in fractional order calculus it had
disturbance rejection and has more degree of freedom in not been used much in practice. Due to this fact, it does not
designing the controllers. In Non-Linear systems like conical have a fully applicable geometrical or physical interpretation
tank, the level changes with respect to cross sectional area, thus in practice. The Fractional calculus involves generalizing
making it highly non-linear. Thus the controller designing for integration and differentiation calculus to non-integer order. It
conical tank is tedious and complex process due to nonlinear uses the operator aDt where a and t denotes the limits of the
behavior. On comparing with conventional counterpart this operation [4]. FOC is used for generalizing Integer Order
method provides highly stable and reliable controller. Control (IOC) to a real or complex order.
Keywords FO PID, Conical Tank, FOMCON Toolbox, d
Feedback FO PID, Feedforward plus Feedback FO PID
I. INTRODUCTION dt 0

The most commonly used and universally accepted aD t 1 0 (1)
control algorithm is ZN PID in industries. The significant t 0
reason for using conventional PID controller in chemical d
process industry is that it is relatively simple ,easier to design

0

and has wide operating range .In the recent years, real physical
model of the systems are characterized by fractional order Where is a real number.
differential equations i.e., these equations involves non-integer The Riemann-Liouville integral is a commonly used definition
real order integrative and derivatives [1]. The conventional for Fractional calculus
PID controllers follows quarter damping ratio resulting in poor x
1


robustness of the system. In addition to that the controller
I f ( x) f (t )( x t ) 1 dt (2)
parameters are designed using S-shaped step response but for
more complicated system this information is not adequate to
( )
design robust controller. Most of the systems are not actually Riemann-Liouville differential equation is
linear but they are normalized as linear models around a
dm 1 f ( )
nominal operating point. At normal operating point, it does not t



show real time changes with respect to process. In industries D f (t ) m 1 m
d , m Z (3)
controller is the brain of a process. It requires wide range of dt (m ) 0 (t )
specification to achieve satisfactory results. Hence to obtain
optimal parameters numerous methods has been proposed. On Grunwald-Letnikov integral is given by
par with conventional PID controller, this proposed technique t
h
( m)
has two more parameters which give extra degrees of freedom
to optimize the process [2]. Thus it provides more flexibility in D
lim h
h 0


m0 m!( )
f (t mh) (4)
design than conventional PID controllers. The main aim of
These equations form the basis of fractional calculus and can
this work is to reduce the maximum overshoot, increases the be used to obtain differential equations for fractional order
robustness and to achieve better disturbance rejection. system.

978-1-4673-9925-8/16/$31.00 2016 IEEE 285


III. FO PID CONTROLLERS
The FO PID controller is generalized form of the
integer order PID controller when it is plotted from point to
plane. It provides much more flexibility and extra two degrees
of freedom in PID control design. The transfer function of
such a controller is
KI
GC (s) K P K D s (5)
s
Where
P - proportional gain
I- integration gain
D - differentiation gain Fig.2. Schematic diagram of conical tank
- real order of integrator
- real order of differentiator TABLE I.TRANSFER FUNCTION FOR DIFFERENT ZONES
The FO PID controller can expand in point to plane it is (s) d(s)
Zone Height(cm) K
shown in Fig.1 (Sec) (Sec)
1 12-24 1 27.66 0.958

2 24-30 0.75 21.79 0.9

3 30-36 0.66 6.13 3.1208

4 36-42 0.66 11.85 1.01

Fig.1. General form of FO PID controller

V. FOMCON TOOLBOX
These controllers show minimal steady state errors and
robustness to high frequency noise and also to variations in the Fractional Order Modeling and Controlling
gain of the output. It is also less sensitive to parameter (FOMCON) is a whole package that offers sophisticated tools
changes while comparing with conventional controllers. for designing Fractional order controllers in control system
design. The goal of this toolbox is to extend classical control
IV. NON LINEAR SYSTEM schemes to Fractional order calculus [6]. The FOMCON
The Conical tank is a non linear system since its level toolbox is superset of existing mini toolbox FOTF Fractional
changes with change in area of cross section of the tank Order Transfer Function. It also supports functions existing in
.Conical tank setup shown in fig.2 is a SISO (Single Input CRONE and NINTEGER toolbox. CRONE Toolbox is used
Single Output) system which is tapered towards its base, to for designing Fractional order controllers for MIMO process
ensure higher drainage efficiency. The measured variable is while NINTEGER is used for SISO process. Henceforth
height of the tank h(cm) and manipulated variable is inflow FOMCON toolbox bridges the gap between CRONE and
rate Q . The conical tank inflow rate is monitored by opening NINTEGER toolbox. Thus FOMCON toolbox supports
and closing of the pneumatic control valve. The motor pumps designing of FO PID controllers for both SISO and MIMO
water into conical tank which in turn controlled by control process. The optimization process involves number of
valve [5]. The input from user through PC is converted using iterations even up to 1000 times. Hence this toolbox is used to
Data Acquisition (DAQ) to voltage, then voltage to current obtain precise optimal parameters. To make optimization even
using V/I converter followed by I/P which converts current easier and sophisticated, number of optimization algorithms
into corresponding pressure. This pressure acts on the has been already feed into the toolbox. It also supports
diaphragm of the control valve. The level in the conical tank Graphical User Interface instead of tedious programming.
can be obtained using sensor Differential Pressure Transmitter Various functions available are fid(), pfid(), fpid_optimize()
(DPT).The DAQ present in the conical tank is ADAM and iopid_tune , fpid_optim(). The relationship of FOMCON
module. The current and pressure range is about 4-20mA and with other toolboxes can be depicted in Fig 3
3-15 psi respectively. The system is nonlinear and hence, it is
divided into four zones to calculate the transfer function. The
transfer function is obtained using Sundaresan Krishnasamy
method as shown in table I

Fig 3. Relationship of FOMCON with other toolbox

286
A. Active-Set Algorithm
This toolbox consists of two modules namely It is based on dividing problem into sub-problem and solution
1. Control System Toolbox of that sub-problem forms the basis of iteration.
2. Optimization Toolbox 1. For inequality constrained quadratic programming

It can be used for both time domain and frequency domain
analysis of the system. In this work, two commands (QP)I min 1 2 X T Qx qT x
x
(fpid_optim and iopid_tune) are used in workspace. Among
the various supported algorithms for optimization in FPID Ax a (6)
Optimization Toolbox, Nelder-Mead method or Downhill Bx b
method is used in this paper. In the Optimization GUI, 2. Assume vector as X k and active index set as Ak

Ak j | bTj x k b j 0, j 1, 2......m2 |
respective transfer function is fed. Optimal design parameters
are obtained from the fractional as well as integer controllers (7)
respectively as shown in fig 4 and 5. 3. Set k= 0 and check for no convergence.
4. If convergence is no compute g k Qx k q (8)
k
5. Using KKT equations obtain step direction d and
Langrage multipliers for inequality constraints and
active inequality constraint .
k

6. When d 0 and then if


k
k 0 , it is optimal
point and obtain k k 1
Else, the index j0 and solve for (QP) I

d
min 1 2 d T Qd
g
k

T
d
aiT d 0 (9)

bTj d 0; j Ak
7. If d 0 then calculate step length
k
k where

bj bj d
T k


k min 1, T k
j Ak , bTj d k 0 (10)

bj d

8. Update x
k 1
xk k d k (11)
9. Update active set index if k 1 and then

k 1 k or else kU { j0 }
k 1

Fig. 4 fopid panel block diagram 10. Update k k 1


Since applying this algorithm manually is tedious and
erroneous , using FOMCON toolbox helps to optimize easier.

VI. FEEDBACK FO PID CONTROLLERS


Feedback loop is formed when a part of plant output
is redirected back as input to plant.Thus it forms a closed loop
system.Hence it is also called as closed loop system[8]. In
feedback system ,the process is a cause and effect relation
which operates depending on the input variables.It is usually
used to control thermal,chemical and mechanically process
plants.Here the distrubance is passed through the system then
an corrective action is taken.The design of feedback system is
based on root locus or frequency response for classical
controllers,state space approach for Modern controllers and
Fuzzy logic for Advanced controllers.It can be positive or
negative feedback control loop. In Negative feedback, the
output is 180 o out of phase inaccordance with the input signal.
Fig. 5 NIPID panel block diagram Most commonly negative feedback is used in the industries.A

287
system can be made to desensitize to unwanted fluctuations by As analyzed from the graph results that the conventional PID
using negative feedback. Meanwhile in positive feedback loop suffers from maximum overshoot and poor robustness due to
,the output is in phase with the input.Feedback loop drives the quarter damping ratio property. However it can be infered that
controlled variable to desired output and takes corrective FO PID controllers shows minimised overshoot and robust
action at once when output deviates. In other words the action stablity.The oscillations is also much reduced in FO PID
is taken only after the distrubance affects the system once.The controllers.The designed FO PID controller parameters also
Simulation for designing FO-PID controller is done using works effectively in rejecting the disturbance which is applied
MATLAB Simulink as shown in Fig 6. at time=180sec. Fig 9 shows the result for disturbance
rejection

Height (cm)
Fig 6 Block for a set point change in Feedback controller
Time(s)
The results for the Simulink block daigram for zone 1 and 2 is Fig 9 Disturbance rejection for zone 1
shown in Fig 7 and 8. From the graph it is analyzed that FO From the graph, it is clearly shown that FO PID controllers
PID controller settles faster .Moreover like conventional show better disturbance rejection than conventional PID
controller it does not suffer from quarter damping ratio controllers. The performance index values are tabulated as
property. shown in Table II.

TABLE II.PERFORMANCE INDEX FOR FEEDBACK FO PID


ZN PID FO PID
ZONE ISE ITAE ISE ITAE
Height (cm)

1 12.67 549.6 0.934 21.95

2 8.151 195.3 0.396 3.37


3 2.738 14.3 0.389 3.98
4 63.01 1.684 1.684 14.26

Height (cm)
From the table, it can be inferred that the FO PID controller is
more efficient in disturbance rejection and error has been
minimized drastically.
Time(s)
Fig 7 Set-point tracking for zone 1 VII. FEEDFORWARD AND FEEDBACK FO PID
CONTROLLER
The main drawback of feedback controller is that no
corrective action is taken until a deviation upsets the
controlled variable. Moreover it does not give predictive
control action for compensating the effects of unknown and
measurable disturbances [9]. The main aim of feedforward
control is to measure the disturbance and to take corrective
action before it upsets the stability of the system. Therefore
the feedforward loop acts in a predefined way. The
feedforward with feedback control is widely used in industrial
process. When feedback controllers are used in nonlinear
process like conical tank, it does not give satisfactory results.
Therefore significant improvements in control can be achieved
Time(s) by adding feedforward control with feedback control. The
feedforward control offers potentiality to take predictive
Fig 8 Set-point tracking for zone 2
action before the disturbance affects the load on the par
traditional feedback loop helps in tracking the set point soon.

288
The Simulation for designing feedforward and feedback FO
PID controller is done using MATLAB Simulink as shown in
Fig 10.

Height (cm)
Time(s)
Fig 13 Graph representing Varying Set Point change for Zone 1

TABLE III.PERFORMANCE INDEX FOR FEEDFORWARD PLUS


FEEDBACK FO PID
ZN PID FO PID

Fig 10 Simulink diagram for Feedforward and feedback Controller ZONE ISE ITAE ISE ITAE

The simulation results for setpoint tracking in zone 1 and 2 1 12.39 516.6 0.927 19.09
using feedforward and feedback controller are shown in Fig 11
and 12. 2 8.08 214 0.395 1.7893
From the graph it can be infered that in FO PID controllers
rise time gets accelerated and thus settles faster, thereby 3 2.64 10.52 0.389 4.05
enhancing the controller performance. It is observed that in
time domain analysis, FO PID controllers settles earlier thus 4 4.25 50.61 1.647 8.928
outperforming the conventional controllers.
From the performance index values as tabulated in TABLE III
shows that FO PID controller is more efficient than
conventional PID controller.

The comparative analysis of feedback controller with


feedforward and feedback controller is illustrated in the fig14.

Time(s)
Fig 11 Set Point Tracking for feedforward and feedback controller for Zone 1

Fig 14 Simulink diagram of FO PID Feedback Vs Feedforward plus Feedback


Controller for Disturbance Rejection

Time(s)
Fig 12 Set Point Tracking for feedforward and feedback controller for Zone 2

The proposed controller can also track precisely the varying


setpoint changes as shown in fig. 13.

289
index values that Feedforward plus Feedback FO PID
controller has better disturbance rejection .Thus from the
proposed paper it can be found that for non linear process
Height (cm)

like conical tank FO PID controllers gives superior results


than the conventional ZN PID controllers.

IX. REFERENCES

[1] Serdar Ethem Hamamci, An Algorithm for Stabilization of


Time(s)
Fractional-Order Time Delay Systems Using Fractional-Order PID
Fig 15 Disturbance Rejection of FO PID Feedback Vs Feedforward plus
Controllers, IEEE Vol. 52, No. 10, October 2007
feedback controller in zone 1
[2] I. Podlubny, Fractional-order systems and PI D -controllers,
IEEE vol. 44, no. 1, pp. 208214, 1999
[3] Sabura Banu U Multivariable Centralized Fractional Order PID
Controller tuned using Harmony search Algorithm for Two
Height (cm)

Interacting Conical Tank Process SAI Intelligent Systems


Conference 2015.
[4] Y.Q. Chen, I. Petra and D. Xue, Fractional Order Control - A
Tutorial , American Control Conference, pp. 1397-1411, 209.
[5] P. Sowmya, G.Balasubramanian , S.Rakeshkumar and
K.Ramkumar A Genetic Algorithm Tuned Fuzzy Controller
for a Nonlinear Process Journal of Applied Sciences, 14:1576-
1581.
[6] Aleksei Tepljakov, Eduard Petlenkov and Juri Belikov
FOMCON: Fractional-Order Modeling and Control Toolbox for
MATLAB International Conference on Mixed Design of
Time(s) Integrated Circuits and Systems, June 11
Fig 16 Disturbance Rejection of FO PID Feedback and feedforward plus [7] Nelder-Mead Algorithm Simplex Animation
feedback controller in zone 2 http://optlabserver.sce.carleton.ca/POAnimations2007/NonLinear

TABLE IV.PERFORMANCE INDEX FOR FEEDFORWARD AND 7.html


FEEDBACK FO PID Vs FEEDBACK FO PID CONTROLLER [8] Hwang Chyi, Jeng-Fan Leu , Sun Yuan tsay Anote on time
Domain simulation of feedback fractional order controllers
FO PID(FB) FO PID(FF+FB) Automatic control conference, IEEE Transactions on Vol:47 Issue
4, August 02.
ZONE ISE ITAE ISE ITAE [9] E. J. Adam, J. L. Marchetti Designing and tuning robust
feedforward controllers Science Direct March 14
1 33.83 4.76 0.9209 146.3
2 3.432 2884 0.395 33.87
3 12.16 6048 0.4029 128.2
4 150.4 11570 2.93 1048

From this analysis, it is inferrred that FO PID feedforward


plus feedback controller shows better distrubance rejection.
For non linear process like conical tank, it takes effective
control action to reject both anticipated as well as
unanticipated distrubances.

VIII. CONCLUSION
The mathematical model of conical tank has been
built in Matlab Simulink. Then the conventional ZN PID and
FO PID controllers are designed for optimal parameters. The
simulation results shows that the FO PID controllers
outperforms conventional ZN PID controllers by showing
lesser rise time and faster settling time.FO PID tracks the set
point with minimum oscillations and in smooth transition
manner. The performance index ISE and ITAE values are
calculated for ZN PID and FO PID controller. FO PID
controller has lower error values. Feedforward plus Feedback
FO PID controllers take predictive action and effectively
reject the disturbance. It can be inferred from performance

290

You might also like