Computers end Chemical Engineering, Vol. I. pp. 11-21. Pergamon Press, 1977.

Printed in Great Britain

COMPUTER AIDED CHEMICAL PROCESS

DESIGN: THE FLOWTRAN SYSTEM*

EDWARDM. RosENt and ALLENC. PAULS

Monsanto Company, 800N.Lindbergh, St. Louis, MO 63166, U.S.A.

(Received28July 1975)

Abstract-A description of Monsanto’s FLOWTRAN (Flowsheet Translator) system is given. A sample chemical
process is first used to illustrate the system. Then the manner in which FLOWTRAN’s design addresses some of the
commonly encountered problems in chemical process simulations is discussed. Finally some reactions of industrial
and academic users to the system are indicated.

Scope-The general field of steady state chemical process simulation has recently been reviewed by Motard et al. [ 141
and a number of the available flowsheet simulation systems have been investigated by Flower & Whitehead[8]. This
paper describes Monsanto Company’s system (FLOWTRAN) as well as some user reactions to it from industry and
academia. A comprehensive description of FLOWTRAN may be found in Seader et al.[19].

Conclusions and Significance-The FLOWTRAN system has been found to be a viable and economically attractive
steady state chemical process simulation system in a number of industrial environments. Although the full impact of its
use in the academic community has not yet been fully evaluated,early indicationsare that it has been useful and
productive in process design courses. Key to the success of a process simulation system is its physical property system,
g simple user- language and its reliability.

1NTRODUCTlON purpose system. A user-oriented language was next

The use of the digital computer to perform material and
energy balance computations for interconnected chemical
processing units is now nearly two decades old. Indeed,
literature on the subject of chemical process simulation as
a design tool began appearing in the late 195Os[12,15]and
by 1960 the Machine Computation Committee of AIChE
held a Workshop on Heat and Material Balances[l]. In
1968 Evans et a!.[71 reviewed the developments in the
field of computer-aided chemical process design and
discussed many of the flow sheet systems which had been
publically acknowledged at that time. This was updated
by Flower & Whitehead in 1973[8]. Among these flow
sheet systems is the FLOWTRAN system of the
Monsanto Company[F5,9,11, 16,191.
HISTORY
Much of the early work in Monsanto Company in the
field of computer aided process design was characterized
by efforts to simulate individual chemical processing units
such as distillation columns, absorbers and flash tanks. By
l%l attempts to tie together simplified models of
individual process units into more comprehensive flow
sheet simulations was well underway[l7,22]. However, it
.was not until late 1964 that a team of six was formed to
attempt to define and implement what has now become
the FLOWTRAN system.
Initially, efforts were concentrated on physical proper-
ties, input/output routines and design of the general
*Presentedat the session on “Computer-Aided Design”, 71st
National Meeting, AIChE 20-23 February 1972, Dallas, Texas,
U.S.A.
Revised May 1975 for presentation at U.S.-Japan joint
seminar on “Application of Process Systems Engineering to
Chemical Technology Assessment” 23-27 June 1975, Kyoto,
Japan.
tTo whom correspondence should be addressed.
developed, a number of the basic simulation subroutines
(blocks) were implemented and a user’s manual was
written. This phase terminated the initial development.
The date was April, 1,966and FLOWTRAN went into
general use in Monsanto.
In early 1969, FLOWTRAN was marketed outside of
Monsanto using external commercial computers for
approximately 70 outside customers and in early 1972,
FLOWTRAN was offered for license. Though the service
bureau business was terminated in early 1973, the
intention to make FLOWTRAN available to Universities
through an alliance of chemical engineering educators (the
CACHE committee) was announced in December
1973[ll]. The continuation of licensing of the FLOW-
TRAN system was reiterated in early 1974.
THEFLOWTRAN SYSTEM
The FLOWTRAN system is a completely integrated set
of programs written entirely in FORTRAN. The use of
FORTRAN makes the system portable, easy to modify
and maintain. It has, in fact, been implemented on the
CDC 6000 series, the GE/Honeywell 600/6080, and the
IBM 360 and 370 series. The major components of the
system are indicated in Table 1.
Central to the use of FLOWTRAN is the information
stream, the design of which is indicated in Fig. 1, and the
operation blocks which calculate output streams from
input streams. Each of the blocks has available a large
array of physical properties, such as vapor pressure, as
well as thermodynamic routines to carry out such
calculations as a dewpoint or bubblepoint determination.
In addition a basic mathematical library to perform such
functions as equation solving and interpolation is availa-
ble. The system also includes preprocessors for FLOW-
TRAN input and physical property development. Finally,
it contains a fully integrated physical property file.
TO carry out a FLOWTRAN simulation the name and
12 E. M. ROSEN and A. C. PAULS

Table 1.Components of the FLOWTRAN system OVERHEAD PRODUCT

(1) FLOWTRAN Process Simulator: This program translates

the FLOWTRAN description of a process flow sheet into FLASH DRUM
FEED
computer programs which it then executes. I *

(2) PROPTY Physical Property Program: This program takes

raw property data and computes constants for physical RECYCLE BOl-rOMS
property correlations used in the FLOWTRAN simulator. PRODUCT
(3) VLE Phase Equilibria Program: This program takes raw
phase equilibrium data and computes parameters for liquid
1
phase activity coefficient correlations used in the FLOW-
TRAN simulator.
(4) INF Information Retrieval Program: This program stores Fig. 2. A flash with recycle.
the physical property constants from PROPTY in a public
or private data file. Subsequently, it retrieves the constants
for use by the FLOWTRAN simulator. Data for 180
chemical species (components) are stored in the public data
file.

Vector index Usage

1 Flow of component 1, lb mole hr-’
2 Flow of component 2, lb molehr6

N Flow of componentN, lb mole hr-’ Fig. 3. FLOWTRAN flow sheet.

N+l Total flow, lb mole hr-’
N+2 Temperature of stream, “F
N+3 Pressure of stream, psia
N+4 Enthalpy of stream, Btu hr-’ ?iFLSH
Ni5 Fraction vapor (molar) Adiabatic Flash
N+6 Stream name DescrzDtioa: AFISH determines the quantrty and composition
of liquid and vapor streams resulting when up to seven feed
Note: The maximum number of components (N) is 25. streams are mixed and flashed adiabatically. The number of
prOduct streams my be one or two. The block can be used
Fig. 1. Information makeup of a stream. to simulate a pressure drop across a valve or through a
pipeline. If two producr: streams are specxfied and the
flash conditions result in a single phase. the appropriate
product is set equal to the sum of the feed streams and the
type of each unit operation in a process, the order of other stream is set to zero. You can also specify heat ad-
calculation, and a name for each unit’s input and output dition to 01 removal from the flash unit.

streams is needed. In addition three other categories of a: The output gives the unit name, the feed and prod-
uct stream names, the flash temperatures and pressure, the
information are required: heat added to or renmvcd from the system and the fraction
of feed which leaves as vapor. An option is provided for
1. The chemical components used in the process. printing stream flows and physical properties including
2. The unit’s design and operating variables equilibrium K-values.

(parameters) such as numbers of trays in distillation Pxmerties Used: Vapor-lrquid equilibria and enthalpies.

columns and areas of heat exchanges. Block List:

3. The composition and condition of each stream which List type BLOCK

flows into the plant. Unit name

Unit type AFLSH
From this information the FLOWTRAN system calcu-
Name of 1st feed stream
lates the steady-state operation of the entire plant. The Name of 2nd feed stream or C
results show the operating conditions for each piece of Name of 3rd feed stream or 0
equipment (e.g. the inlet and outlet temperatures and heat Name of 4th feed stream or 0
duties of heat exchangers) and the complete chemical Name of 5th feed stream or 0

composition and condition of every process stream. In Name of 6th feed stream or 0
fame of 7th feed stream or 0
addition, the capital and operating costs of each piece of Name of lquid product stream*
equipment can be obtained along with a profit and loss Name of vapor product stream or O*
statement for the entire plant. *If only one product stream is specified, Its phase candi-
tion will be determined.

EXAMPLE OF A FLOWTRAN SIMULATION AFISH'

BA-102.2
Figure 2 is the flowsheet of a simple isothermal flash
operation in which one-half of the bottoms liquid is Pa
zzz:e' Lia: PARAM
pumped back and mixed with the feed. To simulate this
unit name
process, FLOWTRAN units corresponding to each of the Index of first entry 1
operations are assembled as shown in Fig. 3. The unit for 1. Flash pressure. psia, if positive.
an adiabatic flash operation is the AFLSH block (User Flash pressure minus the minimum feed
pressure, psia. If 0 or negative.
Abstract shown in Fig. 4) and the PUMP block is used for
2. Heat added. Btw'hr (If negative, heat
the pump. The SPLIT block corresponds to the collector removed)
and the HEATR block is used for the heat exchanger. In 3. Print stream flows, physical properties,
each unit is written an arbitrary unit name (Al, Fl, Hl, Cl, K-values: 0 = NO, 1 = Yes

PI, RI) and the unit type (ADD, IFLSH, HEATR, SPLIT, Fig. 4. Sample block abstract.
 Computer aided chemical process design 13

Table 2. Major features of the FLOWTRAN system FLOWTRAN input information are:

(1) Ability to handle nonelectrolyte process streams of liquid
mixtures that range from ideal to nonideal including three
phases.
(2) Pure component physical property fde for 180components.
(3) Simple control over input and output.
(4) Free format input to the process simulator.
(5) Easy addition of new blocks.
(6) Can accept FORTRAN statements to control simulation
logic.
(7) Has been widely used in American industry.
PUMP, SCVW-Table 3). Also, note that each stream in
the FLOWTRAN flow sheet is given a unique name.
There are two units in Fig. 3 which require special
explanation. First, whenever two streams (such as FEED
and R03) are combined, an ADD block may be used.
Second, it can be seen that because of recycle the
composition of the feed to the flash drum depends upon
the composition of the bottom stream; that is, the inlet
stream composition depends upon the outlet stream. In
this situation the stream convergence block (SCVW) is
used to determine the composition of the recycle stream
at steady-state operation. It does this by iteratively
estimating the composition of the recycle stream until the
estimate matches the calculated composition within a
specified tolerance.
To complete the specification of the problem we will fix
the feed stream (FEED) composition as:
Methane-50 lb hr-’
Ethane-100 lb hrr’
Propane-700 lb hr-’
n-Butane-15 lb mole hr~’
I-Butene-21 lb mole hr-’
1,3-Butadiene-95lb mole hr-‘.
1. FLOWTRAN units-one line is required for each
unit. Each line starts with the word “BLOCK” and shows
the unit name, the unit type and the names of the inlet and
outlet streams with each name separated by one or more
spaces.
Since units are calculated in the order in which these
lines are arranged, the inlet streams to a unit must be
either the outlet streams from preceding units or fresh
feed or recycle streams.
The SCVW recycle block lists, as the fifth data item, the
name of the first unit in the recycle loop. This is called the
“transfer point”. After each unit in the loop has been
calculated, the SCVW block will make a new estimate of
the recycle stream which is fed back to the beginning of
the loop, indicated by the transfer point. The order of
calculation is shown in Fig. 6. Unit Al is the transfer point.
2. Components-since each of the six chemical com-
ponents in this process is included in the data bank, only
the component names are needed. They are preceded by
the words RETR, indicating the data are to be retrieved
from the physical property data bank.
3. Parameters-each block has a list of parameters
(design and operating variables) which are listed in the
User’s Abstract for the block. All parameters are entered
on lines which begin with the letters PARAM followed by
the name of the unit to which the parameters apply. For
example, the PUMP parameter is the outlet pressure
(115 psi in this case).
Each parameter line may list as many parameter values
as can be fit on one line. The index of the first parameter in
the line is entered as the first number after the unit name

The temperature will be 85°F and the pressure lOOpsi.

The heater outlet temperature will be fixed at 120°F with a
15 psi pressure drop. The flash is at 25 psi.
The complete input to the FLOWTRAN program for
this problem is shown in Fig. 5. The input to the
FLOWTRAN program is completely “free format” with
each item of data separated by a blank. This allows the
user the freedom to enter his data without being
concerned with specific card columns. The major sections
of the program, corresponding to the four types of Fig. 6. Order of calculation.

TITLE FLASH WITH RECYCLE

PROPS 6 2 2 2 2
PRINT INPUT TABLES
RETR METHANE ETHANE PROPANE N-BUTANE 1-BUTENE 1.3~EJUTAOIENE
BLOCK Al ADD FEED R03 5.0 SO1
BLOCK F.1 AFLSH SO1 6.0 SO2 OVHD
PARAM Fl 1 25
BLOCK Cl SPLIT SO2 ROl BTMS 5*0
PARAM Cl 1 2 0.5 0.5
BLOCK Pl PUMP ROl RO2
PARAM Pi 1 115
BLOCK Hl HEATR R02 R03A
PARAM Hl 1 120 15 0 0 1 0
BLOCK RI SCVW R03A 2*0 Al I403 2.0
POUNDS FEED 1 50 100 700
MOLES FEED 4 I5 21 95
TEMP FEED 85
PRESS FEED 100
TEYP R03 120
PRESS R03 100
END CASE
END JOB

Fig. 5. FLOWTRAN input for flash with recycle.

14 E. M. ROSEN and A. C. PAULS

Table 3. FLOWTRAN blocks

Category Block Descriptive title Categor) .Block Descriptive title

Flash IFLSH Isothermal Flash Bounded Wegstein

AFLSH Adiabatic Flash (Acceleration Method)
BFLSH General Purpose Flash
KFLSH Isothermal Three-Phase Flash
FLSH3 Adiabatic Three-Phase Flash Control CNTRL Feedback Controller
Absorption/ PCVB Multiple Parameter Controller
Stripping ABSBR Rigorous DSPLT Distillate Fraction Controller
Extraction EXTRC Rigorous Liquid-Liquid RCNTL Ratio, Sum & Difference
Heat Exchange EXCHI Shortcut Heat Exchanger Feedback Controller
CLCNI Shortcut Cooler Condenser
DESUP Shortcut Desuper-heater
HEATR Heat Requirements Cost & Sizing CAFLH Flash Drum
Miscellaneous ADD Stream Addition Estimation CIFLH Cost Estimate
Unit CKFLH Cost Estimate
Operations MIX Stream Addition with no Phase CFLH3 Cost Estimate
Change CFRKB DistillationColumn (‘ost Estimate
SPLIT Stream Split CAFRC DistillationColumn Cost Estimate
EXCH2 Shortcut Partial-Total Vaporizer- CDSTL DistillationColumn Coat Estimate
Condenser CABSB Packed Absorber Cost Estimate
BOILR Reboiler/Intercooler Simplified CTABS Tray Absorber Cost Estimate
Model CEXCI Heat Exchanger Cost Estimate
HTR3 Three-Phase Heater/Cooler CEXC? Heat Exchanger Cost Estimate
EXCH3 Shortcut Heat Exchanger CEXC3 Heat Exchanger Cost Estimate
Distillation FRAKB Rigorous (Ke Method) CHETR Heat Exchanger Cost Estimate
DISTL Shortcut (Edmister) CCLNI Shortcut Cooler Condenser Coct Est.
DSTWU Shortcut (Winn-Underwood) CPUMP Pump Cost Estimate
SEPR Constant Split Fraction Separation CC@MP Compressor Cost Estimate
AFRAC Rigorous Distillation/Absorption
(Matrix Method)
Reaction REACT Chemical Reactor II Report Writer SUMRY Stream Output Editor
AREAC Adiabatic Add/Subtract Reactor Component Physical Properties Table
XTNT Chemical Reactor (Extent of Reaction GAMX Liquid-Activity-CoefficientsTable
Model)
PUMP Centrifugal Pump Size and Power
MULPY Stream Multiplicationby a Parameter Heating and Cooling Curves
GCOMP Centrifugal Compressor
Positive Displacement
Compressor and Turbine Profit RAWMT Raw Material, By
I BPR0D By-Product
PR0DT Produce Stream Value
PRBFT Profitability Analysis

followed by numeric values for each parameter. Each
item of data is separated by at least one blank. The
parameter lines can be in any order (they do not have to
be in the same order as the FLOWTRAN units).
4. Input streams--the composition, temperature and
pressure of each stream which is fed to the plant are
entered on lines which closely resemble the parameter
information. The flow rate of each component, in lb hr-’
or lb mole hr-‘, is entered on lines starting with the words
POUNDS or MOLES followed by the name of the stream
and the number of the first component on the line. As
with parameters, the remainder of the line simply has the
flow rate of each of the components entered separated by
one space. The order in which the components are
specified on the RETR control line fixes the index of each
component (i.e. in this example methane is component 1,
ethane is component 2, etc.).
The stream temperature “F is entered on a line beginning
with the letters TEMP and the pressure psi on a line
beginning with PRESS. As with the parameters, the stream
lines may be mixed in any order.
In addition to specifying the fresh feed streams, it is
desirable, but not mandatory, to estimate the composition
of recycle streams. Good initial estimates can greatly
speed the calculation. Recycle stream estimates are
entered in the same way as fresh feed streams.
The streams are followed by the END CASE and END
JOB lines.
The entire input is preceded by TITLE, PROPS and
PRINT control lines. These lines list information which
is used throughout the simulation. The TITLE line is
printed out as a heading on the output report. The PROPS
line specifies the number of components in the simulation
(6 in this case) and the physical property representation
selected (Table 4). Keywords on the PRINT line control
input and output. The keyword INPUT, for example,
produces a report of the imput data.
The output report is in four parts: unit and flow tables,
unit report, stream report and history.
The unit and flow tables are shown in Fig. 7 and are
printed in response to the TABLES specification on the
PRINT card. These tables make it easy to check whether
the units have been correctly specified. This output is
optional.
The unit report, Fig. 8, shows the names of the input
and output streams and the operating conditions for each
 Computer aided chemical process design

Table4. Physicalpropertyoptions

Vapor Vapor Liquid Liquid

Code pressure fugacitycoef. fugacitycoef. activitycoef.

1 Antoine Idealgas Vaporpressure, Idealsolution

Chao-Seaderif
super critical
2 Cavett Redlich-Kwong Redlich-Kwong Regular solution
and Poynting
equations;
Chao-Seader if
super critical
3 Chao-Seader Wilson
(* components,
T,, < 500°F)
4 Grayson-Streed Van Laar
(* components,
T,, < 900°F)
5 Option 2 except Renon
Prausnitz-Shair
for super critical
N2,CO, A, 02,
NO, CH,
7 Renon with regular
solution for un-
specified pairs

UNIT TABLE occurs (for example, if the pump outlet pressure is

specified to be less than the inlet pressure) a warning
UNIT NAME UNIT TYPE
message will be written in the history.
ADO
:: AFLSH FL0wrRANOF’ERATIONUNW
SPLIT
s: PUMP The use in FLOWTRAN of operation units in which
Hl HEATA stream outputs are calculated from stream inputs subject
Rl SCVY
to selected design parameters was a very natural one. At
FLOW TABLE the time of development this allowed FLOWTRAN to use
much of the technology developed for stand alone unit
STREAM NAME FROM TO
operations. Specifically calculating the output streams
BTMS Cl from the input streams means that each of the items in the
FEED Al output stream vector shown in Fig. 1 must be determined
OVHO Fl
ROl Cl Pl before the block’s calculation is considered complete.
I?02 Pl Ml As FLOWTRAN evolved, it was found that this format
R03 I?1 was very flexible and eventually the unit types shown
RO3A I41 i:
so1 Fl in Table 3 emerged. By and large the unit operation blocks
so2 :: Cl do calculate output streams from input streams but in a
INPUT STREAMS number of instances the design parameter (which
corresponds simply to one degree of freedom in the
FEE0 operation block’s equations) is an output stream item. For
OUTPUT STREAMS example, in the compressor block the exit pressure is a
design parameter and the horsepower requirement is
BTHS
OVHD
calculated. Alternatively the horsepower could have been
the design parameter and the exit pressure could have
Fig.7. Unit and flow tables. been calculated. Indeed many of the blocks are design
oriented rather than rating oriented.
unit. The units are arranged in the same order as they
were calculated. INFOKMATIONFEEDBACK-THE CONTROL UNIT

The stream output report, Fig. 9, lists the streams in
alphabetical order. The composition units may be
controlled by keywords on the PRINT card (default is lb
mole hr-‘).
During the course of a simulation it sometimes happens
that an error condition arises. The purpose of the history
is to give the user a complete picture of the course of the
calculations and to highlight any error conditions. Each
time a unit is computed in a simulation a message
is written in the history. Normally the messages serve to
contirm routine operation of the blocks, as is shown for
the first and last iterations in Fig. 10. However, if an error
When the design parameters used in the operation units
do not correspond to the desired design values, FLOW-
TRAN allows the use of “control units”. These units
monitor any item in a stream and allow the manipulation
of the parameters of any upstream unit to achieve the
desired value in the monitored stream. For example,
consider the distillation column in Fig. 11 in which one
component flow rate in SO02and one component flow rate
in SO03 are desired to have certain specified values. A
control unit (block PCVB) monitors the actual value in
each of these streams and manipulates the reflux ratio and
fraction overhead parameters in the distillation unit until
16 E. M. ROSENand A. C. PAULS

10/13/15
FLASH WITH RECYCLE

Al (ADD) T = 98.05 FI p = 100.00 PSIAr V/F = 0.0193 HOLS/MOL

FEEOS = FEED a03
PRODUCT = SO1

Fl (AFLSH) 1 = 35.50 FI P = 25.00 PSIA* V/F f 0.2308 MOLWMOL

FEEDS = so1
BOTTOMS = SO2 * OVERHEAD = OVHD
hEAT IN = 0.0 BTU/HI?

Cl - SPLIT - INPUT I SO2

OUTLETS = ROl BTMS
FRACTION * 0.5000 0.5000

Pl - PUMP - INLET = ROl , OUTLET = R02

FLOweGPM = 16.55 , DELTA P,PSI 8 90.00
FLUID HP = 0.87 t BRAKE HP = 2.94 9 ELEC KY = 2.63
PUMP EFF = 0.2957 . DRIVER EFF = 0.8318

Ill - HEATR - INLET = RO2 . OUTLET s R03A

OUTLET TEMP E 120.00 0EG F . PRESSURE DROP = 15.00 PSI
DUTY = 0.24450 06 BTWHR

Fig.8. Unit report.

10/13/75
FLASH WITH RECYCLE

STREAM NAHEI BTMS FEED ovno ROI R02

LbMOL/HR LBNOL/HA LRMOL/HR LBNOL/HR LBMOL/HR
1 LIET~~ANE 0.06315 3.05367 0.06315 0.06315
2 ETWbNE 0.40248 x81; 2.92330 0.40248 0.40240
3 PROPANE 5.54239 I;.6752 10.3326 5.54239 5.54239
4 N-BUTANE 10.7226 15.0000 4.27735 10.7226 10.7226
5 l-BUTENE 14.0981 21.0000 6.90183 14.0981 14.0981
6 1,3-BUTAOIENE 64.9872 95.0000 30.0127 64.9672 64.9872
TOTAL LBHOC/HR 95.8160 153.318 57.5014 95.8160 95.8160
TOTAL LWHR 5186.77 8038.45 2851.67 5146.77 5106977
1000 BTWHR -045.44 -1062.00 27.91 -045.44 -845.44
DEGREES F 35.50 85.00 35.50 35.50 35.50
PSIA 25.000 100.000 25.000 25.000 115.000
DENSITY, LWFT~ 39.0724 0.0 0.2430 39.0724 39.0724
HOLE fRAC VAPOR 0.0 0.0272 1.0000 0.0 0.0

STREAM NAME; a03 ROJA so1 so2

LBMOL/HR LBNOL/WR LBNOL/HR LBNOL/HR
1 METHANE 0.06315 0.06315 3.17997 0.12630
2 ETHANE 0.40247 0.40248 3.72827 0.80497
3 PROPANE 5.54218 5.54239 21.4174 ll.OB4G
4 N-BUTANE 10.7226 10.7226 25.7226 21.4453
5 1-BUTENE 14.0981 14.OPRl 35.0981 28.1963
6 1,3-WTADIENE 64.9871 64.9072 159.907 129.974
TOTAL LBNOL/HR 95.8156 95.8160 249.133 191.632
TOTAL LB/HR 5186.75 5106.77 13225.2 10373.5
1000 GTU/UR -600.96 -600.96 -1662.96 -1690.87
DEGREES F 120.00 120.00 98.05 35.50
PSIA 100.000 100.000 100.000 25.000
DENSITY, LB/FT3 35.1713 35.1713 0.0 39.0724
HOLE FRAC VAPOR 0.0 0.0 0.0193 0.6

Fig. 9. Str‘earnoutput 1,eport.

the desired flows are achieved. The desired values are
parameters in the PCVB block. The PCVB block is
entirely flexible in manipulating parameters of units which
may be far upstream of the distillation column.
STREAM TEARING-THE CONVERGENCE
As has been noted above, FLOWTRAN requires that
the user specify the order of computation of the units
used to simulate the process. In the case recycle streams
are present he must decide where to “tear” his system and
make an estimate of the tom streams. He may then use a
convergence unit[lO] as was indicated in the example.
These units calculate new output streams (estimated
values) based on their input streams (calculated values)
(Fig. 12). The procedure continues until the input stream
equals the output stream.
Although there has been considerable literature on the
automatic tearing of chemical processes [ 14,181 FLOW-
TRAN’s philosophy has been that the user can generally
select the tearing points with efficiency since his
knowledge of the process can guide him to select the most
suitable streams. Indeed if the streams selected lead to
UNIT poor convergence, then it is easily possible to tear other
streams and use their convergence properties. Automatic
tearing algorithms, if fact, are still being studied. The
recent work of Upadhye & Grens[21], for example, has
shown that the criterion for best tear for direct substitution
is whether the tear belongs to a “nonredundant” family.
Upadhye and Grens divide different tears (called
decompositions in their work) into three different
families; non-redundant, redundant and mixed. Families
are related by means of the Replacement Rule. This rule
states that all input streams to a unit may be replaced by
all the output streams (fresh feeds can be omitted). A
 Computer aided chemical process design 17

10/13/75
FLASH mITti RECYCLE

l * 10/13/75 HH.MH.S.5
Al (ADO) 7 - 85.00 Ft P = 100.00 PSIAq V/F = 0.0272
Fl (AFLSHO 1 = 29.58 Ft P = 25.00 PSIAe V/F = 0.2162
P = 0.0 BTU/W
Cl - SPLITS SO2 INTO 2 OUTLETS
Pl - OUTLET PRES= 115,OPSIA, FLOW= bO.lLE-MOL/HR
Hl - TEHP= 120.OF, DELP= -15,OPSI. G= 0.1670 ObETU/HR
*I?1 -SCVY-IlER= I* TIME= OSECt MAX ERROR=bO,OGGb9
RO3AfNEU) R03 (OLD) ERROR
1 0,69lD-01 0.0 0,069'
2 0.402 i:: 0.402"
3 4.46 4.4619
4 6.45 0.0 6.452'
5 8.75 0.0 0.749.
6 40.0 0.0 39.955'
7 60.1 0.0 60.089"
8 120. 120. 0.0
9 100. 100. 0.0
10-0.3700 06 0.0 **,***
11 0.6380-02 0.0 0.008

. .

(ADO) B 7 98.05 F, P * 100.00 pSIA+ V/F = 0.0193

:: (AFLSH) I 1 35.50 F, P = 25.00 PSIAr V/F = 0.2306
= 0.0 GTWHR
Cl - spL*Tso SO2 INTO 2 OUTLETS
Pl - OUTLET PRES= 115.0PSIA1 FLOV= 95.GLB-CIOL/HR
Hl - TEHP= lZO.OF, OELP= -15,OPSI. G= 0.2440 06BTU/HR
**Rl -SCVY CONVERGED-I= 5vT3 05ECsMAX ERR= 0.00004
R03AfNEH) R03 (OLD) ERROR
1 0.631D-01 0.6310-01 0.000
2 0.402 0.402 0.000
3 5.54 5.54 0.000
4 10.7 10.7 0.000
5 14.1 14.1 0.000
6 65.0 65.0 0.000
7 95.8 95.8 0.000
8 120. 120. 0.0
9 100. 100. 0.0
10-0.6010 06-0.6010 06 0.000
11 0.0 0.0 0.0
**END OF HISTORY
Fig. 10. History for first and last iterations.
,
, -so03 nately a nonredundant family is one that contains those
decompositions which minimize the openings of the loops
1 in the system.)
Consider, for example, the process suggested by
Cavett[2] shown in Fig. 13 and Table 5. This has been
FEED
studied by Upadhye and Grens as well as Shacham &
I___----
Mot&rd[20], and Crowe & Nishio[6]. Here four flash
Parameter 2 drums are interconnected with three recycle streams, Rl,
R2 and R3. Upadhye and Grens point out that decomposi-
tion (Rl, R2, R3) is a member of a nonredundant family.
(Use of the Replacement Rule shows that (Zl, R3) and
(Zl, S3) are also members.) However, although (Zl, 22) is
Fig. 11. Use of a control unit in design.

CONVERGENCE

ESTIMATED CALCULATED

Fig. 12. The convergence unit.

redundant decomposition arises when a stream may be

estimated from a particular tear in more than one way. A
nonredundant family is one which contains no redundant
decompositions. A redundant family contains only redun-
dant decompositions while a mixed family contains both
redundant and nonredundant decompositions. (Alter- Fig. 13. Cavett problem to study tearing.

CACE Vol. I. No. 1-B

 E. M. ROSEN and A. C. PAULS

Table 5. Data and solution for example of Fig. 13

lb mole hr. ’ K-values at solution

Flash Flash Flash Flash
Component Feed Pl P2 1 2 3 4

N 358.2 358.2 0.0009 5.94 24.6 149.7 620.8

co* 4965.6 4964.7 4.73 1.51 4.64 21.1 72.3
KS 339.4 334.7 5.22 0.89 2.03 8.28 27.1
Methane 2995.5 2996.3 0.17 3.09 10.3 52.9 200.1
Ethane 2395.5 2883.2 14.5 1.00 2.66 11.2 39.3
Propane 2291.0 1899.1 404.2 0.502 0.943 3.29 10.8
Iso -Butane 604.1 198.2 410.3 0.310 0.445 1.34 4.22
N-Butane 1539.9 299.4 1247.4 0.246 0.342 0.99 3.07
Is0 -Pentane 790.4 37.2 753.9 0.155 0.164 0.417 1.22
N-Pentane 1129.9 34.8 1095.7 0.126 0.132 0.327 0.944
N-Hexane 1764.7 10.5 1754.6 0.064 0.051 0.107 0.290
N-Heptane 2606.7 3.72 2603.2 0.035 0.022 0.039 0.101
N-Octane 1844.5 0.55 1844.0 0.017 0.008 0.013 0.033
N-Nonane 1669.0 0.12 1668.9 0.009 0.004 0.005 0.012
N-Decane 831.7 0.015 831.7 0.005 0.002 0.002 0.004
N-Undecane 1214.5 0.007 1214.5 0.003 O.ooO8 0.0009 0.002
Temp. (“F) 120 100 85 100 120 96 85
Pres. (psi) 49 800 13 800 270 49 13

itself a nonredundant decomposition it is a member of a
mixed family. Use of the Replacement Rule gives (Sl, S3,
R2, R2) in the same family which is redundant.
FLOWTRAN was used to follow the convergence
properties of the (RI, R2, R3) decomposition (FLOW-
TRAN input shown in Fig. 14) and the (Zl, 22)
decomposition (FLOWTRAN input shown in Fig. 1.5).
Following Shacham and Motard the error for iteration i
was taken (over all components j) as
li = max )ei,J.
resulting from each decomposition was calculated from
*m=lim%.!
I_ 4 ’
where i was taken as 40. As can be seen the mixed family
decomposition has a A, = 0.940 and the nonredunant
family decomposition has a A,,,= 0.913 and consequently
converges faster.
Acceleration in FLOWTRAN is by means of a bounded
Wegstein method[l3] where for each component
(1)

The solution by direct substitution is shown in Fig. 16on a and

plot of error vs iteration.
The maximum eigenvalue for the linearized systems

10/13/?5
TITLE TnREE STREAM TEAR - Rlt RZt A3 - ZERO START

PROPS 16 2 2 3 2
PRINT INPUT
RETR lN2 *CO2 “H2S *Cl lC2 lC3 l IC4 “NC4 .IC5 *NC5 *NC6 *NC7 *NC6 *NC9
*NC10 *NC11
BLOCK A01 ADD FEED Rl R2 4*0 21
BLOCK FL2 IFLSH 21 52 51
PARAM FLZ i 120 270 i
BLOCK A02 ADO SP R3 5.0 22
BLOCK FL3 IFLSH 22 53 H21
PARAM FL3 1 96 49 1
BLOCK FL4 IFLSH 53 P2 R3A
PARAM FL4 1 85 13 1
BLOCK FL1 IFLSH Sl RlA Pl
PARAM FLI 1 100 Boo 1
BLOCK Cl SCVW RI1 RPA R3A A01 Rl R2 R3
PARAM cl 1 0 0 75 1

uOLES FEED 1 358.2 4965.6 339.4 2995.5 2395.5 2291. 604.1 1539.9 790.4
1129.9 1764.7 2606.7 1844.5 1669.0 831.7 1214.5
TEMP FEED 120
PRESS FEED 49
TEMP Rl 120
PRESS Rl 49
TEMP R2 120
PRESS R2 49
TEHP R3 120
PRESS H3 49
END CASE
END JOB

Fig. 14. FLOWTRAN input for Rl, R2, R3 tear of Cavett problem.
 Computeraidedchemicalprocessdesign 19

10/13/75
TITLE TWO STREAM TEAR - Zl*Z2 - ZERO START
PROPS 16 2 2 3 2
PRINT INPUT
QETR lN2 *CO2 lH2S *Cl lC2 lC3 .1C4 *NC4 *It5 *NC5 “NC6 *NC? *NC8 *NC9
*NC10 *NC11
BLOCK FL2 IFLSti 71 S2 51
PARAY FL2 1 120 ii0 -1
BLOCK FLY IFLSH Sl Rl Pl
PARAM FL1 1 100 800 1
BLOCK FLY IFLSH 22 53 R2
PARAM FL3 1 96 49 1
BLOCK FL4 IFLSW 53 P2 R3
PARAW FL4 18513 1
BLOCK AD1 ADD FEED Rl R2 4*0 ZlA
BLOCK AD2 ADD S2 R3 5’0 Z2A
BLOCK Cl SCVW ZlA 22A 0 FL2 21 Z2 0
PARAM Cl 1 0 0 75 1
MOLES FEED 1 358.2 4965.6 339.4 2995.5 2395.5 2291. 604.1 1539.9 790.4
1129.9 1764.7 2606.7 1844.5 1669.0 831.1 1214.5
TEMP FEED 120
PRESS FEED 49
TEHP Zl 120
PRESS 21 49
TEWP 22 120
PRESS L2 49
END CASE
END Jo13

Fig. 15. FLOWTRAN input for Zl 22 stream tear of Cavett problem.

irregular manner. In these experiments no automatic way

of determining when to accelerate was used as was done
by Crowe and Nishio in their application of the general
dominant eigenvalue method.

THE FLOWTRAN PREPROCESSOR

As shown in Fig. 5 the user of FLOWTRAN uses a
simplified “language” to indicate to the system what his
flow sheet configuration lookes like, what parameters he
desires for his operations units and what chemical
components and physical properties he wishes to use. The
preprocessor takes this information and converts it into a
FORTRAN program made up principally of a series of
CALL statements. This FORTRAN program is then
compiled and linked together with those routines which
must be loaded for the particular simulation. This
procedure has a number of considerable advantages:

1. Only those routines which are required for the

particular simulation need be loaded at any one time. A
10 20 30 40 50 60
great deal of space can thus be saved.
ITERATION NUMBER 2. FORTRAN statements can be used interdispersed
Fig. 16. Convergence of the Cavett problem. among the operations block cards.
3. FORTRAN source decks for new blocks or new
where versions of existing programs can be added to the
particular simulation run.
a = f(x”+l) -f(L)
.x,+1 -L In practice the FLOWTRAN preprocessor has been
a great labor saving device. Rearranging or adding units
If q is calculated to be outside the range to a simulation, changing physical property choices, adding
a component or changing feeds or parameters becomes a
O>qr-5 very simple task.

then 4 is set to 0 (direct substitution) which increases the
stability of the iteration. Convergence in FLOWTRAN
takes place when lx. -f(x,)l/x. < 0.0005 for each compo-
nent.
Acceleration was attempted every iteration on the (Rl,
R2, R3) and (Zl, 22) decompositions with the result that
the error bounced around. However, applying the
accelerator only every 4th iteration to the (Rl, R2, R3)
decomposition gave convergence in 23 iterations (Fig. 16).
Applying acceleration only every 5th iteration to the (Zl,
22) gave convergence in 48 iterations but in a very
THE PHYSICALPROPERTY PREPROCESSORS
AND THE PHYSICAL PROPERTY FILE
Probably the most time consuming and difficult portion
of a simulation is the gathering and development of
adequate physical property data. One of the features
which particularly distinguishes FLOWTRAN is its
physical properties,system. Some of the features of this
system are:
1. A physical properties file containing physical
property data on 180 inorganic and organic compounds.
20 E. M. ROSEN and
2. An automatic retrieval of necessary data upon
recognition of a component name in a FLOWTRAN
simulation deck.
3. Use of advanced correlations and equations so that
properties of all classes of compounds can be described
over broad ranges of temperature and pressure.
4. Capability of accurately handling systems whose
liquid mixtures range from ideal (o-, m-, p-xylene)
moderately non-ideal (methanol, propanol) strongly non-
ideal (ethanol, water) to partially immiscible (methyl-ethyl-
ketone, water).
A list of the physical property options available in
FLOWTRAN which can be specified on the PROPS card
was indicated in Table 4. To use these options parameters
of these correlations must be known and be given to the
FLOWTRAN system. For example, to use the Antoine
equation for vapor pressure, In p” = A - B/(T + C), the
parameters, A, B and C must be known. If these values
are not known a pure component data reduction program,
PROPTY, is used to process raw vapor pressure vs
temperature data to give the required parameters.
In general to estimate the required thermodynamic
parameters for compounds not included in the physical
property tile the user must supply the following data as a
minimum: molecular weight, normal boiling point, critical
temperature and pressure, a latent heat, a density and
structural groups for the estimation of gas ideal heat
capacity. The parameters (output of PROPTY) may be
placed directly into the physical property data file.
If vapor-liquid equilibria data are available, then
activity coefficients based on the van Laar equation, e.g.,
can be obtained by least squares fitting the data in a
vapor-liquid equilibria preprocessor called VLE. Both the
pure component preprocessor and the vapor-liquid
equilibria preprocessor are designed to accept a wide
range of data and units.
The capability of handling highly non-ideal solutions in
FLOWTRAN may give rise to a range of problems not
encountered in simulations where only hydrocarbons, for
example, are found. For example, if the non-ideality is
great enough two liquid phases may appear in equilibrium
with a single vapor phase. This means that ordinary
bubblepoint and flash calculations must be capable of
handling this situation. It is especially important when
energy balances are involved since the enthalpy differ-
ence between a two phase and three phase system may be
considerable. FLOWTRAN allows the user to decide if
three phases may be involved and to insert a block which
will tell the user if three phases are present.
Consider for example, a mixture of ethyl alcohol,
benzene and water at 760 mm Hg (composition 0.23, 0.27
and 0.50) whose condensation curve is shown in Fig.
17[10]). At 65°C two phases exist but at 63.7”C three
phases exist. Note how rapidly the amount of vapor
present changes with temperature. This situation is
practically always present in three phase condensation
curves.
The wide range of K values encountered in non-ideal
flash calculations often require considerable numerical
precautions in an otherwise routine calculation. For
example, the flash equation is well known to be
where (Y= V/F and zi is the feed composition for
component i.
A. C. PAULS
601 , t I , I I , , I 1
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.6 0.9 1.0
Fig. 17.Condensation curve for et&l alcohol, benzene and water.
Now if one were to calculate the denominator as
1-cu(l-K,) rather than (I-a)+(~l( then when a =l
one could run into considerable difficulty if Ki were small
enough so that on your computer 1- Ki = 1.
EXPERIENCESIN INDUSTRY
Through a licensing program FLOWTRAN has been
implemented for industrial use in the oil, chemical,
petrochemical, pharmaceutical, and soap and detergent
industries. Access to FLOWTRAN on computers desig-
nated by the licensee has been both through low speed
teletype-like terminals as well as medium speed batch
terminals. Initially, industrial usage generally starts with
single block applications in which confidence in FLOW-
TRAN’s reliability and flexibility is gained. Subsequently,
licensee’s have been able to modify FLOWTRAN to
meet their specific needs, add in new blocks and build on
complementary software without consultation after their
initial training and familiarization.
Use of FLOWTRAN, industrially, has improved design
bases, has resulted in R & D cost savings, has improved
process knowledge, has helped in debottlenecking, has
improved product yield, has reduced operating costs and
has resulted in engineering cost savings. In addition, it has
been used to generate physical property data for other
design programs.
EXPERIENCESINEDUCATION
In 1969 the National Academy of Engineering’s
Commission on Education established the CACHE
(Computer Aids for Chemical Engineering Education)
committee. In cooperation with this committee Monsanto
implemented the FLOWTRAN system at United Comput-
ing Services in August 1974.In October 1974the user text
describing the system became available[19]. As of early
1975 some 21 educational institutions were using
FLOWTRAN for educational purposes. Also a user
group has been formed within CACHE. Seminars on
FLOWTRAN have been presented through universities
as well as in AIChE’s Today Series. Initial usage in the
classroom has generally been in design courses.
CONCLUSIONS
The experiences with FLOWTRAN both within Mon-
santo and among the licensee’s of FLOWTRAN gener-
ally point to three areas which appear to be especially
important in the design of a chemical process simulation
system.
1. The physical property system should be broadly
based, easily used, easily checked and easily modified. A
data bank of commonly used materials can pay rich
dividends in user’s time and avoidance of costly errors.
2. A simple user’s language which requires no
 Computer aided chemical process design
programming and little data preparation are prerequisites
for wide adoption.
3. A high degree of reliability of each routine,
considerable error checking facilities and dedicated
maintenance of the system are absolute requirements for
giving the required credibility to the system.
REFERENCES
1. Agenda for AIChE Machine Computations Committee Work-
shop on Heat and Material Balances, 27 Sept. (l%O).
2. R. H. Cave& Application of numerical methods to the
convergence of simulated processes involving recycle loops.
American Petroleum Institute Preprint, No. 04-63 (1%3).
3. R. H. Cavett, FLOWTRAN physical properties. Paper
presented at 49th NGPA Annual Convention, 17-19 March,
1970,Denver, CO.
4. Chemical and Engineering News Staff, Process designers get
more software. Chem. Engng News 48, 14 (1970).
5. Chemical Week Staff, To each engineer, his own computer.
Chem. Week 106, 11 (1970).
6. C. M. Crowe & M. Nishio, Convergence promotion in the
simulation of chemical processes-the general dominant
eigenvalue problem. AIChE .I. 21, 528 (1975).
7. L. B. Evans, D. G. Steward & C. R. Sprague, Computer
aided chemical process design. Chem. Engng Progress 64,
39-46 (1%8).
8. J. R. Flower & B. D. Whitehead; Computer-aided design: a
survey of flowsheeting programs-I. The Chem. Engr, No.
272, Part II, No. 273 (1973).
9. R. E. Harris, FLOWTRAN: an approach to a computer aided
21
process analysis and design tool. Paper presented to the
American Chemical Society, Los Angeles, CA (1971).
lb. E. J. Henley & E. M. Rosen, Material and Energy Balance
Computations. Wiley, New York (1969).
11. Institute News Monsanto aids CACHE. Chem. Engng
Progress 70, 75 (1974).
12. M. G. Kesler & M. M. Kessler, Engineering a process with a
computer. World Petrol 29 (60) (1958).
13. H. C. Kliesch, A study of convergence accelerator algorithms
used in steady state process simulation. M.S. Thesis in
Chemical Engineering, Tulane University, New Orleans, LA
(1967).
14. R. L. Motard, M. Shacham & E. M. Rosen, Steady state
chemical process simulation. AIChE J. 21, 417 (1975).
15. M. F. Nagiev, Chem. Engng Prop. 53,297 (1957);The Theory
of Recycle Processes in Chemical Engineering. Pergamon,
Oxford (1964).
16. R. L. Rorschach & R. E. Harris, Process simulation made by
computer. Oil & Gas J. 68, 33 (1970).
17. E. M. Rosen, A machine computation method for performing
material balances. Chem. E&g Progress 58, 69 (1%2). _
18. D. F. Rudd & C. C. Watson, Strategy ofProcess Engineering.
Wiley, New York (1%8).
19. J. D. Seader, W. D. Seider & A. C. Pauls, FLOWTRAN
Simulation: An Introduction. CACHE Corooration. Ulrich’s
Bookstore, Ann Arbor, Michigan (1974). _
20 M. Shacham & R. L. Motard, Application of the theory of
linear recycle systems. Paper presented 78th AIChE National
Meeting, Salt Lake City, 18-21 Aug. (1974).
21 R. S. Upadhye & E. A. Grens, Selection of decompositions
for process simulation. AIChE J. 21, 137 (1975).
22. M. A. Vela, Use fractions for recycle balances--I & II.
Petroleum Refiner 40(5), 247; (6), 189 (1961).