You are on page 1of 6

m:

cla ss and home prOble ms )

Th e objec t of this col umn is to enhance our readers ' co llectio ns of interes ting and novel probl ems in chemical engineer ing. Problems of the type that can be used to motivate the student by presenting a part icular principle in class, or in a new light , or that ca n be assig ned as a novel home problem, are requested, as well as those that are more traditional in nature and which eluc ida te diff icult co nce pts. Plea se submit them to Professor James O. Wilkes (e- ma il: wilkes@engin.umich.edu) or Mark A. Burn s (e-mail: maburns @engin .umich.edu ). Chemical Engineering Department , University of Michigan , Ann Arbor, MI 48109-2136.

AN INTRODUCTION TO PROCESS FLEXIBILITY


Part 1. Heat Exchange
W.E. JONES, J.A. WILSON
Uni versity of Nottingham

University Park N ottingham N07 2R D E ngland

rocess plants need to be flexible to cope with changes in production rates, produ ct specifications, feedstoc k, catalyst deactivation , and heat exc hanger fouling. Tradition ally, once the process structure has been decided , various operating cases are evaluated and one is chose n as the basis for detailed design. However, se lec tion of the design case is not straightforward. Effec tively dealing with all the highly interre lated issues durin g design is a formidable problem. Hence, engineers often resort to the application of ruleof-thumb safety factor s during equipment design (e.g., adding 10% extra area to a heat exc hanger) in an effort to ensure flexibility. Following this strategy, an ex perienced engineer would hope to develop a design that is operable across the anticipated process range, but there is no guarantee that the required flexib ility will be achieved.!':" As the prob lem presented here clearly illustrates, different plant operating modes ca n eas ily lead to equipment design situations that are not cov ered by a simple safe ty fac tor. The above comments explain why no substantial coverage of flexibil ity is found in any of the standard undergradu ate design textbooks, apart from a few remarks on safe ty factors. Despite these diffi culti es, we feel the topic is very important, particularly becau se of the highly integrated plant s

being built today, and that the basic ideas should be introduced to all students. Some students will enco unter flexibility probl em s as part of their final-year design project. These projects are normally simplified from indu strial reality, consider ing only one feed stock and, at worst, a catalyst dea ctivati on or heat
Warren Jones holds BSc and PhD degrees in chemical engineering from the University of Nottingham and is a registered Chartered Engineer. He has a wide-ranging interest in both frontend processes and detailed plant design, developed initially through nine years of experience with a major engineering and construction company. Teaching responsibilities include several design courses, process economics, and engineering thermodynamics.

Tony Wilson holds BSc and PhD degrees in chemical engineerin g from the University of Nottingham. With industrial and consulting experience in process control and batch process engineering, and with active research in both fields, he coordinates the department's research in computer-aided process engineering and is responsible for process control teaching at the undergraduate level.

Copyrig ht Ch E Division ofASEE 1997

172

Chemica l Engine ering Education

exc hange r foulin g cycle. Neverth eless, to design an operable plant, thought must be give n at an earl y stage to condition s under which each item of equip ment is expec ted to operate and to the process-control scheme to be used. Unless the proj ect supervisor is alert, many students wi ll simply size equipmen t for the conditions implied by the desig n mass and heat balances without consideri ng flexibility . This and a subsequent article will attempt to illustrate how se lected aspects of flexi bili ty can be introd uced thro ugh interesting exa mples. In particular, the heat-exc hange prob lem deve loped here may be used direc tly in a desig n course, while the reactor recycle loop featured in the second art icle could form the basis for project work , or for a discu ssio n question in reactor des ign, or simply to indicate to supervisors an area worth discussi ng and developing in future design projects.

Students are familiar with using heat balances to calculate the heat load, Q,

(I)
and the rate equation

(2)

BACKGROUND
Mos t studen ts, if asked , will suggest adj usting steam pressure or hot oil flow rate to a heat exchanger in order to maintain exi t tem perature in the face of process flow change. Slightly less obvio us would be the suggestion to alter the condensate level in the heat-exchanger shell, thereby covering/exposing more heat transfer area for steam co nde nsation. Th e important point is that steam and hot oil are utilities , and changi ng their cons umption doe s not disturb the process. Diffic ulty is immedia tely encou ntered when considering heat exchange between two process streams; changing the flow rate of one will certainly affect the exi t tempera ture of the other. Unfortunate ly, interfering with a pro cess stream flow rate immediately upsets the plant mass balance, whic h is undesirable. Th e difficult y is overcome by using a by-pas s (see Figure I) that does not affect the total flow rate but changes the proportion actua lly passin g thro ugh the heat exc hanger and hence the heat transferred. The probl em presented here is concerne d with heat exchanger by-pass arrangements to ensure satisfactory operation, in the face of aging catalyst, of a reactor at both beginning -of-run (BOR) and end-of- run (EO R).

to determine hea t tran sfer area A, knowing the overall heat tran sfer coefficient U. Here, (mc p)H and (mcp)c are the prod ucts of flow rate and specific heat capacity for the hot and co ld streams. Temperatures T ) to T 4 are identified in Figure I, and for the mome nt we assume the by-pass is closed. But design to take into acco unt flex ibility implies not on ly calcu latin g A, but also looking at the implications for other operating conditions, and this is where Eqs. (3) and (4) become usefu l.P'" They are deri ved for Eqs. (I ) and (2) in order to permi t the writing of expl icit temperature equations:

(1- RB)T2 + R(B- I)T3 + (R- I)T) = 0 (1- RB)T4 + B(R- I)T 3 + (B- I)T) = 0

(3)

(4)

where R = (mcp)C/(rilCp)H and B= exp[ ( UA/( ITIC p)c) R- l)]. To illustra te, for a given heat exchanger (A specified) and known strea m properties (( rilCp )C,(mcp)H' and U specified), we can easily calculate the effect of an inlet temperature (T) and T 3 ) change on both exit temperature s (T 2 and T4 ) using Eqs. (3) and (4). It is a very simple exte nsion to apply sequentially the abov e pair of equatio ns to a heat exchanger network, thereb y evaluating the new tem peratures throughout the network.!" Bearing in mind that various operating mode s are to be accommodated, it is likel y that the heat exchanger sized for the most severe case will be too large for the other modes. As hinted earlier, this difficulty is overcome by ope ning the by-pass. We assume the heat exc hanger will opera te with the same UA (see the Appe ndix), but the effec tive UA for the heat exchanger plu s the partially open by-pass (as indicated by the dotted box in Figure I) is reduce d. Increasing by-pass flow progressivel y reduces the effective UA, whereas maximum heat tran sfer is achieved when the by-pass is closed. Good engineering practice would main tain a minimum flowrate of 5- 10% thro ugh the by-pass. Eq uation s (I ) throug h (4) are writte n for the case of sensib le heating and sens ible cooling of process streams . Special cases result for Eqs. (3) and (4) whe n one side of the heat exc ha nger o perates isoth ermall y. If the co ld-s ide operates wit h isotherm al vaporizatio n at T 3 , then Eq . (3)
/ 73

------------, By-pass I

T2

------

Fig ure 1. Heat excha nge r with by -pass.


Summer / 997

reduced to
T, - BT2 + (B -1 )T3 = 0

(5)

where B = exp( VA

I( mCP )H )'

If the hot-side operates with isothermal condensation at

Note particul arly that C, must have dimension s con sistent with the other variables in Eq. (8). The units used here are chosen for con venie nce in the rest of the prob lem rather than to agree with engi neering practice; hence, the C, values ca nnot be co mpared directl y with , say, co ntro l va lve manufacturer' s data.

T " then Eq. (4) redu ces to


(B - l)T, - BT3+ T4 = O (6)

PROBLEM STATEMENT

where B = exp( - V A I( mcp)c)' For the specia l case of

(mcp)H= (ITIC p)C(i.e .. constant tem -

perature drivin g forc e, dT , throu ghout the heat exchanger), it is easy to show that
(7)

Figure 2 shows the basic tlowsheet for the heat exc hange rs surrounding a catalytic reactor operating at EaR conditi ons. The hot reactor effluent is cooled first by boilin g water at 200C and then by preheatin g the reactor feed. The process operates entirely in the gas phase, and you may assum e a constant specific heat capacity of 2.5 kJ/kgK. At BOR , the catalyst is much more active, requiring a reactor inlet temperature of only 185C. The corresponding process flowrate and reactor effl uent temp erature are 22.5 kg/sec and 296.10C. a ) Calcul ate the VA requireme nts for both heat exchangers implied by EaR operation . Investigate the feas ibility of BaR operation using the VA values ju st determined if no flexibil ity is added to the flows hee t. b) Determ ine VA req uirements for BaR opera tio n if the temp erature to product recovery is to be maintained at
200 C

In the following probl em , Eqs . (5) and (7) are more immediately useful than the genera l Eqs. (3) and (4). Finally, the hydraul ic interac tion between the heat exchange r and control valve in the by-p ass line is important. Sele ction of contro l-va lve type and size is crucia l to ensure it remains operable ove r the range of by-pass flows ex pec ted. Difficult y occurs because transferrin g flow fro m the heat exc hanger to the by-p ass result s in a redu ced pressure drop across the heat exc hanger. The control valve ex periences the same pressure drop and so must accomodate the largest flowrat e at the lowest pressure drop. (To achie ve steadystate by-pass flowrates in excess of 30-35%, if the minimum is 5%, requ ires an unreali sticall y large control-valve size, and it is better to use two sy nchronize d valves, the second being in series with the heat exchange r and compensating for the decreasing pre ssure drop. !" )

Luyben'" summarizes the important properties of control valves . Volume flowrate, q(m 3 I sec), throu gh a control valve depends on the pressure drop , d P( bar), control-valve size, C" fluid den sity, p (kg / m 3 ) , and valve opening, x. The releva nt equation, if we ass ume dP is small co mpared to the operating pressure is
Steam

q = CJ(X )(d P/p)05

or, for mass flowra te

225C
Interchanger

BFW

(8)
where f(x) defines the contro l-valve characteristic in terms of valve opening. For this problem , two valve characteristics are important: I) Linear f(x) = x
125C

Reactor Feed 25 kg/s

To Product Recovery

2) Equal percentage f(x) = 0.' -' (typically, a =50)


/ 74

Figure 2. EOR flows heet.


Chemica l Engineering Education

125C. Wh ich operating mode sets the design for I) the interch anger? 2) the boiler? c) If the minim um flowrate through the by-pass is 5% of the main flowrate, determin e the design UA req uirements for both heat exc hangers. What percentage of the main flowr ate should pass throu gh the by-pass to permit the alternative operating mode? d ) Add the by-passes to the flowsheet and indicate how you would co nfigure the temp erature control loops. How would the plant be operated with your control scheme? e) The interchan ger has a co ld-side pressure drop of 0.6 bar calculated for Ea R flowrate and no by-passing. Select a suitable control valve from the following range of valves with linear charac teristics:

(
a ) Interchanger dut y Interchanger UA Boiler duty

SOLUTION

=25 x 2.5 x 100 = 6250 kW


= 25 K

Tempera ture difference

= 6250/25 = 250 kW /K

= 25 x 2.5 x 75 = 4687 .5 kW = 54 .1 K
= 4687.5/54.1 = 86.65 kW/K

Log mean temp. differe nce Boi ler UA

To invest igate the feasi bility of BaR operation with the above UA values, it is probably best to start with the boiler. Using Eq. (5), B=exp (86 .65/(22 .5 x 2.5 = 4.667 and the process exit temp erature
= 296.1 + (4.667 - I) 200 = 220.60C

c, = 1.0,

1.75, 2.5

4.667

In stead y-state operation, a control valve should operate with an opening between 0.2 and 0.8. You may assume a constant gas density of 20 kg/rrr' and neglect pipin g friction losses.
f)

We can now calc ulate the two exit temp eratures from the interchanger. Using Eq. (7),
VA I(

m cp ) = 250/( 22.5 x 2.5) = 4.444

The boiler has a hot-side pressure drop of 0.4 bar calculated for Ea R flowra te with no by-passing. Why would an equal percentage valve with a =50 be more suitable for this service than a linear one?

and the temp erature differe nce is


(220.6 - 100 ) /(4.444 + I) = 22.15 K

Hence the preheated reactor feed will be at (220.6 - 22. 15) = 198.45C and the strea m to product recovery is slightly too cold at 122.l 5C (but may be accep table ). The preheated reactor feed, however, is certainly far too hot at 198.45C. Conclusion : heat exc hangers sized for Ea R operation will not function satisfactorily at BaR. b) To determin e the UA requirements for BaR, it is advisable to redr aw the flowsheet to reflect BaR operation, as show n in Figure 3. The reactor feed must be heated from 100C to 185C, hence the cross-ove r temp erature of the reactor effl uent from the boiler to the interchanger is 210 C to maintain a temperature of 125C to prod uct recove ry.
Interchanger duty Temperat ure difference Interchanger UA = 22.5 x 2.5 x 85 = 4781 .3 kW =25 K =4781.3/25 = 191. 25 kW / K

185C

Stearn

210C
Interchanger

BFW

125C

Reactor Feed
22.5 kg/s

To Product Recovery

which is 23.5% less than the UA req uired for Ea R; hence, Ea R operation will set the design of this item of equipment.
Boiler duty Boiler UA = 22 .5 x 2.5 x 86. 1 = 4843. 1 kW = 4843.1/38.05 = 127. 28 kW /K
/ 75

Figure 3 . BOR flowshe et.

Log mean tem p. difference = 38.05 K

SlIlII lIIer / 997

which is 46 .9% more than the UA requ ired for EO R; hence BOR opera tio n wi ll se t the des ign of this item of equip ment.
c) We co nside r the interch an ger case in detail here. Fig ure 4A shows the actual interchanger ex it temperature for ; after blend ing wit h the the react or feed at EO R to be TO by-pass, the required preheated temperature of 200 C is achieved. Ou r first co nce rn is to ca lculate TO.

I ) installed UA = 153.2 kW/K based on BOR and 5% by-p ass 2) Z

= 0.2 16 for EO R opera tio n

d ) Fig ure 5 shows the flows heet with by-passes and tem-

200 TO

(0.05 x 100) + (0.95 x TO ) 205.2 6 0 C

perature-control loops added . Th e se tpoi nt for TC I may be set at 125C for all operati ng modes, but the ope rator wi ll need to increase the set point of TC2 from time to time to co mpe nsa te for the fall off in reactor catalys t activity.
e) If a co ntro l va lve with

and this implies a log-m ean temperature difference of 22 .27 K, giving a UA o f 280.65 kW/K , i.e., this UA sho uld be insta lled to give an e ffec tive UA of 250 kW/K for the co mbination of heat exc hange r plu s 5% by-pass. Fig ure 4B shows the frac tion by-passed during BOR operation to be Z . Th e determination Z is a trial-and error calculation, requiring
I ) Gu ess Z

C, = 1.75 is se lected , the following f(x) va lues are calc ulated using Eq. (8):
1) EOR 0.05 x 25 f(x)

= 1.75 f(x) (0. 542

x 20)5

= 0.2 17

where 0.542 bar has been estima ted by 0.9 5 2 x 0.6 2) BOR 0. 1415 x 22.5 = 1.75 f(x)(0.358 x 20)5 f(x) = 0.68 where 0.3 58 bar has been estima ted by

2) Calculate TOto give mix temperature of 185C 3) Ca lculate log-m ean temperature difference 4) Ca lcu late UA and co mpare wi th installed value of 280.65 kW/K . If ag reement is not achieved, return to I ) and repeat ca lculations unt il co nve rge nce is obtained. Th e co nve rge d so lution is
Z = 0.1415, i.e., well below maxi mum by-pass of 0.3

( 0.8585 x 22.5 06 25 ) x .

T he valve is known to have linear cha rac teristics; he nce, f(x) tra nslates to ope nings of 0.2 17 and 0.68 . These are acce ptable, lyi ng within the specified range of 0.2 to 0.8. Usi ng either of the other va lves gives unaccept able or imp ossible so lutio ns. C, = 1.0 means the va lve is not large eno ugh to handl e BOR and C, = 2.5 is too large,

TO= 199.01 0 C
Log-mean temperatur e difference

= 17.05 K

UA = 280.5 kW/K , close enough to installed value o f 280.65 kW/K A similar procedure is used for the boiler , for which the following key va lues are ca lculated:

200"C

IS5't

TC2!)---

-;

/"1;;:--'-_

Steam

T"

T"

r...~----

m"C 125't Z (I-Z)

21O "C 125't

.........'~- BFW

IOO't

25 kg/s
A: EOR Operation

22.5 kg/s
B: BOR Operation

Reactor Feed

To Produ ct Recovery

Fig ure 4. Temperature and flo wra te de tails for the in terch anger.
/ 76

Fig ure 5. Flowsh eet showing by-passes and tem perature control loops.
Chemical Engineering Education

such that the opening for EOR is below 0.2.


f)

ACKNOWLEDGMENTS
Th anks to Ca rl Pulford for drawin g the figur es and John Dos Sa ntos for helpful discussions on heat recovery.

Trial -and -err or ca lculations soon show that it is impossible to find a C, for a linear valve that spa ns the BOREO R flow range, giving both openings between 0.2 and 0.8. Equal percent age character istics enab le a wider range of by-pass flowrat es to be acco mmoda ted; i.e.. ca lculation soo n demonstrates that a C, of 5.32 gives an opening of 0.8 for EOR and 0.404 for BOR .

REFERENCES
1.

2. 3. 4.

APPENDIX
Str ictly, the ass umptio n of constant UA is inacc urate because opening the by- pass reduces flow through the heat exchange r, which affect s the film coefficient and hence the ove rall coe fficient. Th e lower rea l UA mea ns slightly less wo uld have to be by-passed. hence the by-pass flow calc ulated on the ass umption of co nstan t UA is an uppe r bound. For the probl em prese nted, you wou ld expect the by-pass to change fro m 14.15 % to aro und 13% if acco unt is taken for the reduced U. Thu s for design. provided that the most seve re cas e and asso ciated flows have been identi fied. the sma ll "e rrors" for the altern ati ve modes are easily acco mmodated by the co ntrol loop.

5.

6. 7.

8.

Gros sman, I.E ., a nd M. Mora ri , "Operability, Resiliency, and Flexibi lity-Process Design Objectives for a Changi ng Worl d" in Proc. Sec. In t. Conf., "Fou nda tions of Compu te rAided Process Design ," CACHE (1983 ) Morari , M., Compo and Chem . En g., 7, 423 (1983 ) Linnhoff, B., E. Kotjab asakis, a nd R. Smith, AIChE Ann ual Meet in g, Washingt on , DC, Paper 79d (1988 ) Perkins , J.D ., (ed), IFAC Symposium on Interactions between Process Design and Process Control , Perga mon, Oxford , En gland (1992) Kay, J .M., An In trod uction to Fluid Mechan ics and Heat Transfer, p. 313, Ca mbridge Univ ers ity Pr ess , Cambri dge, England (1963) Kotjab a sakis , E., a nd B. Linnhoff, Chem. Eng. Res. Des., 64, 197 (1986 ) Marlin , T.E. , Process Control: Designing Processes and Control Sy stem s for Dy nam ic Performance, p. 801, McGrawHill , New York , NY (1995) Luyb en , W.L., Process Modeling, Simu latio n, and Control for Chemical Eng ineers, 2nd ed., p. 213, McGr aw-Hill , New York , NY (1990 ) 0

. - : letter to the editor

Dear Editor

The universiti es with the eig ht top-ranked doc tora l programs in chemi ca l eng inee ring in 1982. as rated by the National Ac adem y of Sciences, were listed in Changing Times. In a 1991 letter to the editor of this journal (Vol. 25 , page 181). I pointed out that an analysis of this ranking revea led that 63.4 % of the facult y members in these eight "e lite" program s had obtained their doc tora l degrees from one of the same eight top -ranked schoo ls. I sugges ted that these progra ms had ma intained and enhanced their reput ations by hiring their ow n and one another ' s graduates. Doctoral programs in chemical engineering were ranked more tha n one decade later by the Natio nal Research Co unci l (Chro nicle of Higher Educatio n, 42 (4), 1995). I studied this report to find I) the ex tent to which the eight chemica l enginee ring program s that rank ed highest in 1982 retained their high rankings in 1995, and 2) the ex tent to which these program s persisted in hir ing thei r own and one another's gra duates. Th e eight universities and their respective ranking in 1982
Slimme r / 997

and 1995 are: Minnesota, I, I; Wisconsin , 2, 4 ; Califo rnia, Berk eley. 3, 3; Ca lifo rnia Institute of Tec hnology 3, 6; Stanfo rd. 4, 7; Delaware, 5, 8; Massachu setts Institute of Techn ology, 6, 2; Illinois. 7. 5. Eac h of the eight program s that ranked highest in 1982 was again ran ked among the top eight programs in 1995. T he nam es and alma maters of the full-time faculty members in these eight program s in 1982 and 1995 were ob tained from the Internet. The median perc ent age of faculty members who had obtained their doct oral degrees from their ow n school or fro m one of the other seven top-rank ed schools in 1995 was 70.8 % (range, 50.0% to 91.7 %). Thi s is very similar to the 67.4 % (range, 50.0% to 75.0 %) fig ure in 1982. Sincerely ,

Jeffrey H. Bair Divisio n ofSociology Emporia State University Emporia KS 66801


/ 77

You might also like