Available online on www.shts.org.rs, www.cee.

rs
Comput. Ecol. Eng. 4(1)(2008)16-22 ISSN 1452-0729 Scientific paper

A Reactor Fuzzy Control System

____________________________________________________________________________________

Abstract This paper studies the feasibility of using a nonlinear fuzzy logic control for chemical processes exemplified by a continuous stirred tank reactor. The inputs were reactant rates and initial concentrations. The outputs were product rate and concentration of the species. The dynamic response of the control loop to a random disturbance with varying amplitude of the reactants flow was examined. Some inverse models were investigated. The fuzzy logic controller improves quality control, determines optimum set points, updates planning models, and troubleshoots day-to-day operating problems. This capability also allows the fuzzy controller to adapt a system which varies slowly over time. The main contribution of this paper is the nonlinear fuzzy control of the chemical reactor.
Keywords: Fuzzy control, reactor dynamic response, set points, random disturbance

_______________________________________________________________________________

1.Introduction Recently, fuzzy logic controllers have been successfully applied to a wide range of industrial processes as well as consumer products, and show certain advantages over the conventional PI and PID controllers. On the other hand, although, fuzzy controllers have been extensively studied in control engineering( Zadeh,1983;Corn,1993;Bulsary et.al.,1993;Savkovi-Stevanovi et.al.,2008; Savkovi-Stevanovi1992;). there are still rather few theoretical proof that can explain why fuzzy logic controller can achieve better performance. The fuzzy control system in this paper based on qualitative fuzzy variables was used for reducing the number of

membership parameters and input/output rules as much as possible. 2. Fuzzy logic variables and functions Unlike binary logic, fuzzy system do not restrict a variable to be a member of a single set, but recognize that a given value may fit to varying degrees, into several. It incorporate the imprecision inherent in many real world systems, including human reasoning, by allowing linguistic variables classification such as big, high, slow, medium, near zero, or too fast. Fuzzy systems operate by testing variables with IF-THEN rules, which produce appropriate responses. Each rules then weighted by a degree of fulfillment of the rule invoked, this is a number between 0

17

Ivanovi-Kneevi et.al. Comput. Ecol. Eng. 4 (2008)16-22

A reactor fuzzy control system

and 1, and may be thought of as probability that a given number is considered to be included in a particular set. A wide variety of shapes is possible fulfillment functions, with triangles and trapezoids being the most popular. Fulfillment functions for this study were of the form:

a A + b B + c C = 0 where a

=b/a=-1 and c =c/a=1 By setting material and energy balances, and using the relationship for specific chemical reaction rate constant k = k 0 e E / RT , one obtains. Total material balance:

=-1,a

( x, m, s, p) = exp(( x m ) / s) ) (1)
where m, s, and p are user chosen parameters and x is the values to be tested. The function was chosen because of its flexibility, by changing m, s, and p whole families of different functions can be obtained. For p=2 this is an non normalized Gaussian density with mean m, and standard deviation s. A sample of the functions obtains by varying the p parameter. The system operates by testing rules of different types. IF xi is high AND yi is low THEN uij is slow or fast. The degree of fulfillment for such a rule in this study was chosen to be the minimum of the degrees of fulfillment of the antecedent clauses. The total output of the control system is calculated as weighted sum of the responses to all n rules outputs. 3. The case study A chemical stirred tank reactor was used for an irreversible reaction (SavkoviStevanovi et.al.,2005; Savkovi-Stevanovi et.al.,2003)(Figure 1).

F0 F =

dV dt

(3)

Overall heat transfer between process at the temperature T and jacket fluid at the temperature Tj: Q = UAH (T T j ) (4) Components balances:

F0cA0 F cA Vk0eE / RT cAcB = F0cB0 F cB Vk0eE / RTcAcB = Vk0 e E / RT c AcB F cC =

d (VcA ) (5) dt

d (VcB ) (6) dt
(7)

d (VcC ) dt

c p (F0 AT0 + F0BT0 FT) Hr (Vk0eE / RT c A cB ) +

and jacket energy balance,

J FJ C pJ (TJ 0 TJ ) + UAH (TJ T ) =

aA + bB k

cC

(2)

It is assumed the pseudo first order reaction is carried out in a single perfectly mixed reactor as shown in Figure 1. A dynamic model for the reactor control can be obtained using of the first principle modeling approach. Eq.(2) can be restarted in the following equivalent form

where F0 is inlet flow, F is outlet flow, cA0 and cB0 are the inlet concentration, cA, cB, cC, are the current species concentration, V is volume of the liquid reaction mixture, T is the temperature the liquid reaction mixture, k0 is acceleration factor, E is energy activation, R universal gas constant and TJ is the temperature of the heating fluid, FJ flow rate of heating fluid. Q is heat of transfer rate, AH is heat transfer area, U is overall heat transfer coefficient, is density of the liquid reaction mixture, J is density of heating fluid, CpJ is capacity of the heating

A reactor fuzzy control system

Ivanovi-Kneevi et.al. Comput. Ecol. Eng. 4 (2008)16-22

18

fluid and Cp is capacity of the reaction mixture at the constant pressure. The operation parameters of the chemical stirred tank reactor are given in Table 1. Table 1.The steady state operation parameters of the chemical stirred tank reactor
TC

Name Reaction volume V Outlet flow rate F Current temperature T Current concentration cA Current concentration cC Current concentration cB

Value 1.00m3 2.00 m3/s 87.10 C 0.008 mole/m3 0.880 mole/m3 0.005 mole/m3

F0A cA0

F0B cB0

T0

Reactant A initial flow 2.000 m3/s rate F0A Reactant B initial flow 1.000 m /s rate F0B Current temperature TJ 37.5 C
3

FJ TJ0 F, cA, cB, cC, T

V,T cA cB cC

FJ TJ

The step disturbance was incorporated to the inlet flow rate to the reactor and cooling water stream flow. The temperature can be selected as a manipulative variable. The control objective was to operate tank as close to the required product composition values. Hydraulic relationship between reactor holdup and the flow out of the reactor is defined. It is assumed that a level controller changes the outflow in direct proportion to the volume in the reactor. The outflow increases as the volume builds up in the reactor and decreases as the volume drops. The outflow is shut off completely when the volume drops to a minimum value Vmin. F=KV (V-Vmin) (10)

Figure 1. Scheme of the chemical stirred tank where KV is proportional constant. The level controller is a proportional feedback controller. 3. The fuzzy control model The qualitative variables the inlet and outlet reaction mixture flow, the inlet and outlet concentration and the reaction temperature as well as temperature in the jacket were considered. The input variables are the initial reactant flow rate and the initial reactant composition, the cooling water flow in the jacket (which assumed constant) and output variables reaction

19

Ivanovi-Kneevi et.al. Comput. Ecol. Eng. 4 (2008)16-22

A reactor fuzzy control system

mixture flow, product composition, temperature of the reactor mixture and temperature in the jacket are identified. The process can be approximated by a first order plus dead time function based control by conventional proportional integral law. The qualitative model for systematic cause-event analysis was made, and variables discrete state were defined. Initial reactant flow (low, medium, high) product flow(low, medium, high) cooling fluid flow (low, medium, high) reaction volume level(low, medium, high) concentration (increasing, slow increasing, normal, slow decreasing , decreasing). temperature (increasing, slow increasing, normal, slow decreasing, decreasing) From equation (1) for p=1 the meaning of the linguistic values is defined by LLRR-type membership.function:
LL ( B1 x ) /( B1 a ) IF x < B1 (11)

As manipulative variables were considered reactor outlet flow and reaction temperature T. Disturbances were made in inlet rate F0 and cooling water inlet flow FJ. As controlled variables were considered concentration cC. F(t-t) F0A (t) V(t) F0B (t) cA(t- t) F(t) cC(t)

Figure 2. An inverse model to reactor outlet flow loop

FJ(t) TJ(t)

( x) = 1

IF

B1 < x > B 2 IF

T(t-t) F(t) cC(t) cC(t-t) F0A(t) cB(t) V(t) F0B(t) T(t)

RR ( x B 2 ) /( B 2 )

x > B2

where x is current fuzzy value. B1 and B2 are the maximum values of the fuzzy number, and are the left and right spreads, and LL-RR are the appropriate chosen functions. The control system can be defined with two inputs and one outputs. A systematic cause event analysis is defined by fuzzy rules according to the equation: IF ( F1 = B1 ) V is Vmin AND (F2 = B2) THEN (9)

Some fuzzy production rules have shown in the section 5 for extension control system.

Figure 3. An inverse model to temperature loop

A reactor fuzzy control system

Ivanovi-Kneevi et.al. Comput. Ecol. Eng. 4 (2008)16-22

20

4 Some rules sets for reactor control system An advanced fuzzy controller was generated using set rules as following: Rule set number 1: IF outlet flow F(t) is high AND inlet flow F0A(t) is high AND inlet flow F0B(t) is high THEN product concentration cC(t+t) is normal. Rule set number 2: IF outlet flow F(t) is medium AND inlet flow F0A(t) is low AND inlet flow F0B(t) is medium THEN product concentration cC(t+t) is slow decreasing. Rule set number 3. IF outlet flow F(t) is medium AND inlet flow F0A(t) is high AND inlet flow F0B(t) is low THEN product concentration cC(t+t) is slow decreasing. Rule set number 4. IF outlet flow F(t) is low AND inlet flow F0A(t) is high AND inlet flow F0B(t) is high THEN product concentration cC(t+t) is increasing. Rule set number 5: IF outlet flow F(t) is medium AND inlet flow F0A(t) is low AND inlet flow F0B(t) is low THEN product concentration cC(t+t) is decreasing. Rule set number 11: IF temperature TJ is low THEN temperature T is low AND product concentration cC(t+t) is decreasing. 5. Results and discussion The dynamic response of the control loop to a random disturbance with varying amplitude of the reactants flow rates was examined.

The process inputs and outputs are considered during the simulation. The random disturbance were used for control. The investigation is carried out during a time period from 0 to 1200s. 3.0 F0A(t) 2.0

1.0 0.0 120 Sampling time instant Figure 4. Random disturbance in the reactant A flow rate F0A(t) Figure 4. shows disturbance in the reactant A flow rate F0A(t). In Fig. 5 disturbance in the reactant B flow rate F0B(t) has shown. 2.0 F0B(t) 30 60 90

1.0

0.0 0.0 120 Sampling time instant Figure 5. Random disturbance in the reactant B flow rate F0B(t) 30 60 90

21

Ivanovi-Kneevi et.al. Comput. Ecol. Eng. 4 (2008)16-22

A reactor fuzzy control system

40.0 T 37.0

1.00 cC(t) 0.88

36.0 0.0 120

30

60

90

Sampling time instant Figure 6. Dynamic response of the temperature T to disturbance in inlet flow rate F0A(t) The obtained control results are shown in Figure 6. Figure 7. Figure 6 shows response of the temperature T to the disturbance in inlet flow rate F0A(t). Response of the product composition to the disturbance in the inlet flow rate F0A(t) is shown in Figure 7. 6. Conclusions

0.80 0.0 120

30

60

90

Sampling time instant Figure 7. Response of the product concentration for disturbance in inlet flow rate F0A(t) Results of this investigation can be applied in the other domain such as pharmaceutical engineering. Acknowledgment. The author wish to express her gratitude to the Fund of Serbia for financial support. Notation This paper was studied fuzzy modeling for chemical reactor control. An advanced fuzzy control model was derived. The fuzzy controller application was illustrated to the chemical stirred tank reactor with irreversible the pseudo first order reaction. Dynamic responses were studied for random disturbance. The developed model based on fuzzy logic, for stirred tank reactor, performed well for the wider operating ranges considered and can be used with confidence for the on-line measurement. A fuzzy logic controller was illustrates to successfully control the system and to exhibit desirable robustness properties. The system reaches the set point faster, with less overshoot, hence the setting time is the shortest, especially for instable region. Results of this investigation can be applied in the other domain such as pharmaceutical engineering. A-first reactant AH-heat transfer area, m2 B- second reactant B1-maximum limit B2-maximum limit C-product c - current concentration, mole/ m3 Cp heat capacity, J/mole degree E -energy activation, J/mole F-flow rate, m3/s L-left appropriate function k-specific chemical rate constant,s-1 k0 -acceleration factor Q - heat of transfer rate, J/s R-universal gas constant, J/mole K T-reaction mixture temperature U-overall heat transfer coefficient, J/ m2 s degree u-manipulative variable V-volume, m3 w- weight factor

A reactor fuzzy control system

Ivanovi-Kneevi et.al. Comput. Ecol. Eng. 4 (2008)16-22

22

e-error Greek Symbols e- change of error (x)- membership function t- time interval,s Index A-reactant 1 B-reactant 2 C-product H-heat 0-inlet condition J-jacket min-minimum

Zadeh,L.A.(1983) Fuzzy Sets Systems, 11, 1199. IZVOD REAKTORSKI REGULACIONI SISTEM U USLOVIMA NEIZVESNOSTI Maja Ivanovi-Kneevi and Sneana Krsti Tehnoloko-metalurki fakultet Univerziteta u Beogradu 11000 Beograd, Karnegijeva 4, Srbija
e-mail:majai@tmf.bg.ac.yu,snezana@tmf.bg.ac.yu

References
Bulsary A., Palosari S. (1993) Application of artificial neural networks for fuzzy simulation of a chemical reactor, Proceedings of the 35th SIMS simulation conference, Kongcberg, Norway, June,1993. Korn,A.G.(1993) Smulation of a fuzzy logic control system, Simulation, 61, 244-249. Savkovic-Stevanovic,J., T.Mosorinac(2008) A model of the fuzzy controller, Proceedings of the UKSIM2008-10th International Conference on Modelling and Simulation, paper,1-4 April, Cambridge ,U. K.,2008. Savkovic-Stevanovic J.,(1992) Fuzzy supported modeling of fermentation processes, 2ndIFAC Symposium on Modeling and Control of Biotechnical of Biotechnical Processes, Keystone, Colorado, U.S.A., March 29-April 2, 1992 Savkovic-Stevanovic J. ,M.Ivanovic R.Beric (2005) A surfactant production investigation, Comput. Ecol.Eng. 1, No.2, 129-134. Savkovic-Stevanovic J., M.Ivanovic, R.Beric, V.Manjencic(2003) Dynamics of the laurconium-chloride production, The 4th European Congress of Chemical Engineering, Granada, September 23-26.

U ovom radu prouavana je izvodljivost korienja fazi logike regulacije za hemijske procese na primeru kontinualnog reaktora sa meanjem. Ulazi su protoci reaktanata i poetne koncentracije. Izalzi su protoci proizvoda i koncentracije vrsta. Dinamiki odziv regulacione petlje na sluajan poremeaj sa promenljivom amplitudom na protok reaktanata je ispitivan. Istraivano je nekoliko inverznih modela. Fazi logiki regulator poboljava kvalitet regulacije, odreuje optimum poretnih taaka, aurira planirane modele i otklanja operativne probleme iz dana u dan. Ova sposobnost dozvoljava fazi regulatoru da adaptira sistem koji varira sporo sa vremenom. Kljune rei: regulacija u uslovima neizvesnosti, dinamiki odziv reaktora, poetni set,sluajni poremeaj