A Gonfluenc€-Based ExpertSystem for the Detection ofHeat Exchanger FoulingNAIM H. AFGAN and MARIA DA GRAQA CARVALHOInstituto Superior Tecnico, Lisbon, Portugal

Efficiency assessment of heat uchangen ensures appropiate use of the auailable energt' Thb articlepiesens the concept of a heat exchanger on-line expert system based on qualitatiue reasoning. Using'conJluence of the'heai erchanger effectiueness defined by the NTU (number of transfer units) concept

for the simple parallel-stream heat exchanger, the methodologt is deueloped to be used for the'desciption

of the geneic behauior of the heat uchanger system. In this respect, panicular attention was

paid to the iecognition of those situations leading to the degradation of the effectiueness of the heat'uchanger.

The article descibes the selection of the diagnostic uaiables, and their on-line measurements

including the bgtng system for monitoing and acquisition of the data'fo, i tprritri hiaierchanger, the upert system assessment was perfonned for a number of dffirent

sintations. An appropiate diignostic system in the specified time peiod produces a set of readings of the

diagnostic ua*bles, descrtbiig the state of the systeni. The set of diagnostic uaiables conuerted into the

,oi\urrrc porometerc constitutes the qualitatiae desciption of the instantaneous state of the system.

A ietrieual strategt is used to find the conesponding state in the knowledge base with the set of the

parameters descibing the indiuidual malfunction state of the system. Each malfunction state isaccompanied by a recommendation for its correction specified by expen aduice.

In many heat exchangers, fouling is an impor-tant problem that requires careful monitoring. Inspite of precautions taken in the design of theheat exchanger, fouling is an unavoidable prob-lem. Presently, designers are doing all kinds ofoptimization of heat exchangers. Second law anal-ysis of heat exchanger effectiveness is the mostpowerful approach in justification of the totalannual cost function of a heat exchanger [1-3]' Inthis respect it becomes of paramount interest todevelop a procedure to deal with the on-line as-sessment of heat exchanger operation. In particu-lar, in some heavy-duty operations' heat exchang-

Received 4 October 1994; accepted 7 July 1997.Address correspondence to Prof' Naim Afgan, Instituto Superior

Tecnico, Departmento de Engenharia Mecdnica, Seccao de Termodi-namica Aplicada, Pavilhdo de M6guinas' 1 andar Av. Rovisco Pais,1096 Lisboa Codex, Portugal. E-mail: nafgan@navier.ist.utl'pt

28 heat transler engineering

ers are a milestone of total system operation thatrequire monitoring and assessment during on-lineoperation.

Heat exchanger effectiveness is the parameterthat comprises all relevant effects that inducedegradation of the availability cost of heat ex-changer operation. On-line monitoring of the ef-fectiveness is technically feasible with suitable in-strumentation and logging system.

Expert system tools are becoming a velv effi-cient means in the assessment of the functioningof a process system. Several studies [4-8] havedemonstrated that expert system logic is becomingan effective tool for on-line assessment of theparameters leading to the degradation of the sys-tem. However, selection of the expert system do-main is one of the important issues to be consid-ered in the development of the expert system. Inorder to investigate the specific concept of an

vol. 19 no.2 1998

or

It is known that the heat transter coefficient ina heat exchanger is a complex function of a num-ber of variables dependent on the specific designof the heat transfer surface. As the first approxi-mation in this analysis, a simple geometry of heattransfer surface will be assumed in order to over-come eventual problems with nonlinearity of theheat transfer coefficient dependence on the flowrate.

Introducing the expression for heat transfercoefficients as linear functions of mass flow ratefor the hot and cold streams and assuming thatthe changes in flow rate will be substantiallysmaller than the flow rate for each stream, weobtain

are diagnostic paramctcrs rcflecting thc changc ofheat exchanger effectiveness.

QAALITATIW REASONING BASED ON ACONFLAENCE

In qualitative process analysis, the notion of aqualitative differential equation is called conflu-ence f12, 14-1,61. It was shown that confluencecould be used in the definition of state for anyphysical system. Qualitative analysis is based onthe behavioral description of the process in which,instead of continuous real-valued variables, eachvariable is described qualitatively, taking only asmall number of values, usually *, -,0. Thus, themain assessment tool is the qualitative differentialequation, called a confluence. As can be seen inthe derivation of the heat exchanger effectiveness,the confluence is

dw, dw^@. " +O.' dt ' at1de

0-2e) dt

Introduce

7de _-I-(I-Ze) dt-'Then

d6u a6-+K. +K"- dt ' ar

dNn I dN" dh^\t_ _ldt \ano dtl

I aN" d6, \!l--l

laoo dtJ

I aNn dh,

\ ,5. atI dN"+t-\ dh,

(e)ah, \

-tdt I

dNA a6h _d6o dt

where

dNu dhodh, dt

dNu d6,

AUz 1

C^;n hn

AU2 1

C^in kn

AU2 1

dho dh,at 'dtd5^ d6o

K-dt -dt

d6-K.''dt

dh c

" at

u6, :

(10)

(11)

(12)

(13)

d6, at C^in k, dt

NA dh, : nnt I dh,6h, dt C^;n h, dt

dwu aw- d6, d6^O, J+O" +K.J t6.---':f' dt " dl ' at ' at

( 18)

Adopting the notation that d/ dt : d, one obtains

( 16)

Q7)

( le)

(20)

could

6)lA dhh6ho dt

aw, aN, d6 d6-:4 'at d6^dt -dt Ql dwh + Qz aW, + K2 d6h + K., d6, : Iso that the confluence is

dwh+dwc+d6h+d4:foIf it is accepted that the four variables

assume the following values,

;:'lA d6,

d4 dtd6^ dN^ dh ":K.---:." : ':@.' dt dh. At

(14)

dW,dt

(1s)

be deter-where @i, Q., Kz, K3 are constants tomined for a specific heat exchanger and

dtyh aw" d6a d6,.--|-, ---l- ' --:_ , :ot dt dt dt

dwh : *, -,0 d6h: *, 0 (ZI)dW,: *,-,0 d6.:*,0then the number of possible states for Eq. (20) isas given in Tables I,2, and 3. Tables I,2, and 3

vol. l9 no.2 1gg830 heat transfer engineering

d6r, =+ ; d6" = 0

Beside these parameters, the diagnostic systemthe fouling assessment expert system shouldclude:

Wo,W"

In order to ensure confidence in the measure-ments, it is necessary to introduce a validationelement in the diagnostic system, which checksevery reading it enters before the semantic pro-cessing.

Diagnostic systems also comprise a buffer forintermediate data storage and numerical dataconversion into a semantic expression to be usedin the expert shell processing. This element isnormally included in the computer system but isshown as a separate element in Figure 2.

forin-

13@t

Figure 1 State diagram.

The second part of the knowledge base is a setof conclusions with related recommendations tobe given to the user. It will include economicevaluators of the consequences of the foulingstates and will give possibilities for the assessmentof forthcoming development. Also, it is expectedthat in the future development of expert systemsthere will be a pattern recognition procedure thatwill be able to forecast both short- and long-timebehavior of heat exchangers. This includes display-ing messages to the user and imposing emergencyprocedures if the situation warrants.

MONITOMNG SYSTEM FOR HEATEXCHAN G E R E F F E CT IVE N E S S

l0

18

t2

SIGNAL PROCESSING

Qualitative reasoningsignal of measurements

AND ANALYSIS

is commonly based on theof the variables. In order

Heat exchanger effectivenessby on-line measurement of theters:

T11,71r2,74,Tr2

32

can be monitoredfollowing parame-

Figure 2 Flow diagram for on-line measurement and signalprocessing.

vol. tS no.2 1998

d61,4;6"=*

heat transfer engineering

CONCLUSIONS

The conceptual design of an expert system forfouling assessment in a heat exchanger has shownthe feasibility of this tool to increase the effective-ness and availabiliry of the heat exchanger.

The technique of sign analvsis based on qualita-tive reasoning introduced by the concept of con-fluence with trend analysis of the variablemeasurements shows a new possibility in the as-sessment of operational parameters of engineer-ing systems. In particular, the example demon-strated by application to the touling assessment ofa heat exchanger suggests that this technique canbe used in the assessment of heat exchanger per-formance. In spite of the fact that the examplepresented in this article relates to a simple heatexchanger design, it was demonstrated that a heatexchanger expert svstem based on qualitative rea-soning may be funher deveioped for other de-signs.

NOMENCI-ATURE

A heat transfer surlaceC, thermal caPacitv of cold streamC*"*, C-, maximum and minimum stream ther-

mal caPacitY ratesheat transfer coefficients on the hotand cold sides, respectivelvthermal conductivities of fouling la,v-ers on the hot and cold sides of theheat transfer suriace, respectively

overall heat transfer coefficientdefined by Eqs. (17) and (20), respec-tivelymass flow rates of the hot and coldstreams, respectivelyfouling layer thicknesses on the hotand cold sides, respectivelytube wall thickness and thermal con-ductivity, respectivelyincremental change of the diagnosticvariablesheat exchanger effectiveness definedby Eq. (2)uncertainty margin of diagnostic vari-ables: KrRr: KzRz

Uf, fo

Wh, W,

67r, 6.

6^, k^

A

€

(o1o2

hh,h,

t- l-K h, fr,

AU: 1-:- ,')L* niAL:: 1-;- t-u-ro [hAL:: 1

,-L-r" Ka

AL-: 1

C.nnumberEq. (2)

transfer units defined br

proportionality constants for hot andcold sueam heat transfer coefficients.respectivelytimeinlet and outlet temperatures of thehot stream, respectivelyinlet and outlet temperatures of thecold stream, respectivelY

REFERENCES

[1] Bell, K., Heat Exchangers with Phase Change, in J.

Palen (ed.), Heat Exchanger Sourcebook, Hemisphere,Washington, DC, 1986.

[2] Bergles, A., Techniques to Augment Heat Transfer' W.Rohsenow, J. Hartnett, and E. Ganic (eds.), McGraw-Hill, New York, 1985.

[3] Pinheiro, J., Fouling in Heat Exchangers, in J. Palen(ed), Heat Exchangers Sourcebook, Hemisphere, Wash-ington, DC, 1986.

[4] Karni, R., and Gal-Tzur, A., Paradigms for KnowledgeBased Systems in Industrial Engineering, Artificial Intel'ligence in Engineering, vol. 15, no. 3, pp. 125-141,1993'

[5] Afgan, N. H., Blokh, A. G., Radovanovic, P., Zhurovlev,Yu., and Gorb, E. I., Boiler Expert System, in K. Han-jalic and J. H. Kim Gds.), Expert Sysrem and Mathemati-cal Modelling in Energt Engineeing, Begell House, NewYork, 1995.

[6] Afgan, N. H., and Carvalho, M. G', Concept of BoilerEfficienry Assessment Expert System, in M' G' Car-valho, W. A. Fiveland, E. C' Lockwood, and C. Pa-padopoulos (eds.), Combustion Technologies for a CleanEnuironment, Gordon & Breach, 1996.

[7] Yu, C. C., and Lee, C., Model-Based Approach to FaultDiagnosis, Ind. Eng. Chem. Res., vol. 34, no' 3, pp'828-845, 1992.

[8] Kramer, M. A., A Rule-Based Approach to Fault Diag-nosis Using the Signed Directed Graph, AICLE '/', vol'33, no. 7, pp.1061-1078, 1987.

[9] Afgan, N. H., and Carvalho, M. G., Heat ExchangerExpert System, in N. Afgan, M. G' Carvalho, A. Bar-Cohen, D. Butterworth, and W. Rotzel (eds.), New De-uelopment in Heat Exchangers, Gordon & Breach, Newark,1996.

[10] Forbus, K. D., Qualitative Process Theory, Twelve Years1Jrcr, Anificial Intelligence, vol. 59, pp. 115-i23' i993.

[11] Kuipers, B. J., Reasoning with Qualitative Models, lm-ficial Intelligence, vol.59, pp. 125-132, 1993.

[12] de Kleer, J., A View on Qualitative Physics, ArtificialIntelligence, vol.59, pp. 105-114, 1993.

vol. 19 no.2 1998

K.

K.

ta,a\ -

KlNr

Rh, R.

tT6,T;

Tr1,Tr1

34

Eof

heal transler engineering

..!._