pjst8_1_59

8/8/2019 PJST8_1_59

1/9

Fuzzy Logic Modeling of the Fluidized Catalytic Cracking Unitof a Petrochemical Refinery

P.B. Osofisan, Ph.D.* and O.J. Obafaiye, M.Sc.

Department of Electrical and Electronics Engineering

University of Lagos, Akoka-Yaba, Lagos, Nigeria

E-mail*: [email protected]

ABSTRACT

This paper describes investigations carried outregarding the application of Fuzzy Logic Controlto the Fluidized Catalytic Cracking Unit (FCCU)of Kaduna Refinery and PetrochemicalCompany in Northern Nigeria, as a case study.

An optimal control solution where the objective isto determine a well-defined relationship betweenthe vital variables (reactor temperature/riseroutlet temperature, regenerator gastemperature, regenerated catalyst feed rate, andthe airflow rate) through the use of Fuzzy Logiccontrol scheme is the focus of this paper.

In the catalytic cracking unit, feed oil iscontacted with re-circulating catalyst and reactedin a riser tube. The feed oil vaporizes and iscracked as it flows up the riser, thus forminglighter hydrocarbons (the gasoline fraction).

Large amounts of coke are formed as a by-product. The coke deposits on the catalyst andreduces its activity. The lighter hydrocarbonproducts are separated from the spent catalystin the reactor. Steam is supplied to strip volatilehydrocarbons from the catalyst. The catalyst isthen returned to the regenerator, where the cokeis burnt off in contact with air. This is usuallydone by partial or complete combustion. Theregenerated catalyst is then re-circulated back tomix with the inlet feed oil from the crude unit. [1].

The behaviour of the reactor temperature/riser

outlet temperature and the regenerator gastemperature during the chemical reactions in theFCCU were simulated using MATLAB. Theproblem of control, will however involve thecontrol of two outputs (reactor temperature/riseroutlet temperature and the regenerator gastemperature) by manipulating the two inputs(regenerated catalyst feed rate and the airflowrate), which are the critical and vital factors foroptimization of the cracking process in the

FCCU. A relationship was developed betweenthe above stated input(s) and output(s) with thehelp of the fuzzy logic controller. This willfacilitate optimization of gasoline production.

(Keywords: process engineering, control models,

fuzzy logic, hydrocarbon, gasoline production,industrial catalyst)

INTRODUCTION

Optimization of gasoline production poses a bigchallenge in the petrochemical refinery becausethe input and output variables are non-linear,interdependent, and full of uncertainties. Theproblem for this study is to develop a controller(using Fuzzy Logic) to model the response of theFCCU to regenerated catalyst feed rate andairflow rate. The modeled information is in turnused to design a control solution to read thereactor temperature/riser outlet temperature andthe regenerator gas temperature of a FCCU, andadjust the regenerated catalyst feed rate and airflow rate accordingly for optimal performance.

Giving the non-linear and interdependent natureof input and output in the FCCU, the simple falseor true logic cannot adequately deal with theensuing control situation.

Fuzzy Logic is a systematic mathematicalformulation for investigating and characterizing

different types of uncertainties. It is best suitedwhen a mathematical model of the process doesnot exist; exists but is too complex to beevaluated fast enough for real time operation; oris too difficult to encode. In such situations,difficulties arise in using traditional controlmethods [12].

The Pacific Journal of Science and Technology 59

http://www.akamaiuniversity.us/PJST.htm Volume 8. Number 1. May 2007 (Spring)
mailto:[email protected]:[email protected]:[email protected]

8/8/2019 PJST8_1_59

2/9

Due to the lack of a precise mathematical modelfor the process being controlled, rule basedfuzzy control may be adequate because noexact and explicit process models are required.

One such suitable candidate for fuzzy control isthe process of upgrading heavy hydrocarbons tolighter more valuable products by cracking, inthe FCCU of a petrochemical refinery.

Catalytic cracking is a refinery process thatseeks to increase the gasoline and liquefiedpetroleum gas (LPG) production, through theheavy vacuum gas oil and residue conversion inlighter fractions. Because of its impact on overallrefinery economics, the FCCU is the best unit toapply advanced control and optimizationstrategies, and the base for these is always agood mathematical model. The model has to beable to reproduce reasonably well the main

dynamics and stationary gains of the system,without compromising the computational load.

There are many mathematical models for theFCC in the literature. Some of these use asimplified cracking process description, and afew of them present integration betweenregenerator and riser. Most are based on modelwith a high degree of empiricism, and make useof pseudo-components corresponding todifferent groups of species, usually called lumps[6].

Among the cracking kinetic models, there is the3 lumps model of [18], a 10 lumps model by [7],and more recently [15] developed a model with19 lumps, approximating the reactants andproducts according to the crude oil cutscomposition. Among the integrated models, [13]published a well-detailed model based on theobsolete Exxon Model IV with a realisticdescription of the regenerator fluid-dynamicbehaviour; however the combustion reactionswere not considered. It also lacks a detaileddescription from cracking kinetics, making theriser useless for dynamic or stationary control.

More recently [1] developed a model thatprovides a detailed description of thecombustion and cracking kinetics, using the 10lumps model of [7] to represent the mixture inthe riser.

As already pointed out by [4], an importantlimitation in most of these models is the fact thatthey ignore the complex two-phase nature of thefluidized beds in the regenerator. The objectiveof this work is to present a fuzzy logic approachbased on actual plant data.

To optimize the cracking process in the FCCU, aFuzzy Logic controller has been designed in thisresearch work, so as to get a well-definedrelationship between the manipulated input andthe output variables by use of a Fuzzy Model.The Fuzzy Logic controller has been simulatedon a digital computer using MATLAB 5.0 FuzzyLogic Tool Box.

THE FLUIDIZED CATALYTIC CRACKINGUNIT A BRIEF DESCRIPTION

Fluidized catalytic cracking (FCC) is animportant process in oil refineries. It upgradesheavy hydrocarbons to lighter more valuableproducts by cracking, and is the major producerof gasoline in refineries. FCCUs presentchallenging multivariable control problems. Theselection of good inputs (manipulated variables)and outputs (measured variables) is animportant issue, as is the pairing of chosencontrolled and manipulated variables fordecentralized control. A simplified processschematic and instrumentation diagram is shownin the figure below. [11].

Figure 1: Schematic Diagram of FCCU [19].



8/8/2019 PJST8_1_59

3/9

The heavy molecule cracking process occurs ina riser tubular reactor, at high temperatures,building up fuel gas, LPG, cracked naphtha(gasoline), light cycle oil, decanted oil, and coke.The coke deposits on the spent catalyst surfacecausing its deactivation. The catalytic activity isre-established by coke combustion in a fluidizedbed reactor, dominated regenerator.

The system riser-regenerator is called theconverter. Steam lifts the heated regeneratedcatalyst to be combined with the oil in the riserso that the oil-catalyst mixture rises in anascending dispersed stream to the separator.The control valve manipulates the quantity of hotregenerated catalyst from the standpipe to theriser in order to maintain a predetermined outletriser temperature. On the top of the separator,the catalyst particles are separated from vaporproducts by cyclones. The stream transfers the

reaction products overhead to the productsrecovery section. The standpipe transfers spentcatalyst continuously from the separator to theregenerator by a control valve.

In the regenerator, spent catalyst particles areburned in the presence of air. The air flow rate tothe regenerator is controlled by a control valvethat vents portions of the air to the atmosphere.On the top of the regenerator, cyclones performthe catalyst separation from the flue gas stream[10].

A control valve regulates flue gas flow in order tovary the internal regenerator pressuremaintaining the desired pressure differencebetween separator and regenerator. The fluegas goes to a carbon monoxide boiler wherecarbon monoxide is converted to carbon dioxide.There is a recycle stream around the wet gascompressor to control the suction pressure,which maintains the converter pressure at itsdesired value.

The measured variables are riser temperature,regenerator temperatures, wet gas compressor

suction pressure, separator-stripper catalystlevel, separator-regenerator differential pressureand regenerator flue gas temperature. Themanipulated variables are feed flow rate,preheated feed temperature, catalyst circulationrates, combustion air flow rate and wet gascompressor recycle rate. The measureddisturbances are feed characteristics, feedtemperature, and air temperature [2].

FUZZY CONTROLLER DESIGN FOR FCCU

As mentioned earlier, during cracking, feed oilvaporizes and is cracked as it flows up the riser,thus forming lighter hydrocarbons. This leads tothe formation of large amounts of coke, whichdeposits on the catalyst and reduces its activity.Given the complexity of the entire process,traditional system modeling for control design(involving the derivation of a mathematicalmodel to describe the system) that in turnrequires deep understanding of all variableinvolved is too difficult. Hence, Fuzzy Modeling,which deals with the relationship of the output tothe input, considering many other parameters isemployed.

In the design of a Fuzzy Logic Controller, systemadjustments are handled by a Fuzzy Rule-BasedExpert System [17] [3]. Well adopt the

knowledge base approach, which consists of thefollowing components (Figure 2):

(a) Data Base - that contains knowledge usedto characterize Fuzzy Control Rules andFuzzy Data Manipulation in an FLC, whichare defined based on experience andengineering judgment of an expert. In thiscase, an appropriate choice of themembership functions of a fuzzy set plays acrucial role in the success of an application.

(b) Rule Base - that is characterized by

construction of a set of linguistic rules basedon experts knowledge. The expertknowledge is usually in the form cause andeffect i.e. IF THEN. Fuzzy statements canthus easily implement this.

Controller Design

We define (5) linguistic values for the fuzzysystem as follows:

1 = Negative Big (NB) Very Low Consequent

2 = Negative Small (NS) Low Consequent

3 = Steady State (SS) Steady State

4 = Positive Small (PS) High Consequent

5 = Positive Big (PB) Excessive HighConsequent



8/8/2019 PJST8_1_59

4/9

Figure 2: General Structure of a Fuzzy Feedback Control System.

This rule base is derived based on the followingassumptions.

1. The SISO process has a monotonicinput output relationship, not anecessarily linear relationship.

2. The control goal is to maintain acontrolled variable at a desired level by

manipulating process input.

3. The response of the system will not beideal, due to the inertia propertypossessed by the process.

Rule Base

Using derivation based on expert experienceand control engineering knowledge; theexperience of an operator who has been workingat the Kaduna Refinery and Petrochemical

Company located in Northern Nigeria for over 12years was used to obtain the rule base. Theoperator is also our expert in defining the fuzzyrules and the fuzzy set.

Table 1: Fuzzy Rules for Airflow Rate.

Riser Outlet Temp

Reactor CycloneTemp

PB PS SS NS NB

PB 5 4 3 1 1

PS 5 4 3 1 1SS 5 4 3 2 1

NS 5 4 4 3 1

NB 5 5 5 4 3

1 = Negative Big, 2 = Negative Small, 3 = Steady State,4 = Positive Small, 5 = Positive Big

Table 2: Fuzzy Rules for Catalyst Feed Rate.

Riser Outlet Temp

Reactor CycloneTemp

PB PS SS NS NB

PB 5 5 3 1 1PS 5 4 3 1 1SS 5 4 3 2 1

NS 3 3 2 2 2NB 3 3 2 1 1




8/8/2019 PJST8_1_59

5/9

Table 3: Fuzzy Rules for Oil Feed Rate.

Riser Outlet Temp

Reactor CycloneTemp

PB PS SS NS NB

PB 5 5 5 2 1PS 5 4 4 2 2SS 5 2 3 2 1NS 3 1 2 1 1NB 2 1 1 1 1


Case Study Descriptions

The FCCU at Kaduna Refinery andPetrochemical Company in Northern Nigeria was

the case study used for this research paper.

Fuzzy Sets Formation

The membership function plots for the catalystfeed rate, airflow rate, feed oil rate, regeneratedcyclone temperature, riser outlet temperatureand the regenerator bed temperature are asshown below. These form the main parametersin the cracking system in an FCCU. Weemployed these datasets in the design of therule base for the Fuzzy Logic Controller.

Figure 3: Membership Function for Catalyst FeedRate (tons/day).

Figure 4: Membership Function for Airflow Rate(Kg/sec).

Figure 5: Membership Function for Oil Feed Rate(m

3/hr).

Figure 6: Membership Function for the RegeneratorCyclone Temperature (C).

Figure 7: Membership Function for the Riser OutletTemperature (C).

Figure 8: Membership Function for the RegeneratorBed Temperature (C).



8/8/2019 PJST8_1_59

6/9

The Fuzzy Superior Controller was simulatedusing MATLAB

5.0 Fuzzy Logic Toolbox. Our

goal was to establish a functional relationshipbetween the input variables, which are theregenerated catalyst feed rate and the airflowrate, and the output variables (i.e. riser outlettemperature and regenerator cyclonetemperature). The variables in consideration areregarded as the most important with respect tohydrocarbon cracking, energy consumption, andproduct quantity.

The objectives set for the cracking process inthe FCCU are to:

Maximize unit capacity

Maintain product quality whilemaximizing yields of most valuableproducts

Optimize energy utilization

Control conversion

Improve safety and reliability viaoperational stability

It has been found that the set objectives dependon the catalyst flow by minimizing catalystattrition and increasing fines retention.Irrespective of the catalyst type, density, andparticle size distribution, this improves:

stripping efficiency

regeneration capability

the range of catalyst circulation rates

pressure profile

riser temperature control

These improvements are essential tosubsequent improvements in:

unit conversion

product selectivity

catalyst stability in the presence ofcontaminants

And, they minimize:

coke and gas production

catalyst consumption and air pollution

This results in a more stable and flexibleoperation that is easier to operate close to unitconstraints and achieve substantial gains in unitperformance.

RESULTS

Figures 9, 10, 11, and 12 represent the graphicalform of results obtained through this study. Theyare the typical curves of:

i. The catalyst feed rate as a function ofthe regenerator cyclone temperature(Figure 9).

ii. The catalyst feed rate as a function ofthe riser outlet temperature (Figure 12).

Figures 9, 10, 11, and 12 show the input andthe output after modification and enhancementby tuning. To achieve this level of enhancement,the rules and the centers of the input and outputmembership functions were changed.

Figure 9: Catalyst Feed Rate as a Function of theRegenerator Cyclone Temperature.

Figure 10: Catalyst Feed Rate as a Function of theRiser Outlet Temperature.



8/8/2019 PJST8_1_59

7/9

Figure 11: Airflow Rate as a Function of theRegenerator Cyclone Temperature.

Figure 12: Airflow Rate as a Function of RiserOutput Temperature.

Figure 9 shows that the catalyst feed rate wasalmost constant for regenerator cyclonetemperatures below 560 C. For temperaturesranging from 580 C to 680 C, the regeneratorcatalyst feed rate rises steadily. The catalyst

feed rate reaches a peak at 680 C and a sharpdecline is seen for temperatures greater than700 C.

These means that an increase in the inputvariable (i.e. the regenerator cyclonetemperature) leads to a significant change in thecatalyst feed rate. The range of catalystcirculation rates (1.0 tons/day to 2.0 tons/day)

and catalyst stability in the presence ofcontaminants is greatly impacted by regeneratorcyclone temperature. In order to minimizecatalyst attrition and increase fines retention, theregenerator cyclone temperature should be keptas long as possible between the 645 C and 680C range. Once the temperature reaches thatstage, the relationship represents that of apiecewise linear system, expressedmathematically as:

x0(t) = b0x0(t);

b0 positive or negative:xi (t) = input;x0 (t) = output

Figure 10 shows that the catalyst feed rate risessteadily as riser outlet temperature increases. Apeak catalyst feed rate of about 1.7 tons/day is

obtained at a temperature of 648 C.

This implies that optimizing energy utilization byriser temperature control improves unitconversion. The riser outlet temperature, ifsustained at a high temperature (greater than500 C), will lead to a catalyst feed rate of about1.0 tons/day to 1.7 tons/day, a range withinwhich the regenerator cyclone temperature isalso plausible as seen in Figure 9. This is alinear relationship that can be expressedmathematically as:

x0(t) = b0xi(t)

Figure 11 shows that the airflow rate risessteadily to 62x10

3kg/sec until the regenerator

cyclone temperature gets to 600 C, and remainssteady at this temperature until it reaches 650 C.There is a sharp decline in the airflow ratebetween temperature 650 C and 657 C of theregenerator cyclone. The airflow rate thenremains at this level until the regeneratorcyclone temperature of 750 C is reached. Thismeans that an increase in the input variable i.e.the regenerator cyclone temperature does not

bring a significant change in the airflow rate. Airflow distribution remains constant immediatelyafter the regenerator cyclone temperaturereaches 657 C. In order to save energyconsumption, the regenerator cyclonetemperature can be kept constant at 657 C. Therelationship obtained represents a second ordersystem, expressed mathematically as:



8/8/2019 PJST8_1_59

8/9

a2d2

x0+ a1 d x0+ a0x0= b0xi(t)dt

2dt

Figure 12 shows that the riser outlettemperature, if sustained at 550 C, would lead tosavings in energy consumption and minimizecoke and gas production. Riser outlettemperatures greater than 550 C do not provideany incremental addition to the airflow rate. Theinput variable should therefore be kept at asteady state of 550 C as soon as thetemperature is reached. The relationshipobtained represents that of a second ordersystem, expressed mathematically as:

a2d2

x0+ a1 d x0+ a0x0= b0xi(t)dt

2dt

xi (t) = input;x0 (t) = output

From the results of the simulation, we have beenable to establish a relationship in the form ofgraphs for the following:

a. Catalyst feed rate as a function ofRegenerator cyclone temperature andriser outlet temperature (Figures 9 and10).

b. Airflow rate as a function of Regeneratorcyclone temperature and riser outlettemperature (Figures 11 and 12).

We have been able to establish concreterelationships between the input and outputvariables that are required to optimize the FCCUproducts by enhancing the cracking process. Byapplication of a suitable numerical analysismethod to these results, a more linearrelationship may be obtained.

CONCLUSION

The Fuzzy Model that was designed and

described in this paper is capable of managingthe characteristic uncertainties and imprecisionnormally associated with the catalytic crackingprocess in an FCCU. The controller has provedto be capable of providing a workable FuzzyModel, taking into account the objective of theoptimization problems associated with theprocess.

When compared with other methods such asmodel predictive control, etc. this research hasdemonstrated and established the advantage ofrelative ease of design and implementation ofcontrol systems offered by Fuzzy Logic.

REFERENCES

1. Arbel, A., Z. Huang, I.H. Rinard, and R. Shinnar.1995. Dynamics and Control of FluidizedCatalytic Crackers. Modeling of the CurrentGeneration of FCCs. 34:1228.

2. Secchi, A.R., J.O. Ttrierweller, L.A.S. Casali,D.D. Cunha, and G.A. Neuman. 2001. FCCDynamic Modeling: First Principles or SystemIdentification?. In: Simpsio Brasileiro deAutomao Inteligente. Canela: Brasil.

3. Chin-Fan-Lin. 1994. Fuzzy Logic Controller

Design. Advanced Control Systems Design.Prentice Hall: Englewood Cliffs, NJ. pp.431-460.

4. Elnashaie, S. and S.S. Elshishini. 1993. DigitalSimulation of Industrial Fluid Catalytic CrackingUnits IV. Dynamic Behavior. 48(3):567 583.

5. Gross, B., S.M. Jacob, D.M. Nace, and S.E.Voltz. 1976. Simulation of Catalytic CrackingProcess. US Patent 396 707.

6. Grosdidier, P., A. Mason, A. Aitolahti, P.Heinonen, and V. Vanhamaki. 1993. FCC UnitReactor-Regenerator Control. Computers Chem.Eng. 17:165.

7. Jacob, S.M., B. Gross, S.E. Voltz, and V.M.Weekman. 1976. A Lumping and ReactionScheme for Catalytic Cracking. AIChE J.,22(4):701 713.

8. Kosko, B. 1991. Neural Networks and FuzzySystems: A dynamical Systems Approach toMachine Intelligence. Prentice Hall: EnglewoodCliffs, N.J.

9. Krambeck, F.J. 1991. Continuous Mixtures inFluid Catalytic Cracking and Extensions. VanNostrand Reinhold: New York. pp. 42 59.

10. Lansarin M.A. 1999. Modeling FCC Unit:Characterization of Petroleum FractionsAccording to Tem Lumps Model. Florianopolis:Brazil

11. Lee, W. and V.W. Weekman. 1976. AdvancedControl Practice in the Chemical ProcessIndustry: A View from Industry. AIChE J., 22:27.



8/8/2019 PJST8_1_59

9/9

12. Lee C.C. 1995. Fuzzy Logic in Control System:Fuzzy Logic Controller Part I and Part II. IEEETrans Systems Man and Cybernetics. 20:404 418.

13. Marlin, T.E. 1995. Process Control. McGraw-Hill,New York, NY.

14. McFarlane, R.C., R.C. Reineman, J.F. Bartee,and C. Georgakis. 1993. Dynamic Simulator fora Model IV Fluid Catalytic Cracking Unit.CER&D. 17(3):275 300.

15. Pitault, I., D. Nevicato, M. Forissier, and J.R.Bernard. 1994. Kinetic Model Based on aMolecular Description for Catalytic Cracking ofVacuum Gas Oil. Chem. Eng. Soc.49(24A):4249 4262.

16. Secchi, A.R., M.G. Santos, G.A. Neumann, J.O.Trierweiler. 2001. A Dynamic Model for a FCCUOP Stacked Converter Unit. Computers and

Chemical Engineering. 25:851 858.

17. Self, K. 1990. Designing with Fuzzy Logic.IEEE Spectrum. 42 44.

18. Weekman V.W. 1968. A Model of CatalyticCracking Conversion in Fixed, Moving, and Fluid-Bed Reactor. IEEC Proc. Des. and Dev.7(1):9095.

ABOUT THE AUTHORS

Dr. P.B. Osofisan, obtained his B.Sc.(Eng) andM.Sc.(Eng) in Electrical Engineering from theUniversity of Stuttgart, Stuttgart, Germany. Heearned his Ph.D. in Control SystemsEngineering from the same University. He thenworked in a cable manufacturing plant as theProduction/Quality Control Manager for over 15years, before he joined the University of Lagosas Senior Lecturer in Electrical and ElectronicsEngineering Department. His research interestsinclude the application of Fuzzy Logic Theoryand Neural Network in the process control ofindustrial processes.

Mr. O. John Obafaiye, obtained his B.Sc.(Eng.)degree from the University of Lagos and has justconcluded his M.Sc.(Eng.) program at the sameUniversity.

SUGGESTED CITATION

Osofisan, P.B. and O.J. Obafaiye. 2007. FuzzyLogic Modeling of the Fluidized CatalyticCracking Unit of a Petrochemical Refinery.Pacific Journal of Science and Technology.8(1):59-67.

Pacific Journal of Science and Technology


http://www.akamaiuniversity.us/PJST.htmhttp://www.akamaiuniversity.us/PJST.htmhttp://www.akamaiuniversity.us/PJST.htm

pjst8_1_59

Documents