CONCEPTUAL

DESIGN OF
CHEMICAL
PROCESSES

James M. Douglas
of Massodwsttts
McGraw-Hili Rook Company
New York 51 Louis San Frlnciseo !lockland 80101. Cliracu Colorado Springs
Ibmburll LISbon London Madrid Muico M,I.n Monrreal New Deihl
Oklahoma Cily rlns Sail Juan Sao Paulo Sinpporc Sydnq' Tokyo Totonlo
CONCEPTUAL DESIGN OF CHEMICAL PROCESSES
INTERNATIONAL EOIlION 1988
Exclusive rights by McGaw-Hili Boote Co.- Singapore
lor manufacture and export. This book cannot be
(....xported from the country to which it is consigned
by McGraw-Hill.
10 09 Of! 07
20 09 08 07 06 OS 04 OJ
PMP BJE
Copyright G1988 by McGraw-Hill, nco All rights reserved.
No part atlhis publication may be reproduced or distributed
in any form or by any means, Of 510fed In a data base or
retrieval system, without the prior written permission of
the publisher.
This book was set in Times Roman.
The editors .....ere B.J.Oat1l: and James W.Bradley.
The production supervisors were Diane Renda and LDuise Karam.
Ubl'ary of Congress Cataloging-in-Publieation Dala
Douglas,James M.(James Merrill)
CoooeptuaJ design 01 chemical processes.
(McGraw-Hili chemical engineering series)
Bibliography:p.
Includes index.
1. O'Iemical prooesse5. I. Tille. II. Series.
TPl55.7.D67 1988 660.2'81 87-21359
ISBN D-OHl1n627
When ordering this title use ISBN 0-07100195--6
Printed In Singapore
ABOUT THE AUTHOR
James M. Douglas, Ph.D.. is currently a professor of chemical engineering at the
University of Massachusclls. Previously he taught at the University of Rochester
and at the University of [klaware. Before entering leaching, he spent five years at
ARea, working on reactor design and control problems. l-Ie has published
extensively in areas of reacting engineering, process control (including two books).
and conceptual process design. He won the Post-Doctoral FellowshIp Award at
ARea, the Faculty Fellowship Award at the University of Massachuselts, and the
Computing and ChemIcal EnglOeerlOg Award of AIChE.
DEDICATED TO:
The loves of my life,
My lovely wife. Mary E. (Belsy) Douglas,
My mOlher, Carolyn K., and the memory of my
falher. Merrill H. Douglas,
My two wonderful kids, Lynn and Bob,
aod to my colleagues, who have taught me so much aboul
design and control,
Mike Doherty, Mike Malone, Ka Ng, and Erik Ydstie,
and to my students, who have suffered so much.
CONTENTS
Preface ..
Part I A Strategy for Process Synthesis and
Analysis
The Nature of Process Synthesis and Analysis I
I-I Creau,c Aspcc;:u of Proo:;;ess Onlg" 3
1-'
A IhcraKhlCal Approach to Conccplual Iks.lgn

1-3 Summary. Eac:rcue5, and Nomcnclil1urc

"
2 Engineering Economics 2J
'-I Cost Informalioo Requned
"
,-,
Estimating Capital and OpelaliDg Costs 32
'-3 Toral Capital 'nvestment and TOIII r o u ~ Costs 37
'-4
Time Value of Money
"
,-,
Measures of Process Profitabllll)'
"
,-. Simphfymg the economIC AnalysIs for Conceptual Designs ...
'-7
Summary, Eacrciscs. and Nomenclature ..
3 Economic Decision Making: Design of a
Solvent Recovery System 72
3-1 Problem Dclimllon and General ConsideratlODS 72
3-'
Design ora Gas Absorber. FlowshcCl, Malena! and Energy
Balances, and Stream COSIS
" 3-3 equipment Design Conslderauons
" 3-4 Rules of Thumb
" 3-' Summary, eXerCISeS, and Nomenclature 90
x;
xii xiii
Part II Developing a Conceptual Design and 9 Cost Diagrams and the Qujck Screening of
Finding the Best Flowsheet 97
Process Alternatives 28.
'-1 Cost Diagrams 28.
4 Input Infonnation and Batch versus Continuous 99 '-2
CQ5t Diagraml for Complex Processes
'"
Input (nfonnabon 99
'-3 QUl<:k Screenmg of Process Ahemalh"es JOJ
'-1
'4 HDA Process J08
.-, u'"el! DecIsion 8atdl venus Continuous
107
.-, Summary, uerasc. and Nomenclature Jll
'-J
Summary, Elercues., and Nomenclature 111
5 Input-Output Structure of the Flowsheet 116 Part 1\1 Other Design Tools and Applications 317
'-1
Decisions for lhe Input-OulPUI Siructure 116
Deslgn Variables, Overall Malerial Balaooes. and Su"am Costs 123
10 Preliminary Process Optimization JI9
'-J
Process Altc:malives IJ2
10-1 DesIgn Vanabla and Economic Trade-oll"s J20
'4
Summary, EJ.erriscs" and Nomc:oclaturc IJ2
10-2 Cost Models ror Prooess Units J27
Io-J A Cost Model ror a SImple Process m
6 Recycle Structure of the Flowsheet 137
104 Approlimate Optlmiution Analysis J40
10-' Summar). Exercises. and Nomenclature
".
..1 DeCIsions that l)etermme the Recycle Structure 137
.., Recycle Material Balances
14'
II Process Retrofits
.. 3 RuetOf Hut Etreeu
14'
m
... EA:juilibnum L!mltalJOns
14. II-! A Systematic Plocedure for Process PeHofilS
".
..,
Compressor Design and Costs
13J 112 HDA Process Jl8
.-.
Reaelor DesIgn
13. 11-3 Summary and ExerCIses 368
'-7
Recycle F.oonomlC EvaluatIOn
138
"8
Summary. berClSeS,. and NOOlCnclature 13. 12 Computer-Aided Design Programs
(FLOWTRAN) J69
7 Separation System 16J
121 General Struclure of Computer-AIded Design Programs J70
12-2 Malenal 8alaoo: CalculatiOns J7l
7-1 General Slructure or the SeparatIon Syslem 16J
12] Comp1ele Planl Simulation J97
7-'
Vapor Reoovery System 168
124 Summary and Exercises
'04
7-J Uquid Separalion SYSlem 172
14 AZC'olroptC Systems 189
13 Summary or the Conceptual Design Procedure
7-' Rigorous Material BaJances
204
7-'
SummaI"}'. ElerCJSCS. and Nomenclature 211
and Extensions of the Method
40'
13-1 A RevlC"" 01 the HIerarchICal Decision Proa:dure for
Pelrochemical Processes 406
8 Heat-Exchanger Networks
21' 1]-2 Design of Solids Processes and Batch Processes 408
8-1 Minimum Heating and Cooling Requirements
21'
1]-] Olher SignifICant Aspe<:ls or lhe Design Problem 412
8-'
MInimum Number of ElchangeD '30
8-J Area Estlm.ates
2JJ
84 Design of Mlnrmum-Enerl)" Heat-EJ.cbanger Networks '36
Part IV Appendixes 42J
8-'
Loops and Paths
248
8-'
Reducing the Number of
231
8-1 A More Complele Design Algorithm Stream Splilting 231 A Shortcut Procedures for Equipment Design 42l
8-8 Heal and Power Integration
261
A-I Number of Trays ror a Gas Absorber ."
8-'
Ileal and DIstillatIOn '64 A-' Dist.illation Columns" Number or TraY$
4]'
8-10 HDA Proass
27J
A-J Design of Gas AbsorbeD and Dlslillauon Columns
."
8-11 SummaI"}'. EJlerasc::s, and Nomendature: '84 A4 DIstillation Column Sequencing 461
xit' CONTEP'lT$
A'
Complex Distillalion Columns
466
...
Energy Integralion of DiSllllalion Columns
.18
A7
Heal_Exchanger Design
.86
... Gas Compressors
.90
A'
Design of Refrigeration SySlems
.90
A,IO Reactors
S07
A-II Summary of Shortcut Equipment Design GUldel.tncs and
Nomenclalure for Appendix A
S07
B HDA Case Study
'"
C Design Data
543
CI
Hydrocarbon Vapor-Liquid Equilibria
543
C, Temperature Ranges for some Materials
547
D FLOWTRAN Input forms
548
DI Component List
548
D'
IFLSH
""
D3 AFLSH
SSl
D4 SEPR
'"
D'
ADD '54
D6 SPLIT
'"
D7 PUMP
,,.
D8 GCOMP
'"
D'
SCVW ",
D-IO DSTWU '6'
Dll REACT
""
E Cost Data
S6S
EI Operating Costs
'"
E'
Summary of Cost Correlallons
,.,
F Conversion Factors
S78
Indexes "I
Author Index "3
Subject Index
so,
PREFACE
llLis book describes a systematic procedure for the conccptual design of a limned
class of chemical processes. The goal of a conceptual design is to find the best
process f1.owsheet (i.e., to select the process units and the interconnections among
these: units) and estimate the optimum design conditions. The problem is dif-
ficult because very many process alternatives could be considered. In additIOn,
experience indicates that less than I % of ideas for new designs ever become
commercialized. Thus, there are many possibilities to consider with only a small
chance of sUCGCss.
In man} cases the processing costs associated with the various process
alternalives differ by an order of magnitude or man:, so that we can use shortcut
calculations to screc:n the alternatives. However, we must be certain that we are in
the neighborhood of the optimum design condilions for each alternative, 10 prevent
discarding an alternative because of a poor choice of design variables. Hence, we
use cost studies as an initial screc:ning to eliminate ideas for designs that are
unprofitable. If a process appears to be profitable, then we must consider other
factors, including safelY, environmental constraints, conlrollability, ele.
We approach the synthesis and analysis problem by c:stablisWng a hierarchy
of design decisions. With this approach, we decompose: a very large and complex
problem into a number of smaller problems that are much simpler to handle. By
focusing on the decisions that must be made at each level in the hierarchy (e.g.. Do
we want to add a solvent recovery system?), we can identify the existing
technologies that could be used to solve the problem (e.g., absorption, adsorption.
condensation) without precluding the possibility that some new technology (e.g., a
membrane process) might provide a better solution. Moreover, by 'listing the
ahemative solutions we can propose: for each decision, we can systematically
generate a lisl of process alternatives.
In some cases it is possible to use: design guidelines (rules of thumb or
heurislics) 10 make some deciSIOns about the structure of the flowshect and/or to
set the values of some of the design variables. We use order-of-magnitude
"
arguments to denve many of these heunsucs, and we use a simple analys15 of this
type to identify the limllatlOns of the heuristics. In many cases. no heuristics are
available, and therefore we develop shortcut design methods that can be used as a
baSIS for making decislons_
B) follOWing this hierarchical decision pr()(X(Jure. a bt=ginmng designer can
substitute the e\'aIUalion of a number of exIra calculations for experience dunng
the dC\'elopment of a conceptual desIgn Since shortcut calculatIOns are used.
however. the penalty paid in the lime required to screen more alternallves is nol
'cry high Of course, as a designer gaUlS experience, she or he will be able to
recognize what alternatl\es do not need to be conSidered for a particular type of
process and thereby obtalO an increase m effiCiency, Note also that expenence
normally IS required for assessing lhe operability of a design. and lherefore a
begmner should always get an experienced designer to review the resulls of the
design study
of Ihe Text
The text is meant to be used in a onesemester, seDlor-levei course in process design
for chemical engineenng students. We present the material as a Iccture course. A
single case study is carried throughout the te.:l.l 10 illustrate the ideas, and the
homework assignments mclude lhe evaluation of alternatIves for the oc:ntral case
study, as \lIell as sevcral other case studies. The purpose of these Olher case studies
is to Iw':lp the student undcrstand the similarities and differences betwttn \'arious
t)pes of processes (e.g. smgle reactIOns 'ersus product distnbutlon problems. cases
where gasregcle costs domlOate. cases where liquid separation costs dominate.
the choice between recychng or removlOg by-products formed by reverSible
reactions, the economic trade-offs encountered when a gas recycle and a purge
stream is used, etc.). The focus is on scrttning calculations, although a
aided design program is eventually used to verify the apprOllim3liQos
Part I discusses a slrategy Qfsynthesis and analysis.. In Chap. I it is nQled lhat
only about I %of ideas for new designs ever become commercialized, SQ thai we
need an efficient procedure for eliminating poor projects Similarly. sinoc: design
problems arc always underdetined and we can orten generate 10- to 10' alternati'e
processes even for a single product plant. we oeed an efficient way of screening
process alternatives. These discussions provide the motivation for the use of
shortcut calculations. Also, a procedure for decomposing process nowshecls into a
hierarchical sct of simpler problems is presented.
Otapter 2 presents an mtroduction to engineering economics, including a
discussion of various measures of profitability. In addltioo. a simple economic
model thaI is useful for conceptual designs IS de\'e1opcd
Chapter 3 presents a very simple design problem (actually a subsystem of
what could be a larger dC"ign problcm) This example how simple it lQ
generate proocss alternatives. the need for design heunstlcs. the: origll' of dcslgn
heuristics, the IImltallOnS of design heUristiCS, the IOleractiolls among prQOCSSlOg
units, the need for a systems Viewpoint in place of a unit operatIOns viewpolOt. and
shorlcut design melhods can be developed
Part II presents the details of the hierarchical decision procedure for the
synthesis and analySIS of conceptual designs. Chapter 4 describes the infonnation
needed to get started, and lhe decision of designing a balch versus a continuous
process is discussed Chapter 5 presents the important decisions for the input and
output structure, the identification of the important design variables at this level of
complexity, and shortcut procedures 10 calculate the stream cosls and the costs of a
feed compressor (if one IS reqUired). Chapter 6 introduces the additional deciSIOns
required to fix the overall recycle structure of the f1owsheel. i.e., the interaction of
the reactor system(s) with the remainder of the process The reactor cost and any
gas-recyck compressor costs are evalualed in terms of the design variables This
discussion is limlled to single-product plants.
At present. lhe systematic preliminary design procedure is also limited to
vaporliquid processes. For this class of processes. the structure of the separation
system (i.e. the general structure. vapor recovery system alternatives, and the
decisions fQr the liquid separation system) is described in Chap. 7. Chapter g then
presents a synthesis procedure for the heat-exchanger network. At this point, a
base-case design and an estimate of the oplimum design conditions arc available
Our basic design strategy is to develop a base-case design as rapidly as
possible. simply listing the process alternativcs as we go along. to determine
... hether there is SQmething about the process that will make all Ihe alternatives
unprofitable. Provided that our base-case design appcan to bt= promising.
the methods in Chap 9 to screen the process allernativcs. Thus. at this POlOt we
allempt to identify the best process flowsheet.
Part III presents some other desIgn tools and apphcations. In the procedure
presented in Chaps 4 through 9. we used cakulations to estimate the
oplimum design conditions because we were contlOually changing the structure of
the ftowshtt\,.. Once we have identified the best flowshect, we can use more
sophisticated optimization procedures. However, to assess the degree of sophistica-
tion that is desirable, we present an approlfimate oplimizatiQn analysis in Chap. 10.
This approllimate optimization procedure helps 10 identify the dominant econQmic
trade-offs for each design variable, the dominant design variables, and an indica-
tion of how far a design variable IS a..... ay from the optimum without knowing thc
exact value of the optimum This approll:imale optimization analysis is also very
useful for retrofit studies and for oplimum sleady-state control calculations.
In Chap. II we use the same techniques for process retrofits thai we used to
develop a design for a new plant. A systematic procedure is presented for
retrofilting processcs, mcluding completely replacing the ell:isting plant wilh either
the same or a beller process alternative. The approximate optimization procedure
is used to help idenlify the dominant opcratlOg variables and the equipment
constraints that pre\'ent the opcratmg costs from being minimi7.ed. Then, based on
these results, additional equipmenl capacity is added until the incremental,
annualized equipment coS! balances the incremental decrease in operating
"""
In Chap 12 diSCUSS Ihe use of a compuler-alded design program 10
Improve lhe accuracy of the shQrteul calculations Chapler 13 presents a summar)'
of the design procedure, brief outlines of hierarchical decision procedures for
x"iii HUAU
and balch processes. and a brief dIscussIOn of what remams to be done after a
conceptual design has been completed
The appendixes present some auxlhary information. The shortcut mOt.lels for
equipment design are dISCussed in AppendIx A, and the complete details of a case
study aregnen in AppendIX B Some samples of design data and cost data are gIven
in Appendixes C and E.
AcknoMledgmenls
I am \'ery appreciative of the etrorls of A Ene Anderson (formerly with ARCO).
Duncan Woodcock of Imperial ChemIcal Indusllies, Edward C. Haun of UOP
Inc., JetrKantor. University of Notre Dame; Carl F. King from duPont, E. L Sherk
from Euon, R. Hoch (formerly with Hakon International), John Scinfeld,
California Institute of Ta;hnology and J. J Sirola from Tennessee Eastman Co. for
their careful review of the texl. Similarly. I am grateful to the chemical enginccnng
students at the University of Massachusetts and to tbe students from Imperial
Chemical Industries (United Kingdom), Rohm and Haas, Monsanto, Union
CarbIde and Celanese, for many valuable comments concerning the course
material. In addition, I must acknowledge the numerous conlfibutions that my
colleague Mike Malone made to the text, and I want to thank my other colleagues
Mike Doherty, Erik Ydstic, and Ka Ng for their feedback when they taught the
material. The contributions of my graduate students, particularly Wayne Fisher
and Bob Kirkwood. also need to be ackno.... ledged.
Of course, I am especially grateful to 01) lovely .....ife. Betsy, to my children.
Lynn and Bob, and to my mOl her, Carolyn K.. Douglas, for [hel! support dunng
the preparation of the text. SImilarly, Pat Le..... IS. my admmistrative assistant. and
Pat 8arscheoski. who did the typmg, pro\ldcd mucb oeeded support.
James M. Douglas
CONCEPTUAL DESIGN OF
CHEMICAL PROCESSES
PART
I
THE
STRATEGY
OF PROCESS
SYNTHESIS
AND ANALYSIS
CHAPTER
1
THE
NATURE
OF PROCESS
SYNTHESIS
AND A ALYSIS
J.I CREATIVE ASI'ECfS OF PROCESS
DESIGN
The purpose of engmeenng is 10 create new malenal wealth We auempt to
accomplish this goal in chemical engmeering via the chemical (or bIological)
transfonnation and/or separation of malcrials.. Process and plant design is the
creative activity whereby we generate ideas and then translate them inlO equipment
aDd processes for producing new malerials or for significantly upgradmg the value
of cJ:isling materials..
In any panicular company. we might Iry 10 generate ncw Ideas.
To produce a purchased ra..... malenal
To can\trt a waste by-product 10 a valuable product
To create a completely new malenal (synthetic fibers, food. bioproccsmg)
To lind a lIew way of producing an cxlsting product (a new catalyst. a
bioprocessing allernative)
To exploH a new technology (genetic engmeenng. expert systems)
To exploit a new material or construction or hlgh-
prc:ssure-operalion, Specialty polymers)
J
As an indIcatIon of the sucss of the engineering effort, we note Ihal
over oflhe products sold by mosl chemical companies developed during
the last decade or Iwo_
SUCCe<iS
Despite thiS ucdlcnt record of SUct%S5..... e should realize that very few ne.... ideas.
either fljr Improvmg Clllsting processes or fordevdoping new prOttSs.lead 10 new
....-eahh In fact, lhe chano:::s of commercialization at the research stage for a new
process arc only about I to .. at thede\e1opment stage they arc about 10 to
and al the pilot plant stage lhey arc about 40 to 60% Of course, we expect that
the success rale for process modifications Will be higher than that for completel)
new proocsscs. but the economic rewards associated with these safer projects will
have a Significantly lower potential
It is not surprising that so few ideas 10 engrneering ever prove to be fruitful:
the same paltern holds for any type of creative activilY. Sincc experience mdicates
lhat only a small number of ideas e,,'er will ha ve a payout, we see that et'olUalion IS
one of the most slgmficant components of any design methodology. In fact. process
synlhesis. i.e.. the selection of equipment and the interconnections belween that
equipment which will achieve a certain goal, is really a combination of a synthesis
and analysis activity.
S)nthesi.. and Anallsis
Perhaps the major feature thai dlStlllgulshcs dcslgn problems from olher t)pes of
engmeenng problems 15 that they are underdefined: i_c.. only a very small fraction
of the mformatlon needed to define a design problem is available from the problem
statemenl_ For example. a chemISt might discover a new reaction to make an
eKistmg product or a new catalyst for an existing, reaction. and we
want to translate these discoveries to a new process. Thus, we slart with only a
knowledge of the reaction conditions that ....-c: obtain from the chemist, as well as
some infonnation about available raw materials and products that we obtain from
our marketing organization, and then we need to supply all the other information
that we need to define a design problem.
To supply this missing information. we must make assumptions about what
types of process units should be used, how those procc:ss unils will be
nected, and what temperatures., pressures. and process flow rates will be required.
This is the synthesis activity. Synthesis is difficult because there arc a vcry large
number (10 to 10') of ways that we might consider to accomplish the same goal
1-leIlCC. design problems arc very
Thes<: values rerrest.nr the of CSl"nalei supplied by r.>n><b worl I'll In Konom,c
evalua!r('ln I'ours of maJO' chemrcal and I"'I,olwm OOm",,"1tS
SCTION 'I CIlUTIVI! ASl'fCTll I'llocas 5
Nonnally......e wanl 10 find the process alternative (OUI of the 10 to 10
0
possibilities) that has Ihe lowest cost, but .....e must also ensure that the process IS
safe. will satisfy environmental is easy to start up and operate. etc. In
sotnC cases, we can usc: rules of thumb (heuristics) 10 eliminate certain process
alternatives from further conSIderation., bUI in many cases it is ne:ttSsary to design
V.lnous altemati\'es and Ihen 10 compare their costs Experienced designers can
mmlmlze the effort rtquircd for thIS type of evaluation because they can often guess
the costs of a parhcular unit. or group of units. by analogy 10 another
However, beginmng desIgners normally must design and evaluate more altema-
tu-es in order to find the best altemati\-e_
When experienced designers coDSider new types of problems, where they lack
experiencc and where they cannot idenlify analogies. they try to use shortcut (back-
of.the-envelope) design procedures as the basis for comparing altemati\cs. These
back-of-the-envelope calculations are used only to screen alternatives.. Then if the
process appears to be profitable, more rigorous design calculations an: used to
develop a final design for the best alternative. or the best few alternatives.
Because of the underdefined and open-ended nature of design problems, and
because of the low success rates. it is useful to develop a strategy for solving design
problems. We expeet that the strategy that a beginnmg designer would use fOI
synthesis and analysis would be different from that of an experienced designer.
because a beginner must evaluate many more process alteroatives. However. b}
usmg shortcut design procedures .....e can minimize the effort required to undertake
lhese additional calculations
Engineering Mefhod
If we reflect on the nature of process synthesis and analysis. as discussed above, '..e
recognize that process design aClually is an art., ie., a creatn'C process.
\l..e might try to approach design problems in much the same way as a pamter
de\-c:lops a painting. In other words. our original design procedures should
oorrespond to the development of a pencil sketch, where we want to suppress all
but the most significant details of the design; i.e., we want to disco..-c:.r the most
expensh'c parts of a prooess and the significant economic trade-offs. An artist next
c\-.aluates thc preliminar} painting and makes modifications. using only gross
outlines of the subjects SImilarly. we want to evaluate our first guess at a design
and generatc a number of process alternatives that might lead to imprO'o'emcnts. In
this way, we hope to ger...::ratc a "reasonable-looking," rough process before
we Slart adding much detail.
Then lhe artist adds shading, and the details of various objocts in the
painting and reevaluales the results. Major modifications be if the}'
soem to be wflrranted. In an analogous manner, the engmecr uses more rigorous
design and costing procedures for the most expensive e(luipment items, improves
the accuracy of the approximate-material and energy-balance calculations, and
adds detail in terms of the small. inexpensive equipment items that are necessary for
6 SECTJO'II I I CIU. TlVE ASPECTS or PlOCESS OESIGN
the process operations but do not ha\e a major impact on the total plant cost, e g,
pumps. ftash drums, elc,
Thus, .....e s..e thai bolh a palnllng and a process design proceed through a
scnes of successively more detailed synthesis and evaluation stages. Thatcher refen
to a solution sirategy of this type as SUCCeS!MC refinements, and he: call!> It
rngltli'erltlg mf'lhod otc that as .....c malc successivc refinemcnts, ""c should
always mainlalO a focus on the ol}f!rall problrm.
If wc accept thiS analogy betwccn cnglDeering design and art, then wc can
recognize some other IOterestlOg fealures of the design process. An artist never
really completes a painling; normally the work is termlOatcd whenever the
additional effort reaches a point of diminishing rcturns; i.e., if little added value
comes from much additional effort, the effort is DOl worthwhile. Another feature of
art is that therc is never a single solution to a problem; i.e... thcre arc a variety of
ways of painting a "great" Madoona and Child or a landscape; and in pr0ces5
engir-cering nonnally different proc:essing routes can be used to produce the same
chemICal for essentially the same cost. Slill another analogy between cngineering
design and art is thai it requires judgment to decide how much dClaii should be
included in thc various stages of painung. just as it docs in a process design.
Of course, numerous SClcntific pnnciples arc used ID the dcvdopment of a
design, bUllhc overall activity is an art In fact, it is this combinatton ofsocncc and
art in a creative activity that helps 10 make process design such a fascinatlDg
challcnge to an englOcer.
u\els of Enginei.'ring Designs
Now we sec that there arc a number of levels of engmcering designs and cost
estimalc:s that wc cxpect 10 undcrtakc. These vary from very simple and rapid, but
not very accurate, estlmatcs to very detailed calculations thai are as accuratc as we
can make. Pikuhlr. and Diazt e1assify these dcslgn estimates by the categories given
in Table 1.1-1.
They also givc the relativc costs required 10 oblain these eslimatcs. as shown
in Table 1.1-2. From tbis table we sec how rapidly cng!Decrio& costs increase as we
!Delude more detail In the calculations, Obviously, we wanl to noid Iargc dcslgn
costs unless thcy can be economically Justified.
C. M Thald",r. 1M ..1s EttgI-wr,"'fI. Mcrrill Columbus, OhIO. 1962, chap J
I A Ptkuhk and H E. D.u., COSI &umaunl Major P,oec:ss CIlnn. 84(21): 106
(11I17). These accu,aey bounds ..,11 var, "0111 one: eompany 10 anOther, and Ihc .ccuncy of Ihoc
dclalkd Will nO{ be Ih.s good dunn. ofh.gb 1IIftallon (Ihc cnon mlghl be as much as 8
1010'"".. even for a delalled cslmulle) Abo, nOrmall) Ihe ch4ncc 01 oblalmDI poIlu.e cnon Il; I,,:alcr
than lhal for nellamc urorl., 50 Ihal 11M: order-olmaSlUlu<!e esl,malc, I c. Ilcm I. ViOuld be rcportcd itS
+40 10 - 25 %(des.gn enIJrtCI"s seldom Overc:lilUnatc COSlJ,) S,mlla,I" hlpr eolltlrtgcOC' kcs rna, be
tndudcd III Ibc arloc' Icvcls(lhal J$, 10 10 2S}; In lIem } dropplDllO 10% ID .tcm oi) 10 accounl rorCOlls
not Included In tnc anal)'UI (.. hlCh IS somc:.. ta..l d.tre:rcnl r"ltD Ihc KCUnley 011"" c::IlUDalc)
SECTION I I Cll!A1WE ASI'ECT'S Of PlOClSS DESION 7
TABU: I.t-t
Types of desiCn estim.tcs
1_ Ord,u-ol-mapuludc ,""l1m.Ile(nbOc:lilunale) based oa .m"lar preVlOUI Wsl cia.... p,obabk aocunq
uaocdJi
2. Slud, elumalc (rad""cd a.lJm.llc) bued On Ir.DO... J.cdJf- 01 D\.lIJOI I!Cm!l 01 eqwpmUII. probabk
....."'t, up 10
1. Prdlllllnary ail' male (budget aulltonuUon ,""umalc. ICOpc ct.llIIU.!c) based oa sulliocr.l dala UI
pclmll ItIt aiUmalc 10 be budaclcd. p,obabk accuracy WIthin
(. DclinlUVoc c:lillmatc (pro}Cd control e1illma!c) hued OQ almost compkle datI, bul bdorc compktlOll
or dnwlngl and Ipce:,jiQlllolU. probabk a<:cUUC, Wllhln :I.
So DclaiJ.cd eltlffillc (CODtnldOI'1 csumalc) baled on compklc cnll_nnl dn"'mp. Ipco;tfiallOM.
aIld Nle .urve,s; probabl, &CQInq ",.thln J%
F,om It.. f"Ikulilr. ud H E. DIu, COll Eit.....lIDI Majol PtOCCSl Equ;pmcM, CIu-..... bot, IotIll) 106 /1'77)
For the case o! a ncw process. whcrc prcvious cost dala are nOI availablc, it
seems as if it would Dot be possible to dcvelop an order-o!-magnitude estimate
However, an cxpcric:oced designer can o\'crcomc this difticully by drawing analo-
gies between thc new proocss and other existing processes for which some data arc
available in the company files. Procedures for dc,eloping order-of-magnitudc
estimates have been described 10 the literature, but normally it requircs some
cxperience 10 cvaluate the results obtamed from Ihis t)'pe of calculatiOn.
For a beglOning dcslgner, with little or no experience. It would be useful to
ha'c a systcmalic approach for order-of.magmtude estimates We can
usc: order-of-magnitude arguments to Simplify man)' of the design calculations, and
.....e can limn our allcntlon to Ihe major pieces of process equipment as we carr)' out
a prehmmary process design The goal of this lext IS 10 develop a systematic
J. H - Pr(l(;C:$$ Step-Soonn, Mclhod fOf Makln& Quick Capllal EallmalQ,- C-.J' log., P. 207,
J:l1,AuJII5I 1910 0 H AIIc:n. and II; C Pace. PRcV!Kd Tcdtn>quc for P<aiallll Cosl Eaumallo,,-
CM""- 12(5) loi2 (March J, Ins).
TABU: 1.1-1
EagiDeC:ring costs 10 estim.tes (1977)
.........
51 lIIiIIi.- SI-$.! oniltiooo S5-s.5O MiItiooo
Typt of ntimalc ..... ..... ....
Siudy (S lhouurtds)
, "
02)0 ,0<O
Prclunmary ($ IhotlSdrtd.1 IS JS )060 lO90
Dcli... uvt (S .hOWoiln,hl
"'"
60 020 100230
F..- It.. P'ihli\. uod 1I E. o.a.. .ea.. Eou....' .... Ma)Ol Procea Equ,......... ""'"
&rr, Iot(ll) 106 11'71)
8 SECTION II ... HIU....CHlCAl ..,....(M,CH TO CONCEI'TUA.L OI:$IGI<
I J ... ... rPRO..CM TO CONcrI'TUA.L I>EllION .,
procedure of this type and then to show how the results can be extended to a study
estimate.
Detailed estImates are considered 10 be beyond the scope of this text
However, as noted before,the chance that a new idea ever becomes commercialized
is only about I .. so that .....e expect to undertake roughly 100 preliminary designs
for every detailed design. Hence. the methodology of conceplUal process design
should be mastered in considerable detail.
Other Applications of fhe l\1ethooology
Despite the fact thaI our primary focus is dllcctro to lhe design and evaluation of
new processes, much of the melhodology .....e develop is useful for other engineering
tasks. including basic research and technical service. In basic research, we want to
spend most of our effort studying those variables that wjJl have the greatest
economic impact on the process. and rough prooess designs will help to identify the
hlgh--cost parts of tbe process and the dominant design variables. Sinularly. in
technical service activities., .....e look for ways of improving an existing process. To
accomplish this goal, we need to understand the significant economic trade-offs in
the process, and it is useful to have procedures available for obtaining quick
estimates of the potential payout of new ideas. Thus, the methodology we: develop
will have numerous applications in the process industries.
IHeal
Ccmp....... I
Purge
""tH..
I
IHcat I
Reactor CooIllDl
- Ft""
IHcat I IHca' I H
2
, CH",
I
)-
1-
Benune

u
"
Jj
Toluene ].
"

"&
:D

i
Diphenyl T
ncURE 1.2-1
HDA process lAfttrJ M DDt.gloJ. AlOE J. JJ JH (l9&.S).]
Tbe e.xample we consider is the hydrodealkylal1on of toluene: to produce
The: reactions of Interest are
1.2 A HIERARCHICAL APPROACH TO
CONCEPT AL DESIGN
Example: Hldrodealkylatioo of (HDA
Process)
TIlls CIl"" .Iu<!y rC-p'nC'nIS" mod.fiecl "t'ntOll 01 1M 1961 tmhlule 01 Cberwcal En&,,,",,s
(AIChE) Siudenl Problem. J J J McKC"lIa. bc)"C"kJ"rd,o t>f Clw"'lC'oJ PrOCtssutQ _tl/N",,,.
vol "". Dekker, New York, t9n, pin. ror lhe on&,nli problem.oo .mUlIon
The homogeneous reactions take place In the range from IISOoF (below this
temperature the reactIOn rate 15 too slow) to (abo\'e this temperature a
significant amount of h}drocrackmg takes place) and at a pressure of about
500 psia. An e:xcxss of h}"drogen (a 5/1 ratio) is oc:c:ded to prevent coklOg. and the
reactor effluent gas must Ix rapidly quenched to I 15O"F in order to pre\'ent cokmg
m the: heat exchanger following the reactor.
One: possIble: fJowshc:c:t for the: process is shown in Fig.. 1.2-1. The toluene and
bydrogen raw-material streams are: heated and combined with recycled toluene
and hydrogen streams before they are fed to the reactor. The product stream
leaving the reactor contains hydroge:n, methane, benzene, toluene, and the unwant-
ed diphenyl. We attempt to separate most of the hydrogen and methane hom the
aromatics by wing a partial condenser to coodense the aromatics., and the:n we
flash away the light gases We: use the liquid leaviog this flash drum to supply
quench cooling of the hot reactor gases (not shown on the 80wshee:t).
We would like to recycle the hydrogen leaving in the: flash vapor, but the
methane:.. which enters as an impurity io the hydrogen feed stream and is also
produced by reactioo 1.2-1, will accumulate in lhe gas-reeycle loop. Heooe, a purge
stream is required to remove both the feed and the: product methane .from the
proceSS. Note that no rules of thumb (design guidelines) can Ix used to estlma.te the
optimum cona:ntration of methane that should be allowed to m the
gas-rec}"C1e loop_ We discuss Ihis design variable in much greater
Not all the hydrogen and methane can beseparated from the aromatics 10 the
flash drum, and therefore we remove most of the remainlllg amount in a distIllation
column (the stabilizer) to pre\'ent them from contaminating our benzene product
(12-1)
(I 2-2)
Toluene + H
J
....... Benzeoe + CH..
+ Hz
The engulec:nng method (or the: artist's approach) indicates that we should sohe
design problems by first de\'eloping vcry simple solutions and then adding
sua:essi\'e layers of delail. To see hoYt we can this approach for process design
problems. we consider a typtcal fJowshc:c:t for a petrochemical process, and Ihen we
look for ways of stripping away layers of detail until we obtain the simplest
problem of interest. By applying this procedure to a number of different types of
procc:ssc:s, we: might be able to recognize a general pattern that we can use as the
basis for synthesizing Dew processes.
aod
10 5EClION U A HIE.....CHlCAl AI'f'lOACH 10 COI\ICEPTI,IAl OfSlGN
The benzene is then recovered in a second dlstlllation column, and finally, the
recycle toluene is separated from the unwanted diphenyl. Other, alternative
flowsheets can also be drawn, and we discuss some of these as we go through the
analysis.
Energ:r Integration
The process f10wsheet shown III Fig 12-1 IS not very reahsllc because: II implies
that the heating and coohng requirements for every process stream will take place
In separate heat exchangers uSing external utilities (coohng water, steam, fuel. etc.).
In the lasl decade, a new design procxdure has been developed that makes II
possible 10 find tbe minimum healing and cooLlIlg loads for a proa:ss and the heat-
exchanger network that gnes the "best" energy integration. This procedure is
described In detail in Chap. 8.
To apply thIS new design procedure, we must know the flow rate aDd
composition of each process stream and the inlet and outlet temperatures of each
process stream. One alternative f10wsheetthat results from this energy tIllegratlon
analySIS IS shown 10 Fig. 1.22. Now we see that first the: reactor product slream IS
used to partially preheat the feed entering the reactor. Tbc:n tbe: hot reactor lases
arc: used to dnve the toluene: recycle column reboiler. to preheat some more foc:d, to
drive the stabdi7.c=r column reboiler,to supply part of the benzene: product column
reboiler load, and to preheat some more feed before the gases enter the partial
condenser_ Also the toluene column is prc:ssurized. so that the condenslllg
temperature for toluene is higher than the: boiling point of the: bollom Slream in the
benzene column With this arrangement, conde::mmg toluene can be used to supply
some:: of the benzene rc:bollc::r load. ins-lead of using steam and cooling water from
external sources of utllltlcs.
II we compare the energy-lIlu:gratc:d f10wsbect (Fig.. 1.2-2) with the flowshec:t
indicating only the need for heating and cooling (Fig. 1.2-1), then we see that the
energy integration analysis makes the flowshcct morc: complicated (i.e.. there arc
many more interconnections). Moreover. to apply the energy integration analysis,
we must know the flow rate and composition of every process stream, i.e. all the
process heat loads including those of the separation systc:m as well as all the stream
temperatures. Since we need to fix almost aU the Bowshcct before we can design the
energy integration system and since it adds the greatcst complication to the process
flowsheet, we consider the energy integration analysis as Ihe last step in our proocss
design procedure.
Distillation Train
Let us now consider the train ofdistiUation columns shown in Fig. 1.2-1. Since: the
unwanted dlphenyl is rormed by a reversible reaction (Eq. 1.2.2), we could recycle
ThUi wlUllon WH developed by 0 W To..n!iCnd al lmpenal ChcmlCllllndLl$lf1CS, Runcom., Unlled
Kmldom

.l!
" :t
1" u
u
.-
" !'.
] tl
2"u
.
-' "
0-
I;
l"
"u

0-
E
0

"
'T

,f u:
"

u ,

.ll
p
r

1

,
t
uwnjO:) ( ) )
c1
uwnlOO In!llqllflS ;)

..
:t

U
r
;..
I
'"
,
I
1
uwnlOO
,
,
>.
T
e r
u
I
0- L
is
+0
"
'"
" u
"
u

-
E
"
,

"-

t
0
"
II
12 sCTKm 1.;1 Il NIElIllllCHIClll IlPPIIOM'H 10 C'OI'I("(I'Tl'lll DESIGN SECTION 11 Il HlE....IICHlCIll ""PlIOlleH 10 CONO:I'TlJIll DESIGN 13
Benzene
Toluene
(To recycle)
might be chealXr than uSing Ihe configuration shown in the original flowsheet
(Fig. 1.2-1).
The heurisllcs (design guidelines) for separation require a knowledge
of the feed composition of the Slream entering the distillation tram. Thus. before we
conSider the associated v.lth the design of the distillation train, we must
the remalOder of the llov.shcct and esllmate the process ftows. For thiS
reason .... e consider the design of the distIllation train before we consider the design
of the heat-<:ltchanger network
Diphenyl
FIGURE t.2--3
Altcmale dislillalion IralOS,
diphenyl with and lei it build up 10 an equilibrium le\'eL ThiS
alternative would make il powble 10 eliminate one of the distillation columns.,
although the flow Tate Ihrough the reactoT would
If we decide: to rover the diphenyl as Fig. 12-1 indlC3le:s, ....e expect Ihallhc
toluene-diphenyl spill 1,1,'111 be \er)' easy. Therefore. v.e might be able 10 use a
sldestream column to accomplish a benzene-toluene-diphenyl spht. That is, we
could recover the benttne o\erhead. remove the toluene as a sidestream below lhe
feed, and reco\'er the dlphenyl as a bollom Slream (sec Fig. 1.2.3) We can still
obtain very pure benzene O\erhead if we take the toluene sldestream olf below the
feed. The purity of the toluene reqcle will howe\'er, if It IS reco\'ered as a
sidcstream, as compared to an overhead producL Since: there is no specification for
the recycle toluene. the purity might not be important and the savings might be
worthwhile. Similarly, we ellpc:ct that the methane-benzene split in the stabilizer IS
easy. Then, recovering benzene as a sidestream in a H
1
and CI-I.
benzene-toluene and diphenyl spliner (a pasteurization column) (see Fig. 1.2-4)
Condensallon (high pressure, or low lemper-Hllre. or both)
Absorption
Adsorption
A membrane process
To estimate", hether a vapor recovery system can be economically justified.
"C muSI estimale Ihe flow rates or the aromatics lost in the purge as well as the
and methane flow in the purge. Hence, before we consider the necessity
and/or the design of a vapor recovery system. we must specify the remainder of the
ftowshcct and "'e must eSllmale the process flows. We consider lhe design or the
'apor ra;overy system before that for Ihe liquid separalion system because the nit
streams from the options for a vapor rcco\'ery system listed above (e.g.. a gas
absorber) normally include a liquid stream that is sent to the liquid separation
system.
Simplified Flowsheet for dw Stoparation Systems
Our goal is to find a way of simplifying f1owsheets. It is obvious that Fig. 1.21 is
much simpler than FIg. 1.22, and therefore we decided 10 do the energy integration
last. Similarly, since we have (0 know the process flow ratcs to design thC vapor and
hquid recovery systems, .....e decided to consider Ihese design problems just berore
lhe energy integration. Thus, we can simplify the flowsheet shown in Fig. 1.2-1 by
drawing it as shown In Fig. 1.2-5, The connections between the vapor and liquid
recovery systems 10 Fig. 12-5 are discussed in more detadlater.
We now ask ourselves whether all processes can be represented by the
Simplified f10wsheet shown m Fig.. 12-5. Sma: this flowsheet CQntaiOs both gas- and
Vapor Reco\'ery S)'SlenJ
Referring again 10 Fig 1.2-1. we consider the vapor flow leaving lhe flash drum. We
know that we never obtain sharp splits in a flash drum and therefore that some of
the aromalics will leave with the flash vapor. Moreover. some or these aromatics
",ill be lost in the purge stream. Of course, we could recover these: aromatiCS by
IOslalling a vapor recovery system either on the flash vapor stream or on the purge
<ITeam
As a \apor reco\'eT}' SYSlem we could use one of these
FIGURE 1.1-4
A11cnl.ale d,mllatiOfllrams Diphe.nyl
Toluene
(To recycle)
14 SECTlON' H A HfE...ICHk4t 4''''0AQl TO C'ONl:(P'TU4t OESK.N
FIGURE U-S
HDA Jc:par1lllOn s)'$1cm.. [Aft.... J M Dovgla. Aloe J, 31 JjJ (JSlIj).]
Benzene:
I I
tl
2
, CH
4
---.r-'-----'-i-
'-----'-
Toluene
FIGURE L1-7
HDA Slruchne (Af/IO' J M Ou,,!/I,u, Alo.e J, JI JH (/lI8j)]
to understand whal design questions arc importanl to obtalO Ihls simplified
wilhom worrying aboul the addItIonal complexitIes caused by the
separation syslem or Ihe: energy integralion network For example, we can siudy
the factors thai determine Ihe number of recycle: strcams., heal effects in the reactor
equilIbrium limitatIons in the reactor, etc. Thus, conllnuing 10 stnp away levels or
detail, we sec that we want 10 stud) the recycle Slrul.:ture of Ihe flowshcct before
considering the details of the separallon system
,
CH,
phenyl
Vapor recovery
Pu<g
system
H
1
,
, CH
4
Reactor Phase
system Splil
luene:
"'
LiquKt separalion
SYSlem
D;
To
liquid-recycle loops. but somc processes do not contain any gaseous componenlS-,
.....e do oot eXp1 the resulls to be general (Sec Sec. 7.1 for other altemamcs.)
However, .....e can slmphfy Ihe fto\lo'Jhcct stili more by lumping lhe vapor and hquid
separation systems in a slIIgle box (sec Fig. 1.2-6). Thus, \Ioe consider the
specification of the general structure of the separatIOn system before we consider
tbe specificalion of either the vapor or the liquid recovery systems.
Recycle Structurc of the Flowsbeel
Now we have obtained a very simple fto\lo'Jhcet for the process (Fig. J.2-6). We can
use this simple representation to eslimate the recycle flows and their effect on the
reactor cost and the cost of a gas-recycle compressor, if any. Moreover, we can try
Input-OufPUI Siructure of the Flowshecl
1.2-6 provides a \'erJ simple fto\lol>hect. bUI we consider Ihe possibility of
obtaining an even SImpler representatIon Ob\iousl}. If wc dra", a box around the
compkte process, we \10'111 be left Wilh the feed and product Slreams. Al first glance
(sec FIg 1.2-7), IhlS representation might seem 10 be too SImple, but II \10111 aid us m
understanding Ihe deSign ...anablcs lhat affect the overall malenal balances withoul
Inlroducmg any other eomplicalions Since raw-malerial COSiS normally rail in rhe
range from 33 fO 85 of the 10lal product costs, thc o\'erall material balances are
a in a desIgn Also, we do not want to spend any lime investigating
Ihe dcslgn vanables In the rangcs where Ihe products and by-products are worth
less than thc raw materials. Thus, we consider the input-output Slfucturc of the
ftowsheel and the decisions that affcci this structure before we consider any recycle
syslems.
Possible Limitalions
By simplifying a f1owshcct, we can de\e1op a general procedure for
auaclnng design problems. However, our onglOal ftowshcet described a contin-
uous, vapor-liqUId process lhat produced a single product and involved only
SImple chemicals (no polymers or hydrocarbon CULS) There are a large number of
processes Ihal saIIS() these hmllaltOns, and so we Iry 10 develop Ihis systematic
Diphenyl
Benzene
Purge
H,
I I
H
1
, C
-
Rcaao, Separation
-
-
system system
-
I I
Gas
Toluene
Toluene: recycle
FIGURE U-6
HDA,cqdewuclu,IO (Aft,.,.J M JH(/98,S)]
E L Grurm., wScltma b .. Mllenil Utg,19(9) 190 (April I", 1967) Abo
sec II E. Kyle, CMm Eng p,,,,.Il1(Il) 17 (1986). lor SOnIC dall commOOIlY dlCmlQlI
prodlKUQn to d1CmlQlb
16 SfcnOI'l 1.1 A lllUAIlCltICAl A"rIlOACH TO CQNl::ErruAI OBION
procedure In greatcr detail. Howevcr. batch processes may ha\c a somcwhat
difTercnt underlying structure (v,c often carry out mull1ple operations In a single
\essel). and certainly they are described dlfTerentl)' in terms of mathcmatlCal
models (normally ordinary diffcrential or partial diffcrcnllal equations Ins1ead of
31gebraic equations or ordinary differcntlal equations) lienee. our first deciSion
probabl) should be to distinguish bctv,een batch and continuous processes.
Hierarchy of Decisions
If we collect the results dISCUssed abo'e. we can develop a systematic approach to
process design by reducing thc deSign problem to a hierarchy of deoSlons; see
Table 1.2\. One greal advantage of this approach to design is that it allows us to
calculate equipment sizes and to estimate costs as we proceed through the levels In
the hierarchy. Then if the potential profit becomes negative at some level, we can
look for a process alternative or terminate the design project WIthout baying to
obtain a complete solution to the problem.
Another advantage of the proocdure arises from the fact that as we make
dectSlons about the structure of the flo.....sheet at various 1c"e1s, .....e know that if v,e
change these decisions. we .....iIl generate process altematl\cs. Thus, with a
systematic desIgn procedure for idenlifymg ahematives \Ioe are much Icss hkel} 10
O\'erlook some important cholttS, The goal of a conceptual design is to find the
.. best - alternati"e
Shortcut Solutions
Experience indicates that it is usually possible to generate a very large number (i.e..
often 10'" to 10) of alternative f10wsheets for any prOttSs if all the possibilities are
considered. Hence, it is useful to be able to quickly reduC'C the number of
alternatives that we need to consider. We nonnally screen these alternatives, using
order-of-magnitude arguments 10 simplif)' tbe process malerial balanccs, Ihe
equipment design equations, and the cost calculations. These shortcut calculations
often are sufficiently accurate to eliminate the or so, of tbe alternatives that
do nOI correspond to profitabk operation Then if our synthesis and analysis lead
TABtE t.J.1
I-tmrchy of decisions
t. tbleh venus conunllO\lS
Z. InJlul-ouCpul of the 1Io.'sheec
J. Recycle Ilructure of the no,",sheel
.. Geneul sCructurc of che sySlerl\
If. Va po. recovcry
b trquld reoo,ely system
So IIUl-Ucbanl"'r no:lWOTlc
SEcnON I J A HlF.llAkCHlCAl """.0"01 TO CQl'fCl'TU4l OESION 11
to a profitable solutIOn, wc repeat all the calculations more rigorously, because
thcn we can justify the additional engIneering effort
TIle use of shortcut solutions and tbe hierarchical decision procedure also
makes it possible to provide mOTe .-apid feedback to the chemist who is allemptmg
to de\'elop a rrootSs. That is. alternate chemical routes could be used to make the
same product, Yo Ith a laTgc number of flowsheet alternati\'es fOT each .-outc. l-kntt.
quick estimates of the range of con\ersions. molar ratios of reactants. etc.., that are
dose to the economic optimum for the various routes help the chemist to take data
m the Tangc WhCTC thc most pTofitable opcTation might be obtained and to
tcnmnate expcTiments thilt ilTC outside the range of profitable opcTation.
Decomposition Procedures for Existing Processes
Of course, we can also use the approach presented above as a decomposition
procrdure for existmg processes, to simplify the understanding of the process. to
understand the dttisions made to develop the process, or to systematically develop
II list of prooess alternatl\es. The decomposition proocdure we suggest is as follows'
I. Remove all the heat exchangers. drums, and storage 'csscls.
2. Group all the dlSlillation columns (liqUid separation system block).
J. Simplify the general structure of the separation system (similar to Fig.. 1.2-5).
4. Lump (group all units In a smgle box) the complete separation system (similar
to Fig. 1.2-6)
5. Lump the completc process
This dttomposition procedure is different from those that break down Ihe
f10wsheet into discrete subsystems which always retain their identity, i.e" into
individual unit operations. To develop process alternatives, we want to modify the
subsystems. With our approach we accomplish this lask within a framework where
we aJways consider the total plant, altbough the amount of detail included at
various levels changes.
Hierarchical Planning
Our strategy of sucassi,e refinemcnts and our bierarchical design procedure are
similar to the hlerarchlC3l planmng stTategy discussed in the artificial intelligence
(AI) literature. Sacerdoti- states,
The essence of Ihis approach IS to utilize a means for discnminating between
imponanl Information and details In problem space By planning in a hierarchy of
abstraction spaces in which suca:ss.ive levels of detail are introduced. Slgnlficanl
increases in problem-solving power have been achieved
E. D. Sau,doll. -Ptanmn,ln a Hierarchy or Ab!;lradKm Splloes.- AIIr/t..,d. 5 lIS (197").
18 SECTION U SUI.HUn ''"0 XElClSfS
0 c-;::::L
"
'"

~

u
'"
-,
I ~
8

~ lIuml0;) K I ~ ~ )-
E

~
1);-1 11
w

'"
:l:
W
"
U.
U
~ uwnlOO pnpoJd)- r::i
w
~ 1
'"
w
..;
u
l

-0
1l

w
'"
'" W
W
~
uwnlOO B30 )- 'f
y
~
..

I
u.
J ~ r u ~
u
u
"'-
<
W ;; ~
Q ~
<
Ol
H,O

j
<
H,
Cool water
~ I ---.J
.. I ~
1.3 SUMMARV AND EXERCISFS
Summary
Process design problems are underdefined, and only about I ~ ~ of the K1eas for new
designs ever become commercialized Hence., an efficient strategy for developing a
design is initially to consider only rough, screening-lype calculations; l-e., .....e
eliminate poor proFls and poor prOOC$S alternatives witb a mmimum of effort.
The amcepl CIIn be readily eltended 10 a hierarchy of space:li. each deahns wllh
fewer details than the ground space below It and with more delalls than lbe:
abslr1l.CtlOn space above II. By wrwdenns details only when a sl.lCttSSful plan in a
bJ&ber k"eIlipace Jlves Iiuona eVldeQCC of Ihen unportanoc,. beumuc scarcll process
..'I1J mvestlgale a peally reduced ponlOJi of lbe search spice.
In our hierarchy, the ground state represents the energy-mtegrated flowshcet,
and each le\-el abo\-e It contains fc.... er delaib. Moreo\er, if the process appears to
be unprofitable as .... e proceed through the levels in Table 12-1, we look for a
profitable alternative or .... e teminate the project before .... e proceed to the next
level. As noted by Sacerdoti, the hierarchy provides an efficienl approach for
developing a design.
Wlller __
J
FIGURE 1.]..1
IPA planL (A/ler 1947 AIC" $'...., CDftlut Probk... )
19
20 SintoN I 1 nnosF.S
Thcn If the result .. Oflhls preliminary analysIs seem luomislOg. we add del all 10 Ihe
calculations and we use more ngorous compulallonal procedures.
We can simplify the deSIgn problem by breallng it down into a hierarchy of
decisions, as In Table 1.2-1 In this text .... e discuss this hierarchy of decisions In
detail
Exercise<i
Recommended eu,cises are preceded by an aSlensk '
1.3-1. If enginl1ng l'01e C0515 SIOO/hr. cstnnale the worker-hour'i requIred to complete
each type of dCSlgn 5tud) In Table I 1-1 for a smalt plant
1.3-2. Aa:ordmg to the engmeerln! melhod. what would be the best Will' to read a
telilbook thilt OO\tn a field you not sluched before. bloteo::hllOlo&)',
elccuoc:hemlStry, etc.}'!
1.3-3. If the diphenyl In the hydrodealkylahon or loluene (BOA) proca.s is recycled to
ntlncllon, ln51ead of being show one alt"'mallve for the hierarchy of
nowsheets, 'c:.. Input-outpul, recycle, scparauon sySlem, dlstillallon train (do not
consider enerl'S integration).
1.J-.4. A Oowsheel for a proa::ss 10 produce acetone from ISOpropanol is Jlve.n In FiB- 13-1
The reaCllon IS ISOpropanol_ acetone + Ii" and an IIZCOUOplC m.i:l.lure. of
IPA-H10 IS used as the feed stream_ The reactIOn takes place at 'aim and Sn'F
Show the hierarchy of no'"'sheets,
1.3-5. An encr&),-mlegrated IIowshec:l for the production or eth)Ibe:ru:e-nc: IS JI'en In Fig
I \-1 The pH",ary reaCllons ue
Ethylene: + Benzene -, Ethylbenl.cne
Ethylene ... Ethylbenunc Dtethylbenunc
Elhyknc: + D?ethylbcnnne
2Elbylbe:nzenc:;:::. Benzene of
The reaCllon IS ,un ..,th iln ucess of benzene and a1lnO$t complete COn"enlon
of the ethylene, to try to mmlmlze the formation of dt- ilnd triethylbenune. and it
takes place It 300 psig ilnd 82O"F over a catalysL Two re:adOI$ are required (one on
sHearn and the other being regenerated because or ooke formation). There 15
or elhanc: In tbe elhylcnc feed and O.28X water In the bc:nzcne rc:cd.. Develop the
hlCf1lrchy or nowsbccts for Ihis process.
1.3-4. A nowshcct for eth.a.nol synthesis is shown In fiB- I)). The primary reactiOns af':
Elhylene of
1 Eth('r + HlO
The reaClJon takes platt al S60 K and 69 ban., and about 7" conversion or lhe
elh)lcnc 15 obt:uned. The equdibnum constant fo, dicthyl ether produclJon ilt thdc:
oonchtions is aboul K .02. The: feed streams are pure Willer ilnd iln ethylene stream
contamrng ethylene, g" ethane, and 1'7. melhane. Show the hierilrchy or
f1owshects.
Reaction section
HealC'r Reactor &paralor __.., Scrubber Vent
Condensate
Water
Benzoic add
Benzoic acid
rectification
ToluC'De
stripping
Reactor
<
e
"
8
DEE

Feed elhylenC'
FlCUlIt.: I.J..J
Eth..1101 synlliats
FlCURE 1.J.4
Beuooc llad JlfodOChOll. [Afi... lIyJ'ocft Ptoc_ 41(11) 156 (N... IllIU) J
22 Sl:CTlOH U .....UY Ai'<O UOClSU
1.3-7. A nowshcct for bcnl:CIIC aod prodocllon '" liho..... n In h&- 1}-4 (from S IA
VISCOSA P'IJ< . 48(11); 156 (No..... 1964 The prlm.. r)
rea"uon I)
Toluene -t 150: lkJUotCAad + lI
J
O
lIo.. e:\e:r. re\erslble by-pn)(hu:b (beno.1dchydc and beruyl...: .100001) as .. .:II .b
heauer ones (as.sumc phenyl benzoale and benzyl ate also formed al the
reaL1l0n condillons of 16O'"C and 10 aIm Pure wluc:ne and au are used as Ihe ra....
matenalli. and Ihe toluene COD\erSlOn 110 1.epl at 30 10 As shown on Ihe
I\owshc:el. lhe IOluene IS rccovered and recycled in one column. and the re\C:r)Ible by-
products are recyclc:d from Ihe ovcrhead of a second The produclls rcco\'ercd as a
vapor sidestream (""1Ih greater Ihan 99 Xpumy), and the heavy componenrs :.Ire senl
10 fud. Show Ihe hierarchy of 1\0wsheCI5.
1.3-8. Select a I\owshcc:t from HydroctJrf)(JI1 P,ocusing (liOC the Novcmber issue: of any
year). Develop the hiel1Hchy of I\owshccts for Ihe proces.s
CHAPTER
2
ENGINEERING
ECONOMICS
In Chap I Ilo'e descnbed a SYSlemallc approach thai can be used 10 de\clop a
conceptual design In addilion. we hsted the Iypc::s of design esllmatcs lhal nonnal!)'
are undertalen o\er the hfe of a proJCCL The goal of lhese esumates is 10 generate
cost data, although Ibe alXuracy of tbe calculation procedures and the amount of
detail consKkred arc different for each Iypc of csumalt.
Since. COSI estimates arc the dn'lng force for any design studl. Iloe need to
uDdc:rsland the \'anous faclon to Include. We descflbe a procedure for general 109 a
cost estllnate for a concc:plUal deSIgn 10 IhlS chapler We begm by presentlng the
results from a published case Stud}. In order to gam an o"crall pcrspc:cll\'e on the
t)pes of eOSI dala required, and Ihen we discuss Ihe details of Ihe COSI analysis.
Remember Ihat the cost models Ihal we develop should be used OIlly for
screeDlng process alternalives. The cost estimates that arc reported to management
should be prepared by the appropriate economic specialists in the company,
because they will mclude contIngency facton based on expenence and WIll include
the costs of more ilems than .. e consider. Thus, our oost eStimates Donnally will be
100 optimiStic, and they should be kept confidential untillhey have been venfied
2.1 COST I FORMAno REQUIRED
By considering the results of a published case study. Iloe caD I!'et an overview of the
kind of information thai we nttd to de\'dop a cost eSllmate for a design.
Moreover, the framewor\.. rdatmg the material and energy balances, eqUipment
sizes and ulility flows. capItal and operating costs, and procc:ss profilability should
become more apparen!. Tbc: parllcular case sludy we consider 1O.,.01ve5 the
production of cyclohexane by Ihe hydrogenation of benzene"
Iknu:ne + 3H:;:: Cyclohexane (2.1-1)
J R. Fall, Mom.flX,wn. WashlOlloD UDlYC....I)' Dcslp Ca", Stud)' No 4, by B
D Smllb, WllJlua,lan IJm...u'ly. St I OUlS, Ma. Aug. I. 1%7
23
u.."
h_
-
...... a.lt "-,
..... "'pi
,.\ ReooCloc (rool.nt) 10 '.000
Iblb ""b
. \ Waste-beoIl boiler
'''"
0.500

....
C, Feed ODlIIp<UIOr ]16 _000
C, Recydt o;ompreuor ]\ 16.000
p.\
BeIlZlODt ICed pump l.1
p.,
Boiler feed pump II'
".000
p.]
Reaclor reflux pump
"'
l.000
Fillet pump
uthllD8 12 b./daJ'
,
n.ooo
Tor.1 ]]() 2,119,(0)
..... "'pi
E\ Coolef-rondtnscr
'"
128.000
E'
inlt.ooolef
\' '.lOO
E] Comp<cno. allt.ooolcr
"
'.lOO
Tow 1.1JX(1
Operaling Cosls
Once .....e know the slream flow rates and the stream temperatures, we can calculate
the utility flows for the various units sbown on the f1owsheet; see Table 2. I-I. Then
if we know the unit costs of the utilities, we can calculate tbe total utility costs. We
combine Ihese utilities costs with the raw-materials costs and other operating
expenses 10 obtain a summary of the operating cosls; see Table 2.1-2.
One of Ihe most Important items Ihat we dc\"eJop during a design is a
flowshect (see Fig. 2.1-1). The nowsheel shows the majOr piettS of eqUIpment. and
usually each piece of equipmenl is given a special number or name, as in Fig. 2.1-1.
Normally each sHearn on the nowsheet is also lettered or numbered, and a stream
table Ihat contains these letters or numbers often appears 31 the boll om of the
flowshcet. Tbe stream table contains the flows of each component in every slream
as well as the slream temperatures and pressures. In some cases. enthalpies,.
densities, and other information for each stream are included in the stream table.
TABLE 1.1-1
Utilities summlf)': Bur case
FloMSbeel and Stream Table
OUf purpose here is nOI to dISCUSS the details of the design. but merely 10 see
what Iype of results are generated.
Coobng'''ltf
Steam.. .'iO Ib,
....."
lJeo;l"cny

i
!
'"
u:
<

-
..
0
,
si s:s
E
"

Q

f
'"
";0
8

-

,

"
0
..

u
<:>

iii
"
8
0

'"
"- 0 r-:,..:o
.5

-

C!'!
"
.,;
,
'"
oi

:;:ON

Q

'"
"

i:'

"
0
S'
-
N ,

;:; '>

-
<

'"

u
"
"
8
-

,.0

N N
-
"

1-
u c
"

"

c
"
N
E
.l!

;;

"

....
<
c
E

"


"

<
::!5! ..

.. E

II: .. 0;

:I:
"
'"
..
.
>Oz u
(,)-S1!
-
"-u'>
F... J R- hu. W""II"" U........ly Daop c.... !;{-.!J No . 0 s-to.
.......h.d&''''' U"'nsdy. 5. 1-. Mo, 1967
Il:
TARLE 2.1-2
Oper1lling COSI summary: Cydohtnnt-bue cast
ESTIMATED PRODUCTION COST AT ARNOLD, CONSOLIDATED CHEMICAL CO
C.H
II
OUTPUT 10,000.000 GAL (65.000.000 LR)
PRODUCT DELIVERED AS LIQUID, 99.9+ %
PPR "rAR (8322 HOURS)
TOTAL MFG CAPITAL
_ S510.('J('(I
!oj
TOTAL FIXED &I WORKING CAPITAL. 693.000
UNIT
QUflNTlT"
UNIT I'RI(.'li
COST
COST
I'HIt yrAR
PER YEAR
P['R 100 LII
RAW MATERIALS
-
RI'NZENE
"I
SO,D SI.R93.CXXl
HYDROGEN MeF '.100.000 0,23 207.000
CATALYST Ih 10.Mm
'00
21.600
It M HANDLING
TOTAL R M
CREDITS SPENT CATALYST Ib IO,Roo
'SO
- 5.400
NH RAW MATI:RlAI.S 2.116.200 Sl26
DIRECT EXPENSE
Lobo.
"300
9.600
Paytoll CharllU
'''OIl
Steam (SO PSIG-CREDIT) Mlh 4$.SOO
'SO
-22,1100

kwh 2.119.000 001 21.200
Camp Air
Repalrl@4%MFG CAP
20.""
Waler-Coohllil Mgal 141.000 O.OIS
Water-Process
Waler-BOILER FEEDWATFR Mgal >'000 OJO 1.500
Fuel-Ga.-Oil
Fuel-Coal
SUPPheJ}
10.200
2% MF(i C.... p
Laboratory ,
TOT.... L DE- W.OOO 012
(CO",,"wtd)
SEClION U '05T If:QUllFO 29
Profitability Estimate
We combine the operatlng and capllat costs. along With some other costs. and "'e
use these results to estimate the profitability or the process (see Table 2.1-7). The
return on investment IS used as criterion of profitability in the case study, but a
number of othcr critcrla can be used. These arc in Sec 24
Engine-ering Economics
Now thatll.'c can see what types of costs are mcluded in an economic analYSIS. how
can we genenlle thcse cost data" First wc consider some of the mcthods for
"IE')I RnclOf<'Oohnfcoil 41OR'
r._ J R r." """-"_ Oo$>p c.... Sl.tJ No 4......... bJ
0 SmotIt. \IouJ.,.- U",n1'I'. So. l-. 1010.. 1961
Capital Costs
After .....e have determined the stream no101o's and slream temperatures. we can
calculate the equipment sizes; see Table 2.1-3. Then we can use cost correlations
(which arc dISCussed in Sec. 2.2) 10 estimate the delivered equipment coslS. NCXI we
use installation factors to estimate the installed equipment costs (see Tabk 2.1-4),
We must also estimate the working capital required for the plant (see Table 2.1-5)
Combinmg allthe:se costs. we obtain an estimate of the tolal capital requirements
(see Table 2.1-6).
TABLE l-I.J
Equipment schetluk

'.
rqi<llued

Sin (ncb)
R1
,
Rnao, 45-," d,am " 28 r.
CI
,
r d comp'OIOI
<l(lI) bhp. t .. o-I!agc
C, 1 ileo;)ck o;omplcssor bhp
"
,
fl'
E'
1 InlucooIn' 15511'
EJ
,
Aftrn:ookr usn'
PO
,
&nzcnc ked pump 11 IP"'. 860 n
p.,
,
tk"k. fotd pump lllpm.116ft
p.)
,
RenU1 p"mp 13 gpm. 93 n
P4
,
rille' pump 2S !pm, 62 n
T,
,
Iknttnl' 57,000 gat
T'
,
R.,nU1 dlllm 930 pi
T)
,
.... no: tepllralOl' 11.,a.. dlam. " J n
T4
,
Siumdrum 150 pi
,., ,
P,odlJd storal" IS&.OOO 'II

,
FIller link
300 ..'
"
,
Calal}'st filu:. JS ft'

z
l;;
8
o
o

<
u

,
t:
z
"
;;

,
j
,;;

1--1-+-+-+----11---+-+\--1 0

$
o

18
JO 1.1 con INlO"',ATION
TAIU Ie: 2.1-4
1\111nurllcu.ring Capilllll: 8a
II cosr ISrolMUION lEQUIUIl 31
TABU 1.1...
or capilli I requirtemtenls: Base C8 on construction in 1967
f'''''' J It F W.."'n.'.... L"".nol) Do:o.lp ea", Sh.d) No 4 od,,'" b. B D Smllll.. ........""'.1<1.
l:... 51 Louu.. Mo_ 1967
TABLE 1.1.7
Profitabilily Or manur.clure
o../h-UN Ib_
hat>_
f"c' ... ...
..

.. S U,bIkj
C, 76.000
"
21l.llW
c.2 J."'"
"
.""
E' S,l00
"
20.'"
E2
'-""
..
10.
EJ
'-""
.. 10,000
P-Ia
'-""
"
'.
P-Ib
'-""
"
<000
P2a
')00
"
'."" P.2b
')00 .. ,.."
p.,.
100
"
J.
P-lb 800
"
l.700

'.200
"
'."'" T,
.""
20.000
T2 2700

12,qOO
T]
""
..
2300

600 ..
2.100
T-Sa 10,800 ..
".000
TSb 10,800 ..
".000
H m
"
].600
'--I
"00
.. II,SOO
SUJ.l70
SSIO.300
Uiot SSIOOOO
r.OOII J

F".. , "'''''''"'',.... l .... crll,) DeupI e"...

Sl...t) ,.... 4. Wl'od bj. B 0 s..u" '" tln'II011 U....,-
..,_ 51 Me>... 1"7
I. Manufadlmnl Cal"lal
Equ,pment
kuel'"
COllIpres:...>r1i
E.o.ehan.....
PumJ-
Tallu

TOIaI ptOOCU equlpmenl
Tout manufaclunn, capital buc:d on baod f.aclon
Toul UU1aufactunqaxl
2.. NOnm.iUlWadunn, C.p,..1
libare ul'i'ling apow esumaled .1 InlOnwaetunn, ap"..]
l. TOlal Filled Capll.]
Sum of I .o.,l 2
... Y.'orl<lnS Capol.l
k.....m.len..
Goods in prOCQS
FlRlshed p,oduc, ,n""nl0,)'
S,orc liuppha; and all olher Ilcrm a, gross ulcs
TOlal .. orlan, cap".1
S. To,.1 F,.ed.nd C.polal
.........
10' ,.l/yr
TOlal COSI
S 9.700
".000
10.00J
'.000
32.670
2900
10.370
SlO.ooo
....000
'.600
...
29.000
12.000
107.000
S69J,OOIl
From J R F wuJwact_ Umwrw,y Dc..p Caoc 51""y No 4
Wllal by 8 U S "h, W....J""I'OII UBI_W'y, So lo.iol. 1967
F'..... J R F"". WlldulIl'01l UN k>,l" C.",
So""} No ... al".d by II I) SIIU.b, WhIOIIQII Un,r
OI'J, S11.owo, Mo, 1%7
F" W auo.)'l1Ir Coo- 6 d.1>d ..... lrnl p".J
SAkE Coo- la. re:ourch,
"od _nn'
TABLE 1.1-5
WOrkio& capilal
I. Rlw Qlltcnal full)
e"lI. 24,jOO pl@SLt23
1. Goods ID prOCCSl
UI I1SO pi @ SO.u
J. PrOOUCI lo""nlory(SOY. full)
Cyclohuanc; I.. S.ooo gal @SO.2Jesllmaled
... II SYo lVlUIi lilies
10,000,000(024)(0 OS)
'.600
]],000
120.000
SIS9,ooo
Muw..dunnl capital
Total F&.W capol&!'
Grou uIa per year
MllIluf"crunnll;Oll
Grou profil
SAkE' @
.....
Nel profit
on 10111 F&'W
S Slo.ooo
693,000
2""'-000
2.257.400

14,JOO
128.JOO
";00
M.l00
9Jy'
calculating capllal and operating costs, then we descnbe the techniques for pUlling
capital and olXrating costs on the same basis, next we discuss profitability
measures, and fina!ly we pm>ent a simple model Ihat IS useful for screening process
alternatives when we develop a conceptual design
:u ESTIMATING CAPITAL AND
OPERATING COSTS
In Table 2.1-1 the ulihty loads for Ihe various pieces of equIpment on the f10wsheel
Il.-ere itemized, and in Table 2.1-2 the ullhty costs were calculated. Similarly. in
Table 2.1-3 the eqUIpment SIZes for the f1owshoct were listed, and the COSts were
calc.ulated in 2.1-4..Thus, the first costs we consider are the operating and
capnal costs asSOCIated with the equipment on the fJowshocL
Operating Costs
0lXratmg costs are nonnaJly simple to estimate. Once we know the flows of tbe
rawmaterials streams and the utility flows (fuel. sleam, cooling water. power), we
simply multiply the flow by the dollar value of that stream. In companies that
operate their systems. i.e.. steam and power production, as a separate
compan}'. lhe Ullhtlcs costs factors are simple to obtain. If this is Dot the case,
however, an analysis of the total site is needed to estimate tbe cost of steam at
various pressure levels. For our preliminary designs, we assume that a value is
available.
. Care be laken that the utility "alues are {t;1l'en on a thermodynamically
consistent baSIS; . fuel and electricity should be more expensh'e than high-
pressure steam, whIch should be more expensive than low-pressure steam etc.
in prices do occur at times, so that it might appear that there is a
III burrung feedstocks to make electricity or in using electricity to produce steam.
Howeve.r, based on unusual market situations normaUy pay heavy
economIc penalties after a few years. One way to keep utility costs uniform is to
relate utility prices (electricity, various steam k"els, and cooling-water costs) to
an eqUivalent fuel value; see Appendix E.!.
The costs of chemicals can be obtained from the marketing department in a
company. For academic purposes, current prices for most chemicals can be found
in the A{wluting or many of tbe trade publications. Light gases,
for example. 01, NJ , CO, etc.. arc not listed in the
beause most are sold locally on long.tenn contracts. The current prices available io
trade publications are often different from the price obtaiDcd from the marketing
department because of long-term contract arrangements.
Capilal Cosls
As :-'e mIght expect. there are a ... ancty of Il.ays of estlmating the capital costs or
that range from vcT}' quick calculations with Irmited accuracy to very
detailed calculatIons that arc \lery time-consuming bUI more aecurate. The most
accurate estimate is simply to obtain a quote from a vendor; i.e., a heat-exchanger
manufacturer to sell you a heat exchanger that has a specified perfonnance
and that will be delivered on a certain date for 8 specified price It pays to shop
around because a vendor's quote will depend on how much work is on hand. These
vendor's quotes arc used as the costs of a final design.
For conceptual designs we need a faster and simpler approach (i.e" we do nol
want to try to optimize a process based on vendor's quotes). Thus, "loe normally usc
equipment cost correlations. For ell8mple, the capital cost of a heal exchanger
normally is e.pressed in terms of the heat-c.xchanger area, and it is not neessary to
specif)' the number of tubes, the number of baffles, the baffle spacing. or any of the
details of the design. Similarly, tbe cost of. furnace is given in terms of the beat
duty required, and the cost of a distillation column is specified in tenns of the
column height and diameter. Tbe cost correlations are obtained by correlating a
large number of vendors quotes against the appropriate equipment size variable.
PURCHASED EQUIPMENT COST CORRELATIONS. A quite extensive set of
cost correlations is available in Peters and Timmerhaus. Other correlations of this
type have been published by ChiltoD, Happel and Jordan, and Guthric.
f
The
correlations of Peters and Timmerhaus are amODg the most recent. although an
even more recent update is available in ASPEN. Several correlations for various
pieces of equipment that are taken from Guthrie can be found in Appendix E.2.
Of course, we are most interested in estimating the total processing costs.
Therefore, .....e must be able to predict the installed equipment costs. rather than the
purchased equipment costs. To accomplish this goal, we need to introduce a set of
installation factors.
INSTALLED EQUIPMENT COSTS. One of the earliest approaches for estimating
the installed equipment costs from the purchased equipment costs was proposed by
Lang.' He noted that the total installed equipment costs ....-ere approximately equal
to 4 times the total purchased costs, although different factors could be used for
different kinds of processing plants. Hand' found that more accurate cstimates
could be obtained by using different factors for different kinds of processing
equipment. For example. the purchased costs of distillation columns, pressure
pumps, and instruments should be multiplied by 4; heat exchangers should
be multiplied by 3.5; compressors by 2.5; fired heaters by 2; and miscellaneous
equipment by 2.5. The usc of Hand's factors is illustrated in Table 2.1-4.
M S. Pdcrs and K. 0 TUIlIl1QlwlIJS, ''-t DaIfPl-" ELM_,na/or Cltmticlli bog_s. MeG, ...
Hill Nc.. YOlk. 1961. chaps. 13 loiS.
, C II ("h,lton. c_ Oala Correlalm. Cltnro Q,g_ 56(6)' 97 (Jan 1949), J IbJlpCland 0 G Jo,dan.
CIotom.ctJI f'rfKru u-....s. Oo:c. New York. t97S. chap ,. K M GUlhrx., Capolal COSI
&umalllll-- C'-t. 76(6) I J4 (t969).
I H J Ung. S,mpliflCd Arproach 10 Prchminarr Cosl EslJmalct,- Clatm SS(6)' 112 (1948)
, W E- H.nd. Flo.. Shccllo Cosl EsI;male.- rmol. 37(9), 3JI (195&)
34 SECTION 11 ESTlloCAT1N{; C... I'tT... L "'1'10 OPU...TING COSTS
SECTION U fSTlloC...TlNG C... PIT...L "'1'10 Ort:....UNti COSTS 35
-
-
iiiiiiiiiiiiii __.iiii,;;.iiiiiiiiiiiiiii=;.iiiii;';;
---_ _--_ ..
--_ -_ ,
__111__111
__111_-111
_ III_
IIIIII.lIJi!lilll

_. ...._---_....,
_I __lIl
__ ';11__1111
1111 __1111
UIl_1I11
.11 111I111111
.1111111I.111111I1
--- o
---,..-.... , ...
-"'-' .
- ..,0 ,
<- " '.0 __ " u
__ It. ,u
........ u ,.
.......... u ...
..... u "
..... _. n.' ,
_--.,.... 'I'-' ' ,
........,- .. .,
--
.. ..
-.... -..
....
."
_..,-

.".
._--
..
--
.. ..
--- -
,,,.,
-- -
,.
--....
.., ,.
---....
-

--
-
'.
---
o. ,.
-.. __............. _.
"""'_.-...... ---'-

... ... ..

.
-
..
-
.. ..

.. ..
.. .. ..
.. .. ..
.. .. ..
.. .. ..
0' 0'
.,

.,

...
".
.. .. ..
..


,..
-.
_.
.. ..
--

...
-
,.
'"
u
'.
u u

u u
- - -
-
."
-
exchanger can be read directly from the graph. Then a series of correction factors
can be used 10 accounl for the type of heal exchanger (fixed tubes. floating head.
etc.). the operating pressure of the exchanger, and the materials of construction for
both the wbcs and the shell.
Moreover, once the purchased cost of the e}[changer has been estimated,
there is another sct available which can be used to lind the installed cost.
The installation faclors provide separate accountings for the piping requirt:d,
concrete used for the structural supports, conventional instrumentation and
controllers, installation of the needed auxihary electrical equipment, insulation,
and paint. Simtlarly, factors for the labor costs reqUIred to inslallthe equipment are
listed as well as the indirect costs associated with rreight, insurance, taxes, and
other overhead costs.
The installation factors listed in the correlations are for carbon-steel ex-
changers, and we assume that the installation costs are essentially independent of
the correction factors for pressure, materials of construction, etc. Hence, we can
write the expressions
FIGURE 2.2-1
Sbcll.nd-lUbe beal nchaog<:rs. (F,om K. M COli ulimDllng, Chem. Eng., p_ JU,
Mar. 14, /969.)
GUTHRIE'S CORRELATIONS. An alternate approach was developed by
Guthrie," who published a set of cost correlations which included mformation both
on the purchased cost and on the installed cosl of various pieces of process
equipment Guthrie's correlation for shell-and-tube heat exchangers is shown in
Fig. 22-1 We sec that the information for the purchased cost for a carbon-steel
(2.2--4)
(2.2-3)
(2.2-1)
Purchased Cost = (Base CostXF.Xlndex)
Installed Cost = (Base CostXlndexXIF + f. - I)
where IF is the installation factor and Index is lhe correction factor for inflation
Hence,
where F. corresponds to the correction factors for materials. pressure, elc. and
Installed Cost = Installed Cost ofCarbon-Stecl Equipment
+ Incremental Cost for Materials, Pressure, etc.
= (IF)(Base CostXlndc}[) + (F, - IXBase CostXlndex) (2.2-2)
ruE ASPEN CORRELATIONS. Another new set of cost correlations has been
developed by Project ASPEN, using data supplied by PDQS, Inc. These
correlations are part of a large. computer-aided design program. and therefore the
correlations are all in numerical ronn, rather than lhe graphs used in most other
sources. For example, the expression they use for heat exchangers is
Gutl:Jrie's correlations provide much more information than most other cost
correlations. although lhey arc as simple to use as other procedures. Moreover, if
we should wanl a breakdown of the tOlal COSI for piping, or instrumentation, for all
the process units, we could develop this infonnalion on a consistent basis. Some
additional examples of Guthrie's correlations are given in Appendix E.2.
-----,......... _..... _- .. ..
---
-- _ , <>.'CO _ '" a.':U ...,
.-.............
o.
- -
--
............ '-n
...... _- ....
..........-.
""- '.'
.... '" _--
.. .It '"""-
_ '" u _
.. .... .......
."
_e-. _ ,-'_
--
"'-'... "
...... R
._ .
u_ ."
..... _--- ...
-......_,.
0-._ "'<:iI <:iI <:iI W<:iI-...lI Q/ ""
..... a_ ,., 11
.... .. '" ,.. '" '" ,.. ..
... "' u. ,.. ,It ,. , ......
.... u. u , "'''-'l
...... - ,.. '." , p, U> Ul " ..
_. _ ... '-If ." " u. Ul " ......
K M Glllhne, -Capll:;lJ COSI E!;IHllillrng." C/o,-m 76(6). tJ4 (1969).
L B Evallli. ASPEN ProjeCt. Depanmcnl of Chcrrucal Englnccnng &. Encrr;y Labonlory, MIT,
Cambooge, Mus.
.... here C = 1979 ex&hanger cost; C. - base cost for a carbon-steel. floaung-head
eJ;chang...r with a tOO-psig dellgn pressure and 150 and 12.(lOOft
l
of
surface: FIJ - a dcslgn.typ..- correction: FMe - matenalsof-constructlon cor-
rection factor: and F,. = a pressure correction factor. The expression they U5C for
t he base cost IS
SEC"llOI'I lJ 101...L OPl14L INYSTMEPl1 AND 101...L "oouct COSTS 37
separate factors for labor and materials, .... hteh often expenence dlfferentlOflatlon-
ary forces, Guthrie's correlations have the advantage that it l5 possible to update
the material and labor factors at dIfferent rates, or som... kind of uerage faclor can
be u'\Cd to account for inflatIon
In C. = 8.202 + 0.01506 In A + 006811(ln A)l
(2.2-5)
EquatIons for the correction factors arc available as well as the cost ...... pressions fOI
a \-anety of other pieces of equIpment. SImIlarly, the installatIon factors arc gj\'en In
the form of equatIons
Updating Cost Correlations
Chilton's correlations were published in 1949. Guthrie's were published in 1968.
and the Peters. Timmerhaus. and ASPEN correlations are more recent. it
takes about three )'ears to build a chemical planl,and so ....e must be able to predict
future costs. aearly the cost of almost everything increases with lime. and 50 .....e
must be able to update the cost correlations.. Several methods can be used for thIS
purpose., but they are all SImilar in that they involve multiplying the base cost in a
ccrtain year by the ratio ofa cost index for some other year to the cosl index for Ihe
base year.
One of lhe most poplliar cost indlC:C:$ of this Iype is published by Marshall
and (M&S) and is updated monthly in Engilluring. A plol of the
M&S IS shown 111 FIg 22-2 SImilar relationships are Ihe
Rt'cora Indell. the Nelson refiner)' I1ldell. Ihe Cht'mtCol IIglnurlllg plant constrlle-
tion indell. and the materialsand-Iabor cost indell Some of these I1ldu:c:s include
8OOr------------==----,
600
200

Years
FIGURE 1.2-2
Mti lIodel.
1...;'-fIOUSF COST CORREI ATlO....S. Man) companies ha\'e developed their own
cost correlations and installatIon factors, These arc frequently updated hy uSlOg
\endor's quotations and recent construction costs. These company cost correla
tlons should alr"up be used If they arc available We UM: Guthr)c's corrclatiom
because they are avaIlable in the published hterature.
2.3 TOTAL CAPITAL INVESTMENT AND
TOTAL PRODUCT COSTS
There are numerous costs required 10 build and operale a chemical planl other
than the op..-rating costs and the installed equipment costs; sec Tables 2.1-2 and
2.1-6. Some of these costs add to Ihe capital investment, whereas others arc
operating ellpenses. Fortunately, most of theM: costs can be rdated direcily to the
instalJed equipment costs throllgh the use of variolls factors. Aver) conciM:
summary of these costs was prepared by Peters by Timmerhaus" and a modifted
\erslon of their hst for the total capital 11l\'estment is shown 10 Table 23-1 The
corresponding brea.kdo...n for the tOlal product costs IS gi\en In Table 2.3-2.
It is common practlcc in the de\'elopmenl of a design firsl to calculate the
sIZes of all the equipment and to eSlmlate the amOllnts of ulilmes reqllired, ellt.
the equipment costs arc determined. and the utility costs are calculated Then the
other COSI factors are added. and finally a profitability analym is llndertaken
Ho.....ever. for preliminary process design. we prefer to look for processs alternatives
as soon as a design appears to be unprofitable. Therefore., .....e would like to develop
simplified cost models for total invcstment. total processing costs. and process
profitability. We develop a simple model of this type as we discuss the mdlvidual
cost items.
Tot.1 C.pital IOl'e:stmenl
According to Table 2.3-1. the Iota) capital investment (Tot_ Inv.) isthc sum of the
fi\ed capital investment (Filled Cap.) and the working capital (Work.
TOl- Inv - Filled Cap + Work. Cap.
M S Pdcn.nd K 0 '1 'mJDI'rhaus. rlDftJ INJtgtl ..J u_.uft/ll' C"'-_ Eit9UOfff'. 3d cd.
McGra..--HllI. I'lcw Yorl, t%9. chap
38 SECTION 1) TOTAL CAI'ITAL IIft'UTMEl'lT ...,.,0 TOTAL 'lOOUCT COST'S
TOTAL ('A!'ITAL ANI> TOTAL ,lOOOCT COSTS 39
TA8LE 2.J-I
Breakd",,-n of lolal capilal iMeslmtnl ud slart-up cosu
I TOlol C<rfl,ta/ ",",SI_n' equab Ihe sum or lbc. fixed capital plus !he wotllnl c:ap"al
II FlAd copll,,1 ",.,..SI,....n, (FCI) U lbe. COSli reqUired 10 build lbe: plOCCSli, equal to lhe s.uffi ollhe
doree' cosls and lhe Indlrecl cos...
A. D"UI cons equal the. iUm or II"" nlatenal Ind labor COSlli requored to build the complcle.
faclhly, about 10 ISY. or FCI
I tftuUt rosu or ISSL (IIrJJM of /.""u) .rc tbc: C05I.5 or IMla1linilM eqwpmenl
sboWD on lbe. process 1\owihc.c.1 '0 a spea&: FOPaphicalloo::allOD (Ibe. baUery bm1ls),
.boul SO 01 FCI_
" I'IuclvuIti t9"IIJ'"D"I!>duda.aII equ'pmc.ul bIilc.d on complete: span:
pans and nofllnsl.a.lkd spafQ;'wptur; IlQWprMClI, wppba, and eqwp-
menl .11o noes; Inft.lIon COIl- .aIlo..."na, fn,gbl c:ba<JQ, ta.r.c.s. UUUl&lM:lCo.and
dUlICS, .Uo 1lOt' for modlficallon dunn, il.arl-UP; .bout of FCL
II f'urc/oosld",q,,;pmr'" lfIJ'O/lOtlOn mcluda IlUill.llation or &II equipment IlSled on I
complele l\0\/4hec:1 includllliliructural supports, insulation, and painl boUI
7.)-26". or Fel 01 lS4Sy' or purchased eqwpmcnl c.oJl.
C I",-,,-w,,,o/lon WId c.."o/ "duda putdtul', IlliI.allaUOn..and callbrallOll, .boul
0( Fel or 6 JOy. 01 purchased equlplDClll
J P","'fJ l!>dudes COQ 01. plpe.. plpc.lwlJeR. fitUDp. val'C$, IOSWaUOn. aDd eqUlpmc.ol,
:about) 01 FCI 01" 10-10%01 purchased IlQwpmctll CXl5l.
r Ckc"fC'OI ""I"'pmDIl 0Ifd rrtOlt'taols inctu 1M purdI.asc aDd iDslaUallOn olille
rc.qu"cd clc.o::ln;;a1 equlpll>c.nl inctudtnll.-lIcbe:s.. moiOrs. condwl, WIre.. tittulp,
kcders, poundina. Illltrumenl and conl,oI wirin.. "J.hllnll paDC!J. .00 tiiOCialed
libor costs; aboul 2 SilO" 01 FCI o. 1-20% ol pu.chased eqUJpmCllI COSI.
2. OJfSllt COSIJ Of OSBL COJU (0,.", of btl"",,) /,mllS) U>elude C&.ril diroctly .c.laled 10
Ihe. prOCC5li bUl bu,lt ,n l"cOlloM from lhc.m.lll p'OCCSl.,nl equ'rmc.nl
" B"t/d,,.,. (",,:IOOmg iCr'ICCS) .aboul 6 or Fa O' 10 10" or pu,,:b.ascd equ,p.
menl COSI
(I) ""'u bo<,WM.fIS 'nclude supc.ntrucI ..rc.s.
aa:ai ..-ars. cranes. moaorai/$, boasu. dc.>"lI1Gn. (Soaoc OOIl'IpaNCI uodudc
lhex 1a<:IOrt &Ii pan ollM ISBL -u. a.od DOl 1M OS8L"",IS)
(2) A...."I_}- lIldudc adllU-llOtrallOll aDd olficc, mcd>caI at dlSprlli:ll.l)',
eafcl",n.. pOll. product .... "'hOU:SC. pam ... lind s.afcly, fire
51atlOn, chanJC house, pcnonnc.l buildJog, shipping ollioc and p1alform,
rc.sc.areh laboralory, conu-oll.botory
(3) Moinr",rwnu shops include e.le.etne.al p;plDl. ShOCl mel.a.l. m.chine. we.lding.
carprnlJ'Y. IMlrumcnll,
(4) B"t/dong U'nIOCtS tnelude p1unlblna. bc.auna. "c.llldauoo.. dusl ooIlec1lOrt. a"
COOOluonUlg, build,n, bpunl. c.I..... lOn. c.ocaI.alOrs. Ickpbonc.s.. UlICfCOIllJllUru-
caloo-, system. pa.lluna. ipl1I1Uc.r .)'IJe....... liIc: alarm.
II Y",d....,.,---..u sUe docvdopawlll inctuduol Jlle deanna. p"lIdiaJ. roads.
..."ILwap. raiuoads. rc._ pallina areas. .. pen. rtcrUltwul fac;d,ue.s.
la.ndscaptlll; lboul ISS 0" ol Fct
C s,f1J'u IlXi/JlIIS (jllSl.Ilc.d), aboul g 0 ol FCI
(1) UI,/IIII'S Include. siram. waler, powe'. refnrcrallOn. coroprc.s.scd .... fuel, watt",
dlliPOUI.
(2) 1",I",,,,s 10clude. boiler plant. JrK:,nc.talOr, "'elb. liver UluLe, ",aler lTe.allnen!,
c\Klhnglowcn. ,,"'ale' jto.ai:e. rlc.o::lT1C subnallon. p1lnl, a" p1anl.
fuel _IOrage.. ...,,>le d,.poul pbnt. lire p'<>IcetlOU
(l) NOIII"IXIU cornpo$Cd of oltic:c eqwpcnt::ol, ukly .nd
medlCll CQUlpn!llll. shop CQwpmcnl, .ulomOfl'-c. c.qwpmc.nl, yard mllC.naJ-
ha",lhnl c.qulpmc.nl, 1.00ralory eqUipment, bllli. paJkli, hand lrucks.
fire bOK$, fire. c.nI'IU:S, Io"dlng c.qlllpme.lll.
(4) D,wlb"rron wuJ p"d.Gg,ng !Ddude I'lIw-malmal.nd prodtH;! slorlF and
handllnl cqulprnc.nt, plOdUCI pack.llnl equlpmenl. blcndllli r.cihlia;. loadlnl
it.UOI\S..
d '-oJ.bout I
(I) SuO'c)'i.nd recs
(2) PrGpen)
8_ IrulirtCl caw lire. uprMd 110I dJr-ealy lo"'ohul "'IlII matmlll and la.bor 01

I LtgllWlr"'fJ tJNI .bout Y. or Fel or S-ISy' or duM ....1$-
11_ !Aimu,"'fJ COSTS Indude process d""'ICl and geucral c.nl'nnng,
dmung, CO'it c.n&lnc.c.nng. proocsslng. reproducllon. communlCatlonli.
_Ie. modell.. consulI.nl rtts. lra"e1
II ErogutN''''fJ Sf'IN"u_ Qnd
2 CQIIJ'fJIC/I"" U/W'UU; IIboUI 4 8 21.0Y. 01. FCI
/J TtfPIIIO'/J'YIJllln composed 01 COIlSINCIIOll. O-pcra11Oll., .nd IZAUlleMIlOt' 01
IaDpora.ry fllQbues; ofIiot::s, roads. parkllliiou. raliroacb. clec1nca1. p1p1nJ.
commUDOClUOns. kDan...
II Coru'TVCI_ ,oob <r..J
c C""Jlrvcl_ S"JWnu_ aCDDUnl1nJ, Illnd':ttplnJ. upe.o:htm,.
d Wo,rhOOlS<' JW'ltmnrllltld 1J,,'mJs
I Solnl'. ,..,d..-ol. QndIrmgt /H"tfi"
I I't,mm./itld ItJlS. sp"'u/ l'rtflHS
, T".us. UlS"'lltIrt, tmd mlt',tSI
1 C_lttK'ors (tt_ abou' 1 S SO. 01 Fel
" C..t"'9f"'C') 10 for unprc.o:hetablc "enlS 'udl n "arms. ftoods.
pna: dt3nJCI. imall MIlD clgnp:s. ('fron ,n nlltrullCl., rIC boul S Xl. 01 FCI
C AI't't_lt btnJ.:d/w" of FCI
I M-../oc-rllrlTOlJ c"p"oIInl'l'SI......nl-umc U 01Ul1a.
2. H_focIllTlTI(J CtlpluH /TIfIISI......,,1 Ii olfl,nc. plus mdlfCCI >i1.5.
III W....klng coptlat is tM c:apll.al reqUITed 10 aetu.ally oprrale lhe. p1anl, aboUI 0I1be. 101.1
cap"al ,nvc.slmc.nl
A.. Ra...' """",.wl for. on",-momh supply_ (11le supply dcprnds on .vallabo"ly, .KISOII.I
deman4s. cle.)
B. F",1SIwJ III ,'oci. aad wm,filllShod producu, appco..,ma,e production C05I5for oM
monlh (Again. lbe amoUOllZAY vary.)
C Arcoo-u ClCtlC'Obll 10 I"e OISIomrn lO d.ays 10 pay lor coodt. aboul the productlOli
eoslS I'or one montb
o ClUit "'" IoDruJ 10 atttt oprraunl UprlliC5-salarlQ and waF' ra...IMIc.naJ purrnasa.
E. AcC'OMIIu poy<Jbk -" ,urs ""yobl",
IV $,,,,r"P CIUIS; :about 01 Fel
A. PHKtU wKJdljico,w#u needed 10 meel dc.s,&n speclficallolls
B Stan"p labo, more people a.e. nded 10 sra'l up plalll Ihan 1(1 11 rumllnl
C J.,ou In prodwc,,<)tI 111'01'0 lois of re'"(nuo durllli debull&'lIl (f( lhe pr00tli5
... Ir_ M S ""'n...... I.. D l_'ba.... ,._ 0...., -.tIu_",,/- .. Mr:G....-lhll
V",l,I'"
40 SOClION U TOlAl CArrrAI INVE5IMENT MoD TorAl COSH
SECTION l-J rOHI CAMIAl 1'I\E5rMENT MoO rOTAl CQ5ls 41
Start-up Costs
TA8U- 1J..2
Gross earnings lind lolal prodlK"t CCKts
T.k &_ M s. and K. D T......ba.... ,.... Duip...J aw-. ea._... Mc<lr..... H...
N York.. 1963
Many companies also include the start-up costs as part of the capital investmenl
Other companies conSider the fraction of the start-up costs that is allocated to
equipment modifications as part ofttle capital investment, whereas the funds used
for additional workforce: and malerials needed to start up tbe plant are considered
operating The choice among various possibilities depends on the
tax situalion of the company However. for our purposes we include the start-up
costs (Slart-up) as part of the inveslment. Hena:, Eq. 2.3-1 becomes
(2.3-4)
(2.3-7)
Offsite ... 045
Work. Cap. - O.I5(Tot lov.)
Working Capital
The working capital represents the funds required 10 actually operate the plant. i.e..
to pa) for raw materials., to pay salaries., etc. We allempl to replace the working
capital each month OUI of product revenues, NeH:rtheless, we must have money
available before we c0mmence operalions to rill up the lanks and to meet the Initial
payroll For this reason the working capital is considered 10 be part of the total
In\cstmenl.
A breakdoy,n of the working capllalls gl\en m TaMe 2.3-1, and a reasonahle
first estimate of thIS cost can be taken as a 3-monlh supply of ray, malerials_ or
products We can greatly simphf) the mitial In\estment aoal)"sis. howe\cr. if we
assume thaI y,orkmg capital is related to the in\estmenL For this reason, we leI
Fixed Capital 100-eslmenl
From Table 2.3-1 we set" Ihat the fixed captlal tn\estment is the sum of Ihe dIrect
COSt and the IIIdlrecl casls:
The indirect costs dcscribed m the table often are lumped in two categories:
(I) the owner's costs, whleh include the engineenng, supervision, and construction
expenses; and (2) contingencIes and fees (Contlng,) which account both for tlems
The onsite costs correspond to the installed equipment COS!!i for the items shown
on the process flo""'sheet All these items are built III a specific geographical area.
called the bafUr)' IlmllS, We can eSlimate the onsite costs diTC{;II)' from Guthrie's
correlations.
The olfsite COSIS. or OSBl costs, refer to the steam plant, cooling towers. and
other items listed in Table 2.3-1 that are needed for the operation of the process but
are buill in a different geographical area, II is common to ha\<e central
areas for cooling towers. steam generalion equipment. etc. We note from the table
Ihal the variation in the individual olfsite costs is much larger' than that in the
onsile costs. In facl, the offsite costs may vary from as little as 40 to of the
onsite costs for an elfpansion of an existing facility. up to 200 or 400% of the onsite
costs for the conslrUCIIOn of a grass-roots plant (a brand new facilily startmg from
scratch) or a major plant expansion. This situation is analogous 10 building an
addition to a versus building a new home. In OUf studies, we consider only
plant expansions, and we assume that
Fixed Cap. _ Direcl Cost of Indirect COSI (2,3-5)
The direct COSTS mclude the onslle costs (Onstle) or ISBL costs (inside batter)
limits). and the otrslle costs. or OSBL costs (oumde battery limits):
Oncct Cost = Onstte + Offstle (2,3-6)
(23-2)
(23-3)
I Gr0S5 arnlnp 10t.1 UloCOme tot.1 produo;llon CO$l
II Tot.1 product CO$t _ m.nuf.etunnl COSI 1" p:oc:ral nrc:n5eS.
A M.nllr"etunn, eml _ dlrl produchon costJ -+- hed ch.r,es .... rl.nt ovn-MJld
I I)UllC1 f'lodUClIOl1 costs t.boul otlhe tOlal product emt)
" R.... m.atefll" ,aboul 10 Xl" of 100al prodUCI COSl)
b Ullllt," (about 10 20. of tOlal p.oduct CO$I).
C MI,nlrna..... and rrpant (aboul 1 10. or Fel).
J O""rJlllnr: supplors (aboul 10 20.0( COit lor mlonlenance and rrp,,"rs o. 0.5 1".0(
rell
t Opellllnliabor (Iboul 10 0(10111 product COSl).
f Orna su""rvlSlOfl and dmcailibol (lboUI 10-2S" 0( Iabot)
fI labontory chafFS (Ibout 10 20" 0( oprnlu"Ilabor).
4. PatmtJ and fO)"lltlC$ (about 0 oClot" product C05I).
2. Fiud charr;es (Ibovl 10 20% 0( total product cost).
... l)epnaahOn (aboul of Fa).
II Loc:allun (aboul 1 4 ol Fel).
c InsurJlllCJe (about 04 I. of Fa).
J Rml (aboUI 0( ,.II.., 0( rrnled land .Dd bvrldinp).
r (about 0 7" ollolal captal m>a.UDr-Dt).
} P1.anl (about SO ol!hl: oasI ror opcr1Iunr: labor. and ....onlrna.....
or 5 0(10111 product COlli). costs IDd\OCk rmc:ral pLmt upl-crp IDd o..... Mad. r-)"roll
ovnhad, pKkallna. Wl'YlCa, Aft'ly.nd protectIOn, rewt.UnrD!S, ucrca!lOll. S111'"lar.
laboralorxs. and ...orllt' r.alltx:s.
B Gt'nrral _ c:ostJ -+- dutnbuuon IlId sdlml! COitJ -+- Tt'SeIlch and dt'....I
oproc:nt COSls [ali(! ca1It'd SARE (SIlks, adrtlllmlnltioD, IDd mplXftlJl!l]
I Admllu...rlllve QOIItJ (Iboul 1S% r:I for opaUDI labor, supn-v!i1Oll, IIId mllnlrnana
or 2 ollotal produo:1 eost); lOOI.ldt'll COliIS for ua:uti,"t' w-na., c:kno::al ."I@t'S. kpl (<<II.,
offioe svppbr., and commU1lIC.IuonL
2. DttlflbutlOll.nd wlllnl costs (about 2-10" of total prodtJt:I COSI); irlclucks C:OS15 ror SIIkt
offica, S111t'$ sll1l', .lhIPPtnl. and IdVt'ItBUIJ.
3 Rt'teIrdt and dt'Ydoplllt'nl COSIII (aboul 2-S" of evefl' Wo:s dolLu or about 5% of tocal
product cost)
Tot Inv = Fixed Cap, + Work. Cap. + Startup
From Table 2.3-1. item IV. we see lhal
Start-up ... 0, I(Fixed Cap,)
42 RiCTrON 1.) TOTAL CAPITAL 1i'lVI!STMENT AND TOTAL PlOIJoUCT COST'S
SfCflON II TOTAL CAPITAL Ii'lVESTMEN1 AND TOTAL P.ODUCT cosrs 43
overlooked in the preliminary design and funds 10 pay thc contractor. A con-
tingcncy allowance of at Icast S%should be included, cvcn J1 wc have 6rm quotes
on hand from vendors, because something can always go wrong. For our
preliminary designs, where we consider only the most upcnsive pieces of equip-
ment, we include a contingency factor of 20% Thus, we assume thai
Indirect CasU c Owncr's Costs + Canting.
Owner's Cost ::::: O.OS(Onsite + Offsite)
Cooting. - O.2O(Onsile + OlIsite)
With these approximations we can write
Fixed Cap. = Onsite + Offsite + Owner's Cost + Conting.
= I.25(Onsite + Offsite)
(2.3-8)
(2.3-9)
(2.3-10)
(2.3-11)
Total Produce Case
Table 2.3-2 lists a breakdown of the total product cost. Since the total product COSI
(Tot. Prod. Cost) is the sum of the manufacturing costs (Manu. COSI) and the
general expenses (or SARE), we can wnte
Tol. Prod.. Cost = Manu. Cost + SARE (2.3-15)
The SARE costs often are about 2.5%, of the salc:s revenues for chemICal
intermediates. although they may be higher for finished products sold directly to
consumers:
SARE _ 0.025(Revenue) (2.3-16)
The manufacturing cost is the sum of the direct production cost, the 6).ed
charges. and lhe plant overhead (OVHD):
Manu. Cost _ Direct Prod. Cost + Fixed Charges + Plant OVHD (2.3-17)
We caD eslimate the raw-materials costs and the utilities based on OUt
preliminary design calculations. From the table we see that the maintenan and
repairs and the operating supplies depend on Ihe fixed capilal in"estment, and for
our studlCS we assume that
The direct production costs include the raw materials. the utilities, maimenan
and repairs, operating supplies (Op. Supply), operating labor, direct superVIsion,
laboralory charges, and patents and royalties:
Direct Prod_ Cost = Raw MatI. + Ulil. + Mainl. + Op. Supply
+ Labor + Supervis. + lab. + Royalty (2.3.18)
A Simplified Inveslmenl Model
The factors we have selected to use in our analysis should give a reasonable
estimate of Ihe invcstment for the type of petrochemical processes that we are
considering. However, different assumptions should be made for different pro-
cesses, and the choice of these factors is an area wbcre design experience is needed.
Our goal is to develop a simple method for preliminary process design. so other
factors should be used where they are applicable.
When we combine the expressions above. wc find that
TOL Inv. = FiJed Cost + Work. Cap. + Stan-up
- FiJed Cap. + o.lS(ToL Inv.) + a.t(Fixed Cap.)
so that
Mainl. s.: O.04(Fixed Cap.)
Supply = O_IS(Maint.) = O.l5(O.04)(FiJed Cap.)
(2.3-19)
(2.3-20)
Tot Inv. = 1.3O(FiJed Cap.)
Then. from Eq. 2.311.
0'
Tot. Inv. = 1.3O(1.2S)(Onsite + Offsite)
TOLlnv. = 1.625(Onsite + Offsite)
(2.J-J2)
(2.3-13)
Tbe costs for operating labor, direct supervision. and laboratory charges also can
be combintd into a single factor. We assume that
Labor + Supems. + Lab. = (I + 0_2 + O.lS)(Labor) = l.3S(labor) (2.3-21)
The table indicates that the cost for patents and royalties should be about 3% of
Ihe total product cost:
Hence. once we have cstimated the installed equipment costs, It is a simple mattcr
to estimate the total investment, although it is important 10 remember that
the estimate depends on the assumptions made in Eqs. 2.3-3, 2.34, 2.31, 2.3-9, and
2.3-10.
Fixed Chg. ... Tax + Insur. + Rent + Interest (2.3-24)
When we combine these relationships, we 6nd that
Direct Prod. Cost = Raw MatI. + Util + O.046(Fixed Cap.)
+ I.3S(Labor) + O.03(Tol. Prod. Cost) (2.3-23)
The fixed charges (Fixed Chg.) gn'en In Table 2.3-2 include local taxes,
insurance, rent, and Inlerest:
For the case ofa plant expansion, we substitute Eq. 2.3-7 to obtain
Tot. Inv. = 1.625(Onsite +O.4S(Onsile)] = 2.36(Onsite) (2.3-14)
Royalty = 0.03(1"01. Prod. Cost)
(2.322)
44 SECT10N l.l TOTAL CAnrAL INVfSTWUIT AND TOTAL I'llDOUCT COSTS SECTION U TOTAL CAPITAL INVESTWEm ANP TOTAL PI.OOUCT COSTS 4S
When we combine all the expressions above, we obtain an cxpression for thc
total product cost:
Thc allocation for depreciation- may be calculated in a variety of ways, and so we
discuss depreciation allowances in more detail later. With these approximations,
we find that
Thc intcrest chugcs on borrowed capital dcpend on the company's financmg
policy, and for our prellmmary deigns we assume that internal funds arc used 10
finance the venlure. so
(2.3-32) Labor = IOO,<KlOOperators
SimpU6ed Cost Model for the
Total p,.oduct Cost
Whcn wc combine Eqs. 2.3-30, 2.3-31, and 2.3-32. we obtain
Tot. Prod Cost = 1.03(Raw MatI. + Uti I.)
+ 2.13 )( operators + O.I03(1.8IXOnsite)
+ O.025(Revenuc)
or Tot. Prod. Cost = 1.03(Raw MatI. + Util.) + O.186(Onsite)
+ 2.13 )( lOS Operators + O.025(Rcvenuc) (2.3-33)
Hence, we can usc thc estimates of the raw-matcrials cost, the utilities, tlK: revenues,
and the instaJkd equipment costs from our preliminal')' process design, to calculatc
the lotal product cost.
Profits
Now wc would IIkc 10 e1iminatc labor and fixed capilal from this cxpression.
From our previous analysis wc know that
Fixed Cap ... I.25(Onsilc + Offsitc) = L25(1.45)(Onsitc)
= 1.81(Onsllc)
TIle cost for operating labor primarily depends on the complexity of thc
process, and 11 can be M gucsstimated" from an inspection of the ftowshoet (although
somc cxperience is required to makc rcasonable estimates). An allempt to quantify
the reasoning Involved was published by Wessel, who corrclated operatmg labor
in workel'"-hours per day per processing stcp versus plant capacity. The difficulty
witb this procedurc lies in estimating the number or processing sleps; i.e., a batch
reactor may require a full-lime operator, whereas a continuous rcactor may require
only one-half of an operator's time.
For relatively small processes, such as we consider in this text, between two
and four shift positions (operators) would be required. Labor costs per shift
posilion are about SIOO,<XXI (since we operatc 24 hr/day for 7 days/wk, we need
about 4.S operators per shift position):
(2.3-25)
(2.3-26)
(2.3-28)
(2.3-27)
Rcnt _ 0
Interest = 0
Fixed Chg. = O.03(Fued Cap.)
Tax + Insul'". _ O.03(FiJICd Cap.)
Based on thc values In the tablc wc assume that
Plant OVI-iD = O.6(Labor + Supervis. + Main!.}
". O.6(Labor + O.2(Labor) + O.04(Fixed Cap.
_ O.72(Labor) + O.024(Fixed Cap.) (2.3-29)
According to Table 2.3-2. it is reasonable toassurne that the plant overhead IS
roughly 60% of the cost ror operating labor, direct supervision, and maintenance.
Rererring to Eqs.. 2.3-19 and 2.3-21, we obtain
Tot. Prod. Cost = Manu.. Cost + SARE
- (Direct Pl'"od_ + Fixed Chg. + Plant OVHD)
+ O.025(Revenue)
= (Raw MaU. + VIiI. + O.046(Fixed Cap_) + I.35{Labor)
+ O.03(ToL Prod. Cost + O.03(Fixcd Cap.) + (O.72(Labor)
+ O.024(Fixcd Cap.] + O.025(Rcvenuc)
SImIlarly, we assume that do nol rcnt any racliiticl>
PROFIT BEFORE TAXES. Thc gr-oss profit before taxes is the revenues minus the
total product cost:

Tot. Prod. Cost = 1.03(Raw MatI. + Util.) + 2.13(Labor)
+ O.103(Fixed Cap.) + O.02S(Revenue) (2.3-30)
Profit before Tax - Revenue - Tot. Prod. Cost (2.3-34)
The okpoeciahon allo...na:. Included u a hed <;harp' I. ,be lable, bul man)' companies do noc
aorounl lOr depucialoon in .... wa)'
H E. Wead. -New G.aph Condl.'es OpenIJnl Labo, Dala lOr Chcmoeal J'loc:eaes,- CIwM.
.5f(1) 209 (ltS2).
Sf.C1lON U TOTAL Ai'IO TOTAL PlOOlKT (,:QSTS
PROF"IT ."-ITER TAXES, The deprcciallon allowance is subtracted from the profit
before tales because II represents a cost for replacmg C'qUlpme:nt. For most large
corporatiOns, the IOcome tax rate is 48%. so thai Ihe profit after taxes IS
Profit aflerTaxes _ (I - o48XProfit beforeTues - Dcpre:c.)
= (O.52XProfit before TaxC$ - Depree,) (2.]-37)
- 0.S07(Revenue) - [0.536(Raw MatI. + Utit) + 0,S2(Dcpre:c.)
+ O.0967(Onsite) + 1.108 I( lOS Operators] (2.3-38)
The IS-)"r property deductIon allowable depends 00 the month that Ihe lIern was
placed ID service, and IS 8lven in Table 2.3-4 for equlpmenl placed In servIce .fler
1980 but before March 15,1984. The U.S, go\ernmenl has been changing both the:
Iifelime (from IS 10 18 and now 19 yr) and the allowance for Ihis type of property
e\'ery year, so recent tax mformation mUSl be: consulted
The ACRS method is too complex to usc for screening calculations, thus .....e
usc the simpler expression
(23-36)
" 'n,

2 1-1
) "
-I I> 10
7 10 10
Deprec,"", O.I(Fixed Cap.)"'" 0.1(181)(Onsite)
== 0.181(Onslte)
TABLE 2.J..J
Deprec:ialion allo,,-aoctS
J.-J' p.o....n) 5-yr proPft1y
'I'n.
,
,
'"'
"
:s Il
1
"
,
" )
" "
11
TABU
Depreciation allo,,-aocc for 15-yr prope.ny (start in 1980)
l\1oed1 plal:fC1 ill

,
)

,
" " "
I
" "
10
,

)
,
I
,
10 10
"
II
"
II II
"
II
" "
I'
)
, , , ,
"
10 10 10 10 10 10 10

, , , , , ,
) 7
,
7 7 7 7

6

6 6 6 6 7 7 7 7 7 7 1
DEPRECIATIO . Ifwe consIder bUYing a car or truck to use for business purposes,
it is apparent thai lhe: vehicle will wear OUI over time:. Hence. we should set aside
part of our revenues 10 order to accumulate sufficient funds to replace the vehide:
when it does wear out, aRd we sllould consider thc:sc: funds to be one of the costs of
doing business. We could deposit this replacement allowaoa: in a bank and draw
interest., but we hope: that we could gain an even highe:r effcctive interest rate by
investing the funds in another venture: of our company,
Fortunately, the government recogmzes that it is a legitimate expense to
deduct a fraction of the cost of equipment as it .....ears oul, despite the fact that the
funds arc not actually used for Ihis purpose; i.e. they arc: invested in other ventures,
and a portion of the profits of thCS' other ventures is used to replact the
equipment Thus, to prevent a company from establishing completely arbitrary or
unrealistic depreciation schedules, the government specifies the average lifetime
that can be: expecled for various types of processing equipment.
Of course, if pieces of equipment havmg different lifetimes are combined in a
single process, clearly accounting for depreciation can be:comt: quite complicated
Slnct t he:re are often several processes in an integrated plant complex, however, we
can consider thatlhere IS an average hfellme for the process for preliminary design
calculations. In fact, for petroleum processes we oflen assume a 160yr life, whereas
for chemIcal plants we often take an II-yr life.
Onct the process hfetlme has been fixed, the government sull allows us to
choose between methods of computing the dc:prc:ciation: straight-hne or ACRS
(accelerated cost recovery system), Land does not deprc:ciate 10 value, and therefore
the investment in the land should nOI be: considered in a deprCClauon calculation,
Similarly, if we replace the worklOg capital each month, we .....mhave the same
amount of .....orking capit.al al the: end of the project as we started with, so that
.....orkmg capital docs not depreciate. Furthermore:, the equipment may have some
salvage value at the end of the projeci (ofle:n about 10% of the purchased cost,
whIch corresponds to about] %of the fixed capital investment), so we should
account for this salvage value at the end of the projcct (If salvage value is not
Included In the depreciation calculation and the equipment IS sold at the end of the
project life. then a capital-gains tax mUSI be paid on the value of the equipment
sold.)
Straight-line depreciallon Simply means deducting 33 %per year of the value
of equipment having a 3-yr life, 20% per year (or equipment with a S-yr life, 10%
per year for equipment havlllg a 10-yr life, e:tc_ The ACRS melhod is more complex,
II only started lD 1980, and at that time all equipment had 10 be grouped in one of
four C8legones-3-yr property, S.yr property, 100yr property, and 15-yr propeny.
The depreciation allowanoes for the firsl Ihree categories arc gIVen in Table: 2.33.
or, afte:r c:Iiminating the: total product COSI, we have:
Profit before: Tax .. o974(Re:\'e:nue:) - L03(Raw Mal + Ulil.)
_ O.186(Onsile) - 2.13 x lOs Operalors (2-3-35)
To calculate the profit afler taxes, we must consid.:r various depreciation policies
$FCTION H nMt VALUI': OF MONU 49
Profi.ability AnalJsis
Now that we have calculated the cash ftow, we ha\e the information required to
undertake a profitabIlity analysis. HO\l>e\er. sina: a profitability evaluation in-
volves both capllal and operating costs. first we must find some way of pUlling
both types of cost on the same basis. To do this. we need to consider the timc valuc
of money.
We determllle the lime value of money SImply by assuming that "'e Will
ahuys borrow the capital that ",e need from a bank and. of course, that "'e must
pa)' mterest on the money that we borrow. With this approach. we replace the
amount of capital investment by the annual payments that "''C must make to the
bank to repay the loan and the interest on the loan These annual payments ha"e
the same units as opetating which is what we want to achieve
Thus. the key to understanding the relationship betwen capital and operatIng
costs IS mcrcly to de\clop a del,uled undcrstandlllg of the rcpayment of bank loans.
There are two parts to this repa)menl principal and interest.
Consel'"'"8tion or Money in a Bank Account
Banks lend money at compound mterest.. and the simplest way of understandmg
the changing balana: in an aCCQuntls to assume that money in a bank aa::ount is a
consen'ed quantity. That is, the money derositc:d (input) plus the interest paid by
the banl to the account (input) minus the money withdrawn (output) must be
equal to the amount of money that aa::umulates. Thus. the conservation of moncy
in a bank account can be treated just as the conservation of mass energy, etc. Of
course, this conservation pnnciple is valid only ror bank accounts and not for the
federal go\'ernmcnt. the government can simply print additional money.
However. rttogni7jng this restriction, we can write
(2 3-39)
(2.3-40)
or. after substituting Eqs, 2.3-35 and 2.3-36. we have
CF "" O,507(Re\cnue) - O.536(Raw MatI. + Utll.)
+ O,OO98(Onsile) + 1.108 )( Operators
Cash Flo,",
The actual cash flow (CF) that IS relallled bf the company IS the profit after tues
plus the depreaalion allowancc,
CF _ Profit after Taxes + Deprtt.
"" O.52(Profit before Taxes - Deprtt.) + Depree
,.. 0.52(Re"enue Tot Prod Cost) + OAS(Dcprec.)
2.4 TlME VALUE OF MONEY
Accumulation = Input - Output (2.4-1 )
-0"" if we dIvide by AI and take Ihe limit as iiI approaches zero, we oblain
"'here p. is the principal that we put into the bank initially; ie. at t ""' O. S - P"
Thus. our money grows exponentially.
CO"TI"TOUS 1l''TEREST. Some banks are no", offering continuous compound
ing on money, rather than compounding the interest at discrete intervals.. Since the
coDtinuous compounding case is similar to other conservation problems that
chemical engineers study. we consider it first. We let 51. be the money we have in the
bank at orne t. If we make no deposits or withdrawals., the amount we have In the
bank will increase to 51,H.' over a time interval lit because the bank pays us
interest. U we let the coDtinuous interest rate be i, [5 interest!(5 in aa::ountXyr).
then the amount the bank pays us in the time intervallif is i,SI, lit.
Hencc. the conscnatlon equation, Eq. 2.4.1. becomes
(2.4-2)
(2.4-3)
(24.4)
I
dS S
1m
/:I.t dl <
We calf solve this differential equation to obtain
S =
When we consider proass optimization studies. we often encounter trade-offs
between capital and operatmg costs. Forexample, we can recover more of a soh'cnt
entering a gas absorber by mereasing the number of trays. Operating costs are
measured in SItu (or more commonly in S-'yr). whereas capital costs correspond to
a single expenditure of money (i_C-., an investment). Then, to trade off capital costs
against operallng costs, we must be able to place botb costs on the same baSIS.
Thus, we can either annualize the capital costs or capitalize the operating costs. In
this text we report all costs on an annual basis.
Similar Problems and Strategy
The problem of trading off capital agaIost operating costs is commonly encoun-
tered in e\eryday life. For example, whcn I bought my last car, I wanted to
determine if it was to my advanage to buy a VW Rabbit with a diesel engine for
S6400 as compared to a conventional engine ror S5200, when diesel fuel cost 50.89/
gal as compared to gasoline at SO.94/gal and the diesel engine averages 45 milgal as
compared to a conventional engine that averages 32 mi/ga\. There are different
capital and operating costs ror the two choices, and we wanl to find ",hich is
cheaper. A similar problem occurs if we want to assess the desirability of installing
a solar heater that costs SI5.000 m order to save 55% of an oil bill of SIOOO/yr To
solve problems of this type, we must consider the time value of monel'
so SI!CTlOW 10 TlIotE VAl Ul OF MONEY
SECTlO.... " TI\lE VAlliE Of 51
DISCRETE COMPOUNDING. II is more common for a banl to compound in-
terest at discrete lntervals. Jfwe let SI I - SI.. = Ihe accumulation of money in the
account over one compounding interval, I = S iotercstf($ in aceountXI period)).
the amount of interest in one penod - lSI and If we make no depoSits or
withdrawals. then the conservation equation becomes
compounded quarterly We nOie thai II IS CSS('nllal to mclude the eompoundmg
II1teoal In the dcscnpuon of the II1tercM, because \lo'e expect Ihat the
on an annual basIS will be greater than 6-.. If we let r be the nommal
inlerest r3te and", be the number of corresponding IOlervals pcr year, the
expression \lohlCh IS analogous to Fq 24-6 for paymenb for I yr IS
where, again, S _ .p. when" _ 0, so that p. is the inilial amount we deposited with
lhe bank.
Inslead of using finite-difference equations to find the compound interest. we
can prepare a table showing the amount al the beginning of each compounding
period, the amount of interest paid during that period, and the amount al the end
of each penod; sec Table 241. As this table indicates, II IS a simple mailer to
generallU the results and thus to ohtam Eq 2.4-6.
The parameter n takes on only integral values, and thus we call Eq 2.4-5 a
first-order, linear finite-difference equation
Finlle differences are not as common In chemical engmeering practice: as
ordinary differential equations. However, the equations describing the composi-
tions in a plale gas absorber or distillation column, where the composition changes
from plate to plate instead of continuously, have thlS same form. If we use finite-
difference: calculus to solve Eq. 24-5, we obtain
Comparison bet,,-een Discrete and Continuous
Compounding
As we might expect, the interest rate thai a bank would pay using continuous
compounding is different from that for quarterly compounding. Similarly, the rate
for quarterly compounding is different from the rate for semiannual compounding.
Hence, we need to find a way of comparing these various rates.
Suppose that a bank pays 1.5" interest per quarter and compounds the
interest 4 times a year. In this casc: ...... e say that the nomi"al interest rate is
(2.4-7)
(2 4-8)
(2.4-10)
(2.4-12)

(
')"
lim 1+
._... m
5""
(
,)- ( ')'.'" S-P.I-I
m
,.P,I+
m
we sec that
1+ 'of' = (I + i;;f (2.4-9)
For very frequent compounding periods. i.e_,as m approaches mtinlly for" yr.
Eq 2.4- 7 becomes
When ....e comp:ue thiS resullio Eq 24-fi ror I yr, ",here we call1hc interest rale in
Eq 24-6 the effecme mlereSI rale i.
n

Ho.... e\er, ddinition.
which has the same form as Eq. 2.4-4 ir I _ n yr.
Thus. if we write Eq. 24-8 for n yr and compare it to Eq. 2.4-12. we see that
(I + I<lfr = (24-13)
so that Eq 24-10 becomes
so thai
(24-5)
(2.4-6) S =- P.(l + i"f
Siu I - SI. = IS!.
Aher rearrangmg thiS expression and comparing it to Eq 2.4-4 Wllh r = n. we find
TABLE U-t
Discrete compound interest
;.,,=e'-I (2.4-14)
Eump4e 2.4-1. If Ihe llomlnal annual Inleresl rale IS find lhe value or a .5100
deposit idle. 10 yr \0'1111 (IJ) continuous compoundtnio (b) daily compoundmg. (c)
scmUlnnual compoundmg, and (4) Ihe elfccuvc annWlI mleresl rale for conllnoous
compoundIng.
I'ri.;;pool .1 1.lu_
or tanIH dllflac 0( fund u
P...... poriod ..riol end ot ptriocI
I p. p.' 1', +1' _1',(1 +.}
,
1',(1 +I) 1',(1 +.)0 1',(1 -t.J + 1',(1 + ilo _ P,(l +i)'
,
1',(1 + i)' 1',(1 + i)'. 1',(1 + 'l' + 1',(1 + .J', _1',(1 ... ,)'

I',(t + ir 1',(1+i)"
,,
1',(1 -t i)" + 1',(1 -t i1 '._1',(1 +'1
r=ln(i.(f t I)=i.
Thus. we can find expressions thai relate the various mterest rates.
(2.4-15)
52 s.ECT1Ot':U TINE YALUE Of Io4QNn
nrno.. un.. " YALUJP: .,.- "ONn" 53
In the hmlt as At approaches zero. we find that
Again, we can separate variables and integrate, and if S = 0 at r = O. we find that
(_) s _ p.r _ _ 11&2.21
{
')- 00( 006)""'"
(b) S - PI ... ;;; - I 1 + ;65 - 11&2.20
{
')- 00( 006)"'"
(c)S_P If", I l-t 2 _11&061
.... = r' _ I _,,0010 -I -0061&
S_R.(t"'-I)
'.
""hlch IS the contlOuous analog of Eq 24-18
(2.4-19)
(24-20)
It IS ,nlerCllilmg to note lhal contmuous compoundmg is cs.scnlJally the same as d:uly
compoundmg.
5 _ R(I + ;r-I + R(I + i)"-J + R(I + ,r-
J
+ ... + R(I + i) + R (2.4-16)
We can simplify thiS expression by mulliplying by I + i:
DISCRETE CASE. Suppose .... e make periodic payments of SR. for a total of n
periods and the interest rale for each payment penod is i. It IS common practiCC to
make the first payment at the end of the first penod. so It will accumulate mterest
for n - I periodS: the second payment will accumulate interest for n 2 periods.
de. lienee. the mou'")' acculllulated al the end of the /I periods .....iII be
Annuities
If we retum to our example of whelher to buy. car with a conventional or a diesel
engine, we recogOlu Ihal we have to make monthl"'f payments on the car arKIthat
the bank can reinvest this money every month. Similarly, we usually contmue 10
make deposits in a savings account rather than just make a single deposit. Thus,
.....e need to extend OUt analysis of interest paymems to cover Ihese cases. The
method involved IS essentially Ihe same as buying an annuity from a life insurance
company.
(2.4-21 ) PV = S(I + i)-O
GenNal Approach 10 Interest Problems
With a (J(:lailed background in conservation equatIOns., chemical englOccrs mIght
find It simpler to deri\e interest formulas for olher situations In lerms ofcontinUOUS
compounding by maklOg money balanoc:s based on Eq. 24-1. EJI3cdy the same
approach can be taken for discrete compounding.r we usc: fimte-difference calculus
Numerous tables are available in a varkty of books that give the results for the
dIscrete cases. and it is always possible 10 convert from one procedure 10 the other
by calculating the effective annual interest rale.
SohI,iolf. The 101.11 cosl is lhe lum of lhe do...n p3ymenl and lhe: accumulated value of
lhe: monthly payments given by Eq 2.4-18:
1>1.01 [( 0.10)""" ]
Tot. Cost = S20 +-- 1+- -I _ 1681947
010112 11
Presenl
The Interest formulas that we developed earlier describe the amount of money that
.....111 be in our bank account afler a Specified time interval; i e., they indicate Ihe
future value of money. However, we make decisions about investments today. and
so .....e would prefer to know the prese", oolue of various kinds of investment and
payment plans. In other words, we want to ask, How much principal p. would we
need to im'cst today in order to have a certain amount of moocy S available in the
future? Of course, ....e can answer this question merely by rearranging the equations
ie deri\'ed before. The present value (PV) for diSCTete compounding is SImply
2,...1. A fncnd buys a VW Rabbil for S5200, makes a do...n paymenl or
lOY.. and then pays SIS1.01/mo ror ) yt. U the nomina.! intcrCllit rate is
compounded monthly. What is your frjends total cash outlay for the car'!
(2.4-17)
(2.4-18)
is=R[(I +0"-1]
,( "_+'..:.!iY'_--,'
S=R-
i
5(1 + i) = R(I + i)0 + R(I + iy-' +... + R(I + i)
and then subtracting Eq. 2.4-16 from Eq. 2.4-17 to obtain
0<
This same solution can be obtained usmg finite-difference calculus,
CONllNUOUS CQMPOUNOING. For contmuous compounding. the Input term
in Eq. 2.4-12 includes both the payment At and the: interest rate on the: mOlley
accumulated i.SI. At dunng the: time Interval Ar. so that
"llIk thaI ror conlllluous compoundmg IS
PV _ Se-f.l
The lerms (I -+ I) and e-I,. arc called tM: diMoutttfacf(}IFs
(24-22)
54 SECllot" 10 TIME VAlUE or WONI;'(
EXlimpie 1.4-3. On your elghlecnlh bIrthday your nch UliCIe promISeS to &lve )011
$10.000 on lbe: day you arc 25 If lhe nommal interest rale is S-/" compounded
quarterly. how much money would he need 10 pUI 11110 lhe banI. Oil your elghleenlh
bIrthday for hIm 10 be able 10 keep hii promise?
&/"lio. For Eq 24-21 we lind lhal
PV 10.00:{1 + - S574175
2_5 MEASURES OF PROCESS PROFITABILITY
From the dbcusslon m Sec. 2.3 clearly the cost accountmg associated wuh process
economics can become qulle complica led We anlicipate the same kind of difficully
in allempls 10 assess Ihe process profitabllJly Of course. we prefer to use the
simplest possible prOi.:edun:s for our prelllllill3ry deSign calculations, but .....e would
lile the re:.uhs to be as meanmgful as possible Thu:s, before we select a procedure
for estlmatmg profitability, we need to undcf")tand Ihe advantages and
dlsadvamage:s oflhe \anOliS techniques.
Return on Inreslmenl
In Table 21-7, the: profitablhty measure calculated was the return on investment
(ROI). We calculale thiS value by di\'iding Ihe annual profit by Ihe lotal investmenl
and multiplylOg by 100:
Comparing Capilal and Operaling CoslS
If we want 10 compare an investment / I plus annual payments R
1
to anolher
investmenl /1 with different annual payments R
1
, to sec: which is the smaller, .....e
want to compare It.e prescnt values for each case. The present value is given by
(2.4-23)
(or the discrete compounding or
Annual Profit
-;" ROI - T x 100
ot Inv
(2.5-1)
Hence.. It is belter 10 buy the convenlional engine.
for the continuous casc. The appropnate expressions for other payment periods
can be deri\ed in a Similar way
1.4-4. Suppose thai ...e dove a car 15,000 mi/yr, Ihat ...cc leep I car for 7 yr
before II rusts away and 9I-e Junl II. and lhal ....e pay our g15 monlhly If Inc
nomiul mlercstl1Ue 1$ II '"I.lyr eompouooed monlhly, IS 11 better to buy a VW Rabbll
wlIh a convenllonal enJ.lDe or a dlC:Sel engmc (assume we pay Ibe 101a.1 pUfl;hasc pnc:c
In cash and that .....e usc the oml and mileage conditions given earlier)'
Eslimaling Capillil Costs
Now we can use a present-value calculation to compare alternatives Ihat have
different capila! and opcrattog costs. We can use these results to assess the
profitability of a process.
(2 '-2) P )
Fixed Cap + Start-up
ayout tIme (yr _
Profit after Taxes i Depree.
We can base Ihls return on InvCSlmenl on ellher the profit before lUes or the profit
afler taxcs, so we must be: careful to repon the basiS for lhe calculaLion. Also, illS
ImponanllO remember that the ..... orking capllal, as well as the ponion oflbe staTl-
up COSIS considered as an lmestmenl for lax purposes, should be included in lbe
100ai lD\CSlment.
The return on In\cstmentls a \'ery SJmple measure of the profilability, but It
not consider lhe lime ulue o( money Morcover. it must be based on some
kmd of an average lear's opera lion, SlOtt variable deprcaalion allowances (such as
the ACRS method), Increasmg malnlenance costs oyer the pro)CCt life, changmg
WCS Cle., cannol be accounled for except by averaglDg. Dc.spue these
sboTloommg.s, the return on lfl\estment often tS used (or prelimmary design
cakulallons.
Payout Time
Anolher measure Ihal somellmcs IS used to profilability is the payour lime,
...hlch is the time in years It lakes to recover Ihe funds that we invest (after the
payout period we arc playing poker wilh someone else's money, which is a
desirable situation). We recover the working capllal every month, and therefore we
neglect the working capilalln the calculation However. the fraction of the start-up
costs that is comidered to be an investment should be added to the fixed capital
to find the amount of money tied up to the project. The funds thai we
recover from the proJcct are the profit after taxes plus the deprOClalion allowance,
which we call the cashjlow, so thai the payout lime is
(2.4-24)
R
PV = I + -;-(1 - t'-"")
"
DicscJ:
Convenuona.l:
Solllril1ff. According 10 Eq 2.4-23,
(15,000/12)(1/32)(0.94)[ ( 0.11)
PV_5200+ 0.11/12 1- 1+
12
_ 5200 + 2144.48 = $1344.48
(15,000/12)(1/45)(0.89)[ ( 0 ")
I'V - 6400 + 0.11/12 1- 1 + 12
_ 6400 + 1443.85", $7843.85
56 SECT10N lJ MI'.ASUIlE5 OF PIlOCE$$ PIlOFTl ...IUTY
ThIs criterion also is \'ery simple to calculate, but it suffen from the same
limitations as the ROI. Hence, \loe would like to contrast these simple procedures
with a more rigorous analysis that accounts for the time value of money. In this
way we gam a better understandlOg of the additional complexity required to oblain
a more accurate estlmale
Discounted-Cssh-FIO\4 Rate of Return
Still another \loa)' 10 Judge the desirabIlity of in\'esting 10 a ne\4 process IS 10
estimate the maximum amount of Inlerest lhat we could afford to pay if \loe
borrowed all the in\"cstment and the project would just break e\en. Obviously, if
our analysis mdicates that we could afford to pay 120% interest, we know thai it
would be far better 10 in\'est in Ihis project rather than in a bank. However, if the
interest we could alford to pay was only 2OY.. we should abandon the project. When
we consider IOteresl cakulatlons, \loe recognize that intcrest often is compounded at
discrete intenals. and therefore we IKCd to conSider the lime \"alue of mOne)
Hence. we want to evaluale the revenues, costs. depreciallon, taxes paid, and the
investment on a yearby-year basis.
Normally it takes about 3 yr to build a plant, and for this reason we want our
1D\'cstment costs, raw-malerial and product prices, utilities costs. etc.. to reflect the
values al least 3 yr in the future. rather than at the current time Moreo\er. the
calculation of the process prufitabihty should be based on the income and costs
3)r after the decision has been made 10 starl conslructlon. In other words. zero
lime is considerl as 3 yr beyond the prOject approval
which is Just tilt: principal P, required to accumulate an amount of money S after
N yr. The present value of various investment policies can be esllmated once thc
inlerest rate has been specified by using the relationships we developed earlier,
ALLOCATION OF REVENUES AND COSTS. Most ne\lo plants do not reach
their full producti\e capacity in the first few years of operations. often because a
market docs nOI exist for all the product. Experience mdicatcs that the production
rate increases from about 60 10 90 to 95 %dunng the first. second, and third years,
respecti\'ely, of operation After that time, hopefUlly, lhe process operates at full
capaCIty.
Similarly, the depreciation allowance will vary each year, unless a straight.
line deprecIation schedule is used. Thus, with variable revenues and a yariable
depreciation allowance, the annual profits, the income taxes. and the net profit will
change from year to year. We call the sum of the annual nel profil, which is the
profit after taxes plus lhe depreciation aJJowance, Ihe cGYt flo.... because Ihls
amount of money is actually relained by the company each )'ear or eourse, Ihe
cash flow at the end of the first year, and later years. must be discounted to the
present value, again because we could realize the same amount of money al a later
dale by investing a smaller amount of money at time zero_
DlSCOUl'TED-CASII-FLOW ANALYSIS, With the background given alx)\'e we
can sel up Ihe proa:dure for calculating the discountcd-cash-flow rate of return
IDCFROR) This IS accomplished by equallng Ihe present ulue of Ihe mvestment
to Ihe present value of the cash flows. If we consider a falrl) general ca"e where
then we can develop an ekpression for the equality of the prescnl values of the
expenses and the income.
The direct costs, the Owner's costs, and the contingencies are spent over the
construction period, but the working capital and the start-up costs arc requircd
onl)" at startup Thus. the total value at start-up time is
I. The allocation of the direct costs can be represented by percentages, such as
"I = 0.1, ttl = 0.4. "J - 0.4, and ". = 0.1.
2. The re\enues are constant except for the firsl 3 yr when b
l
- 0.6, b
J
- 0.9, and
b
J
= 0.95.
3. The tOlal product costs (or cash operating expenses) are constant.
... We use straight-line depreciation, so that d
1
- d
l
= d
J
= ... "'"
5. We exclude the deprecialion allowance from the lotal product cost.
ALLOCATION OF CAPITAL INVESTMENT. Since it reqUires aboul 3 yrto build
a complex processmg plant. the direct costs will be spent over this total penod At
the outset, we will have to pay for the land, hire a contractor and construction crew,
order the equipment, prepare the Sile. and starl preparing the foundations for the
equipment. Then we start installmg Ihe equipment as it is delivered. Thus, the
direct COSI expendilure at lime mlOUS 3 )'r IS about 10 to 15 y. of the total. During
the periods of both -2 and -I yr, \\e often spend 40 or 35% of the direct costs
each year, and in the last year we normally spend the remaiOlng 10 to
Howe\'er, the owner's costs, which arc for engineering and supervision, and lhe
contingencies and fees may be spent uniformly throughout the construction period.
The working capital and start-up costs are invested at time tero, but
remember that the working capital is recoyered at the end of the project. Similarly.
Ihe saJyage value of the equipment can be realized al the end of the project. and IhlS
often amounts to 10%of the purchased equipment cost or roughly 3 %of the fi;l;ed
capital investment. Of course, money returned after N yr has a smaller value at
time zero, because we could deposit a smaller sum in a bank at time zero and
recei\'e the compound interest on these funds for N yr. Agam, we see that the time
value of money requires us to account for funds m lerms of their t"Cl/llr.
,
L: {["l(Direct COSI -+ Owner's Cost + Conting.))(1 + III
}-,
+ Work Cap. + Start.up} (2.5-3)
58 SECTION U 01' 'lOCfSll rlOnTUILITY
SECTION H IoIl"-SUlU 01 'lOCUS 'lOH1UlIITY 59
The present value of all the cash ftowsdlscountcd bad: 10 Ihe start-up lime plus the
discounted value of Ihe working capital and the salvage value at the end of the
planl life after N yr IS
For a DCFROR calculation, we look for the iDlercst rate j that makes these two
expressions cqualto each other. Unfortunately, there is no s.Ilnplc way to sum the
series involved, 50 that we must usc a trial-and-crror procedurc to find the interest
rate i.
= [OS21(Re\enue -Tol Prod Cost) + o48Dc:prcc][I -(1,+ i) ....]
-t [Wad, Cap t Salv Val](1 -t II ,.
Worl Cap t Start-up
The
capX( I -t
{(
F"'" C'P) }
---:;." ((J + i) - I] + (Worl. Cap. + Start-up)i (I + I)'
- [0 52(Revenue Tot. Prod Cost) + O.48Deprcc.][(1 +it' - I]
+ (Work. Cap. + Salvo Val.)i (2.5-6)
(2.5-4)
{b.;l:0.52(RC\'enUc
1
- Tot Prod. Cost) + O.4&:q
/-1 (I +i)'
+ (Work... Cap + Salv, Val)}
(I + it'
Of course, we still need to use a trial-anderror procedure to find i.
)
.FO"='OdCC=,<pc+.:..:SO'O'C'O'
(No. Years x Revenue> -
0.52
SIMPLIFIED MODEL As we might expect, the analysis becomcs much Simpler if
the investments, cash flows, and depreciallon allowances arc umfonn Wilh
conSlant cash nows we can use our mleresl and annuity rormulas to sum thc series
APPARENTLY UNPROFITABLE PROCESSES. Of course, there IS no sense In
attempting to solve,the problem by lrial and error ilthe total cash flow over the life
of the project plus the salvage value is Dot adequate to pay for the fixed capital
investment plus the starl-up COStS, In olher words, for the interest rale to be
positive, we require that
In practice... we expccl 10 eOCQuQ!cr this limitation quite frcquently; many idcas for
new processes simply arc not profilabk. and Ihe cffects of Inflalion will makc It
appear that we can never build a plant similar to one thai already CXlsts-even If
the: markct expands.
However, we do nOI want to climmale projects thai may become sound
IDvcstmcnts when product prices ose because of supply.and-dcmand considcr-
ations. Thus, if Eq. 2.5-5 is not satisfied, we ortcn k:t i equal 015 or 0.2 In Eqs. 25-3
and 2.5-4; we substilutc our estimatcs of direct costs, owner's costs, contingencies,
working capilal, stan-up cost, 10lal product cost. and salvage value; and then wc
solve for the revenue we .....ould nted to obtam. Ncxt we estimate the product price
Ihat corresponds to these revcnues and undenake a supply-and-dcmand analysis
to dctcrmine how far in the future we migJIt expcctto oblain that pricc.lfthe Iimc
projection is 20 yr, we might as well put the project in the files for 15 yr or so; but if
the time projection is 5 or 6 yr, we might conlinuc 10 work on tbe dcsign. Again,
judgment is required 10 make Ihis dccision.
2.5-1. Cakulale Ihe OCFROR for the allocation of m\'CSlmem and revenue
pallern gJvcn
Soluriofl
j _ O,IS
; _ 0.2
;_0.1611
Oisc....n. 0;';"_1
fuult fa".or ranoe
YUI-4
,,.""
1749 207" 1.861
hall l50,OOO I S21 1728
""
lUI2 no,o..o J 322
'''''
,-'"
YUI-t lSO,OCO lIS 120
"..
.... ollr.m,Clpllll ,,.""
Sarl'upc:mt 100.000

1.680.032 1....9,&40 1.7l9.540
"""' ...
Va'
,
240,000
OI"
O_lll
"'. ,
,.."" 0'"
0.... O.Hl
,
lllO.OOO
0'" Oj7' 0627

""',000 0.572
0."'2 0.",
,
"',000 0."97 0.", 0 ...

""',000 0."32 ons

0.'"
,
""',000 o.J76 0.279 0.ll7

""',000 0.321 0.212 0,219

,
""',000 0,2&4 0.194 0.247
"
"",000 0.247 0.161 0.212
Worlllll capnal 1j().OOO 0247 0.161 0.212
Salvag" value. "',000 0_247 0161 0.212
PV re.rurn 1,869.474 Ull,lH I.H6.l12
Thus, the OCFRQR IS 1681':'.
(2.5-5) - Tal Prod. CoSt (No. Years)
Capital Faclors
We prercr to aVOId trial-and-crror calculations In preliminary process designs. and
yet ",e would hke 10 account for lhe lIme value of money In some way In our
profitability To accomphsh thiS goal. we define a capital charge factor
(CCF) as
Hence, based on lhese simplifying assumptions,. we see that lhere is a dlrccl
relationship between CCF and the discounted-cash-f1ow rate of relurn i If we let
,_0.15 and N .11. "'e find that CCF_O.35I; whereas rIwe let i=<015 and
N _ 16. v.-e obtain CCF _ 0.306. In our prehminary dcslgn calculalions, for the
sake of simphcity. we let
Re\cnuc Tot Prod Cost = CCF(Tol Iny.l (2.57)
CCf = 0.33] yr-
1
(2.5-13)
Nhere lhe total lllyeQrncnl the ..... orkmg capital and the start-up costs
NOll> "'e can usc: this definiuon to ehmlnale Re1... cnue -101 Prod Cost. which
mvohes hourly costs, from Fq 2.5-6, so that .....e obtain an expression contailllng
only ioyestments;
A value of i _ 0.15 is the smallest value we would ever conSider for a new proJect;
i e" a value of 0 2 is more for safe projects. For a proJcct with a high risk,
such as in biotechnology. we might let CCF _ I yr I.
Io\e can subslllute Eq. 2.3-14 for tola) inveslment and Eq. 2_3-33 fOf total
production cost, 10 obtain
[FiXCC: CaP][(1 + it - I) + (Work Cap. + Start-uP)irl + i)'"
= [0.52(CCF)(Tol Inv.) + 0.48(Deprcc.)][(1 + i)" - I]
+(Work Cap of Sah' VaL),
If we solve for CCF. we obtain
(25-8)
A Simplified Profitability Mooel
Using the definition of the CCF
Reyenue - TOL Prod. Cost = CCF(fot. loy.) (2.'- 7)

CCF(2.36 Onsitc) = Revenue - [1.031(Raw MatI. + Uti!. + O.l86(0nsite)
I) + (Work. Cap + Start-up)i - 0.48(Dcprec.)
o52(Tol
+
In\ )(1 + i)'" _ 1)(1 + i) 1<
(Work Cap. + Sa" Valli + 048(Deprcc.)
O.52(Tot Inv.)[(1 + i)' I)
(2.'-9)
0'
+ 0.0256(ReYcnue) + 2.13 x lOs Operators
OAI3(Revenue) - 0.436(Raw Mati. + Util.)
CCF=
Onsile
+ 0.902 x 10' Operators
+ 0 .
nSlte
(2-5-14)
(2.'-15)
Also. we can relatc the other quantities appearing in Eq. 2.5-9 to fi"ed capital. using
the expressions we de\'eloped earlier;
Salvo Val. = 0.03(Fixed Cap.)
In our discussion of depreciation. we assumed Ihal
With these approximations. all the investment terms cancel in Eq 2.5-9. and .....c
find that
Start-up = O.I(Flxed Cap.)
Tot. Iny = L3(Fiked Cap.)
Work Cap. = O.I5(Tot. lnv.)
- a I5(J.3XFixed Cap) = O.l95(Fiud Cap.)
Depree. = O.l(Fixed Cap.)
CCF .l025(1 + i) + 0.295; - 0.298)_(1 + it' - 0.225/ -+ 0.048
0.676((1 + i)' I)
(2.5-10)
(2.3-3)
(2.3-12)
(23-4)
(2.'-11)
(2.3-36)
(2.5-12)
from estimates of the reveoues, the raw materials, the ulilities. and the
installed equipment COSLS (Onsite). we can calculale the CCF corresponding to a
design. tr this result is greater than 0.333. then the project appears promising and
we cao justify undertaking a more detailed design.
trlhe CCF is less than 0.333. we let CCF = i in Eq. 2.5-14. and we calculate
tbe revenue:
Reyenue = 1.06I{Raw MatI. + UliL) + O.998(Onsite) + 2.18 x lOS Operators
(2.5-16)
From tbesc revenues we cakulate the product price required to make the process
profitable. and ....-e undertake a supply-and-dcmand analysis,
Preliminary Design Profitabilily Mooel
For preliminary prooess designs where we are screening process alternatives. we
can approximate the cxpression above by the equation
Revcnue = Raw Mall. + Uti!. + Onsite + 2.13 )( lOS Operators (2.5-17)
Afler we complete Ihe miltenaJ and energy balances, we can eV.llluate the uulllY
costs, and Ilo'e revIse: the:: economic potenllal
Then, as we ealculale the cost of each piece of eqUipment, lIle can SUblraCI them
from the re\lse:d economic potenual
Our first of Illl: poleUllal (EP,), based Oil currenl product
prices, corresponds 10
(25-22)
(2.5-23)
Revenue - Tot. Prod. Cost
Tol. 10v.
d(Re\'enue - Tot Prod. Cosl):dl
d(Tot- InvYdx
0'
d(Re\enue - Tot. Prod_ Cosltdx 0
.]]]
d(Tol In' )fdX
In olher words., to find Ihe opumum design condmons for a case Ilohere the
mlmmum required produci price IS less than the currenl pritt. first we muimize
CCf by solving Eq. 25-22. Then we evaluale CCF at lbe optimum design; and if
lhis value is less than 0.3]]. we sohe lhe problem by using Eq. 2.5-23. If the
oplimum CCF docs exceed 0.333, we might want to consider the possibllllY of
increasing the plant capacHy, since the return on our invcstmenl will then be beller
than for most of our other projects. Of course. marketing consideralions may limn
this alternative.
However, ir thiS mlllllllum reqUIred produci price is less than the current
pnce. it probably is 10 build a larger plant and collect more re\'enues
Smce the CCF is dlreclly related to the DC"ROR by Eq. 2.5-12, we expectlhat the
ma;lIffiUm CCF lIlould correspond 10 Ihe maximum DCFROR. To find a design
variable .l that maXlmlZCS CC'F. we would wnle
dCCI' = d (Re\enUe - Tot. Prod cost) 0
dl dX TOI Inv
(Tol.lnv.)d(Re\eoue - Tot Prod Cost}/dX
(Tol In\)l
(Re\enue - TOI. Prod Cost)d[fOL lnv.)JdX
(TOL Inv.)l
However, close to the oplimum design condition, the incremental return on
an incremental invcstmenl will become \ery small If this is the case, it ....-ill be more
ad\'anlageous to allocate lhat incrementalm\cstment 10 a projcc1 where we would
obl3.l0 a 15% DCFROR or CCF = 0.333. Hence, from thLS consideration of
incrememal relurn on IOcremenlallO\'cstment, .... e r(:qulre thai
(2.5-18)
(2.5-19)
(2.5-20)
EJ>I - Revenue Raw Mall.
EP. -= Revenue - Ralll Mat. - UII!. - L (Onslle)J
i-I
If thiS COSI becomes sigmficamly less than zero al any pomt In the analysis. ""e
might want to redirect our efforts and 100" (or II$S expensive process alternatives.
rather Ihan to complete the design calculations. As in any creattve activity.
judgment is required to make lhls dCClsion.
In some cases it is easier 10 make judgments If we wnle Eq 25-11 In terms of
produci pnoes mslead of fC\'enues. For a process wllh a single major product, .... e
divide both sidcs of Eq 25-17 by the product flow rate (Prod). so that the IcJt-
hand side of the expression JUSt becomes Ibe prodUCI price. C,.:
C
Ra>lo MatI. Ulit Onslle (operalors)
... - Prod + Prod + Prod + 2.18 x lOs Prod (2.5-21)
The lenns on lhe nghl-hand Side are the conlnbullons or the \anous quanlllies 10
the total product poce; I e.. the unus of eacb term can be t/lb product. If any of lhe
terms on the right-hand Side arc \ery large compared to lhe current product prices.
we want 10 consider process alternall\es.
For cases where a process produces multiple products. such as a pelroleum
refinery. the analysis becomes more complex. In these: situations. we consider both
modificalions of the prOClCU thai lead to different product dislributions and
processes thai can be used to convert one type of product to another- We contmue
in Ihis way unliJ we have developed as many cost expresSions as there are products,
and then we look for the optimum process alternative and design condilions.
Optimum Design
In many silualions we want 10 find Ihe values of design variables, such as reaclor
conversion, Ihal maximize the profitability of the process. To do this, first we look
for the values of the design variables lhal wiU minimize Ihe produci price thaI
guarantees us a 15% DCFROR; ie_I we minimize C... in Eq 25-21 (or the more
enCI relalionship given by Eq 2.5-16 divided by Prod). If the minimum product
that we oblam from this analysis exceeds the current producl price, we use a
supply-and-dcmand analysis 10 decide whether we should lemnnate the projecl
Economic Dedsions among Process Alternalin5
10 general, we prefer to select the process allernative that salisfies the producllon
goal and requires the least capital invcstment, because with a specified CCF this
process normally will give lhe smallest producl price. However, if the least
cxpensive process involvcs a lot of unproven lcchnology. highly corrosivc or
hazardous materials, an uncertalll supply of raw materials, or other similar factors,
we must assess the addilional costs that we may encounter in overcoming polential

in addition, in some situalions we can decrease lhe losses of either materials
or energy from a process by installing addillonal equipment For lhese cases we
agam require that the mcremental return on thiS additionalmveslrnent satisfy our
m\e5tment enterion. i e; CCF of 0.333_
output structure of the ftololisheet, Le., 1e\c1 2 In the hierarchy, we can define an
economic potential EP
1
at tbis le\c1 as
EP, ,., Re\"enue - Raloli Mati
2.6 SIMPLIFYING THE ECONOMIC
ANALYSIS FOR CONCEPTUAL DESIGNS
In Eq. 2.5-17 we presented a \"ery simple economic model thal we can use: for
conceptual designs (te. the screening of a large number of flowsheet alternatives by
using order-of-magnitude estimates to determine the best flowshec:t or the best few
alternatives):
Economic Decisions for Process Modifications
or Replacl'ments
If our ne\lo' Idea in1l0h'es the modification or replacement of part of a process by a
ne'" technology. we SIll] ... anl to achie\e a 15 e;.. or more, rcturn on our investmenl
because this project loliiU be in competition With other projects considered by Ihe
company. The m\cstment required is equal to the COSt of the new equipment minus
the ac.ual maTur t"alur of the equipment we are replacing.. Note tbat we sbould use
the aClual market value in the calculalion rather than the original cost minus the
depreciation we ha\c already reco\"ered, becau.$C our original estimates of the
equipment life and the depreciation might have been in error. In other words, we
always base our economic dcrisions on present conditions. and we ignore our past
mistakes. just as we drop out of a poker game if the cards reveal we ha\'e link
chance of winning e\'en If we ha\'e a large stake in the pot.
The 5a1lmgs ... e expect to gam from the replacement are the old operatlOg
costs plus the depreoatlon of the old equipment over Its expected hfe as judged
from the present (and nOI the onginal deprecialion calculation) minus the:
operating costs for the nev. eqUipment plus the depreciation for this equipment
over Its cxpected hre savings provide a 15% relurn on the nel investment,
we might want to conSider the replacement project more detailed design and
costing procedures
Sunilarl)'. when we consider the recycle structure of the f1owsheel, i,e., level 3. and
v.e genera Ie cost tstlnlates for the reactor and a recycle gas compressor (if any). we
can wnte
(2.6-3)
(2.6-2) - (Power + ADn Cap. Cost of Feed Compress. if any)
EP
J
,. Re"enue - Raw MatI. - (Feed Compress. Cap
+ Op COSt) - Reactor Cost - (Gas-Recycle
Compress. Cap + Op. Cost)
If we follow option 3. we e\entually delermine a \'alue of the product price that
would make the process altemau\'e under consideration profitable. If thiS nev.
product pnoe were only slightly higher than the current price, we would probably
contmue wilh the design. (We need to undenake II supply-and-demand anal)-sis 10
sec how far in the future that .....e might e.tpect to obtain this higher pritt)
However, if the product price required to make the alternative profitable
Iil;ere much greater than the current price at any of the levels in the hierarchy. we
would terminate the work on the current alternative and look for one that was
cbeaper. If none of the alternatives were acceptable, we would terminate the
project This approach is very efficient because it makes it possible to termmate
projects with a minimum amount of design effort
I. Terminate the design stud)'.
L Look for a better process altemati\'e.
.l Increase the product pnce so that the economic polentlal IS zcro. and contmue
wilh the design.
Thus. as we add more detail to the ftowsheet. we merely sublract lhe ncv.
utilities costs and the annualized, installed equipment cost of the new equipmenl
that is added If the economic polential at any le,'e1 becomes negative. we havc
three options:
(2.6-1)
Revenues., Raw MatI. + UtiL + Ann_ Install. EqUiP, Cost
+ 2.13 )( 10' Opcratof!i
The: annualized instalkd equipment costs lire determined by multiplying the
installed equIpment costs (sec Sec. 2,2) by a CCF which includes all the investment-
related costs
Economic Potential
In Chap" I we presented a hIerarchical decision procedure that would simplify the
development of a conceptual design The appro.timate COSt model presented abO\'C
fits lOtO the hICrarchical framework \ery nittly Thus. when we consider the input-
Significanf Equipment Items
The case study considered in Sec. 21 is somewhat unusual because one piece of
equipment (the recycle compressor C-I) comprises almost half of the total
purchased (or installed) equipment cost. However. suppose we consider another
ca.sc study, for tbe disproporlionation of toluene to produce benzene and xylene.
R. J Hmplebeck IlDd J T 8.lacheto. Du"optl",_tlOtf t{ WaslunglQII De.p
Caw Study No I. ediled by R 0 Snuth. Wash,nston Un,."...ly. St- Lou!$, Mo JII_ 26. t969
TABLE 1.6.. 1
InHslmenl summary, S
Pumps (1949) "'urP<:ao (I969j
""M
PI
'.900 Reaaor (1969) 19.800
P2 1)20
p.]
1,9SO
Towers (1969)
PA ....,
TI
,,"'"
PA 2.>00
T2 37,600
p.,
'SO
T.'
".
--
Total 98.100
p.

Pumps (19-49).
TrIo)"S (1969)
tncIudllll spates 28,760
TI ',800
Pumps (1%9)
".000
T2 )1,100
T' '2.000
EJ.Cha.llgul' (1968)
TOlal 79,000
E' 140,000
E2 115,000
CompreAOB (1969)
E'
'.800
C,
EA
2.2.000
C, 31l,ooo
E' 26.000 DnulUi (1969) 23,650

"000
IDSlailcd COlA liumllUlfJ
E'
'.<OIl p=,. ItIS,OOO
E' ',200
1.140..000
,.,
,........
121,000
10 17,000
Towers (el< lrays)
'90,000
E11
'.300

""""
E.. 12
'.>00
ComJ'feMOn 751.000
'00,200 Drums 130,000
(1969) ....000
f",n"""
52),000
---
3,142,000
The equipment COSIS rOf the ftowsheet shown in Fig 2.6-1 are listed in Table 2.6-1,
and tbe operal1ng costs are gJ\'cn in Table 2.6-2. A cost summary fo(tbe process IS
preseoled In Table 2.6-3.
When we examine Tables 2..14 and 2.6-1, ..llC sec that the costs of pumps and
drums are only a small fraction of the 10lal costs. If wc neglect lhese costs (or
Simply assume Ihat they are aboul 10y. ofthe lotal), then we can save Ihe effort of
desigmog a large fraction of the total number of pieces of equipment and yet
Introduce only a small error an our calculations. SimIlarly, if we assume that the
costs oflhe feed tanks and product storage tanks will be essentially the same for all
the process alteroatives, then we can omit them from our screening calculations.
Of course, the process WIll not operate without the pumps, drums, feed tanks,
and storage tanks. Ilowe\-er, If our screening calculatIons Indicate that the process
IS not profitable and that the project should be teomnated when we do not include
these costs. then we never need to design them. Thus. for conceptual designs we
FJOm R J HCllplebccl al>d J T B"",<I,.,o, Wuh,nlloa U", nny Owl" CaM: S101dy No. a.
<dncd lot B D Slim'" W""lI"JIOI' UII, nny. 5. Lou... 1%9

.;;

"'

.,;
0
0
"'
N

M
..
...
"'

,
M

f
Ul

"
0

"
,
%
f
...

...

0
<
j;-

- -
+
r<:'

...
"'
<
N

U
-
ol!

u.
... .
11
-;-
N

"'

< ' .
Ul

D'
,

..
N
g

g li'
N

R.

: ,
"

C:cS<:i
66
TABLE 2..6-2
COS, summny, $1000
Ullhlln Taus. ins",..noc
16'
Powr, m Rrpalu 2SO
Slum
'20
MI5OI:lllna>Ull
"
F'"
H1 Plpol d'larp:l
" "ala JO
Totll I SO
lotal 120'
SARF 16.
Lobo.
"
Clu.I)'51
'" SuprrvlSlOn
"
--
Total 2060
F.... k J 1l...1lldood aDd J T Iblldoon-o. U..._y
Deoop c_ S..ldy No .. ed>oeoI by B 0 s.illl. ""......_ 0......
...y. Sl. '-Mo. 1M
include only the COSls of the Significant equipment ilems. This approach is in
agreement with lhe engineering method discussed in Chap I,
2.7 SUM 1ARY. EXERCISF..'5, A 0
'OMENCLATURE
between these t ..... o dIfferent types of quanlitteS. we mUSI consKk:r the time value of
money Thus. by usmg inlerest calculations to determme the: present value of IWO
allematives. we can compare them on the: basis. The present value (PV) of an
Investment I plus annual paymenu R with an interest rate i i..
R
PV=J+ [1-(1+11")

Once we find lhe best alternative, we mUSI evaluate the: lolal cosl aSSOClaled
wtlh the process, to see whe'her additional engincenng effort can be: Jusllfic:d. That
is, we must include the cost of the offsite facilities. maintenance and repaIr..,
.... orking capital, start-up costs, elc. These various factors are discussed in Sec. 2,].
and a profitability modd IS developed in Sec.. 2.5, This modd provides an
ellplanatlon for the simple cost analyses fhat \,l.e use throughout this lext
NOle lbat we still have not considered the control of the process, or
envIronmental factors in adequate detail Any of these faclors might make the
process unprofitable. lienee. the profilability calc;.llat;ons for our conceptual
design merely provide a basis Ibat we can use to Judge .... hether morc detailed
design slUd"es can bejusufied By including rough esllmales of Ihe other
costs, howe\er, W'C are better able: to make Ihis judgmenL
Exercises
2.7-1. Dc:nve an npresSloD for the \'alueofan annUllyahel n yr I'lhe first pa}'mentls made
ill lIme zero. ralhe, llun ill lhe end of the lint year
2.7-2. A fnend of lOUrs joins I ChriSlmas Gub al a Ioc:al banI. She deposlls SID mo
starling on JanWiry I and receIVes SilOal Ihe beglnmng of tlrettmber If lhe nominal
rale available: is monlhly. he... much ,nt(fest does Ihe bank keep
for providmg thrs servICe? (No/t, This payment plan IS dIfferent from the annUity
schedule dISCUSsed in Ihe te.l)
2.7), St Mary's Cemetry in Nonhampton. Massachussctts, charges SIlO for a amelry
plot and ISO for perpetual care of the plot. Al a nominal mterest rate of 6 %
compounded monlhly.....hat are the expected annual maintenanc:c charges"
2.1-<4. Some U1\I\eflillles are fortunale 10 hl\'e endowed chaIrs for lheir outslandmg facull}.
and oflen these cbaJfli provide S50.OlXt/yr. If the nominal mleresl rate 1$ 10%
compounded eaDImuously, bo.... moch money is required to establish a chair"
2.7-5. When we make monlhly payments on a car or a house, the amount of the payment is
a constlnt, but different fractions of this payment represent repayment of the
pnllCJpaJ and lhe iolerest. Also. lbe fracoon anoealed to cadi o\'er lbe loan
pcnod. Using the lennmology below, develop expressions for the pnncipal and
interest payments dunng year r
" - prinCIpal paid at end of lsI month
Summary
When we compare process alternalu'es, nonnally there are different economic
trade-offs bet\loccn capnal and operating costs. To make valid compansons
TA8LE 2.6-3
In.estment and operlling summary
Con"n5>On/pass. JO
LAbor aDd 5llprrvtSIOO' 016
PUfFPS N.
Tua Ind insul"ID(r
017
tDYalmcnls, S IIIlIbons
Repa.., aDd JlllJCelJanc:ous
OJ]
IS8L
'"
Calal)'ll
006
OS8L
112
SARE
0"
-
>06
4.86
WorkIng caPllllo
100
Maleoals. BCDI (60F)
'"
Tolurnc: feed
"..
Calalyslmw:olory
'06
PrDdllCts
j.1l2
.....M
"'"
x,_
2000
OprrallDI coou.. St mlDoon"yr
H, Feed. 10" SCFO'
'"
Uhllloes
120 Fuel p.s. 10" Btu/day 1700
Fr..... "- J ll lcb.d and J T Banc:ho,o. WUbJqlOOl ",""lpI Cur Sr""y I>lo .. od,rod by 8 0
W IP_ U.",,",ly. SL Lou... Mo. 1969
"" .. ! -..d lO_y '" Inod ud ,.ndooaa,.-.lb r-. ,.oducu ......... ' ...
, I......... duo,,.,.
, ICD .. boo"dsIcaInodar day
, SCTD - ."'wrd nobo<: ""'vda,
M _ amount ofmorlgage
R .. nomInal Inlerest rate
, _ R,'12" monthly rale
II "" no monlhs for loan
p" monlhly
r _ annual paymenl
P, - pn1lCJpal paid dunng year J
I, ,. mleresl paid during lear J
70 J1 EXEJlOSES. "'NO NOMENClAnJU
Sl!ClION Jl l;UMIolAIY, uuasn.... ND NOMENCI ..1UllE 71
2.7-6. Estllnate: the: purchased and Installe:d equIpment COlli of. JOOOft\ SS/SS, U-Iubc
heilt e:xchange:r opc:r.atmg at ]80 Ibl
2.7-7. Dc..e:lop an CJlpl"CSSIOD for tbe: total lIl.. eslmc.ot U..1: a.uumc: that 'be: ..orLmg capllal
... ] months' .. orlb oftbc product re:.e:nl>CS
2.1-8. De.clop an uprcsslon for the: profit before: lUes If \Ioe: a.uumc: that the: labor eo)t)
are: or the: lOU.! pcoduct COStS (sec: Table: 2.]-2)
2.7-9. tk\c:Iop an uprCSSIOD fur the: 10lal upst;!l m\CSllne:nt for a grass rOOIS plant
(;tSSume: olf)lIe: eosls are: ] limes onsne:
2.7-10. For an 11)'r plantlifc and a OCFROR .. \Io'hat I) the: capnal charge faC1or'
2.7-11. Calculate: the payout time: and thc DCFROR for thc proocss dcscribed in Sec 2 I U
you need to Introduce: additional assumpllons, c1e:arly state: the:se: assumptions
2.7-12. Calculate: thc ROI, payout time:, and DCFROR for the: process de:scribed in Sec. 26
If you nc:c:d to introduce: addlllonal assumptions, cle:arly state: these: assumptions
Nomenclature
A
C.
CCf
C.
C
N
Contmg.
Depree
d,
EP,
F,
FCI
F.
Fixed Cap

F,
'.n
"
IF
Imur
ISBL
Lab
m
Main!.
Manu. COSI
n
Op. Supply
OSBL
Plant OVHD
Hcal-<xehange:r area
Bast cost for carbon sleel
Capttal charge factor
Exchanger COSt
Cosl of producI, lb
ContlgeOC} costs
Dcpreciallon
Annual depreclallon
EconomIc at ]C:\c:1 ,
Correction factor for pressure:, materials of construclion,
etc.
Fixed capiJaI investment
Design-type correCllon factor
Fixed capital COSt
Malerial-of-conSlrllction correclion factor
Pressure corrttlion faClor
Effcclive InlereSI rate
Conlinuous interest
Imlallatlon factor
Insurance
Inside-ballery-hmlls costs = onSlle costs
Laboralory costs
No. payment penods per year
Repair and mallltc:nancc costs
Manufacturing costs
No.
Operating SUppllCS
Outside-ballery-hmllS COSIS = offSHe costs
Plant overhead
P,
PV
,
R
Raw Mali
ROI
R,
S
Salvo Val
SARE
Start-up
Supervis
,
TC
Tot Inv
TOL Prod_ COSt
Uti!.
Work: Cap
I
Pnnclpal
Presenl value
omlDal interest
Periodic pa}'menl
Raw-matcnal costs
Return on lO\cslmenl
Continuous payment
Amount of money In a ban" account
Sahiage value
Sales, admlnlslratlOn, research, and e:ngineering costs
Starting costs
Direct supervision
Timc
TOlal cost
TOlal invcslmeOl
TOlal product cost
UI,htlCS COSIS
Workmg capllal
Design vanabks
CHAPTER
3
SHTlOW 'I n08lHl OEnNmON ... N" OENEkA.l CO"lSIDElAlIONS 73
Economic Potential
The first step In the anal)-sis of any deslsn problem is to e\aluate the economic
significance of the project Initially we do not know what fraction ohlle acetone we
might aHempt to recover, but rather than spend time on this decision we merely
base the calculation on complete recovery. Thus, we calculate an economic
potential (EP) as
Operating Time
It IS conventional practice to report operatmg costs or stream costs on an annual
basiS. Different companies use somewhat different values for the number of
operating hours per year, and they may even use different values for different types
of projects. We .....iII usc 8150 hr/yr for continuous processes and 7500 hr,yr for
batch. (nus operating time includes scheduled shutdowns for maintenance,
unplanned downtime due to mechanical failures, and/or production losses caused
capacity limitations or lack of fttd.)
ECONOMIC
DECISION
MAKING:
DESIGN
OF A
SOLVE T
RECOVERY
SYSTEM
EP = Prod Value Raw Mati Cost
Or. since .... e are currently burning the acetone,
EP = (10.3 molfhr)(0.27 Sjlb)(58 Ib/moIX8150 hr/yr)
= $1.315 x 106/yr
(3.1-1)
(3.1-2)
To illustrate how process alternati\-es can be generated and the use of order-of-
magnitude calculations to make economic decisions, we consider the very simple
example of the design of a SOlveDt recovery system
3.1 PROBLEM DEFINITION AND
GENERAL CONSIDERATIONS
We assume that as part of a process design problem there IS a stream containlOg
10.3 mol(hr of actlone and 687 moljhr of air that is being fed to a flare system (10
avoid air poUution). The design question of interest is; Should we recover some of
tbe acetone?
Thill "udr I$' moo:hlitd ..,noon olthc 19H AI<:bE Studenl Conl($l PIOI>lcnl. Jet: 1 J McKclla.
... O/C'-wal IIftIl DWp...,( I. Dd:hr. Ne..- YOlk. 1976, P )1"
12
Process Alternatives
The next question we ask is; How can we reco\er the acetooc7 From our
k.nowledge of unit operations, we might list the alternatives shown in Table 3.1-1
We might be abie to think of other alternatives, and so we need to make some
judgment about using conventional technology versus the cost of doing the
de\dopment work required to design and evaluate unconventionlH ahematives.
TABLE J.I-I
Soh-ent recanr, ahematil'C!i
L CoDdr_uoa
.. Hlp prasun:
b Lo... temperaturc
c A combln.lion of boll!
1. Absorption
J,. AdJorpilon
<l A mcmbr'M sepal.llon .""em
$. A ruc:tioa p'0CQI
74 u.cnON u OUIGN 01 A GU AlSOnU
II DUffiN or A GU Al:lSO\laU 75
Acetone
Water
Acetone
Acetone
Solvent
It is reasonable to question whether dl.scardlDg the process water, as is shown
in Fig. 3.2-1, can ever be justi6cd-e\en when a pollution treatment facility IS
available. 1llc only justification IS based on the temperature of the process water
entering the gas absorber If a recycle proa:ss is used, we cool the recycle stream
WiIh cooling ....... Ier. Normally ..... e assume that cooling ..... ater IS available from the
cooliog towers at 9O"F (on a hot summer day) and that it must be returned to the
cooling to.....ers at a temperature less than 12O"'F (to pre\ent exccsshe scale
formalloo on the exchanger surface) T'hen If we assume a IO"F driving force at the
cold end of the exchanger, Ihe recycled process water Will enter the gas absorber al
IOO=F.
However, If we use well water as the whent, theo the temperature of the
water fed to the absorber might be 7rF, or possibly less. It is advantageous to
operate the absorber al as Iowa temperature as possible, which we can achieve If
flCUR..[ 3.2-1
AcetODC absol!>et
Air
We arbItrarily deade to consider lhe design of a gas absorber tirst, although \loe
recognize thai we must also consider alternative designs before we deode whIch
process we might build. This very simple design problem illustrates the use of
shortcut calculations, rules of thumb, and other process altemati\'cs.
3.2 DESIGN OF A GAS ABSORBER,
FLOWSHEET, MATERIAL AND E 'ERGY
BALANCES. AND STREAM COSTS
Now we come to lhis Which is Ihe cheapest alternative? Fair-
suggests Ihat any time the solUle concentration in a gas Slream is less than 5"' ..
adsorption is the cheapest process. However, many petroleum companies prefer to
use condensation or absorption systems because the companies have much more
experience design109 and operallOg these types of units. Furthermore, only a few
vendors sell adsorption equipment Thus, we are again required 10 make a
judgment concefOlng the usc of technology where we ha\e a great deal of
experience \'ersus using a technology where .....e have much less experience. We
should base this decision on the rdatlve costs, as well as the risks, of the various
processes In question.
Of course, we do not know the costs of the various alternatives until we
design each of them. We do not necessarily want to develop rigorous designs
initially, because (at best) we would build only one of them. Henoe. we only want to
include sufficient accuracy in our screening calculation to determine which
ahernallve is the cheapest (or to see whether they have about the same costs), aod
theo we will develop a rigorous design if we decide to build the process.
On numerous occasioos In design we cao develop a number of alternatives
that appear as If the)' Will do the Job If we do oot mow which alternative is the
cheapest, we should consider designing them all By doing quick design calcula-
tions, we can simplify the design effort required to make dccisK>os.
Flo"",sheels Mod Ahermllhes 10 Gas Absorption
Before we can do any calculations, it is necessary to invent a flowshect for the
process. The Simplest possible ftowshcct we might imagine is shown in Fig. 3.2-1.
We use water as a solvent (because It is cheap) to recover acetone from the air
stream in the gas absorber, and then we distill the acetone product from the water
and throwaway the water (environmental constraints may preclude this possibili-
ty, so we include a cost for pollution treatment). Of course, we could recycle the
proa:ss water to the glls absorber (see Fig. 3.2-2), and III Ihis way we would avoid
any environmenlal problems If we used anything other than water as a solvent, we
would always recover and recycle the solvent.
Acetone
Hoo'
J R Fa'r, M,,,.,d 50/"""1 Rrco.,.,,1' "'" Pur,jiUJ/lO". p I, Washlllvon Uruverslly DeI;,&" Cue 51udy
No 7, alllW by 8 D Smnh, Wuhlllglon Umvcnlly, 51 Lou,s. !>to, 1969
FIGURE J.l-l
AlXIODC
well water is available as a solven! and if we do not recycle the ..... ater ThiS
reasoning is the basis for a design heuristic
Thus, we find that we ha\e two alternatl\e ftowsheets. If we do nOI know
which is the cheaper. y,e should dcslgn both, In addition, we must evaluate whether
we really wanl to use y,aler as the solvent. We arbitrarily choose to consider the
f1owshec:t shown in Fig. 3.2-1 first. because it is the simplest. (3.2-2)
If a raw material component is used as the solvent in a gas
absorber. consider feeding the prooess through the gas
absorber
(3.2-1)
Similarly, there is an optimum fracllonal recovery overhead in lhe distillation
column As we add more and more plates in the stripping sa:tion of this column,
the still cost Increases, but the value of lhe acetone losl to the sey,er decreases.
We cannot find the optimum fractional recoveries of acetone in eilher the
absorber or the stili unless y,e carry oul detailed designs of these columns
Furthermore. it is not clear that we want to find the optimum design conditions
unlil we have decided to build an absorpllon process Hena'. for our first set of
calculations. we base the design a rule of thumb (i e. a heuoslic):
It is desirable 10 rcco\er more than of all valuable
materials (we normally use a rcco\ery as a first
guess).
EXACT MATERIAL BAL4.,:..CES. Wllh these rules of thumb. it is a straight-
forward task to calculate the malenal balanocs. For lhe acetone water s)'stem al
7rF aDd 1atm. 1 = 6.7 and fY "" so that
6.7(229)
m = ----p; - 760 - 2.02 (3.2-4)
L = 14mG.., 1.4(2.02X687) = 1943 mol,lhr (3.2.5)
For a rCCO\'ery of acelone III the gas absorber, the aa:tone lost is
We discuss this rule of thumb III greater dctaBlater in thiS chapter.
We also ha\'e to fi,; Ihe water Row rate to the @as absorber before we can
calculate a of material balances. The greater Ihe soh'ent flow rate. the fewer
trays that are required to achic\'e a fi,;ed fractional rcco\'ery in the absorber. but the
greater the load on the distillation column Thus, there is an optimum sohent no\lo
rate. Again, we prefer not to optimize: designs at lhe screemng stage if a heunstic IS
available. and so we use another rule of thumb to fiJI. the sohent flow:
For an isothermal. dilute choose I such that L = IAmG (3.23)
We also dlseuss this rule of thumb later
Also. jf the product of acetone i<; specified 10 be 99"<;. the amounl of
wate, in the product stream will be
Millerial BalarK"e';
DISTRIBunON OF COMPONENTS. Ona: we have specified a flowsheet. we
must try to K1enllfy the components that will appear in ew:ry stream. The: IIllet gas
ftow to the absorber IS given in the problem statement as 10_3 mol(hr of ac:etoM
and 687 mol,lhr of au If we use well water as a solvent, tben the inlet soh'ent stream
is pure water_ The gas leaving the absorber will contain air. some acctone: (we can
never obtain a comp&cu: recovery), and some water. Since water is rclati\'c1y
ine,;pcnsive. we neglect this solvent loss in our first design calculations. However. if
any other sohent is used, it IS essenlial to include the solvCIltloss in the cconcomic
analysis. as we demonstrate later
The feed stream to the distillation column will conlain ....ater. most of the
acctone. and some dlssohed air. Probably we caD neglect this dissohed air III our
first of design cakulations, but it is essential to rccognize that we must put a
vent on the distillation oolumn condenser, to prevent the accumulation of air in this
unit. The overhead from the distillation column will tben contain acetone and
water. while: the bottoms will contain water and some acetone:..
We now sec that acelone leaves the process in three places: the exil gas stream
from the absorber. the distillate overhead. and the distillate bottom stream.
Nonnally, we would be given a product specification for the product stream. which
would correspond to either the specifK:ation that the company established when
they used aa:tone as a raw material or the specification that the marketing group
would indicate when they try to sell this acetone. However. even if we fh the
acelone composition in the overhead of the distillation column. we still cannot
calculate an acetone material balana: until we specify some infonnation about the
aa:lone leaving in the other two streams.
RULES 01' THUMB. Ofeourse, we can recover 90, or 99, or 99.9%, or whalever.
of the acetone in the gas absorber, simply by adding more trays to the top of Ihe
absorber. The COSI of the gas absorber Will continue to increase as we increase the
fractional recovery. but the value of the acetone lost to the flare system 1'1'111
continue 10 decrease. Thus. there is an optimum fractional recovery.
0.005(10,3) - 0.0515 moljhr
and the acetone ftow to the distillation column is
0.995(10.]) "" 10.25 molfhr
U 99.5 %of the acetone entering lhe still IS recovered overhead. we ohtain
0.995(10.25) "" 1020 mol(hr
(
' 099\
0.99 /1020) = 0 10 mol/hr
(3.2.6)
(3.27)
(3 28)
(129)
78 II llUlGN 01 .. GAS AISOUU SECTION 11 UUK.N uF .. OAS 79
Then the bollom Rows of acetone and water are
and .....e can wnte
.nd
0005(10.25) ... 0005 mol/hr
1943 0 I _ 1942_9 mol(hr
(3.210)
(3.2.11)
Y. = P';/Pr
lIence, a quid, way to estImate the soh-enl loss IS 10 write
(12-15)
Stream costs. Once Iloe have esllmaled material balances, ....e calculate the
stream costs
(SO.27/1bX58Ib/moIX00515 molfbrX8150 hr/yr) = 56600/yr (32-18)
Pollution treatment casu (assume SO.25/1b BOD and lib acclone/lb BOD) art:
(SO.25/1b BOOXllb BOO/lb acctoneX58Ib/mol)(00515 molfhrX8150 hr/yr)
= S6100'yr (3.2-19)
Se.....er charges (assume 50.2/1000 gal) are
ACETONE-WATER PROCESS. For the acetone-waler system wilh no recycling
and 99.5 % recoveries, we find from our approximate material balances the
following:
Acetone loss 10 absorber overhead (assume SO.27/lb) is
(SO.27/1bX58Ib/moIXo.0515 molfhrX8150 hr/yr) = S66OO/yr (3.2-11)
AcetoDC loss in slill bouoms is
(
SO.75 XI g.1 \,
1000 gal 8.34 Ibfl8lb/molXl943 molfbrX8150 hr/yr) - 525,600/yr (3.2-21)
Since each of these COSIS is essentially negbglble compared 10 the economic
polential of SIJ 15 x lr per year, we want to contmue developlDg the design. Also
we note thaI there is lillie mCCOlI\'e ID using morc rigorous material balances or in
getting very accurale costs for the pollution treatment system. In fact. we can
tolerate IOOy" errors in our calculations. and the conclUSion of negligible stream
casu will DOl be affected.
(12-16)
(3.220)
m _ JI. GZJG
1 - J.
(
'0.20 )( 1g.1 \,
1000 gal 8.341b! 18 Ib/mol)(1943 mol/hr)(8150 hr/yr) = S6S00/yr
Sohen! water (assume $0.75/1000 gal) is
APPROXIMATE MATENIAL 8ALA1\Ct:S. Even though these material-balance
calculatlons are \ery SImple and quick. accordlDg to the englDeenng method do
nOI want to do any calculations unless they provide us wllh SIgnificant infonnation.
Thus, we want to explore the POSSibility of developmg a set of material balances
that are almost correct (withm 10'1.. or so) with the minimum amount of effort We
need to calculate Ihe solvent feed rate in the same way as we did before, but there is
a simpler way of obtaining good estimates of the other nows
If we assume thaI 99.5% recovery (or any recovery greater than 99%) is
essentially equivalent to a 100% rcco\'ery (actually. we are willing to tolerate 10%
error). then the acetone Row in the disllllatlon column overhead becomes 10.3
versus 10.2. Also since a 99 %acetone composit;'.m in the distillate: is essentially the
same as 100% acetone, the .....ater Row leaving the bottom of the column could be
wnllen 3S 1943, versus 19429. Similarly, Ibe feed stream to the still becomes
10.3 molfbr of acelone and 1943 molfhr of water. versus 10.25 + 1943. Thus, we sec:
that we esseDliall} mal.e no errors m the stream Rows if .....e merely assume 100%
recoveries and pure streams and lhen just wrile down the Rows.
Of course......e cannot base: the: design of the gas absorber or the: still on 100%
recoveries_ Moreo\er, we would hl.e to esllmate Ibe acetone: losses from the
absorber m'erhead and the still bonoms. Howner, suppose .... e use 99.5 %
recoveries for these calculations. Our material balances will no longer quite
balance (although they are certamly wllhin 10% of the exact answers), bUI our
calculation effort will decrease. The savings in time are not significant for this
simple problem, but II can be: for complex plant designs. Thus., .....e use approximate
material balances for screening calculations throughoul thiS lext. although we
would make rigorous material balances as we proceeded with final design
calculations.
SOl.VENT LOSS. For ca-5C$ wherc Il.C usc a solvenl olher than water and we
recover and rccycle this solvent. as is shown in Fig. 3.2-2, II is essenlial to estimate
the loss of this solvent ID the absorber eltlt gas stream very carly in the design
calculations. For a low-pressure absorber, the fugacity correction factors arc
negligible, and the ... apor liquid eqUilibrium relationship for the solvent can be
written as
(3.213)
With greater than 99% recovery of the solute, the soillent composition on the
top Iray will be essenllally
(12-22)
COST OF SOLVENT LOSS. Suppose we also consider using methyl Isobutyl
ketooe(MIBK) as a solveD(, and we recycle the MIBK H1": = 0.0237 atm at 77"F
and P
T
= I aIm, from Eq. 3.2-16 we find that
MI BK loss (assume SO.35/lb = 535/0101)
= (535/mol)(0.0237(687 mollhr)](8150 hr/yr)
= $4.464 x 10D/yr
(3.2-12)
(32-14)
1." 1
If a solvent IS used that is ID homologous senes with the solute. then
80 SECT'ON Il I>SIGN or GAS SfCll0"l " OESlG"l or A GAS "'SOIt.u 81
When we compare thIs value to the economIc potentIal, \loe see that y,e want to
dror any idea of usin@ MIBK as a solvent. Morco\'er, \loe are glad that y,e dId not
take the trouble to dtslgn the absorber and stIli before \loe calculatNi the stream
costs. Agam we find that our dCClsion does not depend to any great elltent on the
accuracy of the calculation Thus, we usc the economic analysis to help us decide on
the accuracy we need for our design calculatIons.
Balances fOI" Ihe Acetone Absorbel"
Our origlOal design problem was undcrdefined, and thus II \\as neccssarl to usc
some rules 01 thumb (greater than reco\ents and L _ I 4mG) to be able: 10
calculate a sel of malenal balances. It should not be surprising Ihat the same
problem is encountered when we try to write energy balances. Thus, we need 10 fill
each of the stream temperatures in order to estimate the energy flows.
Since the inlet composition to the gas absorber is quite dilute, we might
assume that the absorber will opera Ie isothermally. Hence, if thc gas and liqUId
streams entenng the absorber are at 77 F, we assume that the ellit streams are at
Ihls same: temprrature (sec Fig. 3.2-3).
We do not want 10 store our product stream at its boiling pOInt, so we IOslall
a product cooler. With cooling water available at 9O"F and a drivlDg force,
the temperature of the product slream leaving the product cookr ....ill be
Our acetone product contains I %water. But rather than calculate the bubble
point of the distillate, ....e might merely guess that the temperature or the overhead
is essentially the same as the boiling point of acetone (135"F); i.e., we Cltpcct that
the error in calculating the heat load of the product cooler caused by this
assumption will be: negligible. Similarly, we assume thai Ihe bottom stream from
the still is at 212"F (because there is only 0.05 mol of acetone as compared to
Air
77F
Water
nF
90
120
135F

100F Coolant
Acetone
1943 mol of y,alcr) and that \\C must cool IhlS waste stream 10 100 F (cooling-
91ater temperature) pnor to pollution treatment (It ...ould be beller 10 assume that
the bottom of the column IS at 5 to 10 psig, rather than atmospheric prtssure. but
again the error will be small)
We still must spccif) the temperature of the stream entering the distillation
column. Saturated-liquid feeds arc the most common case, and so we might guess a
temperaturc of 20<rF or so (10.3 mol of acetone and 1943 mol of water) Again, we
ellpect that a guessed value WII! enable to calculate the load on the still preheater
\\ithout calculating the bubble point of the feed miltture or correcting for a column
operating pressure of 5 to 10 psig
If \loe do not preheal the feed stream entering the dIstillatIon column to close
to the saturated-liqUId conc.htion, then we will ha\'e a supercooled liqUId entering
the still and the heat load on the still reboiler will increasc. Thus. the lotal energy
demand on the preheater and the still reboiler is essentially constant 1I0wc\er.
usually we prefer to preheat the feed because T10nnally we can usc a hot process
stream that needs to be cooled down as the source of heat, rather than using steam
from a utility supply. Hence. we need to consider the energy integration of the
process as pari of our dtsl1!n activllY
E.....ERGy BALANCES. Once we ha'e SpeCIfied the stream temperatures and we
ha\e esllmated all the stream flows, It is a Simple mailer to calculate the: heat loads
of the various streams uSing the eltpresslOn
(32-23)
Then we could decide on a heat-cllchanger network and calculate the heat-
ellchanger areas, the annuailled heat-cltchanger capital costs, and lhe utility costs
For a simple process, sucb as the one 1;I'e are considering, tbis would be a
reasonable procedure. In general, however, we .....ant to energy-integrate the
absorber-stripper heat loads with those of the remainder of the process, and
therefore we defer tbe energy analysis.
FIGURE J.1.J
5ueuII Il'lllrera1ur<:s-
Hut
?
120 90
276F, 25 psia
120 90
Coolant Sewer
L:':"::::='--' 1000F
PROCESS ALTERNATI\'E. In our selection orthe stream temperatures, we noted
tbal tbe still bollom was almost pure water (0.05 mol acetone and 1943 mol of
water). For this case. the column reboiler uses 25-psia steam to generate essenlially
l5-psia steam Ihat is returned to the column. As a process alternative, we could
eliminate the reboiler and feed live steam to the column. We pay a penalty with this
approach, howe\'er, because once the live steam is condensed. it must go 10 the
pollution treatment system and then is lost to the sewer (whereas the steam leaving
a reboiler would be vaponzed and recycled through the closed steam system). Also,
boiler feed water has a higher quality (it is demineralized, etc) and is therefore
more eltpc.:nSlve than process water. Hence. we must balance the reboilcr savings
8@alOst the incremental cost of boiler feed .....ater. pollution treat men!. and q: ....er
costs to see .... hether this alternative is worth pursuing.
SE<TKJN lJ EQUIPMENT OUI(;N' COt<;S!OEIATlONS 8J
For the: acetone:-water system, the vaJue: Qf m W".tS gl\len
3.3 EQUIPME T DFSIGN CONSIDERATIONS
In addition 10 calculating lhe sizes of the heat exchangers, we must calculale: the:
size and cost of the absorber and Ihe slllI Before we begin any calculations,
however, we wanl to underSland the cause-and-effecl relationships of Ihe: design
variables and to sec: whether we can simplify Ihe normal unit-operalions models.
n-n:cr OF SOLVEI'H.
by
6.7(229)
1/1 - Pl- = 760 - .. 2.02 (3.3-4)
(3.3-1)
Gas Absorber
For isorhermal, dilute systems, the Kremsc:r equation can be used to calculate the
Dumber of theoretical trays required 10 the gas absorber;
N + IX::-::)+t]

We sec: Ihal If we: usc: a sohent such a.s MIBK Ihat forms an c:ssc:ntiaJly ideal
mlXlure: \lonh ace:tone:, so lhal i"; I, then frOIll Eq lJ-J weculthe liquid ralc: by a
factor of 6.7 (and decrease the stlll cost). lIowever, from Eq J3-1 we sec: that the
required number of plales 10 Ihe absorber docs not change
EFFECT OF OPERATING TEMPERATURE. If we change the inlet waler temper-
ature to 4O"C, then 1 = 7.8 and P" = 421 mmHg. Thus, (rom Eq. 3.3-3 we see that
L will increase (so Ihal Ihe stIll costs will IIlcrease), but the: number of absorber
trays (sec: Eq. 3.3-12) Will remaIn the same.
If pure water is used as the solvent, then XI. ""' O. From the: rules of thumb discussed
carlier, we know Ihat
(3.3-5)
N + I::;:N
BacL;-(t(.lbe-Ennlope Design Equation
The Kre:mser equation, Eq. 3.3-1. IS actually a qUlle slmplc: equallon Ihal can be:
used to design gas absorbers for isothermal operalion with dllule fc:eds. However.
In accordance with the engineering mel hod and our baSIC to do calculations
only If we gain some Significant information from this cffon \loe would like to
review the Kremsc:r equation and to evaluate the significance o'f each term. We do
this by examining the orde:r of magnitude of the various terms in the equallon.
First let us consider the left-hand side of Eq 3.3-1. i.e. the term N + I We are
concerned with obtaining accurate estimates of only lhe items Ihat are c:xpensive_
We: upcct that relatively expensive absorbers will contain 10 to 20 theoretical trays
(tbe: COSt or a gas absorber Willi only 2 or 3 trays will probably be ne:gllglbk:
compared to a a gas compressor, a dlstillallon column with JO to 20 trays,
etc.). If we dcode not to undertake a calculation unless II changes the result by
more: than then we sec: thai we: can simplify the side of the Krc:mscr
c:quallon by writing
S)"slems Approach Versus Unit Operations
The simple examples diSCussed above clearly reveal Ihat the Interacllon bc:1\.. c:en
unl1 operations IS the: key feature of process deSIgn. Thus, desIgn cannot be
accomplished merely by connectmg vanous umts and mlslalcnly thinking that If
we properly design each unit, we will obtain a proper deSign of the whole plant
Instead. we must always look at the behavior of the Iota I svstem
(3.3-2)
(3.3-3)
y- :s: 1 - 0.99 = 0.01
Y.
L= IAC::)G
We can usc the Kremser equation and the rules of thumb 10 understand the effects
or the design variables.
and
COLUMN PRESSURE. Suppose we double the tower pressure PT. From Eq. 3.3-3
we see that L decreases by a factor of 2; but since LI(mG) _ 1.4, the number of
plates required in tbe gas absorber (see Eq. 3.3-1) does not change. Lower values of
L mean that the stiU feed will be more concentrated, tbe reflux ratio will decrease,
the: vapor rate in the: still will decrease., the column diameter will decrease, the sizes
of the condenser and reboiler will decrease, and the steam and cooling-water
requirements will decrease. Thus. decreasing the solvent flow to Ihe gas absorber
will have a significant effect on the design oftbe still. but no effect on the number of
trays required in the absorber.
TIle absorber diameter will decrease: (because of the density effect and a
smaller liquid load), and a (eed gas compressor will be required to obtain the
increased pressure. Since gas compressors are the most expensive type: o( prooess
ing equipment, normally it docs not pay to IOcreasc the pressure ohhe gas absorber
with a compressor. (In some cases, a high pressure can be obtained by pumping a
liquid stream to a high pressure somewhere upstream of the absorber.)
For pure solvents, = 0, and the numerator of the nght-hand side: becomes
- A Nod Nr.>iS, 22(2t);.2 (I9lO).
(33-6)
84 SFCfIOW H nUIGI'l COWSIOUAnoWS
SEC.TlON J.' lULU or TIIU... 8 85
The rules of thumb indIcate that
L
-::I;: 1.4
mG
,nd
(3.3-7)
the Sleam and coolmg-waler loads. Calculations of lhls type are discussed m
Appendices A.2 and A.3.
Process Ahernalh'es
Thus,
(
L _ ,)('_) + 1 ::::: 40 + 1 (3.3-S)
mG .'_.
and If ,>,e apply The order-of-magmtude COlenon (I 40), we can write
(33-9)
'The denommator of the right-hand side ofthe equation. In [L/(mG, can be
wrillen as
I n addition 10 completing I he gas absorber design, we must design a condensation
procus and an adsorptIon process (as well as a membrane process) before we can
evaluate whether the acetone is recovcnng and, if so, which process should
be sdeeted Our goal here is merdy to indicate the nalure of design problems and
TO illusTrate how order-of.magmTude analyses can be used to obTain shorTcut
design procedures. These shortcut procedures can often be used to simplify the
evaluaTion step in the synThesis and analysis procedure, particularly during the
preliminary stages of design when a large number of alternatives is being screened
3.4 RULES OF THUMB
In(1 +{;)
Now, from a Taylor senes expansion. \ilie can wnte
L L
In -., In (I + ,) - - 1 04
mG mG
With these: SImplifications, and replaong In by Jog. we obtain
2.3 log{[l./(mG) - I)(Y.JJ'_>J
N'"" 0.4
Within a error. we note that 2.3/0 4 =6 and (2.3 log O.4)/OA =
simplIfied version of the Kremser equation becomes
y.
N+2=6Iog.....!!.
y_.
(3.310)
(3.3-11)
(3.3-12)
-2. Hence, a
(3.313)
Originally ruks of Ihumb weK developed by experienced designers. A designer
IIlIght have optimittd the design of 10 absorber-stripper systems and found that
the optimization always gave values close to L Co IAmG and optimum reco"eries
greater than 99% Then, when The eleventh problem was encountered, the designer
simply wrote down lhe answer Ho,>,'ever. today most rules of thumb (or design
beumtics) are de"e1oped by graduate students who run 5CX) to 1000case studies on
a compuler for a particular problem and then allempl 10 generalize the results
Of course:, the: fact That generahzatlons of computer optimization studlCS are
possibk Implies That The optImization calculations are very insensithe to changes
LD most design and cost parameters If the design and cost equations are Insensitive,
then the engineering method mdlcates that we should be able to simplify the:
equations. Thus. using the order-of-magnitude arguments to simplify problems, we
should be able to derive the rules of thumb. The advantage of a derivation of thiS
type is that the assumptions used in the analysis will clearly indicate the potential
limitations of the rule of thumb.
Now we have a design equation that we can solve without a calculator. For a
99% recovery. Eq. 3.3-13 predicts 10 trays versus the exact value of 10.1. For a
99.9% recovery, we obtain 16 trays from Eq. 3.3-13 versus 16.6 from the ell act
equation. In addition to calculating the number or plates in t.he absor?tr, we
calculate the height and diameter. These: calculations are diSCUSsed m AppendiX
A.3.
Dis.iIIation Column
To separale acetone from the solvent water, we use a distillaTion column The
acetone-water mixture is very nonideal, and we do not know of any shortcut
procedures for hiShly nonideal separations. However, the mixture is only a binary
one and so we can use II McCabc:-l1l1ele diagram to find the number of Trays. We
must calculale the still diameter. the size of the and reboiler. and
Liquid Flow Rate to Gas Absorbfrs
For isothermal, dilute gas absorbers, the Kremser equation, Eq. 3.3-1, can be used
to calculate the number of trays required for a specified recovery as a function of Lf
(mG). A plot of the Kremscrequation is shown in Fig. 3.4-1. From this graph we see
thaI if we pick L such that L/(mG) < I, we can never get close to complete recovery
of the solute even if we use an mfinlte number of plates (infinite capital cost). If the
solule is very valuable or, worse, if It IS toxic (such as HeN), certainly we will want
10 obtain very high recoveries. Thus, we can conclude that we would never choose
the liquid flow rale such that L/(mG) < I
We also see from the graph that if we choose Lf(mG) = 2, we obtain
csscntially complele recovery with ollly five plates. Remember that large solvenl
now rates correspond to dllule fec:dsto the dlslillation column-and therefore large
S6 RCTION H OF TIIV...
UCTlO!'l 10 Of" TIfVIoU 87
OPTIMUM DESIGN. No...... if we use our simplified design equation, Eq. 3,).12 we
obtain
COST MODEL It is common practice to operating costs on an annual
basIs. Thus,to examlllc economic ......e mU)1 also put the capilal cost on
an annualized basIs As discussed In Sec, 2.5, we annualize the capital coSt by
InlroduClng a capital charge: faclor (CCF) of i yr, where the CCF mclude:s all
capllal-related (depreciation, repairs and mallltenantt. etc.). A CCF of
yr corresponds to aboul a discounled-cash-f1ow ratc: of return (DCFROR);
sec Eq 2513"
Suppose we: wrile a IOlal annual cost (TAC) modd as

, ,

6
3
2
I plale
0.4 0.8 1.2 1.6 20 2.4 2.8 3.2
-L
.c
TAC (S yr) "" (C, S/moIXGy_ mol/hrX8150 hr yr)
+ [C", S'(plate'yr)(N plates) (3.4-3)
FlGURa,: 3.4-1
Ii: remKr equauol1. (/'-mm T K, SMrwooJ and R. L AbwrpllOII .1I"t f"ITIICIIOII, MeG'II.... ",'/,
Ntw YO'"k, /9J1.)
TAC - 8150C,GY,.(Y:,,') + C.(610& :i. _2)
),. )-,
(34-')
reflux rallOS, high vapor ratcs, large-diameter columns, large condensers and
reboilers, and high sleam and cooling-water demands. Hence, if we pick L such
thai L/(mG) > 2, we will obtain tin), lIlexpc:nsive absorbers, but very upc:nsive
distillation columns.
Based on these simple argumen15, we find that .....e want to choose L such that
(3.4-5)
(3.4-6)
(3.'-7)
(3.4-8)
(3.4-9)
C" .. S850/plale yr'
Gy.. = 10 mol/hr
C, = $15.5, mol
y_. 6(850) 0004
-, .. = 8150(15,5)(10) "" .
The optimum fractional loss is given by
Fraclional recovcry =
0,
)'_, "" _. 6C",
Y,n 8150C.GJ..
U we consider some typical values
.....e find that
which corresponds to
(3.4-1)
(3.4-2)
L
- _ 14
mG
which IS the: commOD rule of thumb.
L
1< --<1
mG
Of course, 4(mG) = U IS right in the middle of thiS range. Ho.....e\oer, if "'e inspect
lhe sbape of the cuncs Dear 4(mG) _ 1.5 and with high recoveries, we sec that we
might obtalD a better trade-off betwocn a decrc:asmg number of plales required in
the absorber (capnal cost) and the mCfeaslOg capilal and operating costs of the
distillation column by decreasing L lienee:, as a first guess, il soems to be
reasonable: to choose L such that
FraClional Recovery in Gas Absorbers
For a fixcd solvenl ftow rate, we can always lIlcreasc the recovery of the solute
simply by addlllg trays in the gas absorber. Hencc. tbere IS an economic Irade-off
belwocn an increasing absorber cost as we add trays versus a decreasing value of
the solule losi. One of these is a capJlal cost (the absorber), and one IS an operating
cost (the solute loss).
A IS"; DCFROR IS a bare mll1lmllm for<;onoeptWlI <JeliJgns or"ClI-lIl1dCRlood processes, I.e., a
of if2.S - 0.4 Iii mOre reallsllc:. foOl hlpm!; pro,e:c:lii, liu<;h u mbIOtechnology. a of III _ I IS not
unlusonable.
'M SPeIer. and Ii: 0 Tlmmc:rhaus.. flnll/ Des.gn "lid U:tMt"'UCSjOf CIo"ml('W ErtglftUrs, McGraw-
Hill, New Vorl. t96&, (I 389, &lve I)'plal th.1 ,ante f,om 51200 to 52100 per plale depend,n. oa
tbe column d.-merer If we kt the 00Ii1 be S2SSO per (liale and 1IX a C"CF of i )'r. we oblaJll nso/)r per
....'0
From our discussion of the design of an adiabatic absorbe;, we nole Ihat ....e
expect to obtain only a few plates (smaU absorber cost) and large liquid flows (large
still costs). which is not a desirable situation Thus. il is common practice to put
coolin. coils or one or more pump-around cooling loops on tbe bottom two or
three trays of a gas absorber. to force it to behave more as an isothermal lower.
occurs .... ben there is a plOch at the bollom of the tower when)'... _ m... x_. (see
Fig. 14-4).
From Fig. 3.4--4 we see that a "ery small increase in the sohent flow rate
above the minimum will al1o.... us to get from Y;. to y_ with a very few plates.
the upward-curving nature of the equilibrium line means that the
mmimum solvent flo.... rale will be much larger than a correspondmg case ..... here we
could maintam isothermal optration at the same inlet liquid temperature. In fact,
The minimum liquid flow rate for an adiabatic absorber
may be 10 times greater than the rule-of-Ihumb value
L"", J.4mG based on the inJctliquid temperature. (3.4--10)
This example illustrates that the indiscriminate use of rules of thumb may lead to
an inoperable design. In general.
(3.4-11 )
(e)

I Y - ntt
Y I
,
y- '
(b)
:-vr-=
y- ,
X m X X lIUl
flGUAE 3.4-3
Minimum liquid no... RIle CO.-IIIR1led mL'uurfS.
Every rule ofthumb has some limitations.
(a)
G,)"... L,..t
out
Dilute solutions: Y = y, X .. x
G. YOUl L, x...
Minimum liquid flow when X
out
is in equilibrium with Y..
flGURI: 3.4-1
MlRlmUIQ hquKlI'lo.' rile ISOIbennal
Limitalions of Rule'i of Thumb
The rules of thumb .....e de,-e1oped .....ere both based on lile Kremser equation. and
we know Ihal the Kremser equation is 'alid only for isolhcnnal, dilute s)''Slems,
both the operatmg and the eqUilibrium hnes are straIght For a system
satisfying these conditions. the minimum solvent flow rate corresponds to thc
condllion at the bottom end of the tower when tbe exit-liqUid composition (for a
fixed fractional recovery) is in cquilibnurn wilh the entering gas (so that an
infinite number of trays is required at the concentrated end of the column): see Fig_
3.4-14. The economic trade-olTs dictate that we want to operate with L 1.4mG
and a high fractional reco,ery. From Fig. 3.4-2c \1o"e see that these: results
correspond to almost a pinch zone at tbe lOp (dilute end) of the absorber.
For concentrated mixtures of solutes, the equilibrium line might be curved, as
shown in Fig.. 3.4-3. For a fixed fractional recovery, the minimum liquid now rate is
determined by Iheoperating line becoming tangent to the equilibrium curve As the
liquid flow Tate is mcreased. we expect lhal it will require more trays to get from J ...
to y_. for this case than a corresponding change for the dilute case, Fig. 3.4-2b and
c. Hence, we might expect that using a solvent flow such that L/(mG) is somewhat
greater Ihan I 4 and attempting to recover somewhat less solute tban for the dilute
case will get us closer to the optimum design conditions.
An even more dramatic difference IS encountered for adiabatic absorbers. As
the solule leaves the gas slrenm and is taken up by the solvent. the solute gives up
its heat of vaporization This heat effect causes the lemptrature of the liquid stream
to increase as it approaches lhe bollom end of the tower Increasing liquid
temperatures increase the vapor pressure of the solute, and from Eq. 3.2-4 we see
that the slope of the equilibrium line increases_ the minimum liquid flow rate
Sensitivity
The significant feature of this elementary analysis is not the relationship for the
optimum design. Eq. ] which is not exact, but rather the sensitivity of the
calculation. Eq. 3.4-8. From Eq. 3.4-8 we see that we can change any of the numbers
in the numerator or the denominator by 100':-.. and the optimum fractional
recovery onl) c:hanges from 99 2-. to 998_. Thut, the result is in$.Cnslti,e to
any of the design or cost parameters.
This simple scnsilivHy analysis clearly demonstrates thatlhere is no incentl\e
for n=fining the cost data used In the analysis. ThIS same behavior is characleristic:
of a large number of design problems; i.e, the solutions are often ,ery insensiti,e to
the physical property data. the functional fonn of the design equation. and design
parameters such as heat-transfer coefficients. cost data, etc. Thus, good engineering
judgment requires that we obtain some idea of the sensitivity of the solution before
we expend a slgmficant amount of time gathering accurate data or attempting
rigorous design calculations. That IS. we want to spend as hnk time as posslbk
gelling an answer, and we want that ans....er to have only enough accuracy to make
the decision ....e are faced with.
90 SfcJlON H SUMM,UY. EXERCISES. ANI> M.lr.tEN<.JA1Ukl
Sl("TION II SI.'MMAkY. UEkClUS, Ai'O NuMENCLATlIRE 91
F1GUME: 3.4-4
MIn,mum liquId " ...'" adiaballC
d Cooling water IS available al 90'1' from a cooling lower and musl be relurned
to the tower al 120F or less.
e, Assume a approach temperalure for Slreams cookd with cooling water
It IS Importanl 10 remember Ihat e\t:ry rult:: of Ihumb has some limilations!
Exercises
Again, il is essential 10 understand the Inleraction bel ween process unils in order to
develop a close 10 an opllmum design. Similarly, slructural changes in the
f10wsheet (cooling coils in the bollom trays of an absorber) normally have a much
greater impacl on lhe process economIcs than exact oplimization calculalions (the
optimum solvenl flow rate to an adiabatic absorber). We can Iherefore propose
another heuristic:
Avoid the use of adiabalic absorbers (unless there is only
a small temperalure rise across the absorber).
35 SUMMARY, EXERCISES, AND
NOMENCLATURE
(3.4-12)
3-5.1 If we use Ihe rcc)'c1e llowshet:t shown III Fig. 3.2-2, whal an.' the economic trade-offs
thai fiA the n:cycJe COmposition of Ihe solvcnt?
3-5.2 Consider a condensallon process for recovering acetone from an air Slream.
(a) Draw a nowshcct for a condensation process for the acetone recovery problem
(b) If the condensation process operates at 15 psia, Whlll temperalure would be
required to recover of lhe acetone?
(c) If lhe condensation process operates at lOOF, what pressure would be required to
condense 99.5 %of the acelone?
(ei) Discuss your results.
(e) Describe the economic trade-offs involved III Ihe design of a condensation process
(oolh low-tempera lUre and high-pressure).
3-5..3 Peters and Timmerhaus deri"e an CRpression for the optimum diameler of a pipe by
balancing Ihe cost of the pipe (which mcreases with the pipe diameler) againsl lhe
po....er required to dehver a specified amoulll of flUId through the pip" (which
decreases as lhe pipe dlameler increases) For pipes greater than 1-10. dlameler, they
give lhe results
M S Pelers aDd K. D. T'mmerllaus. Plant (H)'lIn and uOI'lonun for- eM"""a/ Eng",,,,,,s, 3d ed,
McGraw_lIdl Ne\\ Vorl, 1980, p. 379
\\here D, = oplimunl pipe diameter (m.). Q
f
= volumelflc flo\\' rate (fl)/S), p =
density (Ib/rt), IJ. - viscosily (cP), K = SO.055fkwh, J _ 0.35, If, = 8760 hr/yr,
E = 0.5, F - 1.4, Kf = 0.2, and X = SO.45/fl. Many mdustnal practlllOnerS use a rule
orthumb that the velocity in a pipe IS a constant, although Ihey US(' differenl values for
liquids and gases. Can you use Eq. 3.5-1 to derive a rule of thumb for pipe velocily?
Wbal are Ibe limitations ofthi5 heuristic? Thai is, for whal cases does it not apply? If
""e change lhe annual charge faclor Kf from 0_2 to 0.4, bowdoes our'eslimate change?
3-5.4. A fnend of mine in induslry tells me Ihat some of the chemists in hiS company estimate
the minimum number of trays in a disttllation column for a binary mixture by laking
the sum of the boiling points and dlvtdmg by] times their difference. Can you show
thatlhe back-of-the--envelope model IS essc:ntially equivalelllto Fenske's equation for
the minimum number of lrays? (HUll: Assume Ideal, c1oseboihng mixtures, lhe
ClaUSIus-Clapeyron equallon. Trouton's rule, and we want 10 oblain 97% purities.)
3-5.5 Several quantitative heurisucs have been proposed for deciding 011 a sequence of
dlsllllauon columns for a lernary mIxture we could recover the lightest
Summary
A number of important concepts are presented in this chapter:
I. Process alternalives
a. A large number of alternatives can be generated even for simple processes.
b. We use shortcut procedures to select the best alternative thai we will design
rigorously, providing lhat the process is profitable.
(I) We want to spend as lillie time as possible gelling an answer.
(2) We only want to include sufficienl accuracy to be able 10 make a decision.
(3) We always consider the sensitivity of our calculations.
2. Shortcut design procedures
a. It is reasonable 10 base process flows on 100% recoveries in separators and to
base equipment designs on 99.5% recoveries, al the screening stage of process
design.
b. Order-of-magnitude arguments can be used to simplify design equations.
3. Systems approach
a. You should always consider the total problem.
b. Changes in the design variables in one unit (absorber) might affect the design
of some other umt (stili), bUl nOI the unit under conSlderallon.
4. Rules of lhumb heuristics
(I, If a raw malerial is used as a solvent in a gas absorber, consider feeding the
process through the absorber.
b. It is desirable 10 recover more than 99 %of valuable components.
c. Choose Ihe solvenl flow for an isothermal, dilute gas absorber as L = 1.4mG.
o Gu. o 0" [O.ggK(1 + J)H'J ".
LOP' f P fl., (I + F)XEK
f
.. ].9Q1"po ,)
(35-1)
oomponcnl In Ihe first column and the:n srhl the remain,"! 1\000, or ... eould rKO'er
1!Ie: he:lI\-1CSI componenl first and Ihen spill the: remilmtng 1"'01 Rod and Marek" usc:
the: allenon
100
SEcnow Jl SUMM"n. UUClstS, ",..0 I'lOMfNCUTV'- 93
F
w,6!JT,267 To 1lO
ckS
90w,
---., 1----, Ac' Uc 1----
1;.. -366 _ _
P
I.\V 0.25Ix.. I 25X("
r "..("-I
...hereas Rudd. f'o\loels, lind Snrola' C:Sllmate Ihe relau'e of a dlsullauon
sc:parallon on lhe: upress,on
FIGURE J.S-I
lint r<'COv,"",
Feed Rale
OlSlIlIallOD COSI - 8o
ilin
s-Poinl Dilrelc:nce
135-3) the: lInnuahzed C()$t of lhe: Iwo uchllngns plus the: cooIlns-_ter eost minus the: \lI1uoe
of the steam J"foouc:ed
Nadgir and Liu
l
propose: a coefficienl of c:asc: of S1Cparation (CES). defined as
(15-5)
...-h,le Nath and MOlard' and Lu and Motard- prcsc:.nl a mole:. <:omplu uprcssion
based on lhe: of Iray, (proportIOnal 10 Inn II) mulupbed by comb,nahons of
flow rale factors. Can you show Ihlll 1I111hc:sc: a.prcsslons have CSSICnlllllly lbe same
depc:ndenc.e On the n:llluve \oIl1lJlllyl 0eri'"C the SImplest aprcsslon Ihlll lOU can for
Ihe vllpor rale In II brna" column. assuming Ih8t RIR.. _ I 2. and compare Ih,s result
to Rod and Marek's lesuh
J.5.4S ConSider the design of a benzene-Ioluene d,stillahon column (assume .2 2.5) for a
case I'.'here the fc:c:d rate IS 100 mol/hr. Ihe ben7.ene feed composition is 005, I'.e want
to rceo"el 99_5% of Ihe benzene, and ."e wanl the: benzene punt) to be: 099 Usc:
Smoker's cqualton 10 c.akulale lhe: number of trays 'cql,ured. and lIS$ume thai
R _ Find both lhe: VlIflO! ,ates and lhe number of t,a)"S rcqullcd In ItIt
recltfyingllnd stnpptng sectIOnS if (a) the ked stream is at 1(1) F and (b) )ou mlltle
ked stream to salur.lled-i1qUld oondlUOn$ (also c:akulate the load on the hellter).
Compare the lola) heat mput and the number of trays requIred for bolh cases.
J-5.7 A rule of thumb commonly used in desIgn is that the approach tempcr.lture In a heal
exchanger should be 10F 10 the range from ambicnllo Ihe boilmg polOl of organics
(lower values.lhat is, 3F, are used ror refrigeration conditions and higher that
is, sooF, are used for high temperatures). To eyaluale this heuristic, consider the
sunplc s)'Slem shown In Fig. 3_S--I, whe:re 1'11:: are allc:mpting 10 lcoo,er 50me heal from
a "'lI$U: stream by productn,; steam The lotallinnulil C051 of IhlS process is the sum of
Wrile the equations for the heal balances for the ellehangers, and show that Eq ).55
can be wrillen as
Sinc.e T. _ 267 and T, musl be gruter Ihan T" we simplify Ihe uprCSSlon by
assuming that (T, - 100)/(T, - 130) _ I. With Ihis applO... imation show Ihat the
opllmum value or '. IS given by
Evro though this equatIOn is rdali\"Cly simple 10 50Ive rot T
1
, suppose lhal we
allanpl to bound lhe ans....er instead Thai is, ... know that T
l
must be Icss than T
ioo
and that It musl be sreatel the T,. Uenoe, we can wrile (after we solve for T
I
- T,.
whICh IS the most .st:nSlli"e term lO Eq 3.5-7)
(B-7)
(H-6)
(3.5-8)
0- _ CAIV.. -+ C.. IVe -+ C.. +-.S..
T, - T, T, 120 JO All,
C.. /V. C.. /V,
C.. /U, c.. C, <: T
l
- T. <: C.. /U, C.. C.
-T.--- 120 -+ JO -+ A-'-'. T.. \20 -+ JO -+ A,
fC.(TI - 100) T, - 120 Fe. Tioo - T,
TAC - C.. U T DO In 100 90 -+ C.. IT In T _ T
, , , , .
Fe. FC.
-+ 30 (T, 1(0) - C. All, (T.. - T,)
(3.5-4) CES _ --.-!!-. AT = lOO( ._0 \0: _1)
F-D F-Of
V Rod.ndJ Marck,C"'fn" Cud. C"""
0 F Rood, G J Po"er$, lind 1 J Surola. P'OCtSS 5)""'''''I$, Plenlicellal!. Fnglc:",ood a,rr\.
1973, P 37
I V M NadJl' and Y A !.m, AlCltF: 29 926{I98J)
, II: Nalh and II: L MOIIld. MholullOlllllY S)'nthrsos 0( Sepalalion Pr()CCSSle5.- AIOF MeellOA,
Ph,laddplua. 1978
I M D 1.lIand II: L MOl81d, /'UI CJw.oI. F..., S,., Mr, !"O 74
Clilcullile lhese bounds for. case: ... ha-e C.. -SIlJl/y". F-5I,100Iblhr.
C."" 1.0 Btu (lbcF). U,_JOBlu/(hr ftJF). U._20Bluj(hrfI
J
oF). C.=
JO.07J881((lb/hr) yrJ. MI, - 933.7 Blullb. T. _ 26TF, T... _ ]66F, and C."
S2122/1.{lb/h') yr]
Undelwood's equatIon for the mllllmum 'e!lux rallo for mullicomponc:nl mIxtures
no{mally requilcs a Irtal-and-error solution. Thai is, for an AB/(CD) split. we first
soh'e the cqulliion below for lhr: value of 8 be:twccn and 1
(3_S--9)
(3 S-12)
(35-10)
94 St,CTION n $UW"AlY, UEiCISEs. AND NO.. ENCUruIE
and then we use result to alcuJale R.. (assunlln! a sharp AB/(CD) split and that
no D goes overhead):
II.C.I:,U' ClteX.D .l:
CII
---+ +--".R +1
11...::-9 11.:-9 1-9 ..
It IS Idvantageous to avoid trialand-error calculations when we are allemptlllg 10
determllle the best separallon sequence bec.:ause the number 01 wJumns thai need to
be designed IIlCfU.ses r"pldl) ;IS lhe number of componenlS mcreues (the column-
sequencmg problem 1$ dlSCtlssed III Chap 7). Malone sugge:ned a procedure for
boundlllg the values 018 In Eq 35-9 He firsl re'Antes the equation U
IIJCX., Xu II
AC
X,." II
I1C
X.,
--+ -- *' I - q- - --- (1S-11)
11..:-8 1-6 1I,fC-8 lI
oc
-6
511lt:e the des.rred value 016 mUSI be IJl Ihe range ofirK <: 6 <: I, be limSUb$ltlutes 11
K
for 8 00 lhe nyll-hlnd sick of Eq 35-11 and lben solves the remiialll' quadrauc
equallOllo. lie repean 1hJ.s procedure wllh 9 = 1 sub$tilutal 00 the n&hl-haod sKie
The actual value of 8 must thus be betwce:n these bounds.
Consader a case where X
A
, - Xu = xc, = xDf" =0.25,lI..c = 4,lI.- - 2. lip(" ...
05, and q - I, and find the bounds Oft 8 For a case where F .. 100, all the A and
99.S" of lbe B are recovered ol'erhead, and all the D and 993 of the Care
recovered in lbe bonoms, lind the bounds on R...
3-5..9, In lhe ISOpwpanollo-aoetone proccu (a smgle, IrreversIble reacuon) sbown III FII-
1.3-1, lhere 'Atli be a large reliCIOr at hIgh bill tbe recyde tIow will be
small (and lherefore lhe: recyde cosu will be small). The: opposne SlIuatioo will hold
for low conVCl"SIons, 50 INit lhere muSI be an optimum co",,'erslon. Suppose that an
esumale of the costs kads 10 lbe nprcsslon
7]) ( I )
TAC"" J082 + t5t
X I - x
when: lbe term 7J3/. corresponds 10 tbe rec;:yde flow, 3082 eorresponds 10
elfecl of Ibe recycle now on the <:om of the equipmenl in the recyde loop, and 151
In [1/(1 - x corresponds 10 lhe ra(:lor cosL Plot rec;:yde COSl,lbe ractor COSt,
and the 10lal Inoual COSI versus cooverslOn. Can you propose: I design heunsllc for
Ihe optimum conversion for lirsl-order, Ic:versible reactions? For wblt types of
kinetIC models ....ould you peet thai your heuristic would not be vaJid?
Nomencllliure
A, Area of water cooler (ft
1
)
A. Arca of Sleam exchanger (ft J)
C.. Annualized COSt of heat ellchanger [S/(ft
1
. yr)]
CES Coefficient of ease of separation
C
II
Annualized cost per plate [S/(platc: yr)]
C, Heat capacity (Dtu/(mol' oF) or Dtu/(Ib. OF)]
M F. MaIORe_ u{C!lertuClll of MaaachwcIl1. penonal wmmun-
""'UOII
C.
C,
C.
D
D,.
EP
F
G
L
m
N
P,
p;
q
Q
Q,
R.
T
TAC
T.
T,
T,
U.
u,
Iv,
Iv,
,
,
,
8
p
".
MI,
OT
6V
1
SIiC1ION 11 SUr.l .... lly. EXHCISf.S. "1'oD NOMENC"I_HlJU 9S
Cost ofsteam (S/Ib/hr) yr]}
Solutc value (S/mol)
Cost of coolmg water (Sj[(lbjhr) - yr]J
Dlslillate flow
OptmlUm ptpe diameter
EconomiC potential
Flow rate (mol/IIC in Eq 3 S-2)
Gas rate (molthr)
Solvent rale (moljhr)
Slope of equilibrium hne (sec Eq ],2-4)
Number of plates
Operaling pressure; must have same: unil as po (atm. psia. mmHg)
Vapor pressure of SOIUIC, must have lhe same units as P
r
(atm. psla.
mmHg)
Feed qualilY
Heal load (Blujhr)
Volumelnc flow rau: (fills)
Minimum reOux ralio
Temperature C'F)
TOlal annual COSI (SJyr)
JnJel lemperalure ("F)
Steam lemperalure (oF)
Temperature of Intermedlale stream (of)
Overall heat-transfer coeffiocnl for the waler ooolant
[8Iu/(hr. fl J OF)]
Overall heaHraosfer coeffiCIent for lhe steam exchanger
[Btu/(hr. It 1. 8F)]
Coohng waler flow rale (Ibfhr)
Steam Bow rale (Ibfhr)
Mole fraction in liquid phase
Mole fraction of oompooenl i in distillate
Mole fraction of component i in feed
Mole fraction in gas phase
Relative volalility
ActivilY coefficient
Small value
Root of Underwood's equallon
Density ""' Ib/II
J
Viscosity (cP)
Heat 01 vaponallon of steam (Btuflb)
Boiling-point dtffercnce:
DIfference 10 vapor rate
Conversion
PART
II
DEVELOPING
A CONCEPTUAL
DESIGN
AND FINDING
THE BEST
FLOWSHEET
97
CHAPTER
4
INPUT
INFORMATION
AND BATCH
VERSUS
CONTINUOUS
As \lot mentioned earlier, the original definition of a design problem is underde-
fined. Not only must .....e develop designs based on ne..... reaction schemes invented
by our own company. but also we must design all our competitor's new processes,
to ensure that Ollf company's technology will remain competitive. Thus, we often
must design a process based on a minimum amount of information.
4.1 INPUT INFORMATION
The information that is normally available at the initial stages of a design problem
(or that must be gathered) is given in Table 4.1-1. We briefly discuss each.
Reaction Information
The reaction information we need to know is listed in Table 4.1-2. Often it is
possible to gather much of this information rrom the patent literature. In
particular, the primary reactions, the temperature and pressure ranges, the catalyst,
and the maximum yield often are available. It is essential to worl.:: closely with a
chemist 10 gather the remainder of the information or atleabt to make a best guess
of the missing data
..
100 SFCTIO"" '* 11'l1'l1! '''''OUlAllOt-
UcnoN" Il'll'lJl 101
HGUR[ 4.1_1
Batdt composition pl061cs
Time
-+- CatCao; -+- CbtCao; __ Cc/Cao
and .... e refcr to lhe con.erslon or A 10 C as a sckcllVlty loss. A dIagrammatic sketch
of our defiflltion of scleclI\lIy for the- !-iDA process (see Sec. 1..2) is sho.....n in Fig.
41-2 (Nole thai a of definItIOns of selCCllvil) and yield arc used in the
IJterature. so lhal II IS ailloa)s nec.essary 10 check the' ddiflltlon)
NormalJ). ,a"-malenals COSIS and selecl1Vll)' losses are the dominant factors
1O the dOlgn of a petrochenncal I"rocess, Raw-materials costs are usually in the
(4.1-1)
S = Mok:s of B Produced
Moles of A Con'erled
p
I'.
V-
I/
1/
1/
/
Q
10 il slOgle predefined cond'l,on Ihill IS used to ehec!.: the reproducibi_
hty of the catalyst. hut Ihat far removed from The Opi101Ulll economic operatlOg
conditions.)
Numerous plOCCSsrS Illne been desIgned 10 operate al the cond,tlOn of
maximum yield. bUI this operallon onen dOC'i nOI correspond to lhe optimum
economic conversion As an example. k:1 us consider a simple:. hypothetical
react,on system A .... lJ C. where 8 's Ihe product and C has only fuel
\alue The of A. 8. and C t,me 10 a balch rcactor or sracc
lime 10 a tubular reactOf arc shown III fig 411 When Blakes on its ma:llmum
conc.enlrallon. polOl P 10 rig 4 I-I, a cons,derable amounl of undesIred C '5
formed and a large amounl of A IS converled !-iollle..'er. if wc consider operaling al
poml Q on Ihe d,agram. only a small amounl of C IS fonned and much Jess A IS
oon\er1ed. Thus. If .....e opera Ie al poml Q. we will require larger recycle Rows (and
higher recycle casa), but we lose less of our ell penSIve raw malerial A Ihal is
converled to the b) product C .... hich has (lnly fuel value
We define St'1'rlll'II, S as Ihe fnlcHon of the reactanl converted Ihal cOOs up
producI
1
0.'
0.8
" 0.7
S 0.6
8. 0.5
E 0.4
8 0.3
0.2
0.1
o
o 5 10 15 20 25 30 35 40 45 50 60 65 70 75 80 85 90 95 100
SIDE R[AnIO-':S. It IS 10 write do"n any Side reaCllons
Ihat mtght take place.. E\en tf 001)' a Irace amount or a by-product is produttd in a
laboratory ellpenment. Ihis bY'product may build up 10 \'ery large 1e\'C1s in a
recycle loop all lhe by-products produced musl be known io order 10
s)'nthcslZe a separallon system. O.-erloolong side reactions has bet:n a common
mislake. and it almosl al\\ays leads to paying large economic penalues
TABU' 4.1_1
Input informalion
I, l"ht reacuons and reaCllon
1. ".., delued production Ule
). The daorcd prodUCI pUrIly. Of _ mformatwn aboUI proce purlly
-4. l"ht raw matenalJ and or JOmc Iniormillon aboll' pro"" '''''"''US runly
Inf...rmlhon aboul I .... fatt 0( I .... ,ean,,'n Ind the rate 0( dracl1U'Kln
6. All) pr<>oc%UlnJ.
7. Od.er plan. and Slle dna
a. Physocal of III componmts
t. Informa\lon oonoelmna the 1.IIk!)'. .nd en'"onmmtll ImpacT of the
rnllCIUIs ,n tM prootS$
10. Cosl data lor b),prodlllC/5. eqlltpmcrll.. and uuhtleS
I'ot,4.XI\IUM nELl). Informahon cona:rning how the product dlstnbution
changes wlth COn\er$lon and. or reactor temperature. molar rallO of reaClaot$. elc..
IS often dllTlCult to obtam A chemISt's focus IS on scoullng different calal)'Sls.
anemplJng to define a mechalltsm. and looking for ways to wrile a broad patent
claim, Dunng lhese .scoul1ng ellpc:dlllOns the normally allempls to find
the reaclion conditions that malllmll;C the yield Thus. ellperience indicatcs thai
the eJJS(lOg data base Will have mOSI of the points grouped in a small range of
conversions close to Ihe mallimum yield. (Often the largest amounl of dala will
TABLE <4.1-2
Reaction informJltion
t, l1tt slo!CIuomct., 01 aU reaC:ItOnl tak... p1ac:c
1. l1tt ranll: 0( temperllules and 'or the rclclions
1. l1tt phasql) of lhe rn"lOn Iysltm
-4. Some: mlOrmlllOn on lhe product d'llnbullon c:onVC!'IIon (and flOJSlbly rcaClOr tempe'"lme,
molar rallo of rclCllInl$. lind/or pressure)
S. Some: ,"(Otm.tlOn aboul C'UuJ Sp"CC Or residence IUne
6. If. call1ni .1 USI'd. lOme: abou' tM Jtale of 1M calalyst (homogcneou... IlullY. p"cktd
heel. powdet. etc:), _O'TIC In'OIm_IIOft abotJl the deactIvation Ind JOme ,du 0' Iht rtgcnalhlhly
of 1M OJ ",dl,Q tht method of (ooh burn.lOlvent w:nh, etc.)
L--
/
- -
/
(a) Rt:a{:lor
(I x) mol
Toluene
1=1
(I mol)
x mol
Toluene
unre3Clw
Toluene
reacted
Rel,;)'Cle
Benzene
produced
Dlpheoyl
produced
1,800.000
1,600,000
1,400,000
1.200,000
.. 1,000,lXXJ

8OO,lXXJ
600,000
400,000
200.000
o
o I
nGURE t-3
Prolil
02 0_3 04 OS 06
Conversion x
0_7 0-8 09
bull of the data which are talco in the neighborhood of the maximum Yield. Agam,
establishing a close relationship with a chemist and providing him or her with
feedback about the optimum processing conditions early in the experimental
program WIll lead to more profitable processes.
Unfonunately, most companies are not organized to operate in this manner
Instead, the chemist's apparatus has been completely dIsmantled, and the chemist
has been assIgned to another project before a design engmttr recei"es the: problem
10 thiS situallon, the designer should altempt to estimate the economic Incentive for
determimng the economi{: optimum conversIon, and therefore for domg some
adduional e.xpenments, rather than just designmg a process 10 operate at maxi-
mum YIeld
henyl =
I - S)x
Benzene = .ll-
lew Toluene - I - x
mol
I mol
Benezenc = Sx f--
Reaetor Separalor
f--
Dlphenyl ==
J(I - S)x
D,p

Toluene recycle = I - x
x
(b) Plant
Toluene
Toluene
FlGURE ....1-2
Sda::uvll)'
range from 3S to 85 % of Ihe lotal produci COSL The opllmum economic
conversion is nonnally fixed by an economic trade-off between large: selectivilY
losses and large reactor costs at high comersions balanced against large recycle
costs allow conversions. Hence, the optimum economic conversion is less than the
conversion corresponding to the: maximum yield, as shown m Fig. 4.1-3.
Often the scouting experimenls performed by a chemiSI will conlain more
informalion about the effeCI of conversion on Ihe product distnbulion than Ihe
E. L. Glumcr. "Selling Pnct n Raw Cost," Clwm 79 (9) 190 (Apn124, 1967).
Catalyst Deactivalioo
Another Pleoe of design dala Ihat orten is lacking al lhe early slages of a design is
the catalySl deaClivation rale. The chemisCs efforts are focused on finding a more
aClive or selective calalyst, so that numerous shorHlme runs with a variety of
catalysts are conSidered. Some catalysts have an operaung life of I or 2 yr before
regeneration or replacemenl IS required, so obviously a lime-consummg expen-
men! IS required 10 find tbe deactivalion rate
In mitlal designs we expect that there may be large uncertamtles m some of
tbe design data. Thus, we examine the sensitivity of the total product cost to these
uncertainties, and we use these results to help guide the experimental development
program in the directIon of the highest potential profitabilily. We use shortcut
lechniques for these imtial design calculations. because we recogmze that it will be
necessary to repeat the calculations as more mfonnatlon becomes avaIlable
... , ~ ... , ,,, ..,,, .,,'V.... ~ I O N
Production Rate
If Voe are to desIgn a new plant to meet an expanding market condition. as a first
guess of the production rate Voe consider the lar:;est plant that has ever been built
With thiS approach Voe obtam the greatest economy of scak. (Nonnally, things arc
cheaper per unit If VoC buy tMm m large quantities.)
The maximum size of a rlant is usually fixed by the maximum size of one or
morc pieces of equipment. Often this maximum size is fixed by restrictions on
shipping the equIpment to the plant silc. Thai is, only a certain size compressor, or
whate\'cr, will fit on a railroad flatcar or truck.
Wc also consldC'r the possibility of exceeding the maximum size of an existlll!
plant. ThiS approach almost always requires the development of new technolog)
and thus has a hIgher risk. !-Iovoe....cr, by gaining a larger economy of scalc. it might
be possible to reduce the product price enough to gain a greater share of the market
and thereby justify the additional risk. However, if a project gelS 10 be too big, ncw
types of management probkms might lead to a SIgnificant increase m COSls..
Ofcourse. Voe must also consKier how oursbare of the market might change if
we bUIld a new plant It makes a great deal of difference whether our company has
50% of the market and $Orne of the largest existing plants versus whether we havC'
S";. of the market and our ellistmg plant has onJy one-tenth of the capacity of the
largest plant.
The production rate specified for the plant might cbange (il usually does)
during a design Market conditions arc constantly changing, and we must be
respoosl\e to thcsc cbanges. By using sbortcut methods for our preliminar}'
designs.. Voe milllmue thc cffort requircd to change all thc calculations,
Product Purity
The producl purit) normally IS also filled by markcting considC'rations. In fact, it
might be possible to produce a range of product purities at different prices. Again.
we must expect tbat the product price \1:TSUS purity predictions might changc as the
design is being developed. It is also csscntialthat a designer infonn the marketing
department about the very high costs thai may be associated with producing some
high-purity products early in the development of a new process., so that the
marketing dC'partment does not raise customer expcctations to unrealistic levels.
Raw Materials
Achemist normally uses '''ery pure chemical reagents in laboratory studies, whereas
natural or purchased raw materials always contam some impurities. Hence, we
need to gather somc mfonnall0n from our marketing group about raw-material
price versus purity in order to decide whether to include a purification facility as
part ohhe design project Moreover, we must work .....ith the chemist 10 see whether
the impurities in the raw materials are inert or will affttt the reactions. We must
Ud"lON .. I 11'<P\Tf I,.roU,UTl(ll'< 105
also ellamine their effect on the separallon system. In panicular, trace amouna of
impurities can build up to large values In recycle loops unless the Impurities are
removed from the process.
Constraints
For thc sake of safety .... e nonnally want to avoid processing conditions thai are
Within the explosive limits of a milliure. (fhere is an unwritten law of nature that a
spark will always find an explosi\e mixture and cause an ignition.) There have been
a disturblllgly large number of serious plant explosions throughout the history of
tbe chemical industry, and we want to avoid another occurrenoc. Howe.. er,
phthalic anhydride plants, which oxidize zylene with air, do operate withlll Ihe
explosive range, although there are many special safely features.
Similarly. we have to know the processing conditions where some of our
malerials mighl polymerize and foul hcat-cxchange surfaces, or where they might
become unstabk and rapidly decompose. Numerous malcnals also form coke and
deactivate catalysts, so we need to know how to minimize this coke formation
Very toxic or highly corrosive materials also affect the way in which we approach a
design problem. Each of these: factors may impose constraints on the design
procedure.
Other Plant and Site Data
lfVoc are p:oing to build our ne.... prooess on an existing site, lhen Voe must design the
process to be compatIble Voith the facihtlCS that already exiSI on that sitc The
battery-limit conditions and costs that we need to know about are given in Table
4.1-3.
Pb}'ial Property Oatil
Conocptual designs oflen focus on attempts to make new materials, so that in many
cases physical property data are not available in the literature- An excellent
TABLE")
Planl ud site data
L UtililJc:t
.. Fuel AlPPy
b Levds of Iteam prasun:
~ CoohDI-watn ,nlet and ootkt tempe",I"'''''
11 Rriril"ntion "'.ds
e Elcdnc power
1. Wasle dlSpoW raalitaes
106 SECTION 41 INPUT INtOMM"nO'"
SECTION <2 LEVEL I OEC1510N ....TCII YElSI'S ("OmlNUOU5 107
I. MUCllOlllllformallOli
..
b RcactlOli 1lI1c1 tcmpc:ralure > tl WC (10 gCI a rCa50nllblc reaCllon ralcl, rC:lctor prCliSlIre = 500
p".
TABLE
Inpul in(!!rmalion for Ihe hydrodealliylalioll of loluene 10 produce benzene
collection of techniques for physical is .tv!ulabh:. New
estimation procedures based on group contribution are an area of aClive
research.
The mformatlon we nomlaJly need are molecular weights. boiling pomts,
vapor pressures, heal capacHleS. healS of vaporization, heats of reaction, liqUId
densities, and fugacity coeffiCients (or equalions of stale)
For conceptual designs. we whate\er information is available (which in
some cases is a guess), and lhen .....e eSllmate lhe sensitivity of the total processing
COSIS 10 these values. This sensitivity evaluation provides a measure of the
economic incentive for making the appropriate measuremenlS. In many instances
Ihe costs are surprisingly insellsitive to the physical property values, but in some
cases they are eXlremely sensitive. The use of shortcut design procedures signifi-
cantly simplifies Ihe sensitivity analysis.
,.
Toluelle + II, 1k1l;Ulle +ClI,
"" Dlphcllyl + II,
Mok:s &mClIC al MeaclOr Omlcl
Sc:lcctlvllY _ .. S
Mob Toluene Convcned
Moles TollKnc Collvcrted 'II Meactor
ConverSion _ _ x
Moles Totuene Fed 10 knetor
(41-3)
Cost Data
R C. RCld, J M PraWimu.,and T K. Sberwood. l'M and L.iquids, 3d McGraw
Hill New Yorlr, t977.
The capital costs of some pieces of equipment are given in Appendix E.2, and some
operating costs are given in AppendIX E.!.
EXllmplt 4.1-1. To Illustrate our deSIgn procedures, we conSider a proocss for
producing benzene from toluene. We develop the design from scratch and discuss
numerous proocss alternatives. The input In[ormation taken from Sec. 1.2 is given in
Table 4.1-4
(41_4) x -< 097
0.0036
S - I - -,-:::Ciio..
(i x)''''
Continuous processes are designed so that every unil will operate 24 hr/day. 7
days/wI: ror close to a year at almost constant conditions berare the plant is shut
down for maintenance. Of course. in some cases equipment or other
reasons. cause unexpected shutdowns. In contrast, batch processes normally
contain several units (in some cases all the units) thai are designed to be started and
stopped frequenlly. During a nonnal batch opera ling cycle. Ihe various units are
filled with material. perform Iheir desired function for a specified period, are shul
down and drained, and are cleaned before lhe cycle is repeated.
Many batch processes contain one Or mOre units that operatc continuously.
For example, in numerous cases the products obtained from several batch reactors
are temporarily stored, and then the producls rrom this intermediate storage tank
are fed to a Irain of distillation columns that are operated continuously. Similarly,
there are cases where a variety of by-products that are produced in small amounlS
arc accumulated conlinuously, and Ihen when a sufficienl amount is available. a
batch still is used 10 separate Ihe products.
d. Gll!> phase
No "",lal)"S1
2. Production rate of benzene: 165 n,ol/br
3. ProdUCl pumy of bcllZClK. x
D
09997
4. lIa.. rn:llen:lls. Pure tolucnc at amblCnt colldlllOIlS, II, '\rcam COlitall\lllg 95% H,. S% CIl, at SSO
ps.... lCOr
s. COllSltolllrs' 1l../aromauc <!o Sal tbe reactor mlcl (10 pre.ert! coLmg); rcaCIOI ollliet leml"'ralure -<
1300 f (10 prCYenl rapIdly quench rcactOl cffilKnt to 115O""F (to prcyenl colin&).
X -< 097 (01 tbe product dlSlrlbulion cornl:ltion
6. O'bel pl:lllt and sue d:lta to be gi"<:11 laler
4.2 LEVEL-l DECISION: BATCH
VERSUS CONTINUOUS
(4.1-2)
Summary
As a general stalement about lhe input information, we can say that
You never have tne right information!
Some important dala will be nmsmg or will be taken in too narrow a range at the
wrong lemperalure and pressure. A chemIst's recipe for producing and isolating the
product might be available, but nOI all the side reactions may be known. The
chemist might have used a solvent for each reaction slep without ever
considering separation costs.
You should never hesitate to ask the chemist for more infonnalion; a close
working relationship is essential to developing good designs. And try 10 use
preliminary design calculations to help guide the experimental parts of the
development program. In particular, we want to determine how sensilive Ihedesign
is 10 physical property data, coking limits or other process constraints, cost factors.
purity specifications, elc.
108 SU IJON 02 lfVfl. I l)fOSION IlAH'!l VUSUS CONTINUOt$
SF.CnON oJ lHfL I OF.OSIO" BATCH VEII;Sll$ CONTINUOUS 109
F=l
Huo Reactor
Scparalor
'--
I I
Multiple Operalions in a Single Vessel
Another fea.ture processes is that It is often possible to accomplish
several operations In a Single batch vessel, whereas an individual vcssel would be
for each of the operations in a continuous plant. For example, suppose that
In a continuous plant we heat a reactant before sending it to a reactor where a
catalyst IS added. and then ..... e send the product mixlure to a distillation column
(see Fig. 4.2-10) In a batch process. we mIght be able to use the reboiler of a batch
still for the heating and reactIon steps as .... ell as for the separation. sa:: Fig. 4.2-lb
Thus, a single pIC of equipment can be used to replace three pies of equIpment
When .....e cany out multiple operations in a Single vcssel. normally the vessel
must be larger than the SIZe required If we had used separate vCS5els for each
operation. For example. If ..... e use one vessel for each operation (heating. reaction
and and it takes I hr for each step, we: can produce P Ib/hr of
by shlftmg the batches from one unit 10 another However, if we use only a Single
The dlStlllcticlO bel .... een batch and conllnuous processes is sometllne<l
"fU72Y M That is. large, continuous plana that exhibit catalyst deactlva
tion may be shut down every year or so, to regenerate or replace the catalyst
Similarly, a large, continuous plant may IOciude a single adsorption unit, which
usuall}' is a batch operatIOn If there are only one or two batch operations in a plant
with large production ratcs that otherwise operates continuously, we normally
refer to the plant as a continuous process.
Guidrlines for Selting Balch Processes
There are some rough gUidelines Ihat help 10 indicate when a balch process may be
favored over 3 conllllUOUS process. These are reviewed DOW.
PROOllCTIO"'l RAT[.';. Plants having a capacity of greater Ihan 10 x 1O'lbiyr
are usually t:onlllluous. whereas planls having a capacity of ItsS than I x 10' Ib, yr
are nonnall}' balch Iypes Large+C3pacity plants can justify a more thorough
de\ dopmenl program, Ie. a more accurate data base, as well as larger engllleenng
desIgn costs In contrast. batch plants are usually simpler and more fleXible. so Ihal
a S3tlsfactor}' product can be produttd with a larger uncertamty in the design
Also. because of Ihelr greater flexibility. batch plants are most common when a
large number of products are produced in essentially the same processing equIp-
ment (e.g. pallltS)
MARKET fORCES. Many products are seasonal; for example. fertlhzer is sold for
onl}' about 3 month III the early spnng. If the fertilizer is produced o\'er the
complele year. then large inv-entory costs are incurred in storing II for the smgle
month ..... hen sales are generated However, if the fertilizer could be produced 10 a
month or so in a batch planl and Ihe plant could be used to mue other products
during the remamder of lhe year, lhen the inventory costs could be dramatically
reduced. lienee. batch plants often are preferred for products with a seasonal
demand.
It requires about 3 yr to build a continuous proce:ss., bUI some products only
have an average lifetime of 2 yr (some organic pigments). The grealer flexibility of
batch plants makes them preferable for products wilh a shorl lifetime
OPERATIONAL I)ROBLf.MS. Some reactions are so slow Ihal balch reactors are
lhe only reasonable allernative. Also, illS very difficult 10 pump slurries at low flow
rates wilhout the solid settling oul of Ihe suspension and plugging the equipment
Thus. II very difficult to build continuous processes when a low capacity of
slurries must be handled, Similarly. some materials foul equipment so rapidly that
Ihe equipment must he shut down and cleaned al frequenl intervals Balch
operalion turns out 10 be ideal for handling malerials of this type, because The
equipment is periodically Slarted and SlOPped, and normally iT is clcaned afler each
balch has been proccssed
(a) ContinlJ()Us
(b) Batch
(I) Feed
(2) Calalysl
FIGURE 0-1
ConlJnllOUS "CISUS balch
Catal}'St
Product
=
-
'"
lleat
I-
Reactor
Separalor I-
Heal
Product
110 LHEIIDt:OSION IHH.HYUSUSt:l.>I'flTNUOVS
\essel, lhen It tales J hr to produce any product, and to oblain the same
productIOn rdle we need to J Ilmej, d$ much m.Henal Of course, when
use larger vessels, v.e often oblam an economy of scale, so lhal of!en there is a
significanl economic incenlive 10 merge processing steps inlo a sing.le vessel.
Design of Batch COnlinuous l)roces!>eS
To develop a conceptual desIgn for a continuous process, we mUSI do the followmg
I. Select the process unils needed
2. Choose the interconnections among these units
3. Identify the process allemames thaI need to be considered_
4. lisl the dommanl desIgn variables.
S. Estimate the opllmum processmg condlIIons.
6. Dctermlllc the besl process ahernatLve
For a batch process we muSI make exactly the same decisions However, we must
also make the follOWing decisions:
7. Which unns m (he ftOlolosheet should be balch and which should beconlinuous?
8. What processing steps should be earned out In a single vessel \ersus USlOg
lIIdlvidual for each processing slep?
9. When IS it ad\anlageous to parallel batch untts 10 Improve the scheduling
of the planl?
10. How much iotennedlate storage is required, and where should II be located?
A S}'stematic Procedure for lhe Design of B.atch
Processes
Clearly it IS necessary to make more decisions 10 design a batch process than to
design a continuous process. For Ihis reason, Malone and coworkers suggestlhal
the best approach to design a batch process IS 10 design a oonllnuous process first
With thIS approach II tS SImpler 10 screen PJoce!>S ahernau\cs and to determmc the
best process Once the best structure of the flollo'Sheet has been deter-
mmed, they SUggesl rollowlRg the systemallc procedure gl\'en In Sec, 13.2 to
develop the best design for a dedicated balch plant.
0 M F Mltonc. A SyslCrt1311C P.-occdllre for Ball:b p,ootSlo S)'lllhtsli,- paper
ptescnlcd al lbe 19U AnnUli AICbE MllnJ,. ClucaIO. 19 5. C M Mynalbros -f1c;ubtlI11 aDd
Tlf!,=Ui 10, Ibleh P.-neas Dcsrp - MS n.cs.... Un,.." ..I)' cI Mluae.:huwu.f" Amhcnr. t916.
4.3 SUMMARY, EXERCISES, A 'D
NO,\1ENCLATURE
Summlilry
The mpul information that we need 10 undertake a deSIgn problem includes
I. The reactJons and reaction conditions. including a correlation for lhe product
dlstnbullOQ, a relationshIp for Ihe conversIon and space velOCIty, and Informa-
lion about catalyst type, dcactivauon rate, and regeneration
2. The deSIred production rate, product purity, and value of the product
3. The raw materials available and their costs
4. Any processing constraints
S. Other plant and site data
6. Physical properties and information about the chemicals Involved
7. Cost data
correct data are nOl available or arc uncertam, However, we do
the best thaI can, and we evaluate the senSIII\lIy of shortcul dcslgns 10
changes lR uncrrtam factors. Estimates of this type can be used to deternune the
economic mOCDli\'cs for undertaking addHional expenmenls. The data on lhe
product dlsmbution and side reactions are usually cnhcalto a good dcslgn_
Lev-el-I Decision Batch versus Conunuous
The faclors lhal favor balch operation are
I. Production nile
Q. Sometimes batch ifless than 10 x lti Ib/yr
b Usually balch if less than I )( IW Iblyr
c_ M uhlproduct plants
2. Market forces
Q Seasonal production
b. Short product lifetime
3. Scale-up problems
Q. Very long reaction times
b Handhng slurnes at low flow nltes
c Rapidly fouling materials
Exercises
4.3-1. Selecl a process from Hydrocarbo" EPIl:yclopedia of Cnemll:ol
and by J. J McKell.. or !he Eru:ychJpcdla T#ttwlog)'
b) Kirk Olhmer. and sec. bow many of Ihe mput dala Ihal )'Otl can find
4.3-1. The 1967 AIChE Student eoolat Problem data sbo....1DI bow the sdecU'IIlY
(5 - mala ofbenzcne al per mole on lbc
SEcnON q St!MMAlY, UUcrsa. AND NONDlCUTUIIE 113
F"u- the 1961 AlOE s..-.., Con'_ I"foblcm
TARLE .t.J-1
Sdcdi,ity dltl fot HDA pt"0SS
Cl ", - elM. + ",
C
,
". - t C1H. + CH.
....
39J
'" 236
c__
\-otW ,.11_ ...-1
H, >00 W >os )$, ..,
CH,
1.28 lOS W U.
".
C,Il,
'89
JH
'"
'"
'"
C,H
o
'"
601
'"
." J.l3
aod some for the product distribulion are given in Table: 43-3. Con,c" the
data I'.-'Crght perc:enl 10 mol percent, and tben develop a correrluon for the
xleccnouy (moles or CJHo at tbe reaclor nil per mole: or C,H, conl'Cned).
and D)"tKbrl' ptCSCnt some product distribulion data for styrene rroduc-
1100. The reactions they consider are
Elhylbenze-ne;Slyrene + H,
Ethyrbcnzene - BenlJ:oc + CJI-I.
Eth)lbenlJ:ne -t 11
1
- Toluene + CII.
and points tead from their graphs are given in Tables 43-4 and 43.S [)e,elop
correlations for these: data
4 :J Consider t_o parallel fiul-order isolherm.' n:&CUOns in a batch (or lubular) reaClor
fed _'lib purc rcactlnl
A - ProduCl A ..... Waste:
and define 5C1ec1l'1I) IS S - mol produCl}mol A convened. Use a kinelIC anaJYSlS 10
delemune. the selectl... l) dcpendt on lbe con''CfSlon. What are the rC'llults If the
first rcacllon IS first-ordet and the second reactlOo is second-order?
Contider two. conSttUlIve, firsl-order, iSOlhennal reae:tiont in a balch (or tubular)
reactor fed wIlh pure reactanl:
TABU: .t.J-J
Product distribulioo for cfhue crlck.irtg
4-3-7.
80T. I aIm
ACl'."tone - Ketene + CII. 7OO"C I atm
Ketene - CO ... iC,II. 7OO
0
C. I atm
Ketene + Acellc Acid _ Acetic Anhydride
and lhe selectivity It defined IlS S _ mol ketene al reactor exit/mol lcetone:
con.erted [)e,elop a correlation for Ihe data. Compare your results to the SlTOple
COrrelattOD S _ I - lx Usc you..- results 10 esllmate tbe conlerslon corresponding
10 lhe maximum }idd
4A A simpLified of I process to produce eth)knc: via ethane aadanl has been
prC$(nled by BoIIc:s..' The reactions of interest are
reaClor conversIOn (Sec T.blc 4 3-1). Plot the data on paper, lind make a
101! log plot of I - S versus I - x. Develop correlations fOI both 5C1S ofdata Why is
,t benel 10 conel.te I - S versus I - x? Also, usc tbe correlatIon gtven by Eq 4 1-4
10 cakulile the yidd of benune (Y - mol benune It relelor nil/rnol of loluene fed
to reactor _ SoT) as I functIOn orconversion. Estinu.te lbe COD_erslon colT'CSpondmg
to lhe muimum yield.
4..J-J. Selectlvily data Jor prooe:ss 10 pl"OdlKZ IIlXhc anbydric:le from KlClone and acclic
acid In: grven in lbe t9S8 AlOE Student Conlesl Problem' The dala are gr'en ID
Table 43-2. The reactions of interest are
s
..,
.."
0977 0.91
09'

.,
06 Q7 0.1.5 0.15
TABU .t.J-l
Selectirity d.t. for aceric ubydridc process
s 0..
."
62 ... OJS Q1lI
."
QIJ

..
.,
OJ

OJ

Q7 O.
A - Product ProduCl ..... Waste
Define tbe: as S - mol product in reactor effluent/mol A converted, and
de,'Clop an ClIpresslOn, based on a kinetic analysis. for ho"" the selectivilydepends on
R.. W. and E c. Dybdal Clom.. bog ,"<4) 2U (1948).

of benune per molt of slynne 'crsus (onvcrsion
'See J, J McKell .. erw:ycwptdla P,oa,.lr\fI <mtl Desig.... yol I, DeHer. New 1976,
p.211
I W L 8ol\el., ,Ioyk"" Pillltf Duigtt MtJl &.--icI, Wuhml10n UDlYel'Sily Destgn Cue SliMly No 6.
Jl 1132, edited by B 0 Smrth. Wulu0lton UniY'ClSlly, 5'- Louis, AUI t. 1986.
Mol bel1U'llCjmol Slyrene
0 o.oos 0.010 0.020 0030 0060 0100 0140
Conyers..", JI
0 0.10
0" 0", o.n
0." OJ' 0'"
r ...... R. "" \It'
114 SfC'IlO'" oJ SUI....... IY. EXUClSE$. .. Nil NONI:NC.....nJ.r
SECTION 0) SUMNAI.Y. EXUOSfS, ""0 NOMfNC1A rlln 115
R- W w.,,,,...1Id E. C Dybdal. CIw", NJ ,.."" .... tal IH
TABU 4.3-5
Moks of 10lueM pet" mole of styrene lersus con,erswo
Mol 101_ 110OI IlymIC 0 0.006 oot'
00" 00"
0.010 0110
"60
COIIv<:n1011 0 0.10
0"
0.20
02'
0.20
02'
,<0
How do the resulls change if the first reaCllon IS KCOnd-order and the
reaCllon is also second-Qrdu?
43-&. To beller undustaDd Ibc sunilanuC5 and ditferCJKlCl bc:lwCCD the dCIJgns of a
COPtlouOUS and a balcb prooc:ss., let us consJdcr a very OVCl"JLmplificd dcuga problem
..here tbe prD0:55 collSisu of oniy II SlOpe reaClor. We desire to produce Pfoduct B
by tbe reKUOII A-B. The: cost of A is C/ (S/mol). we 81SO hr/yr for a
conllnuous plant, the desarcd production rate is PmoVbr, tbe reaction lUes p1aoc by
a fil"Jt<order isotbermal reaCllOn, the separattoD oltbe product from unoonvened
ractlnll is frcc, and we caDnot recover and recycle any uoconvenc:d reac1ants. We
ha"e to pay for Ibc raw malenals and reaCtOr, so our cost model ba;omes
ptodUCIIOll rates of fl l and 8. lire PI and p. moll1lr, tcspc:ctl\dy. both reactiOns are
fir>t-Qrdcl, lSOthcrmat and trrc'erS!bIc wilh rcactk)p rate. constants l and " . the.
lk I
flSltlc:s are P, anJ P.; tbe n:;IIctor downtllnes Ire I" and IJJ; the numbel"J of
bJ.tcbc>. per yen arc", and II" and only one rCilCI()f IS used to make both prodUC4.
Ilow do Ihe. rc:sults for USlDg t... o separale rc:act.us compare 10 tbe results for uSlOl a
slOglc rellClor for 3 C.lle ...-here thc rCill,:tor cost IS Incn by the folio... ing uprCS>olOn?
Reactor Co>t - C. I b < I
4.3-10. SUppo:>c thai t... o parallel reactors arc used lD Ihe process dcscnbc:d In c.>.crcisc: 43-8
I low do thc reaClion limes, number> ofbatchcs per year, raw-materials COSts, reliClor
slle, and reactOf" cost dllJcr from the case of using a Single reaclor?
4.3-11. lSOOCIallC (gasoIIOC) can be ptoduced by tbe reaCllons
BUleoc + lsobulllllC" -ISOOCIaoc
BUte-DC + ISOOCIaooe - C', J
Thc reactions take pboc: In the hquld phase. at 45'F and 90 pua ID a contlOuous
sltrrallank reactor. Assume Ihatthe reactIon klDCIICS agrc:c:s "'lIh the stOichIometry
and develop an u.prcsslOo for the SClecuvlty (l5OOCIane produ.:ed per buten:
coo.e.ned)
(4.3-1) TAC _ elF,8150 + C.V
The prodUCllOn ralC IS related 10 tbe fresh feed rale Ff and lhe conversion x by the
upresslOD

Rc:aclanl
Ptoduci
CO>I of reactants ($ mol)
Armualizc:d COSI of reactor \lolume [S/(fl
'
. yr))
Ftesh feed rale (mol/hr)
Reaction rale conSlanl (I/hr)
Production rale (mol/hr)
Sdecllvity
Downtime per balch (hr)
ReaclioD time per balch (hr)
Reaclor volume (ft
J
)
Com'ersion
ProdUCl purity
Molar deDSuy (mol ftl)
x
A
B
C,
C.
F,
k
P
S
I,
'.
V
x.
P.
(43-3)
(4.3-2)
(4.3-4)
P_ F,x
F, ,
V __
kp..
and the reactor .olume IS gJveD by
Thul, .....e can wrile
TAC_ 815OC/P + C.P l,,{-'-)
x kp.. x '\I-x
SInce tbe tmal annual cost becomes unbounded wben x approacbes dther uro or
unlly, there must be an optimum conversIon.
Suppo$c we do tbe same prD0:55 In a batch reactor, where lI..e produoc: "
balC:bc:s per year fot 7S(X) bt/yr Denve an eaprCSSK>n for the tOlal annual cost io
terms oftbe convcl'$)()o. Let the lImc II lakes 10 emply, dean, and refill the reactor be
I
J
and the re:KIlOO ume per cyde be r, Ilow do the Upre$SIOns roc tbe hitch process
compare to tbe resull lOr the conunuoU$ plant?
COOSldcred tbe problem of making tWO products In a proa:ss thai COnlLSI5
only of a rUClor, I.e.. Jdenllcal to ExerClle 43-8 exoept thai we have two ructions
A, - B, and A. _ B); the COSlS of the raw materials are CII and Cn; Ihe desired
4.3-.
S S...a.au, Prchnunary Deslp and OptUlUZaoon 01 Baldi JotS Theus, U.llvenuy 01
MuucbldClU, Amhcm. 1915
SECTIOl'f 'I DECISIONS FOR II'lI'l,n.()UT'l1l STIlUCTUU 111
TABLE 5.1-1
Bier.relty or decisions
FIGURE 5..1-1
Inpul-<fu,pu' .lrudu't of I\o..-'!htd IF,."" J 101 D<ougltll. AIO.[ J. JI: JJJ (/98J)-J
I. Bau:::b Ya5US conbnuolI$
1. lnpul-oulput Sl'1lC\Ull: of lbe IIowshtt'
J. Recycle of the lIo..-sbr
... ...t of 'M ....
" '".pol' 'ylltm
" LIquid Stl'u.'1Of\ ')"ltm
5. lleal<lLch.an&t. ncr .. o,l
order-of-magnitude argumenl to say thaI this rule of thumb is equivalent to
requiring that we complelely recover and lhen recycle all valuable reaclants. Thus,
Fig. 5.1-la indicates thai no reactants leave the system.
Of course, if air and water are reactants in a process, they are sufficienlly
mexpensive. compared to organic materials. that il mighl be cheaper 10 lose them
10 an exit slream rather than 10 recover and recycle lhem. lienee, in a few rare cases.
Fig. 5.1-10 mlghl not be complele.
The olher sllualion in .... hlCh commonly reaclanlS have been losl from a
process occurs when we have a gaseous reactant and ellher a gaseous feed impurity
or a gaseous b)-produCt produced by one of the reactions. We wanl to recycle the
gaseous reactant. but the componenlS mUSI be purged from the process so
Ihat lh(") do not contmue 10 accumulale m the gas-recycle loop. In the past. It was
so expensive 10 separate gaseous mixIures that some reactant was allowed 10 lea\e
the process in a and purge stream (see Fig. 5.1-lb). Ho.....ever, the new
membrane-separation lechnology, such as Monsanto's prism process, has CUI Ihe:
COsl of gaseous separations so lhat gas recycle and purge might not be necessary
(b)
__P_'-,OC-,= :;:::ucu
(a>
=L
r
-_.c ,,",g,
P,oc,,, Pm""'"
_ By-prOOucls
streams
streams
5.1 DECISIONS FOR THE
PlIT-OlITPlIT STRUcruR
Flowsheel Alternatives
Almost e\ery flowsheel has one of the IWO structures shown in Fig. 5.l-la and b.
There is a rule of thumb in process design (see Sec. 3.4) that it IS desirable 10 recover
more than 99 y. of all valuable materials. For initial design calculations .....e use the
To understand Ihe decisions required to fix the input-output structure of a
f10WSheel, we merely draw a box around Ihe total process. Thus, we focus our
auention 00 what raw materials an: fed to the process and what products and by-
products are removed. Since Ihe raw malerials costs normally faU in lhe range rrom
3310 85y' of the lotal processing costs., we wanl to calculate these costs before
we add any other detail to the design.
In Sec. 1.2 we: descnbed a hierarchical decision procedure for inventing process
flowshccls and base<ase designs" The decision levels are repealed in Table 5.1-1
The batch versus continuous decision of level 1 was discussed in Sec. 4.2, and in the
subsequent chapters we discuss the other decision levels in detail.
INPUT-OUTPUT
STRUCTURE
OF THE
FLOWSHEET
CHAPTER
5
116
SEmON II DC!SJOI'I5 FOil 11lf INPUT-OUTPUT snocnru 119
PROCESS ALTt.RNAltVE. Unfortunately, not all these dcslgn gUldellncs are
quantttallve, so often we must base our Innial decision on our best judgment
(which might be merely a guess)
Wilh Ibts approach, we have a systematic way of generating a Iisl of process
alternall\es
ECOl'liOMIC TRADE-OFFS FOR FEED PURltlCATION, Our deciSion of
(ying tbe feed streams before they arc processed IOvohcs an economic trade-off
bclween bUlldlng a preprocess separation system and mcreasing the cost of the
process because \oloe arc handltng the increased flow rales of men matenal$. Of
course, the amount of mert materials present and where they will efilet and lea\e
Ihe process may ha\e a great Impact 011 the processing costs. Therefore, it IS not
surpnsmg Ihat Ihere IS no simple deSIgn criterion lhat always indicates the corn::cl
decision.
Since the second reaction IS rc\-ersiblc, we could recycle the dlpbcnyl back to the
reactor and lei II build up In the recycle loop until it eventually reached an
equilibnum level. That is, the recycled dlphenyl would decompose to form benzene
at the same rale as ben2ene would be producing diphenyJ.
If we recycle the reverSible by-producl, we must oversl2e all the eqUIpment in
that recycle loop, to accommodate the equilibrium flow of the reversible by-
produci. However, If we remove it from the process, we pay an ecollomlC penally
because of the Increased raw-material cost of reactant (toluene) that was conver1ed
to lbe reverSible by-product (dlphenyl), e.g., the raw-matenal cost of toluene minus
the fuel value of dlphenyl. Since our decision Involves an a;onomlc trade-off
bet""een raw-material losses to less valuable by-products and IOcreased recycle
TABU; 5..1-1
u,'d-2 decisions
I, Should we: punr, ,be: M.c:d SITUIIlli bdOfe: ,be:y C'IIlc:r tbe: Fooas1
1. Should _ umoY<: .... locydc a ,.......tslbk by-prodlK1
1. SbouId ....e: IlM' aau reqo.:k aad pu,..., IIrQm'
... Should we: DOl bo,be:. 10 n:covu and wmr: ruotUnlS'
5.. 110.... /lUDY prOOt>C1 SlfQIDS "In lbe:rC' lx'
6. What arC' tbe: dlOS'ln unables lor lhe: mpul-OUlpul IolnKU1f<:. and ",hal CCOlKlrID<: tndC'o/h .'C'
aUOClatw .. lib lhae ""n.. hles'
We distinguish belween Fig. 5.1-la and b because the presence of a gas
recycle and purge stream adds a design degree of freedom 10 the design problem;
i.e., eilher the reactanl composition of Ihe purge slream or tbe excess
reactanl fed 10 the process becomes a new design variable. We discuss the dcslgn
variables later.
uvel-2 Decisions
The decisions ....e must maLe to fix the mput-outpul structure ofttle flowshecl are
given in Table 5.1-2 and arc discussed belo","
PurifiCliltion or Feeds
A decision 10 purify the feeds before:. they eDter the process is equivalent to a
decision to design a preprocess purificalion system. This is different from a
10 feed the process Ihrough a separation system thaI is required 10 any evenl..Smcc
at this stage of our synthesis and analysis procedure we do nOI what of
separation system WIll be required for the process wilh no feed Impuntlcs, we
cannol always make a definite: decision.
Some design guidelines to be considered are as follows:
If a feed Impurity is DOl inert and l$ prescDt in signlficanl,
quantities., remove it (otherwise, it will lead to raw-matenal
losses, aDd usually a much more complicated separauon
system is required to recover Ihe additional by-produClS). (5.1-1)
If a feed Impurity is present in a gas feed, as a first IUess
process the Impurity. (5.1-2)
If a feed Impurity in a liquid feed stream is also a by-
produci or a ptoduct component, usually it is bener to reed
Ihe process through the separation system. (5.1-3)
If a feed impurity is presenl in large amounU, remove it
(there is no quantitalive cnlenon avaLlabk to indicate how
Mlarge" is large), (5.1-4)
If a ked Impurity IS present as an azeotrope with a reactant.
often It IS beller to process the Impuflly.
If a feed impunly IS lOert but lS easier 10 separate from the
product than the feed, II IS betler 10 Ihe impurilY.
If a feed impunty 15 a C01lotl)st poison, rClllove It.
If we are nol certam that our decision is correct, we list the
opposite decision as a process alternative.
Reco\er or Rec)'c1e Re\'ersible By-producls
The r.r:.actlons 10 produce benzene from toluene are
Toluene + Hz ..... Benzene + CH.
2 Benzene Diphenyl + Hz
(5.1-5)
(5.1-8)
(4.1-3)
120 SECTION,. DCISlON5 rOl mE INPUl-OllTPUl nlUCTUl..I'.
ncnoN 51 DKISIom FOR nu! IN"".()U'T1'VT snUCT\}l 121
costs. it should not be surprismg that no iimple design guideline is available to
make thIs decLSion The result Will also be sensiti\e to the equilibrium constant of
the b)'-product reactIon. So we generate another process ahemative
TABu.: 5.1-J
DestiJUltion coda and compoDnit dassificatiOfL'i
Do Not RKOver and Rcc)'cle Readanl!'i
One of our design gutdelines states that we should recover more than 99 %of aU
valuable materials. Since some materials, sucb as air and water. are much less
valuable than organic ones, we normally do not bother to recover and recycle
unconverted amounts of these components. Of course, we could try to feed them to
the process so tbat they would be completcl)' converted, but often we feed them as
an excess to try to force some more valuable reactant to complete conversion.
Forellample. in combustion reactions. we usually use an uocss amount of air
to ensure complete conversion of the fuel. The greater amount of the ellocss we usc,
the closer to complete conversion of the other reactant we oblain. However, the
capital and operating costs of Ihe blower used to move the air. as well as preheating
and cooling costs. increase as we increase the ellcess amounl of air Thus there is an
Ca!'i Re'C),de and Purge
If .....e ha\'e a Mlight- reactant and either a feed impurity or a b)-
product produced by a reaction. it used to be common practice to use a gas rec)'cle
and a purge stream; Ie. we want to recycle the reactant. but we must purge one or
more components from the process. We define a light componenf as one which boils
lower than propylene (-SS"F, -4S"C). We choose propylene as a breakpoint
because lower-boiling components normally cannot be condensed at high pressure
with cooling water; i.e. both high pressure and refrigeration would be required.
Since gaseous reactants normally are less expensive than organic liquids. and since
refrigeration is one of the most expensive processing operations. it usually was
cheaper to lose some of the gaseous reactants from a gas recycle and purge stream
than it was to recover and recycle pure reactant.
However, membrane technology. such as Monsanto's prism process. now
makes gas separations less expensh-e.. Unfortunately. there is not a sufficient
amount of published information available concerning tbe costs of membrane
to be able to develop new design guidelines. Thus, we treat the membrane
separator as another process alternative.. Because of the lack of published informa-
tion. we would base our initial design on a process with gas recycle and purge,
whenever our design guideline is satisfied:
Numbt'r of PToducl Stream!'i
Example 5.1-1. Suppose we have the 10 components listed ill order of l1lelr boihng
points and WIth the dCSllnatlon codes indlCl.led. Howmany rroduct streams will there
"'?
(5.1- 10)
G:ascout b,prodllClJ alld s...:d .... puntMS
Gueous IUClll!lS pNllIICf1 pscs andlor p_bJ'producu

Rcaettoll Ullermedl.ltcs
Auonopcs "'llb reactlnls (oomchmcs)
RevnsitHc byproducts (somcumcs)
RUela.nUl-lf complete COnVCl'tI(lD 01 ulUllble rc.lCIJOIl ,ntcnncd'llcs
Gueous resetlnl nOI rco;o'fClCld lnd rco;ydcd
Liquid readlnl DOt recovered or recycled
Primll)" Pfoduel
Separlle dcstUl&tion for b,prodIKUI
B,products 10 fuel
By,pfodu.eu 10 "Ule 1R:l1lDC1lt
I. V""I
1. Recyck and purse
, Rydo

5. E1cas-venl
6. E1cea- ...ute
1. Prim.a.1)" produn
L Valuable byproduel (I)
,. Fuel
10- Waitt
optimum amount of eJ:oess that should be used Similarly, at some point. the cost of
noess water used will become significant Moreover, pollution l.tealment costs
must be considered.
To determine the number of product streams that \\i1lleave tbe process, first we list
all the components that are expected to leave the reactor. This component list
usually includes all the components in the feed streams and all reactants and
products that appear in every reaction. Then we classify each component in the list
and assign a destinalion code to each. The component classifications and destina-
tion codes are given in Table 5.1-3. Finally, we order the components by theIr
nonnal boiling points. and we group neighbouring components with the same
destination. The numbeT of groups of all but the recycle streams is then consKlcred
to be the number of product streams.
This procedure for determining the number of product streams is based on
the common seme design gukl.eline that
It is never advantageoU!'i to separate two streanu and then
mix them together.
Also, it is ba!'iCd on the assumptIon tbat the components can be separated by
distillation and that no azetropes are formed. Thus, in some cases (i e.. solids) a
different set of rules must be used to estimate the number of product streams.
(5.1-9)
Whenever a light rcactant and either a light feed impurity
or a light by-product boil lower than propylene (- 5S"F.
- 4S"C). use a gas recycle and purge stream.
A membrane separatIon process also should always be considered.
122 UCTtOfII J I DIlCISlOHS HMt nut INPUT-QUTPVT nalJCTUle
CompoMJtI
Otsri_tioa c.m..-., Dati.luloo.
A Waste F Primary product
8 WaSle G Recycle
C Recycle II Recycle
0 I Valuable by-product I
E }
=L
':iI'",,; H,. CH,
H2 CH4 Process Benzene
Toluene _ Diphenyl
"'IGUR[ 5..1-2
Input-OUlput iUUdUIC 01 HOA pc-oecu. [F,_ J. JJ 31 ljl
$ohniolt. The product streams an:
Evaluation of Ihe F10wsheel
TABL5.J-4
to beo.uoe
Example 5.12 HydfOlilkyladOil of to pl'"odUtt bf,auoe. Find tbe Dumber of
product streams for the HDA process; ie.. sec Enmpk-4.I-l.
(5.1-1 I)
Be nain that all products, by-products, aod impurities
leave the process!
At each stage of the development of a f1owsheel, this is essential:
For example. in our HDA process. there mighl be some other impunlies in the
hydrogen feed stream. We would expect them to leave with the purge. We also
expect Ihat Ihere will be some impurities in the toluene feed stream. We altempt to
identify Ihese impurilies by considering the sour of our toluene, and the:n we look
for where tbey will exil the "process. Furthermore, we need to consider the
possibility of other side reactions (such as Ihe formation of terpbcnyl), and we must
ensure Ihal tbc:sc new by-products kave the process or can be recycled
If even a trace arnounl of a component IS fed to a req'de loop and is not
removed, Ihe process will be inoperable. This is one of Ihe most common mistakes
made by inexperienced designers.
5.2 DESIGN VARlABLES, OVERALL
MATERIAL BALANCES, AND
STREAM COSTS
To calculale: Ihe overall proocss material balances and the slTeam COSIS. we must
first assess wbetbcr Ibe probkm definilion is complete or whelbcr there are any
degrees or freedom that must be specified to calculate tbe malerial balances. In
general. lbe pmblem definition is not complele:. Therefore, normally, it 15 impossi-
ble to develop a unique set of material balances ror a process. Only a process wilh a
singJc reaction and with no loss or reactants from the prooc:ss has no degrees of
freedom.
or course, ir tbe: material balances are not unique, Ihen Ihe stream costs will
also not be unique. Thus. we must develop the material balances and the: stream
costs in terms of the unkno.....n design variables, and eventually "'e will Jook for the:
economie optimum values or these design variables.
Ree:yclc and purac
Ree:ytle and purac
Primary product
."""

-2S3-C
-161"'C
...c
lwe
2Sl"'C
Sob<,iOll. Ill. Example 41-1 tbe components in tbc reacbons are loIUC'lK, hydrogen.
bellUM, mc:lhalX', and diphenyl. No addlUonal oomponents arc prC$CII1 In e:uba of
the: feed streaJns,. U ...Ie arrange Ihese componenlS In order of their nonnal boiling
poiols, we oblain the results shown in Table: 5.1-4. Since both hydrogen (a reaC'lanl)
and melhane: (bolh a feed impurilY and a byprodllCl) boil al a lower temperature
Ihan propylen.c:, we decide to usc a ps recycle and purge stream. 8enune is our
primary prodllCl, and loluene (a reactant) is rcqc:kd. Dipbmyl is a revemble by-
prodllCl, and for lhis uampJc we decided to rc:movc: it from tbe process aDd to we it u
pan of our ruel supply. Tbcsc destinatioD c:odc5 also arc given in Table 5.1......
Thus. lberc: are three prodllCl streams: purge coDtaininl H
J
and CH., the
primary prodllCl stream conlaining benzene; and a fucl by-prodUCt stream conUliniog
dipbenyl. The initial t10wshect is given in Fia- 5.1-2.
I. A. + B 10 waste (do not separate them and tbeD mix them in the sewer)
2. D + E 10 fuel (do not separate them and tben mix them 10 burn)
1. F -primacy product <to stOr.lse lor sak)
4. I by-product 1 (10 storage for sale)
S. J to fud (J mwu be separated from 0 and 10 recover compoocnlS F. G, H, aDd I,
$0 we Ireal J as a separate product stream)
e.-,I I t
H,
CH.
..

O!:pbmyl
Design Variables
1lle vanablcs we select to compkte the: definition 0( the design problems are what
we call the of For compkx reactions., it is usually possible to
correlate Ihe product distribution measured in the chemist's laboratory as a
function of the conversion of the limiling reactant, the molar ratio of reactants, the
reactor temperature, and/or the reactor pressure. U the activation energies of all the
reactions are equal, then temperature will not appear in this correlation. Similarly,
if there: are the same number of moles of reactanlS and products for gas-phase
reactions, or if we are considering liquid-phase reactions. then pressure wiU not
appear in this correlation. In addition. we usually attempt to correlate the
conversion against the space velocity in order to estimate the reactor size..
For preliminary designs, where kinetic data often are not available, we
assume that the reactor configuration used in the process is the same as that used
by the chemist (a batch reactor is equivalent to a plug 80w reactor), because we
cannot evaluale the effect of the reactor configuration on the produci distribution
without a kinetic model. However, if the preliminary design calculations indicate
that additional development effort is justified, part of that development effort
should be directed to determining the best reactor configuration.
In addition to the design variables which affect the product distributlOll, the
other design variables that enter into the overall material balanoes correspond to
situations where we do not recover and recycle all the reaetants. Thus, if we use an
excess amount of air in a combustion reaction, we must specify the amount of
excess. Similarly, if we have a gas recycle and purge stream present, we must specify
either the excess amount of gaseous reactant fed (i.e., in eJlccss of the reaction
requirements) which is lost in the purge stream or the reactant composition in the
purge stream.
A list of the level-2 design variables that might affect the overall material
balanocs is given in Table 5.2-1.
TABLE 5.1-1
Procitdures (Of" ieTdoping malerial balances
I. Start ..uh lhe apcCficd productoon ... Ie..
1. From lhe uo,cl'lIometry(ancl, for complu leIIChOllS,.lhe COtn,bl\>on fo' prodllCl dostnbuuonl find the
byprodllCl nows aDd the ruaant requlrementa (in tcmu o/IM destJJl vanablcs)
J,. Calculatc tM Impurily mlet and outlet Bowa for lhe alreaml wherc the .elIctanlla.c completely
,o>'creel and ,ecycled
.. Calculatc the OUlla 110.... of reactanla ,ft lerm, of a apccified amount of UCICSa (abo\'e tM n:acllon
fCCjouremauaj fOlIllrc:ama ..bien: lbie reacranll arc 001 1CCO'CI..cl and Rqded (rccydc and PU'J'C 0'
.... or ....tcr).
So CaJeWa,lc the: ullcI and oullcllloW1' for the tmpu,nuea enlm.t WIth the reactant lllreams '0 IUCp.
Material Balance Procedure
To de:ve:lop the: overall material balances ror single-product plants, we: always slart
with the give:n production rale.. (If a process in an existing complex is replaced by a
new process, lhen the: fc:ed rate: may be known instead or the production rate.)
From the production rate, the reaction stoichiometry, and (usually) the correla-
Hons for the product distribution, we calculale all the by-product flows and the
raw-material requirc:men15 as functions of the desiln variables that appear in the
product distribution correlations. Then, Knowing the impurity compositions of the
feed streams, we calculate the inlet and then the oullet flows of the inert materials.
For feed streams where an excess of reactant is fed and not recovered and recycled
or a feed stream where the reactant exits through a gas recycle and purge stream.
we must specify the excess amount ofreactant that is lost from the process. Then we
can calculate: the inkt and outlet flows of any impurities from the composition of
these: feed streams. A summary of the procedure is given in Tabk 5.2-2.
TABLE 5.l-t
Possible design variables for level 2
Ovall Material Balances
Nonnally it is possible to develop expressions for the overall material balances in
terms of the design variables without ever considering any recycle 80ws. Any time
the overall equations are underdetioed. it is necessary to look for one or more
design variables to complete the problem definition, and these design variables
always correspond to sig"ificmtl prOCLSS-Cpfindzotion problems. Henoe, the initial
analysis should always focus on the input-oulput 80ws only.
ReaaOf oon'Cf'SlOn, motar rallO of rcactuta,
'eactlOn temperatun and,tu< ptCBWC
R_nU not '_\'enld...- au 'CICY'* aad ...'tc
Limitations
Note that the preliminary material balances described above are based on the
assumption of complele recovery of all valuable malerials, instead of the rule-of-
thumb value of grealer than 99% recovery. There are no ruks of thumb available
to fiJI any 0( the design variables. Moreover, it is aJlIl'ays much more important to
find tbe neighborhood of the optimum values of the design variables which tix the
inlet and outlet nows than it is to include the losses early in the analysis. Howe:ver.
we v.iU want to revise our initial material balances at some point to include the
J=.
Our approach focuses on processes that produce a single product. There are a
number of processes that produce multiple: products, e.g., chlor;nalion or methane,
amine: prooesscs, glycol plants. refineries, etc. The overall material balance ca.lcula-
tions ror these: processes are usually more diffkult than the procedure described in
Table 5..2-1 because the product distribution must match the market demand.
126 II nOI(,N "-S. I'o!AIUIAL llALANCE.S. AI'U HlEAM
SECflON H lH:.SI<iN MATEliAL BAlANCES. AND STlUM COSTS 127
.5.2-1 T"h"'M W knLt'ne. lk...dop lhe m;olenal bal,mCCi (or lhe
II DA process
and we arc 1I\e:o thaI the: dc:smcd proouclioo nlte of bc:nzcoc: 1$ p. _ 265 mol/hr
(ElIImple: 41-1) If we usc a gas recycle: and purge: nrc:am for the: H
J
Ind CI-I.,lnd If
we: recover Ind lhe dlpbenyl, lhen there are three: product Ilrc:ams; see
E10Impie 51-2 Ind tbe: 6o......ihect lI\eo In Fig.. 5.1-2-
We nole from Fl!. 51-2 that aU the toluene we feed 10 the process gelS
convc:ned (I.e.. no lolu.eoc: lea... es the system). However, dUI does DOt mean that the
con..-erswn of loluene: lD the reac!or is unity A low conver-non per pass stmply means
thlt there .....111 be I large inlernll reqck flow of loluene.. H01O'ever, we eVlluale: tbc:se
rttyde: flows III later stage of our desIgn procedure:. Of course:. our lS5umptlon lhlt
aU !he fc:e:d loJuene is con\er1c:d negJc:eu any IOII1C:DC 1055CS lD any of tbe product
II reams, bUI IhlS COtTesponds to Our Initial design assumpllQn of complete: recovery
I!lfClter than reoovenes.
(5.2-6)
(5.2-4)
process, this hydrogen will leave with the purge stream. Thus, the IOtal amount of
hydrogen fed to the: process Will be
p.
FE. + 2S(1 +S)-YFHFG
which IS equal 10 Ihe amount of hydrogen to the makeup gas stream YrHfc;.
Similarly, the methane flow rate leaVing the process will be the amount of melhane
entering the process. (I - YrHWG' plus the amounl of melhane producro by lhe
reaetJons, PIt.CH. = P.IS, or
p.
POI. - (I - YrJl)Fc; +S (5.2-5)
The t01a1 purge flow rate PG will then be Ihe c:xocss H J , FE, plus the tOla! methane
POI.. or
(41-))
Sulurju". The reactions of mll:rC:S1 arc
Toluc:ne + II
J
8c:0zc:0c: + CII.
28c:0zeoc::;:. Dipheo)'1 + II
J
Also, from Ihe StOIChlomelry (Eq. 5.1-3), the amount of melhane produced PIt.CII.
must be
If a fraclion S of toluene is con..erted to benzene, a fraction J - S must be 10SI to
diphenyl. However, from the stoichiometry of Eq. 5.1-3, the amounl of diphenyl
produced P
D
musl be
(5.2-9)
(5.2-8)
p.
YtH)FG + S = (1 Y'II)PG
(1
and the melhane in the fec:d plus lhe methane produced must all leave with the
purge
Rather thao uSing the excess hydrogen feed F as a design variabk, we
normally use tbe purge composJllon of the reactanl 1'H' wMre
F,
)"H = - (5.2-7)
p.
This purge composmon IS always bounded belween zero and unity (aclually there
is a smaller upper bound which depends on the feed composition and sometlmes on
lbe conversion). The use of bounded variables makes the preparalion of graphs
simpler.
We can develop expressIons for the makeup gas rail' FG and the purge rate
Pc; explicitly ID tenns of the purge composition of reactant }'rN either by using
Eq. 5..2-7 to eliminate F from Eqs 5,2-4 and 5.27 or by writing balances for the
hydrogen and methane and then combining them. That is, the amount of hydrogen
in the feed must supply the net reaction reqUirements as well as the purge loss
(5.2-1)
(52-2)
(5.2-3)
p.
PIt.Ol.. = S
SELECflVITV AND REACTION srOICIUOMETRY. We ddioe the ulectivity S
as the fraction of loluene converted In lhe reactor that corresponds 10 the benzene
flow al lhe reactor oUllet Also, we recover and remove aHlms benzene. Hen, for
a production of p. mollhr, tbe toluene fed to the process FFT must be
p.

S
Since the toluene stream contains no impurities, we find Ihc: tnluene fresh feed
rale and the diphenyl by-product flow rale in tenns ofselcctivity from the reaclion
stoichiomelry and the given production rate. A relationship between selectivity and
conversion is given ill Example 4 I-I.
Addmg these expressions gives
RECVCI. AND PURGE. The stoichiometry also indicates the amount of H
J
required for Ihe reaclion If we feed an excess amounl of hydrogen, "-1;' inlo Ihe
r. I - S
P.. -FG + S 2
(5.2-10)
of the desired component divided by thc production of thc undesired component
For the HDA process. in the first case we would ha\c
Enmpk 5.1--2 Tor_ to kurer.. tbe apressioas relauD@ lbe utcnlS of
reacuon to prodtlClion ratc and $ClcctivlI) for tbe HDA process.
(52-20)
(5.2-18)
(5.2-19)
SoIMtioll. From 5.2-15 and 5.2-1 we: find that
p.
C
s
s = -::
. (,
whcreas in the second case we would ha\c
-
S. =
(,
II is essential 10 ensure that the definition of that IS reported by a chemist
(or in the litcrature) is clearly defined. but it is a simplc matter to COfl\'ert from one
definition to another.
(52-II)
We can thcn solve for Fr;:
MATERIAL BAUNCES IN TERMS OF EXTENT OF RUCTION. It is becom-
ina common practice 10 describe material baJance: calculations in tcrms of the
utcnt of reaction I (or fractional extent of reaction). Thus, for tbe HDA process.
we would say that mol (or molesjhr) of toluene react witb mol of H
1
to
produce mol of benzene plus mol ofCH... Also, mol ofbenune produces
C
1
mol of diphenyl plus mol or hydrogen. Then, we combine these statements to
5ay that
If we are given valucs of p S. and either F[or Y'H' we can use the abO\e
equations to calculate the fresh feed rate of toluene. FH (Eq 5.2.1), the production
rate of diphenyl, Po (Eq. 5.23), the makeup gas rate Fr; (F..q. 5.2-4 or 5.2.11), and
the purge f1owrate. PG (Eq 5.2-6 or 5.2-10). Thus. we have detennined all the input
and output flows in tenns of the unknown design variables S and either F or )"H'
Net benzene produced - -
Methane produced _ I
Diphenyl produced ..,
Toluene consumed -
Hydroaen consumed - <I -
(5.2-12)
(5.2-13)
(52-14)
(5.2-15)
(5.2-16)
Also from Eq 5.2-12 "'c find Ibat
Tbu.s \III'e: can \III rilc
(5.2-21)
(52-22)
We can generahze these expressions and say that the number of moles (or
moles per hour) of any component is given by
(5.2-17)
where the II
IJ
arc the stoichiometric coefficients, which are positive for products and
negative for reactaol5.. Normally (for tbe purposes of initial design calculations) no
products are fed to tbe process, II; = O. and no reactants are aUowed to lea\'c,
"J "'=' O. Experience: indicates that it normally is possible to correlate tbe extent of
each reaction <, aaa-iost the perpass conversion of tbc limiling reactant., although
in some: cases the molar ratio of reactants, reactor temperature, and/or reactor
pressure: must be included in the correlation.
EXTENT VERSUS Sf:L.ECT1VITI. For lhc case or only two reactions. It often is
simpler to describe the product distribution in terms or selectivity. Selectivity can
be described in a number of different ways, such as the production of the desirable
component divided by the amount oflimiling reactant converted or the production
Stream Tables
It is common practice to report material balance calculations In tenns of stream
tables. That is. the streams are numbered on a flowsheet, and then a table is
prepared that gives tbe component flows in cach of these streams which correspond
to a particular set of values of the design variables.. An e:xample is given in Fig.
S.2-1. Tbe temperatures., pressures, and enthalpy of eacb stream also are normally
listed_ Since we do Dot consider energy balances until thc end of thc synthesis
procedure. we add \alues to the stream tables later
One major difficulty with this conventional practice: is that the designer is
forced to select values of design \'ariables without knowinatbe optimum values.
Moreover. once a sct of values has been selected, it is orten difficult to remember
tbat they were selected arbitrarily. For this reason, we recommend tbat the stream
tables list the: values of the design variables and that the appropriate material
balance equalions. such as Eqs. 5.2-1. 5.2-3. 5.2-4 5.2-6. etc.. be programmed on fI
such as LOTUS. With this approach it is very easy to change: the
production rate and the design vanablc:s and then recalculate: all the stream nows.
We can use this same table as the basis for calculating the: stream costs a functi(>n
of the design variabks.
130 SECTION S! DeilON V"II"aLES., OVaAlJ.. W"TUI"L ......... NCeS. "NO s.-."" .. COSTS
UCTION H Of.SKiN V"kl"lu.s. OVUAU M"TUI"L II"LANCES, "ND srlEA.M COSTS 131
o12l " 10" lilli/mol
0.383 " 10 Bill/mol
141 " 10 Blu/mol
1.68 " 10" Blu/mol
2.688 " 10" Blu/mol
SO.8S/gal = S904/mol
so SO/gal _ S6 4O/mol
S3 00/1000 fl' .. Sl l>l/mol
Value of
Vilue or
V of HJ feed
Fud Sot 0/10 Blu
Flld H,
CII.
Benzene

DipbCllylt
A..u.... lIP value v...... the
pnca. nl U.l6/pJ.
t w. a.Ioo ......"'" wt the fuel val... nl d,pbe..yl .. SSJI/IDO!.
TAllLE U-3
Cost datil for HOA process
5
Diphenyl
Benzene
Purge
H,
4
5
3
H
2
, CH
4 I I
H
2
, C
I
3
2
Process
4
,
Production rale = 265
ToJuen
Design variables: FE and x
Componem 2
4,000,000
Example 5,2-3 HDA Siream CMIS. If we use Ihe coS( dala given in Table 5.2-3.
""here tbe values of Hl' CH., and dlphenyl in the produci Streams are based on the
healS of combusllon of lhe compollc:nls and a fuel value of S4/I0
6
BlU. we can
calculale Ihe economic polential for lhe HDA process in terms of lhe design variables
(""e use reactor per pass x, instead or S. and YI'H). The results arc sho... nlll
Fig. 522. The graph indicates lha' al high conversions Ihe process is llnpt"ofilable
The economic polential is the annual profit we could make if we did not have
to pay anything for capital costs or utilities costs, Of course, if the economic
potential is negative, i.e. the raw materials are wOrlh more than the products and
byproducts, then we want to tenninate the design project. look for a less expensive
source of ra..... materials, or look for another chemistry route that uses less
expensive raw malenals.
YPH
"'"
...... 0.1
---..
-+- 0.7
___ 0.9
i"..
'li
"
0.7 0.6
Conversion x
0.3 0.4 0.5 0.2
3,000,000
2,000,000
1.000,000
"
0
-1,000,000
-2,000,000
-3.000,000
-4,000,000
0.1
H, F
H
, 0 0 0
FE
CH. F
M
0 0 0 F
M
+ PsiS
0 0 p. 0 0
Toluene 0 PsiS 0 0 0
Diphenyl 0 0 0 Ps(l - S)/(2S) 0
Temperature 100 100 100 100 100
Pressure 550 15 15 15 465
where S I 0.0036/(1 X)U44 F
H1
_ FE
+ PB(I + S)/25
F
M
= (1 - YFH)(F
E
+ P
8
(1 + S)/S))IYFH
Fa = F
HI
+ F
M
FIGURE 5.2-1
Siream ur.bk HDA prOOCSli.
Stream Costs: Economic Potential
Since the Kbest" values of the design variables depend on the process economics, we
want to calculate the stream costs, i.e" the cost of all raw materials and product
streams in tenns of Ihe design variables. Normally, we combine these costs into a
single term, which we call the economic potential (EP). We define Ihe economic
potential at level 2 as
EP
l
= Product Value + By-product Values
- Raw-Material Costs, Slyr (5.2-23)
which for an HDA example would be
EP = Benzene Value + Fuel Value of Diphenyl + Fuel
Value of Purge - Toluene Cost - Maleeup Gas Cost (5.2-24)
We would also subtract tbe annualized capital and operating cost of a feed
compressor, if one is needed. (fhe calculations required are discussed in Sec. 6.5.)
FlGURE 5.22
Eco"omK 2.
TABLE s.J.t
Alternatives for the UOA
p<oc""
I. Punfy the bydrOlC'D r-t JlrcanL
1. Rcqdc 1M dlpheTl)'llo ulJnctK>n.
J. Punfy lhc U.-rceyCH. JlIam
..'C con"crt so much loluene 10 dlphen}l Ihal lhls selccuvllY loss ucecds the
mcrcucd value of the benzene Ihal produce). Also. al hIgh pur!C composItiOns of
hydrogen .. lose money so much hydrogen to cannol mue up
for IhlS )os;s).
Aocordm! 10 the gnlph. the mosl dcsu'lIbie vall.lCS (i.e., the Wellesl profil) of
lhe design vllnables corrcspond to x _ 0 (i.e.. no sc:lCClivily loss) and a reactant
composlIion or lhe purge slream Y'B equal to tero purge pure rm:thane),
As we prooeed lhrough the design, ho...ever, we find lhal a tero conversion per
pass (x - 0) corresponds to an infimtdy large recycle 801" of toluene and that purpng
no hydrogen lI,. - 0) corresponds 10 an Infinuely large gas-KC}'C1e Row Hena:, ... e
develop Ibc:.opumum vall.lCS or x and Y"" as proc:eed througb the design.
5.3 PROCESS ALTERNATIVES
In our development ofa design for the HDA process (see Fig. 5.1-2), we made the
decisions (I) nol 10 purify the hydrogen reed srream, (2) 10 remove the diphenyl
from the process. and (3) to use: a gas recycle and purge stream. If we change any of
these: decisions, we generate process allernatives. It is always good practice to make
a lisl of Ihese alternauves. Such a hSI tS given in Table 5.3-1.
Evaluation of Alternath"es
We could attempt to SImultaneously develop designs that corresponded to each
process alternative Ho....ever. if we remember that less than I of ideas for new
designs ever become commercialized, our initial goal should be to eliminate, with
as little effort as possible. projects that will be unprofitable. Thus. we prefer to
complete the design for one alternative as rapidly as possible before we give any
consideration to tbe other alternatives, for we might encounter some faclor that
will make all the alternatives unprofitable. Then, aner ViC have compleled a base-
case: design, \\'e examine Ihe alternatives.. In the terminology of artificiallOteUigeooe
(AI), we arc using a rather Ihao a
5.4 SUMMARY, EXERCISES. AND
NOMENCLATURE
Summary
The queslions we must answer 10 fix the mput-oulput structure of the flowshect
include the follo....
SECTlQN,. 'l-'loll.MIlY. VIllOSE$, .....-n NOIoIE"ClATU." 133
1. Should we purify the feed stream?
2. Shouk! we remove or recycle a re\'ersiblc by-product?
3. Should we use a gas recycle and purge stream?
4. Should we use: an excess of some reactant that we discard?
S. How many product $Ireams will there be?
6. What are the design variables and the economic trade-ofTs al this Je\d of
analysis?
In some caeS heunstJCS can be used to help make these decisions. When no
heuristics are available. we make a guess and then hst the opposile decision as a
process alternative. We complcle tint desigo based 00 ow original guess before
we consider any other alternatives. (Since Ic:ss than 1 ofidcas for new designs arc
suca:ssful. we might learn something about the process thai will make all the
alternatives unprofitable.)
Some of the heurislics and design guidelines that were presented include Ihc
following:
If a feed impurity is not inert, remo"e it
If an impurity is present in a gas feed stream. as a first guess process the
unpurity.
If an impurity in a liquid ked stream is a product or b)product. usually
the process through the system_
[f an impurity is present in large amounts. remove it
If an impurity is prese:nt as an a7.cotropc with a reactant. process the
Impurlly,
If a feed impurity IS an inert. but is easier to from the product and
bl-product than from the feed. il is beUer to process the impurit}
Whenever there is a light reactant aod a light reed impurity or by-produci
(where ligbt components boil lower than propylene, _48C), use a gas recycle and
purge stream for lhc first design. Also consider a membrane separator later.
If O
2
from air or water is a reactant, consider using an elcen. amount of this
reactant.
For single-product. vapor-liquid processes. we determine the number or
product streams by grouping components with oeigbboring boiling points that
have the same exit destinations; i.e.. we never separate streams aod then remiJL:
th<m.
Be certain thai all by-products and impurities leave the process!
The: significant design variables are those that affect the product distribution
and purge compositions of gas streams
Raw-material arc often in the range from]J to of the lotal
Exercises
5.4-1. Draw the inpul-outpul no...sheEt .nd plot lhe CIODfK)I'Tl;C rolenhal rot'" the IIDA.
rorOOCft for the cue wherr. the dlphc:n)t Os rec:yclaI
134 SECTION,. SUMMU.Y. UUL/Sb.... NO NOMlNCUTUIl.E
SFCTION,. SUMM"'IlY. OiEllCIE'I, "NO I'OMENCLnUIlE 135
11115 probkm IS 1I mOllified VerSIon oflhe AtChE Student COlllesl !>robkm. sec J J McKetlll,
Ent;yelofH!Jla of C,","'lclll ProcnslIIg IlM DC5JfJ". vol I. New York. 1976. p 27t
, R. R Weoner lind E. C Dybdll[, Cltr",. Eng 44(4); (l94g)
5.4-2. Draw the nowshe<:t and plol Ihe economic potenllal for lhe I[DA process for lhe
case where the ".15 separaled from CII. b<::fore It is recyeled and where diphenyllS
removed from Ihe
5.4-3. Acetic anh)"dndc" can be produced by lhe reacuon system
Acetone Ketene + CI I.}
100"C, I atm
Ketene CO + !CJII.
The seleCtlVlly (moles of ketene leaving Ihe pyrolYSIS reactor per mole of acetone
converted) is gIven by S - I - 4x/3 at low conversions. The desired production ratc
of anhydride is 16.58 mo[fbr at a purity of 99Y.. The cost data are: acetone =
SIS.66/mol, acid = SISJlO/moJ. anhydride = $44AI/mol, and fue[ at $4.00/10' BlLJ.
Draw the 8owshcct, and plol lhe economic potenlial.
504-:1. A process for producing acetone from isopropanol IS discussed in Exercise 1.3-4. The
desired production rale is S[.] molfbr. The feed azeotrope contains 10 mol % IPA,
and the costs are acelone - SIS.66/mo!. IPA-HJO au:otrope = S9.S3/mel, H
J
as
fuel S0.49/rnol, and H10 as waste = -SO.001/mol Draw the input-.output flow.
shcct, calculate the overall malerial balances, and ptol the economic potential
5.4-5. A simplified f10wshccl for elhanol synthesis is discussed in Exercise 1.3-6. The
desired prodUClion rate of the azcOlroplc product is 783 mol/ltr (8SA mol %
EIOH). and the caslS arc: elhylene feed mixlUre = 56.1S/mo[, process ....ater =
SO.00194/mol, elhanol as azeolrope "" 5 10 89/mol, and the fuel al S4.00.'IO" Btu
Draw the input-outpul Slructure of the f1owsheer. and plot Ihe economic potenlial
5.4-6. Styrene can be produced by Ihe reactions
ElhylbelUcne Styrene + H
J
Ethylbc:nune -. Benzene + Elhylene
Ethylbelluoc + H
J
-. Toluene + CH..
The reaClions take place al III S"F and 25 psia. We wall! 10 produce 250 molthr
of slyreoc. The cosu are: elhylbenzene ... SIS.7S/mol, styrene = 521.88/mol, ben-
zene = $9.04/mol, loluene = S8.96/mol, and fuel at $4.ll0/1()6 Btu. Wenner and
Dybdal' give colTelations for the product distnbuliOD
Mol Bellzene
:.;::;.:;:c:::::= = OJ]]x - 0.21 Sx
J
+ 2.547.r
1
Mol Styrene
Mol TolueDC
- 0.084x - o264x
J
+ 2.6]8r
Mol Slyccne
where x - styrene conversion. The elhylbc:nune feed stream contains 2 mol %
benzene. Draw the inpul-output 80wshcct and plot the economic poleOlia!.
Ketene t AceliC ACId Acetic Anhydnde
80"C, I atm
5..4-7. Cyclohuane" can be produce:d by the reaclion
Benltne + ] III "'" Cyelohennr:
The re"clion lales place at ]92"F dnd 370 psia. Pure IS used as a feed
stream, but the hydrogen Siream contams 2y. methane. The dcslred produClton raIl.'
15 100 mol/llr, ;$nd the costs art; benzene $6.50jnlol, H, = $1.]2jmol. cyelohel(-
ane = SI2 OJ/mol, and fuel at $400'10" Btu. Draw the mput-.output f1owsheet, and
plot the ecunomlc pOlenlld.L
5A-8. Ethylene can be produced by Ihe thermal cracking of ethane'
CJH.. CJH.. + HI
C
J
H
6
-. !CJH.. + CH.
The reaclions lale place at I500F and SO psia. We desire 10 produce: 87S molfbr of
etby[ene with 7S% purity. Assume that the selectivity is given by
mol CJH. Fonned 0_0381
S ... mol C
1
H" Convened - I - (I _ X)0.101
The ethane fa:d conlains Sy. CH. and costs S1.6S/mol. Ethylene at composi-
tion is worth S6.lS/mol. Fuel is worth $4 00/10" BIU. Draw the mpul-output
f1o ....shcct and plot the eronomic pOlentia!.
5.4-9. Butadiene sulfone" can be produced by the reaction
Butadiene -1 SO, -;:;:: Butadiene Sulfone
The reaC!lon tales place: III the liquid phase at 90
Q
F and 15Opsia. The costs arc
SOl = S0064/mol, butadiene = S6.76/mol, and butadiene sulfone = 58.SO/mol. We
want to make 80 molfbr of product. Draw the input-output f1owsheel, and plot the
economic potential.
5.4--10. Isooclane (gasohne)' can be: produced by the reactions
Butene -1 lsobutane -. ls00ctaoe
Bulene + Isooctane -. C
ll
The reactions take place in the liquid phase at 4SF aDd 90 psia; sa: Exercise 4.]-11.
The desired production rate is 918 mol/hr. and the costs are: ... 6514.S6/0101,
Imbutane = SI8.S9/mol, isooctane ... $]6.54/mol, and fuel = $4.00/10 Btu.
feed stream contams 8% C
1
, 80% bulene, and 12% n-C while Ihe otherconlalOS
12% C
1
, 73% iC., and IS% n-C. Draw Ihe input-output f1owsha:t. and plot the
economic potenllaL
" 1. R FlIir, M"""fact"u, Wasb'"8Io" Design Dse Stud}' No. 4,e.:lItcd by B D.
Snmb. Washlogton Universlly, SI. LouIS, Aug. I. t967
, W L BoIlC:!', Elh}/efli: Pwnt Design anJ EconomiQ. Washlllgion Unl.erslty DesIgn Case Sludy No.6.
cdned by B D Smllh. Washington Unlvcrslly. Sl loUIS. 1970
'llus problem '5 modIncd ,'enaOn of lhe 1970 AlChE Sloocm COlll<:s1 'ICC J J McKella,
E"C}"clofH!tlu, o{C,","'lcal P'OCCU1"II <1M DeHer. New 1977. p 192.
, TIm problem IS modIfied versLon of the 1977 AIChE Studcnt Conlan PfOblcm; MellE Slud"'"1
a"l/cUn. AIChE Headquuler$. 1977_
136 SEcnON H S J ~ M U Y ElCF.lClSES. AND NOMENCLATIJIU!
Nomenclature
EP
F,
F
H
F.
F.,
"
"
,
p.
PClI.
p.
p.
P.II.ct!.
5
Economic potential
Excess 1-1
2
fed to lhe process, mol/hr
Fresh feed rate of toluene, mol/hr
Makeup gas rate, mol/hr
H2 consumed by the reactions. mol/hr
Final moles of component j
Imtial moles of component j
Production rate of ben7-l:ne, mol/hr
Purge flow of methane, molfhr
Diphenyl produced, molfhr
Purge flow rate, mol/hr
CH.. produced by the reaction, molfhr
Selectivity, mol benzene produced/mol toluene converted
Feed composition of H
2
Mole fraction of H
2
in the purge stream
Stoichiometric coefficients
Extent of reaction i
CHAPTER
6
RECYCLE
STRUCTURE
OF THE
FLOWSHEET
Now that we have decided on the input-output structure orthe llowsheet, we want
to add the next level of detail. From earlier discussions we know that the product
distribution dominates the design, and therefore .....e add the details of the reactor
system. Also, since gas compressors are the most expensive processing equipment,
we add the annualized capital and operating COStS of any compressors required.
However, at this level of the synthesis and analysis procedure, we treat the
separation system asjusl a blackbox, and we consider Ihe delails of the separation
system later.
6.1 DECISIONS THAT DETERMINE
THE RECYCLE STRUCTURE
The decisions that fix the recycle structure of the fiowshccl are listed in Table 6.1,-1.
Each of these decisions is discussed in detail.
IJ7
138 HClIO"l.1 TIIA! I.lto'HNII'IL litE
SKTlON 01 OliCiSIONS THAT OflUMINE THE RECYCLE STItUCTUIE 139
TABLE 6.1_1
Decisions for lbe rec:)'cJe Slruclure
,
J
Reactor Rt;actOr
Acelone fe..'l.I
.,
.2
1
Acid recycl
I. flo.. many reaclOf S)$ICIIllI at'l' reqwrcd
T
Is there any scpala"Oll bct,,'Ota the rc:aclor .)'Slc..... ?
1. lIow llally rcc)dc $lIlt'ams.ll1t' KqwrcdT
J,. Do Wit' _Ill 10 WIt' ... u_01 one fCKUllt at lbe: rcaaOt mlct'
.. Is a &as compl"CS$(Jf IcqUIIN
T
Whal arc the o:GIIJ"
So Should 1M rcacaO! be opcntlcd adUlNtlClUy "'lIh d"1 !>Qll"C or co.,;,hn d I
ClImer requucd" &. Or Ii a 'ucnt or twat
i. Du WC """110 duft lhe cqwhbnllm 110....
7. How do the rc.llnOI alfca the CCQOlom", palenllal'
Number of Reactor Sysfems
If 0(. reactions take place at different temperalures or pressures, or if they
require different catalysts, Ihen we use different reactor syslems for these reaction
sels. For example. In the HDA process Ihe reactions
Acetonc recycle
FIGURE 6.1-1
Aczue anhydndc
both place al the same temperalure and pressure withoul a calalyst. Therefore
Ihere IS only one reaelor required. In conlrast, In Ihe reaetlOD system
the first tWO reacllons take place al a high tempera lure. whereas the third reaCtion
takes place at a lo.....er lemperature. Hence, tWO reactor systems would be required
and we could call these RI and R2.
umber of Rec)"ck Streams
From Ihe diSCUSSion above, we see thai we can associale reaction sleps with a
reactor number Then we can associate the feed streams with the reactor number
where Ihat feed component rcacts; e.g., m Ihe aohydride process, acetone would be
re.d the first reaclor, whereas acetic acid would be red to the second reactor
Similarly, we can associate the components 10 rccycle streams with the reaclor
numbers where each component reacts; e.g., in the anhydride process, acetone
would be recycled to Ihe first rcactor, whereas acetic acid would be recycled to the
second reactor (see Fig. 6.1-1).
Acetone __ Ketene + CH..}
700"C, I a(m
Kctene_CO of !CIH..
Ketene + Acetic Acid - Acetic Anhydride 80"C, I atm
(6.1-3)
F. Pnmary produCI
G. Reactalll - recycle to R2
11 Reactant - recycle 10 R2
I Reactant - recycle 10 RI
J Valuable by-product
A. Wasle byproduct
B. Waste by-product
C. Reactant recycle [0 RI
D Fuel by-product
E Fuel b)'.prodUCI
Do nOI separale two components and Ihcn remix them at a reaclor inlet
We also distingUish belween gas- and liquid-recycle streams, because gas-
reClcle slreams require compressors, which are always expensi'c. We conSider a
stream 10 be a gas-recycle stream If it boils at a lower temperature than propylene
(i.e.. propylene can be condensed with cooling water 81 high pressure, whereas
lower-boihng malerials refrigerated condensers, which require a compres-
sor). Liqutd-rttycle streams requlfe only pumps In our IOllIal design calculalions
we do notlOc!ude lbe costs of the pumps because they are usuaUy small compared
to compressors., furnaces, dlsullauon columns, etc. (High-bead. high-volume
pumps can be very expensive, SO 10 some cases we mUSI check this assumption.)
Enmple 6.1-1 Number or recycle ","earns. Conslder tblt' components and the
dcsunauons given below tn Ihe order or Ihlt'lr normal boil1ll8 points:
Now .... e can lake our hst of all lhe componenls ka,mg Ihe reaclor that has
bet'n ordered by the nonna! boiling poinls, e.g., Table 51-4, and .....e Iisl the reaclor
number as the destination code for each reg.de stream. NCllt we group recycle
componenlS having nelghbonng boilmg points If they have the same reaclor
dcstinallon. Then Ihe number of recycle slreams IS merely Ihe number of groups.
This simple procedure is based on IhlS common sense heunsllc;
(6.1-1)
(6.1-2)
1150 1300""F, 500 psia
Toluene + HI - Benzene + CH_}
2Benzene;:: Diphenyl + HI
140 SECTIOl'f'" DECISIONS nlAr OIl!TUMI!'<1' llll! II;f.CYCI.f. STl.lIOl!U
SI'OION II DFnSIONS 1M'" DETH"'lINt: 1H 'ECn:I.E SlllurlUU 141
"Y'
Toluene recycle
Gas reC)'cle
I
Purge
Compress<W
Wl - !kn7.e
Reacior Separator
Diphe Tolue
fWl
Com,....ml NBP. J)f$I;lUlliool
H, -253 Recyck + purSc ...
Cll.
'"
Recycle + pur,c
'U
.. Pnmary prodUCI

'"
Recydo: - I.ju""
o,,,,,",yl
'"
Fad by_prodUCI
En,",w 6.1-2 The components and lheir dallnallon for the ltDA
proc:as are 15 follow5'
There are four product strums (A .. B. D + E. F nd J) .nd thr rn:yclc:
.ilreams (C. G + If. and I). where lhe first .nd 111$1 So 10 RI .nd the 5IeCOnd S0e5 10 R2
Enm"" 6.1.) prl)(:ll'!I5. componenl lisl and the destinalion codes for
lhe .nhydride process arc lvcn'
d
(6.1-5)
Acetone recycle
co + a.-coo:
CO. CH", C
2
H"
I
Reactor Reactor
Aohy

.,
R2
fWl
I
Acetic acid rcqcle
f1GURE 6..1-2
HDA recydt: Rno<:lurc- [F,-. J, /of Doug/#$, JI J51 (1985)J
Acetic add feed
and if the kinetics match the stoichiometry. then the usc of an excess of lsobutanc
leads to aD improved seleclivlty 10 produce isooctane. The larger the excess, the
greater the Improvement III the seleetivily, but the largel the cost to recover and
rcqcle the isobutane. Thus. an optimum amount of excess must be determined
from an economic analysis
The usc of an excess component can also be used to fortt another component
10 be close to completc con\'crSlon For eJlample., III the production or ph<Kgene
Fuel product
Fuel
Fuel b)'-product
IUCUnt - complcldy oon""nrxl
RelICWlI - recyck to RI - Dqll;d
Reacl.l.rll _ to R2 - liquid
Primal')" prodlld.
-)126
-250&.6
-ISH
-"2.1
133.2
2.)
181.9
co
0<.
C,H.
Ketcne
Aoelone
AoetK .00
Aoetic .ohydrilk
ThUs. there Ire three product streams purF- br:tutne. Ind dlphcnyl and ''''0
recycle ilreams. II ... CII" and where the fint is Isas and ,he second,s I.
IKju,d A recycle nowsheet IS given In hg. 6.1-2. and /I shows tbe reaclor and the
recycle gas oomprCSllor.
Thus. lhere are IWO prodllCl Sireams.. CIi" + CO + C;.H" .nd anhydride. and ,wo
Iiqllid_rccydestreams.re returned to chft'ercnt reactors: Ketone is re=yelcd 10 RI, aad
Ia:tic acid is rccydal to R2. A f10wshm is shown in FiB- 6.13.
Excess Ructanls
In some cases the use of an ucess reactant can shirt the product distribution. For
example, if we write a very oversimplified model for the production of isooctane via
butane alkylation as
Butene + IsobuLane - booctane
Butene + lsooct.t.ne en
(6'-4)
FIGURE 6.1-J
Auc anhydnik lec)C1e
SEcnON t..l IlECYCLE MATU.JAL BAlANCES 14]
which is an mtermedlate In the produUlon of dl-isocyanate, the product must be
free of Cll. Thus, an excess of CO is used to force the Cll conversion to be very
high
Similarly, the lise of an excess cun be used to stufl the equilibrium conversion.
For example, in the producllOn of cyclohexane by the reaction
we want to obtam equlllbnum conver)iom very close 10 unity so that we can
ubtain a high conver)ion of benzene and avoid a benzcnecyclohexane distillation
separation (the boiling points are very close together). We can shift the equilibrium
conversion to the right by using an excess of III at the reactor inlet.
Thus, the molar rallo of reactants at the reactor inlet is oflen a design
variable. Normally the optimum amount of excess to usc involves an economic
trade-off between some beneficial effect and the cost of recovering and recycling the
excess. Unfortunately, there arc no rules of thumb available to make a reasonable
guess of the optimum amount of excess, and trerefore we often need to carry out
our economic analysis in tenns of this additional design variable.
Thus, the feed to the reactor is
Purge
Fr(l - x)
Benzene
Reactor
Separator
FT(I - .) System
F
Dipheny
T
Toluene
[=l
F,-,
FIGURE 6.2-1
IlDA, liqUJd recycle.
(6.1-6) BellLcne + 3H
l
Cyclohexane
This same malerial balance is always valid for the limiting reactant when there is
complete recovery and recycle of the limiting reactant.
In some cases, some of the limiting reactanl might leave Ihe process in a gas
recycle and purge stream (ammonia synthesis), or it may leave with the product
(ethanol synthesis), If we consider a simplified \'crsion of the ethanol process, lhe
reactions arc
Heal Effects and Equilibrium Limil3lions
In general, the reactor flows need to be available in order to evaluate the reactor
heat effects. Also, in many cases, equilibrium calculations arc simplified if we have
C<ilculated SOffie of the reactor flows_ Thus, we defer our discussion of these topics
until we have discussed procedures for estimating the process flow rates.
f
,
_ FT
,-
X
(6.2-2)
(6.2])
6.2 RECYCLE MATERIAL BALANCFS
Our goal is to obtam a quick estimate of Ihe recycle flows, rather than ngorous
calculations. We have not specified any details of the separation system as yet, and
therefore we assume that greater than 99 %recoveries of reaclants are equivalent 10
100% recoveries. This approximation normally introduces only small errors in the
stream l'Iows.
CHJCH
J
+
+ HlO
We suppose that we want to produce 783 mol/hr of an EtOH-HlO azeotrope that
contains 8S.4 mol %EtOH, from an ethylene feed stream containing 4 %CH.. and
pure water. A recycle f10wsheet is shown in Fig. 6.2-2 for the case where we recycle
Ihe diethylether and the water.
The overall material balances start with the production rate of the azeotrope
Limiling Reaelanl
First we make: a balance on the limiting reactant. For the HDA process (see Fig.
62-1), we lei the flow of toluene entering the reactor be FT' Then, for a conversion
x the: amount of toluene leaving the reactor will be F,.{l- x). For complete
recoveries in the separation system, the flow leaving the reactor will be equal 10 the
recycle flow. Now if we make a balllnce at the mixing point before the reactor, the
sum of the fresh feed toluene FFT plus the recycle toluene will be equal to the now of
toluene into the reactor, or
p.
zu
= 783 moljhr
This contains
P"'OIl = 0.854(783) = 669 mol/llr EtOH
Then the amount or water in the produci stream IS
(6.2-4)
(6.2-5)
(6.2-1)
(6.2-6)
144 SECTION '2 ROCYClE MATEJJAl IALANCU
SECTION n RECYCU! MATERIAL I\LANCa 145
Other Reaclants
After we ha\-e estimated the now of Ihe limiting reactanl, we usc a spcc1ficalion of
the molar ratio at the reactor inlet 10 calculate the recycle nows of the other
components. For example. 10 the HDA process (sec Fig. 6.2-4). Ihe 10lal amount of
h}drogen entering the reactor is the sum of the fresh feed hydrogen l'fIIFG and the
recycle hydrogen RGV'H We sho.... ed above that the amounl of the hmllmg
reactant, toluene, entenng the reactor IS FniX. Thus. If .... c let the molar rallO of
b}drogen to toluene at the reactor inlet be1\1R. we find that
hHFG+Y,lJRf<=MR(F;r) (62.10)
or R _....!..!....-(MR_ Y,lJ ) (6.2.11)
G SXY'H x Y"H Y'H
we specify the deSign variables X.l'''H. and /of R, we can sohe Ihis equation for
the recycle gas now R
G
.
zeotropc
I
1
C
2
H., CII.
E'
Reactor
Separator
system
A
H,O
I
Ethe<
I
H,O
ThUS. from Eq. 6.2-3 and the rC$ulls above, the required feed rale of water, .... hich is
the hmiting reactant, is
FIGURE 0..2
Elbaoollynl'---
FII,o - y..... p.uo + (I - Y...o)p..u - 669 + 114 >= 783 mOlfhr (6.2-7)
Suppose that we lei the water leaving witb the product be F.... f' - 114 and the
fresh feed waler required for the reaction be F....... Now. referring to the schematic
in Fig. 6.23 for water, we let tbe amount entering the reactor be F ... , tbe amount
leaving tbe reactor be f...(1 - x). tbe amount leaving with the product be f .... f'. and
the amount recycled be f...(1 - x) - f .... ,.. Then a balance at the milling pomt
before the reactor gives
Design Heuristics
There are no rules of thumb available for selecting x for the case of complex
reactions. Similarly. there arc no rules of thumb for selecting the purge composition
y,.. or the molar ratio M R. For the case of single reactions. a reasonable first guess
of conversion is x = 0.96 or x ... 0.98x.,,:
For single reactions, choose x = 0.96 or x ... O.98x." as a first guess. (6.2-12)
This rule of thumb is discussed in Exercise 3.5-8.
This result is identical to Eq. 6.2-2, that instead of the fresb feed rate we
substitute only tbe amount of matenal that enters into the reaction. A Similar result
is obtained for the case where the limiting reactant leaves with a gas purge stream.
F,. + F
R
F
F(J - x)
F
p
Reactor Separator
. p.
"yl
j
. Pc
MR

Reactor
Separator

I
F
T
Diphe
(6.2.8)
(6.29)
(f.... ,. + F......) + [f..(1 -x)- F.... ,.] = F...
F... - F....../x
so that
F(I - x) - F
p
Toluene
f<ed
FIGURE 6.1-3
F.lh.noI5ynl!lesl$.
FlGUR 6..l.-<t
ItDA. PJ .ecydc.
'KIIO!" 6J HEAl EHRTI 147
However, the H
2
and benzene 80ws have been detennined by using the first
reaction and the purge calculations; so we can usc the equilibrium expression to
calculate Ihe dipbenyl 8010' at lbe: reactor exiL
Reversible Byproducis
If we recycle a by-product formed by a reversible reaclion and lei the component
build up to liS equilibrium such as the diphenyl in tbe UDA process
+ U
1
or the dlethylcthcr III ethanol synthesIS CEq. 6.2-3), then we find the recycle flow
by uSing the equilibnum relallonshlp at the reaclor exit That is, at the feal:tor
e1l.11
Enmplt 6.3-2. can be produced by tbe dehydrogenauon of Isopropanol
(63-2)
(6 ]1) Q. '" All. 1"'1" ( 21,5JO)(27l) = x 10" Btullr
.... here All. 1.lhe heal ofre.K'lon at IZOO Fall<! 5(X)pila and hu.lls liberated by lhe
reaction
Elilllnple 6.3-1 IlIlA hom the nlllleriat balances for the /lOA
process, ....'e found thaI only small "mounu of dlphenyt were produccd in the range of
where we obtallled pro61able opcr,lIlon. If .... e want to estimate the reactor
beat load for a case where x _ 0.75, P," 265, and f ff = 27] mol/hr, we might
neglttt the second redctlOn and wnte
If we desire 10 p-oducc: 051) mol(hr of acctone, tben o5lJ mollbr of IPA 15
The heal of rClllcllon at o5-ro-F Ind t 11m 15 205.800 81U/moI., so (be fC.iIctor heat load 15
(6.2-13)
K = (Dlphenyl)(H
2
)
., Benz.cne
2
63 REACTOR HEAT EFFECTS
Q.. - 25.800(51 3)- 1324 x 10"' Btu/hr (6 J.3)
We need 10 make: a dCClSlon as 10 whether tbe reactor can be opcrated adlabatl-
caUy, wilh dlfect healing or coolin& or whether a diluent or heat carrier is needed.
10 particular, if .....e need to lDlroduoe an exlraoc:ouscompooent as a diluent or heat
carrier, lhen our recycle material balances, aod perhaps even our overall material
balances, will have 10 be changed Moreover, we need to make this decision before
we consider the specificallon of lbe: separation system because: lhe decision 10 add
an extraneous component nonnaUy will affect Ibe design of the 5C?3-raLion system.
To make the: decision concc:nllog tbe: reaClor heat effects, 6rst we eslimate the
reaClor head load and Ihe adiabatic lemperature change. These calculations
provide some guidance: as to (be dlffu:ulty of dealing wilh tbe reactor heal effecls.
Similarly, weconslder any lemperature constraiolS imposed on thedcslgn probkm.
and bealiS consumed by lbe endolhenmc relictIOn..
Adiabalic Temperature Qange
Once: we have delennloed the reaClor heal load and lhe rate through lhe:
reactor as a function of the design variables, we can eslimale tbe adiabatIC
temperalure change from lhe expression
(6.3-4)
EliImpk 6..3-3 ItDA pr_ The flows and heal capacina of the reaelor feed
stream for a C3$e .... here x _075 and ",,, - 04 are gwen belo""
Reac(or Heal Load
For single reactions we know that aU Ibe fresh feed of the limiting reaclanl usually
gets converted in the process (Ihe per-pass conversion mighl be small so Ihal Ihere
is a large recycle flow. bUI all the fresh feed is converted except for small losses in
product and by-product streams or losses in a purge stream). Thus, for single
reactions we can state that
Sttum
Mahupps
au
TolucDc reed
recycle
<9,
337t
m
"
C,. 81u/(moI-f)
09S(7) + OOS(IO.I) _ 1.16
04(1) 'i' 0.6(101) - 8 86
...,
48.7
Reactor Heat Load = Heat of Reaction x Fresh Feed Rale (6.)-1)
where the heat of reaction is calculated al Ihe reactor operating conditions.
For complex reactions, the exlent of each reaction will depend on lhe design
variables (conversion, molar ratio of reaclanlS, temperature, and/or pressure).
Once we select the design variables, we can calculate the extent of each reaction and
then calculate the heat load correspondmg to the side reaclions. Hence, il is a
simple mailer 10 calculale reaClor heat loads as a function of the design variables.
Some guidelines for reactor heal loads arc: diSCUSsed laler in this section
Theo. rrom Eumple 6]-1 and Eq 63-4 wllh T.
I
__ I 150F.
-5K78 lC 10" _ (273 + 91)48.7 + 496(7 (6) + -
T.. _,_lt$O I IIS_1265F (63-S)
This ... alue IS below lhe constralnl Oil Ihe reactor Ull lemperature of IJOOfiF (sec
Eumple 4 11) Also, the calcuhtllOn IS nOI very sensitive to Ihe C, values or to the
flows llic: relallvely small lemperature rise is dllC to the large gas-recyck flow
148 SECTION U HEAT F.FFl!CJ1i lIMll"nONS 149
the amounl of heat-transfer surface area Ihat we can fit into a reactor To get some
"feeling" for Ihe magnitude of this area, we consider the case of a high-temperature
gas-phase reaction, and we let U = 20 Dlu/(hr. fl1. oF) and AT=- SODF. Then. for a
heat load of I x BlUthr,
The maximum heal transfer area that fits InlO the shell of a floatlllg-head heat
exchanger is ill tbe range of 6000 10 8000 ftl. Thus, to use a single heal exchanger as
a reactor, when we are altempting to remove or supply the heat of reaction by
direct heating or cooling, the reactor heat loads are limited 10 the range of 6.0 to
8.0 x 10
6
Bluthr.
)'II
J
= 0.8
)"1: = 0.7
Hydro-
710
)'11,
= 0.6
cracking
U
)'u, = 0.5 limit
0

YIl,
0.4
u

680
0]
)'11
1
"
u
'"
)'11, = 0.2
J 650
)'''1
= 0.1
1000 ftl (63-7)
x
FIGURE 6.3-1
Re.aClor cullcmrcralurc.. [From J. M. Doug/D.<. ,fIChEJ, 31 J.B (1985).)
A plOI or the reactor eXit temperature ll5 a function of thc design variables is
givcn in Fig. 6.31, and we see that in certain cases the constraint on Ihe reaclor exit
temperature is violaled
Eumple 6.3-4 IrA process. If Ihe feed stream 10 the acetone process described by
6.J-2 IS I.PA-HJO 117.eotrope (70 mol %IPA) and if ...e recycle an az.cotropic
mlxtllte, then It IS easy to sbow that 22.0 mol/hr of water enters with tbe feed. Also, for
a conversion of,'( 0.96, the recycle Row will be 2.1 mol/hr of IPA and 0.9 mol/hr or
water. If the reactor inlet temperature is 57rF, then from Eqs. 6.3-1 and 6.3-4 Ihe
adiabatic temperalure change is
Q" = 1.324 x 1ij6 _ 51.3 + 22.0) + (2.1 + 0.9)](22.0)(572 _ T.->
or T"._, = 572 - 788 __216F (6.3.6)
Clearly, this is an unreasonable resull. Thus, instead of usin8 an adiabatic
reactor, lJo'e altempt to achieve isothcrmal operation by supplying the heat of the
reaction 10 the process. In p3nicular, I.'e mighl attempt 10 pack the tubes of a heat
exchangcr with a eatalysL
Heurislic for Heat Loads
If adiabatic operation is not feasible, such as in the isopropanol example, then we
altempl to use direct heating or cooling. However, in many cases there is a limit to
Heal Carriers
The reaclOr heal load is oflen fixed by the fresh feed rate of the limiting reaclanl (if
only a small amount of by-products is produced so that the secondary reactions are
unimporlanl). The adiabatic temperature change depends primarily on the flow
through the reactor. Hence:, we can always moderate Ihe temperature change
through the reactor by increasing the flow rate.
If we desire to moderate the temperature change, we prefer 10 do this b)
recycling more of a reactant or by recycling a product or by-product. However.
where this is not possible. we may add an extraneous component. Of course. the
introduction of an extraneous component may make the separation system more
complex. and so we normally try 10 avoid this situation.
In the HDA process (sec Eq. 6.3-5). the methane in the gas-recycle stream
(60% methane) aels as a heat carrier. Thus. if we purified the hydrogen. recycle
stream, the recycle flow would decrease and the reactor exit temperature would
increase. If this exit temperature exceeded the constraint of I 300F, we eoul<l, no
longer usc an adiabatic reactor. !,nstead, we would have to cool the reactor,
increase: the hydrogen recycle ftow, or introduce an extraneous compound as a heat
carrier.
A similar behavior is encountered in many oxidation reactions. If pure
oxygen is used as a reactant, the adiabatic lemperature rise is normally so large thai
problems would be encountered. However, if air is used as the reactant stream, the
presence of nitrogen moderates the temperature change.
6. EQUILIBRIUM LIMITATIONS
In numerous industrial processes equilibrium limitations are important. We
discussed a procedure for estimating the nows of reversible by-products when they
are recycled and allowed to build up to their equilibrium levels at thc reactor outlet.
Our focus here is the primary reaction.
ISO stCIIOH U U)I'IUUJUM UMfI ..TlQNl; SECTKlN.' EQUtUUIUJot UWrr4110NS lSI
he13ne
H,.
H
4
, N
2
Reactor
Separalor
Cyda

Equilibrium Comen,ion
We can use our prC:YlOUS procedure f.:lr cakulaling the process: flows iiS a funcllon
of the desIgn Yanables (com'ersion, molar .alio of reaclanls, elc.). Then \o\,e can
subslllute Ihese flows mlo lhe eqUllibnum rclallonship 10 sec: whether the conYer-
sion we selecled is above or below the equilibrium value Of course, If il cllcc..:ds thc
equilibrium conversion, the result has no meaning.
In most cases, however, II is necessary to delermlllc the exaci ... alue of the
equilibrium convcrsion (as a funclion of Ihe design unables), because Ihis \aluc
appears In Ihe kinetiC modcluscd 10 delermloc the reaClor SIU- Our samc general
approach can be applied, bUI nonnaU) a trial-and-c:n-or solulion IS reqUired We
ilJU5trale tbis Iype of problem by considering Ihe cyclohcxanc process.
Eumpk 6.4-1 CycIGMUM l""ooucrioQ. CyclohcxaDc can be produced by lhe: reac-
lion
FlGUll U-I
Cydobuanc lIo""s.hcci
(6.4-1)
We: cow;Ide:r a cue where: ....e dQ:m: 10 produce: 100 mol/br 01 C.II, J .....llh a 99
punty A pure: benzene fc:cd Slrum IS aval.lable, and tbe: hydrogen makeup stream
cont&l.Jl5 97.5" HJ. 2.0'Y.; CH., aDd 0 5" N J A Dowsbect (()I" Ibc: recyde stroclure: IS
iho....n in Fig. 64-1 for a case \lihere: we recyde 5Omc: of the: bc:nzcnc (",hieh is not
lhe best f1owsbcc:l)
K
Ic "cYc (64-15)
'-'1-.- iii .
J.. P... ".v.y.Y.
From Ibe: H'aJ/unglo" U"iWlrsity Df's/f}" CaJt Study No.4, p. 4-3, Part II,
(6.4-17)
(6.4-16)
(U-14)
(6.4-12)
(6.4- Il)
(6.4-11)
(6-4-10)
.' JC.
J.JlP... - --(MR 3)
J - JC. Yr.
"" - 1
Cyclobc:lIaoe _ p.
TOlal Ilow _ + (MR 3) _'_]
x x Yr.
RlltioaIIl.i,
The.
(6.4-4)
(6.4-5)
(6.4-2)
(6.4-3)
(6.4-1)
(6.4-6)
501,,,io..
O.enU bal.DCb. Auume no losses. Theo
Production of p. = 100
Benzencfrcsbl"ccd: 1",. _ p. _100
Assume ....e: usc: a ps recyc:k and a purse stream. Let
F ., EJ:ccu H. Fed to Proo:;:all;
Total H, Feed .... lP. + F. _ 0.9751"..
.. F.
PurseoomposlllOGofHJ:Y,,,"'" 00 l
F
r
+2F..
1- Y,.
Makeup au rate: FG - lP'
0
.
975
Y,.
Purge nile: PG - F. +O.02SFG
p.
Benzene: fcd to reactor: F. _ -
x
(6.4-8)
Dir-nria ... SlDce benzene and cydoheuDe are very dose boilers, we would like
10 avoid a benzcne<ycloheune distillallon separallon. This can be accomplished by
operating Ihe reactor at a suffiCiently high conversion Ihal we can kave any
leI molar ratio of II, 10 benzene be: At R Then
1 (MRP. )
Recycle: ps now: R" _ --- 09751"..
y,. x
(649) J L Fan, Cydobu..I.pc M.a.nufaetun. WuhmJloa U....'fUSIly Da.lp Cue Sludy No 4. edJlcd
by B 0 51mlh, Wasbllllloa UIlUw:nIlY, 51 L.ows, 1967
152 SI!(."TlON U EQUtUUllIM U""rlAnONS SECTION ., CO""PUSSOIt OUJON' AHO COSH 153
unconverted as an impurity in the product.llo...e'er, to obtain high bo:nzene
conversions, we must foree the equilibrium conversion to be very close to unity.
Equation 6.4-11 indicates the dependence or the equilibrium conversion on tbe dcsign
variablcs.
Economic From Eq. 6.4-17 we see that we can IncreaS(' the equilibrium
converSion by increasing the reactor pressure P_, by increasing the molar ratio of
hydrogen to benzene at the reaClor inlet M R, or by decreasing the reactor temperature
(thO' reaclion is e)(olhermic). 1I0"e"er, high reactor pressures correspond 10 a large
feed comprenor and more upensive equipment becauS(' or an Increased wall
thickness. Large molar ratios of h)drogen correspond 10 larger gas recycle Dows (see
Eq 6.4.9) and therefore a more expensive rt'Cyde compressor. Lo""er reactor
temperatures correspond to larger reactors, because or t1le decreased reaction rate.
Thus, an optimization analysis is required 10 determine the values of P x T:
MR,and'PR' . , '
APflrollim.t.. mod..t If we expect that x. will be dose to unity in Eq 6.4-17, then
we can ""Tite
Re__ersible Exothermic Reacfions
There are several important industrial reactions that are reversible and exothermic.
For example.
High temperatures correspond to small reactor volumes, but for tbese reactions the
equilibrium conversion decreases as the reactor temperature increases. Hence,
these reactions are often carried out in a series of adiabatic beds with either
intermediate heal exchangers to cool the gases or a bypass of cold feed to decrease
the temperallJres between the beds. With these procedures we obtain a compromise
between high temperatures (small reactor volumes) and high equilibrium conver-
sions.
,-"cO'.,. [ MR - 2 J'
x. "" I - I CMR 3)YPH
which provides a SImpler modelta use in preliminary optimizatiOD studies.
(6.4-18)
Sulfuric acid process; SOl + 10l SOJ
In ammonia synthesis,
Water'gas shift: CO + + H
1
Amonia synthesis: N
1
+ 3H
1
2NH
J
(6.4-21)
(6.4-22)
(6.4-23)
DiJuenfs
From tbe discussion above we have found that temperature. pressure, and molar
ralio can all be used to shift the equilibrium conversion. HowcvC'r, in some cases an
extraneous component (a diluent) is added which also causes a shift in the
equilibrium conversion. For example, styrene can be produced by the reactions
wbere the reactions take place at about ll00"F and 20 psia. The addition of steam
(or methane) at the reactor inlet lowers the partial pressure of styrene and H
2
and
so decreases the reverse reaction rate in Eq. 6.4-24. The steam serves in part as a
heat carrier to supply endothermic heat of reaction.
Steam is often used as a diluent because water-hydrocarbon mixtures are
usually immiscible after condensation. Hence, the separation of water caD be
accomplished with a decanter (and usually a stripper to recover the hydrocarbons
dissolved in the water, if the water is not recycled).
6.5 COMPRESSOR DESIGN AND COSTS
Whenever a gas-recycle stream is present, we wiU need a gas-recycle compressor.
The design equation for the theoretical horsepower (hp) for a centrifugal gas
compressor is
Separator Reactors
If ODe of the products can be removed while the reaction is taking place, then an
appareotly equilibrium-limited reaction can be forced to go to complete conver-
sion. Two examples of this type are discussed now.
Enmplt 6.4-2 Acetone prodll(tiOlL Acetone can be produced by the dehydrogena_
tion of isopropanol
Isopropanol Acetone + H1 (6.4-19)
in the liquid as well as the gas phase. AI JOOoF the con"ersion ror
the liquid-phase process is about.l:... _ 0.32 However, by suspending tbe catalyst in a
high-boiling solvent and operating the reactor at a temperature above the boiling
point of acetone. both H1 and acetone can be removed as a vapor from the reactor.
Thus, the equilibrium conveBion is shirted to the right. A series of tbree continuous
stirred tank reactors, with a pump-around loop containing a beating system that
supplies the endothermic heat to reaCiion, can be used for the process.
Example 6.4-3 ProdlKtiOIl of Clh)'1 acrylau'_ Ethyl acrylate can be produced by the
reaction
Acrylic Acid + Acrylate + HaD (6.4-20)
Both acrylic acid and ethyl acrylate are monomen, which tend, to polymerize in the
reboilers or distillation columns. We can eliminate a column required to purify and
rec::ycle acrylic acid from the process if we can force tbe equilibrium-limited reaction to
completion, say, by removing the water. Hence, we use an eJ;OCSS or ethanol to shift the
equilibrium to the right, and we carry out the reaction in the rcboiler of a rcctirying
column. With this approach, the ethanol, water, and ethyl acrylate are taken
overhead, and the acr)"lic acid approaches unity.
Ethylbenzene Styrene + H
1
Ethylbenzene Benzene + Ethylene
Ethylbenzene Toluene + Methane
(
J.OJ x 10-') Q [(P_,), ]
hp = Pin In -- -,
Y Pi.
(6.4-24)
(6.4-25)
(6.4-26)
(6.5-1 )
For a thrcc.-stage compressor w!th IOten:oohng, Ihe ..or'" required IS
liABLE o-a]
Values of l'
MonotoDIC pieS
D'illornlC
More complu JIISQ {CO,> ("II.)
Other a.uc:s
S-u Tilles iro. J aDd 0 (i Jo<<iaa,
Clw.....t "'..uJ,J EL-.c. ld 1. IkLlo.
No.... VorL. !t}S. p Ull
where Pi. - Ibljlll, Qi.. _ltJ/mm, and i' = (C,/C. - I)/(C,/C.). The exit tempera-
ture lrom a compression stage is
The inlermediate pressures that minlluiL.e the war'" arc deicrtlilned from
oWorL-: ;. ;. 0
OP
l
aP
J
which leads to the results
(65-4)
(6.5-5)
(6.5-2)
T_ (P_)'
T. Pi.
(where Ihe temperatures and pressures mUSI be in absolule umls) Values ofy thaI
can be used for firSI estimales of designs arc given in Table 651.
Thus. ...c obtam anolher design heurisllC
The compression ratios for each 10 a gas compressor should be: equal
(6.5-6)
Annulilliz.ed Installed Cost
Thc brake hOT5epower bph IS oblamed by IDlroducing lhe compressor effictency
inlO Eq 6.5-1.
Then, Guthric's correlalion (see Itcm 4 in Appendix E2) or some equivalent
correlation can be: used 10 cak:ulale the installed cost for various Iypes of
compressors:
Operating Cost
By dlvldlllg the brake horsc:po\loer by the driver efficiency, we can calculale the
uulny requirement. Then from the ulllllY cost and uSlllg 8150 hrlyr, we can
calculate Ihe operatmg cost
..... here F. IS gl\cn m Appendix E.2 and M & S = Marshall and Swiflmflation index
(whlCb is published eacb month In
To put the mstalled cost 00 an annualIZed baslS, "'c I.DtrodUlX a capital
charge faClor of i yr. Note Guthrie's correlations and capttal cba.fge factors are
dtsCUSSCd III detail In Chap. 2.
Efficieot}'
For first designs., we assume a compressor efficiency of 90% 10 accouDt for fluid
friClion in suction and discharge vahes. ports, friction of movmg metal surfaces,
fluid IUrbulence, etc. Also \loe assume a dnver effiCiency 10 account for the
conversion of Ihe inpul energy to shaft work.
Spares
Compressors are so expensive that spares arc seldom provided for centrifugal units
(although reciprocating compressors may have spares because of a lower service
factor). In some instances two compressors may be installed, with each providing
60% of Ihe load, so that partial operation of the plant can be maintained in case
one compressor fails and additional flexibility is available to respond 10 changes in
process Hows.
l\1uhistllge Compressors
It IS common practice to usc mullistage compressors. The gas is cooled 10 coohng-
waler temperatures (looeF) belween stages Also knockout drums arc inslalJed
between slages to remove any condensale. It is essential to ensure lhat no
condcnsation tales place Inside the compressor, since liquid droplets condensing
on the vanes wh.lch arc rotating at \ery high speeds might cause an Imbalance aod
wild vibrallons
hp
bhp"'"' 0.9
(
M&S\
Installed Cost = 280r517.5XbhP)OU(2.. 1I + F.)
(6.5.7)
(6.'-8)
-
30

..
180
E
90

2.0
"
4S
6
u
..
,
<
1.0
<

1i
02
FlGLRE '5-1
5ertlll,"17 of nocyde o;Imp..:uor AI'
Sensith'ity
At the preliminary stages of a design.....e do not have a complete Rowshttt.
we cannot obtain a good estImate for the compression ratio P_,IP,. for recycle
streams. Our approach simply to make a guess and then to evaluate the
sensitivity of lhat guess In most cases. the results are fairly Inscnsitlve. An example
for the HDA process with x = 0.75 and Y,.H = 04 is given in Fig. 6.5-1.
6.6 REACTOR DESIGN
At the very early stages In a new design, a kinetic model normally IS not available
Thus., we base our material balance calculations on a correlation of the product
distnbutlon. Also, we assume that we will use the same type of reactor in tbe plant
that the chemist used in the labol1ltory, and we often base a first esttmate of the
reactor size on the reaction half-life measured by the chemist. For adiabatic
reactors we might base the design on an isothermal tempel1lture which is the
average of the inlet and outlet tempel1lturcs or an average of the inlet and outlet
I1Ite constants.
This type of a kinetic analysis is veT)' crude, but ill most cases the reactor cost
is not nearly as important as the product distribulion costs. Thus. again, we merely
look at sensitivities until we can justify additional work and a kinetic model
becomes available We estImate the costs of I"lug flow reactors in the same way
we do pressure (sec Appendix E.2), and we annualize the inslalled cost by
introducing a capItal charge factor of 1yr (we discussed capital chargc factors in
Chap 2).
TABLE ......1
Design guide.lines ror rc.etors
SlD&!c reaCl.ons (_ lutOgllal,IK)
A. ISOlbnmaI _"")'11150" plUI flo.. n:aclt>.
B "duohanC'.
I MUI fIo_ rI the rUdJOll nile docTc:II'ICS ......
2- CSTR llor .... umu... leKUOIl r"" follo_ed by I plUI ""... K'CbOII
II S,nt\l: n:vnSlbk lualOM adlllbahc
A Mlumum lcmp",a,ulc if endolhor:nnoc
8 A !lena of ...hab.tlc beds "lIh .. dccTc;uon'laTIJlC'llurc fll'Ofik II' c10thcmuc
III Para1ld relICUOn5 oomJl'Ol!I,on dfcclS
A. FOf A. _ It and A __ S (WU1C). ",be,t: lh<: nt.o of lhe: rt:lIClwn nICS IS fS -
(1c,/",}C';.' .,
1 If., > II,. kccp C.. hIgh
... UK boItch Of plUI no....
It Hlgb J'l'es!Iu,c, chlnlnJIlc IlM"lS
c. Avo,d Iec)":1t: or products
Can use. tmllI rQC!Of
2 If., <0,. lOltp C.. I.....
1I. UK I OTR ....-1I,h I tllgb
burp reqde or products
c 1.0.. J'fCSiSU.e, add malS
II! Need. btSC reactor
B F.,.. A + B _ R (da-lIed) and A .... B _ S (_"aSIC), _hen lhe: nUO of the ralCS IS '. '5-
(.,It,)C7.' .,
1 lfa,>".alldlt,>b,.bothC.. andC.h'lh
1. If I, < lI, and It, > It,. then C. low, C. hIp-
) 11 "I :> '" and !I, <; !I,.lben C, h,p. C,Io-
4 If '" <; '" and !I, <; !I,. both C. and C.low
5. Se-. FIll- 6.6-1 rOf \aoo.... rUCIO, confiRuauons
IV Col1.KC\ltJve -c:omposlt10n effect.
A. A _ R R _ S lhe of streaml ..lib dlffe,en, compol1Uon$
V PantJlel t ....cllons tempenlllure effects './" - (.,/.,>!(C. C.>
A. If E, :> E. \1.$0: a blSb temper.ture
B. If E, <; E" \I.$C an .ncrusin& lempttalure pt'ofile
V1 Comeolue rudJOftI-t.,...peralure dIecta A It S
A. If E, :> E" u-=' decreasullto;mperalure pt'ofik-aO'l ery KrtIIU.-e
B. If E, <; E,. __ low u:mperalurc:.
Reactor Configuration
Siooe tbe product distribution can depend on the reactor configuration, we need to
determine Lbe best configuration. A sct of design guidelines has been published by
Lelenspicl. For the sake of completellcss some or tbese guidelines are reviewed in
Table 6.6-1. As this table indicates in some cases we obtain complcx reactor
configurations; sec Fig. 6.6-1.
158 su:noi'l" al!cr(;u. UX>I'IOMIl: t:\AlU.-T10N
A-f)
B I
-....4rC, h.gh. C. low C, h,gh. C. low
(c) - S"n wnh A A
o. 0,6 0.' 0.3 0.4
Conversion x
0,2
Ie",,) )
//

YPH
/
---
....... 0.2
/
...... 04
/
'"

........ 06
/
-&- 08
1/
"\"
'\
7
2,000,000
1,500,000
1,000,000
500.000
o
-500,000
-1,000,000
-1.500,000
-2.000,000
-2.500,000
0.\
FIGURE 6.1-1

A
C.bolh low C,. C. bolh low

C6>
FlGUln: UI
Pal1l!ld R:K't101l5.. (F._ 0_ CJo-uaM R..<IC'_ EJog-fItOI, Ztl at.. Woky, "...., Yor.\:, 1911,
c.... 7-" )
rapldl) restnctmg (he range of the design vanables wbere Il>e obtam a POSltl\C
profiL Thus, our calculations for the separalK>n system and tbe hen-cxchanger
network arc Simplified
6.7 RECVCLE ECO OMIC EVALVAnON
Our economic analysis for the input-output structure considered only the Slream
costs, i.e., products plus by-products minus raw-material costs. The results for tbe
HDA process mdlcated Ihal the most profilable operation was obtained when the
conversion was zero (we made DO diphenyl by-product) and when the purge
composition of hydrogen was uro (we both purged and recycled pure methane).
However. when we consider the recycle, Eq. 6.2-2 shows that we nocd an
infinite recycle Dow oflolueDC when the conversion is equal to zero, and Eq. 6.2-11
shows that an infinite recycle flow of gas is required wben the hydrogen purge
composition is equal to zern. Thus, if we subtract the annualized reactor cost and
the annualized compressor costs, both capital and power. from the economic
potential, then we expect to find both ao optimum conversion and an optimum
purge composition. Figure 6.7 1 shows this result for the I-IDA process. (We
would also subtract the annualized capital and operating cost ofa feed compressor.
if one was required and we did not include it in tbe Ievel-2 calculation.)
The values for the optimum shown in Fig. 6.7-1 arc DOt the true optimum
values because we havc not included any separations or heating and cooling costs.
Hence, we expect that the true optimum economic potential will be smaller and will
be shifted to lower recycle nows. However, we can sec. that our simple analysis is
6.8 SL'MMARY. EXERCISES, A D
NOMENCLATURE
Summary
The decisions that need to be made to fix Ibe recycle structure of the Oowsheet arc
as follows: .
I. How many reactors arc reqUlrcd? Should some componenlS be separated
betwccn the reactors?
l. How many recycle strcams arc reqUIred?
3. Do we want to usc an excess of one reactant althe reactor inlet?
4. Is a gas-recycle compressor required? How docs it affect the costs?
5. Sbould the reactor be operated adIabatically, with direct heating or cooling, or
is a diluent or heat carrier needed?
6. Do we want to shlft the equilibrium cOllverslOn? 1I0w?
7. How do the reactor costs affect the economic potential?
160 ncnoN II SUMMAn, '.JlI':llClSf'.I, "NO NOMI!NOATlJIU!
The design guidelmes we use to make some of these decisions for first designs
are as follows:
I. H reactions take place at different temperatures and pressures and/or
require different catalysts, then a separate reactor system is required for each
operating condition
2. Components recycled to the same reactor that have neighboring boiling points
should be recycled in the same stream
3. A gas-recycle compressor IS required If the recycled compoDC'nts boil at a
temperature lower than that of proPllene.
4. If an ellocss reactanl IS desirable, there is an optimum amount of the eJlCC5S.
5. If the reactor temperature, pressure, and/or molar ratio are changed to shirt the
equilibnum conversion, there must be an optimum \'alue of these ariables.
6. For endothermic proc:esses with a heat load orJess than 6to 8 x 10
6
Btu/hr, we
use an isothennal reactor With direct heating. For larger heat loads we may
add a diluent or heat carrier
7. For uothennic reactions we use an adiabatic reactor iflhe adIabatiC tempera
ture rise is less than 10 to 15% of tbe inlet temperature.. U the adiabatic
temperature rise uceeds this \'alue., ...ie use direct cooling if the reactor heat
load is less than 6 to 8 x 1()6 Btulhr. Otherwise., we introduce a diluent or a
heat carrier.
S. For single reactions we choose a conversion of 0.96 or 0.98 of tb.. equLlibrium
COn\erslOD.
9. The most upensive reactant (or the hea\'iest reactant) is usually the limiting
reactant.
10. If the equilibrium coDstant of a re\'('rsible by-product is small. recycle the
re\'ersible by-product.
11. Several design guidelines for reactors are gi\'en in Table 6.6-1.
12. The recycle flow of the limiting reactant is gi\'en by F = F..(I - x)/x, where FII
is the amount of the limitlO& reactant needed for tbe reaction and x is the
conversion.
13. The recycle flow of other components can be determined by specifying the
molar ratio(s) at the reactor inlet.
Exercises
SI'.C11DN II SU\olMIlRY. UEIlClSn, IlJ'oID NOME'oIClATUU 161
6.8-3. De\'elop the recycle struclUre ror the gu-phase process that produces acelOne rrom
Isopropanol (ExerCise 5.4-4). Anume that 611 .. _ 25,800 BlU/mol and that the
feaClor cost can be estimated as a heat nchanger .,.,jtb U - 10 BIU{hr ft
J
oF). The
heat required for Ihe reaction is supplied from a Dowtherm fumaoe \\"Ith Do\\tberm
al 600"F. Piol the economic potential versus tbe sIgnIficant design \'anables
6..8-04. De\c1op the recycle structure for the elhanolsynthesis rroblem (see 5.
4
5)
Assume that 61f
1l
.
EtOM
_ -19,440 Btu/mol and AU
u
[( - Btu mol, that the
forwud reaction IIle constant IS gn"en by
k, _ 14 J( IOnp{-S3,700,1RTCR)J} br-'
and IS fin;l-order In 1Io-.lef: and Ihat
K ., (1679 >< 10 up [10, 1I9;T("R))
Plot the economic potent,,1 versus the silJ1lfic:ant design >-a.riabk:s_
6.8-5. De\dop the recycle sJruclUre fOI the styrene: process 54-6). Assume
Ihal 6H.. _ SO,SJO Blu mol. AU
II
..
u
_ 45.]70 Blu mol. aDd 6U, .. -
-23,380 Btulmol, that the reacuon rate constant rOf the primal}' rea(1lOl1 is
gn-en b) the aprCSStOll t, _ (381) up (-20.440I[Rn'R)]} hr '. and thai K.. ..
7,7].4 up I 27.170 T Plot lhe economIC poIenual >n>;U5 lhe
\--anabb
6A-6. De>elop the ryck Slructure for the cyclohaane process (see oerose 5 4-7) Assume
Ihal AHII _ 93,200 Btu/mol that K.. _ (2.67 )( 10- U)u:p [4.72 )( 10" 'T(R)),
and lhal the fono..ard reac1tan rate constant IS given bJ ('788)(
eJlp ( _ 14.4(0).(R7lR))}, where the rorv..ard reacuon is first-order in bel1lene
Plot the economic polenlial versus the stgnificant design \'ariabks.
6..8-7_ Develop the recycle SIrUClure fOf the ethykne pr-ooess (see oerose 54-8). Anume
Ihat AH
I
_ SS.6SO Blu/mol thai 611. _15.320 BIU mol aDd thatrhc reactor cost
can be eshmated as a pyrolysIS furnace. Pk>t Ihe economic potential .enus rhe design
.ariablc:s.
6.8-3.. De\-dop the reC)"c1e structure ror the butadtcne sulfone process (see oerCI$IC 5.4-9). As
sume that 6H.. _ -48.000 Btu/mol. K... _ (6846 J( 10-
1
') e"p (-36.940/(TfR)).
k
l
_(8.172 J( 10") eJlp (-52.200/T(OR)), and lr, - ...(moINt'.br)):
that the reaction rale corresponds to the stoichiometry: and that we use a CSTR for
the reactor_ ConSider \'ariable denSlIy effeeu and assume that the annuaJizcd, installed
reactor cost lS !Iyen by 315011'" [$I(fl' yrn Plot the economic porential >ersus
the significant design \'anables.
6,.1.9. De\'elop the recycle strUCIure ror thc butane alkylation plocesS (see Ellcrase 5 10)
AS5ume that AH, _ -27,440 Btu/mol, AliI _ -25.180 Btu/mol, t, = (9.56)( 10")
ell' {- 28.oooI1RT{R)]} hr I, and k
J
_ (2.4]9 J( 10") up {-]S.OOOf[RTfR)]}
hr - I. Use a CSTR With the cost correlation gi>'en in E"ercise 6.8-8.
6.8-1_ Develop the recycle structure for lhe HDA process with diphenyl recy<:lod (see
Eq. 6.2-)]). Plot the economic potential \'ersus tbe design nriables, assuming that
K... _ 0.2]96. (1\1so see Appcndi1 B.)
6.8-2. Develop tbe recycle strUCIure ror an acetic anhydride process (5Ce Eq. 6.1-2 and
E1ercise S.4J). If tbe acetone pyrolysis reactor costs are calculated as a rumac:e cost,
and ir the anhydride reactor cost is nesJigible, plot the ecollomic potential versus Ihe
desisn \'ariables. (Assume All11,'0" _ 34.700 Btu/mol, AHII.'" - - 27,000 Btu/mol,
and MI ....... - 20,700 Btu/mol.)
Nomenclalure
A, 8, R,S.
a" h,
A
bhp
C,
Reactive and product components
Order of reaction for component i
Ileat-cxchanger area (ft J)
Brake horsepower
Conoentration of component i
162 Sl'MMAII.'f. uunsu, AN" NOMEtoCLATUl.E
C,
,
f(C.. , C.)
1.
F
F,
F,
F
n
F,
F,
F.
hp
"K,.
M
MR
M &S
P,
Pia.
P
J
p,.
Q..
Q.
"

R,
T., T_
T.
U
x
Y..
Y,
h.
AH.
,
lleat capacity [Btu/(mol "..-))
Activation energy for component i
Function of composition
Fugacity of component i
Rcactor fccd rate (moljhr)
Conection factor for a gas compressor
Feed rate in excess of reacllon requIrements (mol hr)
Fresh f(cd r.lte of toluene (molfhrj
Mal-eup gas rate or III (mol/hr)
Flow rate of compClnent i (moljhr)
Reactor rec:d rate: of toluene (moljhr)
Horsc:po.... cr
Reaction rate: constant
Equilibrium constant
Molecular weight
Molar ratio
Marshall and Swift mdex ror inflation
ProductIon rate of component; (moljbr)
Inlet and outlet pressures for a gas compressor
Pressure at stages of a gas compressor
Total pressure of reactor
Volumetric ftow rate (ft1/min)
Reaclor heat load (Btujhr)
Reaction ratc
Gas constant
Recycle gas flow (moljhr)
Inlct and outlet temperatures from a gas compressor ("R)
Reaelor temperature:
Overall coc:ffiClCnt [Btuf(hr. ft 1. CF)]
ConversIon
Feed mole fraction of H1
Mole: fraction of component i
Purge composition of Hz
Heat of reaction at rcaccion tcmperature and pressure (Btu/mol)
Fugacity coeffiCient of component I
Cp/C. - I
Cp/C.
CHAPTER
7
SEPARATIO
SYSTEM
Here we consider only the synthesis of a separation system to rc:eo\er gaseous and
liquid components Our diSCUSSIOn is broken down mto thrc:c separate: parts.
general struelure, vapor reco\'cry system, and hquld separation system. Also keep
in mmd that .... e need to detemline the best sc:panuion system as a funelion of the
design variabks, I.c., the range or the design vanablc:s In FIg.. 6 7-1 .... here \loe obtain
profitable operation Thus. our previous economic stu(hes hc:lp to simpllfy the
computauonal effort.
7.1 GE ERAL STRUCTURE OF THE SEPARATIO ' SYSTEM
To determine the geDeral structure or tbe: separation system......e first determmc the
phase: of the rc:aelor effluent stream (see Fig. 7.1-1). For vapor-liquid processes,
there arc only three posslbilltlCS:
I. lIthe reactor effluent is a liquid. we assume that we only need a liquid separation
system (sec: Fig. 7.1-2). This system might mclude distillation columns. cxtrac-
tion units, azc:otropic distillation, elc., bUI nonnaJly there will not be any gas
absorber, gas adsorption UflitS. etc.
2. If the reactor emuent IS a two-phase mixture, we call usc: the reactor as a phase
spllller (or put a flash drum after the reactor). We send the liquids to a hquld
separation system If the reactor is operating above cooling-water temperal ure,
\loe usually cool the reaclor vapor stream 10 IOOQF and phase-sphl this stream
163
164 SECTIOI'" 11 GENERAl sn.ucrullE OF nlE SEPAIl.. noN SYSTEM
SF.CTlON 1( GENERAl STRUCTURE OF TIlE SEPA1lAl10N SYSTEM 165
Purge
r't'cd streams
Reactor
Separation
system
/
systt'm
Products
/'
Focus on reactor exil Slream
FtGURF 7.1-1
Pha.sc of the relict". effluent stream
F=J,
Reactor Liquid
Liquid
system
sepanllion Products
system
FIGURE 7.IZ
React". cUI is liquid. [F,om J M. DtNgltu, Ale": J., 31: JH (l98J).J
(see Fig. 7.1-3). If the low-temperature nash liquid we obtain contaills mostly
reaclants (and no product COmponents that are formed as intermediates in a
consecutive reaction scheme), then we recycle them to the reactor (we ha,'c the
equivalent of a condenser). However, if the low-temperature flash liquid
contains mostly products. we send this stream 10 the liquid recovery system. The
low-temperature nash vapor is usually sent to a vapor reco\'cry system. Bu!. if
the reactor emuenl stream contains only a small amount of vapor, we often send
the reactor effluent directly to a liquid separatlOll system (i.e, distillation train).
3. If the rcaclor emuellt IS all vapor, we cool the stream to lOO"F (coollng-waler
temperature) and we allempt to achieve a phase splil (see Fig. 7.1-4) or 10
completely condense this stream. The condensed liquid is sent to a liquid
recovery system, and the vapor is sent to a vapor recovery system.
If a phase split IS not obtained, we see whether we can pressurize the
reactor system so that a phase split will be obtained. (We see whether a high
pressure can be obtained by using only pumps on liquid feed streams, and we
check to see that the pressure does not affect the product distribution.) If a phase
split is still not oblained, then we consider the possibility of using both high
pressure and a refrigerated partial condenser. In case no phase split can be
obtained without refrigeration, we also consider the possibility of sending the
reactor effluent stream directly to a vapor recovery system.
We need 10 ensure that the: same structure: is obtained for the complete range
of design variables under e:onsideration. These: rules are based on the: heuristic that
phase splils are the cheapest method or separation and the assumption that some
type of distillation separation is possible.
,
FleURE 7.1-4
RC<IIdof nIt vapor [From J. M f)Quglru. AfOrE JI J.5J (l98J).J
Gas recycle
I
r-):5- 35"C
Vapor
Vapor
Ph"" recovery
split
system
Vapor Liquid
Liquid I
r--------------
F=J,
Reactor Liquid
Liquid
system
se:paration Products
system
I
FIGURE 1.t_3
Reador uJl IS alld Jiqutd.lFrom J M Do"9fas. Afe,,: 31 JJJ (/98J) l
Purge
I
Vapor
co'x""'')'
system
Vapor
Reactor
Vapor
35C
Ph""
-,
system ;- split
Liquid Liquid
Liquid
separation Product
Liquid recycle
system
Purge
166 $Ecnow 11 l..t;t.IEIl..Il. nllJCWIE 01 TIlE $U,,-"now SYSH:W SECTlON'11 Ouq....l STJ;.l,!C1VIf Of' TN! 1l!.......noN IY5l"I!W 167
Approximale Flash Calcuhllions
To. Ihe phase of the reactor effluenl, in cases, we can usc a sharp.
split approxlmallon procedure to avoid Ihe lrial-aod-error SOluliollS associaled
Wllh lIash calculations. The flash "IualiODs can bt: wnllen as
When we combine these expressions, we obtain
)'- = z,
, V/F + (I V/FXI/K,)
TABU 7.1-1
HDA flub
AIfII'OUaYllt ..
c__
f, ...
,
,
,
,
H,
"..
99_01
""
,
".,
I
CH,
""
,..00
""
"
"B
10
A,....ou.n.11t Un
Com,.-,wl
I,
"
"
I,
"
"
..
26'
0.01040
29.'
235.4
,1.2
236.1
Tol"",ne 91 0.00363 J.'
17.4
J.'
17.4
Dip"""yl

0 '.0
(7.1-1)
(7.1-2)
(7.1-3)
(7.1-4)
F=V+L
balance.
Componenl balance.
Equilibnum:
(7.1-15)
(7.1-13)
(7.1-14)
(7.H6)
fi s; Fz;
Now, we can go back aDd adjust Ihe: vapor flow for thiS loss:
',-1.-
1
,-1.(' - K':fl.)
when:
The cofTcspondlDg expressions for componcnls thai are predomrnanlly 10 the
liquid phase are
Table 7.11 compares Ihe: approximate and exact solutions for Ihe HDA
process. sec Ihat tbe approximalc solution is satisfactory for prdlminary
designs. However, the results are valid ooly if there are no components bUlDg K
values in tbe Bnge from 0.1 to 10
AN ALTER.....ATE APPROXIMATE PROCEDURE .-oR Fl.ASH CALCULATIONS-
Another shoncut procc:dure for flash calculations was published by King. If we
again consKkr Eqs. 7.1-1 through 7.1-3, wc can wnle
L
Ii'" II, + K.Vllj
,
'Dd
(7.1-5)
(7.1-6)
(71-7)
(7.1-8)
(7.1-9)
(7./-11)
(7.1-10)
0'

x _ _
(K
J
J)vIF + J
If K, I in Eq 7.1-4, we s Ihat
Vy, Fz,
and if K, <t I in Eq 7.1-5, we find Ihal
Lxi Fz,
Thus, as a first eSllmale of the vapor and liquid flow rales, we can write
v - [Ji for all components wbere X, > 10
L::: [h for all components where X
J
< 0.1
These express!ons are cqui\'alent 10 a perfeci split. providing lherc are no
components with a K
i
value bel ween 0.1 and 10.
The perfect SpIll expressions Ignore the: vapor-liquid equilibrium_ However
we can Ihese relalionships on our expressions for
flows. Thus, .the liquid composition In equilibnum wilh a vapor component having
a mole fraction
;,
Ji Ii

Y. /,
X_,z_= _
, X. K , EJi
The: liquid flow of Ihis component is Ihell
(7.1-12)
168 SEClION'-l VAreR RF.COVEU'SY5TEw
Purge
"'"
Prevenl loss of
valuabk or
undesirable
material
__,--- ---1 __
, ,
, ,
"--- r--

phase split
PrC'-ent recycle
of certain
l::omponenlS
Gti
recycle r------ r-------
I--.----i
L l
/
(7.1-18)
(7.1-19)
J. L
--I--
v, K,V
By rearranging Eq 7.1-16 we oblam
Now, if we divide Eq 7,1-18 by a similar exprCSSlon for componenlj, we oblain
J;/v, - I K
J
h/vJ - I - K, - 'J
If we specify the fractIOnal recovery vJ/iJ for one component, we can usc Eq 7 1-14
10 cakulale !he fraCllonal reco\'ery for e"'ery other componc:nl_ .
E\'en this analysis is rigorous for constant-a systems., the results for a
SpeCIfied fractional fCCO\'ery of one componenl normally will not correspond 10 a
flash.lemperalure of 10000F and Ihe flash pressure, Thus, some iteration migh' b<
reqUIred.
7.1 VAPOR RECOVERY SYSTEM
onideal Mixtures
arc no calculation procedures for nonideal mixtures. However all
packages,l,e., FLOWTRAN. PROCESS, DESIGN 2000 ASPEN ' '11
handle Ihese problems. ' , etc., WI
When we attempl to synthesize a
decisions:
vapor recovery system, we need to make (wo
nGURE 7.1-1
Vapor recovery klc:at>oa
2. Place the hpor rCl::overy system on the gas-recycle stream if materials Ihat are
deleterious 10 the reactor operation (catalyst poisoning, etc.) arc present in this
stream or if recycling of some components degrades the product distribution.
The g&5.rec)'cle: stream normally has the second smallest flov. rate.
3. Place the \'apor recovery system on the flash vapor stream if both items I and 2
8rc valid. i.e.. the flow rate is higher. but \lle accomplish two objectives.
4. Do Dot use a vapor recovery system ir neither item I nor item 2 are important.
J. What IS the best location?
1. Whal type: of vapor recovery syslem is cheapest?
Location of Vapor Recovery System
There ate four choices for the location of the vapor recovery syslem:
I. The purge stream
1. The gas-recycle stream
3. The flash vapor stream
4. one
Adjust the Material Balances?
Note tbat unless item 3 is chosen, our simple material balance equatioDs will not be
valid; i.e., some materials that we assumed were recovered as liquids will be lost in
the purge stream or recycled wilh the gas stream (which will change the compressor
size). However. in many cases the errors introduced arc small, $OJ,hat our previow
approIimatioDS still provide good estimatC5. We expect to de\'e1op rigorous
material balan<:es if we proceed with a 6nal design, and therefore we use our
engineering judgment 10 see whether I::orrections need to be made at this point.
Eumpk 1.2-1 HDA proce-. Do we need I Vllpor recovery system for the 1I0A
process?
The rules we use 10 make this decision are (see Fig. 7.1-1) as follows:
I. Place the recovery system on the purge stream if significant amounts of
valuable matenals are being losl in the purge. The reason for this heuristic is
that the purge stream normally has (he smallest flow rate.
SobniOIl. For a oon\'Crsion ;r; _ 0.7S and a oomp<mtlon F,. _0.4, the vapor
Ro..s (rom the phase .plitter arc given in Table 1.1-1 The purge and recycle Dows for
this case were 496 and 3371 molfbr, respectivel) lienee, ,,"'e can estimate Ihat the
benzene and toluene nows lost in the purge are 3.79 and 046 molfbr, respectively On
an annual basis and by neglecting the fuel values of tbesot components, Ibis loss
SOJ04 x Ilf/yr. 11us value t5 small compared 10 our ecooorOlc potentIal
(7.2-1)
lind so we dcode not to Include. vllpor recovery system II this POlDt 10 Ihe dCSlgn
development. We mlghl well reconsider thIS dccwon .fier we have deternuncd
whether the BOA process is profitable. I.C.. If we dcQde to .bandon lbe projCCt, ....e
.....ant 10 mlnunw: tbe eoginccnog effort thlt we iDvClit.
Slna: the reaction we ace COn5ldenng is bomogeDcOllS. no COlDpoDCOLS 10 the
gas-recycle slrca.m c:a.n cause QllIlysl delClJvllJO!l. However. there is I slptficant
amount of benzene Itl the Illsh v'por stream of the benzene 8ow); sec T.ble
7.1-1 And mosl of IbIS benzcne (29 6 mollbr - 4 3 mol./hr }ost In tbe purge) "111 be
recycled to the reaetor The benzene IS formed as tbe lntermedilite m a coDsecUtlVe
reacuon scheme
Toluene of H, -> Bcnzcn.c + CH
4
2Bcnzcne Diphenyl + Hz
Therefore, we would eJ.pccl tbal 50me (or most) of &Dy benzeoc thaI is recycled 10 Ibe
reactor will be convened to diphenyl Unfortunalely, lbe selectivity data we Ire uSing
(see Example 4 I-I) do oot include the effect ohny benzene reed 10 tbe reactor. so we
cannot esllmlte Ihe amount of benzene lost 10 dlphcoyJ. This difficulty could be
o,'ercome If "'e bid iI \;metlC model aVlllIbIc.
Wuh lbe 1\"libblc datil, bo".evcr. we wouJd need 10 put I vapor recovery
syslem eIther 00 tbe psrecyclc strum (to JKeveot the loss of 50DX of the benzene by
reaction 10 dipbcoyl) oc 011 Ibe Il.ub vapor Stream (10 preveot!osi of bcnzeoc both 10
the purge strum .nd by reacttOIl). AnolMT Illernative could be to necydc tbe
dipbcnyI 10 ClItDClion.. ratber tban recoverillg and rcmovi0llk dlpbcnyl Witb tblS
altcmaUH\ we would avoid the 5ClccuvllY loss of Ioluene 10 dlpbcnyilltogetber, lod
we C<lD tolerlle the prcscnce of benzene io lhe gas-rccyde Slream
Some of the benzene in the glsrecycle stream Cln be recovered in lbe
compressor \;nod:out drums before the gas-recycle slream enters the reactor The
vapor stream IS a ulUratcd vapor, so that as we false the pressure of [his strelm
In each of Ihe Ihree stages of tbe gas-recycle compccssor. 50me of [he benzene will
condense Normally, we cool the nit from eacb compressor stlge 10 10000F with
cooling water. Ind then ...e include a Imod:oul drum (ic..1 Dasb drum) to collect tbe
coD.denslble m'lenals. We can SCIId tblS condensed benzene 10 lbe: liquid separauon
system.
Rather lhan IlIcmpl Co evallllie IUlhesc VariOWI Iltcruativcs II this umc. ,,",e
merely make some dccisIoo and CODlinue 10 develop a buc c:ue. We list all the olhc:r
allernatives as Ilems that need 10 be considered after we b,ve estimated the
pcofitlbiltt)' of lbe process and bave I betlcr undcntaoding of the allocation of the
costs. Q{ courx., we IWnimJzc our don by suc:ssing thai most of the belU.C'De In lbe
gasrccycle stream win be recovered in the k:noctOUI drums ISSOCIltcd wnh the
comprasor or wdl nOI be converted 10 dipbenyl if II IS recycled to tbe rcaetor
However, it is csscnlillto check IhlS Issumption laler
PROCESS FLOWS. If we do not recover the benzene and toluene from the nash
vapor streams, our assumptions concerning the overall and recycle material
balances are no longer valtd. In particular, the amount of benzene leaving in the
flash liquid stream IS not adequate 10 meel Ihe plant production file, allbough the
HenoN 1J VAl'OllECOVUY SYSlt.... 171
benane m the compressor l.nockout drums WIll decrease the magnitude:
of the error
We could go back and revIse all our matenal balan calculauons. lIo....'e\'er
11 Will be necessary to revISe them agam after we ha\"e spec::tfied a hquld separatIOn
system. Sintt the changes ....e Introduced III the process "o....s are not too large, .... e
decide 10 contmue Wllh the analy!>ls Of course..... e arc startmg to 3ccumulJte
erron., and .... e \;now Ibat If .... e decide nOl to abandon Ihe projeCl. we must rClll:.e
our calculations. We describe a procedure for correclmg the material balances In
Sec. 7.5.
Type of Vapor Reconry Syslcm
The most common choKes (with currenl technology) are
I. Condensation high pressure or low temperalure, or both
2. Absorption
J. Adsorption
4. Membrane separation process
s. Reaction syslems
$honcul dwgn procedures for gas absorbers "'ere dISCUSsed in Chap ], The
economic trade-offs for the design of a condensation prOttSs arc considered In
Exercise 3.5-2. A design procedure for adsorpl!On processes has been presented by
Fair.- Neither a deSIgn procedure
t
nor a oost correialion for membrane recovery
processes seem to be available in the open literature, alihough vendors of
membranes Will provide Ihls service. Reactions arc sometimes used to remove CO2
from gas streami, and H
2
S IS recovered with amlneS
SlTalegy
We design the yapor rcoo\"ery system before .....e consider the liquisJ separation
system because each of the vapor rCCOft'ry proc:csses usually generates a hqllld
slream Ihal must be funhcr purified. For the cascofa gas Ibsorber, where .... e ncc:d
10 supply a solvent 10 lbe absorber. we also introduoc: a new recycle loop between
the separation systems (see FIg 7.2-2). Nonnally we need 10 estimate the StU and
costs or each umt to detennine which is the cheapest
J R hur, SolveOl .nd p l. WlnlulllloD Uwvecolty Dc:5lgn C.se
Sludy No 7. edl\ed by 8 D Smllh, WllShmlllon UnIversIty, St LOUIs, Mo., 1969
, A WDpIc model lhal can be 10 CiUm.aIC lbe are;!, of. membr..., proccu hilS been publIShed by
J E Hopcn and W II MiIlZW. Jlydr""arb Proc AugllSt 1983, P 51
lRI!CTIO,," U UQUJD SU'''kATIOM SYST1!1ooI 173
Light Ends
Some: Lighl ends will be: dissolved in the liquid leaving the phase splittcrs shown m
Figs. 7.1-3 and 7.1-4, and normally some will be dissolved in the liquid streams
leaving the vapor recovery systems. If these light ends ml!ht conlammalc the
product. Ihey must be removed.
Each of Ihese decisions IS dISCUssed below. Remcmber that wc want 10 male
decisions as a (unction of the design variables o'"cr the range of polcntlally
profitable operalion
3. Do we recycle components thai form azcotropc:s wilh the reactanls. or do wc
split the azeotropes?
4. What separallons can be made by distillalion?
S. Whal sequence o( columns do "''C use?
6. How should we accomplish separations if distlllallon is not fcaslbk?
phenyl
2' CH..
Benzene
Purge
Vapor fCCOYCry
H
S)':'ilem
f-
2. CH4
Reactor
.....'"
li):stem spill
oluene
Liquid separation
system
D,
H
tlGURE 1.2--2
Scr-tabOa s)'Slem loop
ALTERNATIVES FOR LIGHT-ENDS REMOVAL. The ehoJCCS we have for re-
moving light ends are these'
Combining tbe Vapor RKonf)' System ,,-jib tbe
Uquid Separa.tion S)'stem
If we use a partial condenser and a flash drum 10 phase-spilt the: reactor effluent.
some of the lightest liquid components will lea ve with lhe: flash vapor (i.e.. a flash
drum never yields perfect spillS) and therefore will nOI be recovered in the liquid
recovery system. However. if there is only a small amount of vapor in the stream
leaving the partial condenser and if the first split in the liquid separation system is
chosen to be distillation, we could eliminate the pbase splitter and feed the reactor
effluent stream directly into the distillation column.
The diameter of a distillation column with the two-phase feed will Deed to be
larger (10 handk the increaseJ vapor IrafflC) lhan a column Ihat follows a flash
drum. this increased cost may be less than tbecosLS associated with using
a vapor system to remove the liquid components from the flash vapor
stream. There docs not seem to be a heuristic available for mak.ing this decision.
and so we noed 10 add another process alternative 10 our list.
7.3 LIQUID SEPARATION SYSTEM
The decisions lhat we need to make to synthesize the liquid separation system
include the following:
I. Drop the pressurc or increase the temperature of a stream. and remove thc light
ends in a phase spliner.
1. Usc a pawal condenser on the product column.
3. Usc a pasteurization section on the product column.
4. Use a stabilizer column before the product column.
The lasl three alternatives are shown in Fig.7.3-1.
The options arc listed in the order of increasing cost, and therefore we prefer
to usc the earlier entries. However, 10 make a decision for ligbt-<nds removal. il is
oecessary 10 kno.... the flo.... rates of Ihe lighl ends and 10 make. some shortcul
calculations or some CAD runs to eslimate II>e amount reco'cred:
I. Flasb calculations. These are discussed in Sec. 7.1.
1. Partial condensers. CAD programs handk these probkms.. or in some cases the
approximate flash calculations given in Sec. 7.1 can be used.
J. Pasturization columns. A shortcut daign procedure has been published by
Glinos aDd Malone (see Appendix A.5).
4. Stabilizer columns. This is a normal distillation column that rcmoves light cnds
1. 1I0w should lighl end. be removed if they might contaminate the product?
1. What should be the deslmation of the light ends? Ie. GliDos and M f F...., Climl. Proc DeJ. 0-.. U 10117 (19115).
1] UOUlD S'.......T10S SYSTEM 175
If produci quality ll> unattepwble, opllOIl5 are:
NOle: Recycle vapor stream 10 vapor recovery syslem if possible.
Applicability of Distilllliion
In general, dlslllJalion IS the lcast expensive means of separating mixtures of
liqUIds.. However, If the relauve volatihucs of two components vo-Ith neighboring
boilmg points is less than 1 I or so, distillation bc::comes very expensive; i.e.. a large
rcflux ratio IS required which corresponds to a large vapor rate, a large column
diametcr,large condensers and reboilers, and large steam and cooling watcr costs.
Wheocver we encounler Iwo nelghbonng components having II relatl'-e volauhty
orless than 1.1 in a mixture, we group these components together and VO'C treat this
group as a single componenl 10 the mixture. In olher words, wc dc\-e!op the best
distillation sequence for the group and the other components, and then vo'e separate
the lumped components by using other procedures (see Fig.. 7.3-2).
the azeotrope nonnally requIres two columns and therefore is expensi\c. Howcver,
if we recycle the azeotrope, we must oversize all the equipment in the recycle loop
10 handle the incremcntal ftow of the extra components. A gencral design heuristic
does not seem to be available for mai-Ing thiS deciSIOn, and so we usually nccd to
evaluate both alternatives. Azcotropic s)stems are discussed in more detail in the
next sccllon.
SL3blhzcr
column
Pasleurizatlon
section
Panial
FIGURE 1.3-1
AhcrualJVCI lor remoVlDI baht ends
DESTINATION OF UGIIT [:"ODS. For lhe deslinauon of the light ends, we can
venl them (possibly to a flare system), send tbe light ends 10 fud, or recyde the 1Ight
ends to the vapor rtto\'ery or the ll.ash drum. If the light ends have u:ry
little value, we wanl 10 remove tbem from the process through a venl. IfthJs ,-enting
causes air pollution problems, we try to vent them through a flare system to bum
the offending component. If most of the light ends are flammable, we try to recover
the fuel value. However, if the lighl ends arc valuable, we waDt to relain them in the
process. If we recycle them to the vapor recovery system, we introduce another
recycle stream into the process.
Column Columns
For sharp splits of a thrcc+componeni muture (with no azeOlropcs) we can cuher
reco"cr the hghlest component first or the heaViest componenl first, and then we
split the remaining two components (see Fig.. 7.3-3). When the number or
components increases, the number of ahernauvC$ increases very rapidl) (see Table
7.3-1). lbe spillS that can be made in the 14 alternatives for a five+component
mixture are lisled in Table 7.3-2.
It appears as if it will be a major task to decide which distillation column
seqUCDCC to sclecl for a panicular prooess, particularly since the best sequencc
SUMMARY FOR L1GIIT [f\,'05. If light ends will not contaminate the product,
we merely recycle them to the reactor with a reactant-recycle stream or remove
them from the process with a by-product stream that is sent to the fuel supply If
light ends will contaminate the product, they must be removed from the process
The method of removal and the destination of the light ends depend on the amount
of light ends. Hence, we must delermine the amount of light ends as a funcllon of
the design variables before we can make a decision.

It 3_2
.... ---......
IS 1.71
Lump----l I
__
D 1.0
E 0.4
A (B,C)
D,E
Separate [ B
design task _ C
r--D
Auotropcs with Reactants
If a component forms an azcQlrope wilh a reactant, we have the choice ofrecycllflg
the azeotrope or spliuing the azeolrope and just recycling the reactant. Splitting
flGURE 7.3-1
Oul.l1lauQn
L-_E
176 UCTl0N U lIQUID S[PAUTlDS SYSTflol
SECTION 7J 1I0lllD SEPAUTION S'($11'r.l 177
TABLE 1.3-2
Column fot ti,". product sunms
Colo_ , C..... ,
CoIooo 3 C.......
I A BCDE. B/eDE. C/DE. 01'
1 A.BCDE. B/eDE. CDE CID
3 A BCDE. BCDE. BrC D/E.

A BCDE. BCDE. BCD CrD

,
A,BCDE. BCo/E BClD B(C

ABiCDE AlB C/DE. o/E.

1 ABiCDE.
A'8 COlE. C/O
I ABC/DE. DIE AIBC
8'C

ABC/DE
D'E AB/C
A'8
10 AIlCD/. A/BCD B/CD C/O
"
A8CD/f. A/BCD BClO
8'C
11 ASCO/f AB/CD
A'B C,D
l) ABeD/. ABCiO A/Be lliC
.. ABCD/E. ABC'D ABIC
A'8
FIGURE 1.3-]
0'$1111."')11 IhemllOves ror I lem.ry mlllUtC
I. ll.emo>'t: 0I>rT0$I>'t: c:omponeal$ all 100II as possobk.
1. Ilemo",", tcao;t,ve componclll$ or .. _ ti possibk
1. Ilanm'e prodUCIJI as
.. Remo", rcqeloc wurns all datillates. plrtJallarl1 .r they an
10. J*d:ed bed l'CKIl)r
GENERAL HEURISTIcs. There are some general heuristics thai can be used to
simplify tbe selection procedure for column sequences (see Table 7.3-3). The first
lKuristic in this list is based on tlK fact that Ihe malerial of construction of the
column is much more expensive than carbon steel if corrosive components are
present. Thus, the more columns that a corrosive component passes through, Ihe
more expensive will be the distillation train
components will cbange the separation problem and tbus should be
removed Il.S soon as possible:. Monomers foul reOOiJers, so it is neoessary to run the
columns at vacuum conditions in order to decrease: the column overhead and
bollom temperatures, so that the rate of polymerization is decreased. Vacuum
columns are more costly tban pressure columns, and we prefer to avoid the
increased cleaning costs.
We prefer to remove products and recycle streams to paeked bed reactors as
a distillate to avoid contamination of the product or recycle stream with heavy
materials, rust, etc., which always accumulate in a process. If it is necessary 10
remove a product or recycle stream 8S a bottom stream, it is often taken as a vapor
from a and tben condensed again. At tbe same time a small, liquid purge
stream may be taken from the reboiler to prevent tbe buildup of contaminants.
TABU: 1..]..]
Gt'nrul beutistK::s for column wquencing
might change as I,l,e aller Ihe design variables To simplify this effort, we might want
to look for heuriSllCS for column sequenCing. There has been a considerable
research effort in this area over the past decade or so, and some of the results are
given below.

Hauistia for column Sl!:C(UfttCing
I. MO!II plentiful lint
1. Lighlcst Iinl
),. H,gh-l'CCXlvcry Jepllt.lIons luI.
4.. DtfficuJI Jepll'It>OllS IIlSI
F...or t:qUlmollr SpillS
" Nest,.."anuon should be cheapesl.
COLUMN SEQUENCING HEUlUSTlCS FOR SIMPLE COLUMNS- A number of
other heuristics for selecting sequences ofsimple columns (i.e., columns wilh ODe top
and oDe boltom stream) have heen published; a short lisl is given in Table 7.3-4.
r--A
'--8
Indirect

.,
,
..
A
B
C

,
3
,
'----c
,-B
Direct
,--A
A
B
C
TABLE 7..]..1
Number of ahemathes
Numoo of comPOIlalIJ 2
Numoo of St:quenc:ell I
Reactor 1----jCoolant
118 Sf:CT1Ot< H UQtJIIl SI'.'AIUTION SnTDot
However, Ihe firsl and fifth heuristlcs III this list depend on reed eomposltlons,
whereas the second and rourth depend on relative volatilities Henee, we ellOpcct
that these beunsucs wiJIlead to contradictions; i.e., if the most plentil"ul component
Ui tbe heaviest, there is a conOiet between the first and second heunsllcs.
A longer hst orheuristic:s has betn published by Tedder and Rudd, and some
Iflvesligatorli ha\e tried 10 order the importance or the heuristics, 10 resolve the
conOlcts.' A suney or the literature has been presented by Nishida, Stephanopou.
los, and Westenberg,I and a detailed discussion or the limitations or these heuristics
has been pubhshed by Malone et all Some additional discussion of the heuristics IS
given below.
We might also note that as we change the conversion in a process, we expect
that Ihe unconverted reactant will go from being the most plentirul component at
very low conversions 10 Ihe least plentirul at very high conversions. Hence, the
heuristics in Table 7.3-4 imply that the best column sequences will change as we
aller the design variables. Similarly, note that tbe studies used to develop the
heuristics were limited to sequences of simple columns having a single reed stream
that were isolated rrom the remainder or tbe process, so thai different results may
be obtained when we eonsider the interacllons betwttn a distillation train and the
remainder of the plaol.
lIe,,1
Reactanl
=
Product
(oj
5t:CIl0N JJ LlQl..'10 n ....ATIOI'l SYSTEIol 179
Flash
Light
,00'
J
INTERAcnONS BETWEEN TIfE SEPARATION SVsn:M AND THE PROCESS.
For ellOample, suppose we consider the two flowshttt alternatives shown In Fig
7.3-4a and b. We might consider these two configurations to be two of the
allematives in a sequencing problem. However, there is a different number of
columns in the liquid.recycle loop ror the two systems, and therefore the recycle
costs will be different. Hence, the optimum conversion, which usually corresponds
to a tradeoff between selectivity losses and recycle costs, will be different ror thl:
two cases. Of course, we should compare alternatives at the oplimum processing
conditions of each alternative, rather than on an identical feed-stream condition for
tbe two allernati.. cs.
From this simple argument we .see that the problem or selecting the best
separation sequence cannol always be isolated from the design or the remainder or
the process: i_e.. tbe least expensive sequence for a fixed feed-slream condition
might not be the least expensive sequence (becauce the feed-stream condition
sboukl be cbanged to correspond to the optimum Oow). In faet, there might be
another heuristic:
F=I
j---1"""'0<
Reactant
Light
.;; 1--__--'
Product
(bj
Select the sequence that minimizes Ihe
number of columns in a recycle loop.
0 W. Toddcr &.lid D F Rood, AlCIIE 104 )OJ (1918).
'I D Seader &.lid A W WClitcrbcrg, A/ChE 13 951 (1917).
'N NlWda, Ci SlcplwlopoulO1.,.Dd A_ W Wale/berg, AICII 19 l26 (1981).
'M F MalllM., K, CiUIIOI., F E. Marquez..Dd I M Doua.W. AICIa J. JI 683 (I91S).
(7.3-1)
flGURE 1.3-4
Scqueooe sde>cl1OG cl\.o.lI&CS recyck
MULTIPLE SEPARATION SEQUENCES. Suppose we consider separalion sys-
lems that correspond to the general flowsheets given in Fig. 7.1-2 or 7.13; I.C., .....e
need both a vapor and a liquid recovery system. A flash drunl never gives shurp
splits, so that some 01 the most volatile. "liquid" components Will leave with the
flash vapor, and orten they need to be recovered and sent 10 a hquld separatl?n
system 1-l0.... eHr. the Ilash liqUid nllghl contain a large number or much heaVier
180 5ECTlOfo; H 1I0UIO SE'UATlOI'I JYS'TDt
components as well as those that are returned hom the vapor recovery system. In
silualions such as thIs. It might be better to split the sequencing problem into (Wo
parts. That IS, we would split the flash liquid 1010 one portion contammg the
oomponenls returne.i born the vapor recovery system and one portion containing
the heavier components. Then we would design one separation system having a
single feed stream for the heavy components and one separation system having
multiple feed streams for the components returned from the vapor recovery system.
AN' ALTERNATE API'ROACII TO SELECTING COLU 1N SEQUENCES. The
reason for allemptmg to develop heuristics is Ihat the number of alternati\'e
sequences increases very rapidly as the number of components increases (sec Table
7.3--1). However, there are a large number of plants where four or less distillation
columns arc needed to accomplish the separalion. Four simple columns (one top
and one bollom slream) arc needed to separate a five-componenl mixture into fh'e
pure streams. but only four columns arc needed to separate sill: componalts. if two
components With neighbonng boiling points leave in the same stream Thus, for
five exit streams and using only simple columns, we need to consider the 14
sequences shown m Table 7.3-5.
An examination of this table mdlCates Ihat 20 column designs are requIred 10
evaluate all the possibihties. Twenty column designs requires a considerable
amount of effort if each of Ihe designs IS rigorous. However, by using shortcut
procedures (see Appendix A 4), it is possible to significantly simplify the calcula-
tions. Using shortcut techniques. Glinos demoostrated thai thC' evaluation of the
14 sequC'nces '''..as almost instantaneous on a VAX 11-780; i.e., the results appeared
as soon as the program was run. Kirkwood' has shown that the 14 sequC'nces can
be evaluated in only a few seconds on an IBM-PC XT.
The results of Glinos and Kirkwood indICate that for modest-slZC' sequencing
problems it is bellC'r to develop computer codes that evaluate the 00515 of sequence
altemalives than it is to use heuristics. Moreover, the running times for these axles
.re sufficiently small that the best sequence can be detennined as a function of the
design variables.
Complex Columns
Rather than consider only sequenoes of simple columns (one overhead and one
bouom stream), we can consider the use of sidestream columns, sidestream
strippers and reboilers, prefractionators, etc. One set of heuristics for columns of
K. Gli-. -A GIot-1 Arll"OKh 10 lbot rrdinunary ","""p altd SYIIlha.ui of Do!Itill:ltooll T.a.II,,-
PhD UlI1tTSlly of Muudtuselts. 1984
'R. L KIrkwood. -PIP Pr.-.,,; III"CIltioll Pl>.O n..s.... of MllSIKh.....u..
Amhenl. 19S7
TABLE 7.J-S
HeuriSlics rM complex columns-Tedder .00 Rudd
CntuUl
I(AI(C 0A.
Eue4 IICp"r.uon IndeX (ESI) "" --- -
1(.". 0.,
If ESI < I. the A B Splil IS harder thall tbe BIC Splll U ESI > I, tbe AlB spbl IS ea$lef than the SiC
_rllt
lleunSlIQ lor ESt < I 6
I. H 40 10 80"; IS mtddlc product and nearly eq....1amounts or overhead.nd bot.toms are presenl.
then favOI desll" S
1. If mon lhan rmddle produd aftd lao Wtt S"; IS bottoms.. t1aal Ca"O< desl.p 6.
l. If_ Os middle prod'llCl and lao Ibn S" 1S0werhea<k. dlra favor dalp 7
4. If lea than I S"; IS middle prodlKt and nearly eq....1 amowns or owahelds aftd bonoms are
present. then r.VOf design 1
So Othc..- (avo< desll" I 01 2. wluehever the most pIelitiruJ cocnJlODe1It lint.
HcunstlCll lor ESI > 1 6
I. If moIC than SO'% II' bolloms produc:l.. then fa"O< danl" 1.
2. II mon: titan SO% IS middle prodUC:I and from S 10 20" tS bono..... lher. fuor doslgn S
1. If more lban SO"; IS nuddle p1ooUC:I and lea than S" is bottoms, tbeD favor dell" 6-
<I. If _e than tS auddk po-odlllCl aad lao thaJt S% .. 0eI'heads. tlom f.VOf desI,n 7
So Otbeno'lse. raVOl delll" J
0l.1ter HeuosllC$ ,
I. Therrna.lly coupled designs J and <I should be considerai all ..hemauves to deslJ1lS I a.nd 2.
resJlCC',ItIt!y, of lea tban hal( tbr ked Os mlddll: produd.
1. D=pK l. <I. 6. and 7 should be coMICkred r.,.. Kparaun, an mulUl'eI _liar alo-lIIUddll:-prodUC1
punly IS a.cgeptabk.
StnltelY
I. Roduco: N-compon<:DI &eplIratiollli 10 oequenca cl pscudot=iilry teparallOllSo, UId perlonn tbe
most ditlicuIl ta1Ia.f}' sep&nllOQ IuL
1. ThIS beunstic does 110I 'Uilrant suuaur.1 opluna..li'y 01 uploal.1y CODAda all complu. colUflin
altertlaUVd
F..... I) w Teddrr.1Id I) F Rudd. AIClte J. 14: J03(I91Il.
this type has been published by Tedder and. Rudd (see Table 7.3-5 Figs. 7.3-5
and 7.3-6). Anolher set has been presented by Glinos and Malooc (see Table
7.3-6). Some shortcut design procedures that are useful for complex columns are
given in Appendix A.S.
COMPLEX COLU tN'S IN SEQUENCrs. Our goal is 10 complete a base-case
design as rapidly as possible in order to make a preliminary evaluation of whether
n \\' Tcdda and n F ROOd. AIChF J. lA: lOJ (1978)
K Ghnos a.nd 1>1 F Malone. Complex Column Ahetnal'ves in Do!Ildlauon rap"'r
submitted to ("lrt... f.rtg Itt, INs., 198.5
182 5EClIO'" 7J LlO'IlI.1 ","AUTIO'" SYfiH,
nCUR 13-5
CoIu..mD dc:ItJllL [F,_ D W.
TdJcr aNI D. F. R.od, ""JCJrE
24: )03119111))
B
B
A
A
c
c
Design 2
Indirect sequence
Design 4
Sidesueaffi stripper
B
B
c
A
A
c
Design 3
SideslreaJn rectifier
EXTRACllON. To separate a mixture o( Band C having a (eed composition
correspondmg to point I on Fig. 7.]-7. we counlercurrently contact the (eed with a
solvent S, corresponding 10 point 2 on Fig, 7.3-7, in an extraction column.
Nonnally, we attempt to recover 99% or more of component C. from the original
feed, which corresponds to point] on the figure. We remove the solvent from this
stream by usmg a dIstillation column to obtain the product stream (or component
B, shown as POlOt 5. The other stream kaving Ihe extraction unit corresponds 10
point 4, and when wc usc distillation to remove the solvent from this mlxturc, we
)btain the conditions at point 6
evaluatiolls, we dcfcr a consideratIon of complcl columns until we coosidcr other
process alternatives, which we discuss in Chap. 9.
Otber Types of Separations
If distillation is lOO expensivc to use to separate IJquid mixtures, that is. (I < 1.1, the
olber chOIces that arc normally thc ncxt least expensIve are lisled in Table 7.3-7. In
most cascs. these: separation procedures require multipk dIstillation columns to
rcplace a convcntional distillauon, and $0 they arc normally more Cllpensivc. A
brief description of eacb type of separation is given below.
TABLE 1.3-4
I-Ie-urislics (or complex columns-Glinos and Mlilooe
5ClJ1ICncc:s
.. 1),.1:> ... I)
b U..c tbt ,lId,reet ""'IlICllU II ZA' 'rL.,. le,) <; I C........ 1>-
e C.Jcub.le lhe .-apor If (:rd - I) \"... l) > -f la) > 1/(".... -+ I).
2. ooluJlUU,
G AI....)s wrwda woln,. SlIksueam wtumn .. hen L .. and 0' L,. <; 0.1
b. ColWder lWnJ' 5,da:m:am column .. hen ,be lO'ermedlllle .. ,ccydcd.nd blVi ptInty IS nOl
reqlllred
c ConsKler 115m, 5100lream wlumD ...hell the WlI,ulJtleS .re 1101 evenly dl51nooted.
d Coostdc:r, lldalrcam 'bove Ihe rced \O'lKn lhe illlermedlllle 15 more difficult 10 rrom Ihe
heavy th'n 'rom Ihe hibl Otherwlsc, consider, siootream bela... lhe feed.
l Sideslream s,nPflC'rs 'nd r"lien
o Cons>der llSInJ' PdClitrcam <:<)1111110 when leu than of Ihe l'ccd IS the
b As :opprlChcs ("A' - 1),.1".oe - I). lhe sllVlnp ,re npo:;ted 10 IDacasc.
e The InUJmllm 5lI"'Dp are jO"_ lI.depeDdenl of the rebu>'e vol,uhllQ,.
4 CO<15ider uwllia udC5Ircam oolumD when lhe VolaWIlICS are nOl C"enIy dWflbulcd.
4 Pellyuk c:ulllfWl.5
T'he DlalUmum vapor ">-..015 for :0 Pellyuk columa I,e SO" aod an approached ..ben
IA -(..... -I)I(a.oe-I)a.ndz. - 0
b The vapoe- rale ... .,np are ht&bcr "" :0 Pctlyul colUIQJI lluln for aa,. olhcl Uad of wmplc.o
rol_
.. for IotrlC' or moderslc l <II Pc'l)"l eolumn 15 b,orcd wbeD
(I) The volallll1lC5 Ire bAlanced Ind bolh splits 're dlfficull, lhal I$, <l
A
"" ll'... S' 2-
(2) The spIn"" 8 IS dIfficult. and lhe 81C spill Ii cuy; Ihal 15, ll' < .,.
d For low X. USIIlI a Pelll'ul column .. hen 1. IS dow: 10 (cr
d
- - I), al1bou&h'
s,de5CCIlon column rna) be" be"lIe, 10 IhlS cue (II .eSlllls In .bolll 11Ie ...me vapor .... VIDg5 bUI has
rewer lrays).
e Pellyul columll5 may be advaD1ageous for mndcrale Or blp Z., ClIpecially ... lKo the ""18 Iplll II
nOI much usic, lhan ,be 8'e or when"L >
f i'leulUhCS r_mcndm, lhe usc of Pctlyul; columos for Iarp: nluuofI. arc DOl always corre<:l,
because: tbe pedonnallU depends 011 lbe volatiblJel. If lhe vollhltuea Ire ew:n.ly dislribuled, the
Pctlyuk column aDd prdracllOnlllCN should be conSIdered..
S PrefraetiolUllOrs
. Do DOl oolWdu a If:o Pettyul c:olwl:ln an bc used.
'" The IDILWDWIl AVlDp dcpcad 00 the volalibt_ nod ..Iticb '-'cd Ii COCltrolbnJ.
(I) H tbc upper '-'cd CODU.... the: lD.UImUIII ..",ap are (a.oe _ '".,},'{a..., - 1)... bIeb OIXUB as
.. -,
(2) If 1M IowCf feed eolttrok, lhe mil.l.lQlllm are (a., _ ll(ll'.oe _ I), ...hoc:h oocurs .. ben
,,-.
the process is profitable Thus, normally we include only SCqucilCC5 of simple
columns in our first designs, Ilowever, a complex column IS often cheaper than two
simple columns, and lherefore we need to consider these possibilities at some poun
In our design procedurc, Since: wc can replace any two neIghboring (;()Iumns ID a
sequence: by a (;()mplex column, .... e can generatc a large number o( process
altcrnatlves. To aVOid gellmg bogged do.... n in a number of alternatIve
184 SECT10N U UQUJD 1\'Sl'l!",
A
A, B
B
B,C
C
Design 5
reduction In the degree of separation must decrease the cost sufT-.aentl) to pay for
the extraction column and the other two distillation columns. In some cases this is
possible.
EXTRACTIVE DISTILLATION. If we attempt to separate HN0
1
and HlO by
extractive distillation. y,e add a heavy component, HJSO near the top of the
tower. The presence of the heavy component changes the vaporliqutd equilibrium
(for this example the actl\-ity coeffiCients y,11I be: changed). which in some cases will
Simplify the separation We obtam a pure component, HN0
1
O\'erhead mthe first
column (sec 7.3-8) Then .....e reco\er the olher component overhead m a sc:cond
column, and we recycle the: extractive entrainer. HlSO back to the first column.
We see that two distillatioo columns are required.
TABLE 7.3-7
AJttmali",s to distillalion
B
(J)B+S
S
B+C
C (+ 11)
CD

B+C
(J)
C
(j)
@)
C+S
(+ 11)
crl
a>
B S
nCURE 7.3-7
E,UI1IclIoII.
AZEOTROPIC DISTILLATION. In azeotropic distillation we: add a rdativd) "ght
component that again changes the vaporliquid equilibrium of the original liquid
mixture, often by formlOg a new azeotrope with one of the feed components. Thus,
to split the ethanol-water azeotrope. we can add benzene, which fonos a ternary
a.uotrope. With tbis modification. we can remove pure ethanol from the bottom of
the first column and reco,,'er the ternary azeotrope overhead (.sec: Fig. 1.39).
Since the ternary lUC:otrope is a heterogeneous mixture when it is condensed,
.....e use the benzene rich la}'er as reflux to the fint column, and ",,-c use the other
layer as the feed to a second column. In the second column, we a@:ain take the
ternary azeotrope O\'erhead, and we reco\"er an ethanol ..... ater mixture as the
nGUJ.: 7M
Columa conti. U'l.tiooL [hem D W.
T,dr-dD F. RMdd,,flOtEJ
JOJ (191111 J
B
A
C
Design 7
B
A
C
Design 6
note: that 6 corresponds to a binary mixture of B and C. which Wll5
tbe: ongmal sc:paratlon tbat llo'e: we:re trying to ma.lr:e, aoc:pt that it is mon::
th,an tbe: original fc:c:d, point I. Also. whe:n we: separate: Band C by
dlstlUatlon, we: can set the: bottom specification to give: us almost pure c.
IX?lnt 7. and. the overhead composition as the original feed mixture., point I. Thus.
we: must carry out the: same: B--e distillation as we would with just
dlstillatron. although the degree of separation required is reduced. or course, this
I. EJltractJ()d
1. EJllr.cuvt! dutillallon
J,. Auouopic dJstillillion
.... ReKbe dlStilb.tJon
5. Crysl.alliDlJOII
LA_
7.
186 SECTlOf' TJ UQUII) SErAaATION SYS'T1ol
Add nonvolaulc component to modify
"'CHON 11 UQt1![) n,...uTlO" 187
Add re<tcmc componem tu modify ")"5
8
r--8,-2S,-
c.g, B. C = xylcnc:s: a "'" 103 FIGURE 7.J-IO
S - organomc:tallic: B, CS: a 0=> 30 dlSl.iJl.alloD.
8
s
)
8
,J
C
r-
y
C
S
S
c
c
C+S
L-__c
Bes ternary heterogeneous azeotrope
Add volatile component that fomu an azeotrope
with one or more of feed componems
c.g., B = HNO)
e = H
2
0
S =
FIGURE 70U
distiII<tlw..
8+C
azcouope
L-_8
e.g., B == Ethanol
e == Water
S = Benzene
HCURE 7.3-'
Auotropt<: dulill.allOll
8
C
B + e az.eouopc
C
bonom strcam Now, in a third column. wc reco"cr purc .... ater. our second
product, as the bottom stream, along wilh lhe onginal bmary azeotrope o\crhead
ThIs binary azeotrope IS recycled to the first column. and .....e obtam pure producls
from the system of three columns
R."crl\'[ DISTILLATION. In some cases it is possible to add an entrdiner that
reacts with Onc component in a miXlUrc that is dlfficuilto For example.
the relative "olatllny between me/a- and para-xylene IS onl)' 103. Howe\er. If
sodium cumene is added \0 a mixlure of the xylene isomers,lt reacts with the para
isomer. and then the relative volatility between the meta-xylene and the organome-
talhc complex that IS produced becomes 30 The reaction can be reversed m a
second column. and the entramer is recycled (see Fig. 7.3-10). Thus, the original
separation is greatly simphfied. but at the expense of handling sodium cumene. If
eutrainers that arc simpler to haDdlc can be found, the reacti"e distillatIon will
become a more imponant separation altcrnative.
CRYSTALUZATION. Thc separation of xylenc isomers IS difficult by dlsullauon.
so often it is cheaper 10 use the difference in freezing poiDts to separatc the miAture_
Thus, by froclJng, separation of the liquid-sohd mlJ:ture, and orten using some
recycle. the desired separation can be achieved (see Fig. 7.3-11).
DISCUSSION. Extraction. extractive distillallon, and azeotropic distillation all
in\'olve the separation of nOOldealliquid mixtures Until reocntly there has!xen no
simple design procedure that could be used for the quick. screeDlng of these
alternatives A procedure of thIS type has recently been de\'e!opc:d by Doherty and
coworkers, and some of the baSIC Ideas of this procedure are discussed next.
188 SEClION JJ UOUID SI>AU,TlOto/ SnTUI
SEC"rlOW ,. -.nonol'lc SYSTEWS 189
.::um,R 7.3-1. IIOA 1lJe f10lVS or lhc flub liquJd ",tream llult are Jed 10 the
dIStillation tTam ale 10 Table 7.1-1 (1 F.Dd 465 psia). If"'e let tbe bpt ends
wtt? the prDduct and lecover all the product, the compositioo of the pt'"OOU<1
stream will be:
FlGUR.- 1J-11
Cry.laUu,atoon.
./

YPH
./ ..... 0.2
........ 0.4
/
""'- I" ___ 0.6
/
'"
-e- 0.8
0.7 0.6
x
0.3 0.4 0.5 0.2
2,000.000
1,5OO,OOCl
1,000,000
500.000

-$; 0
-500,000
-I,OOO,OOCl
-1,500,000
-2,000,000
-2,500,000
-3.000,000
- 3,500,CO:h.1

r--Vapor C (+ B?)
Crystallizer
D.C.
2354
;,." - 2 + II + 235,4 - 0.947 (7.3-1 )
fiGURE 1J-1J
Economic polenlial k\l1:1 4
F1GUR[ 1J-12
HDA pl"0Q::5$.
AZEOTROPIC SYSTEMS
prC!iSun:... remo\'e the and methane: first We send the: light ends to the fuel
supply
The flows of the components remammg after >1.e ro\er lhe light ends are
benu.nc: _ 235.4, toluene: _ 874. and dlphenyl _ 4 lhe heunSII(:S of hghtcst first.
most plentiful firS!. and favor c:qmmolar spltts all favor lhe duect sequence (reco\'er
benzene first). and shorlcut design calcul:ouions verify thIS result
When IIIe add Ihe purge Ioucs and the cost of the disuliallon columns (sec: FIg
73-12) 10 our economIC polcnllal calculations for Ie\el 3. we oblain lhe:
economIC polenllal for level 4 shown in Fig. 7.3-13. The range or the ...
where ..e observe profitable operatIOn hi15 beell further dc:aeascd, ....hich simplifies the
problem of add109 a network (sec Chap 8).
7.4
Distillation Boundaries
For ideal, ternary mixtures we can use either the direct 01 the indIrect sequenoc (see
FIg. 7.33) to oblam three pure products. However. for azeotroptc mixtures, the
feaSible separations orten depend OD the feed composition Henoc, It is necessary to
In tbe previous section where we discussed the sequencing of trains of distillation
columns, we assumed that it was possible to split the feed mixture between any two
components (see Table 7.3-2). However, if azc:otropes are present, it is often
impossible to achieve certain splits. This, for azeotropic mixtures it is essential to be
able to tdentify when distillation boundaries are present that make ocrtain splits
impossible.. Most of the discussion below concerning the behavior of these systems
has been taken from the papers or Doherty and coworkers..
'" g.
whlCb 15 k:ss than the required produa purity or 0.997. Heooe, we must remove tbe
lIght ends.
We wuld aHempt recover the Itpt ends in a pamal wndco5Cr at tbe top or
the column. but SlOce the reqUIred product purity is so hIgh. we expect that
we WIll need 10 use a slablhzer column 10 remove the IIgltt ends. The dc:sign of this
column IS dlscuued 111 Appc:ndlll B. SInce tbe stabilizer mUSI operate at an elevated
.

,
Roocto,
A"'"
H
2
CH.
,L
Denuoc

U

].

h>eoc

:E
'"
S
DiPhenylT

T
T
To
190 SIOCTION 14 ALIOOTIlOPIC SUUMS
understand lhe behavior of these prOLCSSCs In much greater dC:lail than the IdeJ.1
me
For example,. suppose we conSider the ternary mll:turt: of acelone, chloro
form. and benzeoe. We can plot the composItions on a triangular dIagram, and .... e
nOle the fact that Ihere is a maJ:lmum boIling azeolrope for acelollC--chloroform
blflary mixtures. We abo p101 Ihe boiling temperature; see Fig 7.4--1.
ow If .... e suppose Ihal ....e put a blflary mixture having a compo:'lIlon
corresponding 10 poinl A 1fI a simple stIli ,md c;onllnue 10 increase the lemperature
In the sliII. the composilJon of the malenal remaining In the sull ....111 move In the
direction of the arrow shown on FIg 741 (toward the bmary azeolrope). Also
mIXtures neh 1fI acetone would be recovered from the top of the slili.
In contrast, SHirting with a binary mixture correspondmg to point B on Fig.
7.4-1, as the slilltemperature is increased,lhe malerialleft in lhe still will again be
the binary azeotrope, but the overhead will be rich in chloroform. Binary mixtures
of acetone and benzene at point C or chloroform and benzellC al point D will both
lead to final stili mixtures of pure benzene.
Suppose no.... that e conSider lernary mixtures corresponding to points A
and B on Fi@ 7.4-2. As e increase Ihe temperature 1fI a simple the stIli
S!TION 14 AZEOnO',C 191
SO.1
Benzene
I
c
SO. 1
Benzene
56.2
flCURE 1.4-2
Terna., m""UfCS
Bmary
,""".rope
644
Chloroform
612
D
c
Acetone
56.2
A
Binary
azeolrope
64.4
8
Chlorofonn
61.2
compositions for each mixture will approach Ihat or the binary azeotrope, until at
some poin! they will tend to collide. Since benzene has a hIgher boiling pomt than
lhe binary azeOtrope, as ....e continue to increase lhe stiU temperature, the
trajectories (residue curves) w"l bolh lurn toward benzene. Thus, the final
composition In the still pot for both cases will be benzcne. Ternary mixtures
corresponding to points C and Don Fig. 7.4-2 will also yield benzene as the final
still composition.
There arc rigorous proofs for this type of behavior; see Levy. van Dongcn,
and Doherly" However, if we merely say Ihat for every source lbere must be a sml,
Ihen .... e can de\'elop reasonable pictures of the behavior; i.e., each traJeclory
(residue curve) must have a stopping point lhat is eilher a pure component or an
azeotrope (these polOtS correspond to the singular polOIS or the sct of differentIal
equations describing a simple still).
FIGURE 1.+t
AoolOQC-chlorolOl"m bclUl'llC Iyilem-blnary .'ilufd
S. G Levy, 0 8 ..,n DoDf;C"D, and M F Doherty, Ocs>&a &nd Syalbc:w of Azcolropoc DuullauOlL
II M.lD1IIIUDl lIeftw;. /6.C h",,-,,'j>/s, 24 463 (1915)
192 SEctiON 1.4 IYS'nM5
SFCTIOI'IH UEOnOl'lC S"\>TEMS 193
80.1
Benzene
FIGURE 7.4-3
Relidue CUn<e map. [F,om M. F onJG A CQ/holtJ, rue 24: (/985). ""ulr
pr'mLrsiotl of IIIe A_,kan C""micDJ Sod..I),.]
61.2
Chlorofonn
Separalrix or simple
disllllation boundary
Binary
azeotrope
64.4 56.2
Acetone
MORE COMPLEX SVSTEMS. As a more complex system, we can consider tern-
ary mixtures of methyl methanol, and hexane. Now each binary pair
exhibits an azeotrope, and all three are minimum-boiling azeotropes. If we put
binary mixtures conespondiDg to any points on the edges of the triangle in a simple
80.1
Benzene
However, if we start with a composition corresponding to X
n
in Fig. 7.4-4,
then we will obtain pure benzene as a bottoms stream from the first column and a
binary mixture corresponding to point B overhead. When we split this binary
mixture in a second column, we obtain pure chlorofonn overhead and the binary
azeotrope as a bolloms stream. Hence, the products we obtain depend on the feed
composition wherever a distillation boundary is present. .
Another way of stating this result is that pure chlorofonn cannot normally be
obtained in a sequence of two columns if the feed composition lies to the left of the
distillation boundary, whereas pure acetone cannot normally be obtained if the
feed composition lies to the right of the boundary. The relative volatility of
components in a mixture close to the boundary changes from a value greater than
unity to a value of less than unity if we just move across the boundary. Of course, if
we split the binary azeotrope in an additionall;:olumn system. we could recover all
three components in pure form.
FlGt:RE 7.4-4
Sequence of 1"0 ronunuous columns.
61.2
Chloroform
Separatrix or simple
distillation boundary
Binary
",""ope
64.4
56.2
When ",e consider more starling l;:onditions, .....e obtain the results shown in
Fig. 7.4-3. Now we see that there is a distillation boundary going from the binary
azeotrope to benzene that divides the composition triangle into two distinct
regions. Feed mixtures to the left of this boundary produce acetone-rich mixtures
overhead and lead to pure benzene in the boltom, whereas feed mixtures to the
right of the boundary lead to chloroform-rich mixtures overhead and pure benzene
in the boltom_
SEPARATION DEI'ENDS ON TIlE FEED COMPOSITION. Suppose thaI we
now consider the separation of a feed mixture having a composition of.x in Fig.
7 4
. FI
.4- In Iwo contmuous columns using the indirect sequence. It can be shown that
the distillate, feed, and bottoms compositions for a single column must fallon a
straight line (this is the material balance expression). Hence, if we remove
essentially pure.benzene from the bottom of the first column and recover essentially
all. the m the the will correspond to point
A In Fig. 7.4-4. Now, If we spill the bmary mIxture corresponding to point A in a
second column, we will obtain esscDiially pure overhead and the binary
azeotrope as a bottoms stream_
Acetone
690
Hexane
AZeQITopes
SOO
51.8
SITION H AZEOTltOI'I('" SYSTEMS 195
690
lIexane
Azeotropcs
SOD
'1.8 L
Melhyl acetate
57.0
53.5 M<ilianol
64.7
Methyl acetate
57.0
.JL
53.5
Methanol
"'7
(1.4-2)
(74.1)
flGURE 7......5
MelhylaClttale _lhanol MUrK mllal) mUllUQ.
still, the composition of the material remalnmg in the still W1U move In the same
direction as an Increased still temperature. Thus, the arrows in Fig. 7.4-5 corrt:.
spond to the dlrecllon of Increasing liquid compositions remaining in the still pots.
When we consider a set of ternary mlXlures that are close to tbe sides of the
triangles and recognize that the still composition must move in the direction of
increasing temperatures, we obtain the results sbown in Fig. 7.4-6. Since there must
be 8 source in the inlerior of the triangle for the trajectories to behave in this way,
there must be a tt:rnary azeotrope which has a lower boiling point than any of tbe
binary azcotropes.
A residue curve map showing more trajectories (residue curves) is shown m
Fig. 7.4-7. Now we see that there are distillalion boundaries that divide the triangle
into three distinct regions: ADGE, BDGF, and CEGF. Depending on where our
feed composition falls in Ihese regions, we will obtam different products.
Minimum Reflux Ratio
For ideal systems, we can use Underwood's equation to calculate the mmimum
reflux ratio. Then, after we select a reflux ratio of 20% or 50, larger than the
FIGURE 7.....
Te.-ry mululn
minimum \"alue, we can calculate the \'apor and liqUid Rows throughout the
column. We use this mfonnatlon to desIgn the column.
A procedure for calculaung the mlDlmum renux ratIo for Donideal mixtures,
including azcolropic systems., has been developed by Doherty and coworkers.- We
diSCUSS tius procedure now
SYSTEM EQUATIONS. The material balance equallon for the strippmg secllon
can be wollen as
X,"'l-c: I)h.. I) XI.
where s is the reboil ratIO. H we subtract X
t
.. from both sides of the equation, we
obtain
S G. l.c:vy, D B van Donaen, and M F DoMrly. M Iks'Cn and of AuolrOpK
II. Mlllimum Rdlu. Cakulallons, I,fEe f.. 24 463 (1981)
196 SECTION 10 Al[OTllorIC 5YSTF.MS
srcnoN" AU01J.OrlC SYSTEIoIS 197
Now we the left-hand side of the equation by a derivative with respect
to column height:
(74--6) 'l:,
R 1.5
Legend
+ Feed composition
o 8oI:tom composition
Condenser liquid composition
Ii Distillate composition
Mol '1 r=d Dlstillale Bolloms
Ilexane 0.3 0999 0.001
lIeptanc 03 0,001 0,428
Nonane OA 10 x
10-
1
0.571
+
0.6
0.2
0.'
0.4

Non31lC XI Hexane
H<pun<
10
J. D=(' +: + I)(ZV_ X',D) +
Wf'; can use these results 10 de,'clop a procedure for calculating the minimum n:nux
rallo.
x,
Eq 7 I. Thus. the differentIal equatIOn will provide a rigorous calculal10n of the
plOch compositions
An o"erall matenal balance for the column gn'es
IDEAL MIXTURES. To understand Doherty's prOttdure. \loe firsl conSider the
case of a hexane-heptane-nonane ideaJ mixture. If we are attempting to make the
sharp split A/Be ror lhe feed composition shown in Fig 7.4-8. then the distillate.
feed, and bottoms compositIons all must fall on a straightline(thecolumn matenal
balance must be satisfied). Every component will distribule to some extenl, and thc
componenl speclfical10ns are gi\'en on Ihe figure.
If we fix the column splits and the rcAux ralio. we can usc Eq 7.4+6 to
calculate Ihe reboil ratiO, 'ow. we integrate Eqs. 7,4-3 and 7.4-4. starting from the
ends or the column_ If the reflux ratio chosen is below tbe minimum. Ihen the two
traJCClones do not mlersect see Fig. 7 4-8 raJ the beJianc.heptaoc.nonanc: example.
(7.43)
(7.4-4)
(7.4-5)
Residue curve
Az.eolropes
690
Henne
1.0 A
O'
06
02

Methyl aett3((: Xl Methanol
570 64.7
X,
51.8

= (5: 1))0, -Xi + (5 ,)X;..
Similarly. for the recufying section we obtam
= (,; I)Y' - x,- O)Y',D
where r is the reAux ratio.
At mfinite reflux. Eq 7.4-4 reduces 10
which IS jusl the equal Ion for a simple shll Also. pinch pomts exist when
dxddh - O. or x ...... I ... x,.... and for Ihis COndition Eq. 7.4-3 becomes idenhcallo
FIGURE
'-tban lDIDUDum I From S """0 0 B ''OIl <JNI N F OoNt"', IMe Fund"m"u,,!J,
24 463 (/oaj), ..,,11 P't''''W/On fr_ A"",KIlIf CMmocm
198 UCTION H .u:lOTIOf'IC SYirrliWj
Heptane
1.0
08
FlGURE 1.4-9
Gruta than IIlIIWDwD Rft..... S. UrY. D B &>all Doagm. ll1td JtI F
U 46J (/9Ij), ""i,}, ".,mU5,Ott!'om 'M A..."..cmI CMm...ol Sv'J)
If the selected reflux ratio exceed::. tbe: minimum value, then the profiles for the
stripping section and rectifying sc:ctlons cross, and we can oblam the desired
separation (see Fig. 7.4-9). (fthe reflux rallO corresponds 10 Ihe mmimum, Ihen Ihe
pinch 'Zone for Ihe strippmg secllon Just ends on Ihe profile for Ihe recufying seeuon
(see Fig. 7.4-10)
Now if we ..:hange Ihe mole fraclion oflhe heavteSI component in Ihe dlslillate
from xf/ = 0.001 10 x,. = I x 10 II and rc:pcat Ihe calculatIOns, we oblam Ihe
results shown In "Ig. 7,-1-11 We nOle that Ihe mimffium reflux ratio for thiS case IS
R. _ 154 IflSlC:.ld o( R. = 215. We also nOie Ihal Ihe rectifying and strippmg
profiles exhibll very sharp corners, .... hlch correspond 10 pioch 'Zones.
Anolhe:r feature 10 Fig. 7.4-11 IS Ihat the: plOch composition in lbe rcetifylflg
section {poinl P 00 Fig. 7.4-11).lhe plOch point for tbe strippingscctlon (point Q).
and the feed composItion F are collinear. ThIS result can be rigorously proved for
ideal systems, and the result can also be shown to be equivalenl to Underwood's
equalions. However, the same result is approximately valid for nonideal syslems.
COMPLEXITY OF THE PROBLEM. There are 113 types of residue curve maps
lhal can be drawn (or lernary mlxtures.- The: maps Ibal are useful for selectmg
entrainers for blOary mixturc5 with minimum boiling 8zeolrOpc:s, lolal of 35
posslbilitic:s, are shown in Fig. 7.4-12.' The number of possiblll1lcs grows very
rapidly as Ihe number of components prestnl in lhe mixture increases. Hence,
azeotropic systems present a formidable challenge as a separation problem
However, the geometric ideas presented above can be used 10 develop an
expression for the mmimum reflux ratio for azeotropic syslems.
o.
X,
04
0.2
o
NONO'"
Heptane
1.0
02
+
R = 25
04 06
X,
08 1.0
Hmne
o.4'y, 1&.(
NONIDEAl. SYSfEl'>1S. Now suppose: we consider Ihe system of acelone:, chiou>
fonn, and benzene: We plcl a reflul. rallO below the: miDlmum value, \lie: use the
lenninal composllJons shown m Fig. 7.4--13, and we apply the: procedure described
above. Then the profiles for the rectifying and stripping sections do not IDtersect
(see Fig. 7.4-13). However. if we are abcwe the minimum reflux, the curves cross (see
Fig. 7.4-14); and If .... e are at minimum reftux, the pinch region (or the: slripping
section just ends on the profile for Ihe rectifying section (see Fig. 7.4-15).
When we change the end composlIlons, we obtam the results shown lfl
Fig.7.4-16, and the minimum reflux ratio decreases. Moreover, we note from Fig
7.4-16 thaI the pinch zone for the rc:c1ifymg seclion al minimum reflux (point P on
Fig. 7.4.16).lbe pinch point for the siripping section Q, and the feed composilion F
are essenlially collinear. (This example: shows tbe largesl deviation that has been
observed for numerous casc: studlCS.) This collineanty condition provides a
criterion for calculatlng the minimum reflux ralio.
H and II J N'5hunur.. J. Clotm JfHI, 10 111(1977)
'M F Doheny and G A Caldarob. I&I;.C U 474 (I9SS)
, S. G Lev,. D 8. vall DoaJCR, llId M .. Dohen,. -Oc:Np alld S)"Dlbem 01 DlSU,llaUOll
II MlllUnlim Itdilill Cakubloont.,-'&EC U 463 (1915).
08
0.6
R = 2.15
X,
04
0.2
oLL--'of.2'---'0!-,"'-<!0.';;6-10r-.'8--j,"C.O
Nonane: XI Hexane
f1GUIU.. 7.4-1'
MuwDvm rdIu. [From S Lny,D B &>all ... DNJ M F DoIwr'1, IdEC FuruJonw"ftW, 2-1 46J
(lSISj}. ..,A ".nJft1M_!IDM llot A..-K'1JI'I CJwm.cgJ Soocotly]
'99
SECTlON l' Al:I:OTltOPIC SYSTEMS 201
200 SECTION H AZF.oTJIO!'lC nJTQ.lS
Distillate Bottoms
,
,
Heptane
Mol" F<cd

'0
Hexane 0.3 0999 0.001
,
Heptane 0.3 0.001 0.428
,

0'
10 X 10-
11
0.571

., -
--
. ., - Legend

0.'
+ F<cd

o BOHOIUS
"-
Coodcn$("r rIqUId
'-

,
0.6 -
.l!t. Distillate
l.-'"

--
,. -
p
--
,.
-
,
o. _ X,
0.4
R e 1.54
Q
F
0.2
.-
..,
'. -
..
.

0 0.2 0.4 0. 0.8 1.0

Nonane
X,
Hexane
...
.n_

...
'"
_.,
FIGURE 7.4-11
.
Mlnlltlwn rellu1-higher'purily spill. [From S D. B "'"" and}of F fMC
,
46J (/985), "'II1a I't-""u.s"",j,om, A.....nc.... 5.,.....,y)

We can write the equations describing

.. ., -
"
EQUAnONS FOR MINlMIJM REFLux.
..
tbe rtttifying pinch
rz, - (r+ I}", + YD ""' 0
(7.4-7)
and the feed pinch
s)'.-(.J+ 1)x.+ZD=O
(7,4-8)
-
,
..
-
,
....-
... -
-
,
., -
where each of these equations contains eJlpressions for the two key components.
The reftux and reboil ratios arc related by
s = (r + 1)(X.'1 - ZT,I) (7.4-9)
X",I - YD,I
and we IlSC a vapor-liquid equilibrium modd to n:late ;re, and y,. Then. the
collinearity result requires thaI
(74-10)
The value of r thai satisfies this of equalions corresponds to the minimum renux
ratio. Underwood's equalions are a special case of this new general fonnalism
0.001
0.749
0750
0.990
0.009
0.001
R = 74
I.e ,""
+ Food
o Bottoms
Condenser liquid
.6 Distillate
Azctllrope
0.120
0660
0.220
_on,

Chlorofonn
0.2
0.61'1------\
0.4
0.8
Henzene
1.0
X,
Distillation --+__.},
000"""",
0.001
0.749
0.250
R "" 6.5
0.990
0.009
0.001
I.e.<nd
+ Food
0_
ColKienser liquid
.6 Di.sIillale
Az.oouope
0.120
0.660
0.220
0.8
0.4
0.61'-\---\
1.0
0.2
x,
Dislillalion
000"""",
ChI
O'--'0f..2'''''0"i.;;4
omfonn X,
Acetone
FIGURE 1.4-13
l.aa thaD mirumWZl rdlul. [Fr_ S. La,. D B. -. Donge<, INId AI F. {)oM'" I&EC FtuU1DmnI
14 46J (198j}, "'1I1t prrmissiollfrom 1M A",mcaor CJwmkDJ SOd,UY.) '. '"is,
o 0.2 0.4 0.6 0.8 1.0
Chloroform Xl Acetone
FlGURE 1.4-15
Minimum rdiu. [Frpm S UvY. D B 00/1 Dongol. QIld M F. DoN,')', IdEC 14 46J
(J98J). ,,<j,1I pMwiottj,um 1M AJIIl"U:-1Vl CIwmic<l1 SocWI)'.]
0.4 0.6
X,
Food DistillalC 8oIuxn>
0.120 0.990 0.001
0.660 0.009 0.749
0.220 0.001 0.250
1.0
Acetone
1.0 X
0.749
0.251
0.8 0.6
Feed Disciliate Boctoms
0.120 0.990
0.660 0,003
0220 0.007
Legend
+ Fo<d
o Bouoms
Condenser hquid
b. DlSlillate
Az.eoI:rope
Rectifymg
..woo
X,
R = 7.35
\_-----"<;'"' Straight line connecting
, saddle to reed poinl
Aeecone
Benzene
Chloroform
0.8
0.2
0.4
1.0
0.6.1-1''-.10
Benzene
o 0.2
Chloroform
x,
Stripping
..woo
Separatrix ror_-+---'''
simple distillation
1.0
Acetone
0.8

+ Food

" Coodenscr liquid
6 Distillate
Az.oouope
R 8.2
Acetone

Chloroform
0.6,1'T_-\.
0.4
0.8

1.0
0.2
X,
o 0.2
Chlorofonn
Distillation
000"""'" --t--.\
203
202
DlIIoen,. l<fEe
FlGlJaE 1... 1.
MUlUIlWD n:flla _lib doecreucd boa")' cwnPOIiCiU 1Il (be ovnbad. [F,..... S. u.:,. D. B. -
tIIId AI F. DoIw"Y. /d.EC 24. 463 (/98J)" WIlli ..... 1M A.......1t'AII Clt..""ca!
$ocw'y.)
The toluene kaving the reactor TOL
II
._. is the toluene that was not converted in
the reactor
ToIUf:M balance. The toluene entering the reaetor TOL,II"ft is the sum or the fresh
feed toluene TOL,.,. the toluene in the gas-recycle stream TOL
ell
aod the toluene
in the liquidrycle stream TOL
u
:
E:umpk HOA pr_ The procedure is best iIIuslralc:d m Icrms of an
example. aDd for Ibis purpo5C _"e ehoose the !lOA process. The is shown in
Fig. 7.5-1 Now. _"e ....rue balances lOr the component flows or eacb sl.a.rtlng
lbe hmmng reactanl
These: equations arc always linear. and therdore they arc sImple to solve by eIther
matriJt methods or simple substitution. Normally, we stan with a balance for the
hmlting reactant. and then we consider in tum the pnmary product. other
reactants. by-product components. and inert materials.
Not all the fractional recovenes (or losses) of the components in various units
can be chosen independently For example. the sImple flash calculation procedure
dcsc:ribed by King (Eq 7.1-19) shows thai if the fractIonal recovery of one
component 15 filled, then all the other fractional recoveries can be ca1cul:ued.
Similarly, the fractional recoveries ror a producl column must be fi.\cd so that the
product purity specification IS satisfied. and in some cases the fractional recoveries
for purge streams must be chosen so that constraints on molar ratios at the reactor
inkt can be satisfied. Iknee. in some cases some iteration might N required.
(7.5.2)
(75-1)
TOl
lI
_ =TOl..... (I - x)
TOl
lI
,;" ""' TOl,-,- + TOl
clI
+ TOLur
EXTENSIONS OF n-fE METIfOO. This approach ror nonideal systems has beea
utended to multiple-feed streams" columns with aoanegligible heat effects.'
heterogeneous azeouopic systems,' prooedures far seleding eatrainers., and apti-
mum design and sequencing. Once the minimum reflux ratio has beea c;;aJculated,
we c:an let R _ I.2R. and then design the column, roUowing the same procedure as
\o\'e U2d for Ideal mixtull'S. A procedure for cakulating the minimum reflux ratio
for systems with 4. or more. components has fe(XIltly become available
After we have selected a liquid separation system. we have completely filled all the
units in the f10wsheet where the component flows change. Tbc:se units include
mixers (for fresh feed and recycle streams), splitters (for purge. streams). reactors.,
flash drums (phase splitters). gas absorbers (and/or other vapor recovery units).
and distillation columns (and/or other liquid .separation systems). Thus, we can
now develop a set of rigorous malerial balances.
Of course. if our rigorous halanoes differ signiflCantly from our earlier.
approllimate results, then we Wlij need to review the decisions that we made. We
could have revised the material balance calculations at any stage: of our develop-
ment of the design, and clearly there is a trade-off betwccn the time required to
perform all the calculalions and the accuracy of the answer. Our goal is to complete
the design as rapidly as possible. providing that major errors are not introduced,
and to e"plore the alternatives using appro"imate calculations. Then after we have
identified the best ahemati\e, we will use rigorous calculation procedures. How-
ever. remember that it is not possible to make rigorous material balances until we
have compktely defined the pans of a flowshcct where the component flows
change.
7.5 RIGOROUS MATERIAL BALANCES
lfwe Iet!TOl..'" be the fraction of the toluene leaving with the flash \apor TOl,...
then a fractioo I - !TOl.,.. leaves witb the flash liquid TOL
n
:
If ..e let In; be the fractioo of toluene lost in the purge TOl
rc
, tbell a fractioa
I - In; of the toluene will N in the gas-recycle stream T0l..-,II:
Uneu Material Balancing
The procedure we use to develop rigorous material balances is called linear
mlJferial balancing (the set or equations generated is always linear and therefore
easy to solve), and it was first described by Westerberg.
1
To apply this procedure.
first we draw a fJowshcet so that it contains only those units where component
flows change. Then we write matenal balances for each component individually in
terms of the molar flow ratcs and the fractional recovery (or loss) in each unil.
TOl,.. - ITol.,... TOLll ._,
TOl
n
- (I - IU)LJr) TOl,ll_
(7.5-3)
(7.5-4)
If we let ITOl.5T be the fraction of toluene that leaves with the stabilizer distillate
TOLn..o. then a fraction I - ITOt..5T will leave with the stabilizer bottoms
TOL
sr
.:
S.G. LeYy. and M F Doben.,. and DWitlatio.... IV
Minimum Reltu.l Calculalioos lOr Muhiplc Feed Columns,- '.tEe 15: 269 (1915).
I J. R. Knishl and M F. Doberty. MOcsiI" and Synlhaili 0( 1I0m0w:nc0us Azeol.opoc: DislillahOnll. V.
ColumN wilh Nonncl'ipblc hcBl EIJecu.- /dEC FruvltJ_",ah, 8: 219 (1915)
I II. N Pham and M F Doberly. M Deisn and Synlhai, of Ilctero8CIll!OUI AUOlropic DiSllllat/On I
llC1rroaeneous Phase o'a.........- COW... big Sci. (t985).
I A W WClIlcrbcra. -NocCli for a COUI'llC on Chemical P._ IJa>&n,- laUpl al the tnsUlule ck
OcuflOk) TCClt06op:o poOl Ia hodlG<lna QuImic::a (lNTEC). Santa Fe. A'am'",a. A",\lSt 1971
TOL
rc
= II'(; TOL,...
TOl
clI
"" (I -Ire) TOL,...
TOL,.T.D - ITOt..ST TOL".
T0LsT. - (I - IrOt...ST) TOLn
(75')
(75-6)
(757)
(1,5.8)
, CII
4
Flash
vapor
-
Reactor Reactor
on oot
Mox Reactor Fluh
~
Recycle Product Slabilizer
o""head """head ""'th<ad
Flash
I ~ ~
J-
liquid
t
n

,
~
-
-
]5
"
~ S
~
J
T
Now if we comblflc Eqs. 7.5-6, 75-3, and 7.5-2 to soh'e for the gas-rec)'c1c
flow, we obtain
We can use this result to solve for all the other tolueM flows.
NOh that if there is no loss of toluene from the process, i.e-,
(1.5-17)
fTOL.sT = 0 hOL..,. = 0
TOL,.,
TOL
II
,. - ---
>
[TOLIC = 0 frG "" 0
TOL
G
- TOLIl.I.(1 - froXfToL.n--XI ~ x) (7.5-13)
Also, if we combine Eqs 7.5-12. 7.5-10, 7.5-8. 7.5.4. and 7.5-2 to calculate the
liquid recycle flow, we obtam
TOL
u
= TOI
I
..(1 IloL.l,c)(1 1'01.".)(1 - llOl.sT)(1 - fH)I..nXI x)
(7.5-14)
Next we substitute Eqs. 7.5-13 and 7.514 mto Eq 7.5-1, to obtam
TOL.
I
.{1 - [(I - IrGXIToLn)
+ (I - flOL.leXI - flou.XI - JlOUTXI - flOL..fY))
(I - x)} _ TOl,., (7.5-15)
Ihen Eq. 7.5-16 reduces to
Purge
SLabiluer
boll"""
Product column
boctoms
Recycle column
bO(lom$
Gas recycle
To
H,
FlGUIlE 7 ~
IIDA Pf0c:c:s6.
which is the Simplified approximation that we used previously.
We try to select the fraellonal recoveries in these: equations such thatf, will be
a small number. However, the purge split/'G is the same for all components, and
the SpillS of the components in the flash drum are rdated to one another
If a fraClion [TOL.,. leaves with the benzene product TOl,. p, then a fraction
I - [TOl.rlC will lc:a ..e the product column in the bolloms TOl,...:
TOl,..p = [TOl..rl TOl
ST
. (7.5-9)
TOl,I. = (I - lTOL'I) TOl
su
(7.5-10)
Finally, If a fraction/Tol.le is lost with the diphenyl by-product Stream from the
recycle column TOlp,then a fraction I - lTOL.le is recycled to the reactor TOL
u
:
BENZENE BALANCES. The balances for benzene are essentially the: same. except
for the reactor equatIon. That is, at the reactor inlet we obtain
(7.5-19)
(7.5-18)
Toluene Converted ... TOl.x
where Ihe fresh feed flow of benzene 82,., IS equal to zero. Accordmg 10 our
selectivity corTeiation, a fraction S of the toluene converted appears as benzene,
although it is important to remember thai this correlation was based on a pure
toluene feed stream. Thus, we expect that some of the benzene recycled to the
reactor will be converted to diphenyl, and if the benzene recycle flow is significant,
we should revise our correlation Neglecling this discrepancy until we estimate the
benzene-recycle now, we can write that the wluelle converted in the reactor is
simply
(7.5-11)
(7.5.12)
TOllJ "" fTOLIC TOl,..
TOl"l -= (I - fTOL.cJ TOL,I.
208 seCTION n _IOO_OUS MATeRIAL IALA/'fC1!S
The material balances for tbe other componenl.5 are developed III the same way.
with a few exceptions.. Thu IS, ...,-e assume that there is a nesJigibk amount of
diphenyl in the flash vapor stream (see Table 7.1-1). Also, we assume Ihat aU the
hydrogen and methane: in the: Oash liquid that is not rcc::overed in the stabilizer
leaves ",oith tM benzene product, i.e.. there is no hydrogen or in liquid-
recyck stream.
where we can substitute Eq. 7.5-16 for TOL
II
. Hence the ben7.ene leaving the
reaetor is the benzene produced (xS(fOL
II
....plus the benzene fed to the reactor;
BZ
II
__ BZ
II
.,", + xS(TOL
II
...) (7.5-20)
Letting/,z.,v be the fraction of benzene going overhead In the flash drum
(which is related 10/r0l..1"I b)' Eq 7.1- J9) and I,G be the: fraetion of lost 10
the purge (whK:h is the: same for all components), we can show that the gas-recycle:
ftow of benzene IS
methane. the reactor con\'erSlon, the spill fracttons of the components 10 the nash
drum (which are related to each other by Eq 7.1-19 and depend on the tempera lUre
and pressure of the flash drum). the split fraction of the purge stream, anrl the
fractional reco\'enes of the components in distillation tntin.
Optimizatinn Variables
In our pre\'ious approllimate material balaocn..... e specified the production rate of
benzene. the product punty of benzene. the plllge composition of hydrogen (.... hlch
showed ""as cqul\alcnt to specifying the fresh feed rates of hydrogen and
methane). the con\'ersion, and lhc molar ratio of h)'drogen to aromalics at the
reactor inlet. For our linear material balance: problem we can assume that the
conversion and makeup gas flows are optimization variables (since the: feed
composition of the makeup gas stream is filed, specifying the makeup gas flow lilies
the fresh foed rates of both hydrogen and methane) As ....e diSCUSsed earlier_ these
are the dominant optimization variables.
The fractional loss of benzene overhead In the stabilizer also corresponds to
an optimization problem (loss ofbc:nzene to fuel versus the number of trays in the
rcctifying section and the column pressure). The fractional losses of toluene and
diphen)'1 in the Slabili7.er are then fixed by the column design. Specifying
the fractional loss of methane in the stabilizer bolloms will fix the design of the
stabilizer (small losses correspond to a large number of trays in the stripping
section). and ona: Ihe column design is fiud, the h)'drogen loss in the bottoms is
fixed. Ho\o\ever. we expect that all the hydrogen and methane leaving in the
stabilizer bottoms stream WIll also leave with the benzene product. Then the
fractton of toluene that goes in the product column and leaves with the
product stream plus the hydrogen and methane leaVing with this stream is
fued b) the Specified production rate and product purity,
To obtain small amounl.5 of overhead in lhe product column..... e
must include a large number of trays in the rectifying section of this column. Thus,
then: is a trade-off between using a large number of trays in the stripping section of
the stabilizer (to keep the hydrogen and methane 80ws in the produet Slream
small) balanced against using a large number of trays In the rectif)'ing section of the
product column [to keep the toluene (and diphenyl) 80ws in the product stream
small), for a c:ase where lhe sum of the hydrogen., methane, toluene:, and diphenyl
Oows is fixed.
The fractional loss of benzene in the bottoms of the product column is also an
optimization variable (trays in the stripping section balanced against the cost of
recycling benzene through the reactor system), as are the rractionalloss of toluene
in the bollom of the recycle column (toluene lost to fuel versus trays;n the stripping
section) and the fraetionalloss of diphenyl overhead in the recycle column (recycle
COSI.5 of diphenyl back through the reactor versus trays in the rcctifying section).
To avoid all these separation system optimi7.8tions, we fix the fractional
recoveries of the keys to correspond to the ruleof-thumb value of grealer than 99 0/0
and we fu the fractional losses of the arbitrarily as 0.15 to 0.3 times the
fractional losses of the keys Alternativel}. we could use Fenske's equation to
(7.5-21)
(7.5-23)
(7.5-24)
SImilarly, .f we Ic:t I.z.n be the: fraction of benzene lost o\'erhead In Ihe stablhzer
and/
u
.,11 be the fraction of benzene lost in the bottoms of the product column, and
if Wl: assume that all the benzene goes overhead in the recycle: column., then the
liqUId-recycle flow of benzene is
BZ
UI
= fIlJl:,I'II(1 - f'Z.ST)(1 - f.z.,..)[BZ
II
.... + xSn-OL
II

h
.)] (75-22)
Substltutmg Eqs 7.5-21 and 7.5-22 into Eq 7.5-18 gives
nZ'I.I.. [1 - I'Hv(l - f,d -/117..,11(1 - ltu..srXI - f.z '1')]
= xS(fOLII.I.)[flll.n (I - II'G) +l.l.,II(1 -/117..51')(1 - I"z.n)]
BZ
xS(TOLII.lJ(f.z.n(1 - I,G) +ltu..,II(I l,z.ST)(1 - In,Tv)]
1....
, J l'lJ,,{1 f,G) l.z.,,J1 f'Z.ST)(1 IlIl.'-")
We can now usc: this result to calculate all the other benzene flows.
Other Component Flo"'s
Linear Material BalallCe5
From the discussion above we see that by writing balances for the molar 80w of
each component in tenns of tbe fractional recoveries obtained in each process unit,
we obtain a set of linear equations in terms of the conversion or the limiting
reactant and the selectivity (which is related to the conversion). These equations
are simple (although somewhat tedious) to solve for the recyde flows of each
component. Once we have calculated the recycle 80ws of each component, can
calculate all the other flows of that component.
An inspection of the resulting equations indicates that we must SpeCIfy the
fresh feed rate of toluene, the fresh fc:ed rate of hydrogen, the fresh feed rate of
C:Sllmate the fraroonal IQM of the nOllkC)'s Thus. our matenal bala.nccs are 00(
ngorous, but since we e-lpcctlhat these loss tenns to be small we do 001 introduce
much error
Constraints
Most of the flows can be \l.n[len In lerms of [he fresh feed rate of toluene TOl
(see Eq. 7.5-16). However. we want to sol\-e the design problem in terms of
production rate PROD of benzene Hence, we need to sum the flows of benune
hydrogen, methane, loluene, and dlphenylleaving the lop of lhe product
and lben eliminate TOLn from these expressions aod replace it with PROD. This
procedure will remove the production rale constraint.
In addition, we must wrile the expression for tbe hydrogen-to-aromatics ratio
at the reactor inlet, set this value equal to 5/1. aod then solve for the fractional spill
of tbe purge stream!rG that satisfies tbis e-lpression. This procedure removes the
other process constraint.
Unfortunately. the algebra required to remove these constraints IS tediOUS
Thus, it might be easier to solve for the recycle flows of each component, solve for
all the other componenl flows, and then adjust Ihe solutions, i.c., ilerate, until tbe
constraints are satisfied. Allematl vdy. one of the computeraided design programs,
such as FLOWTRAN, PROCESS. DESIGN 2000, ASPEN. etc., can be used to
rt\"ise the material balance calculations. We diSCUSS the use of the CAD programs
to revise the material balances later in the text.
Enmple 7.5---2 IIOA The exprcsslon for loluene feed rate 10 the reactor 15
g"en by Eq. 7.5-16 To IbiS flow, we mus, spcc.:.tfy the tenns 10 tbe equallon
TOl.... Our original dcslgn problem speafics tbe desired prodocuon nne of
benzene, ralba than tbe fresb feed rate or tolueoe. HOllrever, from our sboncut
baJaDCe$ (With DO losses) we fouad tbal FIT _ PJS (see Eq. 5.21). For a case ....bere
p. 26j. x - 0.75, Ind S 0.9694 (see Appendix 8) Fn _ TOl" _ 273.4. We ca.n
UK tbis ut,mate In the finl IOluuon and tben use i'cntion 10 rorrCCl tbe value.
f,.(> Using our shortcut calculations, we found thai the purge ftow rate was
496 molfhr and that the gas recycle flow was 3371 rnol/ltr for a CllSC where x _ 0.7j
and Y,.H - 04. Hence, the fraction of tbe flash vapor tbat is purgc<l from tbe process is
496/(496 + 3371} - 0 128 We use this as. first gucss, .nd then we iterate to m,urn
the problem specifications
fT04.." The retuJts of the shoncut flash cakulations arc liven in T.ble 7.1-1,
.nd we see tb.tfTOL..... - ],6/91 - 0.0]96. AgaIn, we need to iterale to m.tch the flash
drum operaling cooditions.
IHJL.ac The fraction of toluene talcen overhead In tbe recycle column is aD
optimization variable.. For our first design .. e choose f":J(_,ac *"' 099S
11OL.,.a The fracllon of toluene taken overhc<lld III tbe product column is also
an opllmizalloD vanable (lbe: ImouDt of toluc:ne pha mc:lIl11ne t.lien overhead IS find
by tbe product SpeaficatlOM, but ellher composltina can be adjUSled). For our first
design \Ioe might assume tlut the ImpuntlCS ID tbe product jue I SOISO malure or
mc:lluone Ind IOllIoCne
ItOtJT TIt.e fraCllon of loluene leavlllg overhead III the stahlhur depends on
the or lhe spIll belw=n mcth:tne "'nd bcnunc. 511lcc ....e tale some
bcn:.tene overhead in Ii1lS column to ensure :tn adequate Sllpply of rellul, we do not
expect 10 obtaIn a sharp sphl. For a first desIgn .. e fix 'he effluent cooling.waln
temperalure m Ihc partial COndenser used m thiS column as I ..e choose lhe
l:Qnden5mg lemperature as 115 or 120 F; and .... e hi the column prtsSure so Ihatthc k
.... Iuc: or bc:f12eoc IS K. _ OOj If Ihese results arc rea50oablc, 'ben 'lot find the " V1Iluc
of tOluene In Ihe reRull drum, and .. e can e$lImale the toluene loss
For Ihls example, tbe amount ofeffort requIred to iOlve tbe ngorous malenal balanus
by usinglillear malerial balances probably exlXCd5 the effort reqUired to use a CAl)
program
7.6 SUMMARY, EXERCISES, AND
NOMENCLATURE
Summary
The deCISions .. e must malic 10 synthesize: a separation system fall mto three
categories: the general structure. the vapor recovery system, and the hquu..l
separation S)"Slem Tbcsoc dCClsions arc listed here.
I. General structure
Q. Do we need both liqUId and vapor recO'er)' unib, or Just IIqutd'
2. Vapor recovery s),stems
a Should the vapor reco\ery system be placed on the purge slream, the gas
rcc)'ck stream. or the flash vapor stream? Or. is il better OotlO tndude one?
b Should \Ioe use a condensation pr()C(:SS., absorption, adsorption, a membrane:
process, or a reactor system as the vapor recovery system?
3. LiqUId separalloo syslem
Q. How should the light ends be separated If they might contaminate the
product?
b What should be the desllnation of lhe light ends?
c. Do we rec)"CIe components that form azc:olropes with a reactant. or do we
spilt the azeotrope?
d Whal separations can be made by distillation?
What sequence of should we use?
f How should we accomplish separations If dIstillation is not feasible?
Some design guidelines that are helpful in making the decisions above are
listed here.
1. The general structure \I.e choose for Ihe separ,itlon syslem depends on whelher
lhe phase of the reactor effluent IS il- liqUid. a two-phase mixlure, or a "apor
The three types of ftowshc:et arc shown in Figs. 7.1-2 through 7.14 In cases
where the reactor effluent is a vapor and do not obtam a phase spIll when
we cool the effluent to I<XrF, either wc pressurize the reaclor (if the feed and
recycle streams are liquid) or .....e Install a compressor and/or a refrigerallon
system to accomplish a phase spilt If a phase split results in only small
amounts of either vapor or liquid, \lie mIght delele the phase spllller and send
the reactor effluent to either a vapor recovery or a IiquKl recO\er) system
2. We Install a vapor reeoy'ery system on the purge stream If lose valuabk
materials WIth tlte
3. We Install a vapor rccoy'ery system on the gas-recycle stream If some recycle
components would be deleterious to the reactor operation or degrade the
product distribution.
4. We Install a vapor recovery system on the flash vapor stream if bolh Items 2
and 3 above arc Important.
5. We do not use a vapor recovery syslem if neither item 2 nor item 3 above is
imponant
6. Our ehoiccs for a vapor recovery system are condensation (highpressure or
low-temperature or both), absorption, lldsorption, or a membrane recovery
process (A reactor system could also be considered.)
7. If the light ends conlaminate the they must be removed. Our options
are to drop the pressure of the feed slream and flash otT the light ends, to
remove' the light ends by usmg a partial t:ondenser on the produci column, to
remove the hght ends in II: pasleun7.31101l seclion in lhe product column. or to
use a stabthur column to remove the htdlt ends.
8. We recycle components that form a;reotropes with the reactants If the
azcotropic compoSition is not too hIgh. but there is no heuristK: avaIlable to sct
the euct level.
9. We norrnallydo nol use d1stiUatlon to spht adjacent componenls when 2 < I I_
10_ Instead of usmg heuristics to sdeet column sequences, we usually calculate the
costs of all the sequences.
II. If dastlllation 15 100 expenSIve, .....e consider lUZOtrOpK: distillation. extradiYc
dIstillation. reactive distIllatIOn, extractl(ln. or cryslalliution.
Exercises
7.6-1. For one of the design problems thaI you hllye considcred. determine the: follOWIng
(a) Thc general slrudure of the sepllration system.
(b) Whether a V;lpor rCCQvery system IS required and, ifso, wherc it should be IQCaled
If ncccs.sary, detcnnine lhc dCSlgn of Olle of lbe: ahunativcs
(e) Several allernalive one of these: (Cautloll 1 he II'A al1d
ethanol proccsSQ are nOl Ideal and aellvlly coeflicrenl moods and lhe use
ofa CAD progr-am.)
7.6-2. Sletch your best guess of a separation s>,,-tem for one of tbe: procrsscs below (I_e.
the general structure of the q:p:uahon system); gUCS1i whether II vafl'Or .ce<n-cry
s)"stem mIght be needed, It sholiid be placed, and whal type m1ghl be the best.
and I!UCS5 the dlSllllatlon seqllencing a1ternau\es lhal might be the best Describe in
as much detail as you an the reasons for ,our suesscs. and mdlcate in delall what
cakulatlons you nc:c:d 10 do to \"enf} yOllr SlJCSSd.
(II) The: cyclohcunc pcocess (SCl! Exercises 547 and 6.8-6)
(b) 1lJc: butane alkylatIon plocess (sec ExcrClSCS 5.4 10 and 68-9)
(e) The: st)unc prOCtS'l' (5 FlefctSC5 54 6 and fo85)
(01) The acetIC anh)'dnt.le process (sec Exercises 43 and 611-2)
(I!') The acid process (Itt Exercise I 3-4)
7.6-.1. If the BDA process "'Ith dlpbenyl ra-overed ... run at liny hl!h conycamnlo. "'e
mIght obtam a 50/50 mlAture of loluene and d1phcnylthat ..ould be fed to thc recycle
columll. If we select an oyefhc:ad Q)mposilJOn of tolucne as X
D
- 0.9 and we rCCQyer
99% of the tolucne overhead, how many tra)"$ are rcquTred m the diSli1bLion column
(ilS5ume II _ 25)"1
7.6-04. Suppose th,lII the now laIC to the dlSullatmn tram In a butane alkylation prf.)CCU (sec
fuercises 54-10 and 6.89) when x = 0.9. T _ 4O"F. mol i-C./mol D.c. al reactor
inlel _ 9 IS given by C
J
_ 310molfhr. i-C. _ 11892 mol,lhr, o-c. - 143 mol,lhr.
II-C. '"" 419 mol/hr, i.C, _ 918 moljhr. and (", 1 _ 184 mol/hl (..here \\-c do nol split
j-C. from DC.) Usc: heuriStiCS to suggest ahernauve sequenccs of dlsllliallon 10
consider Should sldeslream columns be considered
'
Suppose that in Ell.Crcise 7.6_4 lhere is no" C. in Ihe feed Calculalelhe vapol rates
reqUIred in each column m a scqllcncc where we .emove the hghtesl componenr firsl
Compare IhlS result to a case where ....e rCCQver the C
J
firsl, flash lhe bottoms stream
from rhe C
J
splltler. and send the flash liqUid 10 a distIllation train we lecover
the hghlesl component firsL Assume that thc presSllre or the dep.opanizer is 230 psla
and thai lhe pressure of the debulanlzcr IS 96 psla
7.k. Corn.de. a process Ihat produces 100 mollhr or x}lene and 100 molthr of by
,olue:nt d,sl"'oporlJonauon

lbe reacuon 1$ .aculaJly equ1hbnum-limlted. But, nc&JeCIins tbis IImlla
tion. find the amounl of toluene in lhe fccd to. distlllallon train where the dIrect and
the indlrCCl sequences .. ould have the same cost Would you upccl that a complex
column ""QUid ever be less es.pensive?
7.6-7. A rCSKIuc curve map ror m'XlUIn of acelone. isopropanol, and ..aler lS given In FI&-
7_6-1. For tbe: cond1uons given in Example: 6.]-4, csumate the composition at the
bouom of lhe first lo... el If the dIrect sequenoe is u$Cd and attbe top of the first lO... cr
if the indirect scqllenoe is used
7.6-8. A model for a Simple plant is given m detatl m Sec. 10.] fOI the case where a direct
column sequeoce is used If we ocglcct the optimizallon or the renllx ratio and the
fractional rccoyery in the second tower and if "..e usc the indirttt ralher than the d1rcct
column sequenoe, what Irc lhe opumllm desIgn condluons? How do the costs for the
IWO aiternauves comrarc1 Compare the reactor eJlit compositiol1s at the optImum
of cach alternative At these values of the reactor uit compositions. how
do the Sland-alone dlrecl and mdncct sequences comparc?
7.6-9. The: lC3cllon (sec I.Jlerclscs 5.49 and 6.88)
Butadiene + SOl Rlitadiensulfont
SEC1lON H
Nomenclalure
has a slgmficant reverse rcaC!lon ratC' at IlK boiling POint or the product; $0 we: do not
want 10 usc dlSlillauOD 10 recovc:r aod recycle lbe reactants. Suggest another
scparauon system (not lDcluded 10 our set oIruk5) (or Ihis proau. PIOIlhc:
economte potl:nl1al in terms of tbe Slgwficanl deslgo vanabks.
Bottoms
Distillate
Feed plOch
Fd
Fresh feed
Flash hquld
Flash vapor
Gas recycle
liquid recycle
Plate number
Purge
Product coJumn
Product column bottoms
Product column dislillate
ReaelOr IOlet
Reactor eXH
Stabilizer bottoms
Stablhzer distillate
Strippmg plOch
Tolueoe leaVlog the product column
Toluene Ieaying the recycle column
Toluene leaying tbe stabilizer
RclatJ",e volatIlity of component j with respect 10 component j. a,j =
KdKJ
Rool of Underwood's equ.ilIon
AClivlty coefficIC:nt
Vapor mole fraction
Feed mole fraction
m
PG
PR
PR.B
PR,D
R,m
R,out
ST,B
ST,D
,
TOL.PR
TOL,RL
TOL,ST

F
FF
FL
FV
GR
LR
y,
""""'"
8
D
'.

,
nGURE 1.,""1
rcsldue o;urves
[FrtHIt S Lnly, D B _ [)o"gm.
-" AI F lNIk.ty, IHe FwttIiJ
_'tlls.lA 46J(l98n ..."llpBmu-
_ F-. Amno=n
ScJ<,,,,'y.]
IPA/HZO azco!rope
Benzene: molar flow
Ease-of-scperalioD mdo
Feed rale (molfhr)
Fractional recovery of compooenl i
Component flows of hght and heavy materials (moljhr)
Column height
Distribution coefficient
Liquid flow rale (mol/hr)
Component flows of hquld (molthr)
Reflux ratIo
Reboil ratio
Toluene molar flow (molfbr)
Vapor rate (molfhr)
Component flows of vapor (molfhr)
Conversion
liqUId mole fraction
Mole fraction of component j ID distillate
Mole fraction of component j in foed
0.2
08

Water XI Isopropanol
Acetone
10
06
X,
04
x.
x
x.,
8Z
ESI
F
I.
h.t,
h
K.
L
I,. 'j
,
,
TOL
V
CHAPTER
8
HEAT-EXCHANGER
NETWORKS
Energy conservation has a1....,3)'5 been important in process design.. Thus. it was
common practice to install feed-effiuent eJ.changers around reactors and distilla-
tion columns. However, a dramatICally different approach that lakes into coDsider-
allan energy iDlegratioD of tbe total process bas !xeD developed over the past two
decades. The: basic ideas of thiS new approach are presented now.
8.1 MINIMUM HEATING AND COOLING
REQUIREMENTS
The: starting point for an energy integration analysis is tbe calculation of the
minimum heating and cooling requirements for a heat-exchanger network. These
cakulatioDS can performed without having to specify any beat-exchaoger
network. Similarly, v.:e can calculate tbe minimum number of exchangers required
to obtain the minimum energy requirements without baving to specify a oel...ork.
Then lhe minimum energy requirements and the minimum number of exchangers
targets for the subsequenl design of a heat-exchaoger network.
In any process f1owsheet. a number of streams must be heated. and other
streams must be cooled For example. in the HDA process in Fig. 8.1-1. we must
heat the toluene fresh feed, the makeup hydrogen. the recycle toluene, and the
recycle gas stream up to tbe reaction temperature of 115O"F. Also, we must cool the
reactor effluent stream to the cooling-water temperature 10 accomplish a phase
split. and we must cool the product stream from its boilillg point to cooling-water
216
"
"

"
1-
Ji
u
u

c-"

M
U I'. .e-
0
3 u
"
:rl
" "
{d

M
..
u



.

E:J:!

217
218 SEn ION II WINIMUM llEAlli'lG "'!'IO '::00111'0(. IlEQUIIlfWENTS
(ABU: 8.tt
Fin;t-Ia", c.lculation
".
SECTION 11 HUll"'..... Nl) COOUtolG l.lQUIRMlNTS 219
NO
I:IIJ 140
Sou_ Qaa...

e-diu- H.
r
, Ik./(..... of. f. f_ Ie' HI_JIor
,
'000 'lO '20
Il.
1 .... 100
"" 1 C"'"
JOOO
'"
".
'10

C"'"
6000
'lO
''''
-360
-10
'00
,'"
'"
FIGURE 8.12
ShIfted $ClIla.
temperatures because we do not .....anl to store materials at their boiling points. We
also have heating and cooling loads on the dislillation-column condensers and
reboilers.
FirstLaw Analysis
Suppose: we consider a very simple problem where .....e have IWO streams that noed
10 be heated and 1.....0 streams thaI oocd to be cooled (sec the dala In Tabie 8.1-1). If
we simply calculate the heat available 10 the hOi streams and the heat required for
the cold streams, Ihe difference bct... ce-n these tlkO values is the nel amount of beat
thai we would have to remove or supply to satisfy tbe first law. These results are
also shown in Table 8.1-1. and the first twO entnes are determined as follows:
Flower: and we describe their analyses. If we choose a minimum driving force of
10F bet.... ocn the hot and cold slreams, we can establish two temperature scales on
a graph, one for the hot streams and the olher for the cold sueams..... htCh are
shtfted by 1000f Then .... e plol the stream data on thIS graph (Fig. 8.1.2). eu .....e
establish a sencs of lemperature Intervals that correspond to the heads and the tails
of the arro....'s on this graph. Ie, the inlet and outlel lemperatures of the hot and
cold slreams gIven In Table 8 I-I (sec Fig. 8.1-3).
In each temperature IOlerval Ike can transfer heat from the hot streams to the
cold Slreams because \loC are guarantocd Ihat the temperature dnvlOg force IS
adequate. Of course, \loe can also transfer heat flOm any of the hot streams 10 the
hlghtemperature intenals to allY of the cold streams at lo.....ertemperature
QI FIC,l = [1000 Btu/(hroF)](250 - 120)
- 130 x 10
3
Btu!br
Ql - F
1
C,16T
1
= (4000)(200 - 1(0)
_ 400 x J03 Blulbr
(8.'-1)
(8.1-2)
E. C Hohmanll. Optlmum Nel... orkl ror Ile.l !'h.D T1Iem. Ulll'enny of Southern
Caltfom'a, (1911), T UnKda, J IIOh. and K ShlfOko, CAm! Eng !"flg.14 (9)' 10(1918); 8 L.nnhoff
aOO J R. AlOE 24. 6)), 642 (1978).
Thus, 10 x JO} Btulbr mUSt be supplied from utilities ulhere are no restrictions on
temperalure-driving forces.
This first-law calculalJon does not consIder the fact thai we can transfer heal
from a bOI slream toa cold stream only ihhe temperature of the hot stream exceeds
Ihal of the cold slream. Hence, to obtam a physicaUy realizable estimate of tbe
required heating and cooling duties. a posItive temperalure driving force must exist
between Ihe hal and cold streams. In other words, any heat-exchanger networl
thaI we develop must satisfy the second law as well as the first law.
Temperalure Illlenlllls
A very simple way of Incorporating second-law considerations into Ihe energy
inlegration analysis was presented b)' I lohmann, Umcda el ai, and Lionhoff and
2j()- 240
200-
,'"
,..
''''
100- 90
FIGURE 1.1)
IJlICO'alli
220 SEeriON I I MINIMUM lU!AIlNO AND I"OOt.lNO UQl1I....MENTS
srerloN II MINIMUr-t "FAlIN(; A"U coou,.oo UOUII.fMFf'lrS 221
intervals Ilo..... e\'cr. as a starting point we consider the heat transfer In each Inlerval
The expression we usc is
""
50
The other values arc shown in Fig. 8 1-4 We also nOle Ihat Ihe summation of Ihe
heal available in all the intervals (50 - 40 - 80 + 40 + 20 = -10) is -10 x
IO
J
Btuthr. which IS identical 10 the resull oblained for the first law calcul:uion i e
Ihe nel dlfferencc bel .....een the heal available in the hoi Slreams and Ihat in the
Slreams.
Cascade Diagrams
One way wc could satisfy the net heating and cooling requirements 10 each
temperature interval is simply to transfer any excess heat to a cold ulil,ty and to
supply any heat required from a hot utility (see FIg. 8.1-5). From this figure. we see
thaI we would need to supply 120 x 10) Btu/hr (40 + 80) and that 'o\'e would have
to reject 110 x 10) Btu/hr (50 + 40 + 20). Again, the differena: is the firsl-law
value.
Of course. Ihe arrangemenl shown In rig 81-5 ..... ould correspond to \'ery
poor engineering praclia: because we are Iransferring heal from Ihe hlghesl
possible temperature interval dnecily to a cold ullllly. rather Ihan using Ihis
available heal 10 supply some of the energ}' requirements al lower-tempera lure
inlervals. Thus. instead of using lhe arrangemenl shown in Fig. 8.1-5. we lake all
Ihe heat availab1C' at the highe5t temperalure inlen'al (200 10 and wc
transfer it to the nut lo.....er interval (160 to 200' F) (sec Fig 8,1-6). Since we are
transferring this heat to lower-temperature intervals. .....e always the sct:ond-
law constraint.
From Fig. 8 1-6 we s that there is sufftcicnt heat available 10 the
temperature intcnal to completely salisfy lhe defiCiency in the second interval
(40 x IO
J
Btujhr) and to also supply 10 x 10) of the 80 x IO
J
lequirement for the
third intenaJ. 110'o\c\ler. in this third inten'3\ wc must supply 70 ,. 10) Btu/hr from
''''
'"
,..

,
-40

..
X
80 u
40
20
200
""
""
FIGURE I.I-S
Hal lrln'lfer In Ind I'tom ra. ach ..r.l1urr ,nl..rvaL
(8 1-3)
(81-4)
(8,1-5)
(8.1-6)
QI "" (HXXJX250 - 200) ""50 X 10"
Ql",,(IOOO+4000 60(0)(200-160)= -40x 10'
Q" - (1000 + 4000 - JOOO 6000)(160 - 140)
80 x 10J
for each mlerval. Thus. for the firsl lhree lnlervals we obtain

50

''''
. ,

0:
70
80
10
0
u
,..
0
40
40
20
60
'"
'00
,..
200
,,.,
20
50
-4{)
-80
40
Tolal "'" -10
6000 3000
4000
lSO- 240
200-
''''
,,.,
,..
100_ 90
Fe, 1000

Nee cnc,&y leqUIre<! ., each mltfVal
FIGURE I.t-6
CaSQIdc
Mt",rWuw HfAl1NO .. "'0 COOLIIolG kEQUIU.... UO'5 223
a hoi utjll1y .....e ha\e used all the: heat autldblc at hlgheHempcrature
Inlen'als. Then lhere would be no transfer DC heat bct.... ecn the lhlld and Courth
tempaature Intervals.
For the Courth lemperature Inlenal, we could eilher reJeCt the excess heat to
cold UUIHy, as shown In Fig. 81-5, or Iransrrr Jl 10 the neu lower lernpenuure
interval, as shown m l'lg 8 1-6 1 hen, Cor the lowest lelllpl:rallJre Inlerval .....e reject
alllhc remaining heat to a cold ullht}
We I ig 81-6 a ... dloJgrunt II sho.>..... s how heal 1;,l5(;3des
through lhe temperature Intervals
T_lOO
r .. 120
r .. 140
roo 160
T_1OO
r .. 250
". _ 0
II, .. .aoo(120 - 100) - 10.000
If) .. (lCOO f- 4000)(140 - (20) .. 100,COO
11,,, (100) ... 4000)(160 - 120) .. 100.COO
II. -(100) -t 4OJO)(200 - t60) _ 200.000
If. - 200) -
o
10.""
180,000
280,000
480,000
l30,OO)
600
200
400
Enthalpy, 1000 Btu/hr
a_HOI +-Cold
flGUaE 1.1-1 Fi 8.1 6- ;d:kdallhc
TC'mperalllrc-<nthal"" dQ&ram (ThIS figllrC' '" d ...""" 10 tnc heal In '& - '"
.. turc intcI'-aI)
2'"
240
230
220
210
200
..
.90
0
Ii .80
a
170

8- .60
E
.'"
'"
140
130
.20
110
.00
90
0
Approach lemperature ::: 10
No..........e plot the cumulative H versus T(see fIg. 8.1-7). We call thiS a compoSite
curvt' Cor the hOi streams because it includes the effect oC all the bot streams.
or course, since: the FC. values are constant, we could have replaced the
calculations Cor H HJ' and H. by a sing.le expression
H
1

J
= (1000 + 4OOlX200 - 120) = 40,000
Thus. we only need 10 calculate values .111 the temperature levels ""hen lhe number
oC bot streams changes.
AI the lowest temperature oC any oC the cold streams (9O"F on Fig. 8.1-4), we
choose the enthalpy as Ihe minimum cooling requirement Q<.i.(60 X 10
1
Blufhr
Pinch Temper.aeure
We also nole Crom Fig. 8.1-6lhat there is no energy transCer bet .....een the third and
Courth lemperature intervals. We calilhis the pinch (l40F Cor the hOI
streams and 130F Cor the cokt streams, or sometImes "'e use the average value oC
!3y'F), Thus, the pmch lemperature provides a decomposition oC lhe design
problem Thai IS, above the pinch lemperature we only supply heal, whereas belo\lo'
the plllch tempenllure we only reJecl heal to a cold ullhly.
Dependence on ebe Minimum Approllch
Temperaeure
If .....e change the minimum approach temperature oC IOOF that we used as our
second-law criterion, then we shirt Ihe temperature scales in Fig 8.1-2. The heal
loads in each orlhe inlervals shown in Fig. 8.1-4 will also change, and lhe minimum
heating and cooling loads will alter. It is easy 10 \'isualize these changes iC we
construct a temperalure-enthalpy diagram.
Minimum Utility Loud...
From Fig. 81-6 .....e see lhatlhe minimum healing requirement IS 70 x 10
1
Dtujhr
and the minimum cooling requlremeOi is 60 x 10] Btulhr. The difference bel .... ecn
lhese values stili corresponds 10 the firsllaw rquirement. but DOW our minimUm
heaung and cooling loads have been fiAed also to sausC} the second law
Temperature-Enthalpy Diagrams
To construcl a temperature-enthalpy diagram, first we calculate Ihe mmimum
healing and coolmg loads, using the proceduredescribcd above. Then we define the
enthalpy corresponding to the coldest lemperature oC any hot stream as our base
condillon; i,e., at T = 100 F (sec Fig. 8.1-4), H _ 0 Nexi we calculate the
cumulative heal available in the sum of allihe hot streams as we move to hlgher-
temperalure inlenals. Thus, Crom Fig. 81-4 we obtain the Collowmg
224 SR;TION II MINIMUM AND CODUI\IG kEOlJlkOolENT5
nOION I'
600
Dc:lt ., 20
200 400
Enthalpy. 1000 Btu/hr
+ Cold Delt = 10 <> = Cold o RO Hot
Approach temperature = 20

240
230
220
210
200
190
180
170
160
150
140
130
120
liD
100

o
CUT""e bela.... the pinch (see Fig. 8.19). This grand composite cun'e dearly shows
thai minimum heating requirements arc QH = 70 X IO
J
Btu,lhr and that the
minimum cooling load is Qc = 6O)C 10
3
Btu/hr. The grand composite curve is
particularly useful for profile matching during heat and po.....er integration studies.
nGURE 8.1-8
Tcmperalurc-.o:ntllalp)' dIagram.
",000
110.000
"',000
""000
11
0
.60.000
1/, _ )O(l(l( IJO 90) _ 120,000
II J - ()O()lJ + 6OOO)(ISO - UO) _ uo.ooo
". _ 6lXXl(190 tSOI 240.M:>
T_90
T_ no
T_ tSO
T_ 190
These results also are pIoued on Fig. 8.1-7.
From Fig. 8.1-7 we Dote that the enthalpy of the hot streams tbat must be
rejected 10 a cold utility is Qc "" 60 X 10] Btufhr, and the arnount or heat that must
be supplied from a hot utility is QB _ 70 X 10] Btu,lhr. Moreover, wben TN _ 140
0
and T
c
=- 130", we see that the minimum approach temperature exists, i.e., the
heating and cooling curves are closest together. Thus, this temperature-enthalpy
diagram gives us eUelly the same information as previously.
Suppose now that set the base entbalpy ortbe cold curve equal to 110.000.
instead of 60.000, and repeat the calculations for the cold curve. This shifts the
composite cold curve 10 the rigbt (see Fig. 8.1-8). We nOle from the figure that the
beat we must supply from a hot utility increases by SO x 10
3
to 120 x 10
3
Btu,lhr.
Thus, the increase in the heating load is exactly equal to the increase in the cooling
load. Also at the point of closest approach between the curves (T..... = 150F and
T
cekl
= nOF) the temperature difference is 20"F. Thus, if the minimum approach
temperature had been Specified as 2O"F. tben the minimum heating and cooling
requirements would have been 120 x lOJ and 110 )C IO
J
Btufhr. respectively. and
the pincb temperature .....ould change from T. = 140 and T_ = 130 to T. = 150
and T..w = 13O"F. By sliding the curve for the cold streams to the right. we can
chan8e the minimum approach temperature, Q"._ and <2<-....
on Fig.. 8.1-6). Then we calculate the cumulative enthalpy 10 each temperature
interval '
Grand Composite Cune
Another useful diagram is called the grand compositt cvrw. To prepare this
diagram, we start at tbe pincb condition shown in Fig. 8.1-6. and we say thallhe
beat 80w is zero at the a\erage of tbe hot and cold pinch temperatures T- 135.
Now at the next higher temperature interval, which we again define by the average
T_ 155, we calculate tbat the net heat 80w is ISO - 100 - 80. Similarly, at
T= 195 we find that H _ 80 + 240- 200 = 120, and at T ... 245 we get 11 =
120 - 50 - 70. These points are just the differences between the composite curves
shown on Fig. 81-7. calculated with the pinch as a starting point. We call lhe
results the grand composite curve above the pinch temperature (see Fig. 8.1-9).
Again, starting at the pinCh and moving to colder temperatures, at T _ 115
we let H = 40, and at T a:: 95 we let 11 -= 20 + 40 = 60 These points define the
Relalionship or Minimum Heating and Cooling to
tbe First-La.. Requirement
The first-law analysis indicates tbat tbe difference bet.....oco the heat availabk In the
hot streams and that required by the: cold streams is 10 x IOJ BtuJhr, which must
be remo\'ed to a cold utility_ The second-law analysis with a approach
temperature indicates tbat we must supply a minimum of 70 x IO
J
Blufhr and
remove 60 x !OJ Btufhr. Hence, we sec that any incremental heal that we put in
from a hot utIlity must also be removed by a cold utihty. Moreover, we recognize
that if we put in more than the minimum amount of energy (sec: Fig g, \10). then
we will ba\e to pay more than necessary for both a hoI utllit), and a cold utility
(because 'Ae ""ill have to remO\'e this excess heat)
226 ncrloN II
SlClION II tlNlMU/ol HOllNG ANO COOLl .... UQUllHIENU 227
(8.1-7)
(81-8)
(8.1-9)
OJ
,OJ
""
""
f'
50

5
Q
5
40
.2

70 JO + QE.
..
''''
80
u
40
Q
40 + Qt:
'00
20
60 + QE
flCUR 8.1-11
UotSl healln, and coohnz,
Do Dot transfcr heat across Ihc plOch!
Other rules of thumb that we have developed arc Ihcsc
Add heat only abm'c the pInch
Cool only below Ihe pinch
From Fig. 8.1-11 we see Ihat ifwt: Iransfer an amounl of heat Ql across
pinch. we must put this additional heal iDIO the process from a hot utilit)
somewhere in the network. Furthermore, we must also reJCcl thIS amoun! of heal to
a cold utihtJ. Hence, we oblalO a rule of thumb:
120 100 80 60
EnlhaJpy
40
250T
240
230
220
210
200
190
ISO
170
160
15k_--
140
130
120
110
100

o 20
nr.URE &.I.'
Grand compolll( cun'(
Minimum heat in "" 70
-10
Heal in "" 70 + Q
Net=-IO
Industrial Experience
The calculatlon of the minimum heatlOg and cooling requirements is a very simple
lask. and yet it indicates that significanl energy savings arc possible compared 10
past practlce. In particular, Imperial Chemical Industries ID the United Kingdom
and Union Carbide in the Unlled Slales have both reported the resullS of
numerous case studies Ihal indicale Ihal 30 to 5OY, energy savings, compared to
conventlonal practice. arc posSIble even in relrofil situatIons. Hence. thiS coer&}
IOteg.rallon design procedure is a very valuable tool
MlDimum heat OUt "'"' 60
Heal Out '= 60 + QE
If we put excess heat into the process,
we must remove this excess heal.
Multiple Ulililies
In tbe prevIous analysis, we considered thc case of a slOglc hOI utility and a single
cold utihty. However, the analysis is also valid for multiple utilities. If we shift the
nCUkE ..... ,
Rdat>Olllhlp 10 fint law
D Boland IDdt:: Illlldffilllh, M Hal u.chanlC'r (/0,,,. E.NJ 11Il(1): 47
(19&4), H uMhotf and 0 k V.edcveld, P,nch TochPook>f,y COIUQ or Age.M C/o,m. IOfI 110(7).
H (198-4)
228 5f:CT10J'i'I MlJ'iINUM IlUTING ANI) coot.lNG "EQUIIlf.MErO"S
MINIMl'W !lUTING AND C'OOlI'<G HOUlHMf'N'T5 229
".
(8.1-13)
(8.1-14)
Enlhalpy.
0= HOI
Tbe values of all streams arc known.
Inlet and outlet temperatures of all streams arc known.
FIGURE 8.I-lJ
T-H dJi!QII\-mulllrle utilities
Approach temperature"'" 10
360r--------=.:=====-'-=----,.----,
350
340
330
320
310
..300
290
e 280 ,
e 270
260
250
240
230
220
210
200
1000 Btu/hr
+ - Cold
For the case of ml.ltlUres, where a plot of enthalpy versus temperature is
curved, ....e merely Imearize the graph and select fictitious values Ihat have the
same heat dUly (soc Fig. 8.114). Thus, phasechaogtSsimply increase the number of
intervals considered.
Limitations of tM Procedure
The cakulation of the minimum heating and cooling loads requires the following:
However, the design variables that fix the process flows (i.e.. conversion, purge
composition. molar ratio of reactants. etc.) must be detennined from an optimi7.a-
tion analysis. For each variable. the optimi7.3tion involves recycle costs which
depend on the heat-exchanger network. Thus, the optimum process flows depend
on the hcal-cxchanger network, but we must know the flows to determine the
network. We resolve thIS dilemma by calculating networks as a function of the
Rows to estimate the optImum design conditions
(81-12)
",.
".
50
E

80
10
40
40
20
20
(Q\']
)(J()

5
'"
70
200
FIGURE 1.1.12
Multi" uulillei.
temperature range in our previous example upward by IIO"F, we obain the cascade
diagram shown in Fig. 8.112. Now we see that as a hot utility we need to use Slcam
having a temperallHc in cxcess of270
Q
F Also, wccould use steam at as onc
cold utility and coohng water as a second cold utility. With this procedurc .... e
would reject 40 )( 10-' Btu/hr to the steam and 20 x 10-' Btu/hr to COOling watc!.
Note that lhcre is no heat transfer between the bottom two temperature
intervals whcn v.e use multiple utili lies. Thus, we introduce another pinch, which
we call a urifil)' prnch, iDto the network. An additional utility pinch is added for
each new utility consKtere<t The effcct of multiple utilities on a T-If dIagram IS
shown on Fig. 8.1-13
Also recogmze that there arc some obVIOUS heuristics associated with the use
of multIple ullhties'
Always add heat at lhe lowest possible temperature level
to the process pinCh. (81
Always remo,,"C heat at the highest possible temperature level
relative to thc process pinch. (8.1-11)
= F
where F, and are the fictitiOUS values.
Phase Changes
The procedure requires lhat the values of the streams be constants. We can
incorporate phase changes that take place at constant temperature into this
formahsm simply by assuming a IGF temperature change at the tcmperature of the
phase change and then calculating A fictitIOUS Fe, value thai gives the same heat
duty as the phase change: i.e., if the heat corresponding to the phase change is
F /il/G' we write
SEcnufol IJ IolINI ...I'''' Nu... IU or EXClV.'-C;EU 231
Vilpor
FC,. vapor
Fe,. (I) = AUF
Vapor
FC
p
vilpor

HOI
ulilllY
70
70 110
Stream I
130
20 340
Sucam 2
400
60
FCp liquid
Smks
Stream 3
180
Stream 4
J60
Cold
utility
60
nGURE "'-14
Pba... chan.,.,.
SimIlarly, if a heat exchanger IS used to preheat a stream leaving a flash drum
or a gas absorber and entermg a distillation column, Ihen we must include IhlS
stream In our analYSIS- Moreover, if a process Slream IS used 10 dn\le lhe: reboiler of
a dislillallon column, than steam, then lhe optimum reflux ratio in that
column will change Thus, Ihe energy imegutlon analysis is coupled with the tOlal
design prOblem, and oflen some iteratIve case studIes arc reqUired
(a) Heat loads balance eJtaclly: results of first-law analysis
(
Number Of) (Number of) (Number Of)_ I
(b) EKchangers = Streams + Utilities
nGuJU: U-t
1&". mmomum number or cubanr;cl"fo.
loads salisfy the first-law requirement. We can generalize the result and state that
nonnally
8.2 MI [IMUM NUMBER OF
EXCHANGERS
(
Number Of) _ (Number 01\ + Of) _ I
Exchangers - Streams) Utilities
(82-1)
Our previous anal)'sis allowed us to delennlne the ffilnlnlUffi heallng and cooling
requiremenLS for a heal-txchangcr nellll'ork. We use these results as a starlmg POlOt
to delennme the minimum number of heat exchangers required, The analysis
follows the procedures describro in earlier references in Ihis chapter
First-Law Analysis
Suppose we consider Ihe heatmg and cooling loads for each of Ihe process slreams
as well as Ihe minimum utility requiremenls Ihat correspond to the second-law
analysis (.see Fig. 8.2-1). Now we ignore tbe ffilOimum approach temperature and
jusl consider how many paths (heal c;lchangers) arc reqUired 10 Iransfer lhe heat
from Ihe sources to lhe sinks. If we transfer 70 x 10
1
Btu/hr from Ihe hal utilit)
into stream J. we still have a deficJCncy of 110 x 10
1
Btujhr m stream 3. If .... e
supply Ihis deficiency from stream I, we still have 20 x 10
1
Btu/hr of heat available
in stream I. If we transfer IhlS excess to stream 4, we are left wilh a defiCIency of
340 x 10
3
Btu/hr in stream 4
The other calculations are shown In Fig 81-1, and we find Ihal Ihere arc five
palhs, or that five heal exchangers arc requlfed We nOle Ihal Ihe heal loads jusl
balance. which musl always be the case because our minimum heatmg and cooling
Independenl Problems
Equation 8.2-1 is not always correct, as we see by Fig. 8 2-2. In_thiS
uample we have merely increased the utility requirements. but the first law
analysis 'is stili satisfied. lf we t111ns(er lbe heal belween Ihc. sources and lbe sm,b as
shown m Fig. 8.2-2., then we require only four exchangers Instead Ho.... e\"er.
Ill'e could also redraw the figure so that Ihere are two, compktdy mdependent
pcoblom,. 10;, ;,,1'0' g,w.1 ,,,Ull, .od mo<c ,;go,o", 8.21 "
Number of Number a + Number 0 _ IndependeDt (82-2)
(ExChangers) a; ( Streams) (Utilities) Problems
Loop'
If we return to our original example and consider Ihe arrangement shown in Fig.
8.2-3.... e see thaI we can still sUllsfy the heal-transfer requirements lhe
sources and the sinks for any value ofQ[_ Ho.....ever, for IhlS configurauon we need
SIX exchangers Also there IS a loop in lbe network (i_e., we can trace a palh through
lhe ncl... orllhat starts at the hot utilily. goes to stream 3, goes 10 stream I. goes to
232 1.1 J-IINIJ-IUJ-I Of &llC1INGU.lI
nCTlO1'I11 11.[11 ESllM1ITll 233

"0<
utility
230
Stream
130
Stream 2
400
"" ""
""
No Exchangers '" 4 + I - I

230 130
180
220
PillCh IXI 1.1()
,,,,,-+__-+__ __
stream 4, and goes bade to the hOi ullhty) Any time we can trace a path thai
stans at one pOInt and returns to that same point. we say that we have a loop in thC'
nC'twork Each loop Introduces an eJOtra CJOchanger into tbe nc"'l.ork
Effect of Pinch: Secondu"
As part of our calculatIon of the mmlmum heating and cooling rcquirc:ments \loe
found that thcre was a plOch lemperature that decomposed the problC'm inlo
(a) Heat loads balance exactly; results of analysis
(b) (NUrnbcr Of)_(NUmbcr Of)+(Numbcr Of)_(Numbcr Of)
Streams Utlilhes Problems
FlGLRF U-1
problelTt$
(8.2-4)
(8.2-3)
'"
100
_4+1-1=4
Eklow the pinch lemperalure the:re are only three streams. and so
Eklow pinch: N[ - N
s
+ N(I - I
_3+1_1=3
83 AREA ESTIMATES
Thus, to satisfy the minimum heating and cooling rcquiremelJ,lS requires a
total of $Cven exchangers. However, to satisfy the firsllaw requires only five. Then,
we expect thai the network for the minimum C'nergy rcquiremenls will have two
loops tbat cross the pinch (we introduce an additional cJOchangcr for eacb loop). If
we are willing to sacri6cc some energy by transferring heat across the pinch, we can
eliminate up to two CJOchangers (rom the network. Hence, there is 8 capital-operat-
ing cosl trade-Q1f that must be evalualed.
(Number of Exchangers) '" (Number of Streams) + (Number of Utilittes) - I
::3+1-1::3
FIGURE 1..1-4
Elkc1 uI pondL
distmet parts. That is, above the pinch \\e onl) suPpll heat from a utilit). whereas
below the pinch we only remo\e heal to a ulility. Thus. to include- the second-Ia\\
aDalrsis in our calculation of the mmimum number of eJOchangers. we must appi)
Eq 82-1 (or 8.2-2) to thC' stfC3mS abo\-c and below the ploch. From Fig. 8.24 we
seC' that there arc four streams above the: pinch for our C'umpk. 50 that Eq 8.2-1
(assuming no loops and no independent problems) gi\CS
Above pinch' N( - N.lI + N
u
- I
Cold
utility
220
360
Stream 4
"0<
Stream I
utility
Stream 2
70
130
400
Q, 340
Q,
110 + Q
20-Q
60
Siream 3 Stream 4
Cold
180 360
utility
60
180
Stream J
Sinks
70 -
Sanks
(
Number Of)=(NUmber of)+(Number of) (Number Of)_(Number of)
E:u:hangers Streams Utilities + Loops Problems
FIGURE 8.1-3
Loo..
We have been able to estimate Ihe minimum heating and cooling requirements for
a process withoul even specifying a heat-ubanger network (see Sec. 8.1). We can
use these results to estimate the utility costs for a plant II would be very desirablC'
to CSlimatc Ihe capital costs associated wlIh a heat-C'xchanger network wilhout
234 SECI JoN I.J AIl.EA ESTIMATES
Sf:C1lON Il AlEA ES"fIMATES 235
2
2 - Number of cold streams
in this interval
(b) Match I and 4. 2 and 3
2
FIGURE 8.3-1
TempcralUrc-e-nthalpy dlag......m.
r
_ _r";;;;'---,
250 HOI
utIlity
230
220
210
200
190
180
,
::
E
ISO
140
130
\20
110 1
100 Cold

.L__-;c!;;;,-__.L__,1Yi,-_-'
900 200 400 600
Enthalpy, 1000 Btu/hr
o=Hot +=Cold
(0) Match I and 3, 2 and 4
u.
o
(8.3-2)
1 1 1
(8.3.1)
V hi h
o
where Ihe individual film coefficients include the fouling factors. The area of lhe
heat exchanger is given by
However, if there are multiple streams in any interval. then we must develop
an appropriate expression for the o\erall heat-transfer coefficient. Suppose we
consider the inlerval where IWO hot streams are matched against two cold streams.
If we matched streams I with] and 2 ","'ith 4 (see Fig. 8.1-]), our results would be as
shown in Fig. 8.]-2a. However, if we matched streams I wilh 4 and 2 with 3, Ihen
we would obtain the results shown in Fig. 8.3-2b. The heat loads and the log-mean
temperature driving forces for each of the exchangers will be the same. For the case
given in Fig. 8.3-2a, we find that
EslirDitling Areas
In Sec. 8 I we developed a lemperarure-enlhalpy plot (see Fig. 8.1-7). Suppose now
that we Include vertical lines whene\er there is a change in the slope (see Fig. 8.3-1),
and we considcr thai each interval represents one or more heat exchangers in
parallel. From the graph we can read Ihe heat duty for each exchanger and the
values of the temperature driving forces at each end. Then if Ihe heating and
cooling curves correspond 10 a single stream, we can estimate the individual heat-
transfer coefficient for each stream as well as the overaU coefficient:
having to design a nctworl,. Fortunately, a lechnique for making this estimate has
been prescnted by Townsend and Linnhoff (which is an extension of a previous
result by Ilohmann).'
D. w. Townseod and B. Lmoholr. Aru TUll"IS for Ileal E.chlngcr Ann.... 1
meetIng of ttw: Instll"Uon of Chemical Enginecl1o, Balb, United Kinldom, April 1984.
IE. C. Hohmann. NClworh fo< IlUI Ph.D Thes.... Unive",i,y of SoulMnr
ClIW"omi.. 1911
FIGURE ILJ-l
Overall heal-tllmJer coc:ffiocnls
(8.33)
(8.3-4)
1 I 1
-=-+-
V., h
z
h..
AT.=A.,+A.'=i\T. Q u + AT. Q U
LM., LM.,
Q (1 I I I)
= ATWol' h; + h
z
--t h
J
+ h..
1 I 1
-=-+-
U. , hI h}
so that the lotal area becomes
2J6 SECTIO!<.< DESIGN OF MINIMUM ENERGY MEA"! [XCHANGER N'lITWORX:S SECOON'" Df.5IGN OF MINIMUMENUGY HEArXOtANGEIl N ~ O R X S 237
For the configuration show" in Fig. 8.3-2b ....e find thaI
u"
1 1 1
-=-+
V. , h, h.
(8.35)
Stream
Fe,
1()()()
2
4()()()
,..
3
3()()()
4
6000
Then the total area IS
(8.3-6)
Q = 110 Q
,OI
= 110 + 240 - 60 ~ 360
= 70
which is identical to our previous result.
This result is general, so lhat we can write an expression for the area in any
interval as
200
"0
I ..----f---if----+---+---+--..L--
no
Q = 120
Q = 360
,..
ISO
Q ~ 160
Q = 240
I'
,
(83-7)
-and we can estimate the total area simply by adding the results for all the intervals.
Of course, this approximate procedure does not give the same results as those
obtained by designing a specific nelwork (normally lhere are too many ex-
changers). Nevertheless, Eq 8.3-7 does provide a reasonable estimate of the area
required. This shortcut procedure is particularly useful when we are attempting to
find the effect of the process flows on the capital cost of the heaH:xchanger
network. Once we have estimated the optimum flows, however, we need to
undertake a detailed design of a heat-cxchanger network.
8.4 DESIGN OF MINIMUM-ENERGY
HEAT-EXCHANGER NETWORKS
'00
90
Q
IOI
= 20 + t60 - 120
~
Now that we have obtained estimates of the minimum heating and cooling
requirements and an estimate for the minimum number of heat exchangers, we can
design the heal-cxchanger network. We consider the design in two parts: First we
design a network for above lhe pinCh and then another for below the pinch. We
expect that the combined network will have two loops lhat cross the pinch. This
analysis is taken from Linnhoff and Hindmarsh.
Design above tbe Pinch
As the first step in the design procedure, we calculate the heat loads between either
the inlet or the outlet temperature and the pinch temperature for each stream.
!
J
FlGUR.E &.4-t
Heal load fOI fHeaJIl1I.
Thus. for the first stream (see Fig. 8.4-1) we obtain
Above pinch; Q = Fe, d T = 1000(250 - 140) = 110 x IO
j
Below pinch: Q = 1000(140 - 120) = 20 x 10]
The results for the other streams are shown in Fig. 8.4-1.
Feasible Matches
(84-1)
(8.4-2)
If we attempt 10 match stream 1 above the pinch (Q" = 110) with stream J
(Qc = 60), it is apparent that the maximum amount of heat transfer that is possible
is the smaller of the two values (Q = 60). The approach temperature is just 10"F at
138 SEerION" DESIGN Of NINllolUIol ENUGY ltt..Ar UCItAI'lGU NOWOIIKS
uenON 14 DESIGN or NIN1MUIol [NU(,Y t1UTUCHAN(iU NIn'fltOUS 239
Ihe plOch, so we wanl to transfer the heat from the coldest end of Ihe hot
Then If .... e calculate the temperature of the hot stream Ihat would be lhe Inlel
temperaturc to thc cllchanger, we obl.un
Smce the outkt Ie:mperature: of the cold IS ISO F, the temperalure drivmg
force IS 5O"F and .... e ha\'e a feaslbk heat e:xchanger (:wx Fig 84-2).
Ilo.... e\er, .... c might attemplto malch SlTc<lm 2 .....lIh stream 1 Again from Fig.
8.4-2 wc sec thatlhe Ilmitmg heal load IS Qc = 60 Howe: ... er, when we calculate the
Inlet temperature of the hal slream, .... e: obtalO
".
4
6000
3
3000
2
4000
""
,..
100 190
Q :: 240
U
".
'"
Q 60
FCp 1000
,..
(8.4-3)
(8.4-4)
(8.4-5)
(8.4-6)
T" "" 200
Above the plDch: F
Il
C,1l FcC,c
Below the pinch: F
H
C,1l FcC,
Q = 60 X 10) = fC,tJ.T- 10000T
H
- 140)
Q = 60 X 10):: FC,tJ.T= 4000(T
H
- 140)
Smee the ell.lllemperature of the cold sircam IS ISO"F, we ha...e violated our cnteria
for Ihe minimum approach temperature.
A violation of this type will always occur above tbe pinch If(FC,>c > (FC,)H
Thai IS, the approach temperalUre IS Just lhe mmimum value al Ihc pmch, and the
.6Tbelwccn the two curves will always decrease if Fce,c > FHC,H' Thus, there IS a
design heun5uc for feasible matches at the plOch condition:
Stream
J 2 3
(a) PUI In the matches al the pinch
(b) MiUlmlzc the heal loads 10 c!lm\nale streams.
(c) See .... hat IS left.
1000
"" ".
3000
4000 3000
FIGURE ....)
P'IDCb DVlcbcs
Pinch M8tches
ISO to&O
1.. t40- no
FIGURE ""2
Matches the pinch
Q - 6O.00J = IOOO(T
H
-
T
H
- 200
Outlet T
c
= 150
Malch is feasible
From our feasibility critenon and Fig. 8.4-1, we see that above Ihe pinch .... e can
match slream I with eilher stream 3 or 4, and we can only malch strum 2 with
stream 4. Hence. we match stream I with Siream 3 and stream 2 with siream 4
Also, we transfer the muimum amount of heal possible for each match in an
anempt to ehminate Slreams from the problem These pinch matches are shown in
Fig. 8.4-3.
Nellt we consider the heat loads remammg. The cnteria given by Eqs. 8.4-5
aDd 8.4-6 are not applicable away from the pinch, and we know that above the
pinch we are allowed to add only heat. lienee, we must transrer all the heat
remaining in SlTeam I to stream 4, which is the only cold stream still available. The
heat remaimng m stream I is (110 - 60) )( 10J _ 50 X !OJ, and Ihe remaining
healing requirement of sIream 4 is (360 - 240) )( 10] _ 120 X IO
J
, so Ihal we can
IOsiaU Ihls heal cll.changer (see Fig. 8.4-4) The remaining heating requiremenl of
100 190
Q - 240
Q = 60
140) Q = 6O,00J = 400J(1i, - 140)
1i,:: 155
Oul1el T
c
= 150
Match is not feasible
Vlolales mmimum tJ.T
190 100
Q "" 110
'<COON" OEsl'Ol'l OF MINIMUN F.NF.JlOY HEATr.xC'KAJ<fGfll NF.'TWOUS 241
Q .., 110 - 60 = 50
I--T = 170
,J
4
( H 70
= 178
3
3000
,..
,.,
,,.
4000
2
Q - 240
r'\
1000 FC,
T= 200-
Siream
O--+---c,.,;;;---c,.. ;;;--{J'Q = 60
H = 70
4
3
3000
Q = 360 - 240 = 120
,..
,..
,.,
t ""--:I'"
2
'J__ Q "" 50
'000 FC,
Stream

,..----'----' L __--'- L __--'-__no
FIGURE U-4
Matches .....)' f,om tile pn""h
FIGURE 1.4.5
of 1M I'Indi
70 )( 10
3
which is just Ihe minimum heallO! requirement. is supplied from a hot
utility.
The complete design above the pinch is shown in Fig. 8.4.5. There are four
eJIchangers, which is the minimum required value, and we have satisfied tbe
minimum heating requirement. Thus, we have satisfied Ihe design targets. The
Slream temperatures are also shown on Fig. 8.45, and the temperature driving
force at the ends of every heat exchanger is 10"F or greater
Ahernath'es
For the example under consideration. the ptncb matches are unique. Howe\er. for
the other matches away from the pinch the utility heater can be placed either berore
or arter the heal e:lchanger connecting streams 1 and 4. Figure 8.4-6u shows the
location or the hcater arter the other heal cllcbanger has been inserted, but Fig 84..-
6b shows the result .....ith the last two neat e:lchangers interchanged. Calculalions
show thai both alternatives are feasIble. However, the driving forces for the beal
ellchanger are largest (a lo.....("r (".lI.changer area) when the utility heater is at the
highest temperalure. although beat must be supplied at a higher temperature level.
Thus. in some cases one alternative may ha"e a lower cost tban another.
Design below (he Pinch
We use ell3ctly the same design procedure below the pinch. For a feasible match .....e
require that FrIC,,, FeC;; (Eq. 8.4.6). Therefore, for our example. we can only
match stream 2 ,,;tb stream 3 (see Fig. 8.4.7). We put in this CIchanger and
maximize the load 10 eliminate a stream (rom the problem, Q "" 120 )( tO
l
(see Fig.
8.4.8). When we examine what is left, .....e ha"e only hot streams thai need to be
cooled We are only allowed (0 reject heat to a cold utility below the pinch. so we
install two coolers (sec Fig. 8.4.9).
The compldc." design for below the pinch is shown in Fig 8.4-10 We sec that
the total amount of heat rejected 10 cold utility is Qc = (20 + 40) )( 10
3
=
60 )( 10
3
, which is identical to the miDlmum cooling requirement The number of
242 ucnOlol
OUJON or ... ININUI.II!NU(lY H/!.Of UotANGU ",nwou:s
Slfeam 2 J 4
FC, 1000
6000 1000 6000
FC, 1000 4000 JOOO 6000
,>0 ,<0

,<0
,<0
I
,..
(
Q - 120 J.
-U
Q - 20
Q 120
Q 160
Q = 50
Q = 50
Q: 160 - 120 = 40
'00 90
200 '90
200
'90
t5
/
182

F1GUIl.E .....
Puw:b llIalcllQ
H = 70
Slfe.liRl.
T - 178
2 J 4
FC, 1000 4000 JOOO 6000
"0
,<0
,>0 ,<0 ,>0
'<0
,<0
C - 20
,..
(0) (b)
nCURE U-6
Dc$'i" ahcmau"<:S.
c= 409
'00 90
'-lGLRt,. a.....,
Add rookrs
SU<am 2 J 4
2 J 4
FC, 1000 4000 JOOO 6000
FC, 1000 4000 JOOO 6000
''''
,<0
''''
,<0
,<0
''''
,<0
C-20C(
Q - 120 J..
,..
Q = 20 Q = 120
-U
Q - 160
T-IIO___
C - 40
)
'00 90
'00 90
nCURE .....7
flGUR[ 8.4-10
Dalp bdow tIx p'ncll
ComJ*IC daJ.&1l lbc pInch
24J
244 SICTlON H O$IGN or I>llNlIo1UIol-E/'fElIGY HEAT EXctlANGElI NETWOllK5
u:cnON" DESIGN Of Io1INIMUM ENElIOY III'.ATEl[ctlANGU NElWOllKS 245
exchangers used is 3, which is the minimum number. Also, the temperature driving
force at each end or every exchanger is IO
Q
F or greater, so the design is feasible.
we have established one design alternative below the pinch.
Minimum F:nergy: Complcle Design
A complete design that satisfies the minimum energy requirements and the
minilnum number of uchangers above and below the pinch is shown in Fig. 8.4-11
The total heating load is 70 )< 10
3
Blu/hr, while the total cooling load is 60 x 10)
Blu/lH There are seven exchangers.
As we mentioned earlier, if we apply Eq. 8.2-1 to our example, we predici that
we need only five exchangers (although the minimum energy requirement cannot
be satisfied with less than seven exchangers). Therefore "e anticipated thai there
would be two loops Ihat crossed the pinch. (A loop is a path thai we can trace
through Ihe network which starts from some exchanger and eventually returns 10
Ihal same exchanger.) A loop may pass through a utility (see Fig. 8.2-3). Arter
examining Fig. 8.4-11. "'e find thallhere are three loops (see Fig. 8.4-12). Two or
Ihese loops passlhrough the cold uti lily, and we show later thai if we break ODe of
these loops, Ihe other will also be broken. Hence, there are two independent loops,
and it should be possible to remove IWo exchangers from the network sho"n ill
Stream
FC,
1000
2
4000
3
3000
4
6000
FC, 1000
2
4000
3
3000
4
6000
70
OJ"
""
,.,
()
Q = 50

''0
All
,
Q = 240
r'
,;0=
""
Q 60
Pinch
I
C

Q - 120
C
'''''
'''''
2"iO 240
""
Q = 50
('L
c "" 20
Pinch

Q 120

'00
'"
'00 9Il
FlGURE I..... tl
m,nlmum energy desIgn
FIGURE ..... Ilo
loo,.
246 SEC'lION.O DESIGN Of MINIMUM ENUGY HEAr UOIANGU NElWORKS
SECTION.. DESIGN or MINIMUM ENERGY HEATExCHANGEr. NTWORt:5 U7
Fig. 8.411 by supplying more energy to the process (and removing more). We
discuss thIS procedure in the neKt two sections.
optimum value of the minimum approach temperature; i.e., as we decrease liT.. u.,
we increase the area. Moreover, this optimum value will change with the process
flows We discuss this optimization problem in more detail later in this chapter
Optimum Value of the Minimum Approach
Temperalure
As we noted earher, the minimum heating and cooling loads change as we change
Ihe minimum approach temperature. However, since the heat-exchanger area in
the neighborhood of the pinch WIll change in the opposite directIOn, there Will be an
Additional Complexities ill Ihe Design Procedure
The design probkm IS nOI always as simpk as the example considered. Thus, m
some cases it is necessary to splil streams. These additional complexities are
discussed in Sec. 8.7.
Stream 2 3 4
SU<affi 2 3 4
FC, 1000 4000 3000 6000
FC, 1000 4000 3000 6000
70
,so
'"'
'so
')
Q 50
200
"'''
I
-)H
~
Q ~ 240
.J
'so
""
rV
Q
~ 60
Pinch
C - 201 T
Q ~ 12lrl
,...
C ~ 401 )
''''
70
""
'so
'"'
(')
Q
50
----.:
200
'90
Q ~
Q
~
240
A
'so ,'"
Q ~ 60
Pinch
C - 20) Q = 120
~
,'"
'00
'00
FIGURE U-l2b
Loop'.
FIGURE 1.4-1z.c
Lao",
SECTlOI"" LOOPS AND 149
8.5 LOOPS AND PATHS 2 3 4
Loops and paths provide .....3)'5 of shifting heat loads through a network.
Lool"
A loop is a set of connections that can be lraew through a nelwork that slans from
one exchanger and returns 10 the same exchanger (see Fig. 85-1). A loop may also
pus through a ulilily (sec Fig. 8.5-2). The existence ofa loop implies thallhere is an
exIra l:xchanltcr in the nClwork Thai is. If we breal the loop, we can remo\(: an
exchanger
3
Breaking Loops
FIGURE &.s.1
....'"
Paths
3 4 2
120 - Q
I I
H')- 'JH
{
Stream 2
'00
Stream 3
180
Cold
millt)'
00
Stream I
80
HOI
I
utility
3 4
FIGURE &5-2
lDop lhroup UUhl)
nCURE &5-3
Breakml! Ioop$
(8.5-1 )
(85.2)
Three design heuristics have been proposed by Linnhoffand Hindmarsh:
First. break the loop that includes the exchanger with the smaUest
possible heat load.
Always remove the smallest beat load from a loop.
Consider the cnmple in Fig. 8.5-). The energy requirements are satisfied for any
value of QL' However, If we SCi Q ... 20, one of the beat exchangers (connecting
paths) in the network disappears
or course, the loop shown in Fig. 8.5-3 is the same as onc of the loops in our
deSIgn problem, Fig. 8.4-12. We always satisry the heat loads of each stream by
subtracting an amount Ql from one exchanger, but adding it to another exchanger
on the same stream. An example where we break OM of the other loops in Fig.. g.4-
12 is shown in Fig. 8.S.l
If we break a loop that crosses the pinch, Donnally we violate the
minimum approach temperature in the revised network. (8.5.3)
Of course, if we violate the minimum approach temperature, we must find some
way of restoring it. We use the concept of paths for this purpose.
I-Ieuristics
A path is a connection between a heater and a cooler in a network. Figure 8.5-4
shows two possible paths for our example. We can shift heat loads along a path, as
shown in Fig 8.5-5. We merdy add an excess amount of heat to the hot utilit), and
subtract it from another ell:changer on the same stream (so that the total heat load
for the stream is Of course:, we also reduce the heal load on the other
stream that paSKS through this exchanger 11lu5, .....e must add heat to thi't $lream
in either anotber exchanger or a cooler
2SlJ II LOOt'S AND 'ATHS
SECTION" UOUCINO 'nIE HUMIlEI Of EXCIIANGHS 251
8.6 REDUCING TilE NUMBER OF
EXCHANGERS
We can summ;ln:ce some rules concerning Ihe design procedure:
(
The number of exchangers required for Ihe overall process is
always less Ihan or equdl 10 thai for the minimum energy
nelwork. (8.6-1)
\
If Ihe design procedure for the mmimum energy nelwork is used,
there will normally be: loops across the pinch (8.6-2)
( (
S...earn
FC, 1000
2
4000
3
3000
4
6000
FlGUIitE 15-04
Pum..
( C
'"
o
783
50 ... 20
170
4O-W
".
""
+W
'I.. --2 - SO
200- "" ""
-
I---
( )H 7
-T= I
,
Q;
Q:2(-20
Q = 240
......... T:
Q - 2
A
""
''''
J
Q
_ 60
Pinch
C - 20( )
Q - 120
0-20
110
C - 40
C-40+20 1
"" 90
Break loop conulinlOg smaliesl heat load Q : 20
''''
T-
FIGURE 8.0-1
Break loop 111 mlmmum ene'IY dOli"
H
Q
Q - Q
I. C,.,ll shift heat along a pam.
2. ".nsfer heat across the pinch-more heal in, more heat out
3. 11)(" to restore minimum I1T.
FlGl tU: LS-S
Shll"ll--"'l alon, a path
Note Ihat
\\'heD we add heat 10 a heater and shift It along a p3lh, ,""e must
remo\e the same amOUDI of beal m a cooler. (8.5-4)
We: .,1Ien Shlfl heal along a path 10 restore a minimum approach temperature; thIS
pr"' .....lure always increases Ihe energy consumption of the proo=ss.
SEC'f'IOtl U "EDUCINO mE tNMIU Of" f.XCHANGEU ill
We can break these loops by transrerring heat across the pinch.
but we Will IDtroduce at kast one violation or the: specified
AT..... = (8.6-3)
We can restore AT.... by shifting heat along a path. which
Incn:ases the energy consumption of the process. (8.6-4)
Hence, we have a procedure ror reducing the number or heat exchangers (whICh we
expect Will reduce the capital cost)atlhc ell:pense or consuming more energy (which
Will IIlcrease the operaling COSls). Obviously. we want 10 find the heat""Cll:changer
network (as a runctlon or the process flows) which has the smallest total annual
cosl.
Breaking the Loop with lhe Smallest Heal Load
From Fig. 8.4--12 we see that the smaBest heat load in any or the loops IS thal In the
oookf". where Qc _ 20 X 10'. We arbitrarily decide to break tbe loop shown in the
last diagram or Fig. 8.4--12.. Thus, we start at the cooler and subtract and add
= 20 as we prooccd around the loop (see Fig. 8.6-1). Now "'e calculate the new
heat loads and the new values or the IDtermediate temperatures: sec Fig. 8.6-2.
From Fig. 86-2 we:see that the eIltlemperature or stream I is actually 10"F lower
than the inlet stream ror the ell:changer With Q = 60; ie.. the approach temperature
is -IO"F. which 15 Impossible.
Stream 2 J 4
Stream
Fe,. 1000
2
4000
J
JOOO
4
6000
'000
4000
T,., 250
JOOO 6000
r = 250
II = 70 + Qf
.... 70 - QE
-T-90
T = 110
Q - 220
/ O__---"Q_-:....:.I20=-_--<J Q = 120 - Q,
Q = 120 - QE
T"" 150

-o60 + Q
60 + QE
T = 120
II .. 70
....-T - 178.3
-T-90
T = 190
Q"" 120
-T= 115
Q = 70
\r = 200
C = 60
AT minimum
violatioo

T - 130
T .. 1lO -T-145 T=I30 /
T = 200
""
-T= 100
-T= 100
(120 - QEJ )( 100 = 3000(110 - 90) SO QE" 60
FIGURE &.6-2
Cooler raJi()vN
FIGURE U-J
Ule path 10 1'CSl01'1: Ii. T_
(8."')
Restodng AT....
We restore the mimmum approach temperature at this point 10 the network by
shifting beat along a path; see Fig. 8.6-3. Since the oullet temperature of stre<tm 1is
120F, we want the new inlet temperature to be lIO"F. Then we calculate the
amount of heat that we must shift along the path to obtain this intermediate
temperature. From Fig. 86-3,
(120 - Q) )( !OJ = J(XX)(IIO - 90}
Q = 60)( 10J
Bruking tM Sood Loop
Breaking one loop through a cooler (sec Fig. 8.4-12) has both loops
through tbe cooler. However, there is still a second loop remaJOlDg ID Fig. 8.6-4
The smallest beat load in this loop is 10)( 10J Btu/hr, and so we and add
Q = 10 as we proceed around the loop_ The resuh is shown in FIg. 8.6-5, and the
new stream temperatures are included on this figure. For Ihis case we not
encounter another violation of liT..... (although we orten do). 10 fact, mlOlmum
approach temperature for thIS design exceeds IO"F (120 - 107 = 13 f).
The reVIsed network is shown In Fig 86-4
S=m
Fe,
'000
2
4000
3
3000
4
6000
Stream
\000
2
4000
3
3000
4
6000
T=2SO
....--T"" 250
-----..x) H 130
)10 - Q
T= 240
"
T = 190
H so 130
-..-..A ......
r
_ 168.4
T"" ISO
.30
c = 120

-{J220 + Q
\
Q _ '20 T- .>0-;1..
'20 +
120 + Q
T ..1'211:0,,+_--1'---_---,;;-::-;;;__-+__1<-__
" (').__ .::-..::60::-_-060 - QE
-T:90
C
T = 130
-T-90
T = 110
Q : 220
c - 120
Q - \20
--T.. 145
Q - 60
T - 120
-T- 100
-To;; 100
FIGURE .....
Reviled nelwork.
236 UCTlON U aWI./CIN(l TIft !./lJNUJl Of UCHANGElS
SECTION IT A MOll' COMPI.UII! DESION AUXiJlilllM SPI.ITTII'IG 257
Minimum exchangers - .5
Heat in = 130 Heat out =
Stream
FC,
2
1000 4000
,---T - 250
3 4
3000 6000
120
8.7 A MORE COMPLETE DESIGN
ALGORITUM-STREAM SPUlTlNG
There are some situations where our design procedure does not seem to work
However. these additIOnal complications can always be accommodated by splitting
a stream Several examples of this t)-pc are presented belo... and then a more
general design algorithm IS discussed
Number of Ilot and Cold Sareams
Consider the example shown in Fig. 8.71. The values of f /fC,H for both hot
streams are less than the FcC"c value (or the cold stream HOVoever. if ....e install any
heat exchanger. such as ooe bet.....een streams 2 and 3. tbe.n increase to
Mmimum enf:rgy. exchangers = 7
Heat in = 70 Heat oul = 60

FC, 1000 3000 5000
Use path to
reduer. .GT to
minimum value
I
-T= 168
130
................. 130 + lJ.T
()""'---<1
I
I'D
Pinch
...---l_---.L L--_--"
RGURE 1171
A. CO\Inle"umrlo:.
"V
--T= 190
107--
120
-T= 142..5 T
r'"\ Q-50
T., 120
'"
-T= 100
flGURE 1.6-6
Fillal dcsip-mllumum eu:halllt't1
FC, '000
3000 3500 1500
FIGURE 8.1-2
Srlll the cold sl,u.m
130
-T= 130
t
A
------
I'D
Pmch
Fioal Design
A final design is shown in Fig. 8.6-6. Now the.re are five uchangers, rather than
seven, but the heating and cooling requirements have been increased by 60 )(
10) Btu/hr, i.e., Ihey have almost doubled. Thus, it is essential to consider the
capital and operating costs of the ahemative net .....orks. We could decrease the
utilities requirements somewhat by shifting heat along a path, in order to reduce
the minimum approach temperature to 1000F (rom 1]OF after we break the second
loop. Oearly, we need to optimize the design.
258 Sf(.'110N 11 A loIon Ut;SKiN "LOGIITHIoI_STIEAM SPUTnNG
11 I)6KiN AlOC:ilrnllol-STJIEAIoI 259
4000 2000 3000
Fe, 2000 4000 6000 1000
130

r'
(
T: 130 ""
Ploch
140
Pinch;n --'- -L__
140 130
nGURt: 1.7.)
ADolne, COWllcrnample
FlGUIl[ 1.7-6
Spin a coold
FC, 1500 2500 2000 3000
Pinch:.,,-' "-- L __.L_
140 130
Pioch;;;-__---''''' L __L_
140 130
T ... 140
------
nCURt: 1.7-4
Sphl lbe hOI Arum.
FC, 2000 10,000
)
1000
Stream data
al the p1l'lch
r
rNH s NC'! I
y"
No
r
:s: FcCpc
Sphll
for every
cold stream
pinch match
y"
INO
L
Split a
51ream
Place pioch
(usually hoi)
nlalches
nGURE 1.75
ADOlhcr l;Ounlcrcumplc.
FIGURE 8.7.7
I)Qlgn p.occdUIC above the pmch. (From B UnnJrolf tf Q/. 1981.)
su.-no", It IIUl "''''D P'OW'[a INlWUlIO", 261
Stream data
al the pinch
Splilling HoI "ersus Cold Streams
We usually a hot stream as in Fig. 8.7-4, but nOI always. Consider the problem
shown In FIg. 8.7-5. Jr we: spht the hot stream. then Nil> N
e
, which is not
File,," H'rs-us FcC"c
As .another example: conSider the in Fig. 8.7-3. The criterion Nil < N
c
is
satIsfied. but now File"" > FcC"c. which pre:vents a feasible match. However if
""e spill the hot stream (see Fig. 8.7--4) and adjust the flows such that Fe'S
FcC"c. we obtam a satIsfactory solution. R ,,11
Heal Engines
Every lexI on thennodynamlCS diSCUSses the perfonnance and effiClCncy of heat
engines as mdividual entitic:s Howevc:r, if we lake a total systc:ms viewpoint. we
obtain a lolally different perspecrive. This systems approach was previously
discussed for gas absorber/distillation processes in Chap, 3.
Suppose we conside:1 a cascade diagrnm for the process sho"'n in Fig, 8-81
We pu! heat into the network, no heat is a1l0l01'c:d to cross the plOch. and we remove
heat below the: pinch. Now suppose that we inslaU a heal engine above the he:al
input to the cascade diagram (Fig. 8.8-1 shows this arrangemenl at the hlghcst
possible temperature, although from Fig. 8.15 we sec: that it could be a lower
temperature). Uwe add an incremental amount of heat IYto this engine, recover an
amount of work W, and reject the remainin8 heal QIoo 10 the network (which is the
amount of beat that we: were required to add in any case), lhen the efficiency of the
heat engine based on lhe inCI"eme:ntal amount of eoc:rgy input is lOlly"
ThermodynamiCS tals imply that the efficiency of a heat engine is always less
than 100Y. because some of the heat output must be wasted. However. wilh Ihe
arrangemc:nt shown in Fig. 8.8-1, this waste heat is just whal is required for anolher
task.. Thus., again. a systems viewpoint leads to different conclusions from the
consideration of a particular unit in isolation.
From Fig. 8.8-2 we sec: that if we install a heal e:o:,:ne below the: pinch. we can
also oblain an incremental efficiency of We convc:rt some of the heat that
would be discarded to a cold utility in any event into useful power, and then we
allowahle. Ho",e:\e:r. If \loe: spill a cold stream, We can make: a match between
stre:ams I and 4 away from the pinch and thc:reby &\'OId the: F
u
e,,11 < FcC"r
constraint (see Fig. 8.7-6)
8.8 HEAT A 0 POWER INTEGRATION
General Design Pl"ocedurl'
A gc:neral dc:sign algorithm for conditions abov'e: and belo", the: plllch is sho",n In
Figs 8.7-7 and 8.7-8 Afte:r we put 10 the' plOch matches corrc:ctl). It a
sImple: task to complc:te: the: deSign.
According to the first law. heat and power are related Thus, it should not be
surprising that the energy integration procedure can be extended 10 giv'e results for
heat and power integration. A detailed analysis of heat and power mtegration can
be found m Towsc:nd and Linn hoff. lIere:, we only outline the basic ideas.
(87-1)
(87-2)
te:mperature: of Slre:am 3 will prcvcnt any match for stream I because the mimmum
approach temperature wIll not be satisfted
We: can resolve this dIfficulty very sImply Just by sphttmg Slream 3 (see Fig.
8,7,-2). IIhhou8.h we be careful that F" Cpu S "-cC"c remains afler the
spilt From thiS e:13mple: ....e: rccognize Ihat the:re arc: ne:w design heuristics:
Abo'e: the: pmch, we must hav'e: Nil S N
e
.
Bdo'" the: pinch, ",e: must have: Nil N
e
.
I
I
N
H
S
Nc? I
y"
No
I
FcCl'C
Spilt a
for every
pinch match
hot stream
y"
INo
L
Split a
Place pinch
stream
matches
(usually cold)
FlGUR.- 1.,-1
DeslP ('fcaadUIIe !odmtr the- pnch_ B LmnItoffn J9I])
0 W To-'nKnc! and B llnllholl. -IIe111 .1Id Po-n Nc:tw<wb III Prooess Oarp. rarU I and
A/ClrE J.19 '''1, 7-48 (1983)
262 5teliON II UEAT ArlO !'Own ''''uo.....nOrl
SEerioN.. HEU AND roWEl ISTE<:iUTIOtl 263
Purl

-
Q
EffiCiency - Stand Alone
(b) Across the pIDch
Q
Heat W
engine
1 1
Q.
Energy
=d"
Pinch
Hal
~
engine
Energy
cascades
Q, W
!CI... IQ
+ Q - W
Q_- W
Efficiency == ~
(6) Below 1M pinch
I
Heat
~
engme
IQoo
Energy
cascade)
flGURE .... 1
EfficlC.nCY == 100'-' Hell CIIl'nc abon the poadl.
E!1ergy
cascades
The heat and power integration procedure is also verrusefuJ for lhe design of
utilities systems. Thus, if we energy integrate a whole petrochemical complu,
including the utility system, we can often oblain large energy savings.
Heat Pumps
Heat pumps are the opposite of heat engines. We put work inlO a heat pump to
raise the temperature level of the available heal. From Fig. 8.8-3b we see that if we
place a heat pump across the pinch, we reduce the heating aold cooling require-
ments of the process. However, as shown in Fig. 8.8-3c, placing a heat pump above
tbe pinch docs not provide any benefit. Moreover, as shown in Fig. 8.S-3d, placing
a heal pump below the pinch increases the energy requiremeot of Ihe process and
Ihe amount or energy rejected 10 the cold utility_ Thus, we obtain another heuristic:
discard a smaller amount of heal to the utility. However. if a hc:at engine takes in
energy above the pinch and discards it below the pinch (sec Fig. 8.8-2). then we gain
nOlhing from heat and power integration; i.e_. the efficiency of the heat engine is
exactly the same as it would be if the heat engine were isolated from the remainder
of the process. Hence, we obtain this heuristic:
Place heat engines either abo\'e or below the pinch, but nOI across
the pinch. (8.8-1)
Design Procedures (or Heat lind Power Inlegration
A design procedure for heat and power integration has been presenled by
Townsend and Linnhorr. The procedure is basically an attempt to match the
enthalpy-temperature profile of various types of heat engines with the profile for
the process (the grand composite curve discussed 10 Sec 8 I-I is used for Ihe
matching). The details of the procedure can be found in Townsend and Linnholf's
paper.
flGURE ......1
Ileal CIIjpncl \/Cf'l'" tbc ptDdt
Place beat pumps across the pinch.
(8.82)
264 Sf:CTION It IIP.AT ",..0 DISTILLATION It HEAT ""'0 D1STIU.AOO.. lb5
Column
nGURE U-I
Dlstlllanon cotumn
Heat in
1
I
Heat out
Heal in
Heat out
w
Energy Q + IV
mcode
0-
(b)
Energy '-:;;-
cascade J Q
Energy
cascade
PlOch
Energy
cascade
Q..
En,,,",
cascade
(c)
f1GURE U-J
Heal
8.9 BEAT AND DISTILLATION
Energy
"""""
Q_+ W
(d)
Distillation Columns abOle or below tbe Pinch
Suppose .... e consider Ihe cascade diagram shown in Fig. 8.9-2. We borrow a certam
amount of energy Q[ From Ihe cascade. use it 10 our distillation column. and then
relurn It to the cascade above the pinch WlIh tbis approach we run our column
on borro.... ed energy.....'Ith no e:uemal ulililies. Hence, the operating costs for the
column are reduced significantly. Exactl)' the same behavior is obained if .... e can
borrow the energy 10 fUn the column from below the pinch and lhen relurn il
below the pincb (see Fig. 8.9-2). However, if the column 10 be energy-
integrated bUI il crosses the pinch. then the result is no better than if we had
installed the column as a sland-a1one unil; i.e., we must supply tbe extra energy
from a bOI utility and remove il to a cold utility (sec Fig. 8.9-3).
Of course, if a base-case design indicates tbat a column faDs across the pinch,
lII,e mighl be able 10 sbift it above the pinch by raising the column pressure or to
shift it below the pinch by dropping its pressure. Induslrial case studies at Imperial
Chemical Industries and Union Carbide indicate that substantial energy savings
can be obtained by this pressure-shifting idea.
The results above gi"e us a new design heuristic:
B Lmnboll II Dunlo,d. and R Smnb, -Ileal Inlc&flIuon d Ouldbtton Columns mlo O>rran
V. Sn,J3 I17S(t91J)
Distillation columns are one of the major energy consumers in a chemical plant.
Referring to Fig. 8.9-1, we put heat into the reboiler and remott heat in the
condenser. In all the work on energy integration we plotted temperature
scales with the high temperalure at the top of the graphs. and so we will turn our
distillation column upside do.... n and consider a heat input heat output diagram, as
shown in Fig. 8.9- Ib.
Place distIllation co1umns either abo"e or below the pinch. (8.91)
266 SECTION It HEAT AND OISTlLLo.noN
5(("110"1 It HEAl AND lJoIS1t1lUtDN 267
Purge
RCC)1:k compressor
( ;to
-,
Ii RIA C FI,,.
luene
Benzene
yd,
product 11
2
CH
4
- -
y
'---- '----
To
Ii,
Toluene
nGUR 8.9-1
Cucadc dl&g"'m Iud (F,om B
LuuJ.off, H Dw./onl. und R. S"mh, C4cm.
Clog S-I.l] /17S (/98j))
Q.
I
Q,
r-'-
<
Energy E
cascade
,
"0
u
T
h
Q,
Q,
---'-
<
Enerty E
,
a>cade "0
U
'-r
Q,

Pioc
Dlphen)"l
I
Q,
Energy
a>cade
Pinch
<
E
,
"0
u
Energy
a>cade
Q,
10-
+ Q,
same as a stand-alone column
"'GURE U-]
D.stilbuoll acre- 11M: polICh [F,_ B
1toJT, H o-j... J, -J R S-"A., Ooe... Iff1
&. ]] 117J (1985).)
FIGURE 8.9....
1I0A prlloCC:u
Design Procedure
Adesign procedure for the energy integration of a 'rain or distillation columns WIth
a proocss has been presented by Hindmarsh and Townsend" H we consider ,he
HDA process' (sec: Fig. 8.9-4) and develop the composlle enthalpy-temperalure
diagram for a partIcular set of flows, we obtain the resuhs shown in Fig. 8.9-5. ThiS
diagram shows that the column condensers and reboilt:rs fall across the pinch,
which said was nol a desirable situation.
To sec how to pressure-shift the columns and to best integrate the columns
wllb the remamder or Ihe proces:s. il IS simpler first to remove the columns from lhe:
process (sec Fig 89-6) and to consider Just the energy IDtegralion of the process
With no columns (see Fig. 8.9-1). The correspondmg T-II curves for the columns
are shown ID Fig 8.9-8, and we can move these curves around by pressure-shIfting
E U".dmu"h and 0 W To,,-nloCnd. -lle.a. IDUp'lIlO11 01 O,,"llat>on Sy.u.... 11110 TouJ
A Cooipklc Appuwdl..- P<lpcr p'Cl<CDled allhe Annu.at AlOE Mccun&. S.II Fralla><>. 19&4
'nus cumpk ..-u dc'cloped by lbe ICI Pr<><:Ua Synlhab Tum. D W To........,Dd.l U".d"'.." ..... It
Dunfurd, A P;old, 0 (" Woodood. and A P ROSIlltr
T, C I---QH---
Sf.CT10", 19 IlEAT ""'0 DISTILLATION 269
T. C 1--1 _
Reactor
pinch
,r------
i
Reboilers
Condensers
H. MW
IQ H = 3.1 MW I
FIGURE 8.9-5
Tol..llkl.... T-I/ plOI
H. MW
, compressor
flGURE 0-7
Backgrouod process.
Bc:nzcne H
2
CH.
Toluene L __-, 1 i
,- --- -,
I ,
I ;.'__-,
, ,
, ,
, ,

Diphenyl
L Purge
f------Ic}-..J
1-1
2
Toluene -.l..:'H
Reactor C 1----1 Flash
---,
I
I
I
"'=oe
proouct H
l
. CH.
,
,
I
I
----------_-1
,--
I
I
I
I
I
I
I
I
L __
Toluene '-__
recycle
H'j
Toluene _LJ'T-
Diphcnyl
I'lGURf.. 8.9-6
SimplJlied Bo....sheel.
268
270 SEc-noN U HLA I ANIl PIS'lIlLHlON
,-------- ----------,
Toluene Ikn.tene I
recycle producl H
2
, CH
4
1
I
I
I
I
I
I
I
I
Dlphen)'1 I
L
T. c
T. DC
ncnoN If HEAl Ai'll> OImLlAflON 271
Q",pl
,
I
e--
level by
,
I
Ipressure slut!:
,
I
I
" I u
Toluene
e
I :0
Iknzene

----
Horizontal mOYement
Arbilnry emhalpy basis
H,MW
H.MW
-
1------,I
Ikoue""
FlGUIlE 1.9-8
H. MW T-H p1ot-Kp:an.IOI1.
T. C
Q.." I
I--
, Qoc:p 2 <
,
,
I
Toluene
,
"
mlegrile:
u
"
l
-
Benzene
:0

Q.."
FlGURE &J..-9
IntcgrallOCl of Kp-
:an.uon system.
272 SECllOI'l" RI'AT At<D l>I!iTIu.o.nol'l
Profile Matching
T, C
Reactor
pmch
Toluenef-------,
Benzene
.0 '----t..-="::':;;::':'----.J
,I

nGURE 8.9-10
tl'llelJfllunllhe columns wltb 1M; prooas.
HDA phllnt
H,MW
SEcnoN "0 I1I>A HOCUS 273
T. C

I--===---'
_ Benane

J1
H,MW
QK'p + Qsp z 1.6 + 0.9 MW
= 2.S MW
Cf. base case of 3. I MW
fiGURE 8.9-11

8.10 HDA PROCESS
the columns (sec Fig. 8.9-9). Next we try to usc as much heat as possible from the
process to satisfy the energy requirements of the columns (see Fig. 8.9-10). and ....e
ODe or more columns to minimize the external energy requirements
(sec Fig. This pressure-shirting procedure is dISCUssed in detail for the
energy integration of standalone columns, or for the columns that cannot be easily
integrated with a process (see Fig. 8.9-11), by Andrecovich and Westerberg (see
Appendix A.6). The general approach is to split columns and pressure-shift the
sections to usc the full range of utilities that arc available
!'of J. and A .... Westoerberl- M A Sirlp, Synlbo:sn Me1bod on Ulilll" Bond,nJ r... Ilr:.lll-
OIsuDal1Ofl M AfOrE 31 J6J
A study of the sensitivity of the total processing costs 10 heat-c-xchanger network
alternath'es was undertaken by Terrill and DouaJas." They developed a
exchanger network for a base-case design (x - 0.75, Ypll = 0.4) for the HDA
process. The diagram is shown in Fig. 8.10-1. They also de"e1oped six
alternati\'e heat-c-xchanger network.s, all of which had close to the maximum
energy rcco\ery (see Figs. 8.10-2 through 8.10-7). (Note tbat the quench stream
after the reactor is Dot shown 00 these graphs.) Most of the alternatives include a
pressure shifting of the recycle column, and the oiller distinguishing feature is the
number of column reboilers thai arc driven by the hot reactor products.
D L hmD and J M loI-f;C IUxerll..l' (19 11
174 SECTION 110 tlDA 'IOllSS
SECTION 110 HOA nOC'Us 175
Flash
Water
cooler
Reaclor Furnace
AltemallYe J
Gas recycle
Toluene
recycle
Purge
flGURE 1.16-2
HDA pl"00IlU .. iI fced-dftuenl but uchanl'C' (Nolr The qUCndllllCilIn foDo""nl the rcae,o. 15 DOl
sJKr,r.o 00 Iba.c p1Iphs.) [From D L Tcmlf tIN! J. !If /)Quglos. IdEC Rrsnul'1<, 26 68' (1987), wi/h
P<',mWIOII 0f'lv If",..r..,,,,, Clvm,rlJl Socl<',,.1
40
10
I
I
30
900
700
'"
"

8.
E

500
SCR
I I
I I
I I
I I
II
I I
300
II : I
0 10 20
Enthalpy, 1()6 W

__
I PCC -RCC
i{
-RCR
flGURE 1.11-1
T-H dilJBM for the HDA p,ooas. PCC _ product oolUlDD ooodeDla; RCC _ rccydc oolumn
condenser, SCR, lnd PCR IR the rdxlllcn. for the rcqdc, It.bdu.cr, ud p.oduel QO/wnD$,
respectively; RFS and RPS are !.be 'UciOl feed and p,OOucillTUDll. (F,om D. L TrmJI <JNI J. M.
l,fEC RC"'lJrrll. 26 68' (l98n ..ull P<',mUJion O/IM A_,koll Cltcmirol StK/cly]
276 SEcnON 110 /l0" SECJlON 110 HOA PROCESS 277
PlJrge
PlJrge
Alternalive 2
, roed
oluene feed
FEHFJi"t"m",
Reacl01

yFEHE
cooler
Flash
rn
H
2
CH
4

J-
"
E
]
,
"
"0
81-
.
0
u

"
:c
],
,

]


-J
"-
iphenyl D
H
11
Diphenyl
ToluCIle feed
FICURE 8.10-3
EntllY Inlcgrauon allclnah.c 2. [hom D. L Terril/WIT/ J.}.f DO"g/a. /&.EC 26' 68J (/%'7).
..-illr of .lre A_riC'1UI Cht'miC'o/ SQC'.Ij"y.]
Allemative 3
FIGURE: 1l1(l.4
Energy inlegrahon J_ (From D. L Temll DNJ J. M lJoug/IU./<fEC ReSt''''''''. 26' 68.5 (1987).
...,,1. .....Ofl of ,he Ament',." Clrem.(01 50<:-...1)'_]
278 SIiCIION. 10 Ill)" HOCUS Sl::CTlON 110 IIDA PI.OCUS 279
Toluene feed
FEHE FEIlE h1 Furnace ReaclOr
Toluene
f=l
y FEHE FEIlE
Furnace Reactor
Water
cooler Flash
II I

1-.....,
L.......(\

'- p-
o
"8
u

Diphenyl '(
.-
E
o
"8
<5(1
Diphcnyl t y
nCURt: .. 1.5
Eacrty .nl"lnluon .hcm&t"" 4 (From 0 L Trmll-.l J M Dot.glaI. IdEe RrJlQlclt., 26 685 (/981).
oTtA prrrrun_ of tM Atm'lCt1tl ClwrftKill s..c...ty_J
AlternatlYe 5
FlGURE &.IU
EDc:raY IllugrlllOn allcm&u", S (F,_ 0 L Tarl/1 "'"J M Do.tghu, IdEC Rtf_do, 16 685 (1981),
wuA ,..m1W_ of IItt! A""'rlc,," OwmlCm Sii!fy1
280 st:cnot< ue "0" 'Il:OCESS
SF.CTlON I .. "no. '.OCU$ 281
..
,
J

, ,
,..
'"
']I
'"
...
'0'
'.06 H. 7l'
7;1)
H' ,'"
"
..
"
"
.,
" OJ -, 0

, ,
...
,
J

, ,
,.. ..,
,."
6.11
,..
60)
,OJ
'06 ."
'"
<7,
<7'
" " " "
" "
"0
m
",
,.,
10'
10}
...
H'
,-" UO UO 710
,,,
I. TAC (SIO'/)'r) basc-cuoe 110""5
1. TAC (SIO"/yr) opwnu.ed
1. Cost ....:mp. Yo
... Utihur:s (M\\') oplU!Uttd
So Utihur:s (M\\'). baK-<:asc
TABU 1.1"'1
EMf"gy intqflltion .nd oplimiulion of HDA prOC'tSS
I. TAC (1:10'!yr) fto,",-s
1. U1ilil>el \aUF {MW},

1. E.nerc- saVlnp. %
... COSI saVlnp, %
From 0 L Tt..m _1>6 J M DouI1... I&.C 11,.,_,,- 16 (1"1). "llb pem>I$.Klq from t!I< A_....n Cbe",onJ
S-,
TABLE 1.1... 1
EMfty iDlqflltion fOf the IIDA process
'l
.-
FEHE FEHE Furnace Reactor
4 HRash
cooler "r-
01
H
2
, CH.
J...
LI.- 0
i
<
E E
:;
"0 "0
"
u u .-
"
'--

..
l'.
".."

JI u "8
I}. "
-
..
)-r
y
T
ny'
Toluc
fd
Diphe
Altemalhoe 6
FIGURE 8.1(1..7
Entrl)' Inle.... non _lltm_ti,.., 6 [From D L T,.,JI and J M Douglas, IdEC h_c".16: 68$ (1987).
..il" IM""USIOfl 0/ IN A_,IC_ CNmlCDI
From D L Tcrnll -.l J M Dou&l_.. II,C 1I,......c.. 16 68j (l9In porm_ fro", lbe Amenan
Chen""..1Soaely
Effect o( Energy Integration Optimization
The benefit oblained from energy integration with the base-case flow rates for the
six alternatives is given in Table 8.10-1. The energy savings from the energy
integration fall between 29 and 4] %. which is in the range obtained by lei and
Union Carbide. but the cost savings an: in the range from -I to 5% The cost
savings are not as dramatic because the raw-material costs dominate the prooess
economics.
If we optimize each of the process alternatives, we obtain Ihe results shown in
Table 8.10-2. We oblain savings in the range from 21 to 26%, which is quite
dramatIC. The Impro\ed energy integration has allowed us to increase the recycle
flows, which decreases the rawmaterial costs. The increased recycle nows aClUally
increase the utilities consumption. but the raw-matenal savings more than
compensate (or the increased energy coslJ. Sitmlar results "'ere obtained by Duran
and Grossman. Thus, ....oe sec that we can trade savings flom impro\'ed energy
integration for savings in raw materials.
The relative Importance: of ra..... matenals and energy is simple 10 assess by
looking al a cost diagram (sec Fig. 8.10-8), wherC' the ra..... -materials costs are
expressed in terms of exceeding the stoichiometric requIrements From this
diagram il is apparent that raw-materials savings are mudi more important than
energy savings for the HDA process.
Sf:osith'ily of 1M Optimum S.,iogs
Some of the values for the optimum design '-ariables for the SI;1 altemau..es, as well
as the approltimation procedure discussed in Sec. 8.3, are showl] in Table 8.10-3.
The results, for this case study, indicate that Ihe optimum flows are qulte insenSitive
M A- Duran and I E. Grossmall, AIOlE j92 (1916).
-
'S t<,) x
til

- . "
N

H
$
e:
[
5
;> z. 5
0'
- .
l"l - 0
1-
...
[
5


z 1t
"
"

ii
?
0
."

... :
-
-
"
'"

i-i!
Recycle column - X
ii
'"
Cl
".

E

!
I!'
g
g-
O

"

"
fl
f
."
Z
0

'"
" .


"
i 5?
h
lIO
?
.; E.:

51
r
-l
,
-

0
Ii ?

-
X
N

!'

Stabilizer
?
0
ftn
"
--
m
r ?,

m
ll:

TABLE Uo-]
Oplimizalion rf:llults for HOA process wilh diphenyl by-producl
Ahema,h'
..
App'odm.'t
UN
,
I J

......,
TAC (SlO'fyT) 6.38
,."
<9,
".
.73
'J<
483
Annuiliud eapital
cost (SIO'/YT) 110
'.50
".
1.52 1.4.
,,,
..,
,,.
Anou.bud Opel'luinll
cosl (SIO'/yr) l211 lS2 339
].28 3.26 l2S 3.23
Convcraion ("4) 110 613
'"
..,
'l<
61.5 67.2 67.7
H
J
compositK>n In
JU reey<:1e ("/J
'30
119 31.1 Jl' llO llJ
'"
Jl9
FEHE cncrlY recover)'
C/J. told llream
....
".
..0 936
'16
91.9 81.5 llS! IIS,7
DT_CKl 10 II J8
"
"
,
Il 16
Number of lInllt
,
10 10 10 10 10 II 13
SllIbf/ilrr coli"""
Fraetional 1011 of
1)en1,ene (X)
0' 0' 0'
OS
0' O' O'
..
(olllm"
Fractional re<XlYCry rio)
".0
98.7 98.1 986 989 98.9 98.9 98.9
Rct'llu T.,io 120 \45
'"
158 1.91 185
'"
,7<1
Flied cooler (MW) 0 II
"'
0 0 0 0 0
Ryd, col..","
Frictional recover)',
overhead
...
",
",


'"
Fractional ruovery.
bollonu ""
107 .. .. .. .. .. .. ..
Preuure (kPa)
'01 '01
107 107 107 107 107
'01
eJ
From D L Terrillind J M DovIIU.IIF.C 11.,,.,,;,,, U "., l'C""IlI'/lfI h01ll the Amenelln ChmIonl Soc:Mi11
284 $I!CTIOM Lll SUIo4Io4AIV. ExUCfSES, "''''0 !'lONOlCLATUaE
;
II
"
,
'"
..

M

'"

:l;'!>
- -

X
M- <
2
N
,

<
u
;
'" ]] ..
1
0
..
8
o M
0
u'"
8

"
:l

.s

..

0
""
-5",

;;
-a>
'"
N
E " E z

'"
'!>

"
"" "
x

'"
-> d
.. -;. g <
0
[
8 0

....
>.
..
i
0
0
"
<
"
2
'"

.,
"'
"
<
..
'"

; 0
"'
0
<

0
..

"' ..
ili
,
II

N
a

"

0 .. 2U
u
0

w "
8
..
:l " .-
1.:1 '" ,

Summary
A very simple procedure exists that makes it possible to calculate the minimum
heating and minimum cooling requirements for a process. Also. simple procedures
exist for calculating the minimum number nf exchangen; required and for estimat-
ing the beat-exchanger area required. These calculations are possible without even
specifying a heat-exchanger network and therefore are ideal for screening purposes.
The results indicate that normally a process has a pinch temperature, and
above this temperature heat should only be added to the process, whereas below
this temperature heat should only be removed. Sinoc this problem decomposition
was not widely known a decade ago. it is often possible to significantly reduce the
energy requirements of existing processes (that is, 30 to 50% energy savings have
been obtained in industry).
A design procedure for heat-exchanger networks is discussed as well as the
extension of the basic ideas 10 heat and power integration and heat and distillation
integration. Several new heuristics were also presented:
to changes in the heat-exchanger network, provided that close to the maximum
energy recovery is obtained. Moreover, as expected, the significant design variables
correspond to the conversion and the purge composition.
8.11 SUMMARY, EXERCISES, AND
NOMENCLATURE
Exercises
I. Only add heat to a process above lhe pinch temperature.
2. Only remove heat from a process below the pinch temperature.
3. A feasible exchanger just above the pinch requires that FIIC,H Fcepe. while
the opposite is true below the pinch.
4. To eliminate a heat exchanger from a network. we prefer to break a loop that
includes the smallest heat load.
5. When we break loops that cross the pincb in order to eliminate heat exchangers
from a network, we often violate the 6 T",;. condition.
6. If we add extra heat to a process, we must remove this same amount of heat to a
cold utility.
7. If possible, always install heat engines either above or below the pinch.
8. IT possible, always install heat pumps across l.he pinch.
9. If possible, always install distillation columns either above or below the pinch.
1.111. For one of the design problems that you have considered, assume a value for h T.
1

and include the heat effects in the separation syslem.

(a) Calculate lhe minimum beating and oooling requirements for the process, and
find lhe pinch tempcralUre.
285
Fe" kWrC
T... T_.. C
3 '80 60
I ,>0 30
2 20
'"
,
80
,..
H1'1T _ 1000e, find the minimum heating load, the minimum cooling load, and the
pinch temperature. Design a heat--eltchanger network for the maximum energy
recovery both above and below the pinch. How many loops cross the pinch? Break
all the loops that cross the pinch, and restore 4
(b) 00 any of the: distillation columns fall across the: plOch? If so, pre5Surc:-shif, these
columns, and recalculate the minimum heating and cooling loads.
(c) Design a heat-exchanger networK for above and below Ihe pinch, and calculate
the costs.
(d) 110111' many loops cross lhe How can you break these loops?
8.11-2. Consider the f10wshecl for the hydrosuifuriullion of oil, shown In Fig. 8.11-1. From
an mspection of 1h" llowshc:el, can you suggest a way of savmg energy? Can you
oblam a CSlImatc of Ihe amount of energy lhal you can save? What might
pre\cnl you from r"aJlllng this energy sa"lOgs? If the hydrogen IS recycled, diSCUSS
the effect of Improved energy integration on Ihe optImum purge colllposnion in this
gas-rec.:yde loop
8.1 J.). ConSider the slmphfied 1I0wshcct for producing ethylene from ethane' shown in Fig.
The reac;uon roDditions aDd some features of the design problem are
discuS$Cd in Exercises 5.4-8 and 6.8-7. The reactor dtl.uent stream must be quenched
10 7SO
c
F to prevent roklng. From an inspeclion of the f1owshcct, CaD you suggest
some modifications that could be made to save energy? (In particular, consider the
reactOt feed and product streams.) Would energy savings be expo;;ted to affect the
opumum proces.s flows?
8.11-4. Calculate the opumum value of 1'1T.... for the problem presented in Sec. 11.1. Assume
that the overall heat-uansfer coeffiCIent for all stream matches is equal to U = 50
BlU/(hrft
1
."F)
8.115. For the example dISCUSsed in Sea. 8 I through 8.6, how much can we decrease the
ener@) consumption If we reduce the lJ. T from I J to IO"F after we break the second
loop (sec FIg 8 66)1 IF I'ie consider the same problem but ....e break the other loop
that crosses Ihe pinch first, do we obtain the same results? Estimate the costs or the
U1illlles and the heat euhangers for the minimum energy design, Ihe design where we
break one loop, and the design where we break both loops.
Consider these dala:
-
[
{;

<i:
u.
o

g
v
u.
o
o

N
----Iu.
o
o
j\f
E
:;
v;
u.
8
,-
-( v)-
- -
8. E. 8rown. A Ol.IIlIl"rr lHsu!f"..:t:,. WutunglOn Unlverslly DcsIIO Case Study No. tO,edned by B.
D. Smllh, W....bJn&ton St. I.oUIS. t970.
, W L Bolle&, Eth)'lmt: Plll1Il Dr3/{J" unJ &orlOn"a. Wuhtngton UniversllY Design Cue Study No 6,
edited by B D Smith, Washmgton University. 51 LoUIS, Mo., 1968.
286
288 UcnON III SUMMAItY. UUClSS. AND NOME1'ICl.ATlIItE
Nomenclature
A
C"
F,
H
h,
N,
N,
N.
Q,
To
u,
W
OH.
OT
.:1 TI....
Heat-eJl:changer area
Heat capacity [Btu/(mol CF))
Flow rate (moIJhr)
Enthalpy (Btu/mol)
Film coefficient (Btu/(hr ft
J
CF)]
Number of exchangers
Number of streams
Number of utilities
Heal load (Btujhr)
Temperature (DF)
Overall heat transfer coefficient (Btu/(hr ft2. cF)]
Work
Heat of vaporization (Btu/mol)
Temperature difference CF)
Log-mean temperature driving force (CF)
CHAPTER
9
COST
DIAGRAMS
AND THE
QUICK
SCREENING
OF PROCESS
ALTERNATIVES
Our previous synthesis and analysis enabled us to estimate thc optimum design
COllditions (using a set of case studies) for a single process alternative. Of course,
our goal is to find the best possible alternative. However, before we proceed further,
we want to be certain that we understand the costs associated with the alternative
we have selected. As a first step in this evaluation.....1: need to consider some way of
summarizing the cost information.
9.1 COST DIAGRAMS
Com'cntional Approach
The conventional approach for summarizing cost information is to prepare a table
of equipment costs that is coded to a flowsheet. Similar types of equipment. i.e.,
compressors. heat uchangers, distillation columns, etc., are all grouped together. A
similar table is prepared for operating costs, and all the individual values for a
particular type of utilit)', i.e., electric power, steam, etc., are summed and reported
as a singk enlry.
289
AD uample of a cost summary for a published case study for the dlSpropor
lionation of 101uenc is given in Fig. 9.1- t and Tables 9.11 and 9.1 2. Figure 9.1-)
shows the: ftowsheet, and each piece ofequipment is given a special label. Table 9.1-1
then gives the capital cost ofcach picoc of equipment shown on Fig. 9.1-1, where
aU similar types of equipment arc grouped together. The table also summarizes the
total capital cost for each type ofcquipmenL In Table 9. 1-2 we find the tola) anoual
COSI for each typc of utility used as well as the annual costs for a number of faClors
f._.. J ltecpaebad aDdl T &..1;111<'0. w............. U"'_J 1'<0 ',0<1"011 bt II D WIlll.
WuIw>&u", U.........'J. 51 !Au... Mo. 1969
TABU. 9.1_1
In,eslllleni lloumffllry. S
Pumps (1949)
Toweu (l%9)
P_I
'-""
T-I
P2
1,320 T2
ll,600
P-J
1,9SO T-J lS,SOO

1,680
PI
2.100
TOlaJ 91l.100
P6
,ro
T.:lY5 (1969)
TOlal
14,180
T-'
'-"'"
T-2 ll,200
Pump' (1949). 1I.c:ludln'ipolIQ 28.160 T-J
.,""
p",mps (1969)
10."" Total
"-""
EJ.cballl"n (I968)

E-I
1"il,OOO
C_I
E2
IIS,OOO
C-2 JU.ooo
E-J '.800
22.000
Of\lmi (1969) 23,oSO
E-' 26."" IMtalkd CO$I summa,)
E-6
""" E-1
'....
p-..
IIS.COl
E. ',200
EJ.du'lllftl
1,140,000
E-'
16,000
RektOr 12&,000
E-IO 17.000
To.. crs tea 490,000
E-II
,.m
Tn)'s 395.000
Comprcuon 151.000
E-12
6.100
"""'"
lJO.o.Al
---
Toul 390,100
F._ S23.OOO
---
bchangcn (1969)
....""
3,142,000
fumaca (1969)
209.""
RQClOf (1969) "_800
N
W

+
J
11
R. J Hcnpttbd; and J T Bancbao, Dupr"PO'/>OftDf_ ofTo/wM, Waihlllgton UIIlwt'nlly Ikstp
Case Study No 8. odued by B J) SlIlJlb, Washmcton Un,vcrilly, St. L..oUI$, Mo_ June 26, 1969
291 S(;l"ION JI COSI
SfXTION " COS'1 OIAG..... 193
J M I nd D C Woodcock. /6 EC I',oc t)eJ 0",. 27 970 ( I!J85)
Malnab., bbI 'caknd:u day {6O"F}
ToIlXIK feed 3710

8enut>e 1m
2000
1I,l'ecd. 10'" SCFO I
Fuel ps. 10" Blu/day
Pr.apa1t,. b. 1__"'" ...., "'...., .... prod_
..111110: prgducu n111Cd ., IMr 1M ._Iot)'
dcpaodIoo tM loooct-1I 01110< -.. -.d ...,_1
...... ,.. _a. ""'" '" II.. oh,l'"""'1. nc-
, ItIdudin, po) charJCS.
F.... k- J He1l8"'<'b<d:.1Id J T ea....oo, Wn1l...,.....
u..__ly 0...., e- SolOdr No .. od,ca1 "" I. D
Smllh. w.......pooo UIIi....IJ. 1.- Mo. 1969
,,,
'I'
.06
."
100
1.20
."
.11
.])
.06
."
,.
No
Cost Diagrams
As an alternative way of summari;tjng the most important processing costs, we can
put all the costs on a fJowsheel. The annualized, Installed equipment costs are
placed inside the boxC'l on the fJowshttt. and the operating costs arc attached to the
stream arrows; see Douglas and Woodcock With this representation. we have a
costs. SI(l6/y.
Utiliues
Labor Ind
Tue:s Ind IDJurance
Rqwn IDd IlUK'dIaneow
eo...,.
SAR.
CaIIIYSI lII....nlo')"
TABLE 'U-3
In,"eslmenf lind summary
ConVC:f5IODjpan,
Pu.1'" p!I
SIll"
ISBL
OSBL
UIIlInes
Power m
Slum
,'"
lJ)
Wile.
,.
TOlal l.10.,
lobo<
" SupetYWOn
" Tues. 100000rance 166
Repans
"" M..,.,n.,neous Il
Paymll d11.JeS J2
T.... II,.
SARE
".
eo...,. ..
2060
that we have not considered as yet, such as labor, taxes., etc. Table 9, I) presents a
somewhat dIfferent summary of this information. and it includes the costs for raw
materials. One disadvantage of tables of this type is that it is difficult to visualw:
any type of capital versus operating cost trade-off. Another disadvantage is that we
10Sl': track of the cost of particular operations.
Process Alternatives
Our goal at this point of our process development is to make a judgment (guess) as
to whether there might be a process alternative tbat is more profitable tban OUt
base-ca.se design, rather thaD to undertake a detailed cost analysis. We expect that
most of the possible alternatives will have approximately the same costs associated
with labor, taxes, maintenance, repairs, etc.. and therefore we neglect these factors.
Similarly, we note that the cosls of all the pumps and the drums are small.
compared to the other items (the reactor cost is also small for Ihis particular slUdy),
which is normally tlte case for most plants. Thus. we neglect the costs of all these
items when we are screening alternatives. In addition, we lump the co<;t of the
distillation column<; and the trays. and then we combine these with the costs of the
condensers and reboilelS to develop expressions for tlte total costs associated WIth
a distillation sera ration
TABU: 9.1-2
Operating summary
F..- R J Hmll"C:btd IIId I T
........................-Unt_)
Dts>p e- SoDd, No .. ..so.... '"
1 0 Sn.h.. w..... U"'>nwtl
k MOo. 1%9
SfCTlOI'l , I COST DlAGk,uu 295
(9.1-') Isopropanol ..... Acetone: + H
J
visual picture of how the costs are distnbuted throughout the process. Moreover, a
diagram of this type can often be used to spot errors as well as to visualiZe potenual
Impro\ements that might be obtained by moddying the flowsbcct We dISCUSS both
problems III more detail below
tr we duide each of the costs on a cost diagram by the specified production
rate, we can prepare a cost diagram that Indicates the dollars!Xr pound of product
which each Individual piece of equipment or operating cost contributes to the total
product pritt This type of IIlformation is sometimes for the deSIgn of batch
processes III Industry
A cost diagram for one alternati\e for the dehydrogenalion of isopropanol to
produce acetone: IS shown lD Fig. 9.1-2 The: zaction IS
Using Design Heuristics to Check the Cost
Dislribulion
Suppose \loe conSider the cost distributlon for the: t\loO distillation operallons in Fig
9.1-2 We: might assume that the common rule of thumb that the reflux ratio should
be set to 20-_ larger than the minimum value '4'as used 10 design column.
However, according to Happel and Jordan,- tbe dislribution of costs should be
30% for the column, 35% for the condenser and rebOller, and 35% for the steam
and cooling water at the optimum design conditions. Peters and Timmerhaus'
present some\lohat different results based on a 6.7-yr versus a 2-yr time factor, tbat
is, 15 %for the column, 10% for the condenser and reboilcr, and 75% for the steam
and cooling "ater. '
DeSPltc the fact that tbe: two sets of results are qUite different. it seems as if
both distillatIon column designs in Fig. 9.1-2 arc far from the optimum, the
column cost appears to be much too large. If we increase the reflux ratio. we expect
that the eolumn cost wl1l dc:crease (fcw.er trays, although partially balanced by an
increase in the diameter) aod that tbe costs of tbe condenser. reboiler, steam, and
cooling waler all wtllmcreasc (higher vapor rates). Thus, from just an inspection of
Note that no pumps arc Included in the dIagram, and we do not calculate the sue
or costs of either flash drums or reflu); drums when arc scret:niog alternatives.
Also. Since there are no side reactions, we do not include any raw-material or
product COSt For thiS example, we arc assummg that an existing compressor is
available that we can use to recover the acetone leavlllg the flash drum wllh the
hydrogen by-product, but we arc not allowed to take any credit for the by-product
stream

,;: I f--.l------,
"
" c

0+
- u
E"

5 .
, .
"-
N
-
" .
"-N Ul
,,->C

-"'M
"-
"

E'"'

5
N
I'M
"

e
50! U

E


!
tci

0
N
X
,
;f;
J Happelalld 0 G Jordan. C/WWW<JPrt>USJUI>ft_ua, 2d New p.)iS
I M S Peter> aDd K 0 Tunmerhau. PI",,, lNSI(pl tINI for CIonn...aJ tittilltHrs. 3d
MNnwHdl, 1980, p )
294
296 SfClION'1 COS!
the cost dIagram......e expect that some additional column destgn slUdlcs can
JustIfied
Using Ihe Cost Diagram to Infer Structural
Modifications
We can also uS(: the cost diagram to aid in generatmg process alternatn'C$.. In Fig.
9.1-2 if we consider the cost of the steam and the preheater (6.0 + 27.5). the fuel and
the furnatt (29.3 + I .3). and the partIal condenser and the cooling water
(3.9 + 7.6). then lI>e set' that we are spe.ndmg a large amount for heatmg and
coolmg as compared to the amount we are spending for energy integration (5.8 for
Ihe feed-emuent heat exchanger) Thus, it seems reasonable to try to make the feed-
effluenl heat exchanger much larl!:er. Ho.....e...er. as we try to increase the size of this
feed-effiuent elchanger. the inlet temperature to the partial conden5(:r will ap-
proach that of the steam preheater outlet. so thai the area and the capital cost 11>;11
rapidly increase. Of course, we could a...oid this difficully enlirely simpl)' by
eliminatmg the steam preheater Hentt. our inspection of the cost diagram
indicates that an energy mtegration analysis should be undertaken.
Use or Cost Diagrams to Identify the Significant
Design Variables
Numerous optimizatIon variables exist for the f10wsheet shown in Fig.. 9.1-2.
indudmg the con...ersion.. the reflux ratios in both distillation columns, the
fractIOnal reco\ery of atttone overhead in the product column. the fractional
reco\'C'ry of azeotrope o\"erhead and II> ater in the bottoms of the rce)'c1e column. the
fractional recovery of acetone in the compressor, the approach temperature
the Dowtherm fluid leaving the furnace and the gases leaving the reactor,
and the approach temperature between the steam preheater outlet and the reactor
products leaving the feed-effiuent heat exchanger.
Almost all of Ihese optimization problems involve only local trade-offs. That
is, the reflux ratio in either of the columns affects only the cost of that column, and
the approach temperature for the feed-effluent exchanger affects only the cost of the
feed-effiuent exchangers, the steam preheater, and the partial condenser. Howe...er,
changes m the conversion caUS(: the recycle flow rate to change.. and therefore
changing the con...ersion affects the cOst of every piece of equipment shown on the
f1owsheet. Thus, if the design com'ersion is not dose to its optimum value, we can
pay significant economic penalties, whereas errors in tbe reflux ratios are much less
important; i.e., the total separation cosl of either column is a relati ...e1y small
fraction of the lotal cost of Ihe plant.
A Systems Vieft'poinl
The usc of cost diagrams enables us to look altbe total system...... in Table
91-2 we are constrained 10 look at mdividual pieces of equipment. Similarly. the
COSt diagram helps us understand the interactIons among ... anous pIeces of
equipment. Thus. it is \'ery helpful in screening alternati ...es. although the final
design of the Mbe!iC" alternati...e should be reported by using the convenl1onal
procedure_
With a cost dlagram .....e can also break costs into gas-reC)'c1e effects. fresh
elrcetS-. and hquid-recycle effects. We consider this approach as y,:e look at a more
complex plant in the nut
9.2 cosr DIAGRA:\IS FOR COMPLEX
PROCESSES
The new energy integratIon procedure dcscnbed in Chap 8 lDtroduces a signifK:3nt
amount of additional coupling and complexity inlo a flowsheet. This addItIOnal
complexity makes it more dilflCult to visualil.e the IOteractions in a floll>sheet
Hentt. we oeed to find a way of simplif)ing the cost dIagram
Allocalion Procedures
Suppose we consider our process for the hydrodealk)'lalion of (see Fig
9.2-1). When we use the procedure described in Chap. 8 to energymtegrate Ihe
flowsheel, one of the alternati...e solutions we obtain IS shown in Fig. 9.2-2. ThIS
f10wsheet has so many interconncetions that it is ...er)' difficult to gain an o\erall
perspective of the process. However. this additional complexit)' is primarily caused
b)' the addition of the heat exchangers. Hence, our first task is to remme thIS
coupling Simply b)' allocating the heat-elchanger costs 10 the indi\;dual process
streams passing through each exchanger.
AUneating Heat-Exchanger Costs
Following Townscnd and linnhoft, we allocate the annualized cost of the
exchanger to each stream proportional to the indi...idual film coefficient of that
stream (set' Eq. 8.3-7). These allocations are listed in Table 9.2-1 for the HDA
process. Now the flo"'sheet again appears to be the same as Fig.. 9.2-1 except that
we have established a cost for each of the exchangen.
Lumping
We can simplify the cost diagram further by lumping costs that go togetber. In
other words. for the purpose of e...aluating process alternati ...es. there is no
ad... antage to treating Ihe annualized capital cost and the annual power cost of the
D W TOWTlsccnd and 8 llnnhnlJ. wSurf ..,. Arr. Tilrl!rlJ fOl Ileal Eachanger NrlwOIh: Annllal
mlInl! of tIM: tnsUllmon ofCbc:llltC'l1 fnpnccrs. Balh, Umled Klnldom. AI'1'11 1984
!
Purge
328 K
~
H 311K g89SK
I
89.5 K .
L 895 K. r.I- 311 K
1. I I
"''''
ReaClOr
To!uene 298 K, ~ :I
I
89.5 K
r""

~
~ I 466 K
H
,
. CH.
395 K
~ f
J
Benzene
product

8 ~
~
i2
Toluene
recyde
391 K
rl
__.J
FlGURE Ul
HDA prooeN. [hom J. M. DINg/tIS and D, C Wlll'dcock, IliEC Pro.:. IRJ. Df:lI.].4 470 UMS}.]
PtIrge
H, (eed
(8 & T)
600
- 01
Toulenc feed
Po"",
60
Toluene
recycle
!8
FIGURE 9.2.2
Cost d'JITIm rOt tnetlY.,nlerl,alecl HDA rrooeu rF."", J M "/lIN/I", mId D C Wt>Ddrn{'/r. Idee Proc Dn Df'V 1<4 no (/98S).]
300 SECTION U C()$T Dl ...G .... r.4S rOil N>r.4Pu.x
TA8U: 9.1-1
AllocaCion of COSfS
TABLE 9..2-2
RelicCor cOSI allocalion
Onllnal CO:It, SIO'/yr 6J
" " " 5lream Ructor Toluene Stabili2u
..,
n:boilcr n:boiler reboilcr
.. [kwJ(IlI
'
K)) 0.69
",
1S7
",
II1t. allocaled ar
'"

Strum keacto. RuctQf
diluent cmuenl emuent diluent
Jr, fk_l{m' K) .69 .69 .69 .69
II., a1Iocaled COlt
J"
10 10 I.
F.... J Iol 00qIu ud 0 C w"""""d, '.f;C I'rwc 0... 14 9JU(I9ISl. _h pel
.... _Do.c-..al Soernr
..
ReactQr
..,
",
" kUdQf
efttuenl
.69
29
......
SttBm klmot/1lr SIO'/rr
Gas fer;yck 3J71

LiquId recycle
"

Toluene fttd. 271}
'69 l6
U, feed.. 496
4231 198
F..... J M 00Dp.-.I 0 C "'oodcod./.t:(",,_ THo-
14 91'0 flQlSl. w1'lll p:_ lhoe A_10m.
a..-icaI_r
FIGURE '.2.J
A Simplified COSI dl.",Rm '01 lhe IIDA l'fooe:l:S IF,om J, M Dougl"l aruI D C WfI'O<kocA, IIl/X r'(K
Del /NrJ 2-4 no(l985) I
AUOCIItion to Process Streams
We can gam e\'en more mslght inlo the nature of the interactions in the proce:ss 11'
we nov. allocate the: costs to the processing of the fresh feed. the stream.
and tlK hquld-reqcle stream We base lbe allocation on the 60w rales and the
fractional heat loads or the various sCreams. The reaC10r allocations are given In
Table 9.1-2. and the healing and cooling allocations are 8"ial In Table 9.2-3. The
resull: of Ihis allocation procedure are shown in Fig.. 9.2-4. The raw-material COSI
listed is the value in ellcess of the sloJChlometric requirements for lhe reaction. ie_.
the cosl lI,e might be able 10 reduce by looking for a Dew process allemali'e.
Now we can dearly sec: that lhe: most apensive costs are lhe exoess h)'drogen
and toluene that we fc:cd 10 the process. Thus. any alternative that will reduce the
losses or the purge losses should be considered 1ne nexl most expensi\e
COSIS are associated with the: gas-recycle ftow, which are considerably larger than
the liquid-reC)'clc or fresh fc:cd COSts.
Thus. Ihe COSt diagram provides a useful tool for rank-orderinJ, an evaluallon
of process alternati,'es It is also useful for the gross screening of allernath'cs, as we
show in tbe next section.
TABLE '..2-]
He.ting aDd e:ooling cost allocation
Heal 'olld, Ileal Jo.4,
"'Bm
G,,..
SIOJ/YT GJ,..
SIO'/rr
Gu ""'"
J9.'
"
SlJ 172
l.Jquid rer:yck '.1 Il 1>1 J9
Frah fttd
'"
,.
..,
'60
70.
'"
IISO J7I
F...... J M. DonaJ.u.ncl D C Wooderxl. UlFC P., THs n..,. u
..... h,.............", al' .... Amer_n ('1onDooo1
Purge
600
Beru.ene Diphenyl
compressor separaldy. Hence. we combioc: lbese values into a total annual
compression cost. Similarl)'. we lump the capital and opemting costs for boch che
furnace and partial condenser. and we combitx the column, reboiler. condenser.
sccam. and cooling-water cosls for each distillation column inco a single separalor
cost value. Wllh this approach we obtain the values shown in Fig.. 9.2-3
Recycle compc-essor
109
TolueneL.
Feed and recycle
React..
P"",oct
H, ....
heating
I-
198
I-
cooling
371 153
L.
Toluene column
f-
Benzene column
I-
Srabilizcr

76 174 28
I I
,

,

"

"

6'.
S
" .
i!:0O
i!:

- g
-
.l!


e.
.
E
"-
IS 8
'"
"-
'"
"'
Z ...l H

"
N N
-

f-
u
:E
"-
,
"' .. '"
NOO N
'8
",Z...l H
.!!

..
- -

"' .. '"

..
",Z.J H
" .
" N
..

"' ..
'"
N
"'
"
- -
"'Z
...l H

E
,
I--
0
"
..

..
"
"' .. '"

",Z...l H

- -

..
"
'" '"
.. ",
"'
Z...l H
'"
I
"

E
"
-
, ,
i8
..
"' ..
'"
",Z ...l H
!i
Olo
iii
- -

II


..
" "
"' .. '"

",Z...l H
-
..
e
"' ..
'"

",z ...l H
'"
"0 N
E N

:t
;2:::.
J02
"

.
I

SECTION U QUICK SCIUNTNG or '10CESli 303

93 QUICK SCREENING OF PROCESS
ALTERNATlVFS
1be systematic procedure for developing a process design thai .....e discussed nrher
caD also be used to gc::nerate a lisl of process alternatives. All that we need 10 do is
10 leep a list of each decision thai ..... c make as we proceed through the base-case
design. Then as we: change: these decisions, we: generate: process altc:rnati\'cs. We
.... ant to male some estimate of the economic Imporlance of each alternative,
rather than repeal the: deSign for each case, In order 10 minimize our design draft.
We: use the COSI diagram as a 1001 in making these estimates.
Design Decisions
The decisions we made 10 generate the base-case design for tbe HDA process are
listed in Table 9.31. We can proceed through this list decision by decision and try
to evaluate the savings associated with changing any oftbe decisions. An alternate
procedure: would be to use the cost diagram to identify the largest COSlS and lben 10
identify the decisions thai have the greatest impact on these costs.
TABU '.)...1
Process .altcrn.athes for the HDA pro:ss
Ln-d2 doc:wons: Inpul-DUlpUI iUlIClIlre
I Do nOI punlr Ihe bydrogen feed Ufam.
2. Recover, ralher than recydo:. dlphenyl &0 tbal there arc thn:c pwduct ilrelllli (pur,e. benzene:
product, dlpbenrl by-product)
3 UIoC a pi recycle and puree: Slfam
level.) dceu.IOM' Ra::ycle S1rvaufC
I UK a $IuP readnr.
2. Ute; a p.J (H. and at.) and I bquid (Ioluene) ra::)'Clc 11ram..
l. Ute; a 5/1 ",.10-...010"'" 11100 10 pteW:nt cokul,-.-unun& this 10 be: a dcsip CQIlStratnl
(althouJb 01 eoWd be rorm..lalllCl as In opc.im.ization problem).
4 A PJ'-f'eCJ'dc wmprasor IS -sed.
S. Operate the reador
6. Do DOl ClOtlI>da c:quilibnwa e&cta.
Levd...... Vapor I'a::1OftI)' l)'Itea
I If a vapor ra::overy 1)'IteP! UIOd., plaa: II 011 tbe: lluh Yllpor (II reqdr: II loI1 10 by-
product) or the pUlJC: IIrel1ll (rllbcre IS 110 loA caused by bcm:ene rccydc)
2. Do DOl. l&1li: a vapor recovery l)'SteP!
Levd-4b dcoIiom Uqwd 1)"Ilem
I Mal.e alIlep11l11tions by daltiliallon.
2. Direct lCq noe oJ wnpk QOlumnl IS uscd-prnblblr UK of complu. columns Ihould be.
.........
3. Remove the bght eDds in a Ilabilizer
4, Sc:od the blbl ends 10 rue! DO vapor recovery l)'Skrn.
Lem-S deaoas. EncrC lnlegrallOA. There arc oumerow allernatlves.
F..- J M Doql.aIIDd D C WoodoDo.l, '40C /'roc Va: 14 910 It9tS). wItII per.......... 01 tile .....nc:u
"'--' """"
SlCTlO,,", U QUICK JClfENlNQ or paocf.SS ALna"'AnVEll 305
increased flow rate. From Hougen and Watson- Il.e find that the equilibnum
constant is
K ",,0.248t685 C
The stream flows for our base-case are as follows'
Diphenyl 0
(9.3-5)
(9.3-4)
(9.3-3)
K '"" (DXH1) _ 1>( 1541) .., 0.24
(Benzene)2 273
2
D _ 12 molfhr
1547
2320
91
27J
273 + D
H,
CH.
Toluene
Benzene
n""
and
Purification of thl.' 1I)'drogen F'ud Stream
The hydrogen feed stream containS methane as an Impurity, so that we mIght be
able to cut some: costs by removing the methane from the feed; i_e.. we decrease: the
amount ofineru thai pass through the However. methane is also produced
as a by-product in the reactor. Heoce, there seems to be Iittk iocentiye for purifying
the methane feed stream On the other hand. if we should decide to purify the
hydrogen-recyck stream. then .... e could feed the process through this recycle
separation s)'stem We defer consideration of this alternative until later in this
section.
For our I-IDA process (Fig. 9.2-4), the largest cost items correspond to the
use of the excess raw materials The decisions that affect these costs are made in
level 2. the Inputoutput structure of the flowsheet, so we want to start at the
beginning of our decision list in any event. The next most important COSLS are those
associated with the gas-recycle stream, and these correspond to the leyel 3
decisions. So 3g3in we find that we want to proceed sequentially through the
decision list
Req-cling of Oiphl.'n)1
Since diphen)'1 is produced by a reversible reaction, we can recycle aU the diphen)I
and let the diphenyl bUild up in the process until it reaches iLS equilibrium Ie",::! If
we adopt this approach, we chromate our selectivity 10S5eS, i.e:... the 8 molfhr or
toluene that geLS converted to dlphenyl (the purge losses of toluene and benzene are
oot affected). Of course. we lose the fud credit of the dipheoyl.
We can use the cost diagram to estimate the raw-material savings;
(9.36)
Raw Matt Savings = 1691(1\) - 200
= 840 (x SIOl/YT) (9.3-1)
Assummg that the costs ate proportional to the Increased tlo'" and using the cost
dIagram to estimate the Iiquld-reg-cle costs. .....e find that
Increased LiquKl-Req-c1e Cost _ + JJ + 7 + 43 + 39)
_14(x SIOltyr)
The increased flow also requires that we increase the gas-
rccl"ele 80w, since tbe ratio at the reactor inlet is Specified
as 5/1 Thus. the additional aromatics flow or 12 mol/hr requires ao additional
h)'drogeo flow of 5(12) = 60 molfhr However, the gas-rccycle stream contains only
40% hydrogen, so that the total rccycle flow must be increased by 150 mol/hr.
Again, using the cost diagram to estimate the gasrecycle costs for a base-case gas-
recycle flow of 3371 mol/br gives
which is a significant value compared to the other costs. In addition......e save the
cost of the toluene column ..... hicb was used to separate the recycle toluene from the
diphenyl. Again from the cost diagram we see that
Hence, the total cost savings from recycling the diphenyl is approximately
Savings for Toluene Column = 76 (x SlOl/yr) (9.3.2)
i#r(158+86+109+172)-24(xSI0)/yr) .
Tot Savings - 840 + 76 - 14 - 24 = 878 (x SlOl/yr)
(9.3-7)
(9.3-8)
which is also fairly large.
We should recogoile that this calculation is only qualitativel)' correct
because the toluene column reboiler was integrated with the reactor effluent stream
(see Fig. 9.2-2). the flows in the heat-cxchange and quench systems will
change. 1I0we...er, this "gross" screening of alternatives will help u<; decide (On the
priorities for undertaking more rigorous calculations.
Recycle of diphenyl also incurs cost penalties that are associated with
oversizing all Ihe equipment In the liquid-recycle loop to accommodate lhe
This alternative leads to significant savings and should be evaluated io greater
detail. Note that if we rec)'CIe tbe diphenyl, .....e expect that the optimum conversion
will change dramatically (up to about 9g %) because the economic trade-offs wiU
change. Hence, it is necessal)' to repeat the calculations for this case.
o. A. HOllgeD and Ie M Watson. l'ta.tlplu.1ll K"'l'llCJ ""Ii Wiley. New
VOlk. 1947. P 875.
>H;nON QUIet. SCUENI/<;(i Of PMOCI:.SS AlTUNATI"f.!i 307
Purification or the G.IlIsRecycie Stream
The cost of the hydrogen fed to the process is also very large. Most of this
excess hydrogen leaves the process with the purge stream (40% 1-1
1
), and the
potential savings from reducing the hydrogen loss from the cost diagram IS about
Other Decisions
We can use this same proocdure to obtain a estimate of tbe economic
impact of changing our other decisions. In panieular, the heatiog and cooling costs
are significant, so we want to consider olher energy integration alternatives
Similarly.....e mlghl want to conSIder the use of complex diSlilJallon columns
(sidestream, sldcstream stnppers, elc.) 10 an allempt to reduce: the costs. The
estlmares arc not very accurate. bUI they help to decide whal addit..ional design
calculallOns can be: Justified
therefore we cannot cSllmatc the incremental cost. Nevertheless, our qUick
estimates mdicate that we should explore this poSSibility in more detail
AI.MIliote thallhe large recycle flow of methane (60% CH. In the recycle gOb)
acU; as a heal carrier (diluent), This additional flow hmits the exit temperature from
the adlaballC reactor to below 13()(t'F If 1Ae remove most of the methane from the
reqcle stream. a.:. shown III Fig 9.3-1, we might exceed our reaelor temperalure
constramt If this shoulJ prme to be: the case, ..... e would change the de>lgn of the
hydrogen purificanon unH 10 allow morc IJlCthane 10 be recycled or course. this
would also require lhal we adJusl our eSllmates of the savings
(9.3-9)
(9.3-11)
Saving> .... 2163 - 685 = 1478 (x SIO)/yr)
If we InSlall a hydmgen recovery umt, such as a membrane separalOr, .... e ch,mge
both the o\'erall and the recycle material balances. Suppose we purify the recycle
stream to 95% the same as the feed composit..ion Estimates orlhe new process
flows are shown in Fig 9,3-1 The reduction m Ihe gas-rcc)de flow from 3371 to
1628 leads 10
Gas-Rccycle Saving = (I -OOXI72 + 109+ 158 + 86)
= 271 (x SIOJ/yr) (9.3-10)
Stmilarl)' the reduclion of the hydrogen feed rate saves
(
287.4 + 273\
H2 Feed SaVings = 1- 496 + 273 tl60 + 36 + 56)
= 68 (x SIO)/yr)
However, we must pay for the membrane separator from these savings_ Unfortu-
nately, Ihere is no cost correlatioo available for membrane separators, and
(9.3-13)
Savmgsfrom l-I
J
{foluene =0.5(158 +86+ 109+ 172)
= 262(x SIO)/yr)
Incenthe for Changing Process Constraints
The problem statement Imposes a constramt on the hydrogen-In.-aromalics rallo at
Ihe reactor lOlet of 5/1 10 order to prevent coke formation. Howe\ler. there IS
probabl) some uncertaint) associated with this conslrainl; i.e.. it is unliJ.e1y Ihat a
chemist would undertake numerous experiments to determine an exact cokmg
Iimil for a 1A-ide range of conversions, reactor temperatures and pressures, etc.
Instead, the chemISt would probably set the rallo 10 a sufficiently high \alue that
coking was oe\'er encouotered (or occurred very slowly).
II IS desirable to esumate the economic incentive for reducmg this ralio in
order to Juslify the cost of addirlOnal expenments. This justification needs to be
accomplished suffictently early in the development of a project thai the chemist's
apparatus or the pilot plant has not yet been dismanlled. Again, we can usc Ihe: COSI
diagram ID making these economic eSlimales.
If we reduce Ihe hydrogen-la-aromatics ralio for Ihe basc-<ase desIgn to 3/1
(Rase suggestS a 2/1 rallo), the gas-recycle flow is approximately CUI in half Then
the savings arc
(9.3-12)
CH.
287.4
D
4
B
265
91
l
951. H
2
recycle
951. H
2
/CH.
1628.4
Separator
t
Reaclor Separator
Thluene recycle
Toluene
273
(Note that we do nOI mclude the stabilizer and the benzene column fresh feed
prOCCS$mg COSIS 10 thiS calculallon because lhey aTe DOl affected by the hydrogen
feed flow rate.)
The tOial savings from purifying the hydrogen-recycle stream (neglcctlng the
fuel value of melhane and the 1055 of some hydrogen 10 the purge) are
Potential Savings = 1478 + 271 + 68
= 1817(x SIO)/yr)
951. H
287.4
"-IGUME: 9.3-1
PunrYlnllhc ICCy<;1e IlIum [F'(Im J 101 Do",/IIU QTld D C W<l<Nkod', 1,fEe !',oc Du Dn1 2. 470
(l98J) ]
U F Ra.c, Ck"ucal R'llCla, la, !'.ucns PI""IS, yol 2 C"sr SI"dJtI and D"I", WIley,
New t977. p l60
30S SFellON U HO" ...OCESS
The operability (coking) or the plant must be verified experimentally berore any
detailed design studies are undertaken. However, now we have a justification ror
carrying out additional experimental work. (Coking data ror plant scale applica-
tion are extremely difficult to obtam In a laboratory experiment. and a considerable
amount orJudgment IS required to mterpret the results.)
Revising 1M B.se-Caw Design
Once "'"C ha'"C dctcmuned these screening estimates or the proo::ss alternatl\-es. ".iC
want to use our design procedure again to obtam all improved estimate orthc lotal
processing costs. We contmue to use shortcut calculations ror those reVisions until
we are certain that the project.seems to be profitable; i.e-. we must add the costs ror
pumps. reflux drums, slorage.. control, safety. poUutiOD, labor, maintenance, etc.
Om::e all these other ractors have been included in tbe analysis. we repeat all the
design calculations, using more rigorous procedures and a computer-aided design
program such as FLOWTRAN, PROCESS, DESIGN 2000. ASPEN, etc. How-
e\'er, our goal at this poillt is to find the best prOCieSs alternative. or the best rew
process altematives.
or course, to compare two alternatives. we must detennine the optimum
design conditions ror each. We have used a case study approach to estimate the
optimum design conditions, but there is also an approximate optimi7;llion
procedure that we could usc. ThiS method is described in Chap. 10.
Toluene reed
Toluene
recyde
Gas recycle
Il<nuo<
MellON" HIlA I'IlOCESS 309
Purge
9.4 HDA PROCESS
Our gross scrccntng procedure indlcales that there is a large economIc mcenthe ror
recycling Ihe dlphenyl alld (or reco\'eriog some or the hydrogen rrom the purge
stream. There docs not soem to be a design procedure or a cost correlatioll
available in tbe literatuu ror a membrane separation system. but we can evaluate
tbe savmgs from recycling the dipbenyl using the procedures we described
previously. Of course. Wbcll we compau proocss alternatives, we "'ant to compare
them at their optimum values., and so we want to impro\'e the estimate we obtained
hy using our gross screening calculations.
RcC)"cling Dipbenyl-Optimum Design
The reactions ror the HDA proocss are
Toluene + 1-1) ..... Benzene + CH..
+ Hz (9.4-1)
FleURE 9.<4-1
HDA process ..,th recyded-altcmllUvtc 1 (Fr_ D L TnT;(J antJ J M Dawgltu. I&.C
26 645 (198..... .. .,11 of llttt A_,i(:aIt CItm!I(;.J $wtr]
reC}"ck costs. Hov.e\er. we want to check Ihis hcuristK: because conversion
normally is such an important design variable.
Sensiti"ity or the Optimum Design to Ene:rgy
IntrgTatioD
Several alternatiH: hcat-exchanger networks ror the HDA process With diphenyl
req'ck were developed by Terrill and Douglas and are shown in Figs. 9.4-1
through 9.4--5. The optImum costs for these alternatives as well as the approximate
procedure presented in Sec_ 8.3 are given in Table 9.4-1. The best altemalive ror the
case wilh recycle has a cost or S3.57 x 106/yr compared to S4.73 x I0
6
/yr ror the
case or diphenyl recovery. which is a significant change. The optimum conversion
and purge compoSlllon ror the case with rCl;yde are x _ 0.911 and Y,.II = 0.293 a<;
If we recycle the diphenyl so that II bUilds up to its equilibrium level, we efTccli\cJy
have only one reaction taking platt Our heuristic for the optimum conversion ror
a singJe reaction is approximately :r = 0.98; we trade only reactor costs againsl
l) L Tcrnlllnd 1 M DoUlLlu. J<fEC 26 68S (1987).
310 H 111M P.ocess
Purge
GaJi recycle
SfcnOl'! 9< HI)A rux;ESS 311

FlGl.RE 9.4-3
HDA procas ,,"h dlphcll)1 .h(flUlllW: 3 (F,_ 0 L T#'rrJ! lUId J M IdLe
26 (/yj '',t .. ,,10 of llot .4.......r"" CIw""",GI SocwIY]

H
2

r-
<

E
,

"8

"
i5
< ,

"
,q

r

l(
Gas recycle
H
2
feed
YFEHE FEHE1l1Fumace
Reado,
WaitT
Toluene
cooler
Flash
'''''
H
2

Toluene A
recycle
<
E
,

"8

t--
,
Be"",,,,,
"
:D
,

;;;
r0


..'CURE '.4-2
HOI. prO(:CU: W1th d,pboen)1 Rq'c1ed-a\l(flU,u,'( 2. 0 L J M o-g/;;u. tHC
1UMwdr., 26 M1 (1981), ,..,10 pmmss_ of 1M .4........IC411 ClwmlCai Soc.,.f,.]
Toluene feed
l...i FEHE hrl FEHE hFurnacetlRcaetor
r=::::J .-- '---
WaterL Flash
roolerr.
312 Sl!CTION ,. !lOA nocas
Gas recycle Purge
Gas recycle
HCIlON ,. NO" nocESS 313
Purge
Toluene feed I
Toluene
recycle
FIGURE , .....
HOI. prOQrM .-llh d,pbUlJI 4 D. L T".nJl aNi J M Dowgltu, IdLe
k_rla. l' 6fJ (11187). A_..(.... CItnnKtM SonrfJ',)
Toluene
recycle
FIGURE '-'-5
HOI. prOl::l<SS ....Ih dlphcnyl lU)'ded .hefmluV'C J D L TrrrJI.,J J /of Dougltu, I<lEe
IWNtJlUt. 26 6fJ (l1lI7). ."IIp-r....wlOft of A"",."'.... CltnotJcfl1 SocIf'I}'.)
Benzene flow
Dlphenyl /low
H)drogen flo....
EqUIlibrium conslanl
Exercise
compared 10 .\ = 0675 .. nd )',11 "-- 0323 for the process Wllh a l p h e n ~ by-
product..
If .....e had used the onglllat base-case: flows (that IS, X "" 0.75. J",.H = 04). the
cost would hd\C been S492 )( IO'/yr, which demonstrates the new for optImiza-
tion We also note lhal the optimum design conditions arc rclallvely msenSI!l\C 10
lhe structure of the heal-cxchanger network, prOVIding lhat we: obtain c10lot 10 the
maximum energy recO\cr)
omenclature
Summary
A cos[ diagram IS prepared by wrHlng the: annualized capllal cost of each
slgmficant piece of equipment Inside the: equipment box on the flowshc:c! and by
allaching Ihe: values of Ihe sIgnificant operating costs to the stream arro.... ~ The
cost diagram IS useful for checking heunstlcs. for mfc:rnng the: desirabllll) of
strucluraJ changes In lhe flowsheet, and for helping 10 idenlify lhe Significant design
vanables. The cost dIagram IS also useful for Obfalnlng \'cry qUIck, but nOl ler)
accurate, esumales of cOSts associaled With various proa:ss allernallvcs_
9.5 SUMMARY. EXERCISE, AND
NOMENCLAT R
95-L For an} of thc processes thai you ha\'c dcslgned, devclop a complclc hSI or
ahCmall\CS ror Ihc mpUl-OUlplJl SIrUf;lurc (Icvcl 2) and Ihc recyclc slruC1urc (le\c1J)
Also dc\c1op a COSI diagram for the process and dlSC:uss the economIc trade-offs
Descnbe ho.. dlC' CCOnf'mlC trade-offs for Ihc alternallves diKer from .hose ollM
base-ea.sc dcs.ln, aDd u c thc (:(1 diagram 10 csumale lbe lilt\"mgs (or !0iCS)
assoaaled with one ollM altcrnatlves..
Benzene
D
H,
K
J
u
o
1
f
N
o
N
o

o
o

.Oc c _

E
.,
0-

~ l
-
-
314
PART
III
OTHER
DESIGN
TOOLS AND
APPLICATIONS
317
CHAPTER
10
PRELIMINARY
PROCESS
OPTIMIZATIO
Numerous tellls discuss ngorous optImizatiOn analYSIS- lIo....(\cr, at this stage of
our designs. the goal is to compare twO or more prOttss alternatives \loben each is
-close" 10 its optimum design condillons. Thus, we defer an exact oplimizatlon
until we have selected the best altcrnatj."c and until we are <:erlain thaI we Will
proceed with a final design; Le., we only develop a sufficient amount of accuracy 10
be able 10 make the next decision in a sequence of decisions.
In Ihis chapler we descnbe an apprOllimale optlmization analysis that has
been developed by Fisher. Doherty. and Douglas. The objocth'cs arc 10 dClcnnmc
the dOIJIl"anl economic trade-offs for each design variable, 10 rank-order the
importance of I.hc design variables, and 10 estimate the incentive for optimizing a
design (i.e. is our initIal guess good enough. or should .....e try to improve il?). Thus,
.....e want 10 gel dose: to tbe optimum without necessarily determining the exact
value: or the: optimum. We: again use the shortcut dcslgn and cost moods ror our
analysis.
W II. M F DoMny. aad J M Doupl, S,pdicul Tradc-o& ror
Pr_ Durp aad Slud,-Sla,c CO,lIfOt OpumlUuoa Prol>lcaQ..- AICIl./.]1 IHI
319
320 SECnol'l 10.1 OUION VA'-!AILES ANI> (COHOMJC TVlO'oFTS
10.1 DESIGN VARIABLES AND ECONOMIC
TRADE-OFFS
Much or the optimization literature spends a great deal or time discussing the local
versus global optima. the issue or convexity, constrained optimizations. etc. lienee:.
we need to understand the basic nature or design optimu.ations berore we
undertake any analysis. We begin by considering two simple examples. and then ....e
attempt 10 generalize the results.
E.umpk 10.1-' R"CIOt Ulmpk. Process deugn optlmi7.atl0Il5 .rc al ....... ,.s chal.c-
teriud by economic ludeolrs. For uampk. Lcvell$pielo describes the design or an
ovcBlmpli.6cd proo:::u COOlllOlOl only I reactor. For a reaction A -0 B. rree separa-
rion 01 reactant and product, and I requirement thlt unconverted react ani Clnnot be
recyclcd. tbe total cost 01 tbe process an economk trade-off between vcry
high ra...m.lerial OMt and a 10... reactor cost at 10... coDvef5ions baJanced .s.instlo...
n....materi.l cott and I very higb reactor COlt al high cor",erltons.. ThU$,. .. ..:rite a
total annual cost (TAq model as

U
SClIOl'l 101 OUION V.... 1Jl.LU .1.1"0 TVlO(.()rr'S 321
CSTR; Firstorder reaction
600,--------=:..:..:.:::..:..:.::::...::..:"------,
The constrlmu 011 the system relate the production rate P to the rresh reed rate
F, Ind thc conttef5JOn by
TAC_C,F,+C... V.
p _
(10.1-1)
(IOI-2)
0.3
o "" Raw material
os
ConYeJSion
+= Reactor
0.7
0: rAe
0.9
(10.1-4)
.nd the desIgn equallnn ror the reactor (...e a!oliumc a firsl-ordcr. ISOlhermal rcanIDII
ID a CSTR) is
(101-])
AD design problems hue this same structure, ic.. I cost model and equ.lity
constralnlS. Wc coukltry 10 solve the probkm as a a>nswned optimiution. bUI il is
simpler merely to usc Eqs. 10.1-2 .nd 10.1-3 to eliminatc the variabies F, and .'f: from
the problem. After Ihls drmlD.tioll OUT cost model becolDC$
C,P TAC __+,
kp(1
The rlw-malerial COIl monotonically decreases from In unbounded value .t
_ 0 to iu smallesl possible value ... hen .:J: _ l. Also the reactor CO$t increases
monotonically rrom its smallest possible value ...heD .:J: - 0 to an unbounded value
when _ I. Thus, there must be an optimum valuc or which minimizes the COSl
This optimum must correspood to. global minimum, and Ihe minimum can nCVl:r be
II a constraint. Usiog the values given by Levenspiel, we obtain thc results shown in
Fig. 10.1-1.
Design problems normally are characterized hy this samc type or beh.vior. i.e..
a halanoe between monotonK:ally increasing and monotonically decreasing cost
flGlJRE 1ll.1-1
ReadDf C.=IlIP'
fuoctJons More<l\'er. the) usually SoCCffi to be and 10 ha'e a global opllmum
The c:<:onomic.potenl1al g.phs used for the development or a base-easc dCSlgo and an
C'Stim.te of the optlmum processing conc:htions thai "'crc JK"CSC'lted In Chaps. 6
thtough 8 a.Iso demonstr.te this type of behavior.
Example 10.1-2 Mioimom refluJt ratio ia I distillation
Jl"Jtems. A COSI model ror. dIStillation column in terms or tbe number or trays and
the vapor file can be wntlen as (5CC Appendices A_2. A.3. and A.4)
TAC _ C,N lVO' + C) VO' + C, V (IO.I-S)
where C
1
includes the .noualiud cost ortbc oolumn shell and trays, C, includes the
annualized cost of the condenser and reboiler, and C, includes the annual cost of Ihe
cooling waler .Dd steam.
From a malerial ballnce we know tbat
(10.1-6)
For a binar)" mi.-ture with a saturated-liquid reed.
0 2d eel ,Wiley. Ne.. York. p. no
R _
.. (0: I).x,
(10.1-7)
322 SIl(.TION 1111 OUI(,N V.....IAaLfS "'NO OONOMIC TIlAV.I!_OFFS
SECTIO" 1111
OfSIGN V.... I....LES ... ,.0 ECO,.OMI(: T.... DE_Ons 32.3
and assumlnll complele recO\ery of Ihe light component,
D - FX
F (101-8)
TABLE lo.J-1
Optimum reflux ralio
FIGURE 10.11
0punlum renUl ral'O
R/R
mJJJ
C = Column += Conden.sor <> = Rc:b.ailer 6 = Steam >l _ Cool v _ TAC
ofteo lei RIR.. _ I 1, 10 ensure Ihat ....e arc "bovc tbc optimum. bUI Ihat ....c arc shll
within the neighborhood of the opumum.
Recycle and Product Distribution "ersus Unit
Optimizations
In the examples above we considered optimum design problems for sin.&le-proccss
unils. However, the same type of behavior is encountered whcn we
optimization problems, We define a recycle to be Ihe ?f
a design variable that affects the capilal and operatmg or all equlpmenl m
a recycle loop. In contrast, unit optimizatIons concern deSign thaI affect
only a single piece of equIpment or a few adjacent pieces of eqUIpment ObVIously.
....e expect that recycle opttml,tatlons and Ihose that affect the product dlstnbullon
will be more Imporlanl than umt opumizalions
To illustrate recycle and unll oplimi.t:allons. wc retum 10 the HDA process.
H,
H __
'.
H,_2(N_I) ... H.
Dr - OOI64(,/11l119M..(T, ...
M&S D' ........"
ColulM Co.! - --(J 28)(101 1 n, , ...
,0)
M&..S 6H ) ...
:u.o (J29XIOI} V-"I)
M&S 611, ) ... ....j)
Rebollc,_ 280 (J29)(IOll 11.250 I
(
) X'H" Sleam - lOOJ ill 81S011'
(
O""X ' X
H
"
Cool W"'''r _ -- -- IlSOlV
1000 1.}4 )()
,
R ----
(.. - Ip.,
V-(R+I)D
Sro<>k...s cq"'UOIl to calculale Ihco,ehQ.llrays - N r
Feed ral" _100 nlOljhr. JI, .. 0 SO. Jr... 099.1 ..
JI .. 2.0. E .. 0 S. II .. IS ft, All. _ Il.JOO BUI/enol.
U." 100 IItv/lhr fl
l
,." 1; .. 111 "-, p .... IS pI.a,
101 .. " 92. MaS _ ISO
2
I., 1.6 1.4
Smoler's equauon
1.2
Thus.
1_[ RR. +I]Fx (101-9)
{a 1)1:,. ,.
For muhk:ompoocnt mixtures. IIIe use Ihe results of Ghnos and Malonc (!eC Sec. 73)
for R.. alon& ..... ilh lbe appcopnalC malerial balance Thus. ...c can aht.-ays ...rile Ihe
vapor rale In lemu of RIR... the feed COmposlllons, as. desired spits. elc
Whcn R IS close to R GIllJlaoo's correlalJOIl, or the appro:untallOn N :::= 1N...
gives poor results. However, for bmary systems ....e can usc Smo"e!'s equation 10
caJculalc the: of Inys, ....hile for muhicompoocnl mi.llures ...e can use
Underwood's equalion (!eC Appcodil A_1). 10 bolh cascs. the of theorelK:al
trays becomes uobouodcd u RIR_ approaches uoily. and tbe number or trays
decreases monotonically 10 Fenskc's solution for lbe OlIntmum Dumber tf RIR..
becomes unbounded.
The: various lerms 10 Ihe cost model arc !i.bo...-o 10 FIr; 10 1-2 for lhe values Il\'en
m Table 10.1.1. APJo Ihere IS a Jl,IobaJ mintmum. Note thai tbe column COSts also
ahlbn a mlOlmwu. Ie.. tbe height decreases but the dQmclcr n)(:reases as RIR..
ioaeases. From Fig 10.1-1 .. e thallhe COILS iocruse rapldl), If ... e fall below tbe
opumum.lxlllDCt"tUe only Yef) slowly u weac:ecd Ibcopumum. For thIS reason...e
20
--
':
20 -
:
t-
40
I-
20
00
1'0.
80
60
401-

3
300
2
2
2
2
200
I
I
I
I
I

324 5ECflON Itl DUIGN ......ll;IULf.S ... NO l!COi'IOlollC rllAu'olTS

Sf.cnoo;: 101 DUIGN "'UIAKf.S AHI) f:COt"OlollC tlADf:.OH'S 32S
W R rehe. M F Oohcrlr. and J M DooII.... Calculluons ofOptlmal Rrxoverv
F.aetIOM lOf 1),51lll1tlOlI Cofumns.- I<fEe I'rM [NJ Dn!_ U 955
Eump&r 10.1... A simplific.d ,rt"5toa of bUlaDC IlkylalHMl. We WIsh to illuslrate some
Important design variables that are not encountered in the IIDA process. For IhlS
purpose ..e consider I very simplified ...erslon of a butaneallr.ylallon process. .....here "e
assume that the onl)' reactions are
and "-e assume that the feed streams arc pure C.H. and i-C.ll
lo
' A SImplified
Rowsbcet is shown 10 Fil- 101-3.
Now we assume lhal 1 < 1 and tbalthe reaction klOetlCS are indicated by the
stOichiometry The economic trade-offs for thiS eX3mple arc thell as follows:
(a) Convvsion The produCI distribution is degraded as the conversion of C.I-J I In-
creases, and there is abo an economic trade-orr belWf:Cn high rellC!or cost al high
conversion and la.rge recycle costs al low conversions TIlls is a rceyck ua.deolf
The optlmlzalloll or the Ippmaeh temperature for the fttd-<,fflucnt heat
elchanger, for ellBmple. invoh-cs a trade-off betwccn the size of this Cllehangcr
and the SI1Z of both the furnace Ind the parrilli condenser (Smce only a few units
affect Ihe opumum .pproach temperature. "-c ca1lthls a urnt opllmlUtlOn.) The
appcoac:h temperature bct..ccn the feed 10 the flasb drum and the coollDg-.ater
inlet rempe-ratute to tne partial condenser. bo.e,er, In...ol\"cs a trade-off between
the size of tf10e p3rtlal condenser Ind lhe lOS!! of aromatics 10 the purge stream as
the flash telnpclature (AgaIn. thIS IS a urnl optlmluuon ) The Opllmum
.dTs for these t.. differ by 2 orders 01 mar;nttude In the IIDA process
(1 K for the partl.1 condenser and 100 K fOf the FEtlE) C1eatly.lhlS dl5Cfepaocy
cannot be accounted for by the P\lbhshcd heuristics. or course. If a ... apor reee\er)
syslem IS Included 10 the flowshcct. the Irade-offs will change
Whenever an energy integration an. lysis (see Chap 8) 15 performed. which
is always In important consldeul1on. Ihe millimum !.IT at the pinch is all
opllrniution ...anable that normally m...olves a Irade-olf bct.. ccn e!l"Changc:r area
(i.e.. capllal COSts) and the uultty requlremcnls operal1ng costs)
U) ReJlux ratio. There IS an opumum reflul rallO for each dlStlllauon column that
balances the Incremcmal number of p4atcs agalOSl the eombmcd costs of the
column dIameter, the condenser and rebc:M.ler costs. and the steam and cooling.
..-ater cosu (see Ellampk 1012). ThlS 15 a urnt opllmlutlon. Ind ..-C note that a
rule 01 thumb 15 aVilllabk rOt csltmalln!- opumum reflul callOS
(g) FrtKfiOtlt1/ rtHKl'rln In tlunlltJIlOfl Sif'lCe only the prodllC1 compos1Uon
01 bcruene 15 SpeCIfied. the fraeltOlial recoveries of belW:ne o'erhcad 10 the
product column and the four sphts 10 lhe srab!IQeT and recycle columns
corrcspond 10 opilmlZalJOn ...anablcs. For cxampk, the fradional rcc:o\"Cty of
belW:M In the product column m...olves the trade-off of mcrernentaltra)'li in the
S1npplOg SCClIon and the cost of recyehllg benzene bltck through the reactor we
COMider Ihese trade-offs to be unit opumlUtions. A rule or thumb of thllll
99% reco\enes IS a\-ailable, but a quick cstimate of the optimum can also be
e.... luated (see Fisher, Doherty. and DoUg/liS).
(10.110)
(IO.I-II)
C"H, + i-C"H
1
-0 jC.H..
C.II. + i-C.B.. -CUH
11i
Eumptt, 10.I.J &-,tn,k I....r-<>/f,. .. lIoe BDA ,..oasI
(a) COIfl:w-siolI Normally it is possible 10 correlate tbe product d15tnbuhon fOl a
process in terms oIlbe conve!'$ion otlhe I1mllln! reaClant (althoup to some Cil5C:!i
the reaclton temperature and the molar ratio 01 react.ants are abo Important). For
a filed belW:ne prodocllOn rate in the BOA process (see FiS-I.Il).2). tbe reactor
0051 and the seleetrVity losses (from lXln...enton of OlXSli toluene feed to diphenyl
by-pcoducl) Increase monotonically as the conversion mcreases, Ho.....e"er. the
cost ot aoo lyeling the uncon"crled lolucnc dc:crea5C$ monotonically
15 tbe con.ersien Increases. The selection 01 the conversion dramatICally affrcts
the Ro.....s In bolh the liqUid and the gas (a 5/1 hydrogen-lo-toluene ratIO lS
required at the reactor mlel) recycle loops and hence Bffrcts the design of each
piece of equipment In the flowshccL Therefote, we classify thlS as a recycle
optimintion problem. There are no rules of thumb that can be used to estimate
the optimum conveT$ion for complel reactions, because the op1imum depends on
the selectivity los5C:!i, which are ...ery different for varioWi reaction systems.
(b) rurfle OOntPOSltlO1l. Wbenever there is I reactlnt and eIther a
impuri1y in a foed stre.m or a by-product formed (.....here implies
havmg a 10llier boihn@ point thao propylene). it is ODn"entlonal 10 use a
req"dc aDd purge stram to remove the nonreactants from the pr0CC$5.. As lhe
reactant composlUon In the purge sr.rum increa'5C:$, the raw-material cost of the
reaet.lLllt will U1Cl'easc molJOtouic:llly. as the: reactant composition in the
purge da:7casc:5. the ps-recyde lIow rite and all lhe costs associated WIth
equipment m lhe psrecyde loop ",i11Increa5C monOlonically 10 an unbounded
value. 1bcrefore. the selection of lbe composItion of tbe hght Teat1anl (hydr08en)
10 the purge stream (or lbe ocess hydroaen feed to the process) corresponds to 1
recycle optimrzatlon. Agam. there are no rules of thumb lI"ailabic for
the oplimum composition
(c) Molar rotio of rtar/or fuds. As tht molar ratio of react.nts (II ,IT) approaches
the stoichiometric requlfement for the reaction, the cost of eqUIpment in the
vapor-recycle loop IS mmimized. To pre...ent coling and the producrion of
undesired by-products. ho....ever, I large UlXSli of hydrogen is required 5/1)
Although the selectIon of this molar ratio corresponds to I recycle optimizatIOn.
this design ...anable is often very dtfficult to incorporate into I process economic
model because: of the unknown coking kinetics. Hence. it is often treated as a
design constraint In order to avoid the oplimization analysis.
(4) PrtBWtof lite jftUIt drwnlUld rrocrlJ' prtuvre. As the pressure oIthe flash drum IS
inc:reasc:d, lhe amount of aromatics IosIIO the purge stream dcae:ases monolonl-
cally. However. as Ihe pressure is increased, the waJlthickness Ind Ihe cost 01 all
the equipmcnl In tbe ps-rttyde loop inaa.ses. 'Therefore.. we dassify tbe
selcction of tbe flash pressure as a recycle optimization.
The pressure of the Rash drum ob...iously is related to lhe teaClor prcssure.
In some cases, ch.nging tbe reaoor pressure may affect the equilibrium conver-
SIOn, the produCi dlStribulion, or tbe phase oItbe reactants. lIenoe. purge losses
arc only one factor that might alfectthe optimum pressure. And the trade-offs will
change if we 1051111 a "apor recovery system
(e) Approach on excltottgen There .re rules of thumb lIv311able for
estimatlDgthe optimum approach temperature In heat e;o;changers. These roles of
thumb are not always ....ltd. however, because the selection of the approach
temrcrature can lO\o1\e '-cry different econormc trade-offs for varIOus units..
326 10 I DESIGN VARJABlES AND ECONOMIC nAnf-OfFS SEcI10N 10/ cosr rna paoa:ss l'NiTS 327
C.. recycle We upect that these OptimlUltlOns Will usually correspond to a global
optimum and that normally the optimum will not be at a constraint (the exception
corresponds to coking constraints) The case-study approach for evaluating the
economic potential that was described in Chaps. 5 through 8 can be used to verify
this behavior, if necessary. The case studIes also indicate the sensltlvity, i C.,
"flatness," of the opumum, whIch is always tnformation lhat we desire
Umitations of tbe Optimiz.luion Analysis
For the purpose of screening alternatives, we are only attempting to get in the
neighborhood of the optimum design conditions. Thus, we use shortcut design and
cost models. We also assume that equipment sizes are continuous and that we are
not in a region where the materials of construction change as we change the reactor
temperature. Our initial goals are to screen out unprofitable processes and/or to
make a first evaluation as to whether a few process alternatives appear to be
sufficiently profitable to warrant an additional design effort.
C
O2
c
E
,
"8
'----1
j

,
i-C..
Reactor ..
C
2
- -
4 I
FIGURE 10.1-3
Simphfied lIowsh""'t for bulanc alkylation 10.2 COST MODELS FOR PROCESS UNITS
(b) RttJl:/or ItmpvtJf ..'c. High temperatures correspond 10 large sdeclivily losses, but
small reactors, ..... tbe opposile is lrue al low lempera1l.lres The rcaclOr
temperalure is a product dislribullon oplimization problem.
(c) Mola, 'Illio afreactants. Large decrease Ihe sclecllvlty losses
but lead 10 large recycle costs of and vice versa. Again .....e oblain a
recycle optimization.
(d) Rejfw. ,alrOS. There is an optimum rel'lw: ratio for each column. This IS a unit
optimization.
(e) F,acl;antll ,ecoveries. There are two optimum fractional recoveries in lhe
column and one in lhe product column (the product composition.is assumed to be
fixed). Even though the fractional recovery of I-C. overhead in the first lower
involves a trade-off between incremental trays in the rectifying section and recycle
of i-C, back through the reactor, we often classify this as a unit optimization
probkm because we ellpectthat the optimum value of the recycle Dow of i-C, will
be quite small (ie., we npect tbal greater than 99% recoveries of i-C. arc
warranted).
Once the material and energy balances have been estimated for the process, we can
usc shortcut design to calculate the equipment sizes. Then we can usc
Guthrie's correlations (see Appendix E.2) to calculate the installed eqUipment cost.
We can put these installed costs on an annualized basis by using a capJlal charge
faclor, say,! yr, and we can calculate the utility costs. Thus, we assume that we have
completed a base-case design.
To mlOlmlZe the amount of computation required for process optlmizalion,
we use variable elimination and the appropriate design equations to write the
annualized capital cost of each (i.e., expensive) piece of equipment
and each operating COSt in terms of the process flow rales. Next we use the
approximate material balances described in Chaps. 5 and 6 to relate all the process
flows to the significant design variables. Several examples of cost models of this
type are presented here.
Hear Exchangers
Guthrie (see Appendix E.2) indicates that the installed cost of a beat exchanger can
be written as
We include the capital charge factor of! yr in the base cost so that all quantities arc
on an annualized basis. The heat-ellchanger area can nomlally be calculated from
the equation
Significant Design Variables
The most significant optimization variables involve product distribution or the
recycle trade-offs. They include all the design variables that affect the process flow
rates (conversion, purge composition, molar ratios of reactants, and possibly the
reactor temperature and pressure). Unfortunately, there are no rules of thumb to
select any of these variables. Thus, an optimization analysis of some type is rquired
to fix the process flow rates.
J A)'."
C.c = CA'liI,\A.c
Q = = VA AT",
(10.2-1)
(10.2-2)
SI!CTlON ICI.2 CO$T MODU reI. ,.l.ocu:I ID<Jl1I 329
For constanl v.lues of C, and U. we caD use Eq 10.2-2 10 eliminate A from Eq
10.2-1 to oblaln
Isothermal Plug Flow Reacior
For a firsl-order isothermal reaCllon In a tubular reactor, the design equation is
..{
Flul'J.T )'"
C,.-C,.
. .1.T",F
ac
Af-e
or. Jf the Siream lemperatures are fixed,
(10.2-3) F I
V=-In--
kp I - x
1be ,"stalk<! cost of thiS reactor can be wriuen as
(102-13)
Thus. we have a simple model for heal-exchanger costs In lenns of the "ows.
(10.2-14)
j F)'"
C,. = C"'\Fw:
(102-4)
CIl ,",
We can relate the COSI of the reaCIor for any con\ersion 10 the cost or the reactor at
base-case condllions as follows'
Heal-Exchanger UtilitK-s
For cooling water We can write
(102-15)
Ccr=C
cr
.
\For.-e
Then from a heat balance we find
(10.2-5)
Furnaces
Available cost models for direct fired heaters relate the inslalled COSIIO Ihe furnace
heat dUly onl) For example,
which relates the cooling-water cost to the "ow and temperatures. Again. for fixed
temperatures
and thus we obtain
(10,2-17)
(102-16)
(10.2-19)
(10.2-20)
(10.2-21)
(F)'"
Cr = Cr'''\F-e (10.2-18) 0'
P
3.OJ)(IO-SPleFI'[(P_)' I]
ower = - -
)' 60p P_,
1 (('. - 1)("
CI" C.
.
F t., )'"
C,.".... C,."..
FIIC AI
IC
(
Q, )'"
en. = C,.".IK Q,..IC
The (sensible) heat dUly for the furnace is
Q,. = FC,dt
so that our cost model becomes
The power required for iseolTopic compression of an ideal-gas stream is
Compressors
The installed cost for. compressor (camp) can be related to the required brake
horsepower (B
bp
= power/efficiency) by
j B )'""

where
(102-6)
(10.2-7)
(10.2-9J
(10_2-8)
(10.2-10)
(10.2-11)
(102-12)
C
W,
Sf... = CSTM 8C --
W
S
.
IC
Q = FC, At = W
s
dH
s
c . Ft., )
ITW SlM. F AI
K K
C
cr
= Ccr.-{r:)
The results for steam are similar:
and
So
or. for filled temperatures,
330 SECIION 107 COST NOOEU FOl UNITS
I( the gas composillon IS constant and lhe inlel and outlet pressures are roughly
collstant for a fixed nowsheet, then the cosl model bec;;omes
Sf{110N 101 COST "'OPEU fOl UNITS JJI
Jr we also wish 10 optimize the reflux ratio, we can use the approximate design
model of Jafarey, Douglas, lind McAvoy
although N should be corrected SO lhal N::::;. IN.. when R R. = 1.2
The iostalled cost of lhe column reboiler aod coodeoser can be wnlleo as
(102-22)
For gas-recycle compressors, both the "apor flow rale and composition may vary
(if the purge composition is opllmlzed,. In thiS case, lhe ratio of heal capacities r
may also be mcluded In the cost model
where
N: --
In 10:/[1
C
RR

10 p
..(;""I )jC;("'R"/ "')
I + ru,-)(RjR.. - x
f
)
(RR. _ 1)2
(10.2-29)
(102-30)
Distillalion Columns
The Installed COSt of a distillation column shell (trays or packing) can be wnllen as

A. )'" V)'" C. = C,. -- = Ca. -

A . at: VJIC
(10.2-31)

N )' "'( Om )"" Cd-CM. - --

N
IC
Dia
K
(10.2-23)

..
Ct; = Cr:. -- '"' Cc:. -
A
c
IC VIC
(102-32)
The column diameter vanes as the square root oftbe column vapor rate, so we can
write
the ope:rallng costs for sleam and cooling water can be wrinen as

N)""( V)....
C.. -C... -
""K YK
The vapor rate III the column IS gl\en b)
(10.2-24)
CSlN -= CSlN -{::)
Ce..;; cell. .c(IE)
(10.2-33)
(102.34)
For reasonably sharp spillS With tbe hght component taken overhead, the dlsullate
flow rate is approximately xFF. If tbe outlet composition and reflux ratio are not
optimized, the number of Irays for tbe requlfed separation is essentially constaDt.
For thiS case, our model becomes
If lhe outlet compositions are optimized but the reflux ratio is fixed (at, say. 1.2
limes the: Immmurn), then the COSI modcllS
Y={R + I)D
c = lin SF )0
SF
IlC
"
K
where the separation factor SF is
SF_( x.
I - x
D
x.
(10.2-25)
(102-26)
(10.2-27)
(102-28)
Thc= vapor rale appearing 10 these CApressions IS by Eq 10.2-25 and the
material balance for a penect spltt The reflull ratio in Eq 10.2-25 can be calculaled
by using Unde,.....ood's equauons or the appro:umations ofGlinos and Malone (see
Appcndu A.2). We can rdate the feed composition 10 these e:l[preSSioDs to the
ute:nt of reactions by uSlOg Simple: male:nal balances.
Total Annual Cost
Once the COSts ha\e been wnncn 10 tcrms of the stre:am flows, we can usc the
simplified material balances illustraled 10 Chaps. 5 and 6 to relale the flows 10 the
design \ariables. Heocc, we can oblam SJmple cost models in tenns of lhe deSign
variables. A model for the total annual COSt of the process is then sllnply the
summallon of the mdlvidual capital and operating costs. We usc these models m
our approximale optimization procedure.
A hluey. J M DoUlll.. and T J McAVOy. Thmques rot D,snllallon Cotumn Deslsn
and Conuol,M l&!.( PH"" LHs Dn. til 121, \1/1U1l11/1
332 10 "COST ro. " SlMrtz ...OCESS
'01 "COST MOOI'.L fOR" SlMnf; nOCF.5$ 333
10.3 A COST MODEL FOR A SIMPLE PROCESS
To illustrate the use of cost models in our approximate optimization analysis. we
consider a particular eumple described by Fisher. Doherty. and Douglas. A
flowsheet for the sImple reaction system
different assumption is used for the column design calculations). flow
of the product stream is P, and Ihc amount of product contained in this stream to
obtain a product purily X
o
IS
(10.3-2)
is shown in Fig. 10.3-1. Component P represents the desired product, and Wis a
waste by-product. The kinellcs of both reactions are first-order ....Ith activation
encrpcs E
1
< E
z
The relative volatilltlCS are such that llA> ll,. > ... and \';e
assume that the direct column sequence is favorable. TIle product stream flow rate
and composition are specified, bUI the composition of the waste stream corre-
sponds to a design optimization variabk. The otber optimization variables we wish
to consider are the reactor conversion and temperature as weU as the reflux ratio
for the product column. Several other design variables are available for tbis
prooess, which we have fixed using rules of thumb to simplify the analysis.
If we let 11',. be the amount or product lost in the bolloms of thc product column,
then the fractional rcco\ery of the product f, is
A-P- IV (IOJ-I)
We also 11K sclectlvily S as
s _ Moles of P in Reactor Outlet
Moles of A Convcrted
(10.3-3)
(103-4)
and we are assuming Ihal no A leaves in cithcr Ihe prodUCI or lhc wastc strcams
Th",
From an overall material balance, we know thatlhe frcsh fced rate of A must be the
sum of the flows of the exit streams
We can comblDC these equations to obtain
(10.3-5)
(10.3-7)
(10.3-6)
(10.3-8)
F
r
.
A
= P -I W
P,. + 11',. P,. + "',.
S= =
P+ IV FT,A
F _ P,. + W,. P,. Px"
1,.1 - S = I,.S - f,.S

Stram Cos15
,nd
J.r--
P
c
E
,
"8
L.--j
J
TL..-_w
Process Flo1'\s and Stream CoslS
We assume that a feed stream containing pure A is available and that all A fed to
the process is recycled to extinction. That is, for the material balance calculations
we assume a perfect split between A and P in the recycle column (although a
FlGURF. IOJ.-I
Flowshoel for the reaction sysll:m A _ p ..... W [F,om W. It FUM" M. F. DaM..)'. aNi J. M. Doo.gku.
AIClIE J., 31. IJ..M' {/98j}]
W JI r,sher. M F Doherty, and J DousJas, AIC" 31: 15J8119SS).
Assume thai tbe slream costs are
Product - (S20/mol) (P molfhr) (8150 hr/yr) (10.3-9)
Fresh Feed = (S15.5O/mol)[(Px
o
lf,.S) mol/hrJ(8150 hr/yr) (10.3-10)
By-product Value - (SI/moJ)[p(>:o/f,.S I) mol/hrJ(8ISO hr/yr)
(IOJ-II)
334 SECflON 10 J A COST l.tOlJEI. fOX A SIlo'lPlE I'.oa:.ss
SEClION 10 I A <.'QST Io'IOOL fO. A SIIo'IPLE ,.oeESS 335
An approxImate expressIOn for the mllllmum reflux ralio is given by Glinos and
Malone- 8S
Stlecli"ily and Reaclor Model
We assume Ihat the rate expressions for firsl-order reactIOns In an isolhermal
tubular reaClor are
(10.3.12)
..... here
xr.=I-:t Xr I' ""xS X, '" = (I - S)x
I) (10.3-19)
(10.3-20)
where
From Sec. 10.2. Ihe appropnate cost model for thIs case (we use a fixed number of
trays) is
(10.3-13)
(
-52,500)
k
l
_46lxlO
ll
exp RT hr-
1
The column vapor rate IS calculaled by
(10.3-21)
If ..... e solve the rate equations., \'t"C find Ihat
I ( k ) S = - --'- (I - x)">JI. - (I - x)]
x 1.
1
- 1.
1
From a recycle balance .... e obtam
F,

x
and .....e assume Ihat the reaClor COSI IS gIVen by
(10.3-14)
(10.3.15)
(10.316)
(10.3-17)
II : (I.2R. + I)D
= (I.2R. + 1{(I-:)F
r
.
A
]
1lIc recycle column condenser and reboiler caplIal COSIS arc given by
-{
I' )' "
Co = Co _
LK
The associated opera ling cosls are, for cooling water and steam.
(10322)
(I0.3-23)
(10.324)
(103.25)
(10.3-26)
Proclucl Colunlll. The dcsllcd ceononuc model for the product column shell is also of
the foml
Recycle Column
We arc not mlcresled jll opllllll;ring the design of the recycle column (the rCnllX
ratio and fractIonal recoveries) in this case study, but we want to include its cost in
tile economic model We that the design reflux ratio is 1.2 tllllCS lhe
minimum and lhat lhe theoretical number of trays is about twice the l111nimUI11
(
N )""( V )' '"
... C.hJ.Jt< N I' ]
lie l,K
(10.3-27)
2 lD SF
No-=2N=
.. EolnaAI'
(103-18)
K OlmoS:llnd M ,. P'oc /Hs ()ft).2J 764 (l91l4)
SfCT1Of<t IOJ ... COS'! l<otOO!. fOIl "'UU.5S 337
In thIs casc, we use the desIgn model of Jafarey, Douglas, and McAvoy' to esllmate
the optimum renux ralio and the product recovery fraction Thus, we lei
TABU IOJ-1
Oasc-<ase calculations
where:
N In /IS
-In{o:/[I + (0: I)/(R/R..)]O S}
5F-( Xo
1 - \"8 X.
P= (R/R.. - xFXR/R.. -= I -t llXF)
(RIR.. _ 1)1
(A 2-47)
(10.3-28)
(A.2-48)
UI 1'_100 mol"',. ",.-om, " .. 08. r..-WF.I._099.5 RR -12..nd 0 .. 08 mo",'
From Eq IOJ-H, '.. ..
0390 hr '

',r
"
,- "'0 up --003789hl
I 987(460 + 90)
F,om F...q 100J.I
The condenser. reboiler, cooling-waler, and steam costs, respecti,-ely, are gh'en by
The dIstillate composition for lhe product column is filled. and the bottoms
composItion is calculated from
08)J .. 0901
11061
011)0 "' ..... _ (I
F.e.h Fttd _ r _ 100(0 \199)
'A 0\195(0907)
WUleFlo'"' .. 110.67 100 .. 1061
Prod. Val..., .. 11_5(IOOXUSO) _ S9 10"/)'1"
Feed .. (! 5(11061X81.50) _ S1.666 x 1000jyr
8)'-p.odUC1 V.lue .. 1(l067XSI.50) .. SOOll7 10"1yr
110..61
Flo_ to RQCI D8 J
O.
, l)SJ 1
Rae!. \ oIumt - --_" It14 fl'
0)9(0.1) I - 08 '
.. 6 - -
Dx - {(WxXl1Hlj" .. 5J2 fl
- 6(5.J2) .. JI9 fl
'( OJ'" )
s .. oil 0J90 003789 [(I
r.odUCI no... 1' .. 100
Eq 10.3-7-
Eq IQ.H
Eq IOJ9
Eq IOJIO-
Eq lQ.J-1I
Eq IOJ16
Eq 1OJ-15
Reat'lor CO$l-Quthnt eoue.lallon for a pressure "e$$d
(10-3-19)
(10.3-31)
(10.3-32)
(10.3-30)
(10.3-34)
(10.3-33)
(I - fr)XD,rS
x., = -
, x
o
., IrS
Similarly, the vapor rate becomes
1'1'" (R + I)D
_[ R/R", + I]P
(;7"" I)S
""WIlt rcrovtr 0\19.5 of A ovtrne.d and 0991 of P In tlK bot 10m- .. 0'. _'. '
Me _ 60 .., 0 ' A' .... - ..
Summary
A base-case design and a sct of cost cakulations are presented in Table 10.3- I. The
cost functions are developed from tbe basc-case conditions x = 0.8, r. = 9<rF.
I, - 0.995, and R _ 1.2R",. We usc this process cost modd to describe: our
simplified optimil..ation procedure.
f.q IOJI8 N,
I. I. 09950991
Sepuauon Faaor - -- - .. 66.1J.(
t -1.1-1. Q.Ol).5 0 OOJ
21n66.Il<l
In] -641
I A J M T J. McAvoy. /dEC Proc Dtj III 121, 197 (1979)
f.q 0'\3-2 Towullc,pt If, 2J", .. I.IJ
Eq rOJ20"JA_' '" I 011-02 K,.-..-S .. 08(0907)_0726
"', .. (1 S)x .. (1-0907)08 00144
338 SEOIO.. 11IJ "con ),lOau. Fa.: " SIMPLE PIO',:fil
TABLE 10.3-1
Ba.se-c:ase calcul.,iOllS (COI1t1I1IU'J)
St:CTlO.. 101 .. COST loIlOOt:l 10.... PIOCUl; 339
TA8U 10.3-1
BaSot-<ast alculalions (contll1ui'd)
X .. OW9(IOU)
"'0999(100) t 1067.
0903
0999 I 00469
Eq 10319 ... 0001 0(1469 -10.300
(
2 + onil) 00744
Eq 103-19.R.,_ 4 2}.., 02 "02(4_0"'OS
R," 570
FAI -.I) 11067(1-081
Re<:ydc now 01 A .. -''';;-' _ _ 27 67 .. D
x 08
Eq 103-22 v, - (57 -t 1)2767 .. 185 S1lIOI-'ll(
Eq A 3-1S DwDc:lCf'" OOI64(185.S)'l ... 2.15
APJN:lldlJ' 0.2
Cost O/Columll .. G:o'1019
X
2.l5)"N'(147J)I _ 3;.
= S79,7(0)r
Eq 11.1-48
Eq 11.2-23
(R,R.-.,)(R R.- 1 .,.H,I
ft.. t
RR
I)!
(1.2 I .... 2(090)))
(1.2 _ I)' - 1489
1 In ,l(Sf) 12.61
".-
O-S"'ujl+(2 1)'12] 05(019)
2 In SF In 2O,JOO

0.5 In.. ln2
COfTecllon faclor for mIn, Eq
COIwlcn$C' A5SlJmc lJ./1 A- 13,JOO BIll/mol, ... 14,400 DIu/mol, and ToA ... I SO' F 11Ien
120 - 90
I1T
L
... 90)/(150 12Q11" 412
Qc.I3, I, .13,)00(185,5).2467 .. 10- Blu!b,
Q.. H67" 10-
Ana U 11T_ 100(432) ... 571 fl'
5123

". - O1l844(0467) _ 5123
Eq AJ.2 _ 2.3(S123) _ 1316
, ,
R
8
- (-'-_-'-'-" - - I 107
Eq A 3d!
Ii 2 ClK'. 3X571.0,(3D. 519, ')r
(
"' .. X' ,., X",300\,
CooIWalcrCosl_ 100011
1
8341b ---yj"",185.S)8ISO
.. SJ:!OOlyf
Eq IOJ-22
Eq.AJ-IS
(
R'
R- RJR_. I 2'(1.101) - 132
I:' .. tR - I)D _ (132 .... 1)100 _ 232.9
[
200 + <00]'"
- 0.0I64Jnl.9 379(60) S20 -) 24 [,
AppcDdu D 2
Eq AJ-13
Appcndu E 2
Eq AJ2S,
Q .2.671 " 10" Blulbr
Q. 2611 .. 10"
A". 11.15U -!ITso .231 fl'
COlI. G:oJIOI 3)(237"
0
1 .. Slll.9OOIyr
(
52 OJ X14.<oo\,
Sleam Cost_ IOOOlb 933 tIBSS)8ISO - S65,JOO/yr
Column Cost _ _{1;')
- i142.SOO 11
CondcllJef Aoau....-I11I, _ 14..00 aDd 11f1 15.SOQ Btu/mol.
120 - 90
1I To, - In ({100 _ 9ON'(200 120)) - 94 2"F
Qc" 14,400(2329) - 3.3S4 " 10- BI""'.
J JS4 " 10"
A,_ 100(942) _3S6fl'
.. ... Z. 1111 .. - 15,SOO Blu:'mol, 1;., _1OO'"F, .. 60, aDd 0999 ThUli
Eq 10331
(1- 0995)(0999),0907)
XI.' "0"999 _
('onl JI(3S6"-f;) $73,200')1

340 R<:TlON 10+ ""'.OXI .. "'TI! OPTI .. lz,ol'lON "'N"'LYSIS
SE<:TIOl'lIO+ ", aOXI .....Tt OI"TIMlZAfiOl'l .. N"'LYSIS 341
'"'
UIII CoM .. Coolanl I +Slea... 1 .. 1 .. SICIllI'I1
EJor:ss loral cou. _ IIxfudm, 1M 'Ioochoomelnc reed 'equlremenl
Dominant Trade-offs ror Each Design Variable
After we develop a base-case design. we can change the slightly and
Ihen calculale the incremental cost for each item in Table 10.3-1 divided by the
incremental change that we made in the These resulls are shown in
Table 10,4--1.
In the first column ofthist3ble, we see Ihat as we increase thecon1fersion. we
produce more by-product (the wasle cost decreases because Ihe fuel \alue
increases). but .....e are required to supply more reactant. The reactor cost increases.
bUI the costs of the req"CIe column decrease because we recycle less reactan!. The
product column (;()Sts increase because the rdlu. rallo must be IDcrea$ed (smce ..... e
are feeding morc by-product to the column).
Howc\cr. when .....e compare the positi\e costs in column I, \\C note that the
feed cost is much more important than the reactor cost and the costs of the product
column. We use our trick of neglecting all costs that are ao order of magnllude
smaller than the largest cost, which indicales that wc can neglect the efTect of
changes in coD\'ersion on the reactor and lhe product column costs. When we
compare tbe negati\'C costs. W'e see that the cost$ ohlle column shell, the condenser.
and the steam in the recycle column are Important lienee. in subsequent
optImization calculalions we can neglect all but lhese Iargcsl-cost lerms.
Now if we return to the base-case condition. change the temperature slightly.
and then calculate the incremental cost of each Item in Table J04-1 dl\'ided lhe
temperature change that we made. lOoe obtain Ihe resuhs sho\lo':1 in the second
column of Table 10.4--1. The ",crease in temperature causes more byproduci to be
formed. so we require more feed to make our desired amount or product. bUI v.e
obtain a fuel credit for the b)-product (i.e.. the waste cost decreases). The reactor
no n' A.
,
P,.
boom Feed .. 8 --
J.S
)610 ... I<t
11.2 J
Cost J)(J2Q). Of:') _IIO.900,y.
Slam CO$l - (::X
1
:;j2J2.9X81SO) - S8II,JOO,)'T
Tot. up. Cost .. React. .. Col.1 .. Cond. 1 .. Reb. I + Col 2 .. COP<ll + Reb. 2
- 29.100" 79:700 of- of- 11,000 .. 142.500 ... 7J.200 + 10.900 _ $.365.9OO,yr
3100 +- 65.100 +- + 88, _ SI61,2O'l/yr
ProM_ Prod Feed +- By.pradUC1 Tot. Cap. - Ulil
8.H - 7.666.100 .. 86.'100 - J6s.,900 - 161,.200
Q, 10" Blulbr
TAC .. EJ.or:ss Feed By producl .. TO! up +- TOL VIIL
TABLE: 10.3-1
c.kul.lions (C(JnflnllCd)
TAC _ 7)9)1(10 86,900 +- 3M.900 .. 161.200 - Sl,ISO,OOO,yr
10.4 APPROXIMATE OPTIMIZATION ANALYSIS
In a conventional optimization analysis. the gradient for each design variable is
equal 10 zero. For our simple example .....e would require lhat
OTAC OTAC OTAC OTAC 0 (10.4-1)
ax-:: ilTJII - ilJ,. .0= iJ(RIR_) =
Ho.....ever, for scrccmng c.,lculallons we prefer to simpliry the analysis. I n particular.
we would like to identify Ihe dommant Irade-ofTs for each design variable, lhe 1I10si
important design variables, and the mcenlive for optimi7.alioll. We diSCUSS each
below,
TABLE 10...... 1
Gndienls
lTAC 0"
<TAC <TAC
,0'
<TAC
-.-0' --0 ,0' --0 ---0'0'
TAC
"
ar.
'f.
iJ(R/It.J
..= .... 2.8139 001.9 -16916 0
By-produel -O.JJI -00019 +09Qj6 0

O.lHl -0.0009 -00112 0
Column I -0.011 0 -0.0426 0
CoDtkmeI I -0.023 0 -0.0126 0
CooIanl I 0 -0.00)1 0
Reboiler I -OOIl
0000' 0.0011 0
Sleam I -0119 0000' -0.06j6 0
Column 2
0068' 00006 26261 -0.1166
Condensel 2 0.0001 00419 0.0J41
Coolant 2 0.001 0 00021
Reboiler 2 0001 0 00010 O.tX133
Slum 1 00)]2 OOOOl 0Q909 001S4
50 that we use a scale factor of 0.3 Similarly, we expect the optImum fractional
reco\'ery will be 10 the range
Rank-ordering ,he Design Variables-The Most
Important Design Variables
There IS no ...."ay of companng the vanous columns m Table 104-1 because the}
ha\e dIfferent units, i.e.., the first column is in (S/yr)/conversion .... hereas the second
IS ID lS/}"f)/"F. So .... e .... ould like to find some way of pUlling each olthe calculations
on the same basis. To do this, .... e IOtroduce scale factors, where the scale factor for
each design variable is the m.U.lffium range for that \'ariable. For example, .... e
expecl that the 0pllmum ,'alue of R/R. will be III the range
so ....e use a scale factor of 0.01 The optimum con\ersion will be greater than zero
and may bc bounded abo'e by lhe equilibrium conversion, the conversiOn
corresponding to the maximum yield, or x = I. Depending on how much informa-
tion we have or how much dforl we are willing to we can select an
appropriate range and a scale factor, We ask the chemist for a best guess of the
range of temperatures (or pressures, if applicable) that are reasonabk 10 consider,
and we use this rllnge for the scale factor
Table 10.4-2 shows the results for the dominant cost terms when ....e add the
absoluu values of all the terms III each column and then muluply the sums bt the
(10.4--4)
l
iITACI
' ... ". - AJ
I'-oy.'
Fisher, Doherty and Douglas called thiS result the ram..-
SECOON 10. ...""-OXI ..... IE o..-n..u..o.noJol ....' ...USIS 343
TABU to..4-}
Scakd gradients
Scak fan..
10 60
'"
'J
iJTAC i'TA-C enc
,R)
--0.
H. oJ,
"
". 'I.
R.
Fd 2.S 139 14982 00769

By-pnldUCf -0-3]10 -0116] 000> 0

0.0412 -O.0>4S

Col_I -0.011iO 0.()()41
-0""" 0
,,-,,-, I
-00U2 0.001<4 -0.0001 0
S'eaJD t -0 WlO7> -0.0006

CoI_2 0.0681 00360 00261 -o.Olj()

"-"-,1 0.0366

0.0104
Rcboikr 2 O.llOU oOln

OOOtO
OOB] 0.0177 -0.0009 0.0226
., 1.1i07 o116 0.069
P,
0678 0386
0<,,,,
appropnate scale factors
order parOtI'WUr
This rank-order parameter IOdicates whether we are trading large positive inc;n:-
mental COSts for large negatIve incremental costs or $01all posllne incremental
costs for small negallve !Dcremental costs. The units of each of the rank-order
parameters are S/yr,
The rank-order parameters in Table 10.4---2 indicate that the opllmlzauon of
the reactor converSIOn and the temperature arc an order of magnitude more
important than the opumiutlon of the fracuonal recovery, which IS about an order
of magnilude more Important lhan the optimization of the reflux ratio in the
produci column The results Imply that we will pay only a small penally if we
neglect the optimizallon of the reOux ralio. They also Imply that .... e caD obtain
reasonableCSllmates if we Simply optimize the reactor conversion and temperature
-and then, if we have lime available, optimize lhe fracllonal recovery. Again, we
find that can simphfy the problem.
...... R FIliMt-M f' Dobeuy,.llndJ JI IH8(l98S)
(10.4-2)
(10_4-3)
R

R.
0.99 < J, < 1.0
cost decreases., the cost of the recycle column IllCreases (I.e., we need a larger rdlu).
rallo because the reactant IS more dilute), and the cost of the product column
Increases (i.e., the product is more dilute)
Examining the positi\e \'alues in column 2, we see that the effects of
temperature changes on both of the column costs arc negligible, except for the shdl
of the rt:Cydc column (the fc=cd composition decreases so that more trays
required). Also, the change in feed cOSt is fairly small The fuel credit of the wa:.te
stream and the reactor cost arc bl>th Important. Thus, only three (or four) co!>t
terms need to be considered for temperature changes.
Similar effects arc obscned for changes in the fractional recovery of the
product overhead in the product column and for changes in the reflux ralio In the
produci column (see the last tlo1'O columns of Table JO 4-1). Using Ihe same order-
of-magnitude arguments for the aDd negatne terms separately, we find
that we do not need to calculate all the procesSing COSts. HenlX, we can sIgnificantly
simplify an optimization anal)"sl5 by considering only tbe dominant COSIS in each
trade-of[ In many cases we can ehmmate 75 y. of the calculations, although for the
sake of illustration we retain most of the marginal terms in subsequent calculations
344 SECllON 10.. APPROXIMATE OPTIMIZATIoN ANALYSIS 10_ APPROXlNATE OI"T1MIZATlO,. A,.ALYSIS 345
'-
--
Proximity Parameter-Incentive for Optimization
Initially we said that the criterion for an optimum is that the gradient of each
design variable be equal to zero, but we have not used tbis information yel. The
gradient is simply the sum of all the terms in each column of Table 10.4-2. By
inspection we sec that the positive terms might balance the negative terms for the
renUl\: ratio, but the positive and negative terms are far out of balance for
conversion and reactor temperature. Hence, we ClIpect Ihat our base-case value of
reAull ratio is close to the optimum. but the base-.case conversion and temperature
arc not.
To develop a quantity which characterizes the incentive for optimization,
Fisher. Doherty and Douglas defined a proximity parameter as
1,200,000
1.000.0Cl0
800,000
;. 000
"" 600,
400,000
200,000
o
0.2 0.3 0.4 0.5 06 0.7 0.8 0.9
I(L OTAC/oy)/6y,1
Pi = L IfffAC/oyNAYJ
(10.4-5)
flGURE Ill.A-u
TOll.l excess COSl venus 000,..r5100 x.
Conversion .x
(10.4-6)
This el\:pression is merely Ihe absolute value of the scaled gradient divided by the
summation of the absolute values of the scaled components of the gradient. The
proximity parameter is equal to zero al the optimum because the gradient is equal
to zero. As we move far away from Ihe optimum, the proximity parameter usually
approaches unity because the total cost becomes asymptotic to either the total
positive or the total negative cost components of the gradient (.see Fig. 10.4-1).
The proximity parameters arc shown in Table I for the design.
Figures 10.4-2 through J0.4-5 show plots of the total annual cost and the proximity
parameters for each of the design variables. From these graphs we note that we are
in the region where the optimum is fairly Aat whenever the proximity parameter is
less than 0.1 Hence, Fisher. Doherty, and Douglas proposed a heuristic for initial
screenings;
Whenever the proximity parameter is less than 0.3, there is
littk incentive to optimize.
Of course, unless each of the significant design variables has a proximity parameter
less than 0.3, .....e must optimize all the variables. Also, as we proceed toward a final
design, .....r might require that the proximity parameters be less than 0.1 or 0.1.
From Table JO.4-2 we see that the proximity parameter for reflux ratio is very
small, but Done of the other variables are close 10 Iheir optimum values.
,000
---
.000
.000
,000
....
,000
---
...Cap.
.000
-t> Oper.
.000 ... Tot.
.000
.000
0
"'--
"-
/
/
"-
"-
/
I\.. V
"
/'
1.000
900
800
700
600
"
;';500
400
300
200
100
20 30 40 50 60 70 80 90
0.90
0.80
g0.70
0.60
0.50
0.40
E
.;;: 0.30
0.20
0.10
0.00
0.1 02 0.3 0.4 0.5 0.6 0.7 0.8 0.9
F"IGURt: 10.4-1
eMl behaVIOl
Temperature. OF
tlGURF. IQ.4... l/l
Proumity paramele. x
Conversion .x
346 ncnoN Ill' APPtOXIMA1EOpnMlZAnON A"'Al'rSlS
SEClION 10' Ol'lIMJZAnON ANALYSIS 347
"'"
h::
\
\
\
"-.....
----
1,400,000
1,200,000
1,000,000
... 800,000
5:
600,000
400,000
200,000
0,
20 35 50 65
Temperlilure, of
80
9'
8OO,CXX>
795,000
190,000

<' 785,000

780,000
775,000
770,000
1.02 1.07 1.12 1.17 1.22
RIR...
1.27 1.32 1.37
FIGURE IO.4-lA
TOlal ucess cost versus lempo::lalure
FI G UR 10.4-4<>
Tolal uocss COil versus RI
Calegorizing Design Problems
The rank-order parameters and the proximity parameters for our example are
given in Table 10.42. We can categorize each design variable as follows:
I. Conversion-important design variable, rar from the optimum
2. Temperature-important design vanable, far from the optimum
3. Fractional important design variableS, fairly close to Ihe
optimum
4. Reflux ratio-unimportant design variable, close to the optimum
1.37 1.32 1.27 1.22
RIR",
1.17 1.12 1.07
8
7 \
6 \
, \
4

3
'"
----
2
r--.... ../"
I
'"

0
1.02
Comparisons
To illuslrate Ihe errors involved in an approximale optimization analySIS, we
compared the results of several case studies. In our simphfied model, we deleted Ihe
COS! of the condenser in column I and the cost of the reboiler in column 2 from Ihe
previously simplified list given in Table 10.4-2. Thus, we considered only the eight
most expcosl\'e pieces of equipment as representative of the dominant costs (see
Table 10.4-3).
Table 10.4-4 gives the results for the base-case design, and case I is Ihe result
for a rigorous optimIZation when all the costs arc included in the analysis. We note
that the optimization leads to a 73.3 %reduction in the costs. In case 2 we minimize
o.
o.
:; o.
U
E
O.

o.
'"
E
o.
'0

o.
o.
95 80
6' '0
Temperature, OF
35 20

"
......
"-
/'
/'
"\ /
"\ /
'\ 7
17
0.90
0.80
0.70
0.60
"
[ 0.50

0_30
. 0.20
0.10
0.00
5
I-IGURE 10.4-).
parameler, lempo::ralure
flGUkE 10.4-46
P,o..im,ly parameter
348 SECflOI'l 10' ",,,,..O"(IMATE Ol'TlMtZATlOtf "'''''''LYSIS
770,000
0.992 0.993 0.994 0.995 0.996 0.997 0.998 0.999
Fractional recO"ery
800.000
79S,OOO
790.000
&- 78HXXJ
780,l.XXJ
ns,ooo
........
--
f- ........
"-
I'---.
'------..
TABLE IOA-J
Domin.nl cosis-sc:altd gr.dtenls
mc mc n"AC
n"AC -n
,.
_.
__or.
8,'
-f.
i'{1I:1I:.J 11:.. <T.
,...
211'9 14988 -00169 0
By I""<>dUCl OHIO 01163 00'l90 0
~ ~ 00412
0"""
0 0
C04umll I 00'10 o.ron 0 0
Sturn I _0119' 00012 -0.... 0
Column 2 0068'
..'"
00263 -O.OJSO
Colllkn"", 2 OOIH 0.0082 0 00104
Sleam 2 00H2 0.0611 -0()()l)9 0.0226
~
"
..,
0.99$
ill
'"1.29'
Cue I .. ",.,..OU. Ofl"If\lQhOl' ">clud,,,O,t>c 000"
Cax 2 .. opUnuu.loon "'.- dom"'"n....... 10. oJl.bc cl<slll' ...... bI...
c... J .. opum """ 01 "., domlnln. cos," w"h R/R. .. 1.2
C&x"" optim """ 01 ,"" cIom'DlInt OM" ..."h RjR.... 1.2 Ind I." 0.991
TABLE 10......
Results of optimi1..1'ioo sludits
Deip mllR
..-
0-1
e-, e-,

o. 0.62 0" 0"

T. F
,.
" " " RIR..
"
I lIS 1.19
"
f.
0.99$ 0.... 0'"
0....
Domananl CDSI.5 " 10'
'I'
m
TAe" 10"
I 19j m m m
6TAC, %
"
0 000'
0006
10.5 SUMMAR\', EXERCISES, AND
NOMENCLATURE

8
7
6
5
/
4
J
r----.... /
""-
1/
2
1
0
o
j O.
"
It 0,
"
'E 0
~ O.
o.
o.
FIGURE IO'...S.
To...l ucas cosl ~ n t l f...cfoonal ,ceo......,.
o
o
only l ~ dominant cost terms (see Table IO.4-3) for the four design variables, and
then we add the costs for the other items. In cases 3 and 4 we repeat the
optimization for the dominant costs, but fix the values of the least importanl
variables at the values given by the common design heuristics. We see that the
errors introduced by considering only the dominant cost terms or by fi:a:illg the
values of the design variables with small rank-order functions are quite small Thus,
al the conceptual stage of a process design. we usc the rank-order funclion as a
gUldclme for delermlnlng ho"" man) desIgn variables we need to consider
nGUIl:f. ItA-56
ProWlUI,. parameter, frK:l..,.w rccoCf}"
(10.5-1)
0.992 Q.993 0.994 0.995 0.996 0.997
Fractional ~ r y
0.998 0.999
Summar)'
NonnaUy only a rc", cost terms dominate the economic trade-offs for each
optimization variable.. 8y identifying these dominant trade-offs. we can greatly
simplify an optimization analysis.
We can further simplify the oplimi7..3lion analysis by determining the rank-
order function for eaeh design variable Y"
I
OTACI
r J - L oYj liy;
where the dYI arc scale factors based lR the range: of a dcslgn variable within which
we expecl 10 obser\'c the optimum We negled to opllmi1.e any design variables
whose rank-order function5 are an order of magnitude smaller than the largesl
SECIION 101 NOMEIliCl....-ru.l 351
Exercises
E.Ilpcncnc(; indicates that If P
J
< 0 3....... are in the region where the optmlUm 1$
rdau\-dy flat. although lIoe can requIre J. lighter tolerance.
ranLordc:r functIon Normally. the !>Igllllicant design are those thaI affttt
the product dl)lnbutioll or the Ilow).
We COIn esllmate whether an mllial guess of a design variable IS close tt> ItS
optimum value by calculatlllg a pro.lllllllly parameter
10,5.1. For any process des-Ign that you h;)\-c conSIdered, calculate the nanL+ocder fUllClions
and the prollmll} paramelers. U!oC the rC'ults to Idcnufy the dominant economIC
lrade-offs DISCUSS lhese:: opllmlZ3u,.n probkms.
105-2. Consider lhe eumplc given In St..... 10]. acepl coMlder t ..o tirSt-ordet-.
m5tQd of conSCCUII\c ,e.RIJOns. (Also. oeglect lhe opumlZauon of tbe
rd'iuJ. rallO and tbe fractIOnal r.,.:o...1") ID bOlh columns.) ulculale lhe rank-order
furw;:lIons and the prOJ.unlly pana.nclcn. and find the Oplimum design con<htlOO$
PIOllhc proli/mll} paRlll<:ICn Ihe de$lgn vanablcs.
105-3.. ConSIder lhe o:ample III Sec Ill.i. bul consider Ibe mdlrect column sequence
rather than the direct sequencc. (Nc):........tthc opllminlllOn of the fractIOnal recovef)
and the rellul ratlos in bolh column._ J Calculate the ranl-order funcllons and the
prnJ.lmll} puamelers. Fmd the orumum design oondllions 110'" do the resulu for
the dlrcci and lhe Indlrecl !M;jeucn.:o '-umpare" What do the beunsllcs
indICate al lbe opllmum 110"5 (Of c." h ca..e?
105-4. For the cyclohCJ.ane pr(l(:eSS de>oenhn.lln Eaerases 5.4--7 and 6.86. how many design
vaflilbks are encountered ill Ie,-el ]" Cakulate the ran.k-order funeuon to delermme
Ihe relall'e Importantt of "lnables'! Also. esltmate the optimum desIgn
conditions al level J
Co)t of steam and base-case value (i/yr)
SpeCIfic heat at .. olume [Blu/(mol "!-1]
Annualized reactor CO)t [$/(fl) yr)J
Cost coeffiCIents
DIstIllate flow rate (molthr)
Overall pl.lte elliClency
Fra<:tion recover)' of the hghl.l..ey o\'c:rhc:ad
Flo.... rate (mollhr)
Coohng-water flo..... ratc and basc--case value (mol hTl
Fresh feed rate (mol/hr)
Flo..... to reaClor (molthr)
Feed rate to compressor (molthr)
HClght of column sump and vapor dlsengagtng space (ft)
ReaCHon rale conSlant (hr I)
Reaction rale constants (hr -I)
Molecular ...... eight of dlsllll.ale
Marshall and s.... lrt Index (see Cklrl/col
umber of Irays
umber of Iheorellcallrays
Proximity parameter
Production rale (molthr)
Inkt and outlet pressures for a gas compressor (ps:ta)
Flow or desired product (molfhr)
Column pressure (psia)
Heat dUI)' (Btu'hr)
Furnace heal duty and base-case value (Blu'hr)
Rank-order function (S,'yr)
Reflux ralio
Recycle flow (mollhr)
Minimum reflux ratio
Seleclivity
Separation factor
Total annual cost (S/yr)
Dlstillale lemperature ("F)
Overall heat-transfer coefficient [Btu/(hr ftJ oF)]
Condenser overall heaHransfer coefficient [Btu/(hr ft
l
, oF)]
Vapor ratc (moljhr)
Reactor volume (fl
l
)
Total flow of waste stream (moljhr)
Amount of product in the wastc stream (mol/hr)
flow rale of steam and basecase value Ob/hr)
Conversion and base-case value
DIstIllate composition
Feed compoSition of light key
Design variable
p.... p-
P,
P,
Q
Q,. Q, K
"
R
R,
R.
S
SF
TAC
T,
U
U,
V
V,
W
W.
IYs Ws,.e
x,x"'"
x.
x,
Y,
CSTN ' CS-Ul flC
C,
C..
Ct. C
l
C,
D
,
I,
F
few, Fe,. Ie
F,
f,
F,
fI,
k
k
l
> J..
l
M,
M&S
N
N,
P,
P
(105-2)
IL lTfAC/oyJI
PI = L ,.TfACICY/I
lIeatoCxchanger ar.J. J.nd base--case value (ft l)
Brale horsepo.....er "f compressor and basc--case value (hI')
ConcentratIons of ... ,'mponeots A and P (mol/ft')
AnnualIZed COSI of heat exchanger and base--case value (S/yr)
Cost of condenser ;1nd base--case value (S/yr)
COSt of compressor and basc--casc value (S/yr)
Cost of COOling .....ater and base--casc value (S/yr)
Raw-material cost lS- mol)
Cost of furnace and rose-case value (S/yr)
Jleal capacily rBw lnlOl F)]
Cost of reaClor ($ rl
Cost of reoolkr anJ h.1se-case value (S/yr)
COSI of dlstlllatlon ....'lumn shell and base-case (S/yr)
Nomenclalure
A, A.c
B..p B..
p
.
1IC
C.t> C,
CA , C.....
IJC
Cc C
C

K
C-.p C._p,1IC
Celt" Cew. 1C
C
f
Ct'/i' C,N. 1Ie
C,
CR, C.flC
C GR IIC
C.... C."M:
G.1Pd; 'lm....

..
6.T
u
O'

a
P
7
P
Heat of vaporu.atlon of IIgbt key (Btu/mol)
Temperature change CF)
log-mean temperature driving force (eF)
Heat of vaporization of steam (Btufib)
Relative volatility
Stt Eq 9.2)0
(C,IC" - I)/(C,/C,,)
Density (moJjft
J
)
CHAPTER
11
PROCESS
RETROFITS
In aU previous discussions we assumed that we are designing a new process.
However. euctly the same techmques are useful for retrofitting a process. By
retrofiuin& we normally mean making minor changes in tbe interconnections
between process equipment, the replacemenl of one or more pieces of equipment by
some other equipment. or the change in the sizes of one or more picas of
equipment in an existing process. The first type of change involves structural
modifications or the flowsheet, whereas with the last two types of change the
60wsheet remains the same.
Some examples of where we might want to retrofit a process arc to increase
the production capacity (debollieneck: a process). efficiently process new raw-
material feedstocks when they arc cheaper. utilize new process technologies, or
reduce operating costs, because the optimum operation conditions have changed
SIllCC the plant was originally buill (This changing economic environment IS the
reason Ihat most companies did not usually allempt to optimi7.e in the
pas!.) The last goal is valid fOi a large number of eXIsting plants, and therefore
there is a great interest III retrofit proocdurcs We describe a systematic procedure
for process retrofits in this chapter
'53
354 SEcnON II I ... SYSTEIotAnC PROCEDURE FOl "lOCUS u:rROFlTS
11.1 A SYSTEMATIC PROCEDURE FOR
PROCESS RETROFITS
The systematic procedure for process retrofits discussed here was deveolped .bY
Fisher, Doherty, and Douglas.- The analysis is limited [0 single-product, conlin
uous processes, which are [he same restrictions of our other design methods.
Moreover, many of the steps in the proe<:dure are identical to the steps that we have
completed previously. . .
The retrofit procedure also proceeds through a hierarchy of decISIons:
I. Estimate an upper bound on the incentive for retrofitting.
2. Estimate the economic incentive for replacing the existing plant by using the
same process f1owsheet.
3. Estimate the economic incentive for replacing tbe existing plant with a better
process alternative.
4. Estimate the incremental im'estment costs and savings in operating costs
associated with changing the existing process.
S. Refine the retrofit calculations.
aearly, the first four steps in this procedure focus on the same of screening
calculations that we considered earlier, and for this reason .....e descnbe the retrofit
procedure before we consider the use of Simulators to refine design or retrofit
calculations.
We discuss each of the le... e1s in the hierarchy above in more detail below.
Estimating an Upper Bound of the Incenti\'e for
Retrofitting
Our goal is to reduce the operating cost of an existing process. and so. our first step
is to prepare aD operating cost diagram. That is, we prepare a cost of the
type described in Chap 9, except that we do not include any of tbe capital co.sts.
From this operating cost diagram, we can see the total costs of the raw matenals
lost in the fonn of by-products or lost in waste streams. Similarly, we can see all the
energy costs that are supplied to tbe process. An example ofa diagram oftbis type
for the HDA process was given in Fig. 8.10-8.
For a large number of processes, such as the HDA process, tbe of the
raw-material losses will exceed the energy costs, which means that we will need to
examine the possibility of changing the process flows. If we change the Bows, we
will also need 10 change the heal-cxchanger network. Thus, nonnally we must
consider the behavior of the total process.
W. R. Fisher. M F Doheny lnd J M Dou&las, KScrOCIl.lD& of Proa:s.s Reuo61 Ahemallvcs,w
I&C Rest"'.,,,, in press
sEeno,," III ... SYSTEIotATlC PROCEDIJlE rOl PRC1CES5 355
Estimate the Economic Incentive for Replacing the
Exisling Planl
One of our options is to completely replace Ihe ellsling plan!. ThIS optIon WIll
correspond to the largest capital investmenL Also, if we do not change the process,
this calculalion will indicate thc profitability position of one of our competitors if
they built a process just like ours.
We call use the hierarchical decision procedure descnbed 10 Chaps. 5 through
8 to estimate the processmg costs. With the shorlcut ma[erial and energy balances,
equipment design procedures, and COSI models, it should be possible to complete a
design in about 2 days by hand (or 1 to 3 hr if software is available). One major
advantage of using this hierarchical decision procedure is that it simplifies the task
of generating a complete list of process alternatives to be considered.
Estimating the Economic Incentive for Replacing
lhe Existing Plant by a Beller Process Alternalive
Estimating the profil potential of the best" pOS5ible process alternative will
indicate whether one of our competitors could drive us out of business by building
a process that yields economics better than ours. We can usc Ihe
gross-screening procedures described in Chap. 9 to evaluate all the process
alternatives, to see whether a bener process exisis. If we find a better process, thclI
we refine the calculations, using the techniques descnbed in Chaps. 5 through 8 to
estimate the optimum design conditions for each alternative.
This identification of improved alternallves also indicates the changes in the
flowsheet that we should consider when we examine the retrofitting of our existmg
process. The savings in both raw-material and energy costs that correspond to
these structural changes in the flowsheet can be evaluated. Thus, there is a
systematic way of selecting structural modifications for more de[ailed evaluations.
Estimating the Incremental Investment Costs and
Savings in Operating Casu Associated with
Changing the Exisling Process
Our initial retrofit analysis sbould focus on screening calculations 10 sec whether a
more detailed retrofit study can be justified. Of course, if we eliminate a piece of
process equipmenl in our retrofit analysis, we still must continue 10 pay for that
equipment (unless it bas been totally. depreciated). Thus, the retrofit study should
focus on incremental annualized investment costs and incremental savings in
operating costs.
A systematic way of screening retrofit opportunities is to
1. Elimmate the process heat exchangers.
2. Identify the sigmficant operating variables.
3. Identify the equipment that constrains changes in [he significant operating
variables.
IDENTIFYTIlE EQUIPMENTTIIAT CONSTRAINS THESIGNIFlCANT OPERAT-
ING VARIAHI,FA If we calculate a proximity parameter (except that we retain the
sign of the gradient)
Any operatmg variable whose rank-order functIon is an order of magmtude
smaller than the largest rank-order function is then dropped from further conslder+
ation dunng the Initial screening. Thus., we obtain a set of significant operating
variables.
4. Remove the eqUIpment constramts by addmg e"cess capacity until the 10-
cremental annualized investment costs balance the incremental savings m
operating costs.
S. Evaluate the oplimum energy Integration for the new process flows, and check
the sensitivity of the optimization to energy integration changes.
6. Retrofit the heat-.cxchanger IlCtwork
EUMINA1E '1I1E PROCESS IlEAT EXCIiANCERS."The process flows and
product dlstnbutlon usually dominate the process economICS, and thC'refore we
initially focus on estimating the optimum flows. To accomplish thiS goal and to
avoid encountenng constraints on heat-.c"change equipment......e initially negkct
the heat e"changers (although .....e IOclude the operating costs for fuel. steam, and
coohng .....ater in lhe analysis). Once we have removed olher equipment constramlS
and havC' estimated new values for the optimum flows......e will resolve the energy
integration problem and retrofit the heat e"changers.
90 80 70 60 50 40 30
./
-.......
..... Opcr.
[7 .... Cap.
-Total
we can gain an IndicatIOn of the Incenllve for optimizing the significant operating
variables; ie.., unless PI < 0.3 for all the operating vanables. an optimization is
justified (sa: Chap. 10). Then we change the Significant operating variables lD a
direction opposite to thC' sign of the proximity parameter; ie.. if the gradient is
positive, we decrease the operating vanable to look for a minimum. As we change
the operatmg variables In lhese directions. normally we encoumer an equipmenl
constraint.
Note that the operatmg costs often he on an eqUIpment constramt e\en for
the optimum design of a new plant That IS. the mimmum of the lotal operatmg
costs for a design (for the type of plants we are considenng) nonnally is al lower
values of the operating variables than the minimum or the total annual cost (which
mcludes equipment as well as operating costs); see Fig. II I-I. Thus, after the
equipment sizes have bttn fixed and the plant has been built, if we try to move
toward the lowest posSible operaling cost. we will encounter an equipment
constraint. The addition of eXIsting equipment capaCIty to remove thlS constraint
dunng the design of a process by definillon will cost more than the savings m
operating costs, which IS wh) we fixed thc design at Ihe nllmmum total annual cost
However. the operating cOStS change over the life of a process. so that incremental
capital costs may be justified ror a process retrofit.
900.000
800.000
7oo.<XXl
60('-000

~ 5OO.<XXl
400_000
)()().000
200,000
100,000
o
20
REMOVING EQUIPMF:f\, CONSTRAINTS. Once we encounter an equipment
constraint. we add capaclI) until the incremental annualizC'd capital cost jU'it
balances the incremental Sa\lOgs in opt:raling costs Thus, we detennine the Dew
optimum relroti, trade-offs for Ihe significant design variables. or cour<;e, if we
encounter constraInts in mOfe Ihan one piece ofequipment. we merely add capacity
10 both umu until the lIlcfC'mental capital costs balance Ihe Incremental savings in
operating costs. (If time pennits and the retrofit economiCS appear promising, we
might also consider the optlmlZ3l1on of the other operating variabks that had
smaller rank--order fuoctlons.)
(III-I)
-I
aroc
I
'J'"" L -,--' ~ _
'-I vXj
WENHFY Tin: SIGNIFICANT OI'ERATINC VARIABU:s. The number of
operatmg variables is equal to the number of deSIgn degrees of freedom IIllnUS the
number of equipment sizes that have been fixed by the design. However, somc of
these operating variables must be fi"ed so that the process conslraints (e.g.
production rate, product purity, molar ratio of reactants at the reactor mlet_
reactor mlet or outlet temperatures. etc) can be satisfied. The remammg operallng
vanables should be tixed by an opl1mi7.atlon anal)"sis.
We can detennlllC which oftheoperaung \ariables are the most Important 10
optimIZe by eakulating the rank-order function (sec Olap. 10), except for thIS
analYSIS we include only the operating costs
(111-2)
L I ilTOC,ax, I
,- ,
."tGUME 11.1-1
COIlS ~ e r s u s lemre.alUle
Temperature. OF
358 SECTION I U HO'" PROCm SECTION III HO'" PROCESS 359
FrOID J. J MeK...., ""fC/ortll", PrQUUI1lfJ ""J
LHJI9rt, yol . Do:llcr, Nt... yo,l. 19n, p 182
TABLE 11.2-1
Optimum design v....illbles for exisling phllll
ENERGY INTEGRATION. From the analysIs above, we obtain new t:stlmatcs of
Ihe oplimum process flows, assuming that no constraints in heat-exchange
equipmenl are importalll. However, the new process nows correspond to a new
energy integration problem. I-Icnee, we use the procedures described in Chap. 8 10
solve the heat-exchanger network design problem for the new ltows. Sinee Ihe
optimum flows depend on recycle costs, which depend on Ihe heat-exchanger
network used, it might be necessary to lIcralC a few times to find a new oplimum.
R... ..,ROFIT TIlE HEAT-EXCHANGER Nt-"'WORK, Of course, we can decrease
thc retrofit investment by using as much of the exisling heat-exchanger equipment
as possible. Thus, we readjust the flows and heat loads in an allempt to estimate the
optimum retrofit conditions. We are still screening the retrofit opportunity, so that
our goal is 10 get into the region where the optimum is relatively flal, rather tban to
find the exact optimum.
RaClor con'emon
H, rurlle compo'ltlOn
Inlel lempe-rallHe 10 pilrllJI
OU1Jel Icmpcr.ature from paMial condenser
Bcnunc reco'cry In produc! column
ratio In prodUCl column
Toluene reco>ery In recycle coJumn
DJphcnyl rceo>cry in rcqclc column
Recycle r;uio m rocydc coJumn
0.75
'" 428 K
31 J K
099
" 0986
0801
10
Refining Ihe Relrofit Calculalions
If the economic incentive for retrofining a process is large, we mUSI refine our
screening calculations, just as we must refine the screening calculations that we
used for process design. The use of computer-aided design programs 10 accomplish
this task is discussed in Chap, 12. Before we consider this refinement, however, we
present an example of our retrofit procedure.
11.2 HDA PROCESS
Gas recycle
b-----------.--?urge
Compressor
As an illustration of our retrofit procedure, we consider the retrofining of an HDA
process. The original oplimum deSign variables, as taken from MeKella," are lisled
in Table 11.2-1 for the flowsheet shown in Fig 11.2-1. The steps in the systematic
procedure are discussed below.
11:2: feed
IFEHE Furnace
Flash
drum
Toluene
[oed
"

"'1-----'

Vi
Quench
Benzene
I-T:==-""i c
Toluene E
,
recycle '8
u
I
1--{'}-'
2. Design a New Plant Using tbe Same Process
An optimized design for the HDA process with diphenyl removed as a by-product
was discussed in Sec. 8.11. The smallest total annual cost was $4.7] x 1QoO/yr, which
included an annualized capital cost of $1.47 x IOoO/yr and an operating cost of
$].26 x 1Q6/yr. This optimum operaling cost is lower than lhe values on our cosl
diagram, S4.62/yr, but the investment is large.
I. Prepare litD Operating Cost Diagram
An operaling cost diagram is shown in Fig. 11.2-2. From this diagram we see Ihat
raw-materials costs are much more imponant than energy costs, Thus, we want to
modify Ihe process flows in our retrofit analysis.
J J M.::Ketl3. Proc,u,"fJ and DtJ,gft, vol 4,IXllcI, Ne.... York, 1911, p 182
F1eURE t1.2-1
IIDA process 60\nhcel
360 SECTlON 11.1 HDA n0CE.S5
SECTION 1I.l HDA nOCESS 361
-0.6.58
TABLE 11.1--2
Significant opera ling '.rillbles
t.
iJf.lih
iJf.Ja',II iJf.JQF,. aj",JR
rc
FUmlOl' fud -1134 -1156

Compressor po",..,r -125 -201

Product column ullHUts

'" 8c'nz IOfi In prod col

-'54
PurE" losses 311 7135 -. , 0
Sclecuv;ty 10<56 5571 0

Retyck col utillllo -)82

Cooting walt' pan cond

"
075 0.46 II. W
..,
.,
01 SO .1
"
'SO 'SO SO JO
"
.'"
077 0 0
4b. IDENTIFY THE SIGNIF"ICANT OPERATING VARIABLES. Values of the
rank-order function and of proximity parameters are given in Table 11.2-2 for the
largest operating costs. The results, as we might expect, show that the conversion
and tbe purge composition arc the most important operating variables (i.e., their
rank-order functions 8re an order of magnitude larger than the values for the other
two variables). Also,the large, positive values of the prollimity parameters indicate
that we would like to decrease the values of the conversion and purge composition.
Of course, smaller values of conversion and purge composition Correspond to
larger recycle flows. We might not be able to obtain these larger recycle flows
because of one or more equipment constraints.
ok IDENTIFY THE EQUIPMENT CONSTRAINTS. The equipment most likely
to constrain the recycle flows includes the gas-recycle compressor. the recycle
distillation column, the furnace, and the partiaJ condenser. Howeyer, we do Dot
oonsider constraints in the heat-exchanger equipment (we defer lhal consideration
until we energy-integrate the process again). Also, in the original design. there was
a considerable amount of excess capacity in the recycle column..
FIOCII W R Fllbel. M F OoI><I1Y. _I'd , M Doulla. Sa.....U\A Proass Rouo61
AJtanlllJv<:s, I"'-C Il:u.",c/\. 111 pres!, 1981, ..'nb pt_i&sJoo from the Anxoaa
Cbcmi<:aJ Socie1y
.... ELIMINATE THE PROCESS HEAT EXCHANGER$. We eliminate the feed-
effluent heat (:ll:changer shown in Fig. 11.2-1. although we will repeat an energy
integration analysis after we have revised the process flows.
Purge
0.20
Operating costs
Alternative 1
0.11
4. Estimating the Incremental Investment and
SaYings in Operating Costs
Gas recycle
H
2
reed
2.163
Toluene feed
1.691
ToluellC
recycle
0.01

<
E
0.004 ,
-0
u
u
"
U
,

]
'"
"-
-0.20
Diphenyl
3. Find the Best Process Alternath'es
An optimum design of the HDA prOttss with diphenyl recycled was presented in
Sec. 9.4. The total cost was $3.37 )( 106/yr, which includes an annualized capital
cost ofSl.27 x 10
6
/yr and an operating cost of S2.30 x to
6
/yr. We could probably
find an even better alternative if we recovered some hydrogen from the purge
strea.m. as we discussed in $cc. 9.3. Also. we sbould evaluate the other process
alternatives discussed in Sec. 9.3.
fiGURE 11.2-2
Operaung emt diagram. [F,om D. L Tnnll (md J AI Douglas. fHC 26: 68$ (J987). "'U},
1M Amt"cQIl Clttmica/ $",,)",]
There are at least two altcmatives to consider. ie.. removing and recycling the
diphenyl. To illustrate the procedure. we consider only the case where diphenyl is
recovered as a by-product (although thIS might not be the best solution). We follow
the procedure presented In Sec. 11.1.
J J McKelll. E.. n{ChtoOu,al P'ocus",g aM .. , vol. 4. Dekhr. Ntw York. 1977. II t82
362 SECTlON 11..1 HOA'loa.ss SECTlON 11.1 IIOA raDasS 363
Thus, for our problem, the gas-recycle compressor constrains both of the
operating variables, Lower purge compositions correspond to higher gas-recycle
flows, which are constrained by the compressor. Lower conversions correspond to
higher liquid-recycle flows. However. because of the 5/1 hydrogen-to-aromatic
ratio requirement at the reaclor inlet, an increase in the toluene-recycle flow must
be accompanied by an even larger increase in the gas-recycle flow. Hence, the gas-
recycle compressor also, indireclly, constrains an increase in the liquid-recycle Row.
Because Ihe gas compressor capacity constrains both of the significant operating
variables, we consider this retrofit problem first. The dominant operating cost
trade-off's in Table 11.2-1 for both significant operating variables involve raw-
materials losses balanced against fuel costs for the furnace, so tbat energy
integration will also be very important.
The cooling-water flow rate to the partial condenser and the reflux ratio in
the product column are less important operating variables. Moreover, these
variables can be adjusted to obtain their optimum values with the existing
equipment. that is, Pi = 0 for each. The cooling-water flow rate primarily trades
utility costs for purge losses, and the only retrofit policy possible to reduce these
costs would be.to increase the area of the partial condenser before the flash drum.
The reflux ratio in the product column trades product losses in the bOllom (which
get recycled) for increased utility costs to increase the reflux ratio. Possible retrofit
policies would be to use energy integration to reduce the utilities requirements or to
increase the number of trays in the stripping section to improve the product
recovery (which we consider to be impractical).
4.9
4.8
" 4.7
'"
"
4.6
<
0
u

';;
4.5
"

0
..
{! 4.4
4.3
o 10
"

0.1
"0
u
..
56
.E.

0 u
20 30 4<l 50 60
4d. REMOVE THE EQUIPMENT CONSTRAINTS- Now thaI we have identified
the equipment constraints, we want to remove those constraints. Since the gas
compressor constrains both of the significant design variables, we consider this
constraint firsL We merely add compressor capacity, and the incremental cost of
this capacity, until the incremental annualized investment balances the incremental
savings in the operating costs; see Fig. J1.2-3. From this graph we see that the
incremental savings are balanced by the incremenlal investment when we install a
new compressor with 56% of the capacity of original unit. The incremental
capital cost is S55,OOO/yr, but the savings in operating costs are S480,OOO/yr. Of
course, the large increase in the process Bows will also exceed the furnace capacity,
but we consider energy integration later.
We could also add another beal exchanger in series with the existing partial
condenser and tben trade tbe incremental capilal cost for the savings in purge
losses. The results of this calculation are shown in Fig. J1.2-4. At the optimum
conditions, the new exchanger should have 1.5 times the area of lhe original unit.
The incremental capital cost is S75,OOO/yr, and the savings in the purge losses are
S410,OOO/yr.
The new values of the optimization variables are shown in Table 11.2-3.
Obviously, the changes in the most significant design variables cause the process
flows to change, which will change the energy integration required.
Compressor ovcrdcsign, %
FIGURE tl.2-3
Esumallon of lhe opl1mum relrotil pohcy Co. rhe gali-'ecyclc compussor. (F,om W R. FuM" M F
[)oM"). and J. M. Doug/lU, Se'H,"ng P"xus Relrofit /&.EC Rewtl,rlt. In press. /987,
""lilt petm<ssi(HI from 1M Amr,klln Clwmktl! Soclfly.)
4e. ENERGY-INTEGRAn: THE RE"TROFJ1TED PROCESS.. The 7:H diagram for
the original process is shown in Fig. J1.2-5. The wide separation between the two
curves indicates that there was a significant incentive for improving the energy
integration. However, the T-H diagram for the parlially retrofitted process is given
in Fig. 11.2-6. Now, we see that the changes in the process flow rates have
significantly increased the incentive for energy integration.
The evaluation of various heat-exchanger network alternatives was discussed
in Sec. 8.1 I. The results indicated that all the network alternatives had about the
same costs. Thus, we begin our energy integration retrofit analysis merely by
adding excbanger capacity in series with the existing fecd-effluent heat ex-
changer. The incremental capital cost and savings in operating costs are shown
in Fig. 11.2-7. At the optimum conditions, we need a new exchanger with 68%
of the area of the original unil. The capital cost is Sl60,OOO/yr, while the savings
are S250,OOO/yr.
4.4,------------,
5ECTIO'" II) HIM. PROCESS 365
1200
4.3
$I
'!>
"
8

i!
4 1
-

"
4.0
3.9
0.2
..
o
"
I
E

E
;;
"
800
400
200 r----J
20 40 60 80 100 120

CondeMer O'\Ierdesign. '"
neal dUlY. 106 Blu/hr
nGl."Jl.[ 11..2-5
Tnnprratu""""nlhalpy d'.aram for 1M IIDA process It base-caJl':; opeTaunt mndillOm. (F,om W R
M F """ I M o..,hu, -/S<.trrting P,as Rt/,o.{i, AItNI'IW.-,HC '"
prrJJ. 1987 ..../" ClotmicQ/ Soc""y_)
F,OIII W L F tot F DoM:rty, and J M Doup.. -Sa....."'. 1"1'.-.
f&C JlDIr1rcIl,. ptUl. 1911. 0I'1th"......--. from
tbe ........,... o.o-c.l Socdy
TABLE 11.1-3
Optimum opn1Iting conditioM
RcKtor c:onenioa
PIlrp evmpo.,itiort
Coolin, Wiler 10 pin. condo
rl. of dtsip value)
Refh,,; Iio ret;y. col.
Fum dilly (10" HTU/hr)
onc-l ....
."
0."
"'0
127
11.9
ROJfFIIfiI .....
0.69
OJ,
110
..36
'"
4t RETROFIT TIlE UEATEXCIIANGER NETWORK. With Ihis optimum ret-
rofit of the rced-effllKot heal exchanger. the rurnace load is decreased 10 within
lOY. ofils design value. One way tbat we can resolve this discrepancy is by adding
area to the fccd-effluenl which according to Fig.. 11.2-7 will Dol lead 10 a
large increase in costs. AnothcT altemaLive would be to go bad: and to restrict the
process flows so that the existing fumaa: capacity is adequate. Similarly, we could
add a steam healer before the reed-effiuc:nt heal el{changer. or, belter, we could look
for more complel{ networks (pressure-shifting the columns, etc.) that would satisfy
our requirements.
Discussion of the Relrofit Anal)'sis
Our retrofit analysis indicaled Ihal we can save SI.140.000/yr wilh an annualized
investment of S290,OOO/yr. We could decrease the investmenl required by deciding
not to retrofit the partial condenser. Now we can compare these savings and
investment to those corresponding to building a new plant. Of course. we still
might not have determined the best retrofitted process because we have nol
5I'CllOl"l IU HD.A ...CXDS 361
FIGURE 11.1-6 I a d I\CW optImum
<hoi,> d "rim for the IIDA pfOCltlo" "'lib ,euo lieu cqlllPmc.n II
Temper;lIurc.cn , __J J M _._laJ - Sc,ltllUlfl P,oc.eu htMfil
110""1. (F,,,,,,, W R- FuN,. M F "i.- .
IdEC 1)187. ,,-ult p#muJ_f''''''' 1M A"""ic"" C/turow:41 S#>cwIY)
0.3

0.2
"'
;;;
0.'
0
u
."-
0
U
0
68
20 40 60 80 100 12<>
o
4.0

"' 3.9

8
38

l!.
0
3.7
"{f
3.6
160
'40
'20
'00
106 8m/hr
Heat duty.
60 80
40
20
'200
1000
u.
0
,;
800
-
,
;;;
l!.
E 600
"E
5
.>i
400
200
00
Exchanger overdesign.
nCURE 11.1-7
Estl.al*II01l olrbl: opUrnl,Ull R:lrofil polJr.:J rOllbe 1Ccd-dlllM:DI beal c..o.cM.n&t=r (F,_ W Il Fulln. AI F
Q/ld J M [)o"gw, - Sc'U"UItJ P,oass Rcuofil AIIUNl'fW,- IcfEC /UNa,"..... pten, 1911,
..ilil fH""UJIOft from "IP.or,le"" Clwmit'Q/ $Dc""y.)
368 g;cn()", IIJ SU...........y ... UUCJSES
considered the reC}'c1e of dlphenyl or the rcco...ery of any of the hydrogen In the
purge stream
Once \l;e on a retrofit poliC)" \Ooe must also refine our !iCrccllllIg
calculallons. The UK of computer-aided design tools is discussed In the next
chapter
11.3 SUMMAR' ANI) EXERCISES
Summary
We ha\'e described a s)'stematic procedure: for retrofitting processes that uses many
of the met hods descnbcd carher. The retrofit analysis proettds through thiS series
of steps:
1. Estimate an upper bound on the incentive for reuofitting-prepare an operating
cost diagram
2. Estlnlate the lIIcenll ...e of replacing the exlstlllg plant with an Identical
uSC: shortcut calculations and generate a list of process alternatives,
J. EstImate the inccnti\'e of replacing the existing plant with the best process
ahematl\e.
4. Estimate the lIIcremental Investment cost and the savings In operatlllg costs
associated with changing lhe existing process.
Q_ Elimlllate the prooess heal exchangcrs_
b_ Identify the Significant operating variables.
c. Identify the equipment that constrains changes in the significant opcratlllg
variables.
d RemO\'e the equipment constraints by adding excess capacity until the
Incremental. annualized capllal cost balaooes the incremental savings in
operating costs.
Energy-integrate the process.
f Retrofit the heat-cJ(changer network...
S. Refine the retrofit calculations. if justified.
6. Find the best retrofit alternative.
Exe:rc:ises
11.3-1. Retrofit the HDA process by recycling the dipbenylto extinction.
11.3-2. For any process that you have designed. look up the original case study and
compare ,he nt._material aJ$1s, utlhties COS!s.and M&:S indu "'ltb current values
How would you apecllhe optimum design values to chaoge n the economic ractors
chao!e1 Calculate the ranleorder runctions and the proximity paramelers. using
current pricc:s bUI the ori"n01.I ca.<;c-study values or ,he optimum variables
Retrofit the process.
CHAPTER
12
COMPUTER-AIDED
DESIGN
PROGRAMS
Our preliminary design (screening) calculations enabJe us to sec whether more
detailed design studies can bcJustilled_ln addition. they help us to selcct the thrc:e
or f1owshc:cts that arc the most promising and to estimale the: optimum design
condllJOns for each of these: process aJternatives. i.e. the compositions, tempera-
tures. and pressures of e\ery stream as well as the equipment sizes. However our
screening calculations were based on a variety of apprOllimations Hence. if our
scrc:ening indicates that a more detailed study can be justified. we need to obtain
more rigorous solutions to our design equations; i e., we need to derive a set of
material balances that are rigorous, we need to account for losses of reactants and
products, need to use more rigorous equipment design equations, etc. We use a
computer-aided design (CAD) prograin for Ihis purpose.
Numerous CAD programs are available commercially, including FLOW_
TRAN, PROCESS. DESIGN 2000, ASPEN. ChemCAD. HYSIM SPEEDUP
elc. In ";'any companies have Iheir own proprietary Since
lDstllutlOns use FLQWTRAN than any other program, we use FlOW-
TRAN III our examples. 1I0\Ooe\er. the other programs arc fanly Similar In their
general structure. FlOWTRAN was developed by Monsanto Co at the expe:nse of
60 worker-fears of effort and a cost of more lhan S2 million
369
370 SECTION 121 Gll'ElAl STlU(.T1,lU 01 (.UMPUnlAIOfU I)UI<.N (110...,..... 1>1)
f ...... J D $tIde<...... D
Cambndr<, lola.. 1971
Table: 12.11. The SlalemenlS used in other programs arc slightly dIfferent, bUllhe
same lype of mfonnation is reqUIred
Physical Properlies Data Bank and
Tbe..-modynamics Package
One of the most tedIOUS tasks in process dcslgn IS looking up data for tM phySIcal
properties of Interest and then using these properlies to compulc
the vapor-hquKl relatIOnshIps, the enthalpies of the process streams elc. Com-
puter.aided design programs ha\'e much of thIS infonnalion slored lD'dala banks
for numerous compounds, so thai these routine calculations are greatly SImplified
For example. FLOWTRAN contains the values shown in Table 12.I.J for the 180
chemical compounds (see Appendix D I) It is possible to add similar dala for other
compounds and 10 delermine Ihe conslants of inlerest dIrectly from experimenlal
data (Ihe user's manual" should be cOllsulled for Ihese procedures)
input dala
Any lillie Iv odenllf) lhe (:ilK SlUlI!
numen.; <ounsl;anu. lhe fiol conSllinl IIldlC'illCS lbe D1DIIb:! or
;a!kllbe nc.ll fOUl cOflnpu..J 10 lbe d>o..:a of lhe 01'1_ In Tabk
12 12-
Ti'le Slre.llm 1:0)"" fe.r Ihc SOIUllUn ClIn be ponIed In;a .allel) ol urulJ. lib moLbr I, rile
dcf4ull UDIl). lbc 'llc..m cUnnttllOflS fOf lhe lkt....ha:1 an be pnnlcd. lbe "'pur o,I.. la
Oln Ix reprod""".!, Jon.! '0' 1M phu"':tl prUrrrl) d;al:t rr<;Old; can Ix pnnred
t 'SCIlla reln..e d.. 14 rno"h flom pn, .. re hlc,
Used 10 Idenrrfy Ille cbel11""I.:ompounJ, used In Ihe UK Sfud) (sec Append" D)
UiiCd 10 supply lJquldphllK .... ben .elular solurron Illeory II nOI
Ippltcable_
Used 10 enlel oc:w cqulpmem Sub.oullncs
Idcnufics Ibe Iype 0( dnrl" C<juarrOll. and Ihe ol.eam rollnccllOllS followrll' lhe
krnnalS III Append 0
Spcofic:o the des'JIl pa.Itne:ICO fOI act> cqu'pmcm Subroulme follOO""II""
III Append!>. D
Spcofic:o tbe flo", fllcsol cadi componcnlln lhe Inpul S(.cams and mUlal esumlln ul
ryck Sflams.
Spcofic:s lbe lempelllllle of uc:b ked sr.ram "ncIlJ1lll.lll CSllmll"S ollhe lempc:uI...u
or rydc u.ea,,"
Spcafics lbe preu"re 01 cadl feed scream:tn(( Imllal CSllm:ttcs ntlhe l")Ck ouc...Da
USCIlto .... pp.CS& a !Losh calarLllUon _ a Slream
Pllamclnc casc .ludlCli
TITLE
PROPS
PRINT
FILE
RHR
I'AIR
TAIIU_ ILIJ
fLo\\rrRA
NEW BLOCK
lILOCK
PARAM
MOLES
TEMP
PRESS
NOFLSII
END CASto
END JOB
Computer-aided design programs all hlJ\C essentially the $arne componenlS:
The curren! trend is 10 add optimization rourines, process dynamic capabilities,
and a friendlier user interfacc.
The executive system reads the input data, controls the order in which the
equipment subroutines are calculated, pnnts the results of the calculations, ctc. The
physical propertIes data bank in FlOWTRAN contains data for 180 chemical
compounds, including molecular v.cighlS, critical constants, heat capacllles. acen-
lric faclors, ctc. These properties used by the. thermodynamiCS
to gas and liquid densities and as well as vapor-liquid
equilibrium relationships. for process SHearns In the plant Then the equipment
design subroutlDCS usc this information on the stream properties to cak:ulate
equipment sizes (and costs) for the various process units, i.e., furnaoc:s, heat
exch:mgers, compressors. distillatIon columns, etc.
Our diSCUSSIon of CAD programs will not adequate to use these programs
efficiently. Instead, our goal is to descnbe the deSign mfonnatlon that is reqUired 10
use one of these programs, 10 diSCUSS an approach for solvmg design problems by
using these programs, and to give examplc:s of some results. In addition, .....e c1ariry
the relationship between our preliminary design estimates and CAD calculations.
To accomplish these goals, we restrict our alte-ntlon to design problems
where lhe basic FlOWTRAN routines are adequate for solving the problem. The
program is much more flexible and powcrfulthan our examples illustrate, and an
interested reader slitould obtain a user's manual.
12.1 GENERAL STRucr RE OF
COMPUTER-AIDED DESIGN
PROGRAMS (FLOWfRAN)
I. An cxecull'o'C: system
2. A physical properties dalJ. ban\.
3. A thennodynamics properues pad.age
4. A collectIon of design (and cost) subroUtlnc:s for a \'ancty of process unHS
Executin: Routine:
The executive routine reads the mput data, iniualizes the variables used in the
calculations, arranges the order of calculalJODS to correspond to the connections of
tbe process units, calls for the thermodynamics package as it is needed, converges
the recycle calculations, prints the output results, and repeats the calculation
procedure for parametric case studies. The fonnat for the input data is given in
"J. D. Seader, W D Selda, aDd A C. Pi.... FWWTIUN S......""'_ All Irurotl""liotJ. CAChE
CambridJC, 19n (a'all:tbk 'rom Ulncb's Boot- Ann Albor, MICh.), J P.:lcr Oarl.
cd.. EnrclStf uo f"roc:ns 5_'"",,,, Us"'l1 FI..OH TR.AN, Ibid, R R Uu&bcs. CACJt Uw of
FWWTR.AN M UCS, Ibtd
372 SECT10N ll-' GENE""l 5TllUCTUlE OF COMf11TEIl_"tDEO varON PIlOGIlM'IS (TLO""-lIAN)
TA8LE 12.. 1-2
FLOWTRAN lhermodynamics options
SECTION 'U GF.NEI.AL STIlUCTUkE m- COMP\JTEI._AIDF.O DE$ION Pl<XJIlAMS (TLOWTIANI 373
TA8LE 11.I_J
PhJsicl1 propert)' data record
f'ropnctot) Mon...nto coos...".. compu.ed by PROPTY
''''f'' J 0 Seadn, W D Sndcr.lInd II C FLOWTRAN 5...."I,,' An In"""""" ..... CACbF.Co'l'.
Cambr>d,., M.... , 1917
Key wOld
Number of ehemic:alap:o'" (25 mutmum)
Vapor pressure optIOn
I AnlOine: equation
2. Cavell equallon
V.por fugacily option
I l<kal.!a5 equation
2. equatIOn
Liquid fugacity option
I Vapor pressure; Chao..Seader ,f ,uperalllcaJ
2. Ro:dhch-Kwong and Poynllng equal1ons;
Chlo-Seader if 5upercrilical
J Chao-Seader equalion (used ",ith components)
4 Grayson.Streed equation {used wilh
compone:nts, T_. < 9<XfF)
5 Option 2 uctpt P",,,,nil2-Sh,,,. foupc,cnhcal
N
J
, CO, AR, 0" NO, CII.
activtty coelflO"nt equal,on opHon
I. Ideal solution
2. Regular soluuon
3_ Wilson
4. Van Lur
5. Renon
6. Reno" ""th regul:u solullon fo. unspcatied pans
PROPS
Basic
Molar ...
Normal botllng po,nt
Cnl,caltemretBIUle
Cnl,cal
Cntlcaloompres.<ibihty
uquid v<.tlume oo"'I1.1nl
Liquid v<.tlumc (6O"F and 147 psia)
EDlhlolry
L>quid COnSlant-
Ideal_g.u heat capacily constanlS
Liquid enlhalpy constants
F..quillbna
Solub,lIty
E_pan"on ranor'
Aoentnc faclOr
,'apor pressu", const.lnU
Cavell vapot COnstanta
Common
s)'mbol
MW
T
T,
p.
'.
<J" . , <J,
w
",.".,0,
".. ".
FLOWTRAN
slmlool
MW
NBP 0'
lC
.,
PC
"'..
ZC
Vl
lDEN60 pll(lb mol)
III
CP(i) Btu/(lb mol 'F)
LlI{i)
DELTA (calfml)'/J
EXPF
OMEGA
VPA(i) psla -F
Vl'C(i) psia. OF
From J. 0 Seader, W D Se.du.and II C Pa..... S...,../<:r__A.. , .. "oJ><r
llD
",
CIIOlE Corp.. Cambnd,." Ma... 1917
These: data for pure components are adequate to predict the
properties of the mixtures encountered in process prOVided that the
mixtures satisfy tbe assumptions of ideal or regular sol.utlOD theory (DO
bonding). For more complex mixtures il is necessary to supply esttmates of
the liquid-phase activity coefficienls or to supply to us.e the
correlation routines available in the program to estimate the activIty coeffiCients.
(Again the user's manual should be consulted for these procedures.) .
The options available in FlOWTRAN for estimating the
properties of streams are given in Table 12.1-2. For purposes of we
consider only the simplest case where regular solutIon theory should give reason-
able predictions. However, the program is not limited in this respect.
J. D. Seade" W D Seidel. and A C. raub, FtJWTRAN Sl/...J<JI ..... I""oduc"o... CAChE
Cllmbrid.. Mus. 1911.
Equipment Subroutines
Another time-<:onsuming feature of process design is the calculation of the sizes
(and costs) of the process equipment, once the stream (low rates, temperatures, and
pressures have been specified. This effort is even more tedious if we must usc trial ..
and-error procedures to determine the stream nows or temperatures because of the
presence of recycle loops. However, the computer can solve these:. trial-and-error
calculations relatively rapidly, so that it is the ideal loal to use to improve our
estimates of Ihe material balances and to solve complicated sets of equipment
design equations.
Three kinds of information must be supplied to the program to use the
equipment subroutines: the type of design equations we desire to solve, how the
various prooess units are connected in the process flowshcet, and design parameters
for the particular unit under consideration. A list of the equipment subroutines
a\ailable in FlOWTRAN is given in Table 12.1-4, and some of these subroutines
are given in Appcndi.x D.
We note that there are no equipment subroulines available for calculating the
size of reactors, for the shortcut design of plate absorbers or extraction columns, for
the shortcut design of distillation columns when the relative volatility is not
constant, and for the detailed design of coakr condensers (when the vapor and
]74 s.ECfIOI'l n.l Gfl'lfMl STIuctUlU: Of' COMfVTEl.AJOEO DE$IGN Pl.OGIlAMS (l'LOWlIA"'1
5fCTlON lJl ... ATU.Al IALANCE CAlCUU:noI'l5 ]75
TABU 12.1-<4
FLOWTRAN svlwoulioes
.....
-
,......
TABU. 11.1""
FLOWrRAI' subroolioe!i (CUflliIlU..J)
From J D Sude,_ .... D Sc><lr,Dd A C P.llb. FLOWTRAN.s-t<u_ A..
1..",,,II,,,r-. CAClIE Corp.. ul1Ibnd.... Mau.. 1911
liquid phases are nOI in equilibrium). I I IS possible to develop subroutines ror these.,
and olher, problems and to add them to FLOWTRAN, but we do nOI consider
probkms or Ihis type.
Pump <:OS.
Compresso,
Tray Ilorbcr emt
cool
B)'_prodll" 'IOlIlue
Product val ....
R...-ma.cnal JlIC
Profirab,II.)
P... led abwrbcr ",-1
Heat-ue/uncer <:OS.
Stream OUlpul edllor
Componen. plr)'YCal prOpclllCS table
..codficJCnu; u,bIc
S"um pnnl bklcl
of a stream
HUlin. InJ coollog
ChcmlCll ,UClO'
llJd-.ubrract ,eaClor
ChemICal rUCIO/ (elren! of rUChon model)
(f-RKB
CABSR
CClNI
('[XCI
CEXQ
CEXCl
crUMP
CCOMP
CTABS
CIlETR
8PROD
PkODT
RAWMT
PRon
SlMRY
TABLE
GA\1X
Sf'R''T
AST\1
CLRVE
RIoACT
ARE"C
XTNT
R,pon
Reacuon
Strum adduJ04
Stream alkiruoa ..ub no pllasc dung<:
51fClUD
Ccu.rifupl pump SlU aDd power
Sueam mwupl>caUo.. by a pa<amcter
Coalprcswr aDd tllrt.lIe
S1raD1
SbonCUI bot uchan,ef
Sboncul cuok, wndrnsc:r
Sboncut desupc.hc,J,l(,
Heal
Shoncu. p,amll/totIJ uponzcr.. oondenscr
Shoncu. ,ebo,ler{mlelcooler
Thrcc-phuc: heller/coole,
SlIOncuI be,,! euhaniler

Adllibil.K ftuh
Gc:nefal-purpoK Ilalih
1!iO.hennal .hree-phasc: f1a5h
Ad,alnlie,'l101hennal.h,a:-phtic 1Ih
Bounded WCplelD tuum coavaCUOl:
RigorolU dLSuliallOn (KB melbodl
Shoncut (Edm>n5Ier)
Sboncut dUllllallOn (Winn-Underwood)
Const3.lll spill '",CIIO<l sc:p.anmon
Rigorous dutlilatlOll/ablorpuon
(1IUI1rII ""'Ihod)
EXCIII
CLCNI
DESUP
IlEATR
EXCII2
BOILR
IITR3
EXCtIl
unll opcrallom,
ADD
MIX
SPLIT
PUMP
MULPY
GCOMP
PART
SlfQm Ol)Qw:fFDOC
SCVW
AbIorpuOfl
ABSBll
Othe. scpanuon
EXTRC
FRAKB
OlSTL
DSTWU
SEPR
AFIlAC

IFLSII
AFLSII
BFLSII
KFLSIl
FLStIl
Conol
CNTRL
pcV'
DSPLT
RCNTL
CAFLll
CFLtIl
CIFLII
CKFUI
CAFRe
COSTL
CD.lllIolkr
Muluplc-par1lJllClef CD.IItIOI bIod.
Dwlllale I'a:d rorwud control
RIIIO, slim and dli'rell:tlCC feedbild;- eonuollt:r
Flash drum Ot
column OM!
12.2 MATERIAL BALANCE CALCULATIONS
In Chaps. 5, 6, and 7 we discussed shorlcut procedures, ror eSllmalmg bolh overall
and recycle material balances. We used these shortcul calculalions 10 help fix Ihe
struclure or Ihe nowsheel. One struclure Ihat we discussed III delail for lhe HDA
process is shown in Fig. 12.2-1, all hough, as we diSCUSsed m Chap. 9, this slructure
probably does nOI correspond to lhe best design However, smce our goal here is 10
demonstrale how CAD calculallons can be used 10 refine our approximale
calculations, we develop a more de131lcd solullon for the ftowshcet 81veo In FIg.
12.2-1
376 UCTION IU MATUIAl. ,AUlI"Ct: CAl.CUUlT1Ol'lS
10101" In MATfUAl. UUltfCt: CALCUUlflONS 377
To
p,,,,
Compressor
r
H

Mixer -'0'
Flash
l!ene
Quench
IkT
ne H2, tH>I,
Tower 3 T..- 2
I-
Tower I
e
Simplified Flo"sh( for 1aterial Balance
Cakulations
Our process f10wsheet (\\Ithout an) heaters or coolers) is sho...n in Fig. 12.2-1.
Ho\\c:ver. to perform a set of matenal balance calculations, we need to
onl) thC' umts where Ihe component f1o ..... s change, the reactor, the phase spliller
(flash drum). the columns, and stream mU:lOg or splilling. I n addition, we consider
the locations wherr pressure changes take place. because: most of the equipmC'nt
subroutines require that the pressure be: specified. For our example, we assume that
the feed streams are at 535 psia, that tht: pressures of the gas and Iiquid-reC)de
streams are raised to this level, that the reactor operates at 500 psis. that the
pressure in the flash drum is 465 psia, that the stabilizer pressure is 150 psia, and
that the product and recycle column pressures arc 15 psia (we should use 20 psia
for final design calculations, but here we wanl to check our shoncut results).
Our simplified flowsheet for the process. which shows the unilS needed for Ihe
material balance and pressure-change calculations. is shown in Fig. IU-2. ote
that we must unils \\here strC'am mixing and splitting lake place: (\\'hich In
practice mIght onl) be a tee). and .... e lOcfude vahes that are used to drop the
pressure if a phase: change: mighr occur
Dlphenyl
FlGURF; 12.1.-1
HDA block f\o"Sh1
General Approach to CAD Calculations
Once the structure of the flowsheet has been fixed, then the conventional approach
to developing a design IS to calculate tbe material balaooes, calculate the energy
balances, calculate the equipml;nt sizes, cakulate tbe capital and operating costs,
and evaluate the process profitability. Most of tk CAD manuals imply that all
these problems can be solved simultancowly, but they recommen.d the
development of a solution by breaking the complete problem down Into a senes of
smaller problems, The reason for this evolutionary approach is that it to
make mistakes entering the input data, tbe codes have fairly large runmng urnes.
and It is both expensive and tediow to make numerous rUDS that fail.
Our goal here is to use a CAD program to evaluate tbe approximations we
made in our screening calculations to verify our selection of the most promising
process alternatives. We based our matenal balances on thecomplele
all valuable materials. along With some Olher approximations ror varor-hquld
splits. Smce the energy balance calculations and the calculations or SIZes
and costs all depend on the material balances. our approach IS to revue the
matenal balance calculauons firs!..
Sequential Modular CAD Programs and Stream
Tearing
Man)' of the CAD programs that are commercially available (PROCESS, DE-
SIG 2000. ASPEN. etc.). as ,",ell as FlOWTRAN. ha\"c a sequentIal modular
structure. Thai is. if we know tht: mputs to a process unit, the equipment
subroutines will calculate the outputs. Thus, if we 6x the feed flow rates of
hydrogen and toluene and if we: know the gas- and liquid-recycle flows for the
(10\\ sheet shown in Fig. 12.2-2, we can calculate the reactor feed conditions. Then
.....e can calculate the reactor product stream. the split tbal takes place in the flash
drum, Ihe outputs hom the valves and towt:l$., the purge split, etc.
Strtarn Tearing
Ho...ever. we do not koow the gas and recycle flows initially. So if recycle loops are
present in the f1owshecl, we must lear the recycle streams. That is, suppose we tear
both of the gas- and liquid-recycle streams Fig. 12.23) and we guess the
component 80ws in each of thcsc: streams as well as the temperatures and pressures.
Then ....e can calculate all the outpuls from every unit, and eventually We can
calculate all the component flows in the gas- and liquid-recycle streams H we
change our guessed values until they match our calculated values, we will have
obtamed a converged set of material balances Thus. the sequential modular codes
mclude convergencc blocks that wlllilerate on these guessed aod calculated values
until they agree wuhm a loleraoce spc:cified by the user.
378 SEcnON IU .....TU.lAl , ... UNa CAlCUunO/'1S
SEc;lI".... In IotATUI"'lIlAl...:'CE U1Cl/UlIOI<S 379
Dlphenyl ph< y
Compressor
Purge
Purge
I -
CH,
I
Mu,er
Reactor F1a>h
r
(
Pump
V.J""
Belene
Ih.
.1
Tower 3

Valve Tower I
0;
" I
Tol
Purge
Compressor Purge
CH,
L
Mixer Ileac'",
Flash
"one
Pump V.J""
Denr
ne
H
2
. CH.
1
Tower 3 Tower 2
-
V.J"" fo- T""",
I
Tol
ncURE 11.2-2
HDA pn'IOXa. lknnbcd: shOWlllI allwuu-
FIGURE I1.J,.J
HOA Pfoc:csa. 1eU11l' ro;ydc 5oUcanll.
Coovrrgeoce
Of course, convergence is obtamed much more rapidly if tbe iniual guesses are
close to being correct. Some inellpenenl%d users stan by guessing zero values. Even
though the CAD programs usually con\trge witb zero starling values for the
recycle flows, a very large number oriter.nons nu. mally is required (i.e., it takes a
very large number of iterations 10 build up large recycle 80ws from a zero slaning
value). However, we can use our shorteul material balances 10 supply "good"
starling values.
Minimiring Ihe Number of "Tear Streams"
After some thougbt, we might realize lhat 11 IS very inefficient to tear both the gas-
and liquid-recycle streams. If, instead......e ttar tbe stream entenDg or leaving Ihe
reactor, v,e brea.. both recycle loops (see Fig. 12.2-4). Most CAD packages mclude
an a1gonthm that v,"!U indlcale which streams to tear to tbe number of
convergence blocks that need to be !Deluded ID the 8owsheet. We can also use our
shortcut calculations to obtaIn good estimates of lhe feed to the reactor or the 8ash
drum.
Cbeck tbe Physical Properlies
We prefer to choose Ihe stream enlering the flash drum as the starting polnl for our
calculations It 15 essential to ensure Ihal the physical property dala provide
realistic prediclions for the process under consideration, and therefore II IS always a
good Idea 10 millate a CAD siudy by undeTlaklllg some 8ash Clliculallons. In
addmon, SPCClfYlllg Ihe fc:e:d to a 8ash drum In a t10wshcct almOSI always breal..s
DlphcnyJ
FlGURE 11.1-4
HDA ptootlS, 1e1lnDI xheme..
one or more recycle loops. Moreover, starting with the foed to a flash drum allows
us to evaluate somc of Ihe decisions we made concerning the liquid separation
system.
Other Types or CA0 ProgTa.ms
There 3fC CAD packages QUASI LIN. ASCEND II) that are based
on an equation-solving approach. That IS. they 311empt to solve all the equations
describmg the flowsheel These programs should be much more
efficient if there are a number of recycle loops. because the sequential modular
programs suffer if there are many nested iterations. 10 addition. SlInultaneous
modular Iteration schemes hne been dt'veloped (FLOWPACK and a rcecnl
Compuler Informalion
The milial step m dc\clopmg a CAD program is to translatc out process flowshcet
mto a computcr mformallon diagram The compulcr information dIagram is
essentially thc samc as thc flowshccl. cxccpt that wc must show all mlxcrs and
splillcrs and thc location of our convergcnce block(s) In FLOWTRAN wc must
number cach of thc strcams from SOl to S99 (iflcss that 10 strcams arc prcscnt. wc
number them from SI to S9). Moreover, we give each unit an arbitrary namc of up
to six Icltcrs and numbers (slarlmg Wilh a IcHcr). and .....c indicalc thc lype: of
subroutinc (sec Table 121...4 and AppendiX D) Ihat wc will use for lhc calcula-
tion A computer lIlformation diagram for HDA material balances is shown III
Fig 12.2-.5
Starling Values-Approximale Malerial Balance5
Using thc apprO\lmallOn proclures descnbed in Chaps .5 and 6(and in AppendIX
8 for thc HDA plant). \I,'C can estimate the o\'erall and rec)'clc material balances
For a ..... here the dcsltcd production ratc of benzenc is 265 mol(hr. lhe
convcrsion IS l: = (thc selccll\'ity is S = 0.9694). the purge compo<;lIlon I'>
= 0.4, and the molar rallO of h)'drogen to tolucnc is .5/1. \I,'C obtain the Slream
flows shown in Table 12,21. Thcn in Chap. 6 we descnbed a shorlcut procedure for
flash calculations. and the revised flows arc shown III Table 12.22.
We note from these ren'iCd flows that we do not meet the desired production
rale specification (sec thc flash hquld now of benzene) because so much benzene
.....ent o\'crhead Vlllh (he flash vapor. We also notc that the gasrcC)'de flows no
longcr balaJICC and that a SIgnificant amounl of benzene is recycled wilh the gas-
recycle stream. Of course, the errors arc not as bad as Tabk 12.2-2 seems to
mdicale, becausc the benzene in the gas.rccycle stream will be recyclcd through the
reactor and be flashed again. Al this point. the amount of benzene in the nash
liquid will increase. We could try to solvc this iterative probkm by hand, but
inslead we use our CAD program
It .....ould be undesirable to recycle this much benzene 10 the reactor
because a substantial fraction of this benzene would be converted 10 diphenyJ (if .....c
had recycled alilhe diphenyl. the benzene rCC)'C1e would not be a problem because
modification of FLOwrRAN). whIch use equipment subroutmes but con\crgc
all thc rccyclc oops simultaneously These newcr types of codes also makc it
possiblc 10 optimIZe thc f10wshect at the same: timc as thc rccylc loops arc being
con\crged. Thus, Improved codcs should be availablc In the futurc.
v U t.anJ, I I B..,..... and I E G'(>$5m.ann. 0rtrmoza,... and HeIIl lnlep.".".
.,Ih rroca:l' rare' nt' 72'" 19lt6 Annual AIChF Mee1lnl- ,..haml Be.ch.. Nomloo
...balIlled 10 C_,."J'" ..I "9-'"
g'
Compressor Purge

CH.
I
v-
Mixer Reactor
A"h
!ene
Pump Valve
I
Benr
nc
H
2o
C11
4
I
Tower 2
f--
Valve
Tol
H,.
SECTION 'u ."t..-!'OCI' C"lcuun<)NS 383
TABL ILJ,.I
Slream 80"'5 (or perfect splits in flash drum-Ie,e1 J
II, T...... Ob I ......... Rac:lor
c_,
''''
,...
nerd.. ...
II, 4619 0 13500 0 lllf)'i
Cll,
'"
0 20250 0
Il<=M 0 0 0 0 0
0 2H4 0
'"
J."
Olp/lcn)l 0 0 0 0 0
I)(h, Ruefof fb,h
00>
Compoocol rucliOll n,;l
liquid Puree rK)CW
H, - 269.1 ISC87 15487 0 1987 IlSOO
CU, 2734 23230 BUO 0 29110 20250
Il<=M 265,0 2UG 0
,."
0 0
TolLXnc -273 4
" I
0
'"
0 0
D1pbo:nyl .,
"
0
"
0 0
Program Inpul
We follow the input formal in-en In Table 12.1-1, and as a title we choose IIDA
FLASH. There are the components in (he feed stream (hydrogen, methane,
benzene, (oluene, and diphenyl) and we decide to use the AntOine equallon ror
vapor pressure (opuon I In Taole 121-2), the Redlich-Kwong equalloD for Ihe
IFlSH (sec .'\ppendu; 0.2). The computer Information diagram IS
sho.... n In Fig. 12.2-6. This diagram IS .dcnlical 10 lhe flash ullil (hal has bttn
isolated from Fig. 1:2-2-5.
o Compressor 521 Purge
...E
GCOMP SPLIT
SI5
S02
I.
I
ADDl

Reador

CONVG

A"h
ADD REACT 5CVW IFLSH
!3 SI2 S03
Pump Valve I
PUMP AFLSH
t
SIO
j
jS
S04
Tower 3 Tower 2
1'E-
Valw= 2

SEPR SEPR AFLSH
5!l9 506
S
511
S
fiGURE 11.1-5
HDA proc:as., lIo.,uu 5ubrOUllDcS.
there would be DO selectivily loss). t-Iowevc:r, our correlalion for producl dislnbu-
tion (tbe selectivily data in Appendix B) is based on a pure toluene: feed 10 the
reaclor. so that wc: cannot eslimale the loss of tbe bclUene 10 diphenyl because of
tbe recycle.
If we bad installed a vapor recovery system on tbe fLasb vapor stream, our
estimates oftbe process 60ws Y<ould have been more accurale. However, il should
be a good test of our approximate material balances to compare Ihis worst-case
condition shown in Table 12.2-210 a rigorous CAD solution. Hence, we use the
values shown in Table 12.2-2 as lllltmi values for our CAD runs.
Flash Calculations
We slart our CAD slUdy by Just conSidering the flash drum, I.e., we want to ensure
Ihat the physical properties are reasonable We call the umt FLASH, and ""e usc an
TABLE 12..1-2
Stnam Bash-Ie'el 4
II, T...... 0_
u,.w
a_
"-"'
,... ,...
.ee,'a.
"",-
,...
H, -4619 0 11SO.O 0 t816.9
CH.
".
0 20250 0
""..
8<- 0 0 0 0 0
ToIlM:DC 0 27H 0 911
,."
Dlpbalyl 0 0 0 0 0
Deh. Relcu.. F'.., Oh
COOlpoaeal .uclioa u.il '.po, liquid rurtte ."d,
H, -269 2 1S4' 7 1S47.0 2 1985 1148 S
CH. 271.4 2l2JO 2112.0 II ".. 201S-4
Il<=M 265.0 26S0
,..
US< J8
'51
ToI_ 271..4 911 J6
.,.
"' "
Dlpbnr.)1
.,
"
0
"
0 0
384 $.fCTIO'" ILl CAtCIJLAllO"'S
HCTlO," 111 MATEI.IAlU..... NCf CAlctlLAnom 385
We USC' our earher estimates of the stream Row rates (see Table 12.2-2), and
l\e enter the Tn the same order as they appear 10 the RETR statement
S02
MOLES 501 1549 2323 265 91 4
SOl FJm
IFLSH
The temperature and pressure statements for the feed stream are
TEMP 501 tOO
PRES5 SOI 465
S03
FIGURE 11.1.-6
Fl..h calc\ll.lion
vapor fugacity (option 2 in Table 12.1-2), option 5 for the liquId fugacily, and
option 2 for the liquid activity coefficient. Hence, the TITLE and PROPS
statements arc
TITLE HOA FLASH
PROPS 5 2 5 2
We decide to print the mput program along with the results. so that l\e can
maintain a
PRINT INPUT
All lhe components of mterest arc listed In the ph}"sical propertIes data banl (sec
Appcndlll D.l for the component list and the allowable names to be used in the
program). Note that FLOWTRAN uses biphenyl instead of dlphenyl. a possible
source of conrusion which exists elsewhere in the literature. Thererore we can write
the retrieve statement as either
RETR HYDROGEN METHANE BENZENE TOLUENE BIPHE YL
0<
RETR H2 a BZ TOL B-P
The block statement associates the unit name that we sclected. FLASH. with
the equipment subroutine ill FLOWTRAN, IFLSH. and identifies the illput and
output streams. From Appendlll 0.2, we see that the block stalemenl for an IFLSH
subroutine should be
BLOCK FLASH IFLSII SOl S03 502
Similarly. if the Rash drum temperature and pressure are lOOT and 465 psla for lhe
first case. and we desire to print the K values (the FLOWTRAN K values w;lI be
diITerent rrom estimates using the Hadden and Grayson correlatltlllS 1II AppendIX
C.I). we write
and this compleles one case
To conSIder the behavior of the Rash drum at other pressures. we introduce a
new title statemenl. a new PARAM statement and a new PRESS statement, and
another END CA$Ecard. A program for operatioll at lOO"F and both 465 and 480
psla is shown ill Table 12.23. Note that the PARAM statement for this second case
is
PARAM FLASH 2 480
because l\-e are changing only lhe Sttond eotry in tbe PARAM statement (see
Appendix 0.2).
Program Output
There are sc\eral parts to the computer outpul, including a stream table and an
equipment summar)'. which gives the heat added or removed rrom the drum
TABLE IU-J
Flash caleulaliollS
TITLE UOA FLASH DRUM
PROPS 5 1 2 5 2
PRINT INPUT
RETR HYDROGEN METHANE BENZENE TOLUENE BIPHENYL
BLOCK FLASH IFLSU 501 503 s02
PARAM FLASH 1 100 465 1
KOLES Sal 1 1549 2323 265 91.1 4.2
TEMP sOl 100
PRESS 501 465
END CASE
TITLE HDA FLASH INCREASED PRESSURE
PARAM FLASH 2 480
PRESS sal 480
END CASE
END JOB
PARAfl.1 FLASH
100 46:li
(12.2-3)
(12_2-2)
(12.1-1)
273.4 - 265
Conversion 2 = 2734 = 0.0307
364.5 - 91.1
Conversion I = - 0.75
199.7 + 298.0
Spht Fraction for Purge Stream :co 1548.7 + 2323 = 0.1283
The PARAM statement for the recycle compressor requires lhat we specify
the exit pressure and the efficiencies., while the PARAM statement for the REACT
block requires that we specify the conversion of each reaction. These conversions
are related to one another by our selectivity correlation. Again, we use our
approximate solution to estimate the appropriate values
lIowevtr, the output of pnmary mterest is shown 10 Table 12.2-4. These are the
same values that were used to compare the rigorous and approximate flash
calculations in Table 7. I I.
FLO\VfRAN Program for Material Balaoce
CalcullitiOl13
The ADD block requires that we specify the component ftow rates of each feed
stream (in the order in which tlley appear in the component list m the RETR
statement). The MOLES statement for the fresh feed toluene IS
The: computer IOformallon diagram we Ulie for the material balance calculations IS
given m Fig. 12.2-5 However, since It IS easy to make nlistakes in developmg any
CAD code, it is always a good Idea to bUIld the program in small pieces and to
debug each of these smaller porlions. Thus, we might develop separate programs
for the liquid separation system, i c.., from the flash drum through the pump shown
in Fig. 12.2-5, and another program for most of the gas-recycle loop. i.e.. from the
flash vapor stream back through the reactor. We can use our approximate stream
flows gh'en in Table 12.2-1 or 12.2-2 as staning values for the two slTeams.
A program for the calculation of the material balances is given in Table
12.2-5. We use SEPR blocks inltlally for the dlsu.llallon columns. because they
provide the simplcst way of estimating the component flows for specified fractlonal
recoveries, ane we use the conventional rules of thumb to fix these fractional
recovenes. (If necessary, \lie adjust these spill fractions to satisfy our product purity
requirement) The feed rate to the flash drum is taken from Table 12.22. The
PARAM statement for the SPLIT block on the purge stream requires that \lie
specif)' the spilt fracllon. We can Ulie our approximate material balances (see Table
12.2-1 to esllmate this value)
o

,.,,...0\ .... "0 0

"""'Ol"0'I0'\'Ol" 11'1
.,., 'U'lU'l 0
IOr-Oce 'Ol"
,., .. .-l 0-
,...0 ON 11'1
N '" M
,.,0 rl ON 0 0\U'l 'Ol"
,...,.,..n"'O'IO'>N,...'"
'0 _1Il0\ _...-<I ...
Ul'Ol"ONNOr-r-rl
rl oCl) 'l'""lCl) 'Ul
NO NlI'I N....
..,. lIPD 0 \0 N
I
OOr-ONr-r'l
o 0 0''''''') l/UIl ...,.
'.-4 _0 -0\
OUl 0\0\\0 \0""
010 .r- '.-4
.... ..,.0'Ol"0.... r-

o
o
M
'"
"'
''''
"'"'
""
... '"
0",
"' ...
......
ESES

IaJ r-r--, COeD
:>\D,.,ONO

0(000,..,0
:> .. '.-100
10\0000
1o::000N
000
UUl'Ol"'Ol"VlN
O"O'I,... .... NOI
0\ CO <0 VI 0'>
WVlCON .... O
... . .
OOO'Ol"\ONO
"'M'"
p:;Q)\O\O .... O
:C,... ....,....,.
'.VlCQ '\0
lnN\oo\""
alNN\DO\""
..J ..... r-O,.,'Ol"
(""It", N CO 10
P;:'Ol"OcDO\r-
O\OChNNCO
Po 'Qf"'l""
o(r-NNMO
>'Ol".-I \00
Ul ,., a:l .. 0
.-INN,., ..
I 0
0::00\0,...,... ....
:e .... ,.,,...O\\O\O
,::> "<7 O\r-\O 0'1
"'0\00 0"<70'>
W ..... .., '0,0 .....
..:1..:1 0"''''' ..
o rl .... NCI)'CI'
'"
I..JOOOOO
1<00000
I fol .. 000
100'1'" .... 0
I f-I'Ol"NUl 'N
I 111 ""'''"0.... .
I .... MNO'I'O:I'
I
...
o
'"
,
'"
...
o
'"
o
I
..
;:-
N
o
-'"
..
I
'"
.. "
o:i
0",
.",
"''''

'"
"'
.-,
::>
"M"
0",
'" o
...
I
0"''''

ON
0"'"
0"'"
0
",0
I "
- ::>
" ..
-OW::>
::t:U)QO
'" "

.. ..
.... Of-otaJ
-IaJIl(O::
"'''''' =.. :=Vl
'"
..
...
..
MOLES SI3 4 2734
TARU: IU-5
IIOA planl mal""rial b.laMe'i
TITLE HOA PLANT - MATERIAL - CONTROL SYSTEKS
PROPS 5 1 2 5 2
PRINT INPUT
RETR HYDROGEN "ETHANE BENZENE TOLUENE BIPHENYL
FLASH IFLSH 501 503 502
VALVI AFLSH 503 60 sOC 0
TOWRI SEPR 504 506 505
VALV2 ArLSH 506 60 507 0
TOWR2 SEPR 507 509 508
TOWR3 SEPR 509 511 510
PU"P PUMP 510 512
BLOCk PURG SPLIT 502 520 521 5.0
COKP GCOKP 521 515
TOLFO KULPY 530 513
ADD 512 513 SIC SIS 3.0 516
BLOCk REACT REACT 516 6.0 517 a
CONVG SCVW 517 0 0 FLASH 501 0 0
PCONT CNTRL 508 TOLFO 1
"CONT RCNTL 516 516 PURG 2
PARJ\.K FLASH 1 100 465 1
PARJ\.K VALVI 1 150 0 1
PARA" TOWRI 1 1. .995 .003 22.0 1
PARA" VALV2 1 15 0 1
PARA" TOWR2 I 1. 1 .. 995 .001 21*00
PARAM TOWR3 1 1. 1. 1 ..995 .00520.00
PARAM PUKp 1 535
PARAM PURG 1 2 .128 .872 5.0
PARAM cal'll' 1 535 0 1 0 .8 .8
PARAM TOLFO 1 .6625
PARAM REACT 1 1265 500 0 2 4 .75 -1 1 1 -1 21.0
PARAM REACT 32 3 .03 1 0 -2 0 1 7C.0
PARAJ'I CONVG 1 3.0 21 6.0
PARAft PCONT 1 6 265 1 .3 0 .001 0 0
PARA" "CONT 1 4 1 5 1 .5 .05 0 .001 0 0
"OLES 501 1 1549 2323 265 91.1 4.2
"OLlS 530 4 400
"aLES 514 1 461.9 24.6
TEl'll' Sal 100
TE"P 530 100
TE"P 514 100
PRESS 501 465
PRESS 530 575
PRESS 514 575
END CASE
END JOB
SFrTlON In MATU1Al Ml.AHC'E CAlC1JUTIONS 389
becau5"" this stream contams only toluClle. which is the fourth component in our
list
RELATIONSIIIP 8t::T\\EEN OLR Af'PROXI !AlE MAn::RIAI. BAUr-.Ct:s
A""D TI-tE FtO\\TRAN INPlIT. We used our approllimate malerial balances to
estimate the fresh feed rales 10 the process. the split fraclion for the purge stream,
the reaclor conversions for Ihe 1.... 0 reactions. and as a slarting guCS$ .... hen .... e tore
the recycle loop at lhe feed 10 Ihe flash drum lienee, our approllimale material
balances provide a good starting pOlnl ror lhe more rigorous analysis. It nOI 3
simple mailer 10 make -reasonable" guesses or the values required for lhe
FLOWTRAN input if these approllimate So[utl0ns arc not available.
Need for Feedback Controllers
Also nole thatlhe FLOWfRAN input requires different infonnation from what we
used for our illltial desiJ!.Il That IS, for the FLOWTRAN Input we must specify
both fresh reed rates, the r<:aClor COn\-efSIOn (we can U5"" our 5""lcctlvity correIalion
to calculate the con\'ersion of the by-product reaction). and the split for the purge
slream. In contraSI, In our shortcut calculations. we based lhe desIgn on a
Specification of the production rate, Ihe r<:aClor conversion. the purge compositIon
(which is cqui\'alent 10 Specifying lhe fresh feed hydrogen flow). and a constraint of
maintaining a 5/1 hydrogen-to-aromatics ratio at the reactor Inlet.
To have our FLOWTRAN prograT:l produce the desired production rate, ....e
willnccd to add a rccdback controller to adjust the toluene fresh reed rate until the
benzcne production rale matches the desired value, Similarly, we will need 10
adJusl the split fractIon ror Ihe purge stream un Iii the hydrogen-to-aromatics ratio
at the reactor inlet mll\ches the desired value of 5/1. Each of these feedback
controllers introduces a new level or ileration. l-knee, we normally dder adding
these controllers unlil after we ha ve obtained a converged 5""t or material balance.s.
The order in which FLOWTRAN undertakes the calculalions is the same as the
order specified ror the BLOCK slatements.
Including CODtrollers- ested Iterations
When we atlmJpt 10 install a reedhad: controlkr that will manipulate the
fresh feed rate of toluene 50 that the Row rate or the benzene product stream is
265.0 mol,lhr, we find thai the CNTRL block can only manipulate a parameter in a
block stalernent and cannot adjust a flow rate directly_ To resolve this difficulty. we
can add a MULPY block to Ihe toluene fresh feed stream and then adjusl the
stream multiplier: see fig. 12.2-7,
We also wanl 10 add a feedback controller 10 adjust the split fraclion of the
purge stream, to keep the hydrogen-to-aromalics ratio equal to 5/1 at the reactor
inlet_ However, the RCNTL block will allow us toconlrolthe hydrogen-Io-toluene
flow. bul not the ratio of hydrogen to the sum or the t-c:nzcne and toluene. We could
390 nCTION III MATUIAL SA1.<ONCE CALC\JL>\TION3 sEn ION III M... TEIIAl S... l.ANCECAll"1Jl..... nONs 391
Malerial Galances 'o\hen Benzene Is Recm'ered
from the Flash Vapor Siream
If we include a vapor recovery system on the flash vapor stream (see Fig. 12.2-8), we
obtain the results given in Tabk 12.2-8. Now if we compare the stream flows in
Table 122-8 .... ith the results of the approximate balances given in Table 12.2-1, we
sec that the errors are quite small, i.e., within a 2 to 6% error. Errors of this
magnitude can easil)' be tolerated when we are screening alternalives.
The S1ream flows calculaled by the program given in Table 12,2-5 are given in
Table J 22-6. Of course, we could decrease the error in these material balances by
requiring a lighter tolerance on the controller blocks. However, the results are
adequate for a second-level screemng of Ihe process ahernall\'es.
It is interesting to compare the stream flows shown in Table 12.2-6 with the
apprOXimate results we obtained in Table 12.2-2 (see Table 12.2-7). The approxi-
mations gi\ell in Table 12.2-2 correspond to a worst-case condilion, where we
made no attempt to correct for the fact that a significant amount of benzene leaves
Ihe flash drum "Ith the vapor stream. Desplle this error, most of the approJlimate
stream flows are "uhln 3 of theIr more eJlact values, whIch is adequate for
screening calculallons We base our error anal)'sis on the totat flows because most
or the heal loads and equipment sizes depend on only the total flows.
l\"1aterial Balances for Ihe Base-Case Flo"'sheel
an altempt hl ...:onlcrge Ihe product now controller. This nested Iteration procedure
is oonllllued ulllil both Ihe SCVW block and the CNTRL block arc converged. At
this point. a change IS made to converge the RCNTL block. Then the whole
procedure SUHb again unlil both the SCVW block and the CNTRL block arc
con\erged, a/llithen a second change made to allempt to the RCNTL
block
nC3ted Iterations WIll greatly increase tile computer tIme (and cost)
required 10 obtam a solullon Therefore, we suggest that illitlally tile convergence
tolerances be sel to 0.001 multiplIed by the desired value of the property being
controlled, rather than lhe default value ofOJIOOOI.
Discussion of Ihe Approximate Malerial Balances
Hopdully we have demonstrated the magnitude of the errors that can be expected
from using our very simple, approximale material balances. We use these approxi-
mations initially to see whether the process has any chance of being profitable or
whether the project should be terminated. If further design elforl can be justified,
we use the approxImations to identify promismg alternatives. However, we would
always use a set of more ngorous material bal:lnces obtained by using a CAD
program for final design calculations. Our improved balances are still not rigorous
o Compressor 521 Purge
...E
GCOMP SPLIT
/
515
RCNTL
/
r----
RCNTL
f-_/
502
I
'i'
I
I
I
FMIX
*
Reactor

CONVG

Flash
ADD REACT 5evW
IFLSH
513
512 503
TOLFD h
Pump Valve I
MULPY PUMP
AFLSH
I
510 SO,
530
PCONT
T
CNTRL
-------
S04
Tower 3
I-
Tower 2
I"
Valve 2 Tower I
-
SEPR SEPR AFLSH SEPR
S09 506
511
overcome this difficulty by writing the code for a new block that would sum the
toluene and benzene flows al the reactor inlet and then using a ratio controller.
However, since our goal here is merely to illustrate how a CAD program can be
used to refine our approximate material balances and siDec: we probably would not
want to allow this much benzene to be recycled, we simply add a ratio conlroller for
the hydrogen-to-tolueDe flow; see Fig. 12.2-7.
When we add these feedback controllers lD FLOWTRAN, we nest the
iterations required to converge the calculations. That is, suppose it takes five
iterations to converge the material balances, i.e., the SCVW block. After these
calculations have converged, a change is made to altempt to converge the product
flow rate controller. After this change has beeD made, it might lake another five
iterations to converge the SCVW block again, and then a second change is made III
nCURE 1U-7
HDA oontrol sysleIIl$.
ill
TABLE 11.1-6
N
HDA plant mlltrl.1 l n ~ n
HDA PLANT - MAT!RIAL BALANCES - CONTROL SYSTEMS
STREAM NAM!:
501
502
503
504
505
LBMOL/HR
LSMOL/HR
LBMOL/HR
LeMOL/HR
LBI"IOL/HR
1 HYDROGEN
1577.14
1575.66
1.48202
1.48202
1.48202
2 METHAN!
2399.24
2388.12
11.1190
11.1190
11. 0634
3 BENZENE
296.602
29.7555
266.847
266.847
0.80054
4 TOLUENE
92.8057
3. 47663
89.3291
89.3291
0.0
5 8 I PHENYL
4.61002
0.00040
4.60963
4.60963
0.0
TOTAL LBMOL/HR
4370.40
3997.01
373.386
373.386
13.3460
TOTAL LB/HR
74096.5 44131.3
29965.3
29965.3
242.996
1000 BTU/Hft
71543.00
2808.79
-4621.20
-4621. 20
16.24
DEGREES r
1265.00
100.00
100.00
98.89 142,49
PSI"
500.000
465.000
465.000
150.000
150.000
D!NSITY,
La/rT]
0.4543
0.8637
52.2572
0.0000
0.4292
"OLE ruc VAPOR
1.0000
1.0000
0.0000
0.0202
1.0000
STREA" HA"E:
SO.
507
508
S09
510
LBI'lOL/HR
LB"'OL/HR
LB"'OL/HR
LB"'OL/HR
LBMOL/HR
2 "I!:THANE
0.05560
0.05560
0.0.5560
0.0
o. 0
3 BENZI!:NE
266.046
266.046
264.716
1.33023
1. 33023
4 TOLUENE
89.3291
89.3291
0.08933
89.2398
88.7936
5 BIPHENYL
4.60963
4.60963
0.0
4.60963
0.02305
TOTAL LB"OL/HR
360.040
360.040
264.861
95.1796
90.1469
TOTAL LB/HR
29722.3
29722.3
20685.5
9036.72
8288.36
1000 BTU/HR
-794.97
-794.97
-2663.32
-814.03
-752.64
O!GRE!S P
370.21
193.69
168.83
234.37
231.20
PSIA
150.000
15.000
15.000
15.000
15.000
ONSITY, LB/PT3
42.6642
0.0000
51.0575
49.2994
48.6827
/'IOL! rRAC VAPOR
0.0000
0.5458
0.0000
0.0000
0.0000
STREAM NAME: Sl1 512 S13 514 515
LBPlOL/HR LBI"lOL/HR LBMOL/HR LBMOL/HR LBMOL/HR
1 HYDROGEN 0.0 0.0 0.0 467.900 138].40
2 I"lETHAN! 0.0 0.0 0.0 24.6000 2096.72
3 BI!:NZENE 0.0 1.33023 o. 0 0.0 26.1247
4 TOLUENE 0.44620 88.7936 279.375 0.0 3.05241
5 BIPHENYL 4.58658 0.02305 0.0 0.0 0.00035
TOTAL LBI'lOL/HR 5.03278 90.1469 279.375 492. 500 3509.30
TOTAL LB/HR 748.361 8288.36 25740.0 1337.92 38746.4
1000 BTU/HR -14.09 -752.64 -3821.65 34 3.96 3157. 50
DI!:GRI!:ES P 405.12 231.20 100.00 100.00 124.65
PSIA 15.000 535.000 575.000 575.000 5]5.000
Df:NSITY, LB/Pr3 54.2040 48.6827 5].1678 0.2543 0.9497
MOLE PRAC VAPOR 0.0000 0.0000 0.0000 1.0000 1.0000
STREAM NAME: 51. 517 520 S21 S30
LBMOL/HR L8MOL/HR LBMOL/HR LBMOL/HR L8MOL/HR
1 HYDROGEN 1851. 30 1577.47 192.261 1383.40 0.0
2 METHAN! 2121. 32 2399.74 291.398 2096.72
0.0
3 BI!:N7.!NE 27.4550 296.695 ].63016 26.1247 0.0
4 TOLUENE 311.221 92.8054 0.42422 3.05241 400.000
5 BIPHENYL 0.02340 4.61146 0.00005 0.00035 0.0
TOTAL LBMOL/HR
4]71.32 4311.]2 487.714 3509.30 400.000
TOTAL LB/HR 14112.6 74112.6 5384.89 38746.4 ]6853.6
1000 BTU/HR -1072 .83 11719.65 342.73 2466.06 -5411.70
OEGREES r 126.11 1265.00 100.00 100.00 100.00
PSIA 535.000 500.000 465.000 465.000 575.000
DI!:NSITY, LB/rT3 0.0000 0.4543 0.8637 0.9637 53.1678
MOLE rRAe VAPOR 0.9111 1.0000 1.0000 1. 0000 0.0000
~
_, IOU,., ... ..... , ...""L .......... NCI. CAl CUI.A lIONS
TAill t: IU-1
SCream composition!!> .lIDd
A......o.illl.llt
C_..- II, T...... Gu
Liot-
RuOf
COnlpo6itiotl
,... ,...
reClck rftyd,t ..
II, 0" 0 0 .. 0
0'"
CII, 00' 0 000 0
0'"
Il<=M 0 0 0 0 0
Toll>t:lloC: 0
"'"
0 100 006.
0 0 0 0 0
TOlalllow 492.j '2134 H1S
"1
4231
0-..- Ra.. ......
,....
Gu
COnlpoUl;o.
...... liquid ....... cyck
II, 0.]66 OJ96 0006 0.391 0.391
CII,
..,.,
0'"
003'2 ..". O.H
Il<=M
.06'
0008 o'OJ .000 0.008
Toluene 0021 0001 0.246 0001 0.01
0001 0 0012 0 0
TOlalllow 4232
'"l
JlS
'"
))<J'
n.owrRAN
C_..- II, T...... G.. Liot.. RutlOl'
tomlJ'O'il.....
,... ,...
r)-d. r.t}'ck ..
Il,
0"
0 .39' 0 0423
CII, 0-05 0
0'"
0 0485
Il<=M 0 0 0007 0.15 0.006
Toluene 0 100 0001 09115
0""
D'phcn)l 0 0 0 0 0
TOll" 110. <192 j 119 ,
"'19
""
4371
&ror 10 80... % 0 L<
"
II
"
0-......'
RUtlOf .... Flub G..
COfI\poMlioa
......
""'r

II,
036' 0)9'
0"'"' 0". 0.391
ai,
0'"
0600 0010 0-597 0594
SenuIM: 0068 0001 0715 0.007 0.008
Toluene: 0021 .001 0239 000' .001
o.pbtnyl 0.001 0 0.012 0 0
TOIaillow 4371
""
37)4 487.7 J39J
Error III 80.... "
,.,
L8 10.2 l.J
because we are DOl ocreain Ihal we can design d,stillallon columns Ihal bave exaclly
Ihe splil fraclions thai we have assumed.
An inspecllon of our approxImate and exact resulls Indlcales lhat Ihe major
error in our approximale calcul.uions IS caused by the flash drum; i e. our
shoncul calculations are based on a perfect vapor-hquld spilt. so Ihal we did not
properly aecounl for the benzene and toluene leaVing In Ihe flash vapor and lhe
Compressor 521
Purge

GCOMP
SPUT
/ t 519
RCNTL
/
5'5
_/
BZREC
r
r-----
RCNTL
SEPR ,
sr
I S02t
I
,
FMIX
Is',.
ReaclOr

CONVG
SO;
Flash
ADD
REACT SCVw
IFLSIi
513
512
S03 518
TOLFD h
Pump
Valve 1
-
MULPY
PUMP
AFLSH
I
t
510
S08
530 rPCONT
i'
-------
CNTRL SO<
Tower 3
Tower 2
1;;<'7
Valve 2
Tower I
SEPR SEPR
AFLSH
SEPR
S09
S06
511
FlGURE Ill"
HDI< prOOllSS wllh bc:nzcnc
hydrogen and methane dissolved In the flash hquld. However. our flash
calculations did correctly indicate the magnitude of Ihis error, although II would
ha"'e been tedious to converge the flash caleulalioDs by hand. Despite the error, Ihe
sboncut calculalions arc: sufficiently accurale for screening purposes.
A second source of error in Ihe approximate balances was due to the
assumption of complete recoveries. The losses depend on the column sequence we
choose, as well as design of Ihese columns, which is Ihe reason why we neglected
the losses in our illJlial calculations. However.lhe losses are only a small souroe of
error in our screemng calculations. . .
Now we have developed an improved set of material balances, H IS slil1
oc:cessary to check lhe approlumallons ....e made (0 calcul.. te the energy balances.
Ibe equipment sizes, and the equipment C051S. We briefty diSCUSS these problems In
the remalOder of thiS chapler
SECTION I!J COMPLETE P'LANT S1MUU.II0" 397
..

;:;:g
H
- .
" .

... ... c ....
oc:coc@
...

-
12.3 COMPLETE PLAI'o'T SIMULATION
Once we: have completed a SCI of material balance calculations. v.c can work on
de\'cloping a simulation of the complete plant. Again. it is advisable for a begmner
to pUI the program together in small chunks and 10 debug these small chunks
mdlvidually, rather than allemplmg to immediately solve the complete: problem
We: use the results of Ollr material balance calculations 10 til the nows for
small subproblems
Some examples of the ly1X'S of studlcs thaI norma III are useful 10 undertake
arc: discussed below. Then we discuss a program for the complete !fDA proce<;s
DiSlillalion Column C.kulalions
Our material balance calculations were based on splil (SEPR) blocks to describl=
Ihe behavior of the dlsllllatlOn columns. With these bloch "'e merd} specify the
fraction of each component taken overhead, so the results will not necessarily
correspond to the distributIon of the nonkeys that we would obtam with more
ngarous models of dIstillatIOn columns as the next step 10 the de"e1opmenl
of a complete simulation. we might de,-elop a Simple program for Just the
distIllation seclion. "'here '''..e replace the of split blocks by FenskeUnderwood
Gilliland routiDC5 (or lheir equi,alenl): see Fig. t2.3-1. The Fenske-Underlo1.ood
Gilliland calculations (called DSTWU blocks in FlOWTRAN) will also ghe us an
estimate for t he column de5igns lhat we can use to check our shortcut calculations.
lbe Fenske-Underwood-Gilliland design procedure is limited to syslems
ha,-mg a constant relative "olaulny Thus. the calculalions mIght be 10
error for nonideal mlll;lurcs where there is h}"drogen bonding aDd where the activit}
coefficients are large. Similarly, the assumption of a constant rdali,-c: ,olalility IS
valid only when allihe components in a mixture have essentially the same heat of
vaporization (i.e_, are boilers). Thus. there are often cases where we want 10
check Ihe accuracy of the Fenske-Underwood-GiJliland estimates by using rigor
ous column models.
Most of the rigorous. lray-by-tray programs are simulalton programs. That
is.. the number of trays atxn-e and below the feed must be as well as the
reflux ralio. and Ihen the subroutine calculates the splits. However, when we are
designing a column, we are trying to detennine the number of trays that we need to
accomplish a desired split. Therefore, normally we need 10 usc an iteralive
procedure when simulation subroutines are used for design.
Experience in<!lcates that these traybylray subroutines are orten slow to
oonverge (i.e. they have the longest running times aDd require the most iterations
of all tbe subroutines availabk). For this reason, if we include rigorous column
models in a program that is also using ileration to close the material balances
calculations. the computing costs are orlen excessive. )lence, in sequenlial modular
simulators, we usually use the tray-by-tray routines in separate programs; i.e.. lo1."e
use the results from ngorous dlstillalion calculations to adjust the split fractions III
slITlple models in a material balance program or complele planl simulation
>D
0:<:000--'
-
.
"
00000"'-'
o
00

... - ....... -
"!
00000--

z
<

g
"
--
00

8
ooe-c-

..... <:>

.
i
..
i.
<
,
,

c
.lI
3%
398 UC'llOI'" IJJ COIotI'IUI! ..L.....rr SIIotUl.ATION Sl:cnoN IU COMPlETE PUNT SIIotUlAnot< 399
By-product
u
i
SEPR
Recycle
ReaClor
Recycle
Product
r-
I
I
By-producl
(a) FJowslJ<:<:1
Produci
(b) Computer mfonmlion dIagram

o
511
Compressor 521 Purge
...2
GCOMP SPLIT
515
S02
I'
FMIX
1
51

Reaclor

CONVG

A",h
ADD REACT 5CVW IFLSH
t
13
512 S03
Pump Val ... 1
PUMP AFLSH
510 SOB
1
5
S04
TO'M:r J
_, I I
S07 Val ... 2 Tov,cr I

DSTWU

AFLSH DSlWU
S06
FlCUR l2...>t
HDA pr_ WIth colwnn modds.
5
5
AnOlher approach that we could usc: is to develop SImulation programs that
avoid including rigorous models for dislillalion columns as part of any matenal
balance recycle loop. That is, any time we have a distillation column in a material
balance recycle loop, we can double the flow rate of Ihis stream (using a MULPY
block in FLOWTRAN), IheD spht the stream in half (using a SPLfT block in
FLOWTRAN), and lhen usc a SEPR block to connect one of the exil Sireams from
the spliller to the recycle portion of the f10wshcet and a rigorous model on the
other half of the stream. Since we can COntrollhe order in which lhe simulalor calls
the subroullnes, we can usc thIS approach to complelely converge Ihe material
balances before we call the rigorous dislillation routine. A Simple example of tillS
type is shown in Fig. 12.3-2.
FleURE 12.J..2
A .wmpl( planl
Energy B.IIl.IInces and I-Iut Exch.llngers
In the material balance computer information diagram in Fig. 12.2.7, we neglected
the quench stream thai uscs flash liquid to reduce: the temperature of Ihe reactor
effluent to II50
c
F. An lOspcctioll of the flowshcet wilh the quench stream included
indicales thai the quench sHearn merely provides a recycle loop arouod the flash
drum. Thus, If we mal.e a material balance from the reactor effluent 10 the flash
vapor and the pressure reduction valve before the stabilizer, i.e., if we IOciude lhe
quench-recycle loop completely wlthm this balance, Ihen the process flow rales will
400 Reno,", I1J COM .... nE PL",n SlMUUT)()fol
S02
S06 Q miJ. S07 CONVG SOl Flash
Re3CIOT
ADD

SCVW IFLSH
(SI7)
/QCONT
S03

Q split
SIl5 SPLIT
S04
flGURE Il.3-J
QurrdI C2kulal>oa
not change. However, the heal dUI} of the parllal condenser will depend on the
quench now rate.
To calculale the qUl'nch flov.. rate, and the load on the panial condenser, we
must adjust the flo" r3te of the quench stream to decrease the reactor uit
temperature to Thus, we need to lDstall another controller that \\'111 soh'e
the problem Ilerativel). sec Fig, 12.3-3. Again. a beginner is advised to solve this
problem separalcly and to make certain that convergence is obtained before
attempting to add all iteration loop to a large program. We use our shortcut
calculations as a point to con\'erge the calculations.
CAD programs also make il fairly easy to geocrale the temperaturc:-enthalpy
curvcs for each process S!ream that are needed for the energy integration aDalysis.
In particular, whell there is a phase change in a stream containing a miJ.ture, the
temperalure-enlhatpy calculations are tedious to undertaJcc by hand. In the initial
simulation of a complete plant, we would illclude only heaters and coolers on the
streams and then design the heat-exchanger network using the procedure described
in Chap. 8. However, I proocdure for incorporating the heal-cJ.changer design
procedure into a sequential modular simulator has been presented by Lang.
Biegler, and Grossmann
\' D. Lang. L T end t F GrOWllollrta, Stm\lh.eneow OphmwotlOQ end IIul
W'nh PJoa:ss Sim\llation.- Pern no 12b. t986 AIIIIUlI AtCbE Medin" M\f.m. Bcadl. I98&,
10 e-,..'<'n;" C/wmiC..1
SECTlO!'I IlJ COIolPLTE PlANT SINUL\T'I()l'I 401
Complete Plant Simulation
AfieT we have used simple studies to make certain that OUT simulator subroutines
WIll converge and will give the correct predictions. we can put these subroutines
together to generate a plant simulation. Table 12.3-1 gi'o'es a program for the HDA
process that contains a fced-effluent heat eJ.changer{see Fig. 123-4). Since we ha\e
already calculated all the process flows. we can teaT as man) recycle streams as we
desire in this flo.... -shcct, and "e do Dot need to mclude any controllers. With this
approach, "e can calculate the equipment SIZCS and the required utility flows with a
mmimum of computational cost. Of course, if we want to change the processing
conditions (i.e.. the values of the design variables), we must include the controllers
in the program.
Cosl Models, Process Profitability, and
Optimization
Many simulators, including FLOwrRAN, include equipment cost correlations.
To use these correlations, normally It is necessary to include a factor that accounts
for inflation, such as the Marshall and Swift indeJ. that "'e used in our models.. By
supplying the: unit costs for cooling water, steam. etc.., it is also possiblc to calculate
the costs for the utility Slreams. The capital and operating costs along with the cost
of labor, maintenance and repairs, lUes and insurance, elc.. (see Chap. 2), call be
combined to obtain an estimate of the profitability of the process.
Each Simulator run corresponds to a single set of design variables. One way
that we could estimate the optimum design conditions is to make a set of case-
study runs that correspond to the rallge of the design variables where our shortcut
dcslgn cakulations indicated thai the optimum was fairly flat. Many simulators
also include oplimizalion routines Ihal caD be used to find the optimum.
Remember thai costs change over tbe years, and Iherefore the optimum
desigll conditions will change. Thus, at the end of the 3-yr period, or so, thai il is
required to build a plant., the optimum design might be differenl from the final
design that is approved: for CODSlrucUoD. For this reason, some: Ihought needs to be
giveD to the tkJoibility of tile process to meet changing conditions in the economic
environment. .
Whal Remains to Be Dooe
Dooc a set of CAD calculations bas been used 10 verify the seleclion of the best
process alternative, the conceptual design effort has been completed. Howe\'er, it is
still necessary to de\'elop a control system for the process, to consider the safelY
aspects of the process, and to add a significant amount of detail associated with the
final design. Safely and conlrot problems Ihat arc discovered might require the
basic flowshcet 10 be changed again. For this reason, our initial CAD studies
should focus primaril)' on finding the best prOttSS alternative and the cost penalties
associated with other alternatives. Some additional discussion of safety, control.
etc_. is given in Sec. JJ.3
T"aLE 113-1
I-IDA process
TITLE HDA fLOWSHEET
PROPS S 1 2 S 2
PRINT INPUT
HYDROGEN METHANE BENZENE TOLUENE BIPHENYL
FLASH IFLSH 501 503 502
Q5PLIT SPLIT 503 504 50S 5*0
VALVI AFLSH 505 6*0 506 0
TOWRI DI5TL 506 508 507
VALV2 AfL5H 508 6*0 509 0
TOWR2 DISTL 509 511 s10
PCOOL HEATR 510 540
BLOCK TOWR3 DISTL 511 513 512
PUMP PUMP 512 514
BLOCK PURGE SPLIT 502 530 531 5*0
BLOCK GCOKP GCOKP 5]1 517
BLOCK FKIX ADD 514 SIS 516 517 3*0 518
BLOCK FEBE 518 52] 519 524
BLOCK FURN BEATR 519 520
BLOCK REACT REACT 520 6*0 521 0
BLOCK QMIX ADD S04 521 5*0 522
BLOCK CONDS BEATR 524 525
PARAH fLASH 1 100 465 0
PARAM Q5PLIT 1 2 .2805 .7195
PARAM VALV1 1 150 0 0
PARA" TOMB1 1 .7 6 5 2*150 .0357 0 1
PARA" VALV2 1 15 0 0
PARA" TOWR2 1 1.8 24 12 2*15 .735 .5458 0 0
PARAM PCOOL 1 100 ]*0 1 0
PARA" TQWR3 1 .1 4 ] 2*15 .9472 0 0 0
PARAIt PUMP 1 5]5
PARAM PURGE 1 2 .122 .878 5*0
PARAR GCOMP 1 535 0 1 0 .8 .8
PARAM FEBE 1 1100 20 10 15 2*0 2
PARAM fURN 1 1150 15 0 0 1 0
PARAM REACT 1 1265 SOO 0 2 4 .75 -1 1 1 -1 21*0
PARA" REACT 32 3 .03 1 0 -2 0 1 74*0
PARAR CONDS 1 100 5 0 0 1 0
MOLES 501 1 1577.53 2402.98 ]92.33 124.8 6.24776
MOLES SIS 4 279.375
MOLE5 516 1 467.9 24.6
KOLES 523 1 1577.53 2402.98 ]92.33 124.8 6.24776
TEMP SOl 100
TEMP 515 100
TEMP 516 100
TEMP 523 1150
PRESS sOl 465
PRE55 515 535
PRESS S16 535
PRESS 523 500
END CASE
END ,JOB
402
it::l

.-

:8
.t-
5l

0 :t-
5l
-:z:

i-"2-

'" ".J
0.-
f-
.<e;
.J
u.!:
.u.
0'" ><

L
-
0
g
'"
-
'"

-.J
r-
5l
"t-
r-
Cl

r-;:;-
EO

N
0<
5l
"''''
'"

-
N
'" oUJ
u:z:
Hj
N:Z:
"<



.u.
"''''
><
t
.J'"
g
T
/
8t< '"
"'"


"

5
UJ
I
N.J

u.:z:
"0
N
"t-

'"
p:;

f-

0
-
U '"

:I::I:
UJU
I

N

'"
N



E::1
'"
"t-
'<'"

Q

'"

Eo
-
u.<
'"

-
'"
;;

..
403
11.4 SUMMAR\ AND EXt-.N-CISES
Summar)'
The large CAD programs. such 11\ fLOWTRAN. PROCESS, DESIGN 2000,
ASPEN, etc., are f'Olo\erful tools. tout lhey are somewhat 1lIousto usc. It is easy to
make mistakes in the mput data. :md lhesc= mIStakes can be costly In terms of
compuler lime IIcOO'. the best ;'rproach to de\'e!opmg a CAD program is to
conSIder only small porllons of the plant at one time and 10 debug Ihe code
correspondmg 10 Ihls small part ..I the plant Then we add another small portion,
and gradually we generalc a code rllr the: complete process.
To use these programs effioClltly, I.e., to minimize the number of iterallOns
requIred, it is usually necessary III ha\'e good estimates for recycle flows, the splilS
in purge streams., reactor conver.;iulis. etc. We use the results from our shortcut
,calculations to prm'ide these estinwlt:S.
":xerc.ises
12.4-1. Dc\'dop a m3lcnal balance r ....l/-lam (usin& ellher FLO\VTRAN Of" another CAD
program that you mlghl ha\"C' ",,,,lable) ror the IIDA prooess'" Ith dlphenyl rttJckd
11.... 2. For one mtne f"Il)C('S5t:$ thai ha"e deslgned.de,dop a ngorous matenal balana:
program Also. de'"elop .I CAll rrogram rOt the distillation sequence. and then
develop a program ror the ....n'rJete plant Ho,", do the cakulallons
oompare "'lIh YOUt shortelll arpl,,,lmatlOOS'
CHAPTER
13
SUMMARY
OF THE
CONCEPTUAL
DESIGN
PROCEDURE
AND EXTENSIONS
OF THE
METHOD
We havC' described a systematic prlJlCedurc for the cona:plual design of a limited
class of pt:lTochemK:a1 processes. i.e. conllnuous, vapor.liquld processes that
produa: a single product. Of course, many other types of processes could be
considered. Moreover, numerous other types of design studies need to be under-
taken to complete a final design
Unfortunately, It 15 oot poSSIble (0 cover all this matenal to a one-semester
cours<:. Petrochemical processes an: selected for consideration because IhC'y are the
most common Smularly. lhe emphaSIS is placed on conceptual design becau<;c
the equlpmenlused In the process and the Slructure or the nowsheet are filed
stage or thc design :tctIYltv; 1C'_. all the other design actIvities depend on the ..
of the
406 StCfION Il' TilE lUUA1CWOl. Ole'S'ON rIOCI'OU1" fOI PETltOCllUI!CAl HO(:uns
The syslematic procedure v.e used to develop a concc:ptual desIgn was
hierarchical. A brier review or this procedure is gi\'en in Sec, 13.1 Drier outlines or
hierarchical procedure!> thai can be used 10 develop com:eptual designs ror
processes and batch proces.-.es are given in Sc:c. 13.2, Finally, some olher types or
design problems that need to be solved berore a final design call be dcvdopc::d are
briefly discussed III Sec. 13.3.
13.1 REVIEW OF TilE HIERARCHICAL
DECISION PROCEDURE FOR
PETROCHEMICAL PROCESSES
To Simpliry Ihe conotplual design or a process. we decompose Ihe problem
1010 a hierarchy of decisions. TM decision levels Ihal v.e consKier are gl\en m
Table 13 I-I The decisions lhal need 10 be made al each le,'e1 ror petrochemical
processes are given in Table 13.1-2.
The mpul-oulput ill!ormallOn requued 15 presented ill Sec 4 I. and the
beurislIcs that are available: 10 help male .he dcasions presenled ill Table: 13.1-2
are discussed in Chaps. 5 through 8 Jr no heunstics are a,allable to male a
decision, we merely make a guess. We go through lhe complete design procedure m
IbIS way, 10 generate a base-case design. We try to develop a complete des.gn as
rapidly as poSSible to soc whelher there IS some reason why we should ter:TlInale all
..... orl on Ihe projecl.
As \loe procec'd through the base-casc design. v.e kocp Irad. of the dCClSIOns
.... e male In addition. we prepare a COSI dIagram ror Ihe base-case deSign (!\.CC: Sees
9_1 and 92) as an aid IIlldenllf)mg Ihe mosl expenSI\'e processmg costs Then we
allempt to evaluale hov. changes m one or more or our original deciSIons Will atree!
the procc:ssmg COsts; see 5ec. 93. We continue to evaluate process ahemall\.esin
IhlS way until we oblam the besl process allernatl\e.
It mIght be necosary 10 change tbe flowsboci correspoodmg to the besl
process allemative because or sarcly, sian-up. cootrollabihty considerations, etc.
TABl.E 13-J.1
Hierarchy of decisions
I Balch ycrsus conlmuow
Inpul'OutptlllUUCIUre of.h" 1l0wWccl
3 Il:ccyck ilructurc of 11K llowlhcrl
Level. Gcneralllruc1ure of.be scparlnoll syslem
a. Vapor sys'em
" LIqUid rocovery syilern
L.evcl5 EncllY mlci,anou
F,am J M Oou&l... A/CAE J. JI. Uljl9.Sj
SECTION ll' IIII' IIlUUCIllCAl DEClS,ON PIO<;OUU tOl rnOCllEMICAl. PlOCSSES 407
TABLE 13.t-l
Design de4:isiom. for conlinuoUS
l.cnl I 8illdl COnUnuOw. w""dcr only COnllnl>OUll prOClCSKS
2 Inpul-Otllpul
I we purify .be ...-m;al"nal $!reams be(ore lhey ar" rcd 10 r<:Klofl" If lhe
,mptlnllcs .'c lIlul, tberc no quanllla"'e heunstla.
l by prodUCI be n:co\ered or recyc\cd 10 ClullCtlOn
1
- No quanUla1JVl:
beumllc 'I
3 .." need las rcqck and ptlfJC A beUtullC 10 be
aYOll!lbIc bcfor" lhe rceo:nt In.'cntlQIl or -.nbrane KpantJOIl prOCC-I$CI 10 $rj!MalC
pseous mUlufrl
.. -II 01 from au 01 H10. rcKllnt .hal 1$ nol recovered and (AD ucas amoulll
musl be spcaficd.)
5 -Ho"" m3Dy produci suuml WIll there Rc:uonablc: """,m 10 be anllablc,
uccpl for r1lc: ...... 01 a by-prooUCI-
Levd 3. Ra:ydc ilt1lClun:
I -H_ runy fcacror $)'51rrDS alc T'bc heUrlSUai aern to be (UIOIUIbIc:. -Ii IIxre
an) IoCp",allOll bel....ec. tbe Uluall)' I deowon r:an be made bued 011 Ibr
dala
1 -Ho..- man)' 1CCJC\oc ,nama I'C lleurul""'lr" available.
3 -SlKKild..-c ..... an C.u:Gi 01 one Normally u..: dal.ll .... mdK:::Ile lhe
am;"e'
4 -Ii a lura:ydc c:oroplUMl' A brunsllC iii "-aJlabk.
5 lhe lcacror be oprraled .chabllllcall)'....llh dum bcaWllL (or coobnlll. 01 a
(but carlKf) Some calc:u.btlOlU.arC na:ded 10 U$r the bcuftillC
II -Do .. e ... anl 10 Ullfhe equl1,bnum coaYCBKNl1- Calcub._ Ind jtId nl alC
fc-qw,cd
.... Srparallcn ')ilna
I -What 1i.IM: 0I1he and liquid Iyslem'- H"UllSIIQ lie ..'&liable
.... \'apo' ,),.em
I k 'hal '111M: IocaIIOn (>f.he VlIpor 'CCO\cry A heunsllC 'Iava.ilabloc
2 k lUll u 11M: bQI IH'" oI.-.po, fcalYCry ')'Rail to w.c'- No heunmcs arc
Level 4b lIqwd ....pal1lllOn
I - ...."ha, call be made by dlliuJlauool- A heunsllc that Uiua.ll)' III
"'lIJ.i.l1b\oc.
2. -Whal seqUCllDl: 01 dtllll""on c:olumm ,bould be uocdl- TIle publllIhed halnst"" arc
bmllcd 10 ihatp ,pllli of ldul mu,lurco to. I IInS'" rctd. but ,n man)' arKS thcydo nOI kad
10 the bQt ocquencc Thus, ClIlculauonl arc required.
3 k Ho.. ,hol,lld lhe hahl cnd. be Calculations. and Judgmenl arc reqUited
4 lhe: Irghl cnd. be Knllo f"",l 0' rcc)'cled to the vapor fceo""')' S)lI(IJI'-
Calculanoru and JudiJfICni ale
S -Ho.. ihould wt: aa:omphih lhe other $rplltaIlOlls'- No heuriillcs arc available
Hcal-c.chanlCr a deslgll proc:edurc iii (KC Chap. II)
F.OIll W " '-Uobc M F Dobor"y. and 1 M Doualu. kSc,ec..". c( P'C>CUI Ah.....".a," /.a.tC
,nl"<
Moreover. the accuracy or the prdlnun:lIry desIgn calculallons oc:cds to be:
to oblaln more accuralr estimates or the costs. SImilarly, other c:quipmc:nl such as
pumps. drums, storage tanks. etc.. nud to be added. Thus. the conceptual design IS
merely a startmg poinl ror other design
Changmg economIC conditions normally cause changes In the types or prOtt<;scs
that .....e bUild. At there is a growing Interest in the: design or batch proct'sses
ror both speciality chemicals and biotechnology. The design or bioprocessc:s is
discussed by Bailey and Ollis.- A review or their chapter lIon product recovery
operalions indicates that solids prooc:ssing units (crystallization. filtration. and
drying) are very common. Similarly, solids processing steps are commonly encoun-
tered in polymer processc:s.
In this sectIon. we present brief outlines of systematic procedures thai can be
used ror the conceptual design of solids prooesses and batch proccssc:s The
prooc:dures are arranged mto a hierarchical structure, similar to the procedure
presented in Sec. 13.1. Howe\'er, new types or economic trade-offs are encountered.
and orten new types or constraints must be: considered.
13.2 DESIG
AND BATCH
OF SOLIDS PROCESSES
PROCESSES
TABLE IJ.l..I
Input inrorm_tion ror soltd processes
I. ProdUCl$
D Duucd p<odll(:l!Of1 ralC' porny
II DemrC'd .... rl,ck sue (and dislnbol,onl and propertld
( of produn. 01 pntt vasus punly
, Valuabk byJ'fodocts. rf .ny
1. R.w mall:nllls
Compoontron rb,lIM;:111 SC.. lt of ..11 raw rn,artoak
II Pnee clach ra" m.. '_/' tlr pncc ",,"us punl)
J. Solid, IFnrr"lIotl
D A,,.jl..bk rntlbods for IFntIlIun, sohd product or c:hara<:tm,lJC:II
b Solubility d..l.. fo, product ..nd pouibk impunues
t. Reaction a'olchtometry (if any) .. r>cl dal..
'*- P'<>COSSInl COn!IU&1IIlS (Ihex wilt vary f,om ploceP 10 p'OCZ5a, but Iypically indudt)
... Producl Innp"rahu't connraiob due 10 tbtnnaI ....tabihl)'
" Crystalhz:tr (pncprlalor) dulTJ' dnosrl)' Iimilal>OllS due 10 a da::lioc ill producl ql....lIt)' or poor
no- propertleS.1 Iu&b IOblb cooo;arlrauon
5. Plaol and Slle dala
D. COSI cl steam Inek, ooobn, .... Ia. rd'riIF"'UOn. dc.
" Wasle dISposal f..cilllles .nd OOSlS
Solids Processes
OUf dIscussions up to thiS point ha\e bc:c:n hmited to \apor-liquid processes
Ho.....e'er, some petrochemical processes mclude solid processing steps in order 10
isolate the product. For example. thc separation or lIylene Isomers IS orten
accomplished by using crystallization inslead or distillation because or the close:
bolhng points or the Similarly. Ihe production or adipic acid includes
crystalh7.-3tlon steps. Of courSC'. If a cf)'Slallization step is presenl, nonnally
filtralton (we consider the use: of a centriruge as an a1temath'e) and drying are also
rc:quirro.
To include solid processmg sleps lD our synthesis procedure. it is nocr:ssary 10
moclJfy the structure or the separatIOn sp:tem, level 4. 10 include liquid-solid splits.
Actually. to make Ihe procedure even more general, gas-solid spilts and IlqUld-
liquid splits should be included (e.g.. gas-phase olefin production requires a gas-
solid split). The: process alternatives for crystallization (and/or precipitation), soild-
liquid separation (filtration, centrirugation_ settling, etc.). and drying. as well as the
unit operation modds, and cost correlations must be: added
An initial rramework ror a synthesiS procedure ror solids processes has been
published by Rossiter and Douglas.! The rocos or this initial work was on
.} E Balky and D FOllIS., R_""""",sI F.na"""...... FuttdtJMmltJ1s. 2d cd. MeG.a.. 11lI/, '.,.. "ork.
'986
'A. P RowIC!.nd} M ..... (""""- Cttg Drs.'" IH (1986)_ 60t I'" (19861
moderate- to hlgh-tonnag.=, conllnuous, morgamc processes that produce solId
products rrom liquid and/or solids feeds. The Input information required is
presented In Table: 13.2-1, and the decisions required are listed In Table: 13.2-2 An
example: or the application or the procedure to a design problem has been
published by Rossiter! and an application to a retrofit study was given by
Rossiter. Woodcock. and Douglas.!
Balch Processes
The design of batch prooc:sscs was discussed in Sec. 4.2 The: design of batch plants
rc:quires not only that we select the units to be used in the process and the
interconnections between these units, but also thaI we decide whether we want to
merge batch operations into a single vessel and/or to replace some batch
units by continuous units. Hence, the: design of batch prooesses is more dirrlCult
than the: design or a continuous process.
To simpliry tbe understandirtg of the: design of a batch plant, we start by
designing a continuous process, using the techniques presented in Chaps. 4 through
II r ROSSIter. (",.,., btg /',oe. 6'1 191 (1986)
tAP ROSlUIa. 0 C. W<><.>dcod:. and} 1-4 U"'" aI .. lh.,.. ...dual Procedure fC>!
Rroolil SlI.La or Sahib P!oor:sses,- Paprr preS.'llltel .. I lilt 1916 Anmtal AtCbE M""'hns. MIolIIDI
Beadt. Ncnembc:
410 SfCTION III I>f.SION 0,. SOLIDS l'aOCE.';SU "''''0 IlATCll PROCUSU
TABLE 1J,,2..2
Hierarchiul desision procedure for solid processes
I. Balch conIlIlUOU' prClCal we Q)MJdcr only CODllnL>QUS proc:c:sKS
1.
G Should We punty lbe rllw-malenaJ st.eams bc:fore P/ocesaD& Or should pr()(lCS4; tbe feed

" h a purge strum requIred"
c. 110Ilol mlIoy ploduct ate rc.qllucd1
What IS !he ce:ollOauc poleDtW (I produet value rwnus raw-mlltenal C06t OlJIlUS d"pog.l QKt
for ptuac aDd
). Recycle ,tfUelllre aod aysl&llun oolWlkn.uons: (wcll,dl.llg rcacuon, If any)
.... What l)'pC 01 eryl&&1lu.ct ihoWd be wed?
" Should the prodl>Cl-fof1lllJll reacuon (If any) tab plaa ....lhJD the aysr.aIluu. Of ICparalcly'
c. How lDUIy c:rysllllu.u clrccts Of lIap a.rc rcquucd?
How lDUIy rccydc 1Itc&lDS aK requucd7
Whal IS the :oPOaUC: ptMeaual the cconoauc poImU&l at level 2 IlI.lDUi rJ>c, SIlD1 oIlbc
U1Duahza;! C&pw aDd lbe opcrall.lll C05I 0I1hc c:rysWlua)7
.. xparaUOll 5Jl'C'QficabOll-lCvcraJ sobd-bqwd 5oCpar.lUODS rmghl be
G. flow cao the pnmary Pfoduet be rcal.erall
" What lypc:I 01 sobds recovuy S)'Slems e.n: reqwr-cd?
c. How dIould lbe wule-lOIod scparaUOOI be aa:omploshcdl
Arc aoy bquod-bqwd rcq...nd7
LocaUOol of ..,,.nbOll UDIU (pur. or r-yclc IUItlUI>S Of boc.b}l
f Whal lithe cconomM" poIclllAl (1.c..1hc CCODOIDX poIe.mal1.1 levd 3 mlllltS Ihc KpRnUoa S)'Ileal
(I.olluakud) captaland opcntrng """ minus bqUOf DlInw. .."Ulunl aODllI.lrud Coplll
aod opa;tlwi c:ow)7
So Produci dtyull
G. Whal 1)'pC 01 dfyc:r sbouJd be 1>ICd1
" What Ioucs call be
C Whal. tbc _IepolC!lIW (i.e.. lbe level 4 cc;onolWC OlJnu.s the aoDuaJu:ed capolal
and opctl.1l1l1 CUIi of the df)'er)1
.. EDell)' s)'SIems
G. lhe llUllltIl,ull hcaUOI and
b HDw lll.IJly helll ucban&en 01 whal JU.C ate rcqwtcd?
c Whal J:S the ccollOfllic polc:oual Ihc kvel S CCOIlOImC porcaualllllDus lbe alUluahud (:lpnal
lind DflUlIUlll 0061) of lbe bu.t..,achan&<=1 OC:lwolfr.?
9. Then we use the systemauc approach developed by Malone and coworkers- that
is given in Table 13.2-3. This procedure is also hierarchical, so Ihat a series ofsmall
problems can be considered that eventually lead to the best design.
OTHER STUDIES IN Tl-IE DESIGN OF RATOI PROCESSES. The design of balch
processes is expected to take on growing importance in the future. and for Ihis
0 lnbanell and M F Malone, -A Syslemallc Procedure for Barch p/ocr:as Synthes,s," Papel
p.-escnted 1.1 the 1985 Annllal AICbE Meetmg, ChlClgo.1Il ; C M Mynllbcas, MFlcalb,ltry and TallCU
fOI Balch PrClCal Dulgns.M M S Umvcnuy of MusachrucrUi, Amherst. 1986.
SECTION 111 OUIGN or SOUDS I'R()(;USU AND ....TCH PROCUSI!S 411
TABLE 1).2..-3
A hienlrchical procedure for Ihe cooccpluill design of dNicaled balch JWocesses
I. Design a oonlJOUDUS ptOOCti lir$t (If VOSSlbk). l>Slnllbc procedure: dc:scrihcd lD Scc. 13 I 1111.
p.ocedure 10 !lad lhe besl prOCCM Ille/nluve aad to .dc:oufy lhe domiOll-ilI desip va"lbIes If
ronunl.lOWi umts Ire DOl a ullble fOI WlIDC p.-ot;a",nl sleps, SIll" wltb the bestgucss of a lIo..-shccl
Ihal showli C&l:h step IOlhww:lIll.Uy
2. Repl..cr: eacb c:onunuoWo IIn.1 by. balch uml
ll. Include onI) an .mcrmedlile .Iora.., Iinl. fOI rccyc:Jc
" Calculale the Dpumum C)cle IUllC$ fo. I:aCh unll by IDIIl1mIZlIl' lbe 101al Innlll.l COil of lbe
complete process
(I) T1w; calctibuOlI prowrda I. bound (HI the C05I fOI the case wbcu the IDtamedllte slonlC
rcqutrCd tD sdxdule the plant IS free.. (If lbe plant IS DOl profil.l.bk WIth free IiIOraIC> II "111 ftOt
be profitable .-ben SlonolC IS mdudcd.)
(1) 11le resulll will proVIde IOInc measurc of the """"",,",IC" '1ICCII1IWC' for modaf)'Ul& tbc chemrsfs
=1"-
(3) NomullY. lhe"," of Adl opcrabOllllllbe opUJUUd t..1dl pnx:as will uODCd rhc CO$! oltbe
c:orrupood-inlllDll In lbe conr..._ plnl
(4) Tbc RSll-l1i arc used lala as I. Ill. 10 mupl 1lDlts.
c CaJadate the oplUIlUlll da:lp by SCII'IlIlhc cycle umcs of every UOlI equal 10 each <Mbcl
(I) nus cakuLalJOll I. bound fOI lbe t:mI .,ben there IS I IDUtlDUlIl eqWplDCll1
......-
(1) rhcrc ....n be _ lIc:llbtbty III the: dowp
(3) A....... I lQC:;UUrc: of tbc _ lDCICnll.-a; lot dlaOIPQI; lhe c:baausI'S/CIDPC IS nbllLlDcd
3. CoIUo>dcr D>CIPI adjaCCrll baldl UIl.IU for tbc dalp 111 2b
... Merll< lUlllS "Ith SUIUlaI cydc: IUlXS a.od SIU
ComjM-te the CO$l$ oIlbe 1nC1.,ed IIIUIi Wllb lhe QKIi 01 the compatabk oollUllllOllS UIUIS.
(I) If the C:OSIl 01 the ooOI1ll"'OUS lUUlI arc ebc:llpa. Ilfcll.lII Ihc: COIIUOUOlIS \l1UlS.
(1) 0rJ>cr,0.__ I.p rbe mcracd balc:b WIlli
c, COIIua...,- 10 mer.,. IIIUIi 110111 the (:0$11 u:l':rc:aa.
.. Cnw;odcJ the usc of (<< parallel mer..,d IIIUII).
c The aoaJ Ili 10 lDCIQ$C equlpflloCnt IlIWuUOIl.
" The rauo oflbe o;ydc IUDa mWit be lIIoIlcbcd 10 the Inverse: tal.rO 01 the Ilwt:lher of
c. NonnaJl). USC 1.1 _I Ih/ce parallel IInliS
S. Add Ihc lIucnocdll.le tr(>nle -tal 10 schedule lhoc planl and opumw: lbe dc:i.J.pL
.. Opuouzc the bett fto",st-t aJlcmallVC: IllChlodtDJ IIcraF-
1. Chcd Ihc opcrabilllY 0I1bc PfDCall, lISlnla balch SIDlulalor
F.1lm M F Maloot. penoa.al """"."IIQUOIl,
reason we are including a survey of some of lhe previous work. Most of these
studies consider fixed cycle tImes for the batch units, which makes them different
from Malone's approach.
Kelner- developed a procedure for mmimizrng the capilal cost (by using
linear cost correlations) for single-produci plants with a fixed flowsheet that
contain bolh balch and sc:mlconlinuous units. The cycle limes and size faclOrs of
the balch unilS were held constant, and then Ihe lrade-olf lhal balances lhe balch
S. Kelner. eM,., 121 (AUI 12, 1960)
412 UCTJOt< 03 oruEl. SIOt<lrlCA,n ASPECTS 0' ritE Of.SIGt' P"O.I.EM SEcnot< ".J mItEl. SIOt<lflC...m ASPECTS Of rHE OEs.lGN I'kOUfM 413
The goal of our conceptual design effort was to decide whether an idea for a DCW
process was suffICiently promising from an economic point of view that a more
detailed study could be juslified. If the results of this study appear 10 be promising,
Y I...oonbJ and 1 0 Robulson.. Irw! Or.... 1'rou lJG. 11, 166 (t970)
'J.O Robmson and Y R. Looabr. Pro TnII. 11I_,-t, 11 861 (1972)-
R. E. J Forckr, and 0 w T INJ EIIf Clemo. P'fKQ3 Drs. 14 191(1975).
'I E Grossmann and __ W II Sar,enL 1M. bog Cltmt PrOUH DttJ. Dr... II' 34) 0919)
, T TlkI"""uu.1 Ha.sh,moto. Ind S Hnebe. C"""P"f", J. 1115 (1919); 1",/ E1tg CNm
Pr",O.., 4)rlr9U)andlJ '10(1984)-
I A I Klml and G. V 31 1516 (19115) Ind JI 1528 (1985).
R C- S...... Qw. D \\. T R,pp,n.,d G J. Forder, (Min En{f. P 520 (1974)
.t B W Overturf, G " Ind J M WQQl!s. Jr.d Jog. e........ PHKUf lH, 11 166(1978J
"0 W T R'pl"n.Cl1<'If'",/", ....d(h,... FtIq.1 1\7(1981).nd146J(19lUl
W nlu_ Clontl JtQ P 11 (I d. 19l1fl) and p (July 11. 19l1)
equipment sizes agAlnSI lhe continuous equipment SizeS WAS evalualed Loooicar
and Robinson- considered the same problem, but used power-law e:Jtpressions
The}' also ellended the resuhs to multiproduct planu..'
A ht'uristic procedure to fil lhe sizes of the batch onilS in a muiliproduci
plant ha\mg filro c}"cle tnnes and filed size factors was de\doped by Sparrow.
Forder. and Rlppen I The number of unus in parallel was also used as an
optimization \'anAble In this stud} Introducing parallel units changes the structure
of the f1owshccl. and a very large number of possible solutions can be generated
Grossmann and Sargent
l
related the C}de time constraint. bUI assumed that the
C)'cle lIme .... as a function of the batch size.
Tnkamatsu. Ilashlmoto, and 1lascbc' noted that the SLtc of each batch uml
had to be delermined by taking into account the schedule of the complete plant in
addition to the productIOn capacity. Both single-product and multiproduct
processes with intennediate storage tanks and parallel units were considered, and
the scheduling was used as an additional optimization variable. They also derived
an anal)'tnl elpression for the mimmum \'olume of a storage in terms of the batch
sittS entenng and leaVing Ihe tank. Karaml and Reklailis
l
developed analyhcal
estimates and bounds for the limiting storage volume for plants composed of
:!oeveral collections of batch. semicontinuous. or continuous operations.
Simulation programs for balch planls have also been developed by Sparrow,
Rlppm. and Forder;-- O\'enur[, Reklaitis. and Woods;tt and Rippin.:: These are
panlcularly useful for checllOg final designs.
Flau" presented a shortcut prOttdure for calculating equipment sizes for
multlproduci plants, for generating prooess alternatives, and for estimating the
optimum conditions corrc<;pondlOg 10 standard equipment sius. ThiS procedure
most resembles Malone's approach_
13.3
TfIE
OTHER SIGNIFICANT ASPECrS OF
DESIGN PROBLEM
It is common practice to unpro\e the accuracy of the calculations by using of
the CAD programs., such as PROCESS, DESIGN 2000, ASPFN. etc,. that we
bnefly discussed in Chap 12_ Ho.... e\er. man} other aspects of the 101.11 design
problem stiD remain to be considered
I. Environmental coruilralOlS
2. Control of the process
J. Start-up. shutdo'o\n. and copmg With eqUIpment failures
4. Safety
S. Site location and plant layout
6. Piping and instrumentation diagrams
7. Final design of equipmcnt
8. Planning for construction
Each item on the hst nonnally inlroduces new costs, and these additional
costs may make lhe process unprofilable. I-{CDCC. il is important to try to estimate
when large DCW costs rna} be incurred. Unfortunately, there is not sufficient time 10
a one-semester course to oo\'er aU these topics,. but a brief discussion of some, as
.....ell as some refereooes Ihat may of be inlerest, are gi\'en below The diSCUSSion
emphasizes the factors to be considered just after the conceptual design has bC'Cn
completed_
En\-ironmental Constrainls
The problems aSSOCiated With the release of chemicals into the environment ha \e
received so much attention in recent years that almost everyone is aware of the
importance of environmental conslraints. Hence. it is essential 10 consider the
processing costs necessary to meet any environmental requirements. At the
conceptual stage of a process design......e include a rough estimate or these: costs by
associating a pollullon treatment cost with all the streams tharleave the proocss as
waste streams.
That is., .....e estimate the annualized. installed cost of a pollution
(realmeot facihty and add the operating cosls of this facility_ Next we aUocate the
total annualized cost of this laci.lity to all the process streams that it is Clpectcd to
handle where thiS allocation is ba.Cd on the amount ofeach stream that is handled
and the biologJcal Olygen demand of the materials in that stream. If thiS
infonnation is avaiJabk, then we can relate the pollution treatment costs 10 each of
the waste streams leaving a particular process in a plant complex. Moreover, as
uperience is accumulated, we should be ablc to prOVide lalrly close esllmates of
pollulion treatment costs These arc the costs we look for when we are developing a
conceptual design.
Of course. II we underestimate Ihese costs. our conceptual design results may
be very misleading Thus. it is essential to consull an environmental elpcrt in the
company at the bcglOOIng of a conceptual design study Similarly. after the
414 SUTIOt. II.J
TARtE IJ.J.I
A hierarchic. I .pproMcb 10 conlrol :lo}'lllhcsi:lo
If .. e ClIn ldenllfy Ind ellmln,lIe f;onlrol ploblem. by u''''I.lead) ,.I_le
modd> (whICh arc m .... b IoImpkr llun lhe dylLln"" mooklJ.), we: <;an mHumJ. OUr dCSlpl dolt
A !;:lenllfy the loI&nlfi<3.,1 du.lu,b..11CC>
I Tho>oe Ihil al'l'c:a the pro.:ess OOCbI ....nU
2 Thow: lhal Ilka the Up.".UII,-..I:..
3 If dl,ru,hlflOCl; do nul a .rlI"fiam eflecl 0" euh(r I 0' 2 4bo'e, lhem from
fortller ronslderallon lhrs SImplifies Ihe probltm
8 Make o:na,n lhat the malllpulalr"e \anablcl. l\adable In the RO>"lihccl arc adcqWlle (borh In
.. umbcr and iCnslllvlly) 10 be able 10 ..atlily the prl)CC<i.!. ron'''.lnl. and 10 opumue lhe
DpUaUn& Ylr13blcs oyer lbe OOGlplete (.c:uonable) '1IIp' otlhe dlSlurbances.
I IT the DUrnbet' ot m&D1p"bu\IC .alues IS AOI Idcquare, the n. DO( CODI.ollabk-
2. To renore conltolbbllu), we Cln
" ModIfy the Rowshed ro ,nlroduoc mo.e malllpuhurve ...
b. Modify lhe cqulp<llenl desllI" 10 lhal lOme c.oD$lralnl$ OC\IC' bc:<:omc af;U'e 0111:1 lhe
oompkle r.nge of the dl$lu.barlCtS.
.c.. Ne&illhe: kaol "aport....1opllCllWllloa v.. nabla.
C See ... helhc ..... y cqU.pnlClIl consualnu afe encouflle.cd thai p,ncDf lbe chaDllb 10 lhe
vanables from Atul),nl lbe pr.x;c:liS mmmunb Of opumwng the opa..un,
ove. the mmplcte (reasonable) range of lhe
I If the process roosfrarnlS annot be Audied. lhe conlilralned equlpmenl must be ove,de
...gncd. 10 rafo.e lhe operablbty otlbe pror:ns
2. lflhc prO<XSli IS operabk .. ha1 doc.e lIe cqUlpmcrll lbe .....n.. 10 opefatmr; taSts
by mtroduanl c:<:jUlpnwnl overoo,,.. 'n ocdcr 10 .erno"" cqulpnwnl CO[ll;(/aJnl> 10.&111 be
ceononuc:ally JUs"hcd
o UiC hcuru,llCIto 5C1ca the ronlfollcd ' .. n.bIcs .u"h lhatthe bchln..,r oflhe pfocesS
""Ill be f;IQIlC 10 the oplimum W R fIsher, f Dohcfl), and J
M DouaJu, P,OC'ttdtngl nf AIlV"''''' COtI"'H P 293. Bo,lon. June 1985)
E Sdecr pamngs oellle manlpula,,'e and eo,mollcd ruhla fu' conI/oilers.
I Cntena
G Hip StlUltrVIIJcS.
" Small dead limes. I.e, dme loarther on the no.."bcc1
conceptual design has been complcled, it is essenllal 10 consult lhe ell\'ironmeDtal
C-lIpen again, i.e.. after tbe best flowshttl has been detennined and beller C-Stlmates
of Ihe process Rows have been oblamed.
AliO il may be necessary 10 deSIgn a new pollution trealmenl faCility for Ihe
process, and time muSI be allowed so thai the construclion of this facility malches
thai of lhe process. Similarly, we could undena.ke a conccplual design for a DC"
pollution Ircatment facility and then de\"dop a more detailed deSign laler m Ihe
development of a project.
Process Conlrol
The conceplual design and ("en a ralrly ngorous optImum design using a CAD
progmm are normally based on lhe assump1l0ns that the connections between Ihe
process and its environmenl remain constant. lIowe\'er. the demand ror the
product nonnally changes With time, Ihe compositions of the feed streams will also
2 Eva.1oale pamnls
11_ plO array
b, Smllular value decompoSlflon
J EllmlOale p;lJnnp .. lIh lI,ac. InletaClKHIt
4 conti'" syslCfRI may bc dc'fdopcd
II "'ormal dyumlC rC$pOlhC-snull perturbluoM and b""ar ptnox::>s dynam113
A RcqUlletlW'tltS 10 build .. dynamIC rnodd
I All equ,pment capaCll1CS m...t be: SpeCIfied, I lbe holdup III the fUbes Ind lbc of adl
he31 e:>.chanler. lhe holdup on the tra)", rn a dlliullatron column, ell;.
2 The: $UC$ of relluJ drums, <;<>Iumn sumps, n,sh drums, mlennc:d,ate ere.. mllSl
be .pcctfied
B Assomc pafcel c:onllollO any 1I00f ",hefe there lIe IWo-pba.w: fIllllUflS.
C. Eva.lUlte the otitic unconuo&d aDd coauollcd prOOC$$CS..
D Usc It_I d)"l1llDUC models 10 evaluate lbe ucady-alle pllnl collirol systam. baVlnalhc fewest
IOlel'.(;1I0n$
UK lhe differencc bel""een Ihe lotal ope,allnl COlil of Ibc oplllllom ,ludY-Sflle oontrol
rcsponK Ind tbe dynanuc rcspollSC or tbe controlled planl IS I performlUlQC tneUurc to
compile conlrol s)"Slem a.1fcnl.llivcs fof In U!l;umed plflem of d,oslurbaoca-dlcck the
ilClWu...ly oflhe fCl.ulu 10 lJIe dullarbaDIX lIAtlenl
2 E..lu.afe the al ihc COfll.OI .)Slefll.-
1 If lhe dynalWC response IS I>Ot $llIIlaaory,
o:l Change tbe oonlrol S)'Slem
b ModIfy lhe I\o... shcr.!
E Desrgn the level COllllollcn, and .Cl;beck lhe perform.. nce.
III Abnormal dynamIC operatIOn brae JlCnu.bauons Ind nonbflC.1l' dyn.amx rClopnme.
A Sian-up and s.huldo...n
t !'Ionnall)," l\ollibec:, sbo",-m& an mlc:rmcd.ate SIOfile ,,1.I>Cd as .. Uarona powl
2 The 1I0.. ,h...1 should be dlcckc..land mo.i11icd to oorrcspond 10 lhe:
J The conlrol rcqulled (01 planl ,Ia..t-Up ..nd .. n a.e from the f;onlrol.
used fo. normal opt',al1on
B Failures.
1 A bilore al1lllysa. of lbe 1Iows.1tcc1 DCCds 10 be undc:nllm.
2 Spc:oal ronllol J)uelD$ to bandk failures mlpt be 1lCCdotd.
IV Implcml'n1ltlOn of lbe connol
A Sho.. ld dmobuled conuol be ulied? 110"
B Whal kind of COnlpole. ronlrol-human Inlerfaoc Ili requIred?
fluctuate, the coohng""-aler lempcrature returned from the coobng towers
from day 10 nighl and from summer to winter, the composition and the heallOg
value of the fuel supply will vary, Ihe and lemperalures of Ihe st_eam supply
will flucluate, elc. Thus, as the conflC(:tions between lhe process and liS envlron-
menl ehaDge thcsechangcs win dlslurb the beha\ior oflhe process, The purpose of
a control sy;tem is to ensure Ihal the prOOC:SS will operate despile
Ihe fact Ihat these disturbances occur
A hierarchical approach 10 synlhestzing control systems for complete pro-
cesses has been proposed by Fisher, Doherty, and Douglas. The slepS 10 the
hierarchy are listed In Table 13 3-1 If Ihe process can not be contrOlled, if start-up
Wilt M F l>oherty, a;1d I M Dou&JlI!i, CIltnt Enf} Rn On',,63 lB (1985)
416 SECTlON lJJ OTHU SIGNlrK'A"'1 ASPECTS Of THE [)SIGN 'ROIILOoI
is \ef)' difficult, or if the pr0tt5s becomes unsafe because of a faIlure In or more
pIeCeS of equIpment. then it may be necessary to change the flowsheet or to
the proJeCt: SlOce flowshect modifications normally arc \-ery expensive, II
IS desirable to identify any potentIal control problems as early as possible In the
de\dopment of a design. From Table 13_3-1 'o\e sec that a steady-state control
study can be undertaken as soon a<; a conceplual desIgn has been completw (i_e__
\Ie need to have a flowsheet avaIlable).
Statl-up Consideralions
The conceplUal design produces a small number of process flowsheets that should
be considered further. Normally, these t10wsheets arc not complete because tlley do
not !Delude all the minor equipment needed to operate the plant, we would
.to intermediate storage. pumps., reflux drums, column sumps, etc. In
addlUOI1..lt IS usually necessary to add special equipment to be abJc to start up the
plant easily. For e:-:amplc, if .....e consKkr an adiabatic, exothenmc reaclor with a
feed-effluent heat e:-:changer (sec Fig. 13.3-1) wherc the reactor e:-:Il temperalure IS
high that there is an adequate temperature driving force at the reactor
Inlet. then the process can operale satIsfactorily at steady-state conditions. Ilow-
there is. way to start up the process, and a start-up furnace must be
IIlstalled. to IOmally supply heat to the reactants Similarly, a special piping system
IS oflen lDstalled that makes it poSSIble to fill reflux drums and reboilcrs WIth the'
appropriate materials (pun;:hased malerials., if ncoessaf)').
Thus, a complelc Dowsheet should de\eloped fairly earl} III the life of a
PIO]Cct. and some sh(luld be gnen to de\'e!opmg a stan-up stn'ltcgy.
B} maklllg a prellmmary e\'aluallon of a slart-up stn'ltegy earl}. It IS oflen poSSible
to Identify changes III tnc tlo\lshttl and the deSign that might be required. With thIS
approach .... e can avoid the vcr} large costs associaled with o\ersighu that can
occur laic in Ihe developmcnt of a dUlgn project. A preliminary starl-up evalua tion
of thIs type also simplifies the mOle detailed start-up study iliat musl be undertaken
during the construction of the plant.
Feeds and recycle
"'''''''
1000
FEHE Rea<to<
.----
I-
IJOO
f-
To separallt'll systC'm
nCURr Illi
,--r:al-dnuml heal
u;cnON l)l OIllU SIONfHCUlT ASPECTS 0' THf DESIGW ....OILEa.! 417
Safely
Safety studtes should be undertaken throughout the life of a project The 100tiai
study must be carried out by the chemist, 'o\ho needs to recognl7.e the nature of the
materials being handled In partKular, it is necessaf) to know which raw-material.
Intennediate, and product components arc flammable. unstable. to:-:ic, corrOSI"'C,
highly reactive. esrecialJy sensitl\-e to Impunties, etc_, to be able to handle these
materials safely In the laboratory
The' propertIes of the' matenals also l1e'ed to be con<;idered dunng the
conceptual design If al all possible. we prefer that the design correspond to
conditions outside the e:-:ploslve limits for any stream in the process. HOIll'e\er, we
also need to eumine whether changes in operating pressures, temperatures.
compositions, will cause a stream to mo\e into the uplosive range_
Similarly, if the presence of corrosi\"I': components means that special
materials of construction will be needed for the equipment that processes Ihose:
components, we might want to modify the dcslgn to miOlmlZC the amount of
expensi\'e equipment required For highly reacth'e materials or situations where
the reaction rates are "'cry sensitive to changes 10 impurities or process parameters,
we also might want to modify the design.
When the intermediate storage units. or other units having capacitances, arc
added to the Ilowshect, we want to minimize the inventory of any hazardous
materials and to consider special safety systems that will ensure safe operation As
the design prooeeds. we must add pressure rehef syste'ms on the prooess ,'essel<;. to
be certain to avoKl hazardous operations. In other words, 'o\e want to able: to
predict .... hat might If something goes 'o\rong with the operatIon of the
plallt. and .... e want to be absolute!} sure that '0\-"1': ha\e a safe situation .... hen
something docs go wrong..
An excellent set of gUldellllcs for a safety evaluation was developed by Datelle
laboratories for the AIChE. Table 13.3-2 decribes the potential hazards, possible
Imtiating e,ents, the propagatIon and amelioration actions, and the consequences
of accidents. Figure 13.3-2 shows a flowsheet that describes the steps in a hazard
evaluation. Note that this figure indicates that design changes might be required,
and. if possible. we want to identify these design modifications as early in the life of
a project as we can. Several checklists that are useful for safety studies arc included
III the manual
Site Location
Sile locatioD also has an impact on conceptual desIgn because the utilities available
on a site, e.g., temperatures, will depend on the geographical
location. Similarly. the costs of raw materials will reflect the transportation costs.
G,... /_ "liZ Erolll'''''' 1"0."'. pt""pan'll by''''' Rall"tk CoIuml>us I)msot,n for I""
Cnln for C'hmlicat SakIV_ AtC'hF."'"" YNk.
j

K
,

.E"
1
!
.,
p
< !
4'8
4'9
dependmg on where these materials are produced Moreover, Ihe: site location rna}
depend on the major uses of the product
For eumple. most plants thai produce benzene are located on the Gulf Coast
of Texas. The benzene product is often u<;ed to produce ethylbcllzene, by reaeling
the benzene with eth}lene Often Ihe elhylbcnzcne is deh)drogenaled 10 make
styrene The Gulf Coast IS chosen for all these: proor:sses because there IS a large
suppl} of tolucne (from crude oil) and a large supply of ethylcne (from elhanc
cracking plants). Both ethane and ethylene would be qUite cxpenslve to transport
elsewhere in the countr)' Styrene i<; also produced on thc Gulf Coast. to minimi7c
transportation costs
However, most of the plants thai produce polystyrene are located in the
northeast. Most of the fabricators who use polyslyrene as a raw material are
located In the northeast, so thai the polystyrene planls are located dose: 10 these
fabricators. Thus, the overall costs are mmimized when styrene monomer is
shipped, but not the otller materials.
SECTIOt" IlJ OTHU SIO/'llfk.....,. ASI"ECTS or mE !>SIGN 1't000lHl 421
Othe:r Design Problems
The other desIgn problems. e.g_. final equIpment design. pIping and instrumenta
tion dIagrams. plant layoul. project engineering, etc., are conSidered to be well
beyond the scope of this text. All these problem areas arc very important to the
success of the commercialization of a proJed, and each area poses many new
challenges. An understandmg of the process. is essenllal 10 de\'e1opmg
successful solutions In each area. and lhat basic understanding is most closely
related to the: conceptual design
... "
I
hUA"'"
Hazards
Jdc:ntifil
IRevise design
Step 2. Estimate
I
or procedure
consequences
Consequences
Consequences
deemed acceptable
OpportunltlCS
not IlCCCpOIble
Identified
Identify opportunities
I
10 reduce consequences
No obvious
opponuniLies
ldenlify accidem
inrtialing CYenlS and causes
Initialing eveniS
and causes identified
Estimate probabilities
of initiating cyc:nl$
ProbabilIties
Probabilities deemed acceptable
not accepLabJe
Opponuniucs
opportumties
Identified
to reduce probabilities
of initiating CYenLS
No obvious
opportunities
acei<knt
rYenl sequences
I
Estimate probabilities
and consequences
of sequences
Probabiliues
Probabilities andlor
and consequences
consequences no( acceptable
deemed acceptable
Opponunities
Identify opportunities
identified
to reduce probabilities and/or
COf1SC:IlIUences of eYeDI sequences
No obvious
opportunities
10. Procttd to more
lJRE IU2
quantll.111ve hazard
,.......,
evalualJOn analysis
FIG
....
420
PART
IV
APPENDIXES
423
APPE DIX
A
SHORTCUT
PROCEDURES
FOR EQUIPMENT
DESIG
Normally \lot' usc shorteul equipment-design procedures "'hen .....e scrn process
alternatives We want 10 focus on Ihe most expensive pieces of pro:ssing
equipment during !.his screening activit)', and therefore we usually focus on gas
absorbers, distillation columns, heal exchangers and furnaces, gas compressors
(and refrigeralion systems). and reactors. Some useful shortcut models (or most of
these: units an in this appendix.
We do Dot include a discussion of reaclon; because they are so dependent on
the reaction chemistry Thus, ont of the many tuts on reactor design needs to be
consulted 10 develop a reactor model. Also, note that our list of shortcut methods is
DOl complete. and other models are available in the many texts on unit operations
and design.
A.I NUMBER OF TRAYS FOR A GAS
ABSORBER
The design of gas absorbers and that of distillation columns have many
similarities. Therefore, y,e decribe the shortcut procedures for finding the number
oflrays reqUIred for each type of unit, and we present a procedure that call be
used for the desIgn (i length and diameter) of both types of uniu.
425
n:CflO,", Al NlJJ04I1El .... lOll A GAS AIISOU 427
and we 30CC that the operatlOg hne .... ill al ....'ays be: hnear on a Y =0 X diagram
Accordmg 10 the defimllOIl of a theoretical plate, we assume thaI equlhbnum
IS established on each plate For moderate pressures P
r
of, say, leu than 50 10
100 psig, where we C<ln consider Ihat Ihe gas mlxlures are ideal, lhe equilibrium
relationship can be: .... rillen as
(A I-J) I
A mateliaJ balance around tray n in hg_ A I-I Si"'t30
Procedures for .be of Theorefic..1
Plafes 10 II Gas Absorber
In general, v.c expect that bolh Ihe gas and liquId II
change as Ihc solUle IS Iransferred from the o.... i.n a gas absorber .... 111
wherc Ihe solvenllS Domolalilc and thc car. gas 10 lIquId Slrcam for
molar ratio30 10 describe lhe composlIions v.e often usc
terms of moles of solule per mole of sOI\e' lC.' e lqu compoSHlons are glYcn 10
belv.un the 10lal gas and II :Ida;d for the gas SUeam The
carner flows G
s
and L
s
are q o.... s, and L, respectl\'ely, and the
(A I-I)
(A 1-4)
(A 1-5) 0'
Y yP' '1r X X
p-- =0 _m
I + Y Pi" P
T
I + X I + X
In general, lhe activity coefficient y will change with both composition and
lemperature, and the \apor pressure po will change with temperature throughout
the column (the temperature Will vary because as the solute condenses in Ihe
solvent, it gIVes up ils heal of vaporization). Hence:. the eqUilibrium relationship is
nonlinear on a Y-versus-X diagram Because of this nonhneanlY, normally we need
10 employ some graphical or numerical procedure 10 soh'e the: design equations.
y} Ls=oL<I-x}
Similarly, lhe relalionship between molar ralios and mole fraclions is
y - y X = --.!...-.
I-y I-x (AI-2)
We denole Ihe composl(lOn ofbolh lh
number. e gas and Ihe liqUid leaving a tray by Ihe Ira}
(A.I-7)
(A.I-6)
(A.I-8)
(A.I9)
(A 1-10)
I) I)('.O)];(A
y_. J
where
KREMSER EQUATION FOR PLATE TOWERS. For dilule mixtures the problem
becomes much simpler, and we can oblam an analylical solution of the design
equallons The result IS called the Kremser equation, and II can be: wnllen as
ApproJ:iJnll!ioL Order-of-magnitude arguments can be used (0
simplify the equation (sec Sec 3.3)

/If I 2:::::610g Y,.
y_.
or, if L/(mG):::::: lA,
L

mG
or N+ I =0 In[1 + (A _ mX",)]_I_
\y_ - m.:c:.. In A
A graph of thiS expresotion is shown In Fig. A.I-2.
FlGURE AI-l
PIale &;15 absorber
-I
I
I
N
I
I
I
N
---
n
r..
x.
---
2
I
I
I
I
L.
G, Y.-
G, Y,n
A K,em""r, \",1 1',,,01 \too,. 21(21) .2 (19)0)
.. 28 A' NllotllUor TUYS rOll A GAS Amoalflt
s.EC1l0t< .... 1 MJM.IIU Of fun f'Otl A (lAS AI050UU 429
Also. we arrange: the equilibrium relationship given by Eq_ A 15 to obtain
Applk-bilil} of 1M Equtioa.. We can usc our approximation procedures to
den\'e a criterion for e... aluatmg lhe applicability of tM: Kremser equation:
(A 1-11)
(A 1-12)
(A 1.13)
T
L
- Inlel LiqUid Temp x _0
y=o
{ [
dH: JG,}
li. 1+ 2(A11 2Au )+ RC,.TI 4 sO.1
where All and A
21
are the Margules constants of the solute and sohcnt,
respectively, at infinite dilution.
We obtam this expressIon by first writing a material balana: similar 10
Eq A 1-3 around the top of the lower
and at these conditions .....e let
l' _ Y\X mX
P
r
I + X}
If m is essentially constant. then both Ihe operating and equilibrium lines will be
linear and we can wrile the: Kremser equation in terms of molar ratios and molar
flow rates. Hena:, .....e ellpc:ct the Kremser equalion to be: valid ..... hene...er m is
essentially constant.
To tcst the constancy of m, we ellpand it in a Taylor senes around 1M:
conditIOn of mfinite dilution:
.95
0.8
1.0
A or IfA
0.3
0.5
0.
0.7
1\
1.05

1.1
1.2
L3
I.0
0 8
0

.
0 4
0 3 -
..
-
<
0 2
..
"
"
0 1
! 008
;J ";..
006
" ,
,

0.Q4
003
"
-
c
002
.><
,
, ,
, -
fOPO
mo=--
P,
The Taylor sencs expansion gi\es the result
(A.I-14)
IIA {stnppln
(A 1-15)
From the definition of m,
am I ['OP" ( 1 )JI
P
T
o=m
o
Similarly,
FIGURE .0\.1-2
umhn Co< COUnlCrcv.tT'O'nl caJCadoes. WIth IIcnry'sb.. equiltbnwa and COftSI.l.lll
abIorpIlOrl Of sl"Pf'lnA factors IAftn lIac""',,'1o -' 1'<POt" C","". Jog 48 j1J, j;,(}, 617
(191]) ]
(A.I.I7)
If ......e assume that the activity coefficient is given by the Margules equation
In y - 2x
1
(A
2I
- A
IJ
)] ('\.1.18)
and let
;0 = eXf1 All
(A 119)
n.CfION A I NU"'U 0' 'hYS (0 GAS .. &$Olin 4JI
then we find that
eml till) 2m,,(2-t
u
_ All) (A 1 20)
oX 0
80th y and P depend on temp.:rature, bUI for most )ubstanccs the lempera-
lUre dependence of the "Cll"lt) coefficient is much less than that for vapor pressure,
thaI IS, d(ln i) J1 is proportional to the hcal of mlxlOg, whereas d(ln P-)'JJ I!>
proporuonallo the heat of vaponzation I fence. to simplify our analysis, ""e neglect
the temperature dependence of
r
O,er moderate lemperature ranges, we assume
Ihal Ihe vapor pressure dependence on temperature is gIven by the C1auslus-
Clapeyron equatIon
(A 128)
(A.I-29)
(A 1-30)
L
A =
, m G
,
We estimate the number of trays required based on the recovery of ale)'
component, I.e. normally we want to recover 99 'Y. or so of some component:
N + I = In [I + (Ax _ IIYX.h. - m..:x..:.
I
_)] _'_
\Yx....., - m... xX,bo In Ax
A... ,
.lJ._ - ",JxJ.... = j -
J'J._ mJxJ.bo A
J
- I
Of course, Ihe design probkm becomes much more difficult If heal effects arc
important
where K refers to the key, Then the recoveries of the other components can be
delermmed
Abs<.lrplion
For multlcomponenl absorpllon wIth dilute SOIUllOm. \O\e can use the Kremser
equalion to descnbe each componenL The slope of lhe equilibnum hne for eal,;h
component will be dIfferent. and lhe absorption factors WIU be different
(A 1-23)
(A 1-22)
(A 1-21)
Ml" ]
2(lAIl A!I)X + RTf (T - Tv
i'ml_ MI,
iJT - m
o
RT
l
. ,
po = "p[_ MI, (.: _ -'-)]
R T T
L
and we find thaI
With these apprOJumalJons. Eq A 1-15 becomes
We would M.e 10 obtam a criterion Ihal would mdlcale the maximum
poSSible difference between m and IIlO' For thIS reason, .....e wrile Eq A.I-23 as
11I_ m{I + Y;. + 2(A
ll
- 2A
I1
)X_, + CT.... , - TL)] (A 1-24)
However, if we expecllo design the absorber for 99% recovenes, Ihen
G
s
Y;.,::;:- LsX_ (A.I-25)
Similarly, if we assume thai the heat released when the solule condenses appears
onl)' as a senSible heat change of the liquid, lhen
G
s
Y.. t1H, = C.Ls<,T_ - TJ (A 1-26)
Thus, Eq A 1-24 can be wnllen as
COLBURN'S MEll-jOD. Colburn has presented shortcut procedures that can be
used even when the equilibrium and operating lines arc curved. He de\<c!oped these
e.xpresslOns by using order-of-magnitude arguments to simplify the design equa-
tions for paCKed columns. and thcn he wrote the cxpressions for plate towers by
analogy. In addition, he lists the results for stripping, distillation, and extraction as
well as absorption. His results are given in Table A.I-1.
ornER PROCEDURES. $c\'eral other shortcut procedures for estimatmg the
number of theoreticallrays have been presented in Ihe literature I In general, these
olher procedures are more ledlOUS 10 apply, but they give more aCC:urau: predic-
tions. Also a number of computer programs are available for lray-by-tray solutIons
for absorber problems.
m {[ MI, ]G,}
rn
o
= I + 1';. I + 2(AH - 2A u ) + RC. Ti Ls (A.I-27)
Of course, If the second h:nn IS very small, say, less than 0.1, Ihen m ::;:- ffl
O
and
the Kremser equlltion will adequalely describe the system lienee, a crilerion for
the applic3blhly of the Kremser cquanon is
(AI-II)
A. P c",lbllrll, S'mpllfied Calcuhlllon of D1rrUSlOnal PrODtSSCS,- h.d tAg JJ 459 (t941).
'G Horton _net W B -Calculallon of Absorber Pt:rformancc: and Dnlp,M fnJ Eng Ch,m.,
31 1)&4 (1940), W C. Edmmer, -Design for tlyd.ocarbon Abwrpllon and SlnpPlQg." ImJ &lg Cltrm..
J5 837 (l94)1. B 0 Smllh. Drs.q" or ... Sraqrd 1"(J("rJW's, Mc(i'a.... H,Il. Nc... Vocl. 1%).
chap I, T K 5hc.... OO<I, II; L P,gk'"l.,,,,d<' R W,lle. M<U. TrwuJtr, McG""-Hlil, New Yo.l, 1975,
chap 9, and U V S'ochr and C R W.lle. and Shof1Cu. Dc:sJp C.lclIlal'otlS fo. Gas
AbsorptIOn tnyolyn,s Lar&C Uca, EIJecu 2. Rapt<! 51101'1-<...' tksilf' I"oo:dl)re for Paclal G...
Abu>rlx,.," I"J I:AtI OIlM Fund. 10 9<t {19711

-, -

, ,
-,-
><
E IE-
;-
<> <> 1 1
, - -
'i
I
i
!

i?
"

, ,
:;.;;;
_I " .. r..
..
-,
,
,
,
>

>

..-
, '
,
-'-
:::-:;

I 1 _ I ,

+
,
>

-
E'
;;;0-

,
+
,
>

I, : + +
- - - -
- - E E

,
g ,
> >
E 1<11: i I'; i'1",
<
JIJ
,

>
"
+

c,
, ,

-
"0
,

..
-

"0

;;
8

;;
8
433
432
liUllO"',,1 435
500 1000
--
.5 10 50 100

Plio.
Absorbers
..
;.
0
0
0
"
IE
0
20
0
E
=
'0
"
0
6
.. 4
"

2
0

I
.:r
0.1 0.5
AClual P)IIIe5-PJate Effidencies
HGURi.. A.l_J
O'Conndn cOllelallon for ove/all lIl)' [From 11 OCtJIlMlI. TraM AICIr/::, 42
7J/(/946))
a = Relative volallilly of kty componems
1' ., Molal average: viscosity of feed, cP
I'll "" Molal average VISCOSIt) of liquid, cP
Mil = Avenge molecular weight of hquld
PA = Liquid deosuy. Ib/ft)
", ,. Mole fuadH>n of solute: in gas In
with liquid/mole fraction of solUie in liquid
3\ = At average: column lcmper<:llurc: and pressure
Unfortunately. our assumptlon of equilibrium on each tray IS seldom mCI in
practice, so \lot must find a relationship .....ec:n the number of IheorclK:a1 trays
required fOf a separallon and lbe: number ofaClUaltrays. This relallonship Involves
the plate effiCIency of the column There are a variety of to prcdlct platc
efficiencies, and these range from \'ery simple. qUId. cslJmates to .. cry detailed,
computational procedures
For prcllminary process designs, we normally usc the simplest method
available (although 11 is nOl very accurale), which corresponds to O'Connell's
. ,

..

..

,
..
I
,

"
,

8
E ID::: Ifl..,:'

"-to,

o

'I"" ..
EEl I
I I _

R H Perry and (' It CbIlIOn. fJt(J'''HT'J lIlUlt1boo1t.. nl , McGrn. New Vorl. 197),
p 18 Il
434
436 SECTION AJ OlftlLLAnO," JoJUWIU or UAYS
SH;TION U OlsnLLHION COLUNI'<S l'ol'NIU OF r.... YS 437
the vapors e\en on hot summer days. so we assume that cooling lAater IS avaIlable:
at 90 F. Also. the design of the cooling lowers is based on th(' Ihat all
the streams returned to the tower ha\(' temperatures of 120 For IC'ss, because the
scaling and corrosi\e characteristics of lAaler be1::omc 'ery pronouncrd aOO\'e that
temperature (ie.. calcium and salls are deposited on the exchanger
walls).
H the buhhle POIHt of the overhead IIlIXtur(' frOIll the column IS greater than
130 F hleh allo.... s a l('mperature dri\lIIg force of atlh(' outlel.
then e can use a total cond('nscr and operale the column at atmospheric pre"sure.
(Actuall) ..... (' .... ould opera Ie the column at an excess of pressure. sa)'
25 psI&, so that any leakage would be to.....ard the outside of the column.) lIo.... e\er.
If the condensing temperature of the overhead mlJ:.ture is less lhan I(XTF. we prefer
10 Increase: thc column pressure until we can condense the o\erhead at 115"F or so.
(We reduce lhe eXit lemperature of the cooling water in order to keep the tower
pressure as low as possible rather than using a refrigerated condenseL)
In some cases. e.g.. ifhydrogen. methane. or similar lighl malerials arc present
in the o'erhead..... e usc a parllal condenser The product is taken as a \apor from
the Rash drum. and lhe column pressur(' is set suffICiently high to obtalll an
adequate supply of renux Also. for a fe.... Slluations. e g.. demcthaniHrs. dcclhan-
izc:rs. and cr)'ogeoic systems. we are forced to both high pressures and
refrigerated condensers In cases where the o,erhead bubble pomt exceeds 267 F,
.....e operate the cond('nser at atmosphenc pressure. but we generate steam Instead of
usmg coohng water.
If a reasonable operating pressure will condense the overhead between 100
and J30-F..... e nexl check the bubble point of the OOlloms stream at pressure,
to C5l1mate Ihe temperature of the sleam required for the reboller Dependmg on
whether lempcrature is slighty abo\e (that is, 2S to 70
n
F) the temperature of a
steam suppl) in the plant. for exampJe. 2S psia = 26TF. 115 psia,., 338 F.
420 psla = 1000 psia _ 545"F (although a pressure this hi,!!h mi,!!ht not be
available). we might decide 10 rcduce the column pressure somewhat 10 accommo-
date a lo.....er-pressure sleam supply to the reboiler. For bubble points above 400 to
5(X)F. depending on the sleam supply available in the plant. we might operale the
distillation column under vacuum conditions or use a furnace as a reboiler.
For most distillation column design problems, the selection of the operalmg
pressure for stand-alone columns (i.e. not energy.inlegraled with the fest of Ihe
IS a relatively simple and straightforward task: the primary consideration
is the condensation lemperature of the slream. Howe,er. in some cases
the costs of high-pressure steam musl be balanced against lower temperalures of
Ihe overhead. or refri,!!eralcd condensers are required. For these sltualions lhe use
of heat integration techniques often prove to be particularly advantageous
(A 1-31)
COLUM S,
correlation-(sec FIg. A 1-3). If lA'e fit a straight line to the graph fOf gas absorbers In
the range from JO to efftciencies. we find that
0.377
0'" 0-'09
(mAt LP,JP,J
A.2 DISTILLATIO
OF TRAYS
O\erhead and Bolloll1 Compositions
In many cases enher Ihe o\erhead or the bonoms composHlon from a dlstillallon
column will be fixed by product punty specifications. If this is notthc casc. then the
('nd compositions repr('SCnt opllmi7..alion prohlems. and the) depend on lhe
process economics. For preliminary designs. .....e uSC' Ihe rule of thumb thai we
desire recoveries of the light key 10 the o\'crhead and 99.5% of the heavy key
in the bolloms. Also. we assume that allihe components lighter Ihan the light key
are taken overhead and that all components hea\ier than lhe heavy key leave wilh
the bottoms.
A varlCty of desIgn procedures for distillation columns, rangmg from shortcut
mel hods to ngorous calculations. are descnbed lit numerous books
Spaa' 1IIIlIIatIOns pr('dude a detailed review of these various technique" Inslead.
we present only a bnefsummary ofsome ofthcsc procedures and a short discussion
of some related material.
N
N --
-- 0
To complete the desIgn......e must find the tower height, the lower diameter. and the
cost. We diSCUSS these calculations in Sec. A.3.
or course. at latcr stages in a process development. we would use more
procedures for calculating lhe plate efficiency
Once lAC ha\e C$timated an overall plate efficiency. the number of actualtra)'s
reqUIred is
Column Pressurt
The operating pressure for a distillallon column nonnally IS filed by the economM:
desirabilily of using a condenser supplied ",ojth cooling water from the cooling
towers. We want to ensure thatlhe condenser .... 111 be sullkiently large to condense
H. E O'COllllCll Plait: EKlCJ<"ncy 011 TXt_ling CoI_ and AbsoIbc,s..- Tt-u AIO,F.. <12 1S1

C S Robf_ and E. R Gilliland. EkmitrtU t>{ f'act-J Dut"""1Oft. <llh cd. Mc<n-aw-lltll. "'",'
Yorl. p 161
43S Sl'CTION AJ OISTll-lAnON r-U..IlU 01 I"".
SECTIO'" Al Dlml1..... TION C'OIUMNS M!JolIU Of r.... YS 439
Eslimating Bubble Poillls lind IJCIo'l Points
For preliminary designs, Iol.e Iol.ould IILe 10 have a procedure for eSllmallng
lhe bubble pollli and Ihe de"" poInt of lIluaurcs For mosl hydrocarbon mUilUfCS
(I hose Ihal do not uhlbH h)drogen bonding) we "'rilc Ihe eqUilibrium expressIOn
..
Now if we dl\ide bolh the numerator and the denommator by the K value or the
heav}' key KIIII. and dcfine Ihe relative: \'olatihty as
Slna:
we can also v. rile I:.q A-2-1 as
then Eq. A_2-3 becomes
I, = K,x,
K
, = _ '
Kill(
CA2-1)
(A 2-21
(A 2-3)
(A 2-4)
ca>e ""here: the: rc:lali ...c: volatlllly b reasonably <:onlliant, we: can use a \lane:IY of
methods to bllmalc: Ihe: numlxr of Iheorellcal trays re:quire:d ror the: sepanlllon
The constant value: of CI used In these IS normally taLen to be the
geome:tnc mean of the top and bottom \aluc:s
(A 2-7)
bpcnl>l\c column:. ("'hll-h dorl.': those: v.e want to dl>lgn most .Jccuratel)')
normall) I.':orre:spond to close: bolhng mall.':nals, and for close boiler:. 0:,.. IS usuall)
close 102...,_. and both arc: In Ihc r,mgc: from I to 2. Whclle\er the: boIling points
or the: o\'erhead and bottoms are widely separated, CI'op IS often much gre,Hcr
than lboo'_' The assumption of conSlant CI is 110t valid for Ihese: systems, and
the assumption of constant molar o\er-flow is not usually \lalid. However, Ihe
spllts are easy for these cases, nonnally there are not many lrays and errors 10 Ihe
prehminary designs do not ha\le a greal impact on the tOial processing COSIS. Thus,
In man) cases we can oblain a reasonable firsl eslimate by using Ihe smallest \alue
or CI 10 Ihe shortcUi design equauons
A cmenon Ihal can be used 10 C:SlLmate the effeCI or \ariauons III :J IS
prc:scnted later in thIs appendIX
",X,
j - ._-
. L',X,
BUBBLE POIl'oIS. To estlmate the bubble pamt fo.>r a specified liquid composttlon
)., (.... hich v.e: eSl1mate frolllihe column material balances), we simply calculate
and then use one orlhe K value correiailons, such as that given in AppendIX C.I, 10
find Ihe temperature at the specified column pressure_ Normally. we calculate Ihe K
value for the light key to estimale the lempcrature.
Eslimaling Ibe Number of Theorelical Tra)" for
Sha.rp Splils in Simple Columns
Initially Iol.e consider the case of sharp splils belween the light key and the heavy
ke)' (IC., no componenb betV.C'e1l the and SImple columns (i.e., one o\crhead
and one botloms stream)_ BOlh empirical and analytical procedures are discussed
Then a shortcut procedure for sloppy splits is presented. Shortcut procedures for
evaluating column sequencing are presented in Sec. AA, complex columns are
considered In Sec. A.5. and Ihe energ) Inlegratlon ohland-alone column sequcnces
is dlscusscd m Sec. A_6.
DEW POINT. A simple CJ:pression Ihal can be used to estimate the dew paim can
be derived in a Similar fashion:
)"=K
x, 'L}'d::z.,
(A.'-6)
FenskeUnderMood-GiIIiland Procedure for
Esrimaling tlte Number of Theoretical Tra.fs
One: of the mosl commonly used procedures for oblammg quick CSllmalCS of the
number of lheorelical Ira)'S requIred for a dlstilialion separatIon IS called the:
Fenskc-Underwood-Gilliland procedure
Once we eslimate K( ror the hghl Ley, v.c use a correlation 10 find tbe temperalure.
Relalh'c Volalility
Once we have eSlllnaled the bubble POllltS of the o\crhead and bottoms streams
(we eSlimate the dew point of Ihe overhead ir we use a panJaI condenser) at the
operaling pressure or the column. Iol.t: can calculalt: the relatl\le voialilily of Ihe
components Iol.llh respect to the hea\'} It:yal both the top and the bottom For
GILUUND'S CORRE(.ATJOl'o, Gilliland- developed an empincal correlation for
Ihe number or theoretical trays III terms of Ihe mlllimurn number or trays at lotal
reflux N.. , the minimum reflux rallO R... and Ihe actual reflux ratio R, sec
C S Rob,nloOll Ind F R Glthland. -Fkmenl\ of Fracl100a1 DIstlttallon,- 41h ai, ItGlI.... -1I111, "' .....
Yori. P l-l1
1.0
0.8
0.
t
' ,

0.'

0.2


00
0.2 0.4 0. 0.8 1.0
0
(g)m'" 0
Q+I
0
(A.H)
(A.2-9)
(A.2-14)
heat of
(A.2-12)
(A.2-13)
(A.2-ll)
In SF
10 ,
N-N [(R-R)'''""]
0.75 1- --"
N+I R+I
In {[xo/(l - %D))[(1 - .x..)/x",]}
N", = I
n,
Hence, we can calculate N aftcr we have estimated N.. , R.. , and R.
R... z, + qx
o
a[xo<:q - I) + z,{R... + I))
R..(I - z,) -t = (R.. + IX' Zl") -+ (q IXI xI
q '.

q-I q-I
where q is the heat required to vaporize I mol of feed divided by the
vaporization. Using these expressions to eliminate y and x. we find that
=

1 +(a - I)x
and the q line representing the feed quality
the equilibrium curve
Fig. A.2-t A simple: equation for Gilliland's graphical correlation was developed
by Eduljee
SECTION U OImU-AnOl'l COLUMNS NUMIIEIl Of TIV.YS 441
SF (X,.,,)(x
w
.",) (d")(H:"') (A 2.10)
XD.IlI( x ....
u
: dlU;",u I rue: I 'HIe:
where X
f
= mole fraction. d, = distillate Row, WI = botloms now, 'i = fractional
recovery. lK = light key. and HK = heavy key.
where the composition ratio in the numerator is ortcn called the separatiQIl [acror
SF. The analysis is valid for multicomponent mixtures, and the separation faClor in
Eq. A.2-9 for these systems can be written as
FENSKE'S EQUATION FOR TI-IE MINIMUM TRAYS AT TOTAL REFLUX
N... For constant-a: systems, Fenske derived an expression for the minImum
number of trays at total reflux. The result for binary separations IS
UNDERWOOD'S EQUATION FOR TilE MINtMUM REFLUX RATIO R For
a binary system with constant a. the minimum renux ratio corresponds to the
common intersection of the operating line in the rectifying section of the tower
R... XD
y=---x+---
R... + I R", + 1
1.0 0.4 0.6 0.2 0.04 0.06 0.10
-
Q + I
o
0.02
.0
.8

.4
.3
.2
.1
8
3
2
0.0 0.01
o
I
o
o
o
o
0.(><
0.0

o
0.0
0.06
FIGURE A.Z'
Gilbland co"dalJon. (From C. S Robuuon orod E. R. G,'ilf<Vld, of F'(KlrOtlol DUI,flOlion. 4111
MeG,oM' Ifill. I'o,k. 19j(J, l' 349)
fl E. Edulp. EqUllion5 Replaoo: G,llJlands' 'h'd,oca,bmt 504(9) 120(Seplcmber 1975).
440
SECTK' Al OISTII.LATIO/'I COLUWIG NU'OU ot TlAYS 443
For thc case of a SlllurauJ-lllluid whcre q = I. this cxpression becomes
Moreover, for high-punly separallons .... here xv"'" 1.0 with reasonably well.
balanced columns (where 0.2 < x,. < 0.8) and moderate values of a, we see that
R. (_'\fx. ->(1 - x.)]
a-IJlx,. I x,.
(A.2-IS)
correspond to more cxpensu'e condensers and rebolkrs, along with hIgher coollDg-
watcr and sleam costs. Thercfore, therc is an optimum reflux rallo for any specificd
separation.
However, expericnce has shown Ihat the value of thc opllmum rdlul ratio
normally falls in the range 1.03 < RIR... < I 3. Furthermore, the slope of Ihe cost
curve is \ery Sleep bellow lhe optlllllun, but relatively flat above the oplJmum
Hence, II is common pracllce to select an opc:rallllg reflux ratio somewhal above
the opllmum, and for firsl esllmales we often use the rulc-of-thumb value-
ofthumb:
Similarly. (or a saturated-vapor feed where q _ 0, Eq. A.2-14 reduces to
Thc operallng reflux ntilO is chosen so lhat RIR. "- 1.1 (A2-21)
(A.2-23)
R. - (_I\fax. - -1
a-lA.y,. l-y,.
and wilh Ihe same simplifying assumplions. we nOle Ihal
(A.2-I7)
SIMPUFIED APPROXIMATION Of GILl.Il.A.ND'S CORRELATION. From a
number of case studies, GiJliland nOled that when RIR ..... 1.2.
In SF
N-2N =2--
In CI
Underwood's binary equallon for Ihe mlOlmum reftux ratio gi\'en abovc aJso
provides a conservative (although often a very conservallve) eslimalc for the
minimum reflux ralio for muhicomponenl mU.:lures:
However, an exact solulion for muhicomponenl s)"stems wilh CODslanl a can be
obtamed by using another of UDdeNcod's equalions. For the case of sharp
separations. firsl we solve the equation below for the \"alue 0(0 beween the 11 values
for the light and heavy keys:
OPERATING REt"l.UX RATIO R, As the reflux ratio is increased above lhe
minimum, the number o(lrays required for a given separation decreases. so tbatthe
capital cost o( Ihe column decrc:.ases. However. increasing the reflux ratio will
incrcOise the vapor ratc in lhe lo.....er. and, as we show lalcr, blgher vapor rates
Approxim.le Expressions for (he Minimum Reftux
Ralios
Glinos and Malone' have developed approximate expressions for Ihe minimum
reflux ratio that can be used inslead of Underwood's equations, Eqs. A.2-20 and
A.2-21. Tbeir results for ternary mixtures are givcn in Table A.2-1, and the
expressions for four-component ffilJ;tures are given in Table A.22. They ha\e also
proposed a sel of lumping rules for the case of more than four components.
MODIFICAnONS OF TIlE FENSKE-UNDERWOOD-GtlLlLA I) PRO-
CEDURE. Inslead of Fensl..e's equatlOll for thc mimmum Ilumber of lrays at total
reflux. some ;n\estigators use Winn's equation.' This approach leads 10 good
prediclions o\'er wider ranges of the relalive volatility than FellSke's equation.
Similarly, man)" IOvestigalors prefer to use Erbar and Maddox's correlatlOn
l
for
the number of lheoretical plates inslead of Gilliland's. All these techmques give
quile good predictions, providlDg Ihal lhe value of Cl is relatively constant-
This fC$uh provKJes a \ery simple preocedure (or eslimaung the number of
theoretlalll tra)"s required for a specified separallon_
c. S and E. II; G,llIIand, 01 F,acflonal OU/llfaflOll, 4th ed, McGrawHln,
York, 1950, p 350
F W WlIfn. MNn. Volauilly ror DiSldlat,on "tlro! Rtfinn, 17(5)
216
J H Erbn and R N MaddoA. "mol . ..0(5) 18) (196J>-
IILG"__"" F Mak>nc.ld.tCPNK Oft, D 7641'98<1).
(A.2-18)
(A.2-19)
(A.2-20)
(A.2-21)
1
R :=:: - I
(';I I)y,.
R... = (_I_XXD,u; _oxo.",,)
a - I x,.,.... XT.""

'" a.Jx".i
L..--=I-q
J-I aj-O
Next, we substitule this value of 0 mlo the equalion
ox

i-' Cli - 8
An CIaet solution of this pair of equations requires a procedure.
Some approximatc solutions of these equations are presenled later in Ihis
appendix.
TA8U: A.1:-1
Minimum reRux ratios for lernary mixlures
TABLE A.1-2
Minimum rreflux r.lios for four-compoMnI mi'lllllrrs
AjIlC
AB,C
( .... +- .>:e,K".. - I) +- - I)
(1.. +- 1.,)(1. " ..,1
C
')
ABCD
All'CD
+ lIeXc
'* :r:l(_. _.1 + + :r.,(:o.-=--fj
2c-'. .... +:rei :r
" -.: + -.: ;;::-1 +- ;;=I
'* h, ... :r.)[1 + "..(Ie"" ,,_)) +;r. (x,"" 1'
AnalYlieal Procedures for lhe Number of
Throrelic.1 Trays-Ideal Mixtures
In addilion to the shortcut procedures for calculating the number of trays., there are
somt analylical procedures for the case of Ideal miuures.
ARC/BCD
- ..... + '".-'. +'"<;"c + _
SMOKER'S EQUATION FOR BINARY SEPARATIONS- Smoker de"elopr:d an
exact anal)tical solution for the case: of bmary mixtures. The requilibnum rdation-
ship IS "-nllen as
SCCllon .. e cakul:ue lhe number of tray$, using the
Similarly, for the stnpping section we modify the expressions above and lei
Then for lhe: rectifying
algonthm below:
=
}' =
(A.2-24)
JU, + qx
o
- R + qx.

(R + I)x, + (q - 1);0:0 - (R + q)x.
b _ (x,- - x,Jx
a
(R + I)x, + (q - I)%D (R + q)x.
Bere k IS the same quadratic requation as for the rectifYlllg secllon
(A.2-JJ)
(A.2.34)
(A.2J5)
(A.2-J6)
(A.'-37)
xo - x. - II
{(
x. )(1 DR )-.
= In I _ x
D
----:;;-JL I _ I/R(a: _ I)..., In R + 1
{(
'-X,)[R(.-IJx.-I]}[ .r R )].
N.. =ln 1-"'0 R(1l l):c,-I In,R+ I
Approximale Solution of Smoker's Equalion
An approximate solution of Smoker"s requatioD has been developed by Jafarey,
Douglas., and McAvoy.- The: result for the: rectifying section is
and N
s
is the same expression as for N ... For the case of salurated-liquid feed,
.... = x,: otherwise, x. represents the intersection of the: q line and the: rectifying line.
(A.'-32)
(A.2-28)
(A_'-29)
(A.230)
(A.2-JI)
(A.2-25)
R

R+l
.x;, = - II
N. ('n {x. [' - mD - I)x;,'(. - m<')]})(,n
I - mc(ll 1)"'0/(2 - mc
2
) me
J
I +(Il-I)Ic

R+I
%0 = %D (A.2.27)
k = -[m + b(1l - I) - ] J[m + b(<< - I) Il]! - 4bm(<< - I)
2m(2 - I)
whcreO<k<1
E.1l Smoh TrtIIU AIeA'}4 (19311). A Jafan:y. J M Doupu. and T J I<!F.C P,oc. 18 197 (1979)
446 SECTION .. 1 DIHU.....TJON Nu.. aEI. OF T..... n
PIST1l..UlTlON COlUI<lNS. tlUloIaEi. OF "BAYS 447
This expression resembles Fenske's equallon at total reflux, except that there are
reflux correction teons in the numerator and denominator. As R becomes very
large, Eq. A.2-37 reduces to Fenske's equation.
If wc use Ihe simple approximation for the minimum reflux rate given by
, R _
.. - I)x,
the operatmg IJIles and the equillbnum IJIles will be straIght, and the
Kremser equation should provide a reasonable approlnmatlon for the system. The
denominalOr terms In Eqs. A.2-37 and A.2-42 correspond to the absorption and
stripping factors for the two ends of the column, that .is, A. = L/(mV) = aR/(R + I)
for the rectifying seclion For nonnal column operation, I.e., when. RIR.. - 1.2, we
expect these terms to be more significant than Ihe refl.u.x corrccllon terms III the
numerator Therefore, we SImplify our solutions by wntmE
we can wnte Eq A2-37 as
'0
Nil = (A.2-39)
(
aR )" ( Xo )(' - X')
R + I = I - XI> --;;-
(
aCR + ')")"
Rx,. + I X w I - x,
(A.2-43)
(A.2-44)
Multiplying these two solutions and assuming Ihal Nil - Ns = Nfl, we oblain Now we see that the number of trays becomes unbounded as R approaches R",.
lienee, Eq. A.2-39 has the correcl limiting values. \
The equal ions for the stripping column arc developed in the same way, and
Ihe results arc
(
Rx )'.' (X )(' - xw)
.Y , D =SF
a Rx,. + I - 1 - xI> X ...
(A.2-45)
(A.2-40) 0'
10 SF

- In (aiJI + I/Rx,.)
(A.1-46)
where
This simple e;l;pression nonnally gh'es predictIOns that arc wlthm a few trays of the
e;l;aet solullon.
For the case of a saturated-hquid feed and a sharp separalion, thiS Ix:comes
'0
Ns =
(A.2-42)
Behavior Near Minimum Reflux
If we retam the reflux correction terms m Eqs. A.2-42 and A.1-45 but assume thai
Nit = N
s
= N12, Ihen we can derive the equal ion
We can use this result 10 estimate Ihe sensilivity of the change in the number of
plates as we decrease the reflux ratio below RIll. .. = 1.2. This result is expected to
be conservative. We can "tunc" the approximation by using Eq. A.2-47to calculate
N when RIll.", = 12, and Ihen we compare the result to Eq. A.2-23. If we introduce
a correction factor mto Eq 2-47 to make the resulls agree when RIR. = 1.2, we WIll
probably obtam a more accurate estimate.
This expression also reduees to Fensl.e's equation as R becomes very large and
predicts thai N
s
becomes unbounded as ll. approaches R...
SIMPI.IFIEO APPROXIMATE SOLLrrION. We expect that the approximate
solution presented above will be quite cooservathe, because the assumption used
to obtain Eq. A_2-37 is equivalent to drawmg the upper end of the rectifying line
through the point YI> = XI> = 1.0 (that is, we set XI> = 1.0 in Eq. A.2-28 to simplify
this expression). Thus, the operating lme is moved closer to the equilibrium line,
and more trays will be required.
For expensive columns wllh numerous trays, wc expect that most of the trays
will be located ncar the ends of a McCabeThlcle diagram. In these regions bolh
where
In {[xnf(l - Xl>)l((l - xw)/xwl!Jl
N = In [al(1 + lIRx,.)1I
2
]
(RIll. .. - x,.XRIll.. - I + aX,.)
fJ = (RIR", 1)2
(A 2-47)
(A.2-48)
448 SFM10"l.u I>lSTILl.AllON COI.U.... "'S NU.... RFl OF nAYS
stenON AJ
DISlllt,AnON COLU.... NS "'u.... n. or n.AYS 449
UNDERWOOD'S EQUATIO:,,/ FOR MULTICOMPONEf"o'TSYSfEM5. Underwood
ha5 .1150 proposed an expre$5lon for calculating the number of tnaY5 for mullicom-
ponent mlJ:tuTeS. At the operating vapor rate of the column. we sol.... for the values
or 8 and iT satisfying the expressions
(A.2-60)
(A.2-S8)
(A 2-59)
(A 2-S1)
(s 04
a,-ar SO.4
",
(I, + Ct,
1;[ "" 2
Hence, a criterion for the IWO kinds of 10 be approximately the same is
we sec that Eq. A 2-56 will be within of the arithmetiC mean if
'(J sO.I(ld
0'
Since the arilhmetlc mean is
(A.2-49)
,1I,DlC,D
V-
L
", .
Rcclifpng section
_''''''''1:
2
,8>:,.
a, - 0'
Then \Ooe use these .. alues to calculate Nit. and N
s
from the ClIpresslons for Ihe
rectIfying section
Stoppmg
(A_2-SI)
and the slripping seclion
EFFECT OF VARIATIONS TN a ON TilE COLUMN DESIGN, Gilliland's cor-
relatIon predicts Ihat the number or trays required to a particular
separation i5 approximalely equal to twice the number of Irays at total reflux:
(A.2-S2) (A 261)
A Critl'rion for Constant Rl'lalhl' VolatililY
Most of the short-cul procedules for eSlJmallng the number of theorcllcal trays
reqUIre Ihat the relallve \'olaUhty be constant. To de\'e1op a criterion for constant a,
firsl we evaluate the relative \olatihties of the hght key with respect 10 the heavy
key at the top and the bottom of the column
If \Ooe let
then
(I = (1:.,(1 + ,p) A.2-62)
21nSF 2lnSF 2N.
N - -
.., In (1.,(1 + ) - In Q' + In (I + ) - I + In (I + t;)j1n Ct
(A.2-63)
Then it is common practIce: to estimate an a\'erage volatility as the geometric mean
However, for small changes in Q' we can use Taylor series expansions to write
(
K,)
(I" == -
Kl "."om
(A.2-S3)
(A.2-54)
'"(1+';)-';
, .;
_ I - -
I + t;j1n Ct.. in a
(A.2-64)
(A.2-65)
and then usmg a Taylor expansion of the: geometric mean to obtain
RELATIONSHIP BE1WEE."'l TIlE GEO;\fETRIC lEA.... AND TilE ARITliMETIC
MEAN. To Simplify the analysis, ....-e write the: a\eragc volatility 10 terms of the
arithmellc mean rather than the geometric mean. The relationship bet....een
quantilles can be estabhshed by leuing
(A.2-SS)
(A.2-SG)
so that Eq, A.2-68 becomes
N _ 2N (1 (A_20 66)
""\ In a
Thus, variations in a will introduce less than a lOY. enor in the column
design if we require that
q, S 0.1 (A.2.67)
In a.,
Artcr substItuting FQ A.2-62 with a "" (1:, and the definition of Ct Eq. A.2-G1
becomes
(A.2-68)
450 StC"l'ON 4J DlnlLLArlUN COLUMNS NUMUJIl TUts
ncrioN 4J IHSTILLAriON COLUMNS "'UIoI8U or T....n 451
ratio untd we reach a pomt where we oblaln a sharp AB/Be split. The value of the
mInimum reflux rallo for Ihi) AB/BC split is
At thIS condition the fracllon of component nm the fttd that IS recovered o\erhead
..
Actual Tnl)'s-Plate EfficieoC)'
Asimple, but not \'ery accurate, techmque for estimating o\'eraU plale effiaeDOCS 1$
10 use O'Connell's correlation for fractionators; sec Fig. A.I-3. If l\e wnte an
equation for the curve in the range from 30 to 90 efficienClCS, we.obtalD
(A 270)
,
R -
LIl,X,.,-
0.4983 0.5 (A 2-71)
Eo = (cr:IlF)O lU - (cr:Il,)II.
Also, if we conSIder columns haying saturated.liquid feeds, and If we usc the result
that the viscosity of most liquids at their nonnal boiling points IS OJ cp, Ihen
OIK - I
I- radlon of B Rcco\'ered O\'erhcad ... !. = --I

SO .....e can calculate the composiuon of A o\'erhead X
AD
Then if .....e merely draw a
straight hne connectmg these points., we can estimate the yalue of R. for any value
of x
AU
thai falls between the values calculated abO\'e; see Fig. A 2-2
If we continue to increase the reflux rallO, .....e conllnue to increase the pumy
oflhe overhead as a linear function oflhe reflux ratio unlil .....e oblain a sharp A/Be
splil. We can estunale tbe minimum reflux ratio for this case where l"p = I, usmg
Ihe approximale results given in Table A.2-1. Then we merely draw a straIght hne
connecting the yalues for the AB/BC and A/BC splits (see Fig. A.2-1), and \\,c can
estimate Ihe values of R. correspondmg 10 any Olher value of lAIl. The figure also
ghes a plol of lhe ratio of the fraclional recoyenes of 8, !'h to the fractional
recovery of A,!", overhead, but these results arc nonlinear.
GlInos and Malone also considered the cost of sloppy spirts in lhe 0011001 of
a column. Agam, the results for a SImple flash, tbe ABIBC splil, Eqs. A_2-69 and
A.2470, and an ABle spill, Table A_2-1, prOVide Ihree points that define twO
slTalghtlincs. The graph IS SImpler to construct If lhe reboil rallO S IS us.cd Instead
of the reflux rallo (see FIg. A bUI the Ideas are exactly lhe same.
Sloppy Splits
MOSI of the research on column desl@n and shorlcul procedures has been limned 10
sharp splits in simple columns. lIo\\,ever. Ihe optimum fraellonal recovery for any
stream leaying a column, except for a product stream, represents an economic
trade-off between adding more tra)s In the end of lhe column (I.e., the top OJ the
boll om) where the Slream of mlereSI lea\cs the column balanced against ellher a
decrease m the loss of malerials In ",aSle or fuel streams or a decreased cost 10
recyclc the material through the reaClOr Shor!cul procedures for cstimating
the optimum rccoyenes ha\e been b) Fisher. Dohert). and Douglas' In
addition, a case study Ihal the dcslrabillty of ma},;Ing a sloppy spill (i.e.,
nOI a high fractional ra:u"ery) halO been prescnted b) Ingleby, R0lOSner, and
DougJas.
1
A procedure for cstlmatmg the O1lfllmum reflux ratlO for sloppy splits of
ternary mixlures has been published b) Glmos and Malone' and IS prcscnted
below They noted thai Underwood's equalloD (Eq. A.1.21) for the mlmmum reflux
ratio can be wnllen as a linear funcllon of Ihe o\erhead composilion of Ihe light
key x"p. providing thai only A and B Iea\c o\crhead_ Ifwe merely flash Ihe feed, we
obtain a lower bound on the dIstillate composition X"p, which corresponds to
R.-O.
Then when we add a recufYlng secuon aDd contlnue to increase the reflux
ratio, we will increase the purity of lhe O\erhead as a !lnear funcllon of the reflux
which provides a simple cnlenon for eValUJlJng the efft:CI of vanalJons In a on the
column design.
McCabe-Thiele lind Nonide:.lol
Binary
The McCabeThiele procedure often bUsed 10 calculale Ihe number of theoretlCd.l
Crays (or bmaf) mlxlures., particularly when Ihe relalne \'olallhly IS nOl conSlant
In addlllon, by lnlroducing fklluous molecular .... elghts, the McCabe-llue1e
melhod can be used 10 describe heat clfeus caused by laTJ!:e dlfferenas In Ihe heals
of vaporization of the l..eys. To account (or heal of mixing effects, II IS necessary to
use the Ponchon-Savant techOlque, a method deeribed b) Kmgln and Doherty,
or a computer program that sohcs the tray.by-lray malenal and energy
The details of the PonchonSavant procedure or compuler roultnes arc available
In texts on distillation.
(A 2-72)
'J R Knll'" and M F Doheny. IdtC '''''''.25 27<;1 (1986)
'w R hillel, M F Doherly. and J UOU1141. IdEC P,or fN$. 1'In 24 955 (1985)
's Inlkby, A P ROSSIler, and J Douglas, (Mm Itj/ Rn Du t>4 24t 11986)
I K Ghnos and M F I&.E' P'o< 1I<'t B 164 (198<1)
R It Pen)' and C It Chdlon, CNmuaJ I:JtglniU"s U",,<lI><wA. 51h cd, ,,-kG....... IILII. Nu ...
1911. p }-246
SfCTIOfo/ "1 PUK>N Of" GAS ....SO,UUS "'''0 OI'ITIu.... nON COLUWO;S 453
7
ABIBC
7
ABIC
The number of aClUal trays In the column IS then
N
N -
..,. Eo
ThIs value is needed to calculale the tower as ..e discuss now
A.3 DESIGN OF GAS ARSOR8EH.S AND
DISTILLATIO ' COLUMNS
(A.2-74)
2 3 4 5
Guthne's correIa lions (see Appcndi:l F I) gillC for the cost of plate
columns in tenus orthecolumn height and the column diameter. lienee. we need to
dCllelop a procedure for calculating the height and dlamet"-l. Both the gas
absorbers and distillation columns ale described by the sam..- models With the
Clceptions noted below.
RorS
(I\_J-I)
fiGURE A.l-2
Sloppy .plJt$ Of botloms oompollhon MQ rdluK 01 rtborJ ntio aDd fatio d fnctioDal
1'UlO>'CntS. Datk arde$ 1M cntocal cood,tJooI for a dw'p lpiit. [F K. GJit.s lINi At F
Atm-. IdLC 1', On. On. 23. 764(J9M). .lIl, of flw A 0tnruc.tI Sorry J
Costly separations correspond 10 low values ora. but for a range or ct from 1.3
10 3,the predicted by F..q_ A 2-10 only changes rrom 63 to 51 % Hence. the
overall effiCIency IS relatively insensitive to ct. and we can obtain a conservative
design estimate rOI atmospheric columns by assuming that
(A_2-73)
To'fter l'leight
In Sees. A.I and A.2 we de\'eiopc:d ellpressions ror the number or plates In a gal'
absorber and in distillation columns The most common tray spacing used ror plate
lowers is 2 rt. Hence, ror our initial designs, we assume a 2-rt spacing bet .....een trays
TOen the: height corresponding to the trays is 2(NIE
o
- I),
II is oommon practice to include some addilional space. say 5 to 10 f1, at th..-
top or the tower as a vapor-liquId disengaging space (.....e may even install demisters
at the top or the tower to prevent the carryover or liquid droplets). Similarly, it is
common practIce to Ifldude spatt atlh..- bottom orlhc: to.. er for a liqUid sump. The
hquid slream kaving an absorber or a distillation column orten is red Ihrough a
heat elchanger and lhen to a distillallon column_ Irwe want 10 ensure Ihal the
to another distillation tower is not interrupted, we maintain a 5-min or so supply of
liquid in the bottom or the tower. Th..- height or the sump required to hold this 5-
min supply ofliquid obviously will depend on the hquid now rate and the oolumn
diameter; but to simplify our preliminary calculations we assume that an addition-
al 5 to 10 rt at each end should be adequate.
With this simplification, we can write that the tOlll'er helghl IS
2N
H.., +H
o
E.
where //0 includes Ihe space at the ends of the column ror vapor disengagcment
and the liquid sump. This expression is commonly used ror lower designs.
As an approximation. "'C could express the additional space at the ends or the
to...er as a pcrcentageorthc height that contams trays. For our studies we include a
allowance for thIS excess spaC('. and .....e write Eq 1\ 3-1 as
(A.3-2)
454 SECIION AJ DESIGN 01 (.OAS ....SO... HS Atol> COll1MNS
0.02 0.03 0 05
Plate spa. ing
5
241'
-
18" - 3
12"
--
-
2
I:"-
--
--
,
-
,
I
-

7
5
I
3
20.3 0.5 1.0 2.0
0.7
0.'
o.
0.4
o.
o.
0.0
0.06
0.0
0.1>1
0.0
0.01
Q. O.

"
'3---

-
} .. CURl. AJ-I
EnlralRma>1
This should be a reasonable esllmate for large: lo.....ers ..... uh many trays, bUI II
probably will be 100 low for smaller lo.... ers. We use Eqs A.3-1 and A.J-2 for bolh
gas absorbers and dlSldlallon columns.
Flooding VelocilY and To"cr Diaml.'ler
The lower diameter usuall) is selecled so Ihat Ihe upor is belween 60 and
80% ofille f100dlRg \'e1OCllY. The early sludy of Souders and Brown- assumed Ihal
f100dlllg .....as caused by the eDlrammenl of droplets carned along wllh the gas; see
Fig. A.J-I When a small droplet IR Ihe lo....er IS stationary, 115 .... eJghl will be
balanced b)' Ihe drag force el.erled by the flUid
tIfR)(PL - PG)g = Co<ihl1)nR1 (A 33)
For lurbulent flo....., Ie., large particle Re)Dolds numbers. the drag coeffiCient IS
conslant, I CD = 0.44, so Ihat the \clccit)' is
fSiRJPL PG
v = .J3(OMj PG (A_3-4)
We can use thiS expression 10 estimate the vapor such that all droplels
larger than a gIven siu will fall back onto Ihe tray.
Except ror s)'Stems operating near Iheir cntical poinl, PG -c PL. Hence. we can
write Eq. A.3-4 as
.-IGURE A..l-2
Flo\>4lnI t"D"" for bubbk-e.ap ud pcri'<>r1llM p1aIG t..:(j_bqwd_lo-ps.au nuo al JIOIlll 01
COllSMkJauoa. (J R. Fau, I'mo CWm. DUO) 4j 19t1/).)
G.-\S ABSORBERS-FLOODl G VELOCITY. A more rigorous f100dlRg correla-
tion de\'eloped by Fair- is shown in rIg A.3-2_ that lbe dcnslty
appearing 10 IhlS correlallon are the same as those oblalOed If we wnte the
group appeanng on the abscissa In tenns of molar flow rates and IOsert our rule of
thumb for a gas absorber that Lsf(mGsl = 14, we sec that
(PG)'" L, M. (PG)'" _I4m M. (!-'.)'" _1.4 ('P") M. (PG)'"
S P G
s
M
G
PL Pr MG PL MG PL
L (A 3-7)
Normally absorbers are designed 10 operale at lemperalures well below.lhe
normal poiDls of the solule, so P"/P
T
=s; 0.5 or so. The liquid phase aCllvlly
coefficient otlcn is in the range from I 10 10; and if we CODSider qlSCS whcre the
solute is being recovered from an air stream IDtO waler at Slandard conditions, we
find thai
(A.3-5)
(A.J-8)
v = KoJpdpG
Also, the densities or mosl liquids are about Ihe same, so \o\.e can wrile
v..fP,; F = Constanl (A.3-6)
This constant F faclor provides a quick means ror eSlimaling flooding vclocllies
L ( )'" {18)f/f!-379 ('P")
_ Pi!. -1.4(10)(0.5 _ ---O.Q)- -0.15
G PL 29 62.4 Pr
However, we see from fig. A.32that the ordinate does not change very much
for (LjG)(PG/PLl
ill
in the range rrom 001 to 0.2. Thus, we should be able to
M Sowu!., h G G 8ro,"'II. M!Xs'ln of FracllOnallnl! Col .. lnJ CItjI eM.... 26 98 (19).1)
, It II PClf)' C II Chilion. ('MmWJ' blgm'J lIdnJboDJc. McGn,.H,Il. Ncw Yor1l., 1973. p S62
J R. Fair; II. H PcrT)' and C II Chillon, HUIldbooJc. jlh cd., McGraw-lllli.
New Yor1l., 1973, p 111-/1
456 SECTTON AJ
SECTlON AJ OSlGl'I 0.- GAS ,,1S(M.IEU ",..0 OISnllATlOl'I COL".....NS
TABLE A.J-l
Vapor t"cJocilies in
0'
...... B D Sm,th. -I)mll" 01 rqul1.o"'",", Slit""" r,ooesoa.- Mc<ira... lldl. cw Yorlr. 1963. P
J II. r.u ...., II. fl I'.,n.nd (' " nullon. -Chr""".1
lI.ndl>onl.- loth l M<G.... llill .. V<>rk. 19" f'
F= 1..5 =: c';;;' = IJjM
G
P.. (A.l-IO)
The: same: result is obtained for dislillation columns, and so we usc thiS elpression
for both gas absorbers and distillation columns.
(A,J-14)
(A.3-IS)
[ (
T. X" '))'"
DT = 0.01 64 fi 379MG --j;--
Limitations on the Design Conditions
It is undesirable to build very tall and skinny to.....ers because the towers will bend.
and might buckle, in strong .... Inds. Adesign guideline often used is that the column
height should be less than about 115 ft, bUI a beller dcslgn guideline is Ihat the
height-to-diameter ralio should be less than 20 to 30
Plate low-cn are seldom used If the tower diameler IS less than L5 rt.
area .., 1.77 ft z, because it is not possible to get inside the tower (through manholes
placed every five trays or so) to clean the tower. Instead, packed towers arc used if
the diameter is small. Either plate or packed towers are used ifthe column diameter
is less than 4.5 fl. area "'" 15.9 ftz, although the tray spacing normally is less than 2 ft
in this range. Similarly, if the required height is above 190 ft or so, we might
redesign the tower with a smaller tray spacing.
so Ihat the largest value of D
J
will correspond to the key component for whrch
Mr; 1; is largest When the two estimates of the tower diameter arc significantly
different. we bUild a lower, called a SYl'cdge: in two sections ha\'ing the
appropriate diamelen. We prefer to a\oid these swedge columns. howe\er, as they
arc more elpcnsi\e 10 build.
The dIameter IS retaln'ely ino;ensith'c to changes in the operating temperature and
pressure.
SinCt: yapor rate in a distillation column is different abo\'c and below the
feed Iray, we need to c\'aluate the tower diameler for sliIls at these two locations If
the Iwo values arc nottoa dlfferenl, lA'e design the tower to correspond to the larger
diameter Assuming Klcal-gas behavior and salurated-IKIuid reeds, lAe can IAnte
that
Once we have Ihe tower height and diameler. we can use Guthric's
correlations to estimate the tOlAer cosl; Le., we cost the shell as a pressure vessel.
and then we add the cos! of the"ays. As a quick approximation, we assume
that thc cost of the trays IS about 20'. the cost of the column shell (assuming that
e\erything IS carbon steel)
Shortcut Cost PrOdure for Column Trays and the
Sb<1I
CA.]- J])
_(4',)" 0, _

T,ay SJ*=III&. r,
3D
'0
"
'0
On:hnalc
0425 0.33 .26
."
flood,n&. p<; - F, 3,24 ...
,.,
60% of "OO<I,nl:. F
'"
151
"'
017
obtain a esllmate of the flooding velocity by selecting the values at
(UGXPGiPJ = 0.1 These values arc listed in Table A.3-t The ordmate on the
graph can be wnnen as (If \lie assume that PI.. _ 5SIbjft' and 0 _ 20)
,/
20
)"( P, )IIZ _v, {P;; _ v,J;;; _ F,
'\ 0 PI. PG Vh J58 J58 (A.3-9)
Then If wc consider that lAe desire to operate at of the f100dmg velocity, we
obtamthe other values listed in Table A.3-I. Most of our preliminary designs are
for towers wuh 2-ft tray spacings, and we assume that
TOlAcr Cross-Sectional Area and Diameter
To find the tower diameler. first lAC convert the molar flow rate of vapor 10 a
\olume:tnc flow Q = (Vmollhr)/(p.. mollft'), and then we dhide by the \'elOClty:
A
T
- V'P..
J.5(
3600
ljM
G
P", - 1.5(3600) (A_3-1I)
Actually. the cross-sectional area mUSI be larger Ihan this value because about 12 %
of the area is laken up by Ihe downcomers. Hence, we write 0
AT Z VjMt;/p.. = 2.1 x 10-" V(::Gy
'
2 (A.3-12)
. Once: we have estimated the column area, it is a simple matter to calculate the
diameter
Distillation Column AU:l.iliarie!.
For dlsllllalJOn columns, we also mUSI design lhe condenser and reboiler. In
addition, we must find Ihe coollllg-waler and steam requirements
COLUMN CONDENSER AND COOU1"oiG WATER. We consider the case of a
tolal condenser, which is the most common situation Then the condenstr heat
duty is the heat required to completely condense the vapor passmg o\'erhead Wilh
coohng water available at 90
c
F and bemg returned at 120F, heat balances give
120 - 90
Oc=J1HvV=UcAcln[CT. 90)'(16 120)] = M. cC,.(I20-90) (A_3-16)
REBOILEM A 0 SlUM SLPPL"I. If we us.e steam to supply the heat to produce
V mol/hr of vapor at lhe bollom of the lower, a heat balance gJ\'es
O. _ MI., P 0& UIIA. J1T.... Us Mis
. be k t be conslraJllC'd to be: less than
The dn\'lI1g force In the re I r mus . . I f
be t 301045' F to prc\'i:nt film bOllll1g. We expeci to obtam a high \a uc u
au, . bol because have heal lransfer
the overall heat-transfer coeffiCIent 111 the re I cr h "._
I
I 'd TI we ex.....ctlhatl ere WI .....
between a condensll1g vapor and a bol mg IqUl. lUS, ,--
a hmlting heat ftux III the reboiler, and we assume thai
(A3-23)
We nonnally assume that an overall beaHransfer cocffident for the condenser
U
c
= 100 Btu/(hr ftl. <>F) gives reasonable results. Hence, the required heat-
Iransfer area for the condenser is
With Ehis approximalion the required heaHransfer area is
(A.3-18)
If coohng water COSIS SC." 10) gal, the annual COSE of coolmg water VI,U be
Similarly. we can find Ibe capital COSI of Ihe condenser by using Guthne's
correlation (sec Sec. EI). To develop a shoncul model, we consider an exchanger
with 1000 fl
1
of area, so Ihat the purchased cost is $8900. The cost exponeDl is 0.65,
and we consider a ftoalinghead CJlchanger, carbon-steel conslruction, aD operaung
pressure of less than ISO psis. aDd an inslallalion factor of 3.29. Thus, Ihe inslalled
cost IS
(A.3-29)
(A.3-26)
(A.3:H)
(A 325)
(A3-24l
A.3-20, and we
S}
{
OH )'" s (M& 328 __I va "s
ell 280 11,250
SC, P IbX81SO hr)
Ann. Cost = 1000 Ib 933 hr yr
= 8.74 x 10 l C
s
6Hp P
The mstalled capllal cost of the reboiler can be estimaled flOm Eq
find Ihat
and Ihe reqUIred steam supply is
where
For the case of25-psl Sleam, where T
s
= 267"F, MIs = 933 81ul1b, and the
is SC
s
/t0
3
lb. lhe annual stcam COSE is
The vapor rate in Ihe bouom or the tower depends on the quaIllY of the reed II
P=V+F(q-l) (A 3-28)
(A.3-I7)
(A.3-19)
(0-20)
C
(
SOOIC,.)( "")(0", 'b)( h')
Ann. 051 = -- -- v- 8150-
1000131 8-341b 30 hr yr
= 3.26 x 1O-C.. i1HI'V
[
Mlv r.-90J
A
c
= 30(100)1
16
120 V
and the required ftow of cooling water is
c,. _ '5
C
A
_
Afler subsiliuling Eq. A.3-17 inlo this expression, we find Ehallhe installed cost can
be wrillen as
0'
C _ (M&S\.328/AH, In -r. 90)0.) ItO.'
C 280 f '\3000 T. - 120
(A 3-21)
VI hlCh IS lhe heal rcqulred 10 convert I mol of feed 10 a saluratcd vapor dl\lded by
lhe molal lalent heal of v3ponzauon
460 "J Vt51UN Of a ...s ... aso..IlUS """0 I>ISnll....rlON COlUMI'fS
Summary of the Cosl Model
Ifwc combIne the expressions above, we can write a model for the total annual cost
(TAC) of a d'stiliallon scpar;lIlon (i.e,. a capital charge factor of! yr is used to
annuahze the lII$tallcd eqUIpment cOSt) In terms of the design variables:
TAC "" Column Shell and Trays + Condenser + Reboiler
+ Cooling Water + Steam
I (M&S\ (t.
3 280 ,1201)1
11
)(2.18 of fd
I (M&S\,
+ 3 280 f 101.3)(229 + II' + A: + Steam + Cooling Water
- (101.9)(218 +
x (2':' + 11
0
)"
,."
M&S /611 T. 90)""
+ 3(280) (101.3)(2 29 + In li.
t
-120 6S
M&S {611 )""
+ )(280'-- (101.3)(229 + Fe --'- v""
. ) 11.250
+ 126 x 10 C.. MI, V+ 874 x 17 (A3.3D)
COL8LRVS COST MODEL An alternate cost model was de\elopcd b)'
Colburn The form of the cost modd IS
(A.331)
where CINIS the cost olthe column shell, C
J
mcludes Ihe cost of the condenser and
reOOiler. and C) mcludes the cost of the steam and cooling water. An eumple that
illustrates the applicatlon of the procedure IS gIven in Happel and Jordan.' USlllg
Taylor series expanSIons, \\,e could reduce Eq A.3JO 10 Eq. A.331.
PICked Versus Plate Toft'crs
For prdmunary process design we are concerned primarily with the most
expensive pIeces of equipment We expect that the expensive towers will have large
vapor fates and dIameters grealer than 45ft Hena:, normally the expensive towers
A r Calb",n "nrero an T ASn Sum",.. , "ich''01
fa. 11K Teachmg uf l'herrllCllIIn(l.llltt,mg, S,ale 1936
'J Ibprel a,,.] l) C, O,,,.,,rnl r",."rJ, 2d o.,Her. N..", 1975. p lK'
SECTION... OlSnU.AllOroI SEOUENCTrora 461
will be plate columns. Tray towers arc preferred when the solvent flows are small,
i.e., packing .....ould not be welled uniformly, and where internal cooling is desired
BUI packed towers are used when the materials are highly corrosive or foam
badly, when low-pressure drops are required, and If the to.....er diameler is less than
2 ft. The height ofa packed tower is estImated by calculatmg the number of transfer
units and the height of a transfer ulllL Often, shortcut procedures such as those
given in Table A.I-I Will gIve reasonable pre<hctions for the number of transfer
units. The height ofa transfer unll can be predicled by a variety of correlations that
ha\'e been re\'iewed by FaIr However. to obtain a quick eslimate of the tower
height. Jordan' recommends lcnmg
IIQG _ 2.0 fl (A.3-32)
ThIS estImate should be reasonable for pacling <;izcs from 0.75 to 1.5 in. and for
gas now rates in the range of 200 to 1400 Ib/(hr ftl)
A.4 DISTILLATiON COLUMN SEQUE CING
In Soc.. 7.3 ",.. discussed the use of heuristics and computer codes for selecling the
best sequence: of dIstillation columns. I fowever, as an alternative approach. we can
use our order-of-magmtude arguments 10 SImplify the problem. By Simplifying
problems of this tnlt:, we often Impro\"e our understanding of the most important
features that dominate thc design The procedure we present here is due to Malone
et al. I
Cosl Model
For lhe sake ofsimplicit)'. we consider ternary mixtures. and we ...-antto detennine
whether it is bener to use Ihe direct sequence (i e.. splil A from Band C and then
split B and C) or the indirccl sequence (i e, splil C from A and B and then split A
and B); see Fig.. 7.3-3. We base our deciSIon on the lotal separatIon cosls of the two
sequences. and so we need a cost model for a distillation separation. We use
Guthrie's correlations (see Sec. E.I) to estimale the capital costs of the column,
condenser, and reOOiler, and then we add the cost of the sleam and c'ooling water
Thus, we the cost of a single column as
(A.4-I)
where Co is the annualized cost coeffICient for the column shell, NT = number of
theoretical trays, V = vapor ratc., C
l
is the annualized cosl coefficient for both the
R H '-.. rry and C II Clullon. CA"mi<>,,/ ":;"g",u", I/n"d/IaoJI.. Slh o:d. McGr.",.1I1Il, Ne'" York,
1973, KC. 111.
10 G.Jordan. C'wmfrQI "'<>Un IkwWpm"",. vol ,. New York. t96ll. p 4SI.
'M Jo Malone. K Ghnos. I f Marque!,.OO J M 1>O"ltU, A/CAE J. Jl 68J (19115).
462 SEcnON A' DISTIu..TION COl 11),11'1 nQtJEI'lCING
SECTION A' OISTIu..noN COlUMI'l SEQUENCING 463
condenser and rcboiler, and C2 is the annual cost coefficlcnt of the steam and t.he
cooling water. If a base-case design is available, we can use a Taylor senes
expansion 10 wrile Eq. A.4-1 as
TAC=Ko+K,N,+KlV (A.4-2)
This is the cost model thaI we use 10 eva!llate column sequenceS. Now if we
considcr the difference in cost between lhe direct and indirect sequences, using
Eq. A.4-2, we obtain
,HAC = + - Nk)
+ - + - "k) (A.4-3)
According to Gilliland's correlation (see Eq. A.2-23),
Vapor Rate
The vapor rate in each column can be related 10 the reflux and distillate rates by a
material balance
along with the approximate material balances that correspond to perfect splits
(rather than greater than 99% recoveries)
D,X,.D = FXiJ (A.4-1O)
Ihen it is a simple malter 10 estimate the vapor flows given a knowledge of tbe
minimum reflux ratios.
The common procedure for calculating reflux ratios for constant.volatility,
muhicomponent mixtures is to solve Underwood's equations
(A.4-8)
R, = I.2R._
1
(A.4-9)
II; = (R; + I}D,
H .... e usc the rule of thumb
(A.2-23)
NT - 2Nmn
NUMBER OF PLATES.
when R - 1.1R...in ,
(A.4-13)
(A.4-15)
(A.4-II)
(A.4-12)
(A.2-16)
1- x
c
,-
I
R
(0: - I)x,-
J = I + r/sox.,-
where
Instead of using Underwood's equations for the ternary mixturcs
(Eqs. A.4-11 and A.4-12), however, we prefer 10 use the approximate expressions
developed by Glinos and Malone so that we can obtain an explicit expression for
the vapor rate; see Table A.2-1. Their results indicate that for
where we use the appropriate IJ and feed composition for the last column in either
the direct or indirecl sequence (sec Fig. 7.3-3). i.e.,
For the direct sequence, we solve Eq. A.4-11 for the value of lJ in Ihe range
!%AC < lJ < tX.cand substilute this result into Eq. A.4-12to find R.. ; for the indirect
scquence we determine the 0 in the range O:BC < 0 < I and again use Eq. A.412 to
find R. The minimum reflux ratios ror the remaining binary columns can be found
by using lhe shortcul approximation of Underwood's equation for binary mixtures
.nd
A/Be:
(AA-6)
(A.4-5)
The same result will be obtained ror lhe BIC split. Thus, nonnally Eq. A.4-3 reduces
10
This result corresponds 10 the well-known heuristic
Selectlhe column sequence based on the smallest total vapor load. (A.4.7)
LIMITATIONS OF TilE VAPOR LOAD HEURISTIC. Of course, we see from
Eq. A_4-3 that this heuristic will be valid only if the cost coefficient K I is tbe. same
for each column. However, if C were a corrosive component, bolh columns ID the
direci sequence (Fig. 7.3-3) would ha\'e to be made from
whereas only one column in the indirect sequence would
protection. Thus, the cost codlicients would not be the If the
columns operate at different prcssure le\<ds m the dlrecl and sequences,
then COSI coefficients will be dIfferent Thus, the vapor-load heunsllc IS not always
valid. Also note thai Fenske's equalEon is limlled 10 mIXtures ha\-ing constant
relative volatilities.
where Ihe mimmu.m number of plates is given by Fenske's equallon. If we :-v
rite
N . in tenns of fractional reco\'ery of the light key overhead '1 and the fracllonal
_. ,.
recovery of the heavy key in Ihe bottoms '" lhen for the AlB sp It,
N _ 21n {[r.. /O - ,..)][,./(1 - '.} (A.4A)
A' - In QA'
Howe\'er, the fractional recoveries for the AjB split will be lhe same for both Ihe
dIrect and indirect sequences (Fig 7.3-3)
464 Sf('"lION ...' OlnllUl flOJ' lUlU...N SEQUENCING
(9.6. 1)
x.
D
1\
Va = V,
0" ,
x,
Ie)
SECHON ... OISTIlunON COLU... N SEQ!JEI"ClNG 46.
,
\
\
\
\
\
(9. 3. 1)
\
(9. 15. l)
\
\
\
\
\
\
\
x.
Va =
V,
x. I a=
'I'
\
I
\
\
I
\
\
D
I
D \
\
I \
\
I I \
I
\
0
,
"
0
,
"
x, x,
la) (0)
I
FlGURt: A.4-1
Bounds lor the dircd and lad"cd sequences (F,,,,,, J( GI_...JM. F M"'-,
Coluwon Alrmwllll't'J '" Oo,.1u,1"", SIJlf'>OtJ. ID Cw-. u., Rn 19/17)
(A.4-18)
(A.4-16)
(A.4-17)
Use dlrecl sequence when
Discussion
The rc:suh. Eq. A 4-11, plO\"Ides us with an approximale, explicit expression Ihat \loe
can usc 10 determme the best column sequence as a funclion of lbe feed
compo'tltlon to Iht dislillatlon train. Thus. "'e can use Ihis IC5UIt in place of the
heunstlcs; Stt Sec. 7.3. Malone: et a1. present a comparison of Ihis expression to Ihe
pubhshed heurislics as wdl as discuss Ihe sensiti,ity of the solulion. That is. the)
descnbe Ihe conditions under which Ihe wrong cboice of lhe column sequence
corresponds to large cost penalties. This type of sensilivity information is often
more usdul for prdlmlllary desIgns Ihan obtaining a rigorous solution for a
particular case
Using Bounding
In anolher study, Glinos and Malone' ploued lhe line where the total vapor ratc:
for the direct sequence was c:qualto the lolal vapor rate for the indirect sequence: on
the plane of the feed compositions; sec Fig. A.4-1. The relative volatilities are
shown 111 parentbeses, and Underwood's equalions for sharp splils were used fOl
tile calculalioDs. From an Iflspeclion of Ihese diagrams, they proposed lhese simple
cntena
(.\:Jf + Xn)/(cr
K
- I) + XAFf(crlK: - I)
R = .
.. (\:iF + x.FXI + XAFXCF)
Now, if we combine Eqs A4-6 and A.4-8 through A 4-16, we oblam
Whenever Ihis result is negallve, we prefer the indirect sequence.
AB C
and for
M I K GhnO'l. I r "blql.iC'l_ and J l}(>uSlu. AIO,F 611\
K (;l1nOl and /It F MaloTK fo<' ... Col""", Alrnnarn"" ." Du""...."'"
.'ill"'....: submlllW 10 (h".. fttq R... INJ. 19117
It IS Interesting lhal Eq. A.4-18 agrees with lhc: heuristic to selecl Ihe direct
.sequence when the arnounl of the lowest boiler is smaiL Also, Eq. A.4-19 is a similar
result for the indirect sequence when the heaviest component is a large rraclion of
the feed. However, for 8 case where lite volalililies are (9, 3. I) and the reed
composilions are .x.. XII '" 0.15 and xc'" 0.7, Fig. A.4-1 shows that the direct
sequence is still favored Similarly, we see rrom Fig. A.4-1 that the region where the
mdlrcet sequence is besl shrinks as the AlB spilt becomes more dIfficult than the
Die split. which contradicts the common heunsllc "do the easIest splits first and
leave the difficult splits until lasL" This result again shows the danger of using
lteuristK:S.
(A_4-19)
(A.4-20)
Use the I1Idirecl sequence if
Calculate the vapor rate if
I.l > X
A
> 1.
x
A
+ Xc
they lise bounds to develop \'ery simple expressions for some
'" COMI'LU OImUATlON COU'IotNS 467
A_S COMPLEX DISTILLATJOi'\ COLUMNS
Normally we do not consider lhe: use of complCl dl)ullatlon columns in our mlllal
design, except for the USt- of pasteunlauon columns to remove light ends from a
product stream (sec Sec. 7.3) Instead, 1Ioe: first look for the OOt sequences ofsunple
columns, and we evaluate the profilabllllY of the: process If additional design effort
can beJustified. we examine the posslbl1lty of replacing two adjacent columns In the
SImple sequence by complex columns, 10 see whether we can reduce the separation
COSIS Remember thai if we can reduce lhe recycle: COStS of a reaClant for a reaction
system where Ihere are slgmficanl b}-product losses, Ihen 'he might be able to
translate these: separation s),stem savmgs to rallo-matenals sa vlngs If we reoptlmlze
the process flows.
Mosltexls: about umt operations do not include very complete diSCUSSions of
complex distillation columns Thus, a boef introducuon IS presented here for
sidestream columns, sidestream strippers and rectifiers, and prefractionators and
PetJyuk columns These results are laken from various papers by Malone and
coworkers
(a) DIrect
A.B,C
A

c
(b) IAdirccl
B can be estimated by using the expressions below for the case of.a high-punly
o\crhead
(A 52)
FlGURE AS-t
Cot..mn scquc:nccs.

c
A.R,C
Sideslream Columns
We: can someumes usc a smgle sideslfeam column 10 replaa: two columns m either
Ihe dnect or the indirect sequence; see FIt! A.5-1. If the mdirect sequence IS favored.
we say that Ihe "primary" separatIOn corresponds to Ihe ABle split, and we
replace the two-column sequence b) a single sldeslfeam above the fecd, see
Fig. A_5-2 Ho.... c"er. if the direct sequence is favored, wc say that Ihe "primary"
separauon corresponds to the AIBC spill. and we replace the two-column sequence
by a single column with a sidcslrcam below feed
For a sidcstre:am above the feed (or below 11). we DOte from Fig. A.5-2 Ihat
are: only three column sections compared 10 the four sccllons available ID Ihe
indirect (or direct) sequence. We also DOle that the AlB (or Blq spltttakes place in
only the upper (or lowcr), single-column section. Hence, this "secondary" separa-
tion is limilcd by the vapor rate required for Ihe primary separation.
Moreover, since we have only three column scclions, we can no longer
achieve any purity that we desire:_ The recoveries of Band C or the composlllon of
Ihe bolloms can be specified arbitranl)', and Ihe composition of A 10 the distillate
can be fixed at any value. However. Ihere is a maximum concentration of B (and a
minimum conoentration of A) Ihal will be oblained 10 a sKlestrcam above lhe feed
even ifan infinite number of trays is used in the upper iCClion. Glinos and Malone
showed that thiS minimum concentration of A and the maximum concentration of
where R2 is the reflux ratio that corresponds to the primary split (AlJ/c) at the feed
plate
." Glmos and M F Maloac. J&C /'roc fJrj 14 1011 (I98S) aDd 13 764 (I9S4) (A.5.J)
468 nello", '" COMf'LL'C l>lsr,UHlOt< COLU,,""'S
SFCTIQI'l'" COMPIVl ounU"'O'IIO.... COLUIolNS 469
The value of R
1
. ..,;_ can be estimated by usmg Undcr.....oods equations or the
apprOXImate ellpressions ofGhnos and Malone- (see Table A.2-1) for the: ABle
split
(A.5-6)
(A 5-5)
(A 5-8)
(A 57)
1)
R """ x..,. + x.,.
.. xAf(a.c - J)
From Eq. A_5-5 .....e sec thai if CI... J> I and/or if x ..,. is small. then we can obtain a
product with high purity as a sK1eslream
The correspondmg results for sidestreams below the feed are
a.,u{x.. ,. + X.,-)
,t:cs, "'"
:I...X .. ,. + x.,.
PASTEURIZATION COI.UM....S. For the spec;al caseorsidestream columns ..... here
lhe desned product IS the tntennedlate boiler and there is a lIIoaste or fuel by-
produclthat is either mueh h(!.hter than the produet (so ....-c rcco\cr the: product asa
sKlcslream abo\'e the feed) or else: is lIlueh heavier than the product (so lIIoe reco\el
lhe plOducl as a sldestream belo.... the feed), Glmos and Malone- find thaltheyc:m
obtam very simple solullons_ For the case where CI ..
c
J> ClK and the sldC"lream is
above the rettl.. the IIllnlmUm amount of Impurlly m the product will be
X..,(Cl
ac
- I)
t: .. s, .... = (O:lK
X
.,. + xc,-)(O: ...
plan to recycle thiS stream back to a reactor and that the Impunlles do not affect
the product distribution). we can select t:.... :: 0: XC.II "'" 0; either the overhead
punty of A. t: ..
n
or the fractional recovery of A overhead (but not both); the purity
and the fraclional reco\ery or C in the bolloms: and the composition or the
hEhtest componenl of A in the sidestrcam, t: ...5 NeJ[t .... e R,,_, using
Eq A 5-4. and then we let R
z
"'" 1,2R
z
._' Now. we can evaluate x
AS
...... by \lslOg
Eq A,5-I. and .... e sec .... hether our SpC'ClrJCalion of x,,$ cxceeds lhls value_ Ir It
not......e must chanEe our spcc1fication ror x...5-
Assuming lhat lhe desired pUrlly can be obtained. Gllnns and Malone .....ere
ahle to pro'<e thaI a column .....lth a sKlestream ab<H'e the feed .... ould al ..... ays be
cheaper lhan the correspondmg two columns lor the indirect sequence. About 10 'Y.
fewer trays were required, and vapor savings up to 30% can be obtained. The
largest savings correspond to balanced volatilities, although the sKlestream punly
decreases, For low values of X .. r. high-purily sidcslreams can be obtained, but lhe
\apor savings decrease,
and the minimum vapor ralc is given by
(A.5--4)
FlGl.R r. A.5-2
Sl(\nl,um .bo.'c lhc rd (I "I'" K GIUOO1 and
/It F MalON. fiOpl/malm RnjllHl' (0' Cr>mpla
Column A/lrtIfQIll'CJ In DUlIlIll"lI'I S,"c......- Jub
","'rd'll Cltr", Eng RCJ Dr.'. /981,)
v,

U
L, X,.J}
I
I ,
I I I.,
p
I'
I I
L, X,.P ,
2
F
I,
I I
L,
X"F
I,
t I
L,
J
I'
t,
,I
I
"

X,II
t:.. r!(7-
AC
- I) + (x" + XO)!(ClIIC - I)
R
1
... ,... - -
. (x..,. +x.,.)(1 + x..,.x
c
,)
The approlllmale ellpression is usually Wlthm 4 or Ihe euet result, which is
adequate for screemng purposes
DESIGN OF" COI.UMNS wnll SIDESTRf..AMS ABOVE TIlE FEEl>, If ....eha e
a where a high pUrity of lhe sidestrcam IS nol required (c g. suppose thal e
s x.,. +- xc,- "c,-/x
c
...-
F xcs/x
c
..'
(1\5-9)
470 UcnON" I roMI't..U IllrnU-"lION 5EcnON ,"-I COMI'l.EJr IlISlIU....TIOJ' COlUMJ'S 471
SII)ESTREAM STRIPPERS. For the <:onfigurallon shown In Fig. A.5-J, we use the
specifications and overall malcnal balan<:cs to fix Ihe: c:\ternal flows. Then, to
Thus, we have falfly simple a\'allable for lhe preliminary of
pasteurization columns
Sidesfream Strippers and SidClltrcam
If lhe pUrity reqUIrements of a siJestrcam column are not sallsfactory, we can add
a column secllon, as ellhe:r a sidestream stripper or a Sidestream rectifier (!iCe
Figs. A 5-] and A.5-4) Now we agalll have fOUf column sectIOns, so that any
spc<:ific:d recovencs and purities can be obtained.
,P
o
X"D
I
P
V,
I
X,
3
2
I
V,
I
I
I t v,.
4
W
F
t
V,
L,
V,
V,
L, I I
X"D
o
F
3
2
x"w
fiGURE A5-f
S,deslream rccufu:r. (From K. Glinol and AI. F. -Opumaltfy Rt(J'fJft1 fm C<>mplt< Column
Uf Dilullaua" SYltems,- wbmtlltd ItJ Ch,rn &g Rn 0trJ, J9IJ7)
estimate the internal flows VI and V
3
, Glinos and Malone developed the
expressIOns
4
(A.5-IO)
(A.5-11)
nGURE A.5-J
Sldoueam Mnpper (f,om A Glmos and AI f' M"I""" Il"lWIU (", C_,plo Cmu"'''
AI/e,,,utllJ<!S '" Oi.mll"tlm, 5,.1'trr<1: sYb"'mtd 10 Clotm En" Rrs Or, . IWJ7 J Ii: Ghnos and M F Malonr. I&t-.:c P'M f)es Dno. 24 1087 (t985) and 23 164 (19114)
472 SEOIOW .. , COMPI.P.Jt OISnu.ATJOlII COLUMWS SECnOl" ... , cxn.. "nx OISnI..LAIlON COLU"''''' 473
where 0
1
lS the fOOl of Underwood's equation
(A.5-12)
in the range !lac> OJ > I For (he case: of 5aluratcd-liquld feeds, they also
de\ldopcd approximate solutions for Under-".ood', equalion $0 that no Iteration is
requIred Also. a procedure for c:sllmalmg the: number of trays \Olas dc\<doped by
Gllnos
SlDESTREAM RECTIFIERS, For the confi!uratlon shown In Fig. A.S-4, Ghnos
developed an upres.sion for cSlimatiog the mlDtmUm vapor rate
4
5
6

2
provides the renux for both columns. ThiS configuration is orten called a Pell)'Uk
columll.
If we make a sharp Ale split in the first column and fix the product nows to
match our specificatlon<;, tben two mternal nows must be fixed 10 be able to
calculale all the other If we I.
J
, we will c\'enlually reach a limitlTlp'
condllion al eHher the upper feed or the lower feed or the column.
dependmg on wluch feet! conditIOn IS controlling For the case were we make a
sharp AB/Be or a A/BC m the prefractlOnator. and the upper
nCURE A.5-.5
PW}U colWlllL {F,,,,,, K GfonoJ DNI /of F /ofalOit', -0"1..-1,,, Rrflt/ftS ffJl' (:""'1"0 Col-
AI_,OWl '" DuJJJa,_ S)"Jlrm$.,- la 0",.. bog Rn. iNt'_ 1987)
(A.5-13)
where (J. IS the root of Underwood's equation In the: range (lAC> 0. > el!IC_
Gimes and Malone found thai for tbe case where q = I, the 10lal vapor
gencralcd fOf the two rcboilcrs in a sidcsueam stripper is cuetly Ihe same as the
\'apor requirement for the singk rcboikr io a sideslream rectifier The operating
costs may differ. ho.....ever, Sllla: Ilk- reboiler of lhe sidestream stripping section
operates at a different temperature and tlk-rerore may usc: a less expensive utility.
Thermall)' Coupled Columns-Pt'llyuk Columns
For a ternary mixture A. B, C. the easiest spin possible is bet....een A and C, letting
8 distribute between the top and bottom of the column. Then either we can split
the AlB o\erhead and the BIC bottoms stream in two additional columns, or we
can combine these columns mto a single column (sec: Fig. A.,SS). With this
arrangement. we need only one condenscr and one reboiler for both columns: i.e..
the columns are thermally coupled since: the reboi1er from tbe downstream column
supplies the \'apor for both columns and the condenser on the downstream column
POTENTIAL SAVINGS. GllIlos and Malone' proved that these sideslream
columns require less total vapor than either the direct or indirect sequence. In
addition, they showed that the vapor savings are alll-'ays large when x.
r
is small,
whIch agrees with tbe results of Tedder and Rudd (see Sec. 7.3). The maximum
savmgs possible is 50%, independent of the \olalilities. and tbe maximum savings
is obtallled when X.tr" (<1.11 - 1)/(<lAc - I) and X.
r
approaches 7.ero.
K Ghnos."11 010ball1pp'N1ch to the l'rchm,na'l Del.,,,, and Symt.e.l< of D""IIHh<>n I ,a",,: I'h I)
Thet:IS, Un,,,,,,,,,y of /l.bspchuO'lll. IImhenl. 1984
K (ibn", and M F 01""",,/"1 RrgI(JPI$/tJ< ('omr1u Cal""", AII... """'......n Dnrdlauon
STJlrfftS." f'&['C' submJlled 10 Cltnro f"" RrJ 19S6
474 SKIION .. , C...,... P1UDUTIU .. TIoNtOII .......
steCIIOI'l .. , cu.. p/.U: Ol."UATIOI'l lOl.U....i 475
fud poml IS controlling, GJmos and M.lonc liohol'ted lhill (aSliooumlOg lhal lhe
prefracllonalor b also at liS limning condllln)
> I :.... dnd a shdrp AS Be !>pbll!> accomplbhed In l>CCllons I and 2. For
case Ihe fraction of B reco\ered overhead in Ihe prefraclionalor is
(A 5-14)
aac - I
f.- --
(lAC I
(A 5-21)
where 01 IS the root of lInderlAOod'. equallon, Iq A,5-12, m the range
a AC > 8 > aBC' BUI for euher a sharp -18 Be Spill or a sloppy ABC split in the
pre(raCllOnlHOr (whIch IS operal mg al 1IIIIIIJng conditIOns). and lhe lower feed point
controls, lhe result IS
v
6._ 8__ I
,
and lhe amounl of B fed 10 Ihe downstream column al lhe upper fced Io;;a\lon is
f.x.rF Column sections 3 and 4 operate al a condlllon aoo\e the 1n1l1l1TlUm. so
thaI Ihe)' can handle larger amounts or H. Thus, we could have deSIgned for a
sloppy AB'C splil and talen more B overhead. 1I0wel'er, there is an upper bound
on Ihe fraClional recovery of B overhead In the prefraetlOnalor. say 1.,_.. I'thlch
corresponds to lhe snuallon where the minimum vapor rate III column secuons 3
and 4 becomes equal 10 1' .. _. We wnte thes<: bounds as
Glmos and Malone also nOled thaI A.5-14 corresponds to the mimmum
vapor rate for a sharp A/BC spin, so thoU \Io'e can also lA.nte
where 81 IS Ihe rool of Underwood's equation III Ihe range a.: > 9
1
> I
If q = I. Ihen mass balances can be- used 10 thai I'j = V.. and
(A 502])
(A 5-22)
I't. = (lACXAI' + (l;x.ff.
I.... (lAC - 01 (l.: 9
2

.... here 02 IS again Ihe roOl of Underwood's equation in the range (lB(" > O
2
> I.
Slmilarl)-, if lhe upper column controls, lhen column sections 5 and 6 can
handle more B lhan corresponds to OJ sharp split III lhe prefraclionator ThaI is, lAC
can perfonn a sloppy AIBe split 10 the prefracllonalOr, bUI now we encounter a
I".... er bound on the fracllon of B tale.n overhead I. _. lIence, can wnte
(A.5-16)
(A_5-17)
x" ]}
8
1
_ I F
V
{ [
0"',,
"'.a'a == max
aAc 01
(A_S-18)
(A 5-24)
(A.5-25)
lAhere 1',,_. IS nollt e\'aluated by uSlOg l:.q. A5-23.e:a.ttplthal 8
1
1S replaced b) 8,_
Fidkowsli and KroM.owsljl developed an e>.prCSSlon for this bound on lhe
fractional recovery of B:
f _ (iliAC - a.dVj,aun -
....... - (iliAC - I)I'J ....I.. - (aACXAr + xllf)F
The results abO\e indicate that a Petlyuk column can handle a range of feed
composillons wilhout changing the reboiler dUly, but merely b) changing lhe flows
10 column sections 1,2, 4, and 5. This flexiblhty both for the hmllmgcon<htions and
for operallng conditIons has also been reported by olbers.:
(A.5-19)
where. appro),:Imatel)'.
R
t': _ a..C<x A1 + \If) J.cr
- - + "'''--0;
xA,.(et"c - :r.d .x",{02.t<' I)
Similarly, Eq A.5-16 gIves the minimum vapor rate for a sharp ABIC spin, and so
and I'",.a;.. == + I), (x..
r
+ x.,-)(R::.ra
c
+ I)J}F (A,5-20)
Thus, Simple approximale expressions that are useful for conceplUal designs are
available.
FLEXIBILITY IN Till DESIGf'\. E\en though the vapor rate II: . eslablishes
lhe ffilOlmum reboiler duty of Ihe separallon syslem, II is sull possibi: to choose a
value for VI independenlly (providlOg Ihal lhe chOice for VI docs nOI change which
feed to the downslream column is controillng)_ To understand thIS flexibility
associaled Wllh tbe deSign, we conSider OJ case where Ihe lower Feed IS controlling,
K. GllnGli and M F MaloM:. IdEC PrtK IN. On U 10ll11198S) and ZJ 764(1914)
K Ghnoll and M F MaloM:. IdEC p,tK lNu, U 10111 (I93S)
z and L Krohlo"-$l,. AIC1>E J, J2 H7 (1986)
'W J Stupln and F J t.oclhan. -Thermally Co"pled Dtii',lIallon Columns A Ca"" 64lh
Ann"at AIChl: Mccnng. San haocliCO, 1971, and T L Wayburn and J D Seader, In I',oc !d '""
FOCAPD C-I, A W and II ChIC". eds., CACI E Corp c/o I'rof H Carnahan.
Dtpaflmcnl of C""fIUC1IJ lo"p'-U"L Un'U$'lly of "'hclupn, An.. Arbor. 48t09, P 7bS, J....
"81
Sf.t..IOI'l J COMPLE.Il I>IHILl.AlIO!'l COLUMI'IS 477
BEST Dt:SIGN CONDITIONS. The smallest of all the possible mmmlUm vapor
rates In the prefraclionalor is ... In column sections 4 and 5. the vapor
rates are gIven by '. = = 1') 1'1 For q "'" J. ,) "'" v
6
and so v) IS filed by
Eq. A 5-16 Thus. the column diameter In section 4 will decrease as II Increases. but
will ne\er reach I] (" hlCh fixes the dIameter of section J). We prefcr to bUIld
columns that ha\e lhe same dIameter In each secllon for reasons of so we
prefer to make as large as (although it Will be less than v]) This largest
value of 1. corresponds (0 the .. apor ratc r:...:" In the prefracllonator Glmos and
Malone' that this \-alue is gnen b)' the expression
(A.S-26)
PERFORMANCE OF PETL\'UK COLUM 's. Glmos and Malone' ....-ere able 10
show that the maxImum vapor were agaIn 50% independenl of the relatl\e
volatilities. and that they occur when (A "" (0:.111 - 1)/(0:.1(" - I) and XII -- U...... hlch
are lhe same resulls as for a sidestream stripper Ilowever. they also found Ihal
large S3\mgs could be obtamed In many cases when x. was large. especially when
the AlB Spill "as not llluch easier Ihan the HIC spilt
Prefractionator<;
I( we add a conden<;er and a reboller 10 the firsl column in a Petl)uk configuration.
i.e., we rcmo\e the thermal couplmg. "c call the result a prejrocllollofOr; see
Fi! A 5-6. We agam look for a sharp Ale sphllO the prerractionator. and we let H
dlstribule between lhe fOP and bottom of this column_ The minimum vapor rate for
lhe do,"" nstream column depends on whether the upper or lo,""er feed is controlllOg.
Since both feeds to the do\\-nslream column are essent13l1y binary mixtures and are
saturaled liqUids. it IS simple 10 dnelop expressions 10 esllmate lhelT values. lienee,
F
v,

'-"
J

V, L,
.-,
I I
,

-
I 4
I
L,
P
L,
,
2
v
1
J.,
I I V,
,

L
z
- V
1
I t
6
- V,
L.l
w
D
i' . _ :lACX:M + a:.d
a
.. - 1).x'F
6,_" - O:"C O:IIC (crAc TX"AC cr.,J
(A 5-27)
FlGURE AS-6
Pn:fn'ClOo...lot (F K GI...... -"'" M I" }.I.u-. Opt.-Jur kgltMJ Co-pIr.. CfJI_
AI'.....f...u iIf D ll,,'_ Su'"'": SIIlmlf/"J ,,, C'-" E,., 11.0. 1967 I
and (A,5-28)
lota1 vapor generated by both reboilers is then
The corresponding result for the prcfractlonalor is given by Eq. A 5-26. and the
fractional recovery of B o\"erhead in the prefractionator is given by Eq. A5-21 The
It. GlIno! and M , Malone. f&E<. /,,,,," fHI . 2.4 1087
K. GIJIlos Ind M F Malone, KOrrrmal"y R"g,o... fry Comrlu Col""", AII""""lIl'f" '" /)mllla,mll
ST,feMJ.
K
f'lJ'f'1 sutlmlllcO 10 C/'em f"!/ 11.". OeJ. 1987
(A.'29)
I>ERFORMANCE OF PRt:VRACTtONATORS. TIle maximum possible savings
with a prdractionalor arc never as large as with a Pctlyuk column. n the feed
contlols lhe maximum fractional savmgs arc (O:AC 1l.r)/(aAc I), wluch occurs
when ----<0 1. Whcn the lower feed controls. the nHUlIllUlll fractional arc
(allC - I )/(Il
AC
- I) and lIrc obtaincd when xc ...... O. lienee. the prefracll011ator
may be large for I:trge values of x.
SEcno....... t:NUGV l ..... OlSllU-ATION COLUMNS 479
for Compie),. Columns
Ghnos and Millonc summarllcd the results of their sludles b)' developing a set of
heuristll.:S heuristics :ire presented In $cc. 73
Ilowever, uSlllg our shortcut procedure for bubble point calculations, Eq A2-5, ..... e
can also wnte
A.6 ENEKGY INTEGKATION OF DISTILLATION COLUMNS
Heal Effects in Columns
SIIORTCI.JT ESTIMATES 0" TilE Tl\1P[RATLRE RA!'GE. Shortcut expres-
sions for 4 Tand Q d Thne: been dc\eloped b) Ghnos, Malone, aDd Douglas
l
and
Ihey arc presented here. O\er narrow temperalurc ranges we can use lhe: Clauslus-
Clapeyron equation 10 relale Ihe \apor pressurc 10 lemperature
(A.ft.7)
(Aft.5)
(A"')
(A.6S)
(A.6-3)
(A 6-4)
(
R ')-. T
u
= __In(1r+
liil (1D T
r
(
R , )-.
T.- _In(1-+_
liH 6
r
T
r
Q=M1V
I
T} T
1
6H 0
1
Using this result, we can relate the temperalure of the distillate and the
bonoms to the feed temperature
for Ideal, close-bOlhng mixlurcs, the heats of vaporization of allche components
are aboul lhe same, so that we can replace dH, by an average dH. Thus, If v.e
combine Eqs. A6-I, A 6-2, and A 2-5 and we let
we can obtain an expression relating the lemperatures at two pollllS in a column
Then, by combimng these expressions and assummg that aH RT
r
, we can v.nte
an uprcsslOn for tbe temperature drop across the column
RT: 6
u
11T ln_
6H o.
HEAT INPUT TO TIlE COLUMN. The heat supplied to the reboiler and re-
mo"cd by lhe condenser is merely the heat of vaporization dH multiplied by the
vapor rate V (note that we are assuming saturated-liquid feeds)
(A.6-I)
[
-Ml.(' ')]
If" Pj,oexp T- To
OISllllallon sep3rallons com-ume large amounts of energy IIcnce, there 15 an
economic 10 male the energycom.umptlon as small as poSSible. The baSIC
Idea behmd energy IIllegr.lllon (see Chap 8) IS 10 use the heat Ihat must be
removed from hot streams whIch need to be cooled to add heat to streams ..... hlch
need to be heated we would hke 10 achlc\'e as much energy llliegralion
as can be economically Justificd v.hen v.e design dlstillallOO columns.
For mosl dlsllllauon separations Ihe feed cntcrs as a satuuted lIqUid, and lhe IWO
product streams arc also salUraled liquids. ExpenSive separations correspond 10
relalhely close boiling mil.lures, and for these cases an enerH balance can be used
to show that appro:umalely all the heat supplied to the rebollcr must be remo\'cd III
the condenser Thus. m most dlslillallon separations, heal IS supplied to lhe
rebOller at a high tempenllurc. and thiS same amount of heat IS remo\cd from the
condenser at a lower temperature for this rcason \\,e sa) that heat IS degraded
across a temperalure range, which lS equal 10 the difference III bOiling points We
would Iil.:e: 10 ha\e a Simple procedure for esllmallng thc hcat load and Ihls
temperature range.
For low-pressure system:., the K value is gwen by
combining Eqs. A.6-7 and A.6-8 we find that
P:'
K=-'
, p-
(A.62) (A 6-9)

we that Q tr. T is approximately a constant for a given separation task, i.e., II
depends primarily on lhe compositions,
K GlIl106 and M F C""'pk.. Col""", Alle,""uues '" Il/s'illmion
SrI/to....." poll"'" 'ubmlHcd 10 Cltt'm bofi Rn 1987
, K GllnO, M r Malone. and J M Dottllla1. AICltE J,. 31 t019 (198S).
V"" (R + 1)0 _ (1.2R.. + 1)0 (A.6-IO)
480 nc:TION ... , EI'llOY II'lTEGunol'l or Dlmlunol'l COU''''N5
SECTION ...., (I'IUO" INTf.o....nOI'l or D1SnUATlO" COlV... m 481
1uhicffecl Distillalion
Qc,
FIGURE A.6-1
MuJ"dJ..;t cotumD_lFrom M J. JI""'...........,.II. rmd JI W IfrJlf'rb"'fI. AfC/i, J, JI J6J (lfI85).l
contmue to increase_ AndrecovK:h and Westerberg showed that there was a very
simple procedure for estimating a lower bound for the ulility consumplion. By
defintng d T. as the difference bet\loetn the temperalures of the highest hoi utlhty
-"'
avaIlable and lhe lowest cold utility, the minimum utihty required is simply
(A.6-11 )
/f----B
High pressure
Low pressure
I---'---A
2
AB----I
A LOWER BOUND ON LITlLln' CONSUMPTION. As we introduce additional
effects in our dlsllilation separation, we continue to decrease the energy requtre-
although the tcmperature le\el through which the energy is degraded will
Muhleffect distillallon has been discussed numerous {'laces in the literature.
although it is not widely ucsed We consider muhidJect columns because they both
pro"ide a simple cllampk of the energy integration of distillation columns and
Illustrate the effect of the pressure shifting of columns_ As we Increase the pressure
in a column..... c increase bolh the overhead and bottoms temperatures. so that in
many we C<ln make energy matches that otherWise .... ould not be possible. Our
of mulliefTect distillatIOn IS nOI complete. and an Interested reader
should the hterature. Our goal here is to illustrate some efTects encountered
In the energy Integration of columns.
Suppose we consider the distillation of a binary miJtlure. but we splil the feed
roughl} in half, raise the pressure of one of lhe streams., and st:nd each Slream to a
separ.l.te dlSlillation column: see Fig. A,6-1. If our pressure shirting was such lhat
the o\'erhead temperature in the high-pressure column is greater than the reboiler
temperature In the lo.... -pressure column, then we can combine the condenser of the
high-pressure column WIth the reOOlkr of the low-pressure column. Hence., we only
need to supply steam to one reboder In the high-pressure column, and we only Deed
to supply cooling water to the condenser in lhe low-pressure column: i.e.. the heat
that must be removed to condense the overhead of the high-pressure column can be
used to supply the heat needed in the rcboiler of the low-pressure column.
Since only one-half of Ihe fted is supplied to the high-pressure column, the
distillate Aow from this column will be one-half of the value for the case of a single
column Also. from Eq A.6-IO the vapor rate will be cut in baH, and from Eq. A 6-8
reboller heat duty will be cut in half Thus. \loe caD accomplish the same
separatIOn VI-;th a multieffect column configuration (soc Fig. A.6-I) as we can in a
conventlonal column, but we require only half as much steam and cooling water
Of course. when we use a muilleffect configuration, we must use two separate
columns_ Hence, muluefTcct s)"stems will be of interesl onl)' wheD the energy savings
are adequate to pay ror the hIgher imcslmenl. In addition, however, we must
supply the heat to the reboiler of the high-pressure column at a higher temperature
than we would need for a single column. The condenscr lemperature for lhe
mulliefTcct and single columns will be Ihe same. but we degrade the heal required
for the mull1effecl s)"stem over a larger temperature range.
To estImate Ihls temperature range. we remember that Eq. A.6-9 indicates
that Q d TIS essenlially a constant. Hence, if we cut Lhe heal load III half, the
temperature range will doubk. Our shortcul procedures enable us to estimate aU
the quantities invoh'ed for a p3r1lcular systcrn
where Q 6. T calculated ror a colullln by Eq A 6-9
C S Roht ......... 00 F R Gllhbnd. Flr--.u ttl F' ....'-J nmdl,........ 4th m. MC<;, ... HIll Nt..
"OIl_ IQ<,(' .nd C J 1',/1(1. ....,... ..,.".. I'rpr<'l"". 2d m. MeG.... 1!l11. N.... Y.... l lonIn
M J AOO'tt...och.nd" W 41(-II.F J. 'I '161
482 UCTlO"'" "N.. l ...n ... lION 01 0" r'lI.A 110/'0 lUI UMNS
Sland-AlollC Column
Andrecovich and WCl>tcrberg showed thai lhis same stackmg procedure useful
for evaluating a bound for lhe utIlity consumption for column sequences. ThaI is,
for a five-componenl mi"ture, there are 14 possible sequences. We find the
(Q 6T)'o',1 for sequcnce I by addlllg thc values of Q6Tfor each column in that
sequence, and we do this calculation fur each sequence. lienee, we can find which
sequence has the l>mallesl valuc of
Of course, we must allow for a 10 1-, or so, temperature drivmg force between
cnergtintegrated condensers and rebOllcrs, and we can change both Q ilnd l\ Tby
only lIltegral values as add effects, l>0 that our bound is not quite correct
However, il does provide an indication of the IlIcentive for undenakmg a more
detailed analysis.
T-Q I)IAGRAMS. In the dISCUSl>I.:>n of energy lIltegratlOn in Chap. 8, we usc.l
temperature-enthalpy dragrams as an aId III undastandlllg the behaVIor of the
process These diagrams arc '1lso us<:ful \\ hen we consider the energy integration of
dIstillation columns The dlects of Illullleffect and the lower bound on
the utililY consumption are illul>trated in Fig A.6-2. We see from the dIagram that
we minimize the utilities consumptIon if"e shirt the pressures so that we can stack
column sections on top of one another
(Q AT),. L (Q AT),
Ihen the minlillum uullly bound is estlmatcd by USlllg the expn:sSlOn
Q (Q AT),.
""n 111;,;1
(A.6-12)
(A 6-13)
T
CONO
T
CW
Q
FIGURE: A.6-2
Mmlmlzlna Ullhl1C5 by uSing a mullldfccl 11,_, If J A"d'u-Olld, .",.1 II
A{CItE 3t 163 (/985) 1
On a T-Q dlagram, .....e diVide the Q l!.Tfor each separallon task into widlhs
equal to QmiD' and then we Slack lhe various plea:S that we obtain between the
ulllity levels. An illustration of this procedure- is given in Fig. A6-1 Thus, there IS
an easy ..... ay of eSlimaung the utilily bounds.
Of course, as we add effects in a multieffect column, we are increasing the
capital cost, even though the energy costs are decreasing. Nevertheless, the pressure
shifting 10 allow the stacking of columns will provide a lower bound on the utility
requlCcments, even if we forbId the use of any multicffect columns For e"amplc, III
Fig A.6-4a consider three columns_ By pressure shifting and stacllllg the
columns, we can reduce the utility consumpllon. The lowest uIlIllY bound then
becomes equal to the "alue of Q for the column with the largest heat load (see
Fig A 6-4b)
We can reduce this load by replacing the column with the largest heat load QI......
by a multieffect column (see Fig. A.6-4c). Normally, we prefer to pressure-shifl as
little as possible when we stack the columns.
Integralion of Column Sequences wilh a Process
If heat is available from process streams that muSI be ejected to a cold utility, we
prefer to use this heat rather than a hot utility to supply the energy 10 a distillation
system, In this situation l!. T"il depends on the nature of the composile curves (sec
Chap. 8). We never want to have a column straddle the process pinch, and our goal
is to fit the columns in between the hot and cold composite curves (see Fig. A.6-5)
or below the grand composite curve (see Chap. 8).
M J ... "d,eo:o.och dud A W Wo<>,'erbcrg. ,HCltt. J. 31 36] (1985)
T
Q
(e)

I ilAI
(b)

CD
Gl
0
I
(a)
FICURI: "'.&4 II
Vlro'tnl (..) Three columns, (b) s.adtd te) ",,,Iudkel. [from M J A,w'rro"'c
<VId".. It K'ts,..,bv,fj'. MO' J, 31, J63 (198$)
7
FIGURE ... .6-]
MinImum uulllln. mulueffttl confi!ural'On for rollt :oq>arll'ons lr,om I-f J A"n.Ut1tKJr dnJ A It
14"$'"""'g. AfOrE J, JI J6J (/M-'n
110<
Slreamm,,__-
Cold
screams
2
2
T
Q
+--+ .1.--'-
IQ"",,1
I I
I I
! !
Limitations or
H .....e energY-integrate columns. then .... t' will change: the: optimum reflux ratio
because we do nol have to pay for UllhllcS. WecJlpcct that the oplnnum renux rallo
will be shifted 10 higher \3Iu(:'I. thatlhc operating reflux ralioand the vapor raIl,"
Will lIICrC3'\C: A8 the v3[)(.r ralc IS lIICIC3SCd. Eq A6-8 indicates that the column
heat load .... ,11 lIIueasc lIenet', some ,lerallon Will be rcqlllTcd to find
the best design
Q
FIGURE A.6-S
1 /I d..."am and column tud;,,'g. I' ....... M J An,h".......A and A If' II ... ,...'!I. AfCAE J. Jl
J6J (1M') J
'85
486 llEArU:C1I\N(;El DUI(;,.,.
A.7 HEATEXCIIANCER DESIGN
GUlhnc:'s <,;OIIC:I.H1ons (see AppenoJu: l 2) gl\e the cost of a he"l o:\changc:r In ternl)
of its area The heal-Iransfer area of a heal eJ,ch.. ngcr is cakulaled from an
equallon such as
TABU': "'1
hnl-lrlllSfer coefficients
K...-.bk ol_lb of onraU IoaHnooJ... c.ollicic8os
,...... .,1 .. reW.la_1 10 _ f..- preli-i-

heat-tran)fer coefticlent, and .. .... lug me.1R "'here Q = heal dUly, U = overall
lemperalUre dnnng forre
Heal-Transfer Coefficient
Q (A 711
Condc:llIolRl! III boiling hqllld
Condcmml III fk>"-ID& hq\lId
Condcmomg h) pi
Lkj;Uod 10 liquid
l.Iljuld 10
pi
Panla] cnndc05Cr
1lO
\lO
1O
lO
1O
10
lO
The overall heat-transfer coefficient III thIS expression is rdaled 10 lhlt mdl\ Idudl
film coeffiCients b}
where .... e neglecl the wall resistance: and we include the fouhng factors 10 lhe
IOdividual coefficients. Also, the IOdmdual codficients, excluding Ihe fouhnll
faclors, normall} are gIven In terms of empirical corrdallons, such as the DUlLls
Roeher equallon
I I I A.
- + -
L h, he All
10,0 _ (OG)"(C",),,(,,), ..
--0.023 - - -
k j.l k /-l ...
(A 7-2)
(A.7-3)
Simplified Modds for
(or Feed Vaporizers)
There ilre desIgn problems where .... e cool a vapor, condense II, and then subcool
Ihe liqUid. The normal heat-e:xchanger deSign equallon, Eq A.7-1, IS nOI vahd for
thl) case Inslead, first we calculate tbe: heat dUllCS for the process stream for lbe:
thr sccllons of Ihe exchanger (see Fig. A.7-1)
QI = Fe,dTI - Td Q1 0:- F lJ.H. QJ = FC1'l(Tc - Tl ) (A 7--4)
from lhltse: eSlimales. we can calculale Ihe tOlal heat dUly and the requIred flo....
rale of cool1lJg water
(A.7-5)
Thus. to obtam an accurate estimale of U. even for the simple case: of a
double-pipe heat exchanger, we mUSI evaluatlt Ihe (hennal conductivilies, lhe heat
capacities. and the viscosilies of the fluid mixtures on both the tube: and the shell
of the exchanger This proocdurlt ....ould reqUire a significant amount of eWon,
which might not be warranted for prdimlOary design calculations. In parhcuJar, If
lhe heal-e:xchangltr area required for a specific unil turns out (0 be fairly small, so
Ihal the cost of thaI unil contributes only slightly 10 Ihe total processing COSlS, "'e
prefer to obtain only a rough apprOXimatIon of tile size rather Ihan a ngorous
design.
To avoid tedious calculations m our first estimates of the lotal proccssmg
COSts, it is common practice 10 base the inwal design ofa heal exchanger directly on
an overall heal-transfer coefficient, such as the values listed in Table A.71 An
equivalenl approach is to use valults, such as those given in Table A.7-2, for lhe
individual resistances Then if lhe process appears to be: profitable, so lhal an
additional deSign dron can be Justified, more accurate values of lhe overall
coefficients can be calculaled In masl of our sludu:s we usc the values gIven to
Table A.7-1
TABLE A.1-1
Design shcll-and-Iube hUI nchangers
Boiliolc
e_
n...
_......
liot-
.....
FIled pses
...,
ughl hydrae.fbou gua oon 000<
ArQClUlllC loquMls 0.001 0011 0.001
UPI byd,OClI'bon tiqllld> 000< 000'
Chlallo_led bydrQC:;lfbons 000< 0009
Slum 0'"
000\
Bo"er ... 000'
0.00<
CoollOlilower Wiler 000'
\ ... R.....'anco: ,0000udoD ....... nl ..... -w wall aIIow__ 0."",_ U
.rc lor fll 'F B,u F,ORI II F R.... ud lot II BoI1Ow. -"'Of'
In_'''1 01 """"'"' Pw.... w,k,. New Vor\. 19n

T
c
. h!if to be the smallest driving
To develop a criterion of thIS type, we c. oose 1
force at one end of the exchanger. and we wflte
120r-----;; --.: 1;
" 90
"
61
1
""1+(
A"
The arithmetic mean driving force: is simply
(A.7-1I)
FIGURE A.7_t
Conckmcl rrofiJu
(A1-12)
Next we calculale the coohng-water temperatures corn:spooding 10 the discon-
tinuities in the condensation profile
Once these iOtermedaalc temperatures have been evaluated, we can usc the:
Dormal desIgn equallons 10 find 1M: areas in each of the three sections:
(A.7-18)
(A7-17)
(A7-14)
(A_7-13)
(A 7-16)
M, 16
-" .
M,
U ( is small we can write
In contrast, the log-mean eltpn:sslon IS
btl -Al
J
6 (
A, .. - In (AtdAl,) - /, In (I + ()
Multipass Exchangers
. ,., be n"cU"<: and/or multiple shell
Most industrial exchangers IOciude rou tIp e ,,_.
However for conceptual designs we !tmlt our focus to sImple
pa5'<", "hS"- The correction factors for multipass exchangers are gIven to
curren ex... .... '1
most lexts on heat transfer as well as (hal by Perry and ChI lon.
Furnaces
Guthrie's correlation gives the COSI of furnaces in terms of the heat by (he
process fluid. Thus, no design procedure IS neoessary for conceptual desIgns.
In (I + () _ ( _ i(' + +
. E A 7 13 and then uSlOg syntlletic division. or
Substituung this result into. q. .
another Taylor series eJ:panslon. gIVes
!ill. _ bl,{l + i( -1'2(J + ...) (A 7-15)
Now we see that if the: (l term IS 'er) small, the resull will identical to the
. . b Eq A 7 12 Henet: suppose we requm: that
arithmetiC mean !l"en y . - ,
n(l S olilt>
( :S 0,6
0'
.' . drivins forces to be
Then. from Eq. A.7-11 for the anlhmetlc and geometnc mean
about the same, we require that
(A.7S)
(A.7-9)
(A 7-10)
w..(120 - '.) Q,
(A.7-6)
.....(120 90)
Q. + Ql + Q]
W.(I
I
- 90)
Q,
(A 7.7)
:
w.(l20 - 90)
Q, + Ql +Ql
Ql'" UIA lln [(1; (1)/er. tl)]
1; - t
1
-(T
z
-90)
QJ - UJA
J
In ((1; I
z
)/(1; 90)]
Also, we can usc: the heat-transfer coefficients given in TaMe A.7.) or A. 7-2 for 1m:
three sections of the ('xchanger.
The probkm ofdesigning a feed vaporizer-where we first heal the stream 10
its boiling point. next vaporize Ute: stream, and then superheat the vapor-is
essentially the same, exccpt that II IS slmpkr because the temperature of the heating
medium, which usually is steam, is comaant throughout the exchanger. For this
casc we usc the fractional heat duties in the three portions of the exchanger to
estimate the quantity of steam condensed in the three sections.
Arilhmetic Mean Versus Log-Mean Driving Force
Of course, it is simpler to use the arithmetie mean rather than the Jog-mean
temperature as a driving force to calculate the area by using Eq. A.7-J. Thus, we
would like to develop an expression that will indicate when tbe arithmetic mean
provides a reasonable approxImation of the log-mean driving force.
490 )lCTION...,
DESIGN Of IlEH,(,t.UIIUN H1TU.!)
un'"'' OE\lGN 01 UIlIGlUlIO" 491
A.S GAS COMPRESSORS

I I
tllll.1ell;"cr Fla.)h
I
('001 lit
77"1'. I aim
(
At:clOllC -A Ir
(A 8-1)
GUlhoc's corrdauons 8'" ,h---, or
......... a gas compre!>SOf In lerms of Ihe br:a.le
horscpo",er bhp (see Appendu E).
. M&S
Inslalled (OSI = 180 (S 17 S)(bhp)o 11(211 + FoJ
Design
Assuming a o:ompressor enlclcnc)' of 0 8 gives
hp
bhp_ -
08
The horsepower is by Eq 6.5-1'
(A..B-l)
AcelOne
fiGURE A.....
Ag:tOllC coltdcnullon Id'lJr1llloa
s.letch of a t10wshecl b sholll n In Fig. A 9-1 If we guess that we v.ant 10 recoler
of the J,cetooc. so that .... t: can dlr:ll)' compare tile cost of our condt:nsallon
preas!> 10 the absorpllon syslem thai lIle discussed in Chap 3, then we reduce the
mole fraction of (he acetone from the mici \'alue of 015 to
hp ('OJ: 1O-}.Q{G:), IJ
and Ihe gas exil temperature is
(A 8-3)
)'z_(1_0995)(015)=7.5)( 10 s (A.91)
Operaling Cos.
The operalmg COSIS are based on Ihe bhp and a mOlor dfioency of 0 6
"'hlch IS qUlle small Tbere must be an OpillT1Um frJ,ctlonal reco\'ery, but lOre use thi)
\alue for our basc-casc dOI!!n We also assume that Ihe acetone le;,Hlng as Ihe
hqUld siream from Inc phase !>phnc'f IS pure, despile the fact thai a !>mall amouOl of
aIr b dlssohw In IhlS acc:tone
Muhisfage Compressors
For multlslage compressors, an equal compression rallo IS used for each Siage; see
Eq. 6.5-5.
A.9 DESIGN OF REFRIGERATION SYSTEMS
Even Ihough refrigeration cycles are dISCUSsed in numerous thermodynamic
normally il IS .nol a tnvlal mailer to usc Ihe basIC Ideas 10 dc\'elop a
Hence, Inslead of mcrely describing tile basIC ideas. we presenl a
fairly sImple desIgn case siudy Ihat iIIustrales Ihe economic lrade-orrs enCOuntered
In Chap 3 we discussed Ihe usc of a gas absorber 10 recover a solvenl from a gas
stream (I_e.. acelone from air). and we nOled Ihat a condensation process would be
a process allernative. lienee, we ehoose Ibe recovery of acc:looe from air as the
problem 10 conSider.
LSTl\IATING THE TEMP[RATl.Rl A"'D PR.SSURE OF TIlE I'IIASI:-
SPUTTER. The temperature spcf.1.fied III Ihe problem slatemeni IS thai of the
feed Ho..... e\el. "'e expectlhat the vapor and hquid Ieavlllg the phue spltller
will be in equilibnum, so thai from Ihermodynamics we expecl that Ihe partlal
pressure of acetone In the O<lsh vapor will be equal to its vapor pressure.
(A.9-2)
OESIG"'I. Iflll'c operate the condenser al atmospheric
pressure P
r
- I atm, from Eq A.9-2 .... e find Ihal
P:.,.. 7 5 x (A.9-J)
If "'e plot the dala for the vapor pressure of acetone given In Perry and Chllton's
handboo.l as p. \thUS liT, where T IS In we find that the temperature b
-121:1" F, which is qUllc a low temperature
Ini.ial Flo\tsheCI and SCI"ccning Calculations
We want 10 up 10 IOJ mol/hr of acetone from 687 moljhr of all. where the
feed stream IS 81 amblenl condulons If we usc a condensation process. our first
R II l'ury.Dd c: H Chll'on. &tjUJN'" S,h cd.. MeG...... lloll. !'k... Vorl
1913, I' )..49
neTlo!" A' DESIO!" or IF.llUOUJl.nol'l SVSTF....,$ 49J
Air Coolant
1-140
0
F I I

Air -128F
FEHE Condenser Auh
71F
I
67F Acetone
_128F
FIGURE A.9-]
Aoelonc: co...xnsatJon- Idnltt"tloa.
Planning a Rd"rigeralion S)"!Jlem
If we examine the graph of the pressure versus lemperature for varioll!i rdrigeranlS
shown in fig. 1230 of Perry and Chilton- (also see Fig. A9-4). clearly only a few
materials caD be used at -14O"'F, Letting Ihe lowest operating pressure in the
refrigeraot prooess be about S psI&, SO thaI air docs not leak inlo the system, we see
that only ethylene and Freon14 are accc:ptable rOf the low-temperature operation
We also require that the refrigeration system reject heat 10 cool;ng watc:r at
90 F, which corresponds to a refrigerant temperature or 10000F if "ie allow a JO"F
approach temperalure. either ethylene nor Freon-14 will reach Ihis temperature.
and therefore we must use a two-stage process; i.e., the ethykne or Freon-14 will
reject heat to another refrigerant, such as propane, ammonia, etc., which will then
rCJCCt the heat to the cooling water.
If we suppose that our acetone recovery system is a pari or a petrochemical
complex, then propane and ethylene may be readily availabk, and so we choose
these materials for our hase-ease calculations. To obtain a quk:x checlr: of whether
these fluids are accc:ptable, we use the Hadden and Grayson conelaiions given in
Appendix c.1. From these correlations, and assuming 1000F approach tempera-
tures., we obtain the results given in Table A.9-1, which appear reasonable.
Modified Flo""sheet
When we add Ihe eqUipment for the refrigeration system to the flowsheet shown in
Fig. A.9-3, we oblain the result shown in Fig. A.9S. Note that we have considered
(A.9-4)
mollhr
!-140' I
I
687
Acetone-Air -128F
Condenser
Flash
77F, I atm
I
so that P
r
- 4(M)()atm This IS a \'eQ high pressure; I.e., most high-pressure
processes operate in the range: honl tOO to 500 psia, although hydrocrad..e:rs
opera Ie at 1000 psia and processes operate at 4].000 psia
Also note that Eq A92 IS no longer \ahd at this pressure le\'e1 and lhat we must
mclude fugacity correction lerms abo\e 50 psla, or so.
Coolant
H1GU-PRESSUR[ DESIGN. An alternate approach for condensing the acetone
would be to increase the pressure of the feed stream until the partial pressure of the
acetone were equal to the vapor pressure At 77"F, the vapor pressure of
O.}] atm, and from Eq A9-2 we find that
CHOOSING A 8ASE.cASE CONOrrIO:,,\. There must be an optimum pressure
and temperature for the phase sr1ilter. I.e., as we increase the pressure (which
increases the cost ,of a feed compressor), we can operate the phase spliner at higher
lemperatures (whICh decreases the COSI of the refrigeration system, whkh indudes a
compressor), As a first design, suppose thai we operate at atmospheric pressure,
which us one bound on our deslpl (and places the maximum load on the
design of a refngeratlon system).
If we choose: a IrF approach temperature (Ihe rule-of-thumb value is 2 or
]OF for low but we should evaluale the optimum for a final design),
we can now put temperatures on our sec Fig. A.9-2. However, we would
nOI want to exhaust air to the atmosphc:re at this low temperature (it would cause
condensation and ice fonnallon). Hc:nce. ""e Install a feed-effiuent exchangc:r, which
also reduces the load on the refrigeration system. If ""'e pick an approach
temperature or JO"F, ...e obtam the re\ised flo\losheet sho""'o in Fig. k9-] or
course. \Ioe would also not send aorlOne to slorage at this low temperature, so that
actually we would mstall another feed-effluent However, since the
acetone product flow is small c.\lmpared 10 the olher flows, we neglect Itm our first
set of calculations.
10.3 lIlolthr
Acetone, -128F
F1GURf.: A.'_2
Alo;",,, o;ondc:llullon
R. H P".ry .nd C II ChIlton. CIw"'K"o/ Eng".urJ IIt1NHHJoJ<, 5th McGla..lhlJ. Ne.. York,
1'7J. Jl 12)()
SEnlQ,. H OUIGN Ot IInr.IG[UltON SYiiTEMS 495
K _ 1 it 10000F .. hell I' _ I"'" p5.... A. _ I .t -l2'"F _hen P- 2jO puI
K_I,uI7 .......hclll. 42-F K_latI7pPa""bc.r
F
"'
Air, 67F
-128
Feed-effluent - Refrigeration
128"F
Flash
clorlC( )
exchanger condenser

-150F,
17 psla AcelO
Compressor I
C
Vah-e
-32F,
250 psla
I Evaporator
-42F,
17 psia
_ Compressor I C
OValVI,':

'1
100F,
190 pSla
EVlIponllor
Coolmg w3lcr
120F 90F
TARLE A_9-1
Refrigel1ltion system conditions
the posslbllily of uSing aftercoolers after Ihe compressors. 5'lDCC compressIon causes
a nse These aftercoo!en; lise cooling waler supplied by the cooling
lo....ers
Base-Ca.w Design for the Refrigeration Process
In our of a process flowsheet, we estimated tbe malerial flows or the
proctss slreams, ahhough nol for Ihe rerrigeration fluids.. Moreover, we selected the
An-Ac

- "
.-
1

FIGURE .4.3--5
p"X"'" n"...heel
EnC'rgy Balances
Hence. the incremental heat load used to cool the feed stream and condense some
of the acetone is
Fn:O-EFFLUENT EXCHANGER_ From an inspection of the ftowshc:et
(Fig. A.9-5) we know the flo\\ rate and the temperature change of the air returning
through the beat exchanger. The heal duty of this exchanger is (the
heat capacity values were taken from Pert) and Chillon, pp. 3 134 and 3 126)
(A,9-1O)
o
120.\100
431.300
128)) + 243(58XIO,3 - 9.58)
o
.S,IO
291,600
All. _
5045(-21-1,1
o
75.SOO
1)),600
Aile -
14,100.
Q. = [600(0.506) + 687(0.238X29))[ - 80
= 242,200 + 10.200
.. 252,400 Blu/hr
21 OOts 0
40 00012 HS
80 00019 '.8
TABU: A.'2
Tncrefl1flt111 hnl IOld
Now. If we select temperalUres. use a vapor-pressure curve 10 find the vapor
pressure at these lemperatures. and apply Eq. A.9-2 10 find the mole fracllon of
acetone and Eq. A.9-8 to find x. we obtain the results in Table A.9-2. From these
calculations we sec that Ihe effluent of the process stream is about _gOF.
Moreover. we note from Tabk: A.9-2 thBllhe lalent heat effecl is a fairly small
fraclion of the 10lal heat duty of this exchanger (133.61925 _ 15%). so thai we base
lhe log.mean lit on the inlel and oullet slream lemperatures. The desIgn calcula-
liOns are gillen m Table A.9-3
Rt:FRIGERATION CONDENSER, Now that we halleeslimated the temperature of
the process stream entering the refrigeration condenser, we can calculate lhe heal
dut}' of lhe refrigeration system:
Refrigeration Cycles
Similarly. we see that the: lalent heal effcci is only a small fractioD of the tOlal
The Simplest procedure for designing the t .....o refrigeration c)des is 10 use pressure
emhalpy diagrams. since the expansion through the valve takes place at constant
enthalpy and the compression is approximalely isenlropic. We expect the estimates
we obtain from these diagrams to be somewhat different from our results using the
Hadden and Grayson correlations. so we repeat the analysis. The diagrams we
(A 9-5)
(A_9-6)
B,u
:= 925.000 ttt"
(
mo')( Btu )( 'b )
QrEE = 687 0238 lb. OF 29 mol (67 - (- 128)]
lemper-lIIures and pressurc:!i of most of the (.treams Based on tlus information. we
can no.... use energy balances 10 calculate other stream temperatures and flows.
and then .... e can calculale Ihe sizc:!i and costs of the equipment.
ObvIOusly. our results depend on lhe assumpllons we made for the Inlet
acetone composition. lhe fraclional recovery, the operatmg pressures. and the
approach lempcralUres. Thus. If our results im,licale thaI refrigeration is an
allracli\e reco\'ery process. we .... 111 need 10 determine Ihe sensilivit}' of Ihe dcsign
to lhese: assumpllons. Ho e\er..... e must consider the absorptIon and
adsorpllon alternau\c<l. and e develop a final drslgn only for the least CJIlpcns.n'e
process
which is a fairly small heal load We expect that acelone might Slartto condense on
the: feed. rather than the return Side of this exchanger. and so we run the feed slream
through the shell Side The de\\ pomt of the: feed mixture can be estImated
from Eq. A,9-2 and a vapor-pressure cun'e for acetone; po - Pr}' "" 0015 and
T __ 21F. Then the heal reqUired to cool the fud 10 the dew pomt (DP) is
Qor = [(600 0.506 + 687(0.2J8X29)(77 - (-21))
= 494.000 Btu
h,
Q..... = 925,000 - 494,000 = 431,000 Btufhr (A.9-7)
then we can write
l: "'" 10.3 687)'.+
If we kt x = moles of acetone condensed, .... here
Q._ = [600(0.506) + 687(02.111)(29)]( 21
- 5045( 21 r
l
)
(A.9-8)
t,)" (243 BIU,moIX58)(IOJ)
(A 9-9)
TABLE .U-J
Exch.nger
Lnil Q, 10' 1l1a(hr
"'. "'.
A,_ V A, ft'
092S 10
.,
'"
20
'909
Rrin,..t:l1l<lft condense' 0252
'" "
<t. 20 lO"
r ..ror..... ,n k-<>p 0.,.
'"
10 '20 20
".
hl,-IIOO In J"0l"ne kx>J1 0611
"
'0
12] 20
,...

,
I
.,
J
i

j
I
,
j
,
1
,
,
.. ,
,
,
,
""!!"

. ," 1
, I
,
,
, , ,
'I
,
..
,
,
,

,

'"
,

.
, ,
,
1

,
,
,
I
,
$
,
, ,
, .
I
"
,

-_.... -
I ,
:'1-
, ,
,
,

-
, ,
'''..,
,
,

..
'.
.
,
, ,
'-
'i.!
. ' ,
,
...... "',,-'
,
,

F
-
.
..,. -,-
,1
1
1l! I I 1 11111 1 ! I! ii'"'' ,,
"
I,'""1++
'
l+l+fjI ItI.. .1
I rT' " .
'IJI I. Illlf -'
, !." :;
'" y, ' 1
, ,'.I.-
" <
,
, III > .:;'. . .'
\ II
(A.9-13)
(A.9-1I)
252,400
86 _ ( 44) = 19411blbr Ethylene Flow - w[ - Q..
H
l
- 1/)
The evaporalor duty in the e:thykoe loop is then
Qn = 1941(H. - HI) - 1941[166 - (-44)] = 401,600 Blufhr (A.9-I2)
Similarly, the circulation rate of propane is
Propane Flow = "', - 1/) = 1:
J7
2) - 3996 lbfhr
and the heal duty of tbe final evaporator is
QCl' = 3996(fI. - Ill) _ 3996(151 - (-2 - 634,000 Blu/hr (A.9-14)
Mater-ial and Energy Balallces for the
Refrigeration System
w C. EdmlSlel and B I Loc. lIydlOc,lfbon vol 1," 2d ed Gulf
PubllshJn& Houslon. Tua.s. 1984
I W C. Reynolds. TMo-rtWdynamoc P.optrtln UI Sf. Dcpilfllnelll of Mcch.onlCill EllplICCnnll. Siallford
Un,ven,lIy. Slanford, uhf., t979
used for the calculations were published by Edmister," but a more fcceol set
published by Reynolds' is available
To estimate the: temperatures, pressures, and cOlhalpies of the: various
slreams in the: rc:rigc:ralioD loop (see Fla. A9-8), we start our analysis at point I
(see fig. A.9+8) 00 the propane diagram (fig. A.9-6). where the: fluid is a salUraled
liquid at 10000F, and we read PI and HI ()eC Table: A.9-4). Nc:xt ....cCllp3nd the fluid
through the: valve al constant enthalpy 10 a pressure: of 11 psla 10 POlOt 2. and we
read temperature T
1
from the diagram Then we evaporate the Itquld al constaDl
temperature aDd pressure until poinl 3, Iloherc it is a saturated vapor, and we find
H) and S). The compression tales place al consclanl entropy unul .... e reach the:
original pressure: at pomt 4, and we find T
4
and H4 If T. 1$ much larger than
we install an aftc:rcoola. By allowlOg a IcmperalUre rise of Ihe cooling water of
IO"F and a IO"F approach temperature=. tbe temperature at point 5 will be IIO"F,
and we can find 11 s' Then the vapor is condensed agam. The ethylene refrigeration
loop is calculated in an identical fashion by starting with a temperature IO"F higher
than T).
Using this approach, we obtain the values given in Table A.9-4. We neglect
the: aftercoolers in both loops because temperatures are not excessive. The
diagrams are difficult to re:ad, 50 we: might expeclto obtain more accurale
by using a computer-aided design program.
ill::FRIGERATION CONONSER. The heat dUly of Ihe: refngcrallon condenser IS
given by Eq. A.9-IO, and the enthalpy charge of ethylene is given 10 Table A.9-4.
Hence; we can calculate the Circulation rale of ethylene:
499
,
'"
'"
'"
'"
'01
"'
'66
I
p
"
..,
'00
'31
"
-'SO
86
1615
PoitM .3
-128
"
..,
-2
" -ISO
_u
P.iM 1

2SO
JO
_u
>9,
"Xl
2
P.....
Evaporator
Refngcrallon
condenser
TABU: A.""
Rdriceralion cottdiliom
Edo, ....
SECllON ", DESIGN Of" lEnlGEUno", SnTENS .501
G) _'-_---l---;'000F
60if
90 Cooling
EvaJl'f'I3lnf
105r WlIfl'T
flGl ME ,'-9-8
sys.IaIl
P, "'".
TT
II. Btu1b
,
-80
P
T
H
,

-
:i
,
..
!
!

,
I
I
I
,
,
,
"
e
j

'.
Irtttll , ,
!'"" ,
! ! !! ! ! !
502 A' DUiKJ.. 01' .. EFlKiEUIJON SHILUS
Equipmenl Design
If the cost of cooling water IS SO 075/1000 gal and Ihe plant operates 8150 hr/yr, Ihe
annual cost is
1IE.4.T EXOtANCERS. ow that the heat dUlles of all the heat cll:changc:rs ha\'c:
been eslimated and Ihe tempen.lUres of the slleams ha\e been fi.ed. It IS a
stralghtfor..... ard lasL. to calculate the size and cost of the exchangers The rcsults arc
given 10 Table A 9-5.
The coohng-waler flow rate IS detcrmmed frolll an energy balance
""here we ha\c assumed an mstallatlon factor of 40 Also, the po.....er COSI, assummg
a blower effICiency of 70}' ... a mOlor efficiency of 90%, and a power cost of
S0042fkwhr, is
(A_923)
(A.9-2s)
COSI = 5S,000(4XUS) = S683,600
(
50042X Hw , {h')
Cost = _.- 466.3 hp 8160-
L. whr 1.342 hp yr
= $1J9,200iyr
Aoo
BLOWER. We also might need a blo.....er for the feed slream. TIle air flo.....' rate is
687 mol/hr = 687(359 fI
1
/moIX537/-492)j60 = 4500 fll/mlO. and Peters and Tim-
merhaus the purchased cost of turboblowers 10 terms of this voluD1etnc flow
rale If we guess thai Ihe pressure: drop IS 5 psi on both the lube side: and the sbcll
side of heat eJ.changers, then the pressure drop of Ihe air slream is 15 50 that we
neW. a 3O-PSI blower The I;OSI IS
(
'b)( m.'X mOIX r,')( I )(' hp.,) I
H,. - 15 in.l 144 l"jT 687""hr"" 392 mol 3600 s 550 f1.lb 0.7(0.9)
= 466.3 (A 9-24)
(A9-ls)
(A.9-16)
(A 9.17)
611,400
W = -- = 20400 lbfh'
30
Qu = 611,400 .. IOC,.(120 - 90)
Cost = X I. gal )(20,400 IbXSI50 hr)
1000 gal 8341b hr yr
= SI440 )r
Aoo
so that
COMPRESSORS. The horsepo'AU for IS gl\ien by Ihe
expression
The capital COSt of blo....-er IS \ic:ry high, and we nOle from the COSI
thai we could cui IhlS COSI by a faClor of 10 if we could use a ]-ps.!
lurboblo'Aer, I.e.,
(
"'")( 'b)( Ih, )( r, 'h)( I h
P
) I
hp... H.. - 11 J Ib wI: h---;: ]600 s 778 Btu 550 ft Ib.'s <E...
(A.918)
COst - slOO(4)(US)" S6],400 (A.9-26)
where E. "" compressor efficiency - 0.8 and E", = mOlOr effICIency - 0.9. Thus,
Hence, we must estimate the pressure drop of these exchangers fairly carefully lind
lool.. for ways of mlfltmlZlng Ihe pressure drop If we guess that a J'pSI turboblowu
'AlII be adequate, then the: power costs will be
./1941)(778) I , 7
hp = (166 - 86\3600 550 (O.9XOS) '" 4
(A 9-19)
Ann. Cost ... 119,2OO(t's> = S23.8OO/yr (A.9-27)
Similarly. the results for the propane compressor are
At a cost ofSO.042/kwhr,lhe annual power COSI for the eth)leoc loop compressor IS
{
3996X77') I
hp-(Isi-Ioo - - _1112
3600 5'" (0.9XO ')
Aon
(
50042)( Ilw ) {h') Cost = -- (847 hp 8t60-
kwhr 1341 hp yr
= S21, 7oo/yr
(A9-20)
(A9-21)
Otber Equipmenl COSIS
The capnal cost of the heat exchangers is csumatro from GUlhric:'s correlation,
AppendiX E.2, based on lbe values given in Table A.93. We assume that lhe:
IflSlallalion factor IS 3.29. that we use floating-head exchangers, that the: pressure
correclion faClor for Ihe evaporalors is 1.1, thai we use stainless-steel (or low-Nt
steel) exchangers at temperatures below -2O"F, Ihat the M&S ratio is U8, and
that the capital charge faclor IS !. Stnularly, we consider mOlor-dn\ien, reclprocat
mg compressors with an installation factor of 3.11
while that for the propane loop IS
00(
1tt.2)
Ann ('OSI = 21,7 847 = S28,5OO yr (A.9-22)
M S and K 0 Tlmnx,lLau... MPI"n' fkl;,&n and EcunOfnIQl fur Chcmx:al En..,na:l'l.w 3d cd,
Mc<ir:a.. thll. Nc:"" Y.,.l.t980. p .no
SEctlON ".f DESION Of" UfllIOU... nON SYSTEMS 505
The results of the COSI calculations are given in Table 1\.9-5. 10 addition,
Ihe utility cosls are listed We see that roughly an investment of $1,304.000
(S434.800jyr) aod an COSI of S75,500jyr are needed to recover acclone
valued at S881.300jyr. H we oplimize Ihe: process. wc might be able: to impro\'e lhe
profitability IiDWC\cr. we must also complete the dcsigns for a gas absorber and an
adsol'plion process before we can decide which prooess might be best.
ShO... Design Procedurrs for RefrigeratKJn
Loops
From the casc: siudy presented abo\c. we see thai the design of refrigeration
processes is quitc tedious. FOrlunately. Shellon and Grossmann- have preseotcd a
shorlcut procedure that signtficaolly simplifies the caJcu]ations. From Fig. A.9-9
we see that the heat load for the evaporator (points I 10 2) can be written as
(A.9-2')
where AHi is the molar latcnt heat ohaporization andAJ/r, is Ihe molar lost heal
of vaporization. Since
and Eq A.9+28 becomes
(A.9-29)
(A.9-JO)
if we assumc that the enthalpy of a saturated liquid is independent of pressure and
that the heal capacity of Ihe hquid is constant over the temperalUre range, then we
can write that
Instead of attemptlOg 10 integrate Ihis expression by using an equation of state.
Shelton and Grossmann- noted that the curves for constant entropy on a Mollier
diagram, where log pressure was plotted against enthalpy. are almost linear.
Therefore. writing Eq. A.9-32 in terms of In P and using tbe ideal gas law give
(A.9-32)
(A.9-lJ)
(A.9-31)
(
Olf) V
oP
(
.!!.!-) = PV = RT
oinT',
For the isotroptC compression. points 2 to 3. we can write that
M R Shelton.nd I F (j,os.\m.nn. and CAmt Fftg." 61S (t98S)
506 SEC1'lO,," At IlUIO,," Ot- Ut"JJGUAno,," HSTEMS
nOlO,," II SIJMMAI. A"O NOMu.n.AlUI.( 101 A 507
The t;ondenl>er dUI}, 1X>lllb 3 to 4, can be Wflllen liS (see FIg. A.9-9)
Q , "" Q.,.P 1 W
Il
(Mli +AH71)FIfF"
our r':lIou[ts, we oblain
(A 9-37)
[
,T,
Q. Mi,
(A 9-38)
Summar)
Thus, we ha,e obtamed sImple for Ihe quantlllcs of intercst
A mellC modd muSt used to calculale lhe required reactor "olume For lubular
reactors, Ihe lenglh-Io-dlameter rallo musl e).ttCd 6, but preferably it should be
grealer Ihan 10 The COSt or a tubular reaClor IS eSllmalw b) using Guthrn's cost
corrclallon for \cssels
For conlinuous stirred lanl. reacton;, v.e usually set the L-D rallo 10 aboul
10 1.5, COSI correlations ror CSTRs are avaIlable III Peters and Tlmmerhaus
P
2
1---------"":
H
Evaporator.
Po.... er requirement

A.lO REACTORS
Q.,., =- [AHi - T1)]FAF
Alii (T
2
- T.)F.F
n'll .c:
1;
Q._ =' [AHr - - T1)]F.F
FlGURl. 0-9
RanacralKHI plOCftl.. (Ttl"" AI R ShrII"" <INI J . G,tlS_ CompIIlIUf ""d Chnn Cng _, 6H
(l9&S)J
EvaluatIng Tal T
2
and Ifllegraung then give
P,
tJ.Hi2 = RT
2
1n - (A.9-34)
P,
We can rewrite Ihls expression in terms of temperatures by introducing the
Clausius-Oapeyron equation
In P2 = t1Ur (A.9-35)
PI R T
1
T
1
Then the power requirement for Ihe compressor becomes
A_I I SUMMAR\ OF SI-IORTCUT
[QUIPME TDESIGN GUIDELINES AND
FOR APPENDIX A
We have presented a number of design guidelines during our It 1$
con,eDlent 10 comblDe lhese mto a single list which we can rerer to when we attack
other problems. The hst presented below is not complete, so il always pals 10 asl.
eApenencc:d englnccrs if they know of any other guidelines thaI may apply to a
partIcular problem
Perhaps we should emphasize thaI It IS dangerous 10 blindly usc any
published rule-or-thumb or design value, since anyone or these generalizations is
hmited 10 cerlain types of applications Therefore, it is always worth understandlDg
Ihe origin of the rule 11\ order to gain some feeling for thc potential limitations
," 'II. F _ Alii (T2 - T,)FIlF
Yn- U 1211F"-
T,
(A 9.]6)
M S anJ K 0 1 l'lunl Jhj'iJ" und Eronom.{'f for t"ntJtnU,f, ).J eli
\lcOn.. 11111. Ne... Yorl, 1980, pp.19O-19t
" II
su......" ..v M'''U NOMEN('I....lUl. fOR "F'f'fNDlX"
509
Moreo\'er. ror final desi{ln .... e ahuys prefer to use ngorous calculation
procedures httause .... e must be as certalll as ....e can that the final desltln .... 111
correspond to an operable process. U'. the cost or modlrymg an eXisting planl to
IT.ake It meet de<>lgn specificatIons enormous compared to the cost or making
changes at the design stage. 1I0.....e\er. lime pre... and budget constraints
somellmes roree us 10 lake ris"s. so good judgment is always requITed
I PhySIcal PropPrt/(!s
Procedures for estimalll1[t. physleal properties of materials arc: descnbed by'
R. C Reid. J M Prausllltl. and T K Sherwood. Th,. ofGast's and
Uquith. 3d ed. McGra.... 11111. Nc.... York. 1977. A condensed \crSlon of
of lhese methods IS g"en in R " Perry and C. H Chilton. Chetl'l/cal
Engmeers lIandbook. 5th cd.. McGrawHiII. Ne....' York. 1973. p. 3 226. The
API Data Book and the Int,.rnallOl'/ol Critical Tables, as ....-ell as many other
handbooks, also contain much valuable inronnation.
II IS always belief to base designs on actual experimental data rather
Ihlln esllmates obtained from correlations. However, we often use the
correia lions for prelImInary designs when data are not In-allahle
A simple procedure for estimating K values for vapor-hquld equilIbria IS
presenled III Se\: C I or Appendu. C
II DUldlalllNJ Column DeSIgn
Disllllalion columns are one of the most common Ulllt operations, and
normally the distillation costs are a large fraction of the lotal processing cost(
A Phast Eqllll,brlllm
I General
a Careful allentlon must be gl\en to hIghly nonideal mixtures In
parllcular. It is essential to determine ..... hether azeotropes rna) be
formed An interesting discussion or nooideal effects has been
presented by G. J Picrolll, C. H. lXa!. and E. L IXrr. Ind Eng.
Chem. 51: 95 (1959).
b If the relative volalihty is 1.1 or less. extraction or some other
oreralion may be cheaper than distillation
c:. The presence of heat-sensitive compounds or polymerizable mate-
rials may require vacuum openuion.
d_ Corrosive materials require more expensive materials of construc-
tion.
2 Shortcut design procedures
0. Most shortcut design procedures. e.g, that of Fenske, Underwood.
and Gilliland. require the assumption of constant relati\e volatility

(I) Constanla is obtained for similar materials

(II) The hC<lts of vaporil.ation of the lIght and hcavy are
the samc.
(h) The boiling points of the right and heavy are doo;e
together
(c) Use Ideal liquid mu:tures
(d) The separations are relatl\'e1y diffICUlt and are
(2) For SimIlar materials (Kieal solUtions) ..... here the bolhng POlllts
are wKlely separated. heat effects are Important.
(a) Base the prcliminarydesign on the Kremser equation for the
top and bottom sections.
(II) Usc fictitious molecular .....eights and a McCabe-Thiele
diagram
(e) The separations normally are simple, and Ihe are
less eXpenSI\-e, so that great aceuracy is not required for
preliminary designs.
(3) For some nonideal liquid miJ.turcs (WIdely separated bollmg
points).
(0) Base preliminary designs on the Kremser equation for the
rectifying and stripping sections.
(b) The separations often are easy and the COIUffillS relatively
irn:xpensive, so that great accuracy is not required
B CO/11m" Mmerial Balances
I Normally only one compositIOn. which may be either lhe top or the
bottom, is fiJ;ed by product SpecificatIOns.
2. For first estimates assume bet.....ccn 990 and 99.9-;' rccO\leries of
"aluable components for normal separations.
"_ These estimates normally must be modiftod if az.c:otropcs arc
present.
b, Lower estimates often are used for high-pressure columns and
columns wilh parlial condensers or refrigerated condensers,
3. Use Fenske's equation at total refluJ; to estimate the splits of the
DOnkey components.
C. Column IHrigrl
I. Estimate the column pressure.
o. Condense the overhead at 100 to IJO"F.
(I) Use of a total is prderred_
(2) Make certain that there is adequate renux if a partial (;ondenstr
is used.
(3) Compare bolloms temperature to available steam temperalures.
(a) We desire a 30 to 45F lit. or more.
(b) Readjust pressure if necessary.
2 Estimate minimum reflux rate.
o Use one of Underwood's equations if a is constant
b. Use McCabe-Thiele diagram for bioary miJ;tures.
3. Let R _ 1.2R. for first estimate.
4 Estimate number of theoretical plates-use sieve trays (or val\le trays)
o N _ 2N.., or GIll1land's correlation.
b. Use Krcmser equation or McCabe-Thiele diagram.
510 SEClltoN ... II
SECTION" U
5 bumale the actual number o( plates
a Assume an overall plate eftlcleney of
(I) It should be hIgher (or c1osebOllm&. Ideal mutures.
(2) It should be lo.... er (or mixtures "'Ilh large difference>. 10 their
boiling points.
b Use O'Conneil's oorrelallon (allhough It b nOI \cry reh.. ble)
(I) Usc: a if a is approximately oon>.lant
(2) Usc different CSllmates for the top ,lOd bollom I( the
Kremser equallon IS used for dOlgn
6 Esllmate the column height.
a Use a 2.0(t !ray SpaClllg (a 2-fl 6-10. trlly spacmg for manhead
locallons).
b. Add g to 15 fI at top for vapor disengagement.
c. Add eXira space al bottom as a liquid surge tank.
(I) Normally use 3- to 5min holding llme based on bottoms rtow
(2) Use 6-mlO holdmg lime I( bollom product IS (ed to a furnace
(3) Use 8- to 15-ft space for first de:l;lgns.
d If column height 15 grealer than 175 rt or If LtD IS greater than about
20 10 30, special foundations are requITed
(I) Redesign wilh 18-111. (or 11-10.) lray spaCIng.
(2) Build lhe rectifYlDg aDd stripping sections separalely, if neces.
sary.
7. Esumale the column dlameler for both the recllfymg and slnpplng
Stttions
(l Estimate lhe vapor rates In the top and bottom In ft
3
/s.
b. ESllmate the vapor ve!oclIY
(I) Guess lhe vclocuy.
(0) For atmospheTlc operation assume the superficial VcJOCllY IS
3 (t/s and the pressure drop is ] in 1110 per tray.
(b) For pressures below 100 mmllg use 6 to 8 ftfs. Be careful
thai the pressure drop is not so high thai the bonoms
temperature becomes excessive.
(e) For high-pressure operation, tbe velocity should be reduced
from 3 to I fils as lhe pressure II1creascs..
(2) Use an F (actor of F - 1.5 for prclimlOary designs.
(3) Use Fair's floodlOg correlation., and setlhe velocity at 60 to 85-;.
of the flooding value.
c. About 12r. of the area is required for downcomcrs.
d. Tower diameters should be specified m 6-10 IOCTements.
Use the largcst diameter (or the rectifYing or stripping sections tf
they are dose 10 lhe same or build a 5wedge tower if they are very
different.
f Limllations.
(I) If the tower diameler is large. Ieo. greater than 12 to 15 ft.
consider increasmg the Iray spacing 10 36 10
III.
(1) If the to..... er diameter .... Iess lhan 18 m., redesIgn the to.... .:r as a
pad.cd column
(0) Usc 1-10 pall nng.-. ofa suitable metal.
(b) For 3tmosphenc operatlon assume a superficial vapor
of 3 (t s, a pressure drop of 0.5IDJi
J
Ojft. and
HOt> = 12 for prehmlOary designs.
(c) For thdn 100 mmHg. assume a \'c1ocity of6 to
t; rt s. a pressure drop of 1.0 m.H
1
0fft, and 11ex; = J fur
prehmmary desIgns
(3) !'or towers In the range from 1.5 to 4,5 fl, use either a plate or a
packed tower.
(4) Other applications for packed lowers.
(0) Corrosive materials
(b) Materials which foam badly.
(c) If a low-pressure drop required.
D Sequences of Distil/OliO" Co/umllS
The heunstics may be contradictory.
I. Remove corrosive and hazardous materials early In the sequence
1. Remo\e tbe most volatile component firsL
3. If a component IS predominant 10 the feed, remove It first
4. Make 50(50 SplitS whenever possible.
5. Male the easiest separation first.
6 Male the cheapest separation first.
7. Remove the components one by one as column o,erheads.
8 Sa\e difficult scparallons unlillast
9 Prefer to operate as dose 10 ambient conditions as possible, but prefer
hIgher pressures to vaccum operation and higher tempera lures to
rerngeration condItIOns.
10 Prefer not 10 mtroduce a component absent from the original mixlure
(absorpuon or extraction); but If we Illtroduce a foreign species..... e
desue to recover it as soon as possible.
Heat ExcJwngers
Heat exchangers also are very common pieces of process equipment, and the
total costs normally are a large fracltOD of the total processmg
costs. Some heurntics are given below for approach temperatures. but. beller
results can be oblamed by usmg the energy mtegration procedure descnbed. In
Cbap8
A Heat Transfer CoeffiCients
Use the overall heal-Iransfer coefficients gIven In Table A.7-1 or A,2 or
similar values.
B. Dutgn Condl/ions
I W;Her coolers.
/.! Assume coohng water IS available al 90' F (on a hot summer day)
and must be re1Urned to the cooling towers at less llian to
pre\enl scaling o( exchanger surfaces SolutIons of opt1mWlllOn
problems that eslJmate the effluerll coolmg-water temperature nor-
mally ,ue \alues greater than 120 F. because scaling problems are
not considered
b. Assume an approach temperature betll.ecn the stream being cooled
and the mlet cooling water of 10 F (for very clean materials) or

2 Condensers
(l Dlstlllalion columns
(I) Usc a lotal condenser If posSible
(2) Pure comroncnts condense at comtant lemperatures in a Iota]
condenser
(3) For ml,'ll;tures calculate the dell. pomts and bubble pomts, and
use these to estlmale lhe log-mean l\. T for a tOlal condenser
(4) For parllal condensers make cenam thaI there is an adequate
amount of renux
b. Cooler condensers in processes.
(1) Use the same approach temperature as for water coolers
(2) Cool the \'apor to ils dell. point. condense and cool to liS bubble
point. and then sulx:ool the liquld_
(0) Base the log-mean d T on the IOlet and outlet temperatures if
the lalent heat effects are small compared to the senSible
heal changes.
(b) Base the log-mean l\. Ton the dew point and bubble pomt if
the latent heat elTecls arc lar,e compared to the heat
chanpes.
J. Reboilers-to prC\'enl film boihng assume l\. T "'" and U = 250
Btu/(hr fl
l

0
F). so thai UdT= 11,250 BIU[1
1
(up to 13,000 IS
satisfactor}')
4. Wasle-hear bollen-use an approach lemperature of 6(rF for first
estimates.
5. Air-cooled exchangers and condensers.
o. Assume U,.. 70 Btu/(hrrtJ.oF) based on nOGfinned surface.
b. Use an approach temperature of WF based on the inlet high-
temperature fluid.
c. Assume the air temperature is 9O"F (on a bot summer day) toensure
that the area is adequate.
d Assume 2Q-hp power IOput per 1000 ft
l
of base surface area
6. Countercurrent elchangers.
o Assume an approach temperature of 30 F betv.-etn the en!enng hot
stream and thc exiting cold stream for multipass elchangers,
(I) The opllmum value will depend on the thennal capacitances of
the hot and cold and the number or tube-and-shell
passes
(2) The mmlmum value IS about 10 I
ROlON "-II SU......... IlY "'ND NOMf"lCL4TUIlE rOIl "'I'PENOIll... S13
b, Multipass exchangers,
(I) Use one of the published charts to corrcel the log-mean 6T
(2) Uthe correction factor is less Ihan 08., redesign the exchanger to
obtain a h1lher correction factor.
c. Uthe area is less than 200 ft
l
, consider the use of finned tubes
7. Feed-cffluent exchangers.
o. Rcco\er about 75 % of thc available heal for liquid-liquid u-
changers
b. Recover about of the available heal for gas-gas or gas-liqUid
exchangers.
c, The heuristics abO\e often gIve poor resull5. Usc tbe energy Inlegra-
tion procedure presented in Chap 8
C Stondord /Rsign.f
I. Tube diameter
a. In a given shell. Ihe smaller the tubes. the grealer Ihe surfaa: area.
but the greater (he difficulty keeping the tubes clean.
b Usc join. tubes for clean materials, I-in for mOSl nuids, and 1.5-in
for \'ery dirty nuids.
2 Tube lengths
0_ 10st chemical companies usc 8-ft-long exchangers
b. Most petroleum companies usc 16-ft-long although III
some cases 2O-ft-long exchan,ers arc uCd.
c. For 16-ft-long elchangers. the m8llimum surface area space shell IS
(I) ]500 ftl ror join. tubes.
(2) )200 ftl for I-ln. tubes.
(3) 2000 ftl for IS-in. tubes,
3 PIlch
o Tnangular pilch ghes smallest shell SIZe for a g.t\en area
b S<juare pllch IS much easter 10 clean.
4 Pressure drop-assume a 5-psi pressure drop for bolh the tube Side and
lhe shell side of the exchanger for first estimates.
IV Furnaus
A. Assume that the nUld \'eloclly in the tubes for a nonvaporizing liquid IS
5 fils.
B Assume an efficiency based on Ihe net heating value,
V
h 30J '0
' (JOJ x '0 ',Xl A Design equation; p"'.)( or
p.. Qi.[(P_IP;.)1' - I]lK and T_fT... :III: (P_IP.t. where hp =
horsepower, P,. = Ibfn
1
Q is In drn. and K "" (C.IC" - IXC.IC.).
I for monatomIc gases usc K "'" 0,4.
2. For diatomic gases usc K _ 0.29.
3 For morc complex gases (COl' CH.) usc K = 0.23.
4 For other usc K _ where R the gas
An excellent recent colJcclion of shorlcut equipment-design procedures
has been published by F. Aerstin and G. Street. Applied Chemical Process
Design, Plenum, New Yorl, 1978. Also the Manual of Economic AnalyJis of
Che""col ProcesseJ, published by the Instilul du Pel role, McGraw-
Hill. 1981, contaLnS much uscfullOfonnatlon.
B. Efficiency.
I. Assume a compressor efficiency of 90 %10 accouni for ftuid friclion In
suction and discharge valvcs and ports, friction of moving metal
surfaces, ftuid lurbulence, elc.
2. Assume a driver efficiency of9O% to account for the conversion of the
Input energy inlo shaft work.
C. Spares.
I. Comprcssors are very cxpensi\c, so spares arc seldom provided
2. A common practice is 10 provide IWO compressors with each handling
60% of Ihe load, so that partial operation of Ihe planl can be
maintained if ODC compressor fails.
VI. Gas Absorbers
A. Plale columns-similar to dislillation.
J. Assume 99% or greater recoveries.
2. Use I../(mG) = 1.4.
3. Use lbe Kremser equalion for design for dilute conoeolntions.
4. If beat effects are large. use sidestream coolers.
5. Assume vapor veiocilY is 3 rtfs, or estimate it from F factor, F _ U.
6. Assume 2-ft tl'1ly spacing-use sieve trays.
7. Assume pressure drop is 3 in.HzO per tl'1lY.
8. Use O'Conocll's correlation for ltay c:fficiencies-lhe effiClenoc:s 10
absorbers are lower Ihan in dislillation columns because Ihe liquKt IS
oot boiling.
B. Packed columns.
I. Applications.
Q. Corrosive materials.
b. Foaming materials.
c. If low-pressure drop requited.
d. If tower diameter is less Ihao 2 fl.
2. Design.
D. Assume 99.9% recovery.
b. Usc I../(mG) ... 104.
c. Kremser equation gives NO(; = 20 for 99% recovery.
d. Assume HOG a< 2 ft.
e. Packed heighl _ 40 fL
f. Assume vapor velocity is 3 ft/s.
g. Use I-in. Pall rings in metal or plastic.
h. Assume pressure drop is 0.5 in.H10/ft.
VII. DIMr Equipment
A more complele collection of shoncul design procedures is available 10 J.
Happel and D. G. Jordan, Chemical Process &onom;cs, 2d ed., Dekker, New
York, 1975, appendix C. Another sel is given in W. D. Baasel, Preliminary
Chemical Engineering Plant Design, Elsevier, New York, 1976, pp. 114, 211.In
addition, numerous textbooks on unit operations lisl values.

1.1, b, C
A
A
c
All
A.. AD
A,
Au, All
b
8
e
C.
C,
C,
C.
D
D
D,
E.
F

Fe
9
G
G
G,
H
H,
H.
11.. h
o
H,
H,

k
k
K,
Cocfhclents In Smoker's equatlon solutlon
... L/(mG), absorpuon faclor
IleaH:Jl.changer area of condemc:r and rebOller
Inside and outside areas
Tower cross-sectional area (ft
l
)
Binary interaction palllIDClers In Marguks
equations
Inleroept of opel'1lllng line
&Uoms ftow (mol,lbr)
Design parameter
Ora, coefficient
I-Ieal capaClIY [Btu/(mol oF)]
COSI of steam (S/Klb)
COSt of coohng waler. SfKgal
Distillate ftow (mo)/hr)
Tube dlameler
To\O,'Cr diameter (ft)
Ovel'1ll1 plale efficiency
F factor
Refrigeranl flow (mol/hr)
Correctlon faclors 10 COSI correlations; see
Appendix C
Gravitalional acceleralion
TOlal gas rate (mol,lbr)
_ PI). mass Row rate
Flow of carrier gas (mol/hr)
Heighl of a plate column
Enthalpy of feed (Btu/mol)
Enthalpy of vapor (Bw/mol)
Film heal-transfer coefficients
Enthalpy of liquid (Blu/mol)
Disengagement plus sump heights
Height of a transfer unit (ft)
Parameter
Thennal conduclivily
Constant
516 KClION" II SUMMARl' AND NOMU<c.... ru." '"a. "f'Pf.NOIX A
SEcrlOt< A I I SUMMARY AND 1'IOMUICUT1JRI! ,"0. A"ENDIX A 317
L
L,
m
M
M,
M&S
"
po
P'PT

Qc. QIII
Q.
Q.
R
R
"
R.
R,
S
SF
To
T,
T,
U. U U.
o
0,
r
v. v
"
"
'v,
X. JC.
X.
X
o
X,
X,., X....,
TOlalliquld now (molthr)
flow of solvenl (molthr)
Slope of equilibrium line
Slope of operating line
Molecular weight of liquid
Marshall and Swift index
Plate number
TOlal number of plates
Actual number of p1atcs
Mmlmum number of trays
Tbcorehcal trays in rectifJing and SloppIng
sectIons
Vapor pressure: (atm or psia)
TOlal pressure (aIm or psia)
Feed quality
Heat load of condenser and rdx)lkr
loads (Btu/hr)
Volumerie now rate
Gas constanl
Drop radius
Fractional recovery
Mmimum reflux ratio
Term in Smoker's equation solution
Reboil ratio
SeparatIon factor
Bubble point (OF)
Condensate temperature ( f)
Inlet liquId temperature
Heat-transfer coefficients [Btu/(hr fl
'
. OF
Gas (fI/s)
Weight fraction of solute in solvent phase
Vapor flow ID stripping section
Vapor rate in rectifying and stripping sections
(molfhr)
flow of cooling water
WClght fraction of solute in nonsohent phase
Sturn flow rate (Ibjhr)
Mole fraction of solute in hquid
Bottoms mole fraction
Distillate mole fraction
Feed mole fraction liquid
Molar ratio of solute in the liquid phase
(mol/mol)
Composition variables
Bolloms mok fraction
Greek

dill
dr.
/J.H.
Y
,
IJr. IJL
p,
P.
0.8
1
,
Mole fraction of solute in gas
Feed mole fraction-vapor
Inlet and oullet molar ratio of solute in the gas
phase (moVmol)
Molar ratio in the gas phase
Feed composiHon
volatility
Heat of vaporizalion (Btu/mol)
Log-mean temperature driving force: (oF)
Heat of vaporization of steam (Btuflb)
_ (C,IC. 1)/(C,/C.)
Liquid-phase activity coefficient
Viscosity or feed or solute (el")
Density (lb/ft
1
)
Molar density (mol/fl
1
)
Root of Underwood's equation
Small parameter
APPENDIX
B
TABU: 8-1
or I Ill" proc:e,)
S .99 098S 0971 .97
'93
,
"
.60
07'
.7>
."'
HDA
CASE
STUDY
fr.- the J'Ml AJell.!:. 51odta' C_1at I'f,*""", Sec
I J M_" -Eac)d..ope.ila '" o.c-:..oJ 1',......... a""
... Y..k 1911pll2
The 1967 AIChE Student Conlcst problem prc:sents the scleclU"lty data
In Table B-1. Ifwc: plol the data as Inti - S) versus In(1 -.x). so that .....c make the
data 35 sensitive: as possible, we oblam the resuhs shown In Fig. B-1.
I. Now, If we fit an equalloD (0 the dala and rearrange the results. we obtain
00036
S = I - (I _ X)I
(B.3)
CondItions: Gas phase, no catalyst
LEVEL 0: I PlIT I FORMATION
Soe J J -Eac:ydopcd.. of ChmlICllI PIOCc:wDI.Jld 0eRp,- .01 .. Dd..la. Nco> "wi
1911, P 182
The definition of the problem was given In Example 4.1-1. The infonnallon of
importance IS as follows:
Reactions:
ReaCilOn com:hlions: Reactor lOki temperature"" 115O"F (10 get a high
enough rate) and reactor pressure = 500 psia
Seleclivity: Moles being produced per mole of toluene: converted = S
I 0 0.1
0
I
'\
i\
"-
i\
f\
01
\
I-x
o.
o I
nGntl 8-1
5ckct.o'lly ollhc IlDA
(0-1)
(B2)
Toluene + Hz ...... Benzene + CH.
2 Bcnzcne;::Dlphenyl + Hz
The design of a process to produce: benzene by the hydroafkflalion of toluene has
been discussed in numerous chaplers in thiS ICX!. The purpose of this appendIX is
simply to colke! and present a SCl of sample calculations for thiS case study This
problem is a modified version of the 1967 AIChE 5IUdent Conlesl problem.
518
520 OF.cISlONS 'N1'\11 OUlPllT naUCluaE
lEVEll IlEOSIOHS IN'U1.()lfT'U1 SUUCTUa.E 521
2. Production rate = 265 mol/hr of ben7ene
J. Product punty:
TA8LF 8-2
Component dfStinatiollS
5, Number of product streams: The component boiling pomts and destinatiOn;
given m Table 8-2 The III and CH. are lumped as one product stream. an e
benzene and arc t"'O additional product streams, Thus, the nowsheetls
gIVen in fig. 5,1-2
6. Material and stream costs: Let
4. Ra'" materials: Pure toluene at ambient conditions; 115 f1
1
CII. Ilt lOOT.
550 p<>13
5. II. aromatics"" 5 at reactor inler (to pre\enl coking); reactor
outlct lemperature:$ IlOO F (to pre.'ent h)'drocraekin!!- re3ctlom),
reactor cmuent 10 I' 50 F (IO pre\ent eoklnB-)
6. Other plant and site data are gi ...en where needed
LEVEL-l DECISION: HATCH VERSUS
CONTINUOUS
II.
Cll,
"'- 10ft......

-.n
'" 17t> 2
Hil
.,..
G.,. 'CICJdt and P'''at'
Gas .nd purat'
Prunl'f) product
Recvele
B)'pfOdUCC
Choose a continuous prOCt"ss
p. - 265 mol'hr bc:n7enc
'C =0,75
(8-4)
LEVEL-2 DECISIONS, INPUT-O TPUT
STRUCT RE
0,0036 _ 0 969'
5=1- '15"--
(I - x)
(B-3)
(B-8)
(8-9)
(0-6)
(B-7)
(8")
(8-10)
Euent 2 =4.1 8
p. 265
F,--=-- -1734mol'hr
5 0.9694
P
D
- p.(1 - 5)/2S
_ 265(1 - 09694)/2(0.9694) = 4.18
Add lhe<;c 10 obtain
Extent I = 273.4
Makeup gas
Dlphenyl produced
Frnh feed toluenc
Sa: 51-2. 521. and 5.3-1
I. Punfy feed stream5 The toluene feed stream IS pUlc;do nOI punfy Ihe hydwgcn
feed stream because the methane Impunly small. Also, mcthanc IS a by-
product of the reactions. and the separation of gases IS expensive.
2. Reversible Diphenyl is a by-product fonned by a reversible
reaction. Thus, we can either recovcr diphenyl from the process or lei it build up
to its equilibrium level in a recycle loop. If we remo\'c the diphenyl, we hne a
selcctl\ity of toluene to produce the dlphenyl. However, if we recycle the
dlphenl'l......e can avoKt the selectivity loss. but .....e o\"ersi7.c all of the
equipmentm the recycle loop to accommodate the reqcle now of dlphenyl. Wc
guess that It is cheaper to reco\'er the diphenyl
3. Recycle and purge; Since a reactant (hydrogen) and both a fen! impullty
(methane) and a reaction h)'-product (methane) boil lower than propylene,
Will need a gas ra::yde and.a purge stream; ie., we want to recycle the hydrogen.
but the methane will build up in the ra::yde loop so that it be purged. A
membrane separation process, "'hich could be used to separate the h}"drogcn
and methane, might be less cxpensive than the leKS of hydr(lgcn m a purge
strcam. Unfortunately, however, no deSign procedure or cost correlation is
available for membrane separators. and so we base our first design Oil a gas
recycle lllld purge
4. Excess reactants: Smce neither 0
1
from air nor 11
1
0 is a rcactantthcre are no
reaClan'"
522 llVel.l INI'UTO\J1I'Ul snlUCTV.E
lHt.1 J 1>1.<15101'<) Sl.L"CT'U or T111IW.... UUl 523
Combine Eqs 0-8 and B-IO
LEVEL-3 DECISIONS: RECYCLE
STRl"CTURE OF TIlE FLOWSHEET
Rc:move lhe CII. from the:: III reed Slte3m This IS probably not desirable
because .... e pr<>duce lb a by-product anyway
dlphen)1 ralher th.1I1 removmg II \Vuh lhis approath ....e avoid
In)) of lOluene to dlphen)I, bUI .... e muSI "II the
equipment In Iho: "hphen)l-reC)c1e loop 10 accommodale the c'Iulhbnum flow
of Jlphenyl
Rcco\er some liz from the purge stream We mUSI delermlne .... hether lhe liz
reco\ery IS Jusllfied b) delermlnmg the cost of the recovery syslem
proc(: ./) we develop II II is easier 10 revie.... lhese decisions
afler "c complete a design The alternatives al k\e1 2 arc as rollows.
(BII)
I.
I.
(B-12)
3.
(8-13)
Excess HI
fa" 1'.(1 - (I - JI'II)(I - S)/2]
S(YFH )/1'11)
_ 265[1 - (I - O.4XI - 0.9694)/2]
- 0969-l(0.95 _ 0.4) = 492 5
p. :: 4925 + 265(1 0.9694)_
(; 2(09694) - 4967
= YI'"F
G
= 0.4(492.5) = 197.0
"
R
p. p.
1 cacted = S - 2S(1 - S)
= 273.4 -418= 2692
Slream COSIS
Design Decisions
The deslgn deClSlon) for the recycle Slructure ....ere dbCussed In Example 6.1-2
arc bndl) reHe.... ed no....
J. one reaclo, I) fI:quucd (lhe rcactlons tale place al the same temperature
and pressure)
2. There arc ' .... 0 reC)c1e streams-a gas reeyle (and purge) of H
1
+ CH. and a
lJquld toluene-reqcle Slream
3. \\ e musl use a 5, I rallO al lhe reactor Inlet accordmg to
the problem stalemenl
4. A gasrccyde compressor is requtred
Recycle Material Balances
The rcC)'cle material balances werc developed as Eqs 6-2-2 and 6_2-10.
Toluene to reaclor:
Before .... e can decide on lhe reactor heat effects, we must calculate the recycle
matenal balances The reqcle ftowshect IS shown In Fig. 6-2-1.
(8-14)
(8-15)
(8016)
(8-11)
(8-18)
(819)
(8-20)
(8-21)
Heal values:
Value of Benune = SO.85, gal = S904 mol ... C.
Toluene - SOSO'gal = S6.40 mol = C
T
(assuming a capllve mlernal prioe)
H
l
Feed = SI.31,mol = C
H
,
H1 : 0.123 x 10" Btu mol = aH
c
.
1I
CH.:: 0.383 x 1()6 Btu/mol = liH
<.M
Benzene"" 1.41 x 10' Blu/mol = aH
c
..
Toluene - 1.68 x 10
6
Btu/mol = aH
<.,
Dipheoyl _ 269 x 10' Btu/mol = al1
cJJ
(We assume thaI the fuel value of diphenyl is S5.38/mol.)
Economic potential:
EP - C.p. - CTFF + CF{liHc.DPD+ [afl
c
.
1I
Y'II + 6H
o1
(I - YI'H)]PGI
(8-22)
We can use this relationship and Ihe expressIons above to prepare
Fig. 5.3-1.
uvel-2 Allernatives
(B-23)
We made several decisions concermng the structure orthe flowsheel and If we had
made dlfTcrcm deciSions, we would have changed the flowsheet .If we Jist lhe
265
- 09694(075) = 365
524 LVEL J DEClSIOI'l' RlC\'C1.l!. "RUClURIll Of" THill FWW$HT
llllyE....J OCUIOPII UCVcu: mucroRIll Of THII': FLOWSHEET 52.5
(8-30)
(B-29)
(
p_')' (SSS)"'" (T,)
- 1.046 __
PIn 465 T
I
and for our first design we assume that Ihis value is constant. The gas density at
100F and 465 psia. given an ideal gas, is
_(_1_ mOl) (460 + 32) 465 _ 0.0774 moVrtJ
P.. 3S9 fl' 460 + 100 1.7
Also,
(8-24)
Recycle gas:
_ 265 (_'_ _ 0.95 ) = 3376
0,9694(0.4) 0.75 0.95 0.4
Reactor Heal Ell"f'Cts
and lhe inlet pressure is
The reactor heat effects were diSCussed In Examples 6.3-1 aod 6.}3. The reactor
heat load is
(B-3I)
(B-2
The volumetric flow rate is
RGtmoljhr) Rr;
Q... = (p.. mOI/fl2X60 min,lhr) - 6Op.
(8-32)
and the adiabatic eXit temperature IS gi ...en by
p.
Q.=AIl
S
= (7.16Fr; + (7)"", + 10.1(1 - J'plI')]R
to
+ 48.7F
T
}(T .- - 1150)
21.530P./S
T._-II50+ -
7.16F
G
+ (10.1 3.1YplI')Rr; + 48.1P./Sx
". 1150 + 21.530(265/0.9694)
7.16(492.5) + [10.1 - 3.1(0.4)]3376 + 48.1(265)/[0.9694(0.75)]
= 1265 (8-26)
These results are shown in Fig. 6-3-1. and we find that we expect that an adiaballc
will be aa:eptablc.
Then since
273 ( 5 0.95)
-- _3371mo r
R
G
0.4 0.75 0.95 0.1 Ifb
3311
Qlo "'" 6O(0.On4) - 127 rt'/mln
Also, hp = (6.70 x lO
o
X721Xl.046 - I) = 267
and the brake horsepower is
2.7
bhp - - = 334
0.8
Guthrie's correlation (Appendix E.I) gives
(B-33)
(8-34)
(8-36)
Recycle Compressor Costs
The reactor pressure is given as 500 psia. We guess that the pressure at the phase
spliller is 465 psia and that the recycle compressor must increase the pressure to
555 psia This allows a pressure dropof90 psi though the gas recycle loop. We need
to assess the sensitivity of our design to changes in this guess..
The design equation for a gas compressor is
U M&S _ 792 aDd Fe _ 1.0 (a DtDtrifugal compressor), and introducing a capital
charge factor of! yr to put the installed cost on an annual basis, we find that
h
(
3.03 x 10- ') [(P_)' )
p_ - p. Q. --1
l' '0 0 Pl.
Using Table 6.5-1, we can wrile
y - 0.29YPH + 0.23(1 - y,.H) - 0.254
(B-27)
(B-28)
(
792) 0.82(3.11)
CompressorC05t= 280 (517.5)(334) ] =SI77,800/yr
For optimization calculations. we can write
(
R, )"" [(",/4.')' -1)""
Compo Cost... 3]71 1.046
(B-37)
(B-38)
(B-39)
326 J PEelIIOl'lS UCYCU STlUCTUlE at TIlE nOWSItf.T I SUUCTUlE Of TIlE fl-OWSHEff 527
The operating cost is based on a mOlor efficiency motor) of
08 and a po.....er cost of S0045!lwhr:
Reaclor COSI
The kiQ(:tics for the pnmary reaction ha\'e been discussed by Silsby el al,-- and
Zimmerman and Yor" 1 Sinu the: amount of dipbcnyl produced IS \ery small, we
base tht: reactor design only on the primary reaction. (Ho\loner, a "lOetie model for
the b)'-produci reaction IS avaliable 10 Hougen and Watson. I) So
where TJI =: (1150 + 1265) 2 .. 1207. from Eqs B-42 and B-43
klll/l =: 6.] x 101<1(8 mOI)"[( I )(28.32 L)J-1'l
l 4541b mol ft
l
x [04(0.0279) mOlJI
1
2CAp [-
ft 1 I 987( 1685)
;. 2-522 )( 10[04(00179)0 S up [ 9],600 ] = 001421 ,
1987(1667)
Then. from Eq 8-44,
_ (492.5 + 2048 + 364)(mollhr) In (I{(I - 075)] _ ]
.- (00142s-
'
X3600slhrXO.0279mollft) -4090fl
If we assume a cyhndrical reactor wllh L.ID. '"" 6. then
(
334h
O
)( Ikw )(0045)( h')
Ann Cost- 08 1.341hp kwhr 8150
yr
=SI14,OOO/yr
We can also wflte tins expression as
Po,,"er COSI - = S3]8IRG/yr
(8-40)
(8-41 )
and thus,
,0'
',c __
4
D.=952ft
:; ]
L. - 4(6D.)
L. - 56.96 fl
(847)
(8-48)
(8-49)
wher':, the molar density at the reactor conditions (assuming an ideal gas) is
Smoc there IS a large excess of hydrogen, \Io'e MSUmc lhal Il' 1 IS a constanl
Also, ""'e assume lhat .... e can esllmalc the reactor \'olume based on isolhemlal
operation, although \Ioe .... '11 base the Isothermal reactor temperature on a mean
value bet .... een Ihe reactor IOlel and outlel temperatures T. = (1150 + 1265)/2-
1208 Thus, .... e wrlle the reactor volume V. as
VIle neglect the cost of the insulalton In our first esllmate of the reaclor cost. hopmg
thai the additional COSt associated with oversizing the reactor will compensate for
the tnsulalion.
(851)
(850)
D._ lOft
The ml.el and outlct t.empcratur.es for tbe: reactor ar.e 115O"F and
which IS very high, and thercCore we will add 6 In of Insulalion on lhe inside: of the:
rcaclor to try to tbe inSIde walllempcraturc below 900"F. The addmon orth,s
insulation requires thai .....e: add anotber foot to Ihe diameter of our reactor shell,
and 10 be somewhat conseJ"\au\e .... e Ic:I
Then. from Guthrie's correialion for pressure vessels,
Ann. Cost = (792) (101.9X101 OCi.X600 11) 2.18 + ].67 + 1.6
282 3
= S239,254fyr
(8-11)
(8-43)
(8-44)
" _ _-__ x."JJ
kp...
r= _k<D(H)I'l
10' 0 (g. mol 1..) - 111(5 I) exp [-52,000 calf(g
RT.
.l ;. 6.3 )( where
(
I mOIX460 + 32)(500) ,
P. - 359 ft] 460 + T. 14.7 '"' 0.0279 mol/ft
(8-45)
Economic POlential
R I S,llbyand E W Sawyer;J Appl Clw:m.6 )41(AullW;1 ISlstI); W D fkllf," I'opper.lnd R I
Silsby, J Appl Ckm. 1 491 (Seplcmber 11151)
, C. C Z,mmennln and R Yml. I&. f I'roc Ve. On:.,3 254 (July 111(4)
lOA Ilougcn lOO K M WllfQn, Clw:rrllc"ll'rocru PrI".,lple5 "' .....un ""J ("UIIJyflJ, Wiley,
New Vorl, 1941, p an
By sublracting the annualized reactor cost and both the annualiz.cd capilal and the
operating cost of the compressor from the stream costs (i_e" the levd-2 economie
potenlial, Eq 822), we can calculate the economic potential 31 levc:1 3 liS
a function of the design variables (conversIOn and purge composition); see
Fig 6.7-1. We note that there: IS all 0pllmUnl value for bOlh con\'ersion and purge
composilion. Of course, these optima arc not the lrue optima because we have not
considered the: separatIon COSIS or the heat exchanger costs yet However, we do
lEyu.... YAI'OIl SYSTEM 529
Dote that the range of the design variables where profitable operation IS possible
has been significantly reduced. Since the process is still profitable. we continue to
the next leYe!.
plant group to rUIl some additional experiments (which is why we want 10
undertake the conceplUal design study \ery early in Ihe life of a project). The value
or the benzene: in the flash vapor stream. assuming that all of it is lost. is
LEVEL 4: SEPARATION SYSTEM
Now ",e add the details of the separation system. To determine the general
structure of the separation SJstem, .....e must decide .... hich of the ftowshec:ts sho.....n In
Figs 11-2 through 7.1-4 is applicable The reactor effluent is at 1265"F and
therefore is all vapor Then, providing we obtain a phase split at 1(lO'T and 465
psia, the ftowshec:t gl\'en in Fig. 7. 1-4 will be correct.
We can use eIther the Hadden and Grayson method given in Sec. C 1 or a
FLOWTRAN program to find the K
f
yalues for the flash drum. The flash
calculations were dISCUssed in Sec. 7.1, and as shortcut expressions .....e use these'
Light components, K
j
> 10:
(8-56)
(B-58)
(B-S7)
496
496 + 3376 - 0.128
Benune in PUTge = 0.128(29.6) = 3.79 mol/hr
Toluene in Purge = 0.128(3.6) = 0.461
Fraction Purged .., PG
P
a
+ R
a
so that increasing the pressure will decrease the K yalue for benzene and will
decrease the amount of benzene in the Bash vapor. Since the number of moles is
conserved in the reactions, Eqs. 8-1 and 8-2, pressure docs nOI appear to affect the
reaction rate or the equilibrium. Howeyer. the possibility of coke formation or
h}'drocraeking reactions (both of which were mentioned as the causes of some
conslfaints in the problem ronnulation) again makes it essential to undertake
additional eJ:pcriments before we undertake II design al R higher pressure.
Rather than get bogged down in details about lhe benzene-recycle loss or
pressure effects, we go ahead and complete a firsl design In particular, if the
process is not profitable in any e\ent. so that we decide not to proceed to a final
design, then we do not want to waste money on experiments developing a data base
for a plant that we are not going to build.
LEVEL 4a: VAPOR RECOVERY SYSTEM
\Ilo'hich IS \ery large. Thus. there IS a greatlncentl\e rOl undertaking the experiments.
As an ahemati\e,..-;e could allemrtto Increase the pressure ollhe flash drum
Since PrYI = ;,P'!x" we often can obtain a re:asonablc estimate or the effect of
pressure by using the expression
Potential Benzene Loss.., (29.6 mol/hrXS9.04/moIX8150 hr/yr)
= $2.18 x Il)''/yr (8-55)
If ""Ie are going to include a vapor recovery system, we must decide on the location
(flash \'3por. purge stream, or gas recycle) and the type of system
(absorption. condensation, adsorption, or membrane system). In the discussion
above, we estimated the value of the total amount of benzene leaving in the flash
vapor. If it shouk! be acceptable to recycle benzene. we still must estimate the
blenzene and toluene loss in the purge stream. to see whether we should place a
rcco\ery system on this purge stream. The flash vapor flows of benzene and toluene
are gtyen in Table 8-3, the purge flow is given by Eq. 8-10. and the gas---recyele flow
IS gi\en by Eq. Thus, the rraetion or the flash vapor that leaves as purge is
The benzene and toluene losses are then
(8-52)
(B.S])
"_ _L{!)
Kf L,/;
I
f- K
I
'[/;
Heayy components, KJ < 0.1:
-
With these eltprcssions, we calculate Ihe vapor and liquid flows leaving the flash
drum (sec: Table B-])
'[II"" 1549 + 2]2] = 3872 r..l1 - 26j + 91 + 4 = 360 (B-54)
From these re5ulls, we see tbat we obtain a reasonable phase split. Ho.....ever,
significant amount of benzene leaves with the flash vapor. Some of this benzene
will lost in the purge slream, and the remainder will be recycled with the gas
stream. From the reactions in Eqs. B-1 and 8-2, we would expect that50me of this
recyeled benzene will be converted to diphenyl Ho-..'Cyer. our selectivity correla-
tion. Eq. B-3, does not indicate that there will be any loss.
We suspect thatlhe data given in Table 8-1 were for a pure toluene feed. so
lhat no loss wouk! be apparent. Hence, from the informatioo available: it is not
possible to estimate the benzene loss. and we should ask the chemist or the pilot
TABLE 8-3
Flash tak:ulations
Com..- ft II
K,_K,

H,
"..
9901
".,
2
CH,
""
21700 2312 II
....
'"
0.QI04 29.'

ToIliellC
"
0.00363 ).. 87.4
Diphe:nyl

0."""'"
0
,
(8-59)
530 u;vn 41> LlOUIl) u,........ flOf" sysrUI
and lbe values of these strC<ilms arc (sec Eq 8-54)
Benzene Loss "" (59.04 molX179 mol,'hrX81SO hr,'yr) = S279,200'yr
Toluene Loss _ (S6,40,'mol)(0461 mol'hrX8lSO hr/yr) = S24,000/yr
Now we muSI guess whelher Ihe addulon of a benzene and toluene recovery
system on Ihe purge stream can be justified Ilowe\er, If the coSIS of the hqUld
separauon system and Ihe heat-exchanger networl.. are large thdt Ihe
process loses money (and we deCide not to bUild il). 'ftc do nOI want 10 spend much
dron on Ihe design of a vapor reem'ery sySlem. Thus, lentallvely, we accepl thIS
purge loss., ahhough 'fte might relurn to Ihis problem laler (For example, we could
use the toluene feed as Ihe sohenllD an absorber.)
TABU a...
Rash
C__
I. or f
j
A,OIK
J
II,
,
921
19'
00>
CII, II
'"
911 1,49
Bcnuoc 2354 00906
08'
'U455
rO]llCOC
."
00337 011
"',
Dlp"""yl

0,000074 0 0
usc Ihe lIadden and Grayson correlations gIVen In Sec. C.I for such a large
pressure change). Thus, we obtain the K values and the component flows from
Table 8-4:
which IS sull less Ihan our desired punty of X
D
= 0.9991 (e\en If we neglect the
toluene III thiS slream). Moreover, there is a fallly large benzene loss from Ihls low.
pressure ftash; ie., from Eq 8-59
LEVEL 4b, LIQUID SEPARATION SVSTEM
From the results of the flash calculalion (see Table B-2), we have an estimate of the
amount of H
2
and CH. dissolved In the flash liquid. Hwe assume Ihat we recover
all the benzene, then the feed 10 the liqUid separallon syslem is
Feed: H) = 2 CH. = II Benzene'" 235.4 Toulene = 874
Diphenyl _ 4 mOl/br (8-60)
Le,eJ-4h Decisions
The decisions we must make concerning the liquid separallon system include
r./j = 235.4 + 81.4 + 4.0 = 326.8
Now If we recover all the H
2
, CH.. , and benzene, we find that
234.55
XD = 234.55 + 0.05 + I 49 = 0.9935
j085)
Benzene Loss = (S483,ooo/yr'\316 = SI09,200
1
).,
(8-62)
(8-6J)
(8-64)
(8-61)
I. How should the light ends be removed If they might contaminate the product?
1. What IS the best destination of the light ends?
3. Do we recycle components that form az.eotropes wilh the reactants?
4_ What separations can be: made by dislillatlon?
So What sequcnce. nf columns should we use?
6. How shnuld we accomplish scpirations if distillation IS nOI feasible?
We discuss most of these decisions below. However, no auotropC5 are
formed with the reactants and distillation separalions are easy for all Ihe
components. so Ihat items 2 and 6 are not considered.
UCIIT ENDs' If .....e recovered allihe benzene as well as the: H) and CH. overhead
in a product column, the product punty 'Ilould be
= 235.4 = 0.948
xI) 2 + II + 235.4
which is well below our product purity requirement of XI) = 09991. Of COUfSC,
some toluene musl Icave 10 thiS stream.
If we attempt 10 drop the pressure to 50 psia and flash off the 1-1 I and CH.. , we
can obtalO a rough esllmate of the K values by USlOg Eq 856 (It would be beller 10
Hence, our estimates Indicate that .... e probably will need a slablhur 10 obtalO a
product stream with the required purity (a pasteurization section on the product
column might be acceptable, and we list thiS as an
If 'fte recover Ihe hghl ends (Hz and CH.) in a stabihzer, then we would
normally send lhese lighl ends to the vapor recovery system 10 rcco\erany benzene
or toluene that lea\'es with this stream (or to any unit thai recovers and recycles
some of the hydrogen from the purge stream). However, since we have nOllOciuded
units of this type (at least at this time), we would probably send the hght ends
to tbe fuel supply.
COLI: L,,"SEQUENCINC. To use a stablhzer column to remo\e the Hz and CH..
from the benzene product. we normally pressurize this column to make It easier to
condense the o\erhead and thereby 10 oblalO an adequale amount of reflux_ Thai
!S, the HI and CU. arc: remo\ed as a vapor slream from Ihe reflux drum after a
panial condenser, but il is nece.s.sary to lake some benzene overhead to provide an
adequate amount of liqUid reflux. From Table B-3 the K value of bcn.tene is fairly
high at 50 psia, so that we might SCi Ihe stabilizer pressure at ISO psia, or so.
H we use a column sequence thai does not remove Ihe light ends III the first
column, then ellcrycolumn that totally condensed the HI and CH. in Ihe overhead
.... ould ha\e to be opera led at high pressure. Thus, lhe capital COSt probably would
532 LEvu.. *' UOIJlD nPu.A'OON SYsrEM
LIQ\IID SYSTEM 533
TOLUENE COLUMN. From a plot of the vapor pressures of loluene: and dlphen)l
we find Ihat Ihe sk>pcs of the vapor-pressure: curves for lhese 1.....0 componenls are
somewhat dlfferenL This result, aceordlOg to the Oausius-Clapeyron c:qualK>n.
imphes that the heats of vaporization of the 1.....0 compounds are drlTen:nl, whIch In
turn implies that the common assumptIon of equal molal o\erflow in the column
.....iII nOI be correct We could correct for Ihis difference 10 latent heats b)
introducing a fictilious molecular weight for one of the components and then uSlOg
the McCabe-Thiele procedure to design the column. However. for our prelimmary
calculations .....e Ignore thiS potential difficulty.
From the vapor-pressure data.. .... e find that
be greater !.han If we had removed light ends In first column Similarll,
operation of 5C\'eral cDlumns al high pressure Increases the bubble poinls of lhe
bol1oms SHeams. so Ihat a higher-pressure sleam might be required 10 dri\e Ihe
re:bollers. For these reasons. we assume that it is cheapest to remove the: light ends
in the: firSI column
When we then I:Qnsider the separation of benzene (234.5 rnolfhr). toluene
(87.4 moljhr), and diphenyl (4 molfhr). almost all the hcurislics (i.e, lightest firs!.
most plentIful first, favor equlmolar sphts- but not the easiest first or save dlmeult.
hIgh-purity splits until lasl). favor the dlP:ct sequence Hena:, for our base-ca<;l,"
dcslgn we Will choose the hghtcst first scquena: 10 evaluate Ifowe\cr. comple\
columns might pro\'ide a cheaper separalion system.
Now we want to cslimate the SizC5 and the costs of the columns. Smce the
recycle column, I the loluene-diphenyl split. involves only a binary milture, we
consider the design of this column first. SImilarly, we consider the design of the:
product column before we consider that of the stabilizer. We design both of the
recycle and product columns to operate at slightly above ambient pressure because
it is easy to condense both toluene and benzene with cooling water at this pressure
Assuming a saturated-liquid feed, Undcrwood'sequatlon for mlmmum renuJI
(8-72)
(B.7I)
(B-7])
(B74) R:::::: J.5R.. = 1.5(0,0]47) = 0.05
R = I [0.9997 25(1_ - 0,9997)J
- 25 - I 0,956 I - 0.956
= h (1.0456 - 0227) = 0,0347
In [xaJ(1 - XD)]((1 - x.J/x..J In[(0.9996/O()()()4XO.9051O,095)]
-= =31]
.. loa In 25 .
(8-75)
If we recover 995% of the toluene o\-erhead and of the dlphenyl In the
boUoms. then we find that
d
r
= 0_995(874) = 811 d" =U.005(4) = 0.021 (B-67)
,nd x _87.1 = 09996
(B68)
" 87,1 + 0021 .
Also,
"'r .." 0.005(87.4) = 0.4]8 """ = 0 995(4) = o. '98
(R-69)
.nd
0438
=Q438 + 398 = 0095
(8-70)
which is \'ery low: the feed composition is very high. and III is \'ery large With a
low of renul such as th,s. .....e should also consider the usc of only a strippmg
column as an alternatIve. Ho"'e\er, w-e contlllue wilh Ihe design, and we let
According to fenske's equallon, the mlOlrnum number o[thc:orellcaltra)'s at
total reflul needed for the separation is about
(B--65)
10.000
11_= 405 = 24.7
160
"- - - "" 102.7
1.4
This is a large vanation in a. as well as a very large temperature gradient across the
column, thai is, 110.6C "" 231F at the top versus = 492F al the bollom.
Thus, we el:pect that some of the simplified design prooedures, such as Fenskes
equation, Gilliland's correlallon, or Smoker's equation, which we often use to
estimate column designs. may give misleading predK:tions. Nevertheless. we might
be able to get some idea of the column desIgn with these shortcut design prooedurc:s
if we choose: a conservative estimate of a:. For thiS reason we let a: "" 25
Given no losses of aromatics anywhere in lhe process (which, of course, is not
really consistent with our other calculations.. but the error is small), the feed rate to
the toluene column is 87.4 mol/hr of toluene and 4 mol/hr of diphenyl, so lhat
The o\erall plate efficiency is given b)' O'Connell's correlation, Eq A2-72 For a
quicl C5tlmate we assume thallJr = 0.3, and we write thai
We can obtam an estimate of the number of theoretical trays requ,ired at the
operating renuJ! ratio by using Gilliland's approlimation
(R 7R)
62
"" 21 6 = 22
B30
0.5
Eo <:l:: (0.3a:)0 H = 0,302 (8.77)
Then the actual number of required lrays IS
(1l-66) x, = 87.4f(R7A + 4) - 0956
Thus, the feed composllion of toluene IS qUIte high
I 'b IIVUm 5ErAI'.ATKll'< nSlHl SJS
For a 2-fl tray spaclllg and an additIOnal 15 1 at the ends, Ihe tower height IS
The area Ihen bc..:omelo
1/ ... 2(22) of 15 = 59;:::; 60ft
The lower cross-sectional area can be estimated by using Eq
(6-79)
A 3-12
A
I9,600 {91 7) 60f 2
11,250 =1 I
(8-911
and '0\.(' wanlW base Ihe deSIgn on lhe oonom o(lhe tower, ie., the dlpbenyl, where
At = 154 and r. = 492 F The vapor rate b wrillen as
C
(
S225)[19,6OO(917)lbJ( .,) 96000
Ann_ ost = UX)() Ib 667.5 hr 8150 yr ;. S4 , yr
(B-93)
(8-9)) Anll
and lhe cost is
COlot = UlC:9) = S8600}r
The boiling POlllt of dlphen} lIS 492 F, and so we must use 1000-JhI Sleam (or
some high-pressure level} III the reboller Of course, we could use 42Q-psI steam If
we (lperated Ihe tower under a \acuum, so that the bol1l11g POlDt of dlphenyl 'o\.'as
reduced 10 about 42O"F (.... hleh 'o\.ould allow a 4 Tof 30"F} 1Iowe\-er. smce the costs
associated with thIS column arc reasonably small up to this point, 'o\.e usc: IOOO-psi
steam Then the steam COSIS are
(8-8')
(8-83)
(8-82)
(6-80)
f4A
D= ,r;: = 2.7;:::; 3ft
Vg L + D - (R + I)D = 1.05(87.1 + 0.021) = 91.73
Now we can use Guthnc's correlatIon to find the COSt
and the area becomes
A - 2.124 )( 1Q-JI54(492 + 460)(91.1) =7.5 ft
J
The column diameler IS
TOLUENE COLUMN CONDE 'SER AND COOLING WATER. A condenser heat
balance gives
Ann Cost = (101_9X3
1
102(3.;8) = S26,300Jyr (B-84)
(B9-1)
235.4
XT - =0720
2354 + 91 4
ThIs cost is fairly high, so thaI ,,"c mIght want to cxamllle som(' a!t('rnatl\es lal('r
BE.... ZE:-;E assuming perfect separations and no wc liCC'
Ihal lhe: flo,," ral(' 10 Ihe bc:nzene column contains 235.4 mol hr of benzene.
874 mol 'hr of toluene, and 4 mol br of dlph('nyl. Since the dlphcnyl flo.... rale IS so
small .... (' assume Ihal we could oblalll a reasonable e5tlmale oflhe column dcslgn If
\l,-e lump II IOgelher 'o\.llh the toluene Then
In addition. our produci speClficallon for benzene requires thaI the punty bc:
9997-;,. and we want 10 99.5% of Ihe bc:nzene overhead. Wilh IheS('
reslnetlons
(B-86)
(8-87)
Qc - dll, V= UcAcdT.. = w.C,.(I20 - 90) (885)
aDd if AH. = 14,400 Btu mol U
c
= 100 and r. _23'-F, then
120 - 90
AT - = 125F
- 10 [(231 - 90)/(231 - 120))
14,400(91.1) 1
A
c
- 100(125) = 105ft
Thus,
TOLUENE COI.UMN REUOILER ANn STEAM. A heal balance for Ihe reboller
gives
c ( $0.06 ) ( I 8.
1
) [14,400(91.7) IbJ ( .,)
OSI - 1000 gal 8.341b 30 hr 81SO yr - S2600/yr
(B-89)
(0-97)
(B-96)
(8-98)
(
' - 0.9997)
d'
r
- 234.2 = 0.Q7 (895)
0.9997
w
T
= (87.4 + 4 - 0.07) = 92.33
]C. .. _ 0,013
d. = 0.995(235.4) "" 234_2
w. = 235.4 - 234.2 _ J.2
From Ihe vapor-pressure data we find that 0I :::l;:: 2,5, and then Eq. 8-71 giv('s
I I
R..,;:::; = = 0.926
{a - I)x, (2.5 - 1)0.72
R =:. I.2R.., _ J .2(0.926) _ I II Hence
'0
(8-88)
(8-90)
(
792) (3.29)
Ann. Cost = 280 (101.9)(105 ") -3- = S6500/yr
Aoo
and Ihe annual cost IS
Also, the coolingwatcr costs are
LIQUID n'....... T1Dl'I SYSTnl SJ7
Also, from Eq 8-15. and the cost is
In [(O.9997ftl00J3)(O.981/0013)]
N..... In 2.5 = 13.6
"Then, by Gilliland's apprOJItmation
Ann. Cost = G:)(IOI.3)(9280
= S26.700,yr (8-109)
The cross-sectional area or the column is given by Eq. 8-80, where we
evaluate the area at the: boUom of the colum and V is gi\'en by Eq 8-81. Hence.
For a values in lite range of 2 to 3, Eo 15 msensHIVe to a (sec Eq A.2-72). and
therefore we can guess that Eo ., 0.5. Hence, the tOlal number of tra}'$ required is
A = (2.124 x 1O-).j92(231 + 460) [(1.11 + 1)(234.2 + 0.07)
., 19.73
N
r
:::::.2N.. =2(Jl6)-272
(8-111)
(B-1l2)
(B-1l4)
(8-113)
(
792) j329)
Ann Cost = 280 (101.3)(708
06
\.3
= S20.800/yr
_ ( $1.65 ) [14,400(494.7) Ib](SI50 hr)
Ann. Cost 100J Ib 933.7 hr yr
= SI02,600/yr
Morco\'er, the steam cost is given by
Also, Ihe cooling-water cost 15
and the COSI is
Q. _ MIl' P _ V.A. aT. _ lJ.1/
s
so that with .6Hr- 14.400, P = V, and VIII aT. = 11.250, then
14,400(494.7) 708 f 1
A.= 11,.250 = t
BEl'IlZEl'IlE COLUMN RE80lLER AND STEAM COSTS.. A reboiler heat balance:
gives
STABu.lZER. The design of the stabilizer is not as simple as that for the other
columns. FlI'St, it is vcl)' diflkuh 10 lind reliable thennodynamic data 'for mutures
of and methane with aromatics. The vanety of points that could be used
on tbe Hadden and GnY50n charts (see Sec. Col) illustrates this difftculty_
In addition, the millOture of hydrogen and methane that we desire 10 remove
from tbe lop of lhe column is essentially noncondensible, 50 to obtain a sufficient
amount of liquid ror reftullO, we musl allow some benzene to go ovcrhead. However,
as I'o-e saw in the purge-loss calculation (Eq. 859). even small flows of benzene are
quite valuable. Thererore, we want to pressurize the stabilizer column, both to
minimize the benzene loss from the partial cQndenser and to be able to use normal
cooliog water in this condenser. or course, as we increase the operating pressure,
we increase the capital costs of the column, of the condenser, and or the reboiler
because the wall thickness must be increased lienee, we anticipate that there will
be an optimum operating pressure for the column.
( ,"06 ) Ib] (8150 h')
Ann. Cost 100J gal. 8.341b 30 hr yr
_ SI2,900/yr (8-110)

(B-I04)
(8-103)
(B-IOI)
(8-102)
(8-105)
(B-I06)
(8-IOS)
(8-107)
beat balance
4A t:F'9.73)
11 ""'..;4 \.----;:--.- - 5.01 ft
27.2
N 54.4:::::. 55
0.5
1/ _ 2(55) + 15 - 125 ft
D_
120 - 90
6T. - 70.9
13,300(1.11 + 1)(234.2 + 0.07) 8 r'
A
c
- 100(70.9) - 92 t
Ann. Cost - G:)(101.9)(125
1
066)(5.0I
U01
) c'; 8)
_ SSI,300/rr
and the column height (allowing 15 ft at the ends) is
Then, from Guthrie, the cost is
and the diameter is
Qc= lill" V "" UcAcdT.= w.C,(I20-90)
Witb 6H" - 13,300, U
c
= 100, and
BENZENE COLUMN CONDENSER AND COOUNG WATER. A
00 the condenser gives
we find that
338 L(V1[L'. LMJUID 51oI'AILt,TION sysnw LU'El'. I"JUID nl'AUllO!' S\SnM 539
which is fairly large. Of course, this loss is compensated, at least in pan, by the fuel
valoe of the benzene in the overhead stream
After eumlDlDg the sensili\;ty of the benzene loss from Ihe flash drum of the
slabilizer, we decided to choose K. = 0.04 with a condensing lemperalure of
115F, so thai for our finl design lhe operaling pressure O(lhe stabilu.er is ISO psia
At these condilions
(H-125)
(H-1201)
(8-127)
(8-128)
(8-129)
.l. "" 0.%1
3
_. _ -_900
......... 004
3.
... _- = = 36
,
'CH. = 00324
70
a" =-=70
, 1
1 [0812 dO.<3')]
- 36 I 00324 3,0961 - 0689
and Ihe feed COlllpo)lIIon IS
XII> = 00055
F== 2 + 11 + 326.8 - 3791:1mol'hr
BOllom.
Top.
With thIS as)umplion, Ihe effluent from lhe bouom of the lO"er v.11I be
essentIally pure benzene boihng at 150 psia Then from the Hadden and Grayson
correlalions for K, = 10 and P _ ISO psia, we find Ihat lhe temperlllure IS 360' F.
However, we expect Ihat Ihe presence of the toluene and dlphenyl .... llllOcrease IhlS
boiling POlOl, thus we guess that Ihe bollomstemperature is about 4OO'F. Also, at
ISO psia and 400F, .....e find Ihal
K
H
, "" 70 K
CH
_ 36 (8-126)
Nov., from Eqs. 8-119 and 8-125 we can eSllmate Ihe a "alues,
Clearly to introduce an assumpllon of constant rc!:Hi\'e volanht)' .... ould be
rather "shal..)'" bUI all the shortcul design procedures rely on this ahumpllon
Hence, for our fir)t design we assume thai :l
j
IS cunstant, but we use the smallest
possIble value, Eq. 8-127 at the botlom, In our calculallons.
Underwood') expression for lhe minimum reflux rates for binar)' separallons
should give us a conservath'e estimale for multicomponenl separalion), so
R _'_[X
u
.
c
, - a.... X
D
,,]
a-I X,-.c, ,- X'.M
We !l.ln\plify the dOlgn cakuldllons If .... e lump all tho: aromilUCS and
conSider Ihem 10 be benzene, so thai the benzene feed rale 10 the column becomes
(sec Table 1J11 2354 + tl74 + 4 = 326.8 mol,lhr. lienee, Ihe feed rate I)
(8-115)
(B-116)
(8.117)
(8-118)
leaving in lbe
Ann. Loss = 9 04(0.7)81 SO = S51,6(XJlyr
From Table 8-3, we expecl tbe h)'drogen and methane flows
stabilizer overhead to be about 2 and 11 molfhr, resptttl\e1), so
0'
'=K= ".
J. , 2+11+n.
13K.
n.=---
I - K.
If K. = 0.05, then lhe benzene loss calculaled from lhls cxpressioll IS aboul 07
mol/hr, and from Eq B59 Ihe value of the hcnzelle loss IS
To fix the opcrdllng pressure of the slablhzer, ....e ought slart by examining
Ihe behavior of the Rash drum when K, <01 for benzene For example, by
considering the value Iisled in Table 8-3 for benzene. we might inlllall)" guess Ihal
.... e want K. _ 002 Then, from the Hadden and G,-ayson correIalion (Sec. C I), Ihe
p,-essurc of the Rash drum bas to be 280 pSla 10 condense the at 100"",
and II is not nen possible 10 condense lhe: o\crhead at 130 F SImilarly, If .... e lei
K. = 0.05, .... e find lhat the pressure I) 80 p5ia at 100 F and 165 psia at IJO'"F, so
lhe pressure calculation is very sensime.
No.... , 10 determine the importance of choosing values of K. thai arc \'ery
small, ie., in tile range from 002 10 005, we mUSI eSlimate the value of Ihe benzene
that is lost with Ihe vapor leaving the flash drum. We expect thaI the values of K
H

and K
CH
will be very large compared to unity, and so we elpectthat the liquid in
the flash drum will be essentially pure benzene. Then Ihe equilibrium relationship
when x.:::::: 1.0 gives
K
CH
= J6 (8-119) Then, (or an operating reflux, we: choose
and the benzene loss is R 1.5(0 689):::::: 1033 (8-'30)
The distillate flows are lhen
13(0.04)
n. = 1.0 _ 0.04 = 0.54
Ann. Loss 2; 9.04(0.054)8150 = S39,800/yr
X....
II
= 0.148
(8.131)
(0-132)
N. _In [(XC/X.)D(X.,IXC,)..]
In Il"C,_1
If (x,/xc,) .. _ 10-
6
, then
In [(0 812/0 04) 0,719{(6_62 x 10 '
N.= In36 -3.4
We esllmale the mlDlmum number of trays by uSlOg Fenske's equation, which we
.... rite as
(8-120)
(8-12')
(8-122)
(8-123)
X D - 0.040 x....
CH
= 0.812
d. = 054 d
r
= II
ood
so lhat
.s4O LEVV. ... LIQUID SYSHloI
USing Eq 8-7910 estlmale lhe overall plate effktency (note thai we are not certam
that Jl-r ... 0.3 centipOisc at high pressures) rues
Then, from Gilhland's apprOlumallon
NT = 2N. - 6.8 (8-133)
We lei U
c
= 100 DIU/(hr f,l F). and we aSSUrTl<: thai AH
H
- 389 and AHol .-
3520 Btu mol lIence
(8-143)
(
'006 )( 11)[ 83,600 IbJ ( ''') Ann. Cost = - 8150- - S500/yr (8-145)
1000 gal 8.34lb 1(100 9O)hr yr
05
Eo = [03(36)]0 a '" 0276
so that Ihe tOlal number of required trays is about
6.86
N = 0.276 = 24.6
lind the column height is
Ii = 2(25) + 15., 65 rt
The vapor rate is
If = (R + I)D :::: (1033 + IXl + II + 0.54) ::::: 27.5
and lhe tower dIameter IS (for an ideal gas)
[
i15)J'" A = 0(XK)214 78(400 + 460\150 (27.5) = 0478
(8134)
(8-135)
(8.136)
(8-137)
(8-138)
and the cost is
(
792) ." (' 29) "600 Ann eml = 280 (1013X
4
18 ) -,- = iyr
Also, the cost of coohng water IS
STABILIZER RE80lLER AND SI"EAM. The reboiler heat balance is
QII ., 6Hp P "" VIIA
II
6T,. = IVsAH
We assume Ihat P = Vand AHp = so lhat
A = 13,300(27.5) = 32Htl
II 11.250
(0-144)
(8-146)

STABIUZER CONDENSER AND COOUf"oG WATER. A heat balance for the
condenser gives
Qc"" MI, V",. Vc-AcAT.. = W,C,(I_ - 90) (8-141)
Since we arc condensing the overhead at we cannot allow the temperalure of
the cooling water to increase form 90 to 120 ,- I for our firSI deSign we
Ihal ,. lOO"F. 1 hen
Unfortunately, this diameter is less than the allowable minimum value of
1.5 fl, so we should consider the possibility of using II packed tower. lIowever.
dearly we have based our design on a number of questionable assumptions. so at
some point we must undertake II more careful analysis. Nevertheless, our primary
focus at this time should be the processing costs of the total plant. ralher than an
accurale design of the stabilizer. and sino:: we do not expect that the stabilizer ",;11
be very expensive, we estimate the COSI
and the column diameter IS
f4A
D = .J---;; = O.8ft I ft
Ann. Cost = G:) (IOI.3X65
1
066Xlo 'Ol) (218 ; 1.15) = S5400/yr
(8-139)

and Ihe COSI IS
(
792) (' 29) Ann COSI = - (101 1)(32 SO 6') - = S3000/vr (8-148)
280 - 3
The bollmg pomt of the boltoms is about 400 F. so lIle u: 420-psla steam m,hc:
reboiler. Then. using the \alues in Sec. E.J. we find that the cost IS
(
'2 )[I,,3OO(27.5)lb]( b')
Ann. Cost.. -- - 8150 - - S1700/yr (8-149)
1000 Ib 775.6 hr yr
I EVEL-4 ECONOMIC POTENTIAL To calculate the economic at 1e\c:1
4. we subtract the annualized capllal costs of Ihe columns. conl1enscrs. and
rehollcrs. as well as the steam and costs. from \alucs of the levc:l-]
economic In additIon. we must subtract the purge losses of benzene and
toluellC (as \leU as any recycle losses. but we may talr.e fuel credit for the benzene
and tolurne in the purge and the stablhzer overhead). The results are shown in Fig.
7313. and we see that the range of the design variables corresponding to profitable
operation has been dramatically reduced
Of course. wilh a dllferr:nt alternative. the profitability might be
However. "e want to complele the design before ....e examine any
alternatlles.
AT ..."
.. In[(115
100 - 90
90)'(115 - 100))
19.6::::: 2O"F (1)-142)
Ltv.. ,...... ALTt'RNA11\ r.s. 1 here are numerous aiternatl\'es that we could con-
atlC'\'c!4 A vapor f('covery system nl1ght prove to he profitahle on ell her the
542 LEVIOL' """T NElll.Olt.
vapor (if we U1(;ur recycle losses) or Ihe purge Amorpllofl. condensa-
tion, adsorptIon, or a membrane: rooo\lery system could be used as this vapor
recovery proc:es.s
There are also numerous alternatl\cs for the disllllalion system. SlIlce the
reflux ratios arc so small for both the stabilizer and the recycle column, we might be
able to usc just suipplng columns., I e,. no rccufYlng so.:tlons. SImilarly, Slllce both
the CII.-benzene and the toluenc-dlphenyl SplHS are very easy. we should conSider
the posslbhty of uSlIlg pastcunzatlon seclions to decrease the number of columns,
However, we defer an attempt 10 evaluate the costs of Ihese alternatives until we
have evaluated the beat-eJlchanger networl for our base-case desIgn.
LEVEL 5: HEAT-EXCHANGER NETWORK
The procedure for desiglllog a network was presented in Chap 8.
To develop Ihe data we need for this procedure. we must calculate the lempcralure-
eDthalpy curves for each process stream. We would recommend that Dormally this
Information be developed wilh the use 0(.1 CAD package such as FlOWTRAN,
particularly for processes where miXlures exhibit a phase change (i.e., a set of flash
calculations is needed along the length of an ell:changer 10 calculate Ihe T-11
profile). Similarly, for cases where a pbase change lales place at hlgb pressures, illS
usually Simpler 10 use a CAD paclage 10 calculate the ph)"Stcal propcnies required
Ihan It is to use handbooks and emplflall corrclallons The use of FlOWTRAN
for lhese calculallons IS dlscussc:d in Chap 12
APPE DIX
C
DESIG
DATA
CI HVDROCARBON VAPOR-LIQUID
EQUILIBRIA"
USC Figs C.l-l and C.l-2 for compounds wllh lower boIling pomts than heptane,
and use Table C.I-l with the graphs (or compounds or cuts wilh higher boiling
points than heptane.
Ta1eo from 5 T Ibdden a.nd II G Cinl)'soll, -Ne,", Charll for lI)'droca,bon VlPOI_Uq.oo
Eq\llilbna. lIyJTi1ClUb PTDr' lJ'" "rlnJ' Rrf",,", 40 91, 201 (SrplnDbcr 19(1)
543
.
..
w
w
..

,
: I
: i
,
,

.,
."
..

"

..
-
-
....-
, .......
--

--
,. ........
---
...-.. ...'-
""...
...'_ ')0 ;0;," fOto_
.. co,. c,,,,,.-:;
-
...,
e, "' ...
"'''''","
.,......
.....'-
(0, ... ,
.,.....-

o 0 <I 1IO 10 00 100 .-.0

'........"'..,

e.'..'..I_....I'" -,..-r'-
, .. _ ....';'"4
,

/'

(1 i
0'< -1
j

W
,.
W
.W
.,
.,
5ECT'O...- (' 1 IlYO./X'.... IIUJ'l V.. POIl LIQUID EQIJIUII.JA 545
,.
,.
.w

,
.:,; l'
.




,
Oll'" ..;

-e

10 .,.:-::::...-=
lC
e

w'
00 I

w,
wI
.-
...s .. _
_......,
'\""
.-...-
-
-
-
.
.-
-
-
.
.- '_1_
... -1 ......
.--,"
-
I':::...
,
"
..

..
..
"

..,
-
-

-
-
.
-om'
.,
..
, ......... of
8

..

FlGURE
Var<>l-h'lU,d equ,hl>"a.. "0." 800 I 1 flodok.. a"d fI G !f',,,,,.,,,,,. 1(,,1'0<:,,,1> 1'.Qt", "",I
Prlt"l. Rrfi"". 01(1 107 (.vp,,.,.,J. 10(,1) J
FIGt Hr C.l 2
\'I.r<'"liqUld C<ju,lihlia. 2(,() l('J f- 100 I If'"..., S 7 lI"tIM.. "",J II (; GrM , ..... Ihthocmlo ,..rH; ""J
POI'''' Rtf"'''. 40 1m (S,,.,,.,"I,,. /9'\lll
544

TABLE C.1-1
VOI'lilily exponent rOT pure Ind true boilinJt point sooF eul$ al 5F Inltr".ls on
boiling point
Vohtilily UpotM'nl /I
Ttmp. Pun NBI', 5O"F CUI., THr rnidbtliling-poini ItmPf'tllu.1!l, OF
la,,,,.\,
'F
'00 JOO ... 100 >00 JOO ... .. ..
'00
00 0.360 0.&3. -0.081 0326 0.792 1.348 2..016
a"
"
0.381 0861 -0.061 0.)48 0.817 1.379 2053 2.843
10 0.003 0.4(13 0.818 -0.041 0)70 0.842 1410
'.C90
2.886
"
0.021 0.425 0.916 -0021 0,)92
,...
""
2..126 2..928
'"
0...
.", 0'"
-0.002 0414 un
""
2.912
"
0.0"
0.... -0.)76
COl! 0436 0.921
,""
2 202 3.014
'"
0.079 0491 -0,)56 0.038 0948 1536 2240 1m8
lS
00'"
0.514 -0.336 00S8 0481 0.97' l.569 2.2711 J 102
.. 0.117 0.537 -0.316 0.071
0""
iJXl2 , "2 n16 J 146
"
0.137 0.>60 -02%
01'"
o.m
"29
16J4 2)SS 3190
'"
0'"
0.5111 -0.276 0111 1.0S7 1668
UQ4
3.'"
lS 0.176 0.607 -0.257 o.m 0.$12 l.OBS 1.701 24J' 3.280
"
0.196 0.631 -0.231 0.159 0.596 !.In J.'H 2474 3,324
"
0,216 0.636 -0218 OlIO 0620 1.142 1,769 1514 3370
,.
0236 0680
_01<18
0>00 0... 1.170
'''''
B34 3413
"
02S6 0,., _O\1K
0221 0... 1199 l.lllB
3<6'
.,
0,277 0,730 -O.ISH 0242 0692 1228 1,874 26]5 3,SQ7
"
0.736 -0139 0262 0.716 1.2SB 1.908 1676 3,S33
90 0.)18 0.781 -0.120 0284 0742 l.288
,...
2,718 3600
"
0339 .108 -0.100 0.303 07M\ 1,)18
"10
27S9 3616
K
II
K,/(KdK ,)'
whcrt K
II
.. K "Iluc of thc cut or high.hOlllng [lurc hydrocArhon
K
1
K "Ilue of cthlnc II thc lcmpefllure. pressure and c:nn"crBcntt of lhe ayllcm
K," K "Ilue of /Iheplane Illhe. lemperllure.. pressure l/lod cOnycfJCnoc ptUlurc or lhc .Ylrem
/0, .. Vollnllly uponenllfllm TAhlt C,!.!
;; -


'::.'E.,t
... ;I ; ... -
;; "
" i li
i

t: t:: t: IS
o 0 0: S 0:

<

"
1

;;!>!
3
1<'

_ N

"

r.

e'
"

..
'" 7" ):;....
-; ...
;.
;;;- :z;
- 7
" >-
,... r- "'0
, '" "'
"
>
...
co
"
'"
"
>
Z
"
[i.

C
"

..
""

"
"

_.
-
-
,

:
,

-
,
>
,

,
"
'f:.

D.I CHEMICALS IN THE PUBLIC DATA

FILE
S"lf", d.o,,,I<:
Sulfur l"olld"
/II.!'fOp8Rol
T,undhylam,ne:
V,nylacclykr>e
Thiophene:
Melhac,ylonnrik
BUlad""m:s
Thmelhylacclykne
Elhylacelykne
l,2Rulad""".,
1.3 1l11lad",ne
RUlylene5
I-Ilulcne
(",,Ihulene
twbulcne
T.",.. lbulene
l.wbUl}faklehIOc
MFK
N-lluII'K and
Ethyl a",'a'e
Methyl
!'ropll [onnl<le
O,melh)'1 acelamide
!sobulanc
NBulane
IwbUlanol
""_Butanol
ToRtllyl alcohol
O... ,hyt dh.t
O""lhyle"" ,glycol
Furfural
!sep,ene and l"5
2-Me-tbutene
lMe_2_buleRe
JMe_l_bulene
CyciopenlaRc
IPcnlene
CiJ-2-p""lene
T."....2pcntene
D1etbyl kclnnc
",-P,opyl acelale
lwpenlane
..... Pentane
Conti""ed
C,ll.O
C,II.N
C.Il.
C,II.S
CII,N
C,ll.
e,ll.
C,II.
C,H.
C.H.
e,H.
C.I/.
C.II.
C.H,
C.ll.
C.II.0
C.II.0
C.H.O,
C.H.O,
C.Il.0,
C,II.0,
C.ll.NO
C.H..
C.II,.
C.II,.O
C.I\,.O
C.Il,.O
C.H,.O
C,II ,.0,
C,il.O,
C,II.
C,II
C,H,.
C,II,.
C,l1,.
C,II,.
C,II,.
C,II,.
<-",11,.0
(',11 0,
l",II"
C,II "
Chloroform

Fonnaldch}dc
MeThyl chlonde
Melhyl IOdIde
MeThane
MnhanoJ
Melh)'1 amHl"
lnchlo,,,,",hylcnc
O,chlmoalyl-CL
Attl!'k:04"

\ In' I chlonde
A<'el\l chlonde
lllTfI-CL-<:lhsne
ACl''''OIlnle
lene
I IDichk>.O"lhone
I,: D.chloJOelhaoe
ACO'laldehyde
Eth,lene
Aomc aCId
Melhyl rannale
chlonde
Elhane
Dimelhy1 ether
Elhnol
!lycoJ
Dimelhyl $ulfide
Elh) I mercaptan
Ethylam",,,,

Melhylattl}1ene
Propad ...nc
Propylene:
AttlOne
Eth)'IIOImale
Melhyllt<'elate
ProplOn", lIcld
D'mcl hylf(>rmam,de
Pr0r"'ne

DIp'''''' chemicals
Compo_". lUI"'"
0,5
NiH,,'" " ..de U,S

[mpi..... l I",mlli.
N,
N,O
0,
T.... BLE: 1).1-1
CliO,
COIN
CH,O
Cil,O
Cll,'
ell,
CH,a
CH.N
(,Hel,
C,HO,O
C,H,
C,H,C1,O
C,H,Cl
C,H,ClO
C,11,0.
C:I1, '"
C;!',
C;II.O,
(,H.O,
(,11,0
(,H.O
(,H,O,
C,H,O,
(,H,G
(,II.
C,H.O
(,11.0
C,H.O,
(,Il,S
C,H.S
C,H_N
C,Il,!"
(,B.
(,H,
C.H.
e,ll,o
C,Il.o,
C,!l.O,
C,II,O,
(,11.1'10
C,II.
C,ll,O
CompoM1ll Ita.....
Chlonnc
Hydrogen iodide
Ilyd,olen
Water
HydrogeR 5ulfitk
Amn",n'lt
Neon
NUllc
Nu.ogen dlOude
0,
HI
H,
11,0
",5
H,N
N,
NO
NO,
Empiriaol lonnul.
A,gon
R'ommc
Carbon lelradllonok
('a,bo"
PhosgMK
Caroon dio..de
Carbon d"ullide
Tnchl<>lilllttl )1..('L
lIyrlrog<'n chlonde
Com,.-.,I na_
The tables in this appendix are taken from J. D. Seader, W. D. Seider, and
A. C. Pauls, ..... trall IntroductIOn:' 2d ed., CAChE Corp.,
Cambridge. Mass. 1977. CopIes arc: available from Ulrich's Book Swre 549 E
University Ave" Ann Arbor, Mich., 48104. .
Physicial properties are available for the components in Table D.I-I.
TARLE 0.1-1
Chcmiuls in the Public Dala File
APPENDIX
D
FORTRAN
INPUT
FORMS
Empi..... l rormllb.
"
B,
=.
co
COCl,
co,
es,
CJ OC1.
0"
548
550 SE.. ,-,oN P I WLSII
TAbLE: D.I-I tCOIIl",ued)
..... __ ... __._'_u_- __
OJ "-FUll 551
occurs, lhen the appropriate product composition is set equal to the feed and the
other slream is set to zero. The block also calculales the rate of heat addilion
(positive) or removal (negatjyt.:) from lhe flash vessel in order to maintain the
specified temperature and pressure.
Physical Properties
Vapor-liquid equilibria and enthalpies are required.
OUlpUI
The outpul gl\'es Ihe unit name, the feed and product stream names, the flash
temperature and pressure, the heat added to the yessel, and Ihe fraction of feed
..... hich kayes as \'apor. An opllon is provided for printing stream flo..... s and physical
properties including eqUilibrium K-values.
C,U"
e.HjCl,
C.ll.O,
C.H.a,
C.H.O,
C.H,llr
C.H,Cl
C.U,!
C.H.
C.Il.O
C.H,N
C.H"
e.H"
e.H"
C.H..
C.H..
C.II,.
C.ll ,.
C.H..
c.u,.o.
C,li.
C,H.O
C,H,.
C,lI,.
C_II,.
C,H,.
C.H.
C.H,.
C.H,.
C.H,o
C.H,o
C.lI,.
C.H"
C.II ..
C.H,.
C.H
II
C.H,.O,
C.H.
c"H,o
Nto-pcnlanc
1.2.4 Tn-CL-8Z

O-DlChlolobcnune
PDlChlo,obcnunc
BromobcnleIlt
Chlorobcnzenc
lodobenum.
Benzene
Pbcnol
AnilIDe
Cydobcnne
Melhylcycloprnu.nc
l-Heunc
2.2 Dimethylbulane
- 2.3D,methylbutane
N-Hnal>t
2-Melhylprnlallt
3Melhylpcn,ant
Tnclhylcne glycol
Toluene
O-Crnol
MClhylcydoheune
Elhylcydoprnlane
l-Heplenc
N_Heplane
SIYlenc
Elhylbcnzc:ne
Xylenes and ETII
M-Xylene
OXylem.
P-Xylcne
N-PropylbclllCDC
Elhylcyclohennc
N-Propylcyclopcnra
N.{)cllne
Terraelhene glycol
Indent
Indan
C.",o
e.ll"
C,II,)
e,1l "
C.U,.
C,.U.
e,oll,o
C,.H,o
C,.Ii"
C'G
H
..
C,oll,.
C,oH,o
C,oU"
C"H,o
CIIH,o
C"Ha
C"H.
C"H,.
CuH"
CuH
u
C"H,.
CuU,.
C"H,.
Cuu,.
C.. Il,o
C,.H.
o
C"H"
C"H,.
C"U"
CnH,.
CnH,.
C,.""
C,.H..
C,.'i"
Methy"tyrene
C,jAlkylbcnzenc
l-El2-Me-bcnZ"l>t
N-Propylcydoht.llan
NNonalle
Naphthalene
J_Melhylind"oc
2Methylindene
DscydopcllladlCoc
N-llulylbcllulK
1)-D1MElETHBZ
NButykydoheuDc
NDcciJ.ne
J-Methyinaphlhlllcn
2-Meth) Inaphthalen
N-Undccane
Accnaphth)lcnc
Biphenyl
2..7[)mlCthylnaphlh
1,2.3-TR!M E-Indcnc
NDodccane
Fluorene
C,3-Alk)lnaphtha
""'

,h.
2..3,5-TRIM ENaph-
lh.
NTndccane
Phenanthrene
N_Tetradccane
l-Phenylindene
2-ElbylJluoreIlt
N-Penladcca.oc
Fluoranthene

J- Phenylna phlhalcn
N_Hendeeanc
Qrpene
Block List
List Iype
Umt name
Umt type
Name of feed stream
Name of liquid product sHearn
Name of vapor product stream
Parameier List
List tHI<:
Unit name
Index or first enlry
I. Flash temperature, OF
2. Flash pressure, psia
3. Prinl Slream flows, physical properties,
and K-values: 0 =no, I =yes,
D.3 AFLSI-I
BLOCK
IFLSH
PARAM
D.2 IFLSH
Description
IFLSH (Isothennal flash) determines the quantily and composition of liqUId and
vapor streams resulting when a reed stream is flashed al a specified temperature
and pressure. If the flash condillolls arc such that only a single phase product
Description
AfLSH (Adiabatic flash) determines the quantity and composition of liquid and
vapor streams resulting when up 10 seven feed streams are milled and flashed
adlabatlcally. The number of product streams may be I or 2. The block can be used
to simulate a pressure drop across a valve or through a pipeline. H IWO product
streams are specified and the flash conditions resull in a single phase, the
SECTlO,," OJ An,5H
appropriate product is.set equal to the sum of the feed streams and the other stream
is set to zero. You can also specify heat addition to or removal from the flash unit
Output
The output gives the unit name. the feed and product stream names. the flash
tempc:ralUre and pressure. the heal added 10 or remo\'ed from the system. and the
fraction of feed which leaves as vapor. An option is provided for printing stream
flows and physical properties including equilibrium Kvalues.
Properties Used
Vapor-liquid equilibria and entbalptes are required.
D.4 SEPR
Description
SEPR (Constant splitfraction separation) can be used to simulate distillation or
other separation processes when details of the process are unknown or irrelevant.
The basis of the model is that for each eomponent the TallO or the moles in the
overhead to the moles in the feed IS constant These split rractions. one for each
component. are input values The bolloms product may be assumed to be a
saturated liquid. and the o\erhead product may be either saturated liquid or
saturated vapor. Product temperalUres are then detennined from the saturation
requirement. The product temperatures may be alternauvel) set equal to the foed
temperature..
Used
Vapor-liquid equilibria and enthalpics are reqUired
Outptlt
The OUlput gi\'es the unil name, the I}-pe of condenser (total or partial) if one is
simulated, the stream names. the: tolal fraction of feed taken as O\'erhc:ad, and tbe
splil fraction for each component.
Block List
ust Iype
Unit name
Unit Iype
Name of lSi feed stream
Name or 2d feed sHearn or 0
Name of 3d feed stream or 0
Name of 4th feed stream or 0
'lame of 5th feed stream or 0
'lame of 61h feed stream or 0
'lame of 7th feed stream or 0
'lame: of liquid product stream
"'arne of vapor product stream or O
ParameIer List
List type
Unit name
Index of firsl entry
I. flash pressure. psla. if positive.
Flash pressure minus the minimum feed
pressure psia, if 0 or negative.
t Heal added, Btu/hr (if negative, heat removed).
I. Print stream flows, physical properties.
Kvalues: 0 - no. I _ yes.
If only one prodUd Slrum is lpec:lrled. ju phase oondillon will be dlennined
BLOCK
AFLSH
PARAM
Block List
List t)PC:
Unit name
Unit
Name of feed stream
Name of bottoms product stream
Name of overhead product stream
Parameter Ust
List type
Unit name
IndeJI of firsl entry
Splil fractions, moles j in overhead/moles j in feed
BLOCK
SEPR
PARAM
I. Componenl I
2. Component 2
3. Component 3
4. Component 4
5. Component 5
D.5 ADD
De:scriplion
ADD (stream addition) adds up to seven process streams and detennines the
produci stream temperalure and phase condJlion which satisfies Ihe condilion of
zero enlhalpy change. The pressure of the oUllet slream IS taken as lhe mlOimum of
lhe nonzero pressures oflbe IOlet streams less a specified pressure: drop. ThIs block
IS similar 10 the adlaballc flash block: AFLSH.
Output
The OUlput gives the unil name, the feed stream oames and produci stream name,
temperature, and pressure. An option is provided for printing Stream flows and
phySIcal properlies IOduding equilibrium K-values.
Properties Used
Vapor-liquid equilibria and enlhalples are used.
J. dP, pressure drop, psi (sublraclc:d from
mlOlffium IOlel pressure)
2.. Ponl stream flows, properlles, and K-values:
o ... 110, I = yes_
0.6 SPLIT
Descriplion
SPLIT (stream spill) separates an input stream into as many as seven o.utput
streams, each having Ibe same compositlon, temperalure, and pressure as the
You must specify the amounl of each produci stream as a fraction of the
stream If the sum of the fractions does not equal unllY, the fractions Will be:
normalized_ This makes It possible for a gh'en fraction to serve as a mampu1ated
variable 10 a conlrolloop
Output
The OUlput gives the umt name:, Ihe IOput and outpul stream names, and the
normalized spltt fracuons.
Properlies
are used
Block List
List type:
Uoil name
Unit type
Name of lst feed stream
Name of 2d feed stream
Name of 3d feed stream or 0
Name of 41h feed Slream or 0
Name of 5th feed stream or 0
Name of 6th feed stream or 0
Name of 71h feed S1re3m or 0
Name of product stream
Parameler Ust
List type:
Uoil name
Tlldex of fint entry
BLOCK
ADD
PARAM
Block List
List type
Urnt name
Unu Iype
Name of mput stream
Kame of 1st oulput stn::am
Name or 2d output stream
Name of 3d output stream or 0
-arne of 4th output stream or 0
Name of Sib output stream or 0
Name of 6th output stream or 0
Name of 71h outpul stream or 0
Parameter List

Umt name:
lode). or first entry
BLOCK
SPLIT
PARAM
f!\RAM
I

l'olYlropic clflCtt'oc)'
EnlhaJpy eh.nMc r mOll!
(whh 7opfllOPfiliC 'III ItS)
100011
"." r(r,).-., ]
AJt= . - -J
.... - ,. p
-'. .
ibp.- w
blip - ill II
,.
0.8 G O.\1P
Par. met"" List
type
Unit name
I"dt" flf ri'$1 enll)'
I. Oudel rrc.'msre. Nl'"
I. POlyu(lflic compl\".,siol)
K = C
,

c.,

GCOMP (ttQtrifUlal CI.'mpfCS50f, ..1t compICS!iQr. Olnd lUrhlne)
computes the ...orlo: rtquhed lor e:umpleuion Of 1M W()f" a.nd
tbe oull Slf\::im tcmpcnllurt. phase ooodition. eOO enthalpy
TIle bioct \lIiII pe,fl1fnl rOu, differenl oJ
I. (,entrifufal compressor us!nJ!. lhe polY'Ulric dftlXllcy. cnlkd the N
_,W
Z. rOJIII\'('.dh"lacemenl CUl1lllle5..'OOr usittt 1he poIYHt?Pic CR"JcicllC)'
J. TurbilK (I' Cltpandn usinilihe isent,opic.! cfficJell4.:Y, also allw lbe ilJfabtlllc
800 lhe M"tlltr /f'Ur/lQd
4, Cenlnfll,al d'Jfflpteswr lhe i!etlUopic c:ffKtr.ncy
Wilh eaeb DlC:lh,lll. )'OU mo)' :pc..;Jf)' lhe apptUllllare cfflC..erlor; nil nullet "IICliIO
aDd shaft wbrt. will be oompulM I h;.wcycr-. If the olltlcl lemreral\Jfe. ft
koo'A'n. it may Ixelltered for 1he: p".ltmeIN, amllhe prPltnlm will oompulc
1M efflCicnc}. This option useful for an c)tisring compressor
The t;lllc\lbtion must nol Ilt: fOr lIlJtamli thnl are r:nlilllly
liquc6t.d at iolel ur oUlkt coodiliol'l$,
The. c:qUltLiOnli used afC u 'nlluWlI I'here the subscript I rders to Wei
and subscnpt '2 r(fCN to OUlSet conditions:
I
BLOCK
I' Mr
0.7 PUMP
Deteription
PUMP (cenlrifu(t<1J pump size and (lOwer) , .isc& tbe pressure of a suum to
rpecirlCd value:. It should be' used anly ror iii IUlall)' liqllid TypiCllI Qlrva
ru)m lhe lilcnlture tile Ie) (1tlrnp and dri"n etrK:Xooes so Ih31 driwI
$1lC' and electric po",-e, n:quiremc:nls I;.iln be ClllculalM_ Thcf< t'llf"\'C$ are hll:\Cd nn
wale,: pum dficieocy be for morc "iscous Ouids,
OUIP'lI
The QulpUI g1VIS ullil nlune. lhe inJcl :lnd llullcllllu::rm name.\., tiM! aClu:t1 Oow
In @:pm. prcS$ure chtnl-C in ... nmd bOfSC'pov.'Cr. ptlmp effacleoo. bnlke
power, dri"e:f elncic:ncy. and kilowau..,

M. S. PeleB. Nunl Dws;gn and (;hmrkm Eng;n"",
New York, p, '29J.
BJock Llo!lc
List ITpe
Utili ftll.mc!
Unil lype
r"lan1l.' of (ttd ,IM(:lm
N"MC (If plodUCI
I. Number or output slreanu (2 (0 1)
L Split {or ht output
3. Spilt r,..ction (OJ 2d output Itee.m
.. Split fl1lelion leu 3d oulpul ,tream
Spill for 4th output stream
6. Splil hattinn ror 511'1 Qulpul ,,-lff'./Im
7. SpUI rfaeUon ror 6th O"'pl.ll jlreJlm
IJ. Sphl fraction for 7th OUlp'll $IIQrn

Pr.p<r'!<s UII<d
Uguid lJlClW voJunw is uted,
2. Iselilropic
SECTION 1)9 scvw .559
OIock List
3. Iscnlroplc expansion
h
_ h, - h,
,---
,.
List type
Unil name
Unit Iype
:"lame of JIllel
Name of oUllet stream
BLOCK
GCOMP
4. Volumctnc cfficlcnC)
", = 10-0.01 Pl-t 1
J
_ VI)
PI V
l
5. Outlet temperature 1$ from the outlet enthalpy;
h
l
= h, + "'h
OUlpU'
outpul consIsts of the: tnlel and outlet stream names, discharge pressure,
actual and constanl-entropy discharge temperatures, constanl-entropy horse-
power. horsepowcr. and brake horsepov.er. Work done by the flUid is
negalive. If Ihe posllive-displaecment-compression option is chosen. the OUlPUt
also includes the \olumetnc effiCiency and the required displacement in ftllbr.
Properlies
Entropies, enthalples. speC1fie heats, specific \olumes, and \apor-hquid equilibria
arc used.
Nomeoclature
Parameter List
LISt type
Unit name
Inde:x of first entry
I Outlet pressure, psia
II. Outlet temperature,F (computed if 0)
III. Type of unit
1 Centrifugal compressor by polylropic method
2 Positive-dlsplaecment compressor by polytropic method
3 TurbIne by isentropic method
4 CentrifugaJ compressor by Iscntroplc method
IV. Ckarantt fraction for posillve-displaccmcnt compressor
V efficiency (I if 0)
\'1 Polytropic effiCIency for options 1 and 2;
isentropic effiCIency for options 3 and 4
(0.71 If 0)
PARAM
C CJearance fraclion
C, Heal capacity at constant pressure
C. 1leat capacll)' Sl constaD( volume
h. Enthalpy on cooslanH:nlropy path
n Polytropic exponent
w Siream flow
". Mechanical efflClCnC)'
'" Polytropic efficiency
'I. lsenlroplc efficiency
fl. Volumetric efficiency
References
Dresser Industries, Inc.. Clark Mlii/israge Ci'ntrijllgol Compressors. Franklm
Park. Ill., 1969.
Ellioll Company, E/lioll Compressor Ri'fresher. Jellnnelle, Pa
D.9 SCVW
Description
SCVW (bounded Wegstein stream convergence) is a stream convergence block
..... hich is capable of converging one. IWO, or three streams simultaneously. The
accelerated con\ergena: method of Wegslein with a modification by Kleisch and
Su!li\<an is used (sec References).
The followlOg options are mcludcd:
I. The convergence tolerance may be specified. The same value is used in each
stream for every variable except enthalpy and fraction vapor. If a value is not
specified,0.0005 is chosen (see parameter 2). For each variable in a stream, the
error is defined as the smaller of either lhe difference between the calculated
(inpul) and eslimated (output) values or the dilferences divided by the estimated
value Errors which exceed the convergence tolerance are indicated In the
history by an asterisk The first N variables m the stream are the molar now
Outpllt
There' is none.
BlO K
scvw
Bind: USI
List type
Unit OiJUC
Unillrpc
X.mc or first input $utam
N-amc orstClOOd input SlIOlIll orO
ame ollMrd inptll Slfe<lm or 0
TntJufc. pOinl (unit ";lme or dlC' 6nt bkH:l; III tile rcc)'Cle

or fiNoI oulput $llCflllO
Name (l(S(lCrmd output $ITeam (,Ir II
Name mthird (JUlpul WCifm or 0
Rdeft'f1Cf'S
H. C. Kkisch.... A Stud)' of A(;C('1e.3IQI AIg.orilhmJ I)seJ in StcDd)'
State Process Silllulltion." Master's Thes.'h.. T\J!.;11\e Ullhrnity. Ne'" orleans. 1967.
J. 1-1, WeJ;$tf5n. .. Aocelcf:'ling ConYClvnl"e 01 lleJltion 1'(OcoseS." Comm, A ,M.
1: 9
Used
ud .... pof-equilibtl3 dala :m' needed II 1I.e uUlput SIJCllm I!C n..,hC'd.
Set: panuneteTtI fila II. )'ou yt\ur ')'$Icm is c).ccvtionaUy suabk (or
unstable). you should the ddault ,al",C$ 01 Q.l.. - - j.o Mild Q..... - 0.0. If the
.solUlinn I, found 10 u&(:lItale or dh'elgt'. 01)1: or both Of Illest' \'/llues IIho",kJ be
lncre;rSffl. Uowt'\'u. if the J)'stcm h Jlabk, dCCfe,nlna Q..... and(or
Q_. rna)' kivc oonvusencc in Pet helations
If Q.III and Q.... are: both nonnegarivc, IMlt: \Io'ill be no o-ecdcfalion.
c(lnllergcnct being equi\'aleotIQ dire:ct itelation with d.ampingin (be ,anre betwcm
Q.'II and Q,_., Fluther. unlhS II. is .,.... (i'lnlC'whnllcs-Slh.n tn.lhe: block will
rc:re:uedly lJ'y the ,.arne \'llh/IC of)( , (and thus 3PrCal 10 h:t\c Selling
<l.... < 1.0 forces lh4: bloc:lc to slW*ys a 1.0 - Q_. friiClioa or the
ile:'8ti.oo step.
In addilion 10 ehantin@ Q..... ;'lnd Q...... can be IflRueoctd by
changins the il)itial number 01 direCl i1rratiom. 6. and tlte anH)unl.o'
damp''''' parameter 7. In gtOl'tat this will also decruse Illt t.te of C()n\'tfg.ence

Slop< =!('J - !l.r._,)

x. -
.t.il .. .l.. +(1-'1.)/('.)
In Ihe method, In ..deration q. Q. 4:Alcut.:.ted lind
used to weigh! lhe pre,'ioos s1rl:am cstiltUltes to oblain a nc..... e:sumatc. nw-
I;u:lor b dclintd IIstbe.slope by the slop:minU$ I. 11If:: slort: is
df'tnmilled fbr e:tch variable in (rom
,.,'here the: y.. .",en to the IOF ptlrticul"r !>lR'... mYamlblc.lbla<;k
:'Ifldj( l._.) IS the e"rl'C$.pt1ndina "Dlue alcul;dtd (.n," Ihe prQCt:s\ tblock
llIpul), The: valuCJ _v. 8Iklf(\'J are the pair (\1 inpul aoo It)pOnse stream
\'.11.1"- Smcc t....O \,:tluq. uf (x) lne nceded to alculJllc q., :It one diu:\-"!
itet"tJQn must be. malk I}cfoFe ifOlXkr.. tiQn CllIn be: applied
Ihc initial ditct1 itcrations. die new estimate for t:ath strc:llJn
y.riilblc 1!4 c-.kut:fle:!J rtom -
rlIttll or ClIlcll or dlt .v eOlnp6MnB, tI + I ill the lotal ntul:u Rf)w IlUe. N + 2
lemperalurc, N.. i$ pressure, N + .c IS the t:f1th.IPT noW', and N :.. SIs the
rnu.'tion vllp<'lr
1: TIle .Ul&lIirminl Ilwilber or ilennionJ lUlly be ipttified. If 11K' number nt)1
spccdied. 30 ilerltUOllli art automatically thosen (see panllnele. J). The bloc,"
DUly operate as a direc.1 tlentioo bJoc:k by lhe number or ditc:
10 be cqlJ1lllo tbe- mAXimum numM. 01 ile:ratioM. Tbe
di.1Xl 1!oahons Ulay be cbmped by a OOnttllnl lactl)l (pu/lmetu 1). lIr YllIri.. bk
damping mil)' Aprlkd. .. eap'flined bd(tw
J. YOll C:IO continue Ihc 51lnul,Ulon CYC" the mvamum numtx-r ol
IIc.atH>ni Ia.a.'l bn IcsdlC'd. Jr )'01.1 dl), lI)e I:b"t C$timlfle nf lhe output ,t.hbll1.t
and a message indiealing that Ibc simulation comioued I';th uuconvergtod
outpul art pnnled in Ihe bisto1). Set' paramder 4-
.. You \l;helhcr Ihe OUlput stream is to be Oasllc:d. You may 001 WIsh
10 Oa!Jllllc $ue:am if lllett' I:S a possibilily of llm::c phases e:... or ir a heat
b3l1lnce: not or S<'C' pllr1lmetn S.
If q. :' O. theQ)kulnlloo IS (hreClller.. tion Iro < q. < I, illS.dlfCt'1 itctlfliQn ...ilh
damptng......f\lk 1/. < 0 applies ae:k(lliioll. yaluC$ or q. g.i\'e >10..... MaiM
oonve:rgencc: nC'galJ\l: cnn con,sidnobf)' Jfl('Cd up bul int'fcuc
lbe likelihuod "f
T IrluJ(o-off t.ell''ftfl n"d ,,'.. hllil)'. eh q. Im\llcd
co Ihe l'anllP
561 SECTION I) II O5TW1J
SEc:nON I) 10 pnwu 563
All pan.....,tcn for IbIS blod, a.c opt"....1 Default mlJoCl; a.., 10 and un.pc<:tJied
1"n>c:tCn.
Description
DSTWU (shoncut disliUalion) uses the Winn-Underwood method (sec Refer-
ences) 10 calculale Ihe overhead and holloms Slreams for a slOgie feed dislillalion
column with either a tOlal or a panial condenser; tbal is, the overhead product is
cieher saturated hquid or saturaled vapor. The desired column perfonnanoe is
specified as splits for lighl- and heavy-key components. The GiUiland correlation
(sec References) is used 10 cakulale eilher Ihe actual rdlux ratio for a given number
of trays or Ihe aClual number of lrays for a specified reflux ratio. The reflux TallO
may be given as either a fixed value or a mulliple of the minimum reflux. If Ihe
specified reflux ratio or number of trays is less than Ihe mlDtmum that will produce
the desired separation, Iwice the minimum is used.
Properlies Used
Vapor-liquid equilibria and enthalpies arc employed,
DSTWU
PARAM
BLOCK
List 1)"JlC
Vml Dame
Index of first enu'Y
I. Heavy-key component number
2. Lighl-key component number
3. Split for light key, distillate mole fraction!
bottoms mole fraction
4. Spilt for heavy key, bouoms mole fraction!
dlsllllate mole fraction
Parameler List
List type
Unit name
Unil type
Name of feed
Name of houoms produce
Name of overhead product
Block Lisl
References
F. W. WlDn. Perrot. Rrfiner, 31 (5): 216 (1958)
A. J V Underwood, eM"'- Eng. Progr.. 44 (8): 603 (1948).
E R. Gilliland and C S. Robinson, Elemenuof FrM'ional Dlstillotion,4th ed,
McGraw-Hl1l, ew York, 19SO, p. 347.
Outptll
Mmimum reflux ratIO, mmlmum number of trays, IClual reflux ratiO, actual
number of trays, number of trays in rcclifying sectlOD for optimum feed tray
location, and condenser and reOOiJer tempera lUres are output_
PARAM
Parameier- Lisl
Lin type
Unit name
Index of firsl entry
D.1O DSTWU
l. Print the q vcclor aner each Iteration
o"" no, I ::::: )'/:S.
2. Comergencc lolcrantt (0.0005 if 0).
3. Maximum number of ilerations (30 if 0).
4. AClion if unconverged after the maximum number
of iterations: 0 z= lenninale simulation,
I = continue with last estimate of OUlpul streams.
5. Flash outpul slreams: 0 = yes. I _ no
6. Number of dlTCCI iteralions to be perfonned before
acccleralioD is applied (I if 0).
7. Damping facior for dirccl iterations (0 = no damplUg).
8. Code for delennlnmg upper hmil on q, Q_.
a. If 0. Q... IS SCi 10 O.
b. If I. Q_. is sct by parameter 9
9. Maximum value of q, Q_. (sec parameler 8).
10. Code for delennining lower hmil on q, Q......
a If 0, Q__ IS set to -5.
b. If I, Q..;_ IS sel by parameler II.
11. Minimum value of q, parameler 10).
1>.11 Rt:A<;r
, Ttrw 6.<N qll.lit," the ""',*,11 01 bul MlrlIo:I 0 "I'ltrln Ilit d M,.,.( to,. It.>
..IM fc'i'f<:,cno: <k- I1lII 1I11"tW"J'Oinl ""Ih.lpe..1 (l( Iht kt<1.
I rr Ipml. , .... ,lni, .... ".ft...."... 1l1oi dmt", n_IlII1" (ltt'l)1o .ue tllttl ,... dultN ,ftl... i\
IllQttoi II l!l(t .,# tl<>th lIlIllI, lk 8'\OClI do: ,dhI. Iv, I""U ",ioom,,,, ,,"""'" ('I
UJt.
OUlptrl
GIro' prinlcd in lhe hu'lory arc u
WARNINO RE/iCnON x CONVERSION RljSI;;T"'O = V
If Jl!iSel II) 0 and )'OU specify if con",n$ion '01 leaction Xhi&Jtc:r Ihlln the
muimum block uses the: m:uimulll Yand continue,.
REACTION X EXCI!WS M..\ J"OSSIBLI! CONY Of V
If J issei to I ,nd ,)'Ou spe<:if)' 3conversion 'or t'Cllctioo XhigMr 1han the
muimum V,lhe slmllialion is tenmnated.
REACTION X KE'" COM_I' FLOW. 0
f(Jf ,uclioo X Dow r.ale kn Ihe key OQmpooent " Uln. The
simulatit;m OQnllnut9 011 10 the nut rtSfcIM>n in the itqUCDt'e.
FEED STREAM IS ZERO
AU 'ec<U It) lilt' bkd: ate zero The m\IlalK>n eonlinuell .'ilb !he' rtCll blOCk.
OUlput SUt3mS are $(,1 to lClQ"
ERROR IN MFlSH
The prOIt'IIn1 I:arfllOI fjJ)Jj Ibe enlhlllrr 0' lhe ail meams. Simulation I.
Icmrinlted_
MAX NO. REA :nONS EQU"L
Simulilton is terminatttl
KEY COMI' COEF I'OSITIVE OR ZERO
Kt) eOfTlf")flCfll ct'ocmcitul nU1S1 be SimulatiOn it IC'ltllnalaJ
uUlpoll Or!he btock IUnpCo 1M ItlK.'IIOII Jln4
-tnt' Iht nalllc.'C ", Ihe input and tlulpUI !I,ca,n'll
Tile ml.almum numl;lcr ul ,ea-etlons Ihnl elln "<C. MCHC Ihllli 'OUI

re",ettolU ('41n be h"odled by plac:iltJ reactOr bloekll jn series. Up 1054:''(0 feed$ ClIO
be h:lnJkd by Il.EACT whM:h adds (he component Row ",let i.l\ one I'd 10 the
.nClOI 0 belt of rr....ClJon is calculated (or the tbe reactof Itmpel"lu'e
and p,essule must be spcr;ilic:d. The cmue(1I is ftllsbed iolo Hquid and vapor
prodllclS 11.1 lhe Icmper:Hule !tnd p.cuure 111"-"(1 rn:lduclj spodfard
The output ciln he 3 tillgle
1'lUlIrnele.t J is used 10 dclcillunc the blocl Iletton whetl lhe sptriraW
col\\trsion ls "rClIler th.n Ihe maximum po"sible oom'elsiol\ (when all we rellClll'll
it. deplclcd). A....Ine or Ue'tllto(S ItI<'. bk>cJ: to !($C! Ihe 10 Ibe fNuJmum
possibk and u value or I causes llle-simufatM'lt1 10 lermioste wilh an eflor
Ill""""
The parame-ter hslJ allow 2S l,'OI)tJ)CKlltnu: '0' each of lhe 'our ,eactions.
Stoktnomenlc eoefflcicllu ro' conlrnnent, nO! ,n a luction should be 7.Cro,
Componenl$ DtC nunlberw f,qm t 10 1S tn Ihe l)ukr in which Ihey
oor:ur in the comrollenl list.
I

lIT
FEED
DeuripcHln
REACT {themlc;lt is II whkb COtllputt.t. the ('ompOl;IJon nf lin
strum fcom n reactor. The first reaction use5lhe cO",ponenl Row$ found in
the Ittd SHellm. A 5t'eOnd ("C.IIciiOD lloolJld usc lbe ftnaJ rtaCtion mass 01 the Grt
rcactlon IU. feed $1ream. This procedure is applied in a rimil;ar rnllOnne, to tbe thred
.',Id (ourlh ""ction,. The cuet compu18lion is as (O(lOWl:
5. Q\,.I'ly of Iced (<<unpuled if 0)
Ii. Dc:slrC'd rcBul nll'O if pofitiyc: 11 Ie" ell.n -I. tbe
,dlu.1 181io u 1p(6}1lilhd lilt minImum rc:nUA
7. Oc$irtd number or nays I
. Top pU:lOiure,
9. Botwnl PCr5Sule. Ii.
10. CQndelucr ttre: 0 - 11)1:.l, t - parllJlI
ROUT, -- - .,. )(' __)
whc;re i is Ihe Ith component '1'1 lhe alld
FEED. ... a,mount or'th oompolltl'll io Id stream
ROUT. = amount of lth oonlpOllcnl in emoenl Sinaln
PE.E.O..,. IImount <.rltey oompoocot io rced
key .,.. key componrnl (8 (ClldalllJ
CONY frllclu:mal COn\'crflon (lr key component
COE.Fj = nlolat sfoichiom('ltle eceffideol of ; (or the
IIJI prndtJdll, nCI"t,ve 1m- reac:lllntf
COEF........ Molar sloichiomeuic coeffiMt or key oompol'lenc '<It (CkhO"
(1lIw'lly:S ncltlJ1h't, - I)
566 SRTKll< 0 II nAC1
PrOIH"r.ies
I-.nthalples an: used.
B1ocL.
LISt tyJ)C
Unit namt
Unit type
arne of lsi feed stream
Name of 2d feed stream or 0
Name of 3d feed stream or 0
Name of 4th feed stream or 0
Name of 5th feed stream or 0
Name of 6th feed stream or 0
Name of 7th feed stream or 0
Name of liquid product stream or singh: output
Name of vapor product stream or 0
Paramc(cr List
List tyJ)C
Unit name
Index of first cOlr)
I. Temperature of reactor, of
2. Pressure of reactor, psia
J. Terminate simulauon if specified conversion exceeds
maximum conversion: 0 = no, I = ycs.
... Number of reactions..
Data for 1st reaCllon
S. Key component number
6. Fractional conversion for ley component
7. StOIchiometric coefficient 1st component-
8. Stoichiometric coefficient 2d component
9. StoichIOmetric coefficient 3d component
10. SlOichiometric coefficient 4th component
II. Stoichiometric coefficient 5th component
('omponenl order 151M Older In .. lhe)' OC<:111 In lhe componc:ol IIsl
BLOCK
REACT
PARAM
12. StOIchIOmetriC coefficient 6th component
31. Stolduomelnc codficlent 25th component
Data for 211 reaction
32. Key compolKnl number
33. Fractlonal comersion for le)' componenl
34. SloichlOmetnC coeffiCient Jst component
58. Stolchlometnc coefficIent 25th component
Data for 3d reaction
59_ Key component number
60. Fractional conversion for key component
61. Stoichiometric coefficient 1st component
83. StOIchIOmetric coefliClelll 25th (;omponent
Data for 4th reaction
86. Key component number
87. Fractional com'euion for key component
88. Stoichiometric coefficient 151 (;omponent
112. StOlchlomctnc coeffiCient 25th component
SECTION D,I ..UCT 367
APPENDIX
E
COST
DATA
E.I OPERATING COSTS
Otemteals
Jbe costs of raw millrnals. produ(;'IS. and by-products can normally be found in the
:::n,mical Marluflflg Reporter. The values lisled arc the current market prices,
Nhich may be: significantly different from the price used in a particular company
xcause of long-term contracts. The costs of light gases usually arc not lisled in the
'::hernical Mar/Wing Reporlu because these materials often arc sold "over the
"coee" (3 vendor builds a special plant to produce these materials which is located
leX! to the site Ihal will use them) or a long-term contract is negotiated.
Ulililies
rhe best way to estimate the cost of utilities is 10 relate Ihe costs of any utility to its
fuel value by using thermodynamics and Iypical efficiencies of (}Ower
:llants, turbines. boikrs., ciC. Market fluctuations might occur al times which makC'
.he valuC' of steam less Ihan Ihal orrue1, but large cost penahles can be eocounlered
568
TABLE .11
UtililM5 cosls
Ulilil)'

Pritt
ruei (0,1 Of IU) 10 S4.00/10
f
Blu
Slum
600 I'l"I at 7WF UO 1'5..20/1000 Ib
SlI1uratcd $ttam
.......
'"
..
-
,"' .... 0"
H2
ISO f"S'I
0"
H
SO PSII 0.'"
" IS PS'.
057
2"
E\eclrlat)' 10 SllO<lj1<whr
Coolin. walt.
07'
SO.OJ/looo pI
If a IS based on dislorted prK:cs and Iheo the costs fC\'ert 10 1hC'lr normat
pauern.
A reasonable sel of factors 10 use IS given in Table E.I-l. Once the value of
fuel has been specified, Ihe costs or the other utilities can easily be calcut:ncd. Note
Ihallhe values in Table EI-J were not used IhroughoUllhis le"t Similarly,
IhC' costs used in different probkms arC' solllC'times different Howe\C'f. Ihe cos's
used in \arious problems arC' IdC'ntified as Ihe solutIon is dC've1oped.
.2 SUMMARY OF COST CORRELATIONS
The 19705 have been a period of rapId cost escalllilon (see Fig 2.2-1 J), and so vC'r)'
feu cosl correlations were published during this period. We use Guthrie's cost
conelalions in this lext, whene\'er possible, 10 iIIuslrate costing procedures, bul
DOle thai Ihesc correia lions arc oul of dale. We updale Ihe correlations from the
mid-I968 valucs by using a ratio of the M&S indices, but this is not II
rocommC'oded practice: for such a long lime span. Instead, if an updaled sel of
company cost correlalions is not available. a designer should consult onC' or more
vC'odors C'arly io the cosling procedure 10 obtain more recent cost data
For our preliminary process designs, we use a simplified version ofGulhrie's
correlalions, The nonnal material (the base costs assume carbon steel) and
pressurC' correction factors are used to C'$timale the purchased cost, bUI the most
collSC'o'ari\'e base module cost factor is used to estimate the installed costs. This
approrimation corresponds to a conservative cost estimate.. For more accurale
estimalCS" Guthrie's book should be consulled,'
K. M vu(hr"". M uplla! Cost E5.lIl1U1ull",-c,,- 76(6) 114 (Mafch 24, 19(9)
K. M G\lIhrir. PrOJ.J PI<IIfl ur...or.., l ....lllllt_ Mi Contro(, Craflsman Boot Co. Solana 9cach,
Caltf. 1974
370 SECTION I!.I SUlollolAn Of COST ronELAnoNS st,cnuN "-I SUIolMU\ 0'" COSI Cl.lklHAlk)NS 571
Process Furn.l6ces
Mid1968 cost, box or A-frame construc!lon with multiple lube banks, field-
erected.
(
'1&5)
Purchased Cost, S = 280 (5.52 x
when; Q = adsorbed duty, 10" Blufhr, 20 < Q <]00
F,,,=,f.+F.. t F,
Installed Cost, S = x 1O)Qo 8)(1.27 + FJ
Oirecl-Fired Ilealer.!>
cost, cylindrical field cra:;tioll
Purchased $ = X IO
l
)Qo
where (! IHborbed dUly, 10" Hlu/hr, 1 < Q < 30
f',=F.-if... t/<"
1,000
100
FIGURE 1:.2-2
Direct-fired healer. [X M. Gu,Ar/#!,
Eng., 76(6) JJ4 (Mwrh N, /969).]
100 30 50
t()6 BlUthr
] 5 10
Absorbed dUlY.
.
..-
I
ttttI
..-
5
I
,
I
1
3
10
50
30

FIGURE: 1':.1-1
PrOCCS!i rurnao::s. [K M GUlllr/#!,
CIonn. Eng; 16(6) 114 (MlJull If.
1969}.)
30 SO 100 JOO 500 1,000
Absorbed dUly, 1()6 BlUthr
..-
J:::
..-
1.--
I UJjJ
10
JO
50
30
500
300
100

TA8LE E..l_l
CorrecliOD faclors F< for Pf"0ces5 furnace
Jla;ip ,.._re,
Desip 1Yl'" F. bdiu! .... Ien.t F.
""
F.
P,QOe$S heater UXl Carbon m,el 00 000
110 Chrome/moly 035 1000 0.10
Rc:ronnc:r (no c'llalys,) US StalDless 075
"00
0.11
2000 025
2500 0<0
JOOO 060
TABLE E.2-2
Correction ractors F, (or direct-fired healers
Oesip pr_e.
F, Rad;"nt tu!loe malen.1
".
""
F.
Cyhndncal
'"
Carbon Sl[ 00 Up to 100 0.00
Dowtbcrm 133 (brome moly
"5 '000
."
0>0
"00
020
lIeat
and lube. compkte fabricalion
Gu Compressors
Mld-1968 cost. centrifugal mach me, mOlor drive, base plale and coupling
(
"&5\
Purchased Cost. S - 280 rI0l_JA06SF,)
.-t arl"a fl '; 2f10 <': 4 < 5000
fl. j
Shell-andTubc 1alcnal = F.
Purchased S".., IlF,
",here bhp brake m< bhp < 10.000
F, = F.
(
"&5}
Inslallw ('OSI, S - 280 'i17<i)(bhp)ou(211
100
Surfaor a,u.
" "
<S, <>
" "

'"
,
ro'
'"
Bra" "'0
SS SS Mond
"'>md
,
"
I ()"Yl In "OOn
'00
".
'"
",
11'i no
'"
19'i 11 (l'i
("&5) Cost. S = 280 1011A
06s
(229 + r,)

'.000

"
500
c
0

300
/'
c

" 100
c

so
o
t 30
,
"

U 30 SO 70
Bake horscpowe:r. bhp
n(a:R[ J...4
Process Pi COUlprtSSOn and d .. (K At G"'N", a.m.. fj,q_ 76(6)- I U (M.......' 14. .'9I'tfJ)]
FlGURE: E..l-J
Sbe:U-arod-llibe beal (Ao U
GIIIMw, 0-.. 16(6)- 1/. (Uoo-d1
U./969).]
JOO 500 1.000 3,000 10.<XX>
Surface area (calculated), ff2
100
<l
0
..-
,
0
/1
S
3
,
,
1
,
,
TABLE: E.l-J
TABLE E..2-4
Correction f."lors for heal c\':chanli:eI"S
Correction facfOf'S for
Compressol"!i

0,
..
0,
0e0p 'y,.. F,
"-"'10<
KClIk. robo,kr III Uplol50 000
('nllnfllpl. motor 1.00
FloaIHlI! head
""
100 000
RCC1proall,ng. steam
00'
U-Ilibe
0"
.." 02'
turbrne
'"
lube
0""
""
0"
.reap molm
'"
''''''
'"
S'"
'"
574 SHTION U SlJIoolM... n Of" con COUli ... no...s
E-l SU.uo..... n or COST COIlEU.noNS 575
Distillation Column Truys and Inlernab
(
M&S\,
Installed Cost, S = 280 j.7D
I
HHF,
."
'"
'"
.,.
225
'62
100
'00
TASU:U5
Correclion faClOrs for pressurr 'tssel;,
(
M&..S\ 90' o""lIu uf)
Purl:haJ.ed COSI. S = 21) r 101
\\here I) = dliJrncter. rt
II = helghl. rt
F. = r.f.
Pressure VcsSt'h., Columns, Reaclor.,
""I"'"
..,
soo
"'"
'00
'"" "'"
""00 f>,U>ule L'p '0 j() '00
'"
115
'20
IH
'"
,so
I."
190 230 250
F. '00
Installed \0;,1 S = (M&S)IOI 91)' o"Ho IOJ(2.18 + F.)
2S0
where D "" diameter, rl
H "" tray slack heighl, (t (24-10 spacmg)
F, = F. + f, -f F.
FlGUR.E .2...
DwdbuOClcoIWllGUlIY'LA '" G"I/r
C"- 76(6) 114 (M.... ,. U

I
Tra) slack helghl, Ii (2-1"10 spacing)
i

TABLE U--6
COrTeclion fMCIOD for column Ir..,.!>
Tn.y .. In
" " "
F,
10
"
22
Tra) Iypc
God !'Lile SIC\"e
TrOUllh Bubble
(00 do.... n
o. valve
ap KQCllde
cOIner)
F,
00
00 00
" "
"
TUlltUlenal
Ci SS
MoDel
F.
00
"
..
.".
6-
;SO
5
.'"
1//
,.
/
4 , 0'

//
VV I
'"
1- i..... <::O
rV
I
.
I
2

I
I
,
I
50
40
3ll
2
V/
V
I
7
5
Vertical f",brtcation
3
V
2
v--- Honzonl'll labricatlon
I
07f-<V
05
=

576 Sl'M'l<lN EJ SUMIolUl' rouFl....not<S 5E("TION P2 SUMMARY or C'OSl ronELATlOt<S 577

TABLE .:'1-7
To... er picking"
where Q = cfm and
where Q "" am and
From Peters and Timmerhaus." Januar) 1967 cost. sec Fi@. E.2-7
3-PSI maxImum .
Purchased CO'iI - ( 'J"
100 < Q < 10.000
Turbo Blolo"ers
100p<;1 maxImum discharge:
Purchased Cosl =
1000 < Q < 30,000.
3O-psi maximum discharge;
'"
n.
,
" 212
"70
702
14S
159
M ft'i... ....
....... sm'
"'c.. ....bon
"'lun"nl
Coh
C,ushed .... ne
S1hc:1 reI
I-In Rasch" Ilflp Stoncw..r
Porc..... on
St ..",1e5s
1.1n a.:rl Sln...,..':t,e
l'OIcd.,n
Purchased Cost =
.... here Q = cfm and 2000 < Q < 15.000 Assume IOstallallon raclor ., 40
0'
i!
"
E
=
<
2
]

0'

<
0'
o
10'
,
""'"'
_Turbo blowers

1
3O="psl max. dIscharge
100psi maJl. dlschaIg
JKi max discharge
"TITTTT
Rotary blo...-er
,V
Centrifugal blov.'Cr
10-p.si max. dischar
I
V- 17
/
17 ,
I
m
Heavy-duty blowel
J-in. .stalK: P'"
""'!"
Jan. 19671
10'
Capacity, ft'/min
10
f"IGLIl. 1:..2--7
8lM.en (I,,,," J.f .OJ Pn,,, nnd K D TUJlnW:'htN.. PIQ1II nr"q" aruf
Erottom.o /t1If Enoj,,,",,. td rd. J.fcG,n.. ..' I", . 1980." J,61)
M S Peters aDd K D T'lllrMlh.u<, MPbnl Ilf-<>,:" Ind rCt>oomlCS ror OlCrmetl EngHlecrs.M
MeG,. .. Ne... YOlk, I9'lQ P 562
APPENDIX
F
CONVERSIO
FACTORS
Area
I (II = 0.0929 m
J
= 144 in J
Density
Ilb/fl) ""' 16018kg/m)
.. 1/62.4 g/cm)
I Ib mole of an ideal gas, 0"<:' I aim'" 359.0 (t]
lib mole of alf, O"C I aim'"' 0.0807Ib/{t)
Energy-Also see Work
I Btu ""' 252 cal
= 1.055 kJ
= 7779 {(Ib{
_1929 x 10 4hphr
= 29307)( 10- kwhr
51.
4PI'NI)IJr r COKYUSION r 4ClOU 579
Force
I Ibf = 4 4482 (kg m,o'sl)
_ 32.174 Jbmft/s
J
_ 4 4482 x 10' dyn (I' cnu
2
)
Heal Load-Also see PO""er
1 Btulhr _ 029307 w
Heal-Transfer Coefficient
I Btu/(hr (tl"F)=5.6782w/(ml."q
= 1.3571 x IQ-ca!f(cm
J
s"C)
Length
Ift=03048m
Mass
I Ibm "" 045359 kg
I ton (shorl) = 2000 Ibm
Pressure
I atm .. 14.7 psi
I psi _ 6894.76 N/m
J
(dyn/cm
l
)
POl\er-AIso see Heal Load
I hp ... 550 {(Ibfjs
- 07457 kw
... 2546.7 Blu/hr
Specific Heal
I Blu'(lbm .0F) _ 4.1869 kJ/(kg"q
Work-Also see Energy
1 ft Ib{ a 1.2851 X 10 ) Btu
= 17662)( 101"whr
AP'EJ'lOIX , ('ONHUIOtol fAClOU
Velocity
I fIls""" 03048 mls
Vicosil)'
IlbmlCrt s) = 14881 J..@.'(m s)
I Ibm (fl hr) "" 4 IJ '8 'I( 10 19/(CIl1 sl
Volume
I rt
J
_0028lI7m'
= 28.32 L
= 7481 gal
INDEXES
581

AUTHOR
INDEX
Aerstln. I ..
AIChE, 417
Allen. D 11.7
Andrecc'I\ich. M J. 272. 481 485
Baa'it'l. W D_.514
J E.40R
Banchero. J T. 65 68. 291
Barro.... , M. 1-1. 487
Biegler. L. T.. 381. 400
Boland, D., 227
BoUes. W. L. 112. 1]5, 285 286
Bro.....n. B E.. 285 -286
Brown, G. G; 454
Caldarola, GA. 192. 198, 201
Chien. 1-1., 475
Chilton. C. H. ]]. 4]5, 451. 454-455.
461.491-494. SOB
Colburn. A p.. 4] 1 434. 460
Deal, c. H., 508
Den. E. L.. 508
Diaz. H. E.. 6
Doherty. M. F; 191-192. 195 204.
214. ]19, 325. ]]2, 336. 34]. 345.
354. 364-367. 407. 415. 450
Douglas.J M_.14 15.117.164 165.
178. 273-282. 293-294. 298 303.
306.309-314,319.325.3]1 332.
3]6.343.345.354.364 367.407.
415.445.450,461.464,478
Dunford. H. 265
Duran. M A. 281
D)-bdal. E. C; 113. 134
EdmIster. we. 4-' 1. 498 5(l(J
Edul].11 F.441
Erbar. J H; 443
E\ans. LB. 35
Fair.) R_.13 31.74 135,151,171.
m
Fldko.... ski. Z. 475
Fisher. W R. 319. 312. 336. -'4'.
345.354.364 367. -J07, 415. 450
Flatz, W. 412
Flower, J. R., 219
Forder, G. J., 412
Franklin, W 8.,431
GIlliland, E. R_. 437. 439 440,443.
485
Glinos. K_. 173, 178. 180 181. 184.
335,443.450.461. 464 465.
468-478
GrayOn. H G.., 543 546
Grossmann. 1 E. 281. 381. 400. 412.
50S
Grumc:r. E. L., 15, 102
Guthrie. K. M., 33-34, 569-577
Hachmuth. K. H. 42F
Hadden. S. T. 543 546
Hand, w. E. 33
J., 33. 295, 460. 514
Hasebe. s.. 412
lIashimoto. I.., 412
Hengstebed:. R J.. 6568.291
Hindmarsh, E. 227. 236. 267
Hogsett, J. E. 171
Hohmann, E. C, 219. 2_'4
1I0rton. G. 4]1
Iiougen. 0 A .305. 516
lIughn. R R_. -'70-371
5115
586 AUTHOl 1i'lDL\
Inglo.':by, 5, 450
Irnbarren, 0,110,410
Itoh, J., 219
Jarare)', A, 331, 336, 445
Jordan, D G., 33, 295, 460 461,51..
Karaml, A I.. 412
Kentar, 5_. 411
Kmg, C J. 167,448
Kirkwood, R. L, 180
Kmght, JR., 204, 450
Kremser, A, 82, 427
Krolikowski, L,475
Kyle, H. E., 15
Lang, II J, 33
Lang, V D., 381, 400
Levenspcil,O., 157-158,320
Le\y,5 G, 191, 195-204,214
Lmnholr, B., 219, 227, 234, 2]6,
259 261,265,297
Llu. Y A., 92
Lockhan, F. J.. 475
Loonkar, Y R.. 412
Lu, MD., 92
McAvoy, T 1,331,336,445
McKella, J. J., 8, 72, 112,134 135,
358,361,518
Maddox, R. N., 443
Malone, M. F., 94,110,173,178,181,
184, 335,410,44],450,461,
464 465, 468 478
Marek, J., 92
M.arque:t., F E., 178,461,464
Malsuyama, H. 198
Mazur, W H, 171
MOlard, R L,92
Myri:Hbem, eM, 110
V !\I tJ2
R, 92
N.. 178
l"lshlmura, II J.
O'Connell, II "',435 436
Ollis, 0 F, 408
Overturf. B W, 412
Page, R. C, 7
Pauls, A C, 370 375,548
Perry, R H" 435, 451, 454 455,461,
491-494. 508
Peters, MS. 33. 37 40,87.91,195,
S03, ;07, 577
Pham. H N.. 204
PlCrO!tl.. G 1,508
PIgford, R L.. 431
Plluhk. A., 6
Powers, G J., 92
Prausmtz, J 11.1., 106. 508
Rase, H F., 307, 487
Reid, R C, 106, 508
Reklailis, G V.,412
Rlppin, D. W 1.,412
Robinson, C S. 437, 439 440,443,
480
Robinson, J D., 412
Rod, V., 92
Rossller. A P. 408 .. 09,450
Rudd, D F,92, 178. 181
Saccrdoll, I:. D, 17
Sargent, R W 11,412
Saw}.:r, E. W, 526
O,I7S,370 375,475,548
Seider, W 0, ]70375,548
SheltOn. M R, 505
Shn\\ood. T 1\ 106, 431.
Shuoko, K, 219
Smola, J J.,92
SJlsb), R I, 526
Smlth,B 0,431,456
Smith, R, 265
Smoker, E H., 444
Souders, M.. Jr., 454
Sparro\\. R E.,412
Slcphanopoulos, G, 178
Stoekar, U V.. 431
SUCCI. G. 515
SlUpm, \\' 1,475
S\\3ml, S. 114
Talamal5u. T" 412
Ii. 7
Tedder. 0 W., 178, 181
Ternll.O L 19,273 283.309-314
Thau;:her. C M" 6
Tlmmerbaus. K 0 .. 33, 37 40.87,91.
295,503.507,577
TOloloDscnd. W 0, 10,234,261,265.
291
AmHOl INDU 581
Umed,l, 1., 219
Underwood, A J V. 56]
\all Dt>negen, 0 8., 191, 19" 203.
21 ..
Vrede\tdd, D R, 227
Walson, K. M.. 305, 526
W8ybum. T L,475
Wcnner, R W., 113, 1]4
Wessel, Ii E.. 45
Wel>lerberg, A W., 178, 204. 272, 475,
481 485
Wilke, C. R., 4]1
Winn, F W.,443
Woodcock, D. C. 29] 294,298303,
306, 409
Woods. J M, 412
Yorl, R_, 526
Zimmerman, C C.. 526
SUBJECT
INDEX
Il. 73, 1611
adldb;itk;., 88 90
ahernatnc:s.74 II')
b.. ..... -e..f-the-en..dopc model, IB,
427
melhoJ, 43 I 434
cos!. 457
design equ3UOnlt., M2 M4,425 430
design problem, 74 90
procedure, 514
diameter, 454 457
energy 80 81
flooding, 454 457
fractlonal recovery, 77, 86 89
heat effects, 429 430
height. 453
equation for Ira}:>. 83.
417 430
matenal balaoces., 76-79
mtnlmUm sohenl now. 88 90
mulllcompooent.431
lIumbc:r of plales., 83. 425 436
pdcled w.... ers, 431 435,460,514
plale efficiency, 435 436
pressure. 82
of thumb, 76 77,85 89
shoncul design for plates., 83.
436
!>Ohent.
sohent flo.... rate, 77, 85 86
schent los5., 78 80
lempcrature, 74 76, 83
AcetiC aCid, I' 2, 134, 138, 140, 160
ActllC anb}dnde, 112. 134, 138, 140,
160,213
Acclone. 20. 112. 134, 138, 140,
147 148,152.160 161,213
absorplloo, 72-81
condensation, 490 507
hc aCid, 152
Adlaballc temperature change,
141 149
13,73. 168
AII.xallon of costs. 297 302
Alternatnes
cost diagrams, 2li9 315
economIC evaluation of. 63.
289315
flo.... shecIS, 74, 116. 303
heat-elllhangcr networllt.. 140,
3{)9-)15
IdentlfYIng. 119, 30J
number of, 4, lb
retrofit. 354 368
scrc:c:mng: companson b)
desJglllng each,S
cost diagrams, 289 315
elimlllallon uSlllg beunSllcs, 5
oplimlzatlon, 319-349
solvent recovery. 73, 81
(Ste 1l1so Hydrodealk)lallOn of
loluene proo:ss)
Ammoma s)othesls- 153
Annuity, 52
Approach temperature, 222, 196, 324
Area of heat exchangers, 233-236
ArtIst's approach (set' Englllcrnng
method)
Azeotrope, 21, 143, 174
AzeotropJc dJslJllallon. 185 1&6
dlstillallon boundanes, 189 194
feed compositiOn, 192 194
mlllimum rcflux ratio, 194
Back-of-the-envelopc calculations
(see Shorlcul designs)
BasJC research, 8
Batch proo:.c:sses, 16, 107 110.
114 115,409-412
Benzene, 8 15.20 21.23. 113.
126 1l2, 134 135, 138, 140.
142,147 148. ISO, 153.
160 161,207 208,213,
518 542
Benzoic aCid, 22, 213
Blower co:.t, 516
591
392 SUIJECT INDEX
Boundmg solullons, 90 93,464 465
Bubble points, 428
Butadiene (see Butadiene sulfone)
Butadiene sulfone, 135, 161, 213
Butane alkylation, 135, 140. 161, 213.
m
Butene (set' Butane alkylation)
Capital charge factor (CCF), 60
Capital costs, 23. 29, 32
contingencies, 41
correlations, 33-37,
cost diagrams, 289-315
depreciation. 46
direct cost. 41-42
fixed capital, 37. 41-42
Guthrie's correlations, 34. 569 577
indirect costs, 41 42
inflation factor, 36
inside battery limits (ISOl). 41 42
installation factors. 29, 33 35
offsite C05t, 41-42
onsite cost. 41-42
\s, operating cost. 48 54,354-368
outside battery limits (OSBl),
41-42
owner's cost, 41 42
start-up cost, 40-42
tolal capital investment, 37-42
vendor's quote. 33
working capital, 29. 37, 41-42, 70
(See a/50 Cost: Profit)
Cascade dia8-ram, 220-222
Cash flow. 48
Catalyst deactivation. 103
Centrifuge.408-410
Complex distillation columns,
ISO 182,466 478
Compressor cost, 155 156,329,573
153 155,490.
513
computer-aided 557
refrigeration processes, 490 507
Computer-aided design (CAD),
369 404
computer inFormation diagram. 381
controllers, 389
convergence, 378, 559
equipment subroutines, 373-375,
550-567
roxecutive. 370
flash calculations, 382-387,
550-553
Iterations, 389
physical property data, 371, 379
sequential modular, 377
starting values, 381
stream tearing. 377
thennodynamics, 371
Computer infonnation diagram. 381
Conceptual design (see Process
synthesis)
Condensation, 13,73,168
Condenser, 458, 487-488
Consecuti\'e reactions, 113
Conservation of money, 49
Constraints. 105.210,357-368
Continuous interest, 49-54
Control,414-416
Convergence, 378, 559
Conversion, 17, 94, 124, 145, 150.
157,209,296,320,324-327,

Con\'ersion factors, 578-580
Cooling water, 75. 458
cost, 328-332
Cost:
capital (see Capital costs)
cooling water. 328-332
data, 106, 569-577
diagrams. 289 315
distillation, 330. 457, 460. 461,
574 577
equipment (see Capital costs)
estimates. 23 71
opcratlllg (see Operating
product. total. 37. 43
revenue. 44 48
Cost diagrams, 289 315
cost allocatiOltS, 297-302
heat exchangers. 297
process streams. 300
heuristics, 295
operaling costs. 289-315. 354,
358 360
significant design variables. 296
structural modifications, 296
Cost models, 327-332
Creative activity, 4-5
Crystallization, 187-1 88, 408--411
C)"c1ohexane, 23. 134, 142, 150. 161,
213
Decomposition, 8, 17
Design:
costs of developing. 7
optimum (see Optimum design)
types, 6
\ariables (see Conversion; Molar
ratio of reactants; Pressure;
Temperature)
De..... points, 438
Diethylbenz.ene (see Styrene)
Diethylcther (see Ethanol)
Diluents,153
Diminishing returns, 6
Dipbenyl (see Hydrodealkylation of
toluene process)
Discounted-cash-flow rate of return
(DCFROR), 56-59. 70
(See 0150 Profit)
Discrete compounding, 50-54
Disproportion of toluene, 66, 213,
290-293
Distillation:
alternatives, 182 188
applicability, 175
azeotropes, 189-204
bubble point. 438
complex 180 182.
466 478
compoSItions, 436. 450, 509
SUBJf.n 593
DistillatIon (COTII.):
computer-aided design, 397. 553,
562 564
condenser, 458
cooling water requirement, 458
costs, 330. 457. 460. 461, 574 577
dCl'lgn equalJOlls, 436 453
deSign procedure. 508 511
dew point, 438
diameter. 454-457, 510
energy integration, 264 272,
478 485
FellSke- Underwood-Gillila nd
procedure, 439-444
Fenske's equation for minimum
plates. 441
flooding, 454-457
Gilliland's correlation. 439 441
heat integration, 264 272
height, 453. 510
heuristics for sequencing, 462
McCabe-Thiele method. 450
material balances, 436, 450, 509
minimum reflux ratio, 194-204,
441-444, 447
number or plates, 439-453
packed tower. 460
pasteurization columns. t 73,469
Petlyuk columns, 472-476
plate efficiency. 451-453
prdractionator, 476--478
pressure, 436-437. 509
reboiler, 459, 512
reflux ratio. 197-204,296,321-325,
341-349,441-444
relative volatility. 438, 448-450
sequencing. 10, 175-182,461 465,
511
sidestream columns, 12, 466
sidestream rectifiers and strippers.
470 471
simple 175 180
Smoker's equation. 444 447
splits, 436, 450, 509
stream requirement, 459
594 SUIJECT II'lDVl
Distillalion (ConI.):
Underwood's equations: minimum
reflux, 441 444
plales, 448
vapor rate, 463
(Su also Azeolropic dJslilJation)
Dominant design variables, 319 350
Drums:
COSt, 67
ftash (set' Flash)
Drying. 408 41 I
Economic pOienlial, 61, 64, 73
)evel 2. I I 32
level 3, 158-159
4, 188-189
EconomIC trade-offs, 5, 319-350,
356-357
Economics, englnccnng. 23-71
Economy of scale, 104
EffCClive interest rate, 51-54
Electricity COSI, 32
Energy conservation (Slit Heat-
exchanger networJu)
Energy inlegralion. 10
distiJlalion, 264-284.478-485
heat and power, 261-264
(SI!I! also Heat-exchangcr
networJu)
Engineering method, 5-g, 20
Environmental constraints, 5, 143
Equilibrium limitations, 15, 141,
149-153
Equipment COSt (UI! Capital costs)
Equipment sizes, 23
Equipment subroutines, 373-375,
550-567
Ethane cracking, 112, 135, 161,
286-287
Ethanol, 21,134,143-144,152,161
Ethyl acrylale, 152
Ethylbenzene (sel! Styrene)
Ethylene, 21, 112, 134 135, 143
Extent of reaction, 128 129
Extraction, I85
Extraclive: dislillalion, 184
Feasible matches, 237
Feed distribution, 157 158
Fenske's equal Ion for mimmum
plates. 441
Filter, 408-41 1
Finite differenct: calculus, SO-54
First law, 218, 225, 230
FiJl:cd capital investment, 37, 41-42
Flasb:
cakulations, 166 168
computer-aided design, 382-387,
5SO-553
splits, 9
Aowshccl:
absorber (stripper), 75
decomposition. 8, 17
input-output structure of lYe
Input-oUlput Slructure of
llowsheet)
processes: atttone from
isopropanol dehydrol!:enation,
18,294
benzene from toluene
hydrodealkylation, 8-17,

benzoic acid rom loluene
oxidation, 21
cycioheJ.ane from benzene
hydrogenation, 24
disproportionation or toluene 10
give benzene and xylene, 66,
213,289-293
ethane cracking to ethylene, 286
ethanol from ethylene and water,
20
ethyl benzene from benzene and
ethylene, 19
hydrodc:sulfurization, 285
recycle struClUre of (see Recycle
structure of ftowshcet)
FLOWTRAN, 369 4O-t, 548 567
FLOWTRAN (Com.);
equipmenl subroulines, 373 375,
550-567
IIlput data, 371
physical propeny data, 371 373
thermodynamICS opllons, 372
Fractional recoveries, 296, 325,
341-349
Fuel cost, 32
Furnace, 329. 489, 513
cost, 570-571
Gas absorber (su Absorber)
Gas compressor, (sa Compressor
desIgn)
Gas rec)cJe and purge, 9,126-128,
209, 324, 520 522
Gasolll)e, 135
(Sel! abo Butane alkylauon)
Gl\hland's correlauon for dIStillatiOn,
439---+41
Grand composile cune, 224
Grassroots plant, 41, 70
Guthnc's correlatIons, 34, 569 577
Hazards, 417-420
Heat:
and dlsuJlation integration,
264-284, 478-485
and power integration, 261 264
Heat carrier, 149
Heat effecls, 142, 146-149
Heat engines, 261
Heat-exchanger design, 486 489,
511-514
COSI, 572
Heat-exchanger networks:
area eSllmates,
heaHransfer coefficienls, 234,
486-487
computer-aided desIgn, 399
cost model. 327
$UIJECT II'lOU 59S
Heat-exchanger networks (COrll.):
design of minimum-energy
236-261
algonlhm, 257-261
alternatives, 240
capital vi. operatlOg cost trade-
off,251-256
complete design. 244
design above the pinch, 236-241
dC'S1gn below the pinch, 241-244
distillation columns. 264-271,
478-485
eliminating CJ,changen, 251
energy relaxation, 251
feasible matches, 237
heat engines, 261
heat pumps, 263
heurisucs, 238, 248, 251, 260
loops, 248, 251-256
optimum approach lemperalUre.
24.
paths, 248
plOch matches, 239
stream splilling. 257-261
mmimum healing and cooling
rcquired, 216-230
approach temperature, 222
cascade diagram, 220
first-Ia.... analysis, 218, 225
grand composite curve, 224
limitations, 229
minimum approach temperature,
222
minimum utility loads, 222
multiple utilities, 227
phase changes, 228
pinch temperature, 222
lemperature-c:nthalpy diagram,
222-224
temperature intervals, 218-220
minimum number of exchangers,
230-233
first law, 230
independent problems, 231
loops, 231
second law, 232
retrofit, 354-368
596 SUBJECT INDFJC
Heat pumps. 263
HeaHransfcr coefficiellts, 234,
486 487
Heuristics,S, 9, 85. 90 91
approach temperature In heat
exchangers, 92 93
batch processes, lOS
compressIOn ratio, 155
94. 145
cost diagrams, 295
dlSllllalion column scquencll1g.
91 92.177 178.180 184,
461-466.511
fraclJonal recovery. 77,
hcal-exchangcr fKtworks, 227. 231.
238,248,251.260.284
heal loads, 14g
II1pul-output s!ruClUre. 13]
hmllallOM,
minimum trays In a dlsullallon
column, 91
numtX'r of product streams,

optimization, 345
pipe: H:IOCIty, 91
reactor (smglc:
reaction), 94. 145
reactor dcsign. 157
recycle structure, 160
separallon system. 163 165,
21
solvent flow In gas absorber. 77,

Hierarchical design proccdure
batch processes, 409 410
pertochemical processes, 8 16, 407
solids processes. 408 410
Hierarchical plannmg. 17
Hydrodealkylalion of toluene (I"IDA)
process, 518 542
alternativcs: feed purificatIon, 304
punfy gas-recycle Slream, 306
recycle of dlphenyl, 304
dlphenyl recycled, 20, III, 1J3,
161. 213, JOJ JI5
dlphenyl removed ad13balle
Hydrodealkylallon of toluene (IIDA)
process. diphenyl removed
(Conl.):
temperature risco 14 148, 524
alternativcs. 132,522.541
benzene oolumn, 535 -537
case study, 518-542
compressor, 524
computer-aided design, 375 40]
.:onstramts, 307
cost diagram. 297-303
decisions. 520, 530-541
decomposition of fIowshcct.
."
distillation, 397. 530 541
distillation column sequencln@.
531
economic pOienual. IJO 1]2.
ISS 159,188-189.522. 527,
54.
ecooomic trade-offs. 124,
158-159,18g 189
energy integralion. 216 284
flash calculations. 167.528
heat-ellchanger net\l.ork.
216 284.399,542
heat load. 147, 524
hydrogen purification. 134, 52'0
mput data. 107.518-520
input-output structurt:, 520523
levels of detail, 8 15
number of reactor systems. 138
operating cost diagram, 360
overall material balances.,
126 130,521-522
reaclor size and cost, 526
rt:CycJe and purge, 120, 520
rcqde compressor, 524 526
recycle material balanCt:S,
142-14],145,523
retrofil. 358 368
revcrsible by-products, 149,520
rigorous material balances.
204 211
sclt:Ctlvlty data. 'i19
5]7 541
Ilydrodealkylation of toluene (IIDA)
process, diphenyl removed
(Com.):
loluene column. 532-535
vapor recovery system, 169 170,
519
H)'drodesuUuriz.ation, 285, 287
Hydrogen. 8-15, 20. 23
Independent plOblems. 231
Input information:
constraints. 105
COSI data. 106
ph)"slcal property data. 105
plant and site data, 105
product purity, 104
production rale. 104
ray, matenals, 104
reaction information. 99 103
Input-output structure of ftoVl:shttt.
15.116-136
dt:51!D 124
c:xccssrcactants,l20 121
gas reqde and purge. 120
DumtX'r of product streams.
121-123
material balances. 123 130
purification of reed streams.
118-119
reco\-er or recycle re\'ersiblc by-
products, 119 120
Inside ballet')' limits (ISBl), 41 42
InvCSInlt:Dt (st'e Capital COSIS)
lsobutane (see Butant: alkylation)
lsooctane (Sl'e Butane alkylation)
20, 134, 147-148, 152.
213
Ileralion, 389
K 543 -546
Ketene (set' Acetic anh)'dride)
II"DX 591
Levels of designs (St'l! Design, types)
Limitations:
heat-ellchanger nety,ork dcsign,
229
opliml:lation procedure, 327
synthesis procedure. 15
llmitmg reactant, 142
liquid separation system. 10 13.
172-204
alternath'CS 10 dIstillation
azc:otroplc diSlillation. 185
crystallization,
ClItractlOn, 182-185
ClItraCtlH: distillation. 184
rcacthe distillation, 187
;azco(ropes with reactants, 174
distillalion: applicabitit). 175
complex oolumns. 180 182.
466-478
sequencing of simple columns.
175-182. 461-465. 511
mteraClJon with process. 178
hght ends. 173-174
multiple- dlstlltalloD sequences, 179
2]1. 248-256
McCabe-ThICk method for
distillation. 450
Marshall and Swift (M&S) Index.
3"
Material balances:
absorber (stripper). 76-79
approximate, 78
computer-aided design,
equilibrium conversion. 150
linear, 204-211
overall (reeds and products).
123130,521522
recycle, 142 145,523
reversible by-products, 146
rigorous,2Q4 211. 375 396
solvent loss from absorber, 78
vapor system, 169 170
598 SUlIJlocr INOD:
Malerial lemperature ranges, 547
Membrane 13,73, 168
Melhane.8 15,20
Methyl isobutyl kelone (MIBK), 79
Muumum coohng required, 216 230
Minimum-energy heat-cxchanger
networks, 236 251
Minimum healing n:qulrcd, 216230
Molar ratio of reactants, 17, 124,
140 142, 157,324 327
Monomers, 152
MOrlgage payments, 69
Multielfecl distillalion, 48
Muhipass heal exchangers, 49
Nominal interest rate, 50 54
Olfsile cost, 41
Onsite cost, 41-42
design problems, 4
Operating cost diagram, 354,
358-360
Operating costs, 23-25, 32
vs. capital cost, 48-54, 354 368
cooling water, 328-332
(See afso Utilities)
cost diagrams, 289-315, 354,
358-360
direct product cost, 43-45
insurance, 43-45
interest, 43-45
labor, 43-45
laboratory charges, 43 45
manufacturing costs, 43 45
opera lor. 45
overhead, 44
raw malerials, 15.43 45
rent, 43 45
repairs and maintenance, 4]-45
royalty. 43 45
Operaling COSts (COnt):
sales, research, admmistration and
engmeering (SARE) cost,
43 45
start-up cost, 40
steam. 328 332
(See also [JI1I!IY COSIS)
superVtSlOn, 43-45
supphes,43 45
laxes., 4] 45
utilities. 25, 43-45, 328-]]2,
568-569
(See a/so Cost)
Operating time, 73
Oplimum design, 62, 319 350
3pproach temperatures, 246
retrofit, 356-368
\ariables. 124, 209
Order of magnitude, design
estimales, 7
Outside ballery limits (OSBL), 41-42
Oxygen. 2, 120
Packed absorber (see
Distillation)
Parallel reaclions, II]
Partial condenser, 173
Pasleurization dislillation column,
173,469
Paths. 248-256
Payout lime, 55
(See also Profil)
Pellyuk distillation columns, 472 476
Phase changes, 229
Phosgene, 141
Physical property dala, 105,371.379,
SOH, 543- 546
Plllch:
dIstillation, 264
heat engines, 261
heat pumps, 263
Pinch matches, 239
Plllch lemperature. 222
Planl and sile data, 105
Planl design, ddinitiun of. 3
Plale ellklency:
absorbers, 435 4]6
distillation columns, 451-45]
Plate gas absorber (see Absorber)
Power and heat inh:gration, 261 264
Power cost, 32
PrefractlOnator
476 478
\alue, 53-54
Pressure, 124, 157. 324
absorber. 82
distillalion columns, 436-437, 509
Problem definition (see Input
infonnation)
Process ahernatives (see Alternatives)
Process constraints, 105
Process control, 414-4 J6
Process design definition of, 3
(See also Process synthesis)
Process flowsheet (see Flowsheel
processes)
Process retrofits, 64. ]53-368
Process synthesis.
batch \s. conllllUOUS processes, 110
creat1\e activity, 5
definillon of, 4, 110
hierarchical procedure, 8
heatcxchanger network, 216-218
Input infonnation, 99-107
input-output structure, 116-136
recycle structure, 137-162
separation system, 163-215
Product purity, 104
Production rate, 104
Profit, 23, 47-48
measures: discountcd-cash-f1ow
rate of return, 56
payout time, 55
return 00 investment, 31, 55
model,61
after tales, 47-48
before tales, 45 46
Proximity parameter. 344, 356
Pumps, 67
Purge (see Gas recycle and purge)
INlJlX .599
Purge composition (see Gas recycle
and purge)
Rank-order parameter, 343, 356
Raw materials, 104, ]20
15,32
(See (llslJ Operating costs, raw
malerials)
Reaction information, 99-103
Reaclions:
consecutive (see Consecutive
reactions)
parallel (see Parallel reactions)
Reactive distillation, 187
Reactor:
computer+aided design, 564-567
configuration, 157-158
cost model, 329, 334, 574
design, 156-158, 507
design guidelines. 157
design variables: conversion (see
Conversion)
molar ralio of reactanls (see
Molar ratio of reactants)
pressure (see Pressure)
temperature (see Temperature)
equilibrium limilalions (see
Equilibrium Iimilations)
heat effects, I, 142. 146-149
heat load, 146
separator reaclOrs, 152
Reboiler, 459, 512
Recycle and purge (see Gas recycle
and purge)
Recycle S!fucture of nowsheet, 14,
137-162
case study, 523-528
compressor design and costs,
153-156
economic evaluation, 158 159
equilibrium limitations, 149-153
excess reactants, 140
material balances, 142 145,523
number of reactor systems, 138
600
Recyele structure of nowshcct
(Co"r l'
nUlllhcr of rcC}'CIe streams.
Il8 140
reactor dign. 156 158. 507
reactor Mat 146 149
Renux lallO. 197 204.296.321-325.
341 ."\49. 441 444
Rdnger3110n dcslgn_ 490 "'07
Rdatl\c H,lallht). 438. 448 450
Relrofit (.su Process retrofits)
Return on m\t'slmenl (ROI). 31. 55
Reversible byproducts, 10. 119. 146.
520
Rule of thumb (Mf' Heunsllcs)
SaFety. 5, 417 421
Salvage \<lluc, 58
Scalc Factors, 342
Second la ..... 2.12
Sdtivity.101 102.111-114.
126 119.334
88
Separation s)stem'
general structure. 13. 163-168
liquid separation system (_
Liquid separatIOn system)
vapor rcC()\cry system (SoU Vapor
rcco,-ery system)
Separator reactor", 151
Sequential modulJ.r Simulation. 377
Shortcut designs, 5,16.73
absorbers. 426 436
backofthe-ene,dopc modd, 83
boundmg solutions. 92 95.
464 465
cost models. 328 340
nash calculatiOns. 166 16R
optimizations, 319 350
overall matcrial balances. 123 130.
3l]
reC)'cle balances. 142 145.))]
rdngeration 'i05 507
Side reacllons. 100
Sidestream distillation columns. 12.
466
Site location, 417 421
Smoker's equation for dlslillation,
444-447
'iO,. 135. ISJ
Sal' 153
Solids procc:sscs. 408 410
Sohent rceo\'er). 71-81
alternall'-cs. 73. 81
Stabilizer. 9. 173,537 541
Stan-up considerations. 5. 416
Steam cosL 32, 331
Stoichiometry, 126 129
Stream cost. 32, 3J3
Stream spliuing_ 257
Strum table. 15. 129
Stream teanng, 377-]79
Styrene. 113, 134. 153. 161. 213
Success rates, 4
approximallons (UfO
Fngineering method)
Sulfuric acid. 152
Srnthcsls and anal)sls (Sf'f' ?rocess
synthesis)
Systems approach, 82 83. 296
Tanks:
reed,67
storage, 67
Temperature, 124. 157.314 ]17.
341 349, 547
Tempcrature-enthalpy diagram.
222-224
Temperature intervals. 218-220
Thermodynamics, 371 ]7],543 546
Time value of money. 48 54
Toluene. 8 15.20.22. 113. 126 132.
134.138.140.147 148.151.
161. 205 207. 211, 29710l,
]75 403.518 542
Toluene disproportion3110n. 66. 213.
290 293
Total annual cost, 331
Trade-offs (Mf' EconOOlte trade-offs)
Tricthylbcnzcne (ye Styrene)
Underddil1ed (st"l" Open-ended
design problems)
Underv.'oo(fs equauon ror mimOluOl
reflux. 441-444. 448
Utility costs, 32, 328 332, 568 569
(Sa also OpcratJOg costs)
Utility flows. 23-25. 216 230.
480 482
minimum, 222
multiple utilities. 227
Vapor-liquid equilibnum, 54) 546
Vapor rccO'ery system. 13
combined ..... ith liqUid scparation
system. 172
localiOn, 16R J71
typc. J71
Watcr-!as shift reactl"n. 153
Workmg capltal.:N. 41 42. 70
Xylene (sa Toluene
disproportional ion)
Yicld.IOO.111 114