Professional Documents
Culture Documents
Abstract— Continuous Stirred Tank Reactor (CSTR) plays a engineers is to ensure that the reaction will give relevant output
significant role in chemical processes. The continued progress with good efficiency, i.e. maximum profit and minimum cost.
and growth of industry has increased the need for chemical and A continuously stirred tank reactor (CSTR) with concentration
control integration with artificial intelligence. This paper control transfer system is an interesting dynamic phenomenon.
presents the optimal design of PID & FOPID controller based on Because of the immense use of CSTR in process control
soft techniques i.e. Genetic Algorithm and Particle Swarm industries, control engineers are taking much interest in
Optimization (PSO) approach for concentration control of intelligent control mechanism to get desired result. In a CSTR,
isothermal continuous stirred tank reactor (CSTR) model, one or more fluid reagents are introduced into a tank reactor
irrespective of the disturbances introduced in the system. The
equipped with an impeller. The impeller is used to ensure
prelim step is introducing of the mathematical model which gives
relations between state variables in the mathematical way (state
proper mixing of reagents, while the reactor waste is removed
space matrix). Mathematical model is obtained from mass and from the system. The design of chemical reactor depends on
heat balances inside the chemical reactor. Time domain analysis following factors:
will show the performance of different controllers and it is • Size of reactor
noticed that Fractional order PID controller with PSO approach • Nature of reagents
gives better result than the other techniques. • Temperature of reactor
• Pressure of reactor
Keywords—CSTR; PID Controller; FOPID Controller; Genetic
Algorithm; Particle Swarm Optimization (PSO)
• Rate of chemical reaction
• Catalyst
I. INTRODUCTION • Stability of reactor
During the past decades, many design strategies have been II. FRACTIONAL ORDER PID CONTROLLER
developed to comply with regulation and transient
specifications [1]. Numerous control methods such as adaptive The PID controller algorithm have three separate
control, neural control, and fuzzy control have been studied. parameters; the proportional, the integral and derivative values.
But, Proportional integral derivative (PID) controller is still the The proportional gain depends on current error, the integral
most common controller in process industries because of its gain based on the sum of recent errors, and the derivative gain
simplicity & to tune less number of parameter [2]. PID determines rate at which the error has been changing. The
controller implementation has been recommended for the performance of PID controller can be further improved by
control of processes of low to medium order. Controller tuning proper settings of fractional-I and fractional-D actions [5]. In a
is used to determine the controller parameters, which helps to PIλDμ controller, besides the proportional, integral and
get optimum output and minimize the error of the system. It is derivative constants, we have two more adjustable parameters:
difficult to determine the optimal value of PID controller the powers of s in integral and derivative part [6,7].
parameters through conventional tuning method [3]. The (1)
performance of PID controllers can be further improved by
proper settings of fractional-I and fractional-D actions [4]. To
get optimal result, many soft techniques have been employed to where F(s) is transfer function of Fractional Order PID
enhance the performance of controller. Controller. It provides robustness to the plant model against
gain variation, noise and disturbance. Hence, it possesses
A chemical reactor contains two types of chemical better transient response as compared to integer order PID
reactions: exothermic and endothermic reactions, which is Controller. The fractional order PID controller expands the
mainly used for heating and cooling of one or more than one integer order PID controller from point to plane form as shown
chemical in reactor. But for isothermal condition, temperature in figure 1. This expansion adds more flexibility to controller
is constant throughout the process i.e. heat will neither evolve design and we can control our real world processes more
nor absorb in chemical reactor. The main task of chemical accurately as compared to integer order PID controller [8,9].
(2)
Figure 1. Expansion of PID controller from point to plane form
(5)
(6)
(7)
(8)
Figure 2. Temperature control model of CSTR
The elements of state space A matrix is found by: IV. PERFORMANCE INDICES
A performance index is a quantitative measure of the
performance of the system. The performance assessment is
usually used in research environment to minimize the value of
error between the input & desired output.
(9)
A. Integral Absolute Error (IAE):
Use for systems that need to suppress all errors equally.
(16)
The element of state space B matrix is found by:
B. Integral Square Error (ISE):
It is suitable to minimize initial large amount of errors.
(17)
(10) C. Integral Time Absolute Error (ITAE):
It is used to reduce the initial large amount of error to, as
well as to diminish error occurring later in response.
(18)
(11)
D. Integral Time Square Error (ITSE):
D = null matrix (12) It is used for systems where persistent large errors are
problematic.
Using reactor parameters value, we will get:
(19)
V. GENETIC ALGORITHM
Genetic Algorithm (GA) is a class of stochastic global
search method that mimics the process of natural evolution. It
is an optimization technique introduced by John Holland, one
of the founders of evolutionary computation [14]. The
continuing performance improvement of computational
systems has made them attractive for optimize result. The
Converting the state space model to reactor process genetic algorithm starts with no knowledge of the correct
transfer function: solution and depends entirely on reactions from its
environment and evolution operators such as reproduction,
crossover and mutation to get the best solution.
Genetic Algorithm is search and optimization techniques
(13) inspired by two biological principles: the process of “natural
selection” and the mechanics of “natural genetics”. GA
Similarly, the transfer function of input flow disturbance examines not just one potential solution to a problem but a set
due to reactant A & B is: of potential solutions, known as population. The potential
solution in the population is called chromosomes. In each
(14) generation, the fitness of every individual in the population is
assessed, multiple individuals are selected from the current
population (based on their fitness value), and modified to form
(15) a new population. The algorithm terminates when either a
maximum number of generations has been produced, or an
acceptable fitness level has been reached for the population
[15]. The evolution operators of GA are:
A. Reproduction:
During the reproduction phase the fitness value of each
chromosome is evaluated. All selection methods are depends
on the same principal that is giving fitter chromosomes a
larger probability of selection.
B. Crossover:
The crossover operation swaps certain parts of the two
Figure 3. Block diagram of CSTR process model selected strings in a bid to take the good parts of old
chromosomes and create better new ones. A probability of 0% VI. PARTICLE SWARM OPTIMIZATION
means that the offspring will be exact replicas of their parents, Particle Swarm Optimization (PSO) is a population-based
where as 100% means that each generation will be composed evolutionary computational technique, used to determine the
of entirely new offspring. global optimum value of objective function. PSO is based on
C. Mutation: observations of the social behavior of animals such as bird
flocking, fish schooling and swarm theory, which is originally
Using selection and crossover on their own will produce
proposed by Kennedy and Eberhart in [16]. In the PSO
large amount of different strings. However, there are two
algorithm, instead of using evolutionary operator: mutation
main problems with this:
and crossover to manipulate algorithms, a flock of particles are
Because of the initial population, there may not be put into the d-dimensional Search space with randomly chosen
enough diversity in the initial strings to ensure the velocities and positions knowing their best values.
Genetic Algorithm searches the entire problem space.
The swarm is initialized with a population of random
The Genetic Algorithm may converge on sub- solutions. Each particle in the swarm is a distinct possible set
optimum strings due to a poor choice of initial of the unknown parameters to be optimized. Each particle
population. adjusts its flying toward a potential area according to its own
These problems may be overcome by the introduction of a flying experience and shares information among particles [17].
mutation operator. Mutation is the occasional random The goal is to efficiently search the solution space by
alteration of a value of a string position. The mutation rate is swarming the particles toward the best fitting solution
normally low because a high probability of mutation would encountered in previous iterations with the intention to get
destroy fit strings and degenerate the genetic algorithm into a optimum solution through the course of the process.
random search. The process of genetic algorithm is
summarized in figure 4.
TABLE II
Figure 6. Tuning of CSTR Model using Zeigler-Nichols Method Comparison of various tuning methods for CSTR Model
Zeigler-Nichols method is a well known conventional Tuning Rise Peak Peak Settling
technique, which have less complexity. From figure 6, system Techniques Time time Overshoot time
achieves steady state after 3.35 min and have peak overshoot (min) (min) (%) (min)
of 134.6 %. Z-N 0.052 2.346 134.65 3.358
Method
The response of CSTR Model using GA having Genetic 0.106 0.268 25.39 1.204
parameters: Population size=100, iteration=100 and ITAE as tuned PID
fitness function (see figure 7). Through genetic algorithm, PSO tuned 0.107 0.271 25.14 1.164
system takes 1.204 min through PID controller and 0.368 min PID
through FOPID controller to achieve steady state. Genetic 0.135 0.275 5.15 0.368
tuned FOPID
PSO tuned 0.136 0.276 3.80 0.321
FOPID
VIII. CONCLUSIONS
This research explains a design of PID & Fractional order
PID (FOPID) controller by using Genetic Algorithm & PSO
technique to control the product concentration of isothermal
CSTR process model. Time response analysis shows that
FOPID controller with PSO provides a good result as
compared to conventional PID controller. The demerit of a
single stage CSTR with conventional tuning method is that it
can be relatively wasteful on product during start of reaction.
FOPID controller with soft techniques shows satisfactory
result to achieve steady state in short time. Because of the
importance of soft techniques in control industries, researchers
are taking more interest in this field.
Figure 7. Tuning of CSTR Model using Genetic Algorithm
REFERENCES [13] W.L Luyben, Process Modelling, Simulation and Control for Chemical
Engineers, McGraw-Hill, New York, 1989.
[1] K. J. Åström and T. Hägglund, PID Controllers: Theory, Design, and
Tuning, 2nd ed., Instr. Soc. of America, 1995. [14] Michael Negnevitsky, Artificial Intelligence: A Guide to Intelligent
System, 2nd Edition, Pearson Education.
[2] J. G. Ziegler and N. B. Nichols, “Optimal settings for automatic
controllers,” Trans. ASME, vol. 64, pp. 759-768, 1942. [15] L Davis, A Handbook of Genetic Algorithms, 2nd Edition, Van Nostrand
Reinhold.
[3] Yun Li, Kiam Heong Ang and Gregory C Y Chong, “PID control system
analysis and design,” IEEE Control Systems Magazine, pp. 32-41, 2006. [16] J.Kennedy and R. Eberhart, “Particle Swarm Optimization” Proc. IEEE
International Conference on Neural Networks, vol.4, pp. 1942-1948,
[4] Podlubny, “Fractional-Order Systems and PIλDμ Controllers”, IEEE 1995.
Trans. on Automatic Control, vol. 44, pp. 208–214, 1999.
[17] M Clerc and J Kennedy, “The Particle Swarm: Explosion, Stability, and
[5] Petras, “The fractional order controllers: Methods for their synthesis and Convergence in Multi-Dimension Complex Space” IEEE Transaction on
application,” Journal of Electrical Engineering, vol 50, no. 9-10, pp. Evolutionary Computation, vol. 16, pp. 58-73, 2002.
284-288, 1999.
[18] Riccardo Poli, James Kennedy and Tim Blackwell, “Particle swarm
[6] YangQuan Chen, Ivo Petras and Dingy u Xue, “Fractional Order Control optimization-An overview,” Springer Science, Business Media, LLC.
- A Tutorial,” American Control Conference, USA, June 10-12, 2009.
[19] Yanzhu Zhang and Jingjiao Li, “Fractional-order PID Controller Tuning
[7] K.Sundaravadivu, B.Arun and K.Saravanan, “Design of Fractional Order Based on Genetic Algorithm,” IEEE, pp.764-767, 2011.
PID Controller for Liquid Level Control of Spherical Tank,” IEEE
International Conference on Control System, Computing and [20] S M Giriraj Kumar, R Jain, N Anantharaman, V Dharmaingam and K M
Engineering, pp.291-295, 2011. M Sheriffa Begum, “Genetic algorithm based PID controller tuning for a
model bioreactor,” Indian Chemical Engineer, vol. 50, no. 3, pp. 214 –
[8] Mithun Chakraborty, Deepyaman Maiti and Amit Konar, “The 226, 2008.
application of stochastic optimization algorithms to the design of a
fractional-order PID controller,” IEEE, 3rd ICIIS, pp. 1-6, 2008. [21] Liu Y, Zhang J and Wang S, “Optimization design based on PSO
algorithm for PID controller” 5th World Congress on Intelligent Control
[9] Deepyaman Maiti, Ayan Acharya, Mithun Chakraborty and Amit Konar, and Automation, Vol. 3, pp. 2419-2422.
“Tuning of PID and PIλDδ controllers using the integral time absolute
error criterion,” IEEE, ICIAFS 2008, [22] J.Y. Cao and B. G. Cao, “Design of Fractional Order Controller Based
on Particle Swarm Optimization”, International Journal of Control,
[10] Mohammad Ali Nekoui, Mohammad Ali Khameneh and Mohammad Automation and Systems, vol. 4, no. 6, pp. 775-781, 2006.
Hosein Kazemi, “Optimal Design of PID Controller for a CSTR System
[23] B. Nagaraj and P. Vijayakumar, “A Comparative Study Of PID
Using Particle Swarm Optimization,” IEEE International Power
Electronics and Motion Control Conference, pp. 63-66, 2010. Controller Tuning Using GA, EP, PSO AND ACO,” Journal of
Automation, Mobile Robotics & Intelligent Systems, Vol. 5, no. 2, pp
[11] O. Levenspiel, Chemical Reaction Engineering, 3rd ed., John Wiley and 42-48, 2011.
Sons, New York, 1999.
[24] Rania Hassan, Babak Cohanim and Olivier de Weck, “A Comparison of
[12] MA Rahman, “Chalcone: A Valuable Insight into the Recent Advances Particle Swarm Optimization and the Genetic Algorithm,” American
and Potential Pharmacological Activities,” Chemical Sciences Journal, Institute of Aeronautics and Astronautics, pp. 1-13, 2004.
pp. 1-16, 2011.