JOURNAL REVIEW

Steady State Chemical Process Simulation

In recent years process simulation by computer has become an accepted

tool in the design and control of chemical processes, but there are still un- R. L. MOTARD
certainties and conflicting views about the real benefits relative to cost in M. SHACHAM
both chemical industry and in education. Department of Chemical Engineering
This article surveys the literature of the last three years, exploring the University o f Houston
answers to the question: How has chemical process simulation by computer Houston, Texas 77004
fulfilled expectations in industry and education and what are the major
areas of research identifiable at this time? and
It is concluded that process simulation remains a very dynamic field with E. M. ROSEN
a continuous expansion in the areas of application and in the number of
users. New subjects requiring research are replacing older areas which have Monsanto Company
matured. St. Louis, Missouri 63166

SCOPE
The objective of this study is to investigate the different a rational topology supplied by the user, sequences the
techniques for process simulation. It may provide valuable computations, and supervises data input and output of
introductory information to potential users, prospective results as well as the iterative recycle computations.
developers, or beginning research workers in the field. A complete simulator combines expertise from three
The development of simulation applied to chemical areas. The preparation of accurate mathematical models
processes began in the mid 1950’s and the first simulator of chemical processing units and good correlations for
was published in 1958. In the mid 1960’s the most popu- thermodynamic and physical properties is an engineering
lar current structure for process simulation, namely, the problem. The design of efficient input and output pro-
modular program, was crystallized. According to this cedures and program/library organization is a data proc-
approach each chemical processing scheme can be repre- essing problem and the iterative calculations are numerical
sented by linking together a set of standard mathematical problems.
unit modules in arbitrary and flexible networks.
The main components of a modular simulation pro- Several simulation programs are in routine use in service
gram are: the unit module library of subprograms, the bureaus and largely within companies which have devel-
thermodynamic and physical property estimation library oped them. In the field of education they are used to
(TP), and the executive program with its library of teach process simulation, in support of teaching other sub-
services. The unit module subroutines compute physical jects such as stoichiometry and unit operations, and in
outputs for each standard processing operation given the the preparation of design projects.
physical inputs and the design (or decision) parameters. Recently, most research has concentrated on the areas
The necessary TP data are supplied automatically to the of dynamic simulation and the optimization of process
unit modules with some user control over the estimation simulations. The research literature of the last three years
optiom. The executive program manages the data libraries is reviewed and the directions for further progress are
(data bases), connects between unit modules according to noted.

CONCLUSIONS AND SIGNIFICANCE

Process simulation programs have indeed acquired tool extensively and in the heavy demand for consultation
some stature in process engineering. One major benefit, from the developers of the program. Growing interest in
in addition to better designs at lower cost, is the improved the use of such programs and in research at universities
communication between different engineering groups will help in providing more trained manpower and stan-
using the same simulator. It also acts as a focal point for dardizing and improving performance.
the dissemination of collective expertise from many The senior design project is the most important educa-
quarters: engineering, scientific, and mathematical. There tional use of simulation programs and the availability of
are still some problems in training engineers to use this this tool assists the student in translating all of his funda-
mental knowledge into practical and economically feasible
designs. The instructor in such a course faces difficulties
Correspondence concerning this paper should be addressed to E. M.
Rosen. M. Shacham is with the Department of Chemical Engineering, because of lack of realistic, nontrivial problems which are
Ben-Gurion University, Beersheva, Israel. simple enough to be mastered by the students. Industrial

AlChE Journal (Vol. 21, No. 3) May, 1975 Page 417

collaboration in project work, where it can be developed, of process engineering is also underway, Graphic, inter-
can improve the realism by drawing on resources of solved active displays will gain more importance because of the
designs in company files. latter as well as reduced cost. Academic research will
In industrial process simulation, a major effort will concentrate on process optimization, applications to re-
probably concentrate on extending the physical property source conservation, and dynamic simulation.
data bases and the reliability of estimation methods. This The rapid development of computer hardware, com-
will be particularly true of polar components, solids, and munication, and software will suggest new types of prob-
polymers. Work linking flow sheet simulators to the lems to be tackled and new ways of organizing projects.
mechanical design, specification, and construction phases The field will remain dynamic for some time.

INTRODUCTION information flow between the unit modules.

In recent years, process simulation has become an ac- Although the specific program may be more efficient
cepted tool for the design and understanding of chemical from the computational point of view than the modular
processes and almost a requirement in chemical engi- one, the use of specific programs is limited to special
neering education. The rising interest in this area is well cases because of their lack of flexibility.
exemplified by the number of process simulation programs A simulation program may be used for steady state or
which are offered for sale to the chemical industry dynamic simulation. In the first case, a material and en-
(Hughson and Steymann, 1971, 1973a, b) and by the ergy balance for steady state operation of the chemical
sharp rise in the number of published research works in plant must be done. In the second case, the time varying
this field. operation of the plant is simulated. Dynamic simulation
It is time to look at the progress which has been made is the more complicated and computationally expensive of
in the area of steady state process simulation since the the two.
publication of the first process simulator in 1958 (Flex- Although the aim of the process simulation program is
ible Flowsheet) and to explore two questions: to be a tool for design, most of the simulation programs
1. How has chemical process simulation fulfilled es- work in the simulation/performance/rating mode. It is
pectations in industry and education? characteristic of such a mode of calculation that all sys-
2. What are the major areas of research to extend the tem inputs and design parameters for the units are spe-
utility of the computer in this field? cified, The information flow in the simulation program is
Our aim is to review the literature for the past three in the same direction as the heat and material flow in the
years in order to answer these two questions. Some tutorial chemical plant. Ideally, in the design mode of calculation,
material is presented to resolve terminology problems. the system inputs and/or design parameters are calculated
Where references are not explicitly given, the reviewers from specified outputs.
are attempting to distill the significance of the literature The simulation mode calculation is generally more
or their own experiences. The literature of the previous stable numerically than the design mode, but engineers
years is well covered by the recent review articles of are more interested in design problems than in simulation
Flower and Whitehead (1973a, b ) and Kehat and Sha- problems. The usual solution to this dilemma is to per-
cham (1973a, b ) . The topic is also discussed in Perry form design calculations by iterated simulation. Iterated
and Chilton (1973). simulation may take place over an entire process with the
Terminology user scheduling the cases to be computed or, it may take
Process simulation is the representation of a chemical place internal to the simulation through the use of con-
process by a mathematical model which is then solved to trol blocks. In the latter mode, a program is really a hy-
obtain information about the performance of the chemical brid between simulation and design. Control blocks iterate
process. The mathematical model is usually a computer subsystems consisting of one or more process modules by
program. This computer program is generally known as a manipulating specific design variables to meet various de-
chemical process simulation program or a flowsheeting sign requirements imposed a priori by the user, such as
component ratios, degree of vaporization, temperature,
program. The chemical engineer mostly deals with con-
tinuous, deterministic system simulation, but may also total amount of a stream, or recovery of a product as in
simulate discrete events and stochastic systems ( Briddell, distillation or absorption. Obviously, there are at least
1974~). two classes of design variables. There are those that are
Kehat and Shacham (1973a) classify process simula- fixed by noneconomic factors and derived from physical
tion programs according to their structure and possible or chemical constraints and those that may be manipulated
uses. The simulation program may be a specific one pre- freely on the basis of the overall economics of the
pared to simulate a particular process with a fixed plant process. One variable that is always constrained in a proc-
layout. Such a program may be solved efficiently using ess simulation is pressure. One can compute pressure
sophisticated numerical techniques since the simulation drops at various points in the system according to current
may be treated as a mathematical problem with fully-de- operating conditions. However, at other points one must
fined equations and constraints. A more widely-accepted restore pressure datums via a pump/compressor or valve/
type of simulation program, however, is one which is expander. Otherwise the simulation becomes meaning-
constructed using a modular approach. According to this less.
method each chemical processing step is represented as a Some process simulation programs perform optimiza-
separate mathematical model called a unit module. The tion, equipment sizing, or economic evaluation. In many
unit modules are connected by data sets which represent cases, the connection between the programs which per-
the streams of material and energy flowing between the form these tasks and the process simulation program is
units of the plant. An executive program supervises the weak, frequently ad hoc.

Page 418 May, 1975 AlChE Journal (Vol. 21, No. 3)

 TABLE1. GENERALREFERENCESTO PROCESSSIMULATION

Hughson & Kehat & Flower &

Briddell Steymann Shacham Whitehead Wells & Robson Crowe et al.
Reference/subject Mah ( 1974) ( 1974a,b,c) (1973a,b) (1973a,b) (1973a,b) (1973) (1971)

1. Historical review Yes No No Yes No No No

2. List of existing programs No No Yes Yes Yes NO Some
3. Terminology No Yes No Yes Yes No Yes
4. Description of structure Yes Yes No Yes Yes Yes Yes
of simulation program
5. Description of simulation Yes Some No Yes Some Yes Yes
techniques

Classification and Scope The design of a process simulation program has three
Since the modular steady state process simulation pro- major aspects, which may be classified as follows:
gram is the most widely used, the discussion is limited to 1. Data processing. All of the tasks dealing with the
this type. In addition, as we are interested in the problems input data itself, the best methods to test, store, and re-
of preparation and use of these progams,'we do not give trieve them, and the possibility of changing the data in an
details of their historical development nor the charac- interactive environment are in this category. The Struc-
teristics of existing program. Discussion on these subjects ture of Simulators discusses various aspects of data proc-
may be found in the general references summarized in essing.
Table 1. This table may also be useful for someone who 2. Engineering models. The preparation of an accurate
wants to be more familiar with the terminology, the struc- mathematical model of the chemical process units and
ture, and the techniques which are used in process simula- preparation of good correlations for thermodynamic and
tion programs. physical properties requires in-depth knowledge of the
In a typical process simulator the input data contains chemical processes. In Unit Module Algorithms and Ther-
the process topology, design and operating parameters of modynamic and Physical Properties the different ap-
the units, and the composition and state of the process proaches to engineering models are discussed.
feed streams as well as perhaps an estimate of any re- 3. Numerical procedures. These situations appear in
cycle streams. It is read and checked by the input sub- the computation-ordering algorithms, in the recycle cal-
routine of the process simulator. The input data is trans- culations, and also in the iterative solution of the more
ferred to the executive program which determines the complicated unit modules. This aspect is discussed in
order of calculation. Then the executive program calls the section on Computational Efficiency.
the unit module subroutines according to this order, and The user of a process simulation program may face a
passes the required input information to these sub- number of difficulties. Sometimes the physical property
routines. When one or more recycle streams are detected, data for some of the components used in the simulation
the composition of which cannot be known in advance, are lacking. In addition, the different word sizes, operat-
the executive program gives an estimated initial value ing systems, and compilers used on different computers
for the composition and state variables of these streams may lead to difficulties in using the same simulation pro-
and proceeds to obtain the correct values for these streams gram on these different computers even when the pro-
by some iteration method. gram is written in a high level language.
The unit module subroutines calculate the physical In Experience with Simulation Programs, we discuss the
output from the units when the physical input and the uses of chemical process simulation in industry and edu-
unit design parameters are known. For this calculation, cation, and we explore the problems which limit further
the unit module subroutines needs physical and thermo- extension of their use.
dynamic data. These data are obtained from the thermo- In Future Development, we propose some possible solu-
dynamic and physical property package which contains tions to the problems raised in this paper, and we forecast
estimation parameters for the required properties. (Esti- the direction of future development in the field of chem-
mation parameters are those pure component parameters ical process Simulation.
used in correlation equations for physical property rep-
resentations, for example, the Antoine coefficients for va- STRUCTURE OF SIMULATORS
por pressure, critical properties for use in equation of General simulation programs have been constructed in a
states, etc.) After the chemical process has been com- number of ways. Although they all use unit modules for
pletely calculated, the output unit converts the large their basic calculations, they can be viewed as having one
amount of information obtained into the form of an easily of two basic internal structures-fixed structure or vari-
readable technical report. able structure. In fixed structure programs, the executive
The preparation of a computer simulation is usually program is exactly the same no matter what is being simu-
controlled by the requirement that the program be able to lated. The matching of streams to unit modules and the
solve most process problems with tolerable accuracy for path of calculations through the program is determined by
a reasonable cost in user and computer time. The state of data supplied by the user. In variable structure programs,
the art is such that it is often possible to perform a large a different executive program is written or generated for
each different flow sheet. If the executive is written by
number of simulations and still achieve both of the above the user in FORTRAN, this will generally consist mostly
requirements. However, there are process simu1at:on situa- of calls to the various unit modules or subprograms. Al-
tions where this is quite difficult. Still, the judicious use ternately, a problem oriented language (POL) may be
of a simulation package can greatly increase engineering used which is interpreted directly into executable code, or
productivity, although at higher cost per person-hour. the POL may in turn generate a procedure-oriented lan-
The cost per project and the elapsed time, however, can guage (often FORTRAN) which in turn is compiled,
be substantially lower. linked, and executed.

AlChE Journal (Vol. 21, No. 3) May, 1975 Page 419

 Fixed structure programs have one signscant advan- COOLER
tage: they generally require less computer time for short FIRED REACTOR HEAT DISTILLATION DIVIDER
HEATER EXCHANGER COLUMN
simulations. They have a host of disadvantages, however.
Among them are lack of flexibility and overcomplexity
as they grow to incorporate more unit operations, broader
physical property capabilities, and wider output options.
Further, they require that all unit modules be loaded into
the simulation no matter how trivial the simulation prob-

-A-
lem being solved.
A variable structure program that uses a POL and

vq!)
generates a FORTRAN main program has a number of ad-
vantages. Only those modules actually needed are loaded,
FORTRAN statements may be added directly by the
user, and data storage space may be allocated only as
needed by the simulation. In return for the extra over-
head that the translation, compilation, and linkediting ~

requires, much greater flexibility and modularity can be

obtained.
Regardless of the structure used, we can distinguish 8

four main phases in modern simulation programs: FIRED DIVIDER

HEATER
1. Input phase Fig. 1. Chemical process flow sheet.
2. Preprocessing phase
3. Calculation phase
4. Output phase
13
In this section we discuss some aspects of each of these
phases and some implications of the structures on these
aspects.
input Phase
This phase is the most critical interface between the
user and the simulation program and therefore has re-
ceived considerable attention. The less the input data
required, the easier it is to use and the fewer the user
mistakes.
In the input phase the following steps are executed:
data input, data checking, and data storage.
The data required by all process simulation programs Fig. 2. information flow sheet.
are :
1. Process topology
2. Feed stream information, including physical prop- UNIT NO. ASSOCIATED STREAMS
erty data
3. Design parameters of the units I I -2
4. Convergence criteria 2
3
Some programs may require: 4
5
5. Calculation order list 6 6 -0 -7
7 10 -11 -12
When economic evaluation and optimization is being car- 8 9 -10
ried out, there may also be a need for:
Fig. 3. Process matrix.
6. Cost parameters
7. Optimization criteria.
The required information may be entered in a variety In programs with variable structure, the process topol-
of ways, depending on the approach used and the internal ogy may be explained by sentences like (Nott, 1971) “In-
structure of the program. Generally, however, the proc- put 1 of column 4 is output 2 of heat exchanger 3.” A
ess flow sheet is translated into an information flow sheet POL approach is obviously employed here.
(Figures 1 & 2 ) . In the information flow sheet, one uses Process topology may be expressed directly on the
the name of the subroutine which will be used for calcula- FORTRAN level by use of stream (information) vectors.
tion of the appropriate equipment instead of the name of (The stream vector contains at least the flow of each
the chemical equipment. The interconnections between component, the stream temperature, pressure, phase ratio,
the unit modules may now represent information flow as and generally total enthalpy.) Mixing of several streams
well as material or heat flow. may be expressed as
One way for mapping the process topology into com- CALL MIX (Sl,S2, S3, NONE, NONE, SOUT)
puter data is the use of a process matrix (Fig. 3 ) . In this
matrix, for every unit there is the list of the identification Here the first five arguments are stream names which are
numbers of the input streams (positive) and output input streams to a MIX module. (The stream N O N E is a
streams (negative). This method is used in simulation zero stream vector.) The resultant stream SOUT is the
programs wi,th fixed structure. output stream. The input and output stream definitions are

Page 420 May, 1975 AlChE Journal (Vol. 21, No. 3)

part of the unit module definition. 2. THEINPUTFORMOF DIFFERENT
TABLE
The mapping of unit modules or subroutines onto the SIMULATIONPROGRAMS
actual process is not always one-to-one as indicated in
Figures 1 and 2. It is often necessary, or desirable, to use Simulation
fewer or more subroutines to model a particular plant sub- program Input form Reference
system. This may be necessary because the functions de-
scribed by the modules must be collected to represent a PROVES Fixed Klumpar, 1970a,b
hardware unit, or it may be desirable because the user PRIMER FORTRAN Wells and Robson,
wishes to constrain the simulation during the early de- NAMELIST 1973
velopment of a model. For example, a flash module at a CONCEPT I11 Conversational CONCEPT, 1973
specified temperature and pressure might be used in place CAPES User’s Written Maejima and Shindo,
Main Program, 1973
of a compressor and interstage cooler. Assembling a suit- FORTRAN
able model from the functional unit modules may be a Extended U.P. User’s Written Hanyak and Klaus,
difficult task and requires experience and skill to make the PACER Main Program, 1973
most effective use of the functions provided in a par- POL
ticular simulation program. SLED Users Written Nott. 1971
The feed stream information contains the following Main Program,
data: Components specification and flow rate, tempera- POL
ture, and pressure. Often the component’s specification FLOWTRAN POL Seader et al., 1974
must be given by a code number (Motard and Lee, 1971).
This number connects the internal dictionary of com-
ponent names, the physical property package and data If a POL is being used, the data may be in the form of
bank, and the input stream. Alternately, the component key words.
specification may be given by name, and the ordering In Table 2 the input form for some of the more recent
provides the connection between the physical property simulation programs are summarized.
package and the input stream (Loibl et al., 1973). The Although the fixed-format input form is the most primi-
permissible units (for example, O F , lb./hr.) for feed tive one, this form has often been used for routine in-
stream data vary among simulators. Virtually all allow dustrial use. Well-documented standard forms must be
English units; some allow optional use of units such as prepared, and the engineer must introduce the data into
BPD, 1b.-moles/hr., etc., or SI units. these forms. This form of input is convenient when there
The design parameters of the units are the necessary is a large amount of input data.
data which make it possible to calculate the physical
output from the physical input. These are parameters For educational purposes, the conversational input
like heat transfer area for heat exchanger or the number form is preferred since there is considerable prompting
of equilibrium stages for a distillation column. This data for the inexperienced user. However, this form of input
must be given according to a specified location in a vec- is time consuming (Loibl et al., 1973) so one should
tor for each use of a particular module. provide an additional input form which may be used for
The convergence data refers to recycle calculations. large simulation problems by experienced users.
The user must specify which convergence-acceleration Data Check. In most of the simulation programs the
method to use on which streams, the convergence cri- data checking must be done by the user. The input data is
teria, and the maximum number of iterations to be at- printed out by the computer and the user must check it.
tempted. Users rarely have enough experience and knowl- This check is often not sufficient, and more sophisticated
edge to choose reasonable values for the convergence checking must be done by the simulation program. Flower
data. This problem will be discussed in more detail in and Whitehead (1973a, b) have found that some of the
the section on Computational Efficiency. industrial simulation programs (FLOWPACK, 1970;
The problems in the input of additional information are PACER 245, 1971; FLOWTRAN-see Seader et al.,
the same, so we shall not discuss this further. More details, 1974) have comprehensive checking facilities
however, may be found in the User’s guides of simula- Much more attention must be given to the logical check-
tion programs (Motard and Lee, 1971; Seader et al., ing of the data. It is often impossible to discover from
1974) or in textbooks (Crowe et al., 1971; Wells and the computer results that the source of some error is due
Robson, 1973). to absurd input data.
For the simulation of the same rocess, all simulators Storage of the Data. The input data as well as cal-
P
require about the same amount o data. However, this
input data may be entered in different modes and forms.
culated values are usually stored in arrays with fixed
boundaries. In fixed-structure programs, due to the limita-
1. Batch mode or time-sharing mode. tions of the FORTRAN computer language, the storage
If entered in the time-sharing mode, prompting may be space must always be allocated according to the maximal
present or available on request. If the user can change capacity of the simulation program with no connection
the sequence of execution logic during the calculation with the actual problem size. This leads to the situation
phase, interaction is present. If both prompting and in- that relatively large simulation models (like the Natural
teraction are present, it is called conversational interac- Gasoline Plant simulation, p. 119 in Motard and Lee,
tive. 1971), actually use less than 20% of the allocated stor-
age. Such use of the storage space reduces the efficiency
2. Fixed format or free-field format and flexibility of simulation programs. This problem may
By fixed format it is meant that the data are entered in be solved partially by the use of main programs written
fixed columns on the input cards. In free-field input, the by the user or POL generated main programs (Seader et
data need not be entered in any specific columns. Free- al., 1974), since the boundaries of the arrays are declared
field input is always list-directed. according to the actual problem size. Another approach is
3. FORTRAN NAMELIST to use secondary memory for part of the data (Worley
The NAMELIST feature of FORTRAN is a free-field and Motard, 1972).
form in which selected elements of a list may be supplied. A further approach to the solution of this problem may
4. Numbers or instructions be the use of dynamic storage for the data. This approach

AlChE Journal (Vol. 21, No. 3) May, 1975 Page 421

makes it possible to allocate the precise amount of main The Calculation Phase
memory space for every model. Creation of such data Most of the problems associated with the calculation
storage may be difficult in FORTRAN, but it can be easily phase will be discussed later. We are interested at the
done in PL/1 (with some increase in overhead execution moment in discussing how the simulation program must
time). be constructed to be flexible enough and to execute the
calculation phase most efficiently for a wide selection of
The Output Phase simulation problems and computers.
There are two types of output from a standard simula- The calculation phase is executed by the unit module
tion program: intermediate output and final results. The subprograms with the aid of the thermodynamic and
form of the final results pages are usually fixed (Flower physical property package and the supervision of the ex-
and Whitehead, 1973a, b ) . A stream table is printed con- ecutive program. To solve a wide selection of simulation
taining the composition, flow rate, temperature, enthalpy, problems, there must be a large number of unit subpro-
pressure, and perhaps density and other information for grams for different equipment items and also some differ-
every stream. ent subprograms for the same item with different levels
The calculated operational data of the different equip- of sophistication. This approach may lead to a very large
ment is also printed (for example, like heat duty and pres- number of unit subroutines (150 for PACER 245). As
sure drop for heat exchangers, usage of utilities like cool- indicated, this is a disadvantage for a fixed-structure pro-
ing water, steam or fuel, etc.). When there is also an eco- gram since, when the simulation program grows to such
nomic calculation, the results of this calculation are a magnitude, it cannot be stored in the main memory of
printed in a standard report format (Klumpar, 1970a, b ) . even a big computer. One approach to the solution of this
If the required results cannot be obtained for any rea- problem may be the use of multipurpose modules which
son, it is very important that the simulation program have contain different calculation possibhies in the same unit
a tracing option. This option makes it possible to find the subprogram. However, this approach does not seem to be
exact source of the failure. Although most of the existing very promising due to the complexities introduced in the
simulation programs have some type of tracing options module.
(Flower and Whitehead, 1973a, b ) , the standard tracing A widely accepted method is the use of overlay structure
options are often not adequate, and the user must intro- (Worley and Motard, 1972). In every time period, only
duce tracing statements of his own in the system to lo- that fraction of the simulation program which is really
cate the problem. More attention must be given LO this being executed is in the main memory. All the other frac-
difficulty. tions are on a secondary memory device and transferred
to the main memory when needed. With a good overlay
Most of the existing simulation programs also give error structure, large simulation programs may be executed in
messages and warnings about difficulties that the simula-
a relatively small storage space (PACER 245 may be run
tion program cannot solve, but these may reduce the ac- in 36K words of storage), but the penalty for using less
curacy of the calculation or may lead to the failure of the storage space must be paid for by an increase in computa-
program in a later stage. tion time. The transfer of fractions of the program to and
In addition to the final printout, several programs may from the main memory may be a time-consuming process.
write final or intermediate results on files, or may punch Time-sharing executive systems which swap fractions of
them on cards (Motard and Lee, 1971). This makes it the simulation program both ways can result in one-half
easier to use this information later on to restart the pro- the memory requirement of a batch operating system
gram. which only retrieves (Worley and Motard, 1972).
The Preprocessing Phase For the execution of most simulation problems, only a
The boundary between preprocessing and calculation small portion of the simulation program may be in use. It
phases may be a little fuzzy, but some activities clearly is better to load only those subroutines which are actually
belong to the former. Where the input data has been in- needed. This may lead to a reasonable improvement in
troduced via a POL or user-written FORTRAN main storage and time efficiency. So far only a few simulation
program, the precompilation (POL) and subsequent com- programs have a limited possibility of this type. These
pilation and linkage editing of the target simulation pro- programs are usually constructed from several subprograms
gram to job specifications is done in this phase. which may be separate or linked together to execute a
more complicated task. Examples of this type of opera-
. Other procedures during this phase include the retrieval
tion are PROVES (Klumpar, 1970a, b ) , and the CHESS-
of physical property estimation coefficients from an archi- CHEEP-PRODYC system (Worley and Motard, 1972).
val data base and the computation of missing estimation The simulation programs in which the user must write
coefficients for nonstandard components or special estima-
the main calling program either directly or via a POL may
tion procedures from raw data supplied during the input be a good basis for extending the use of this approach. In
phases. If an archive of phase equilibrium data is avail- such a main program only the subprograms actually used
able, this may be searched for the required chemical are declared, and subsequently linkedited.
species and temperature polynomials curve-fitted to the Implied in these comments is the dependence of the
retrieved data. Road-map options will be set in the theimo- structure of the simulation program on the operating sys-
dynamic and physical property package. tem, hardware, and language technologies available to the
Finally, if the system includes automatic calculation development team. The current virtual memory technology
ordering this will be done in the preprocessing phase, even (Parmelee et al., 1972) completely eliminates the need
prior to precompilation (POL) since the resultant ma- for overlay or program structuring by the developer since
chine generated FORTRAN main program is structured the code is paged into memory from secondary storage
according to the ordering results. automatically as needed. However, virtual memory con-
Since the preprocessing phase is executed only once cepts are now viewed as not the only solution to large
for the entire simulation run, it is convenient to transfer program processing although they will remain important to
as much of the calculations and logical decisions to this minicomputers and small machines. Very large computers
phase as feasible. This will improve the efficiency of the may also enjoy very large main memories using cheap
calculation phase which most likely will be iterative due semiconductor technology which can hold entire active
to process recycles. programs in main memory at once.

Page 422 May, 1975 AlChE Journal (Vol. 21, No. 3)

 UNIT MODULE ALGORITHMS TMLE3. THEBASICMODULESET
Equipment, rocess, or unit modules are designed to
F
provide most o the functional operations one might rea-
sonably encounter in a typical chemical process. In their
Function

1. MIXER Mix n input streams to pro-

elementary form they are simulation-based, that is, they
duce one output stream
compute physical output streams from given physical 2. DIVIDER Divide one input stream into
input streams and equipment design parameters. They n output streams of the
are designed to be used at several nodes in any process same composition according
network, and each instance of such use is associated with to specified stream fractions
an equipment design parameter vector specific to that 3. SPLITTER or SIMPLE Separate one input stream into
node, such as a heat transfer coefficient or a number of SEPARATOR two products with speci-
equilibrium stages. The equipment parameter vector may fied component distributions
also be used to store output from a module, such as a 4. FLASH or PHASE DE- Perform isothermal or adia-
heat duty requirement. TERMINATION batic flash or phase separa-
More complex modules may be design-based, that is, tioddetermination
ou ut streams are computed to meet some criterion of 5. HEAT EXCHANGER Add or extract heat to reach
8
pe onnance by manipulating design parameters, such as
reflux ratio to meet recovery specifications or heat duty
5.1 Simple
5.2 Two-stream
specified temperature or
by exchange with second
stream with specified ge-
to meet a temperature specification or degree of vaporiza- ometry and heat transfer
tion. No simulation program today operates in the pure coefficient
simulation mode since some design conditions are so com- 6 . PRESSURE CHANGE or Adjust pressure upward or
manding and generally accepted that there is no point in PUMP/VALVE downward to specified out-
ignoring them even at the early stages of modeling. Slight put pressure
inconsistencies in mass and energy balances at this stage 7. REACTOR Stoichiometric, isothermal re-
may be resolved in later computations. These notions will actor according to specified
be discussed later in this section. Three books have ap- conversion level
peared on the steady state modeling of processes-crowe
et al. (1971), Wells and Robson (1973), and Seader et
tent. The more common failures must be anticipated and
al. (1974). In addition, the seven CACHE Committee
handled in a rational manner, such as modules that at-
computer program books ( 1973) should provide resources
for unit modules. tempt to compute with zero inputs and module con-
vergence tests based on relative errors with zero divisors.
Bosic Concepts Every module will be used in some process recycle situ-
Input stream information and output stream identifica- ation where the common assumption is that all intermedi-
tion is communicated to a unit module in at least two ways, ate process streams will be initialized to zero unless spe-
either ( 1 ) as a list of unique stream code numbers or cified by the user. Here, the input specification of inter-
names in a subroutine calling sequence, or ( 2 ) through a mediate process streams must be done with full knowledge
COMMON data interface identical for all unit modules. of the calculation order since initialized values may be
The former approach is characteristic of POL or user-writ- destroyed when a stream appears as an output of a zero
ten main program systems such as FLOWTRAN (Seader input module before it is used as an input in the intended
et al, 1974) and the latter is used, for example, in systems sequence. Thus, it is useful to make one pass through the
such as PACER 245 and Concept 111. automatic ordering routine, perhaps with dummy equip-
Each simulation program generally includes a universal ment modules, or to override the automatic ordering fea-
basic set of modules shown in Table 3 which should allow ture with the user’s own calculation order so that the ef-
any process to be modeled at an elementary level. fect of the initialization of intermediate streams is clearly
appreciated. The ruling principle in all cases is to mimic
The basic set must be augmented by a few services per- the start-up of an actual plant. A heat exchanger with one
formed by the thermodynamic and physical property pack- input stream undefined will pass on the alternate stream
age as shown in Table 4. These functions are called only unchanged. A fractionator with latent condenser or re-
by the unit modules. boiler operation will dump its feed out the top or bottom
The most important characteristic of a unit module is depending on feed condition. Thus, a well-designed simu-
that it should compute mass and energy balances accord- lation program will chatter with diagnostic messages as
ing to the designer’s intent. Stream compositions are almost the recycle streams fill up but should run smoothly as
universally transmitted in moles/unit time rather than steady state is approached.
mole-fractions. The next most important quantity is total The most difficult processes to start up in simulation are
enthalpy content. It is less important whether temperature those with strong energy-coupled loops (see hlenzies and
and phase condition are consistent at intermediate stages Johnson, 1971, for an example). Initialization of inter-
since these can always be determined by interposing an mediate conditions is almost always demanded, other-
adiabatic flash module at some critical point in the net- wise parts of the process will remain dry if the mass flows
work. However, the best module writers will always verify are not there to carry energy.
phase condition and temperature before releasing the out- The pressure variable must be controlled by the user in
put streams to the executive routine. This can be done by simulation, with changes programmed at the appropriate
calling the adiabatic flash routine from any unit module points in the process. A useful rule in writing a module
and the former should be capable of returning an all-liquid with several input streams is to operate the module at the
or all-vapor phase with a consistent temperature. It is ex- lowest nonzero pressure. A stream with zero pressure
tremely im ortant that the flash routine be stable and should default to a finite pressure, such as 0.01 lb./sq. in.
P
absolutely ool-proof for both single and multicomponent
streams. It must also be very efficient since it is called so
abs., otherwise the TP system will fail.
Finally, the guiding principle is that the modules and
very often; in practically every unit module in the system. system should run on if at all possible despite all of the
The most demanding task faced by the developer is the apparent diagnostic difficulties, in the so-called “fail-soft”
design of user diagnostic messages, their location, and in- mode. Flashes, dew points or bubble points that do not

AlChE Journal (Vol. 21, No. 3) May, 1975 Page 423

..
.
converge, even computation underflows and overflows, TABLE
4.BASICPROPERTYFUNCTIONS
should be intercepted by the module writer to provide
default values. These may not be fatal errors entirely. The Function
best judge is the user who must decide what stream or per-
formance data are physically meaningless so that he can 1. ENTHALPY Compute enthalpy of desig-
adjust his modeling concept for the next pass at the com- nated stream at indicated
puter. temperature, pressure, com-
position, and state (liquid
Complex Units or vapor)
There are two areas of chemical engineering coniputa- 2. TEMPERATURE Compute temperature of desig-
tion that have consumed huge quantities of computer nated stream at indicated
time: fractionation and reactor analysis. The most im- enthalpy, etc.
portant unit module in any simulation program is the 3. DEW POINT TEIM-
rigorous fractionator package, and no system today at- PERATURE
tracts much attention without a good, multi-option, multi- 4. BUBBLE POINT
TEMPERATURE
stage equilibrium module. This unit is usually the largest 5. DENSITY Compute density, specific grav-
module in the system and it alone, along with supporting ity, or compressibility of
services such as TP and phase equilibrium, dictates the designated all-liquid or all-
total active main memory requirements on the computer. vapor stream.
The modern fractionator package often relies on matrix 6. PHASE EQUILIBRIUM Compute K-values at designated
techniques for mass and energy balancing with assumed RATIOS temperature and pressure, or,
vapor and temperature profiles and the Newton or quasi- of two co-existing streams for
Newton method to converge total vapor and temperature composition-dependent ratios.
profiles (Orbach et al., 1972). It is stable for all applica-
tions including rectified absorbers and strippers and in-
cludes options for multiple feeds, multiple products, inter-
stage cooling and heating, various condenser and reboiler the simulation package. Most fully developed systems will
options (reflux ratio, boil up ratio, withdrawal or boil up include at least three classes of heat exchange modules:
or reflux rate), and multicolumn operation (see MULTI- (1) A conventional two-stream exchanger with specified
COL in CONCEPT 111).While the simulation-based frac- overall heat transfer coefficient and geometry ( a simula-
tionator module is the most tinie-consuming phase of tion-based module) ; ( 2 ) A temperature controller which
computation, some of the modern codes are efficient computes the heat duty to reach a specified temperature
enough to allow design specifications on components to or which computes temperature for a given heat duty ( a
be imposed through a control block which solves the design-based module); and ( 3 ) A phase fraction con-
rigorous simulation module iteratively (see FLOWTRAN, troller which computes heat duty required to reach a
Seader et al., 1974). specified phase fraction including dewpoint or bubble
I n addition to the rigorous fractionator package, other point conditions ( a design-based module). Additional op-
modules for shortcut design-based approaches need to be tions under the first class would include water-cooled ex-
included for preliminary analysis. These would be macro- changers where the module computes the amount of water
scopic techniques like Fenske-Underwood-Gilliland for required between specified temperature limits and steam-
distillation and Edmister absorption-stripping factor or refrigeration-driven modules where the amount of
method for absorbers or strippers. steam or refrigerant is computed for operation between
The more complex reactor simulation modules are so fixed enthalpy limits at constant temperature.
s ecific to each process that many systems provide only
Control Blocks
tX)e basic module which solves one reaction at a time based
on stoichiometry or the multireaction stoichiometric sys- The design-based aspects of simulation programs find
tem based on extent of reaction vectors (Henley and their epitome in the concepts of control blocks (Henley
Rosen, 1969 and FLOWTRAN), isothermally in either and Rosen, 1969). Some control block (or design-based)
case. Adiabatic reaction modules involve only a simple ex- applications have been indicated above within individual
tension of the equipment parameter vector to provide modules. In the general case, a control block is a unit
heats-of-reactions. The adjustment of properties to tem- module placed in the process network, with a span of in-
perature and pressure specifications can then rely on the fluence over one or more simulation-based modules, for
TP system. Extension of the reactor analysis modules to the purpose of manipulating one or more individual equip-
equilibrium systems may be done with the help of a non- ment parameters to meet user-defined module output
linear equation-solving subroutine in conjunction with the specifications. These act effectively as iteratpje feedback
stoichiometric and adiabatic versions above (Sanderson controllers by nulling the error in the desired output and
and Chien, 1973) or alternately from a free energy mini- mathematically are equation solvers. They may include
mization point of view (Dluzniewski and Adler, 1972) convergence algorithms of the Newton, quasi-Newton,
but with caution (Bjornbom, 1974). Wegstein, or steepest descent type.
One approach that has not been generally provided is Some characteristic applications of control blocks are
to simulate a reactor by specifying reactor volume and the given in Table 5.
kinetic mechanism. This is one area where simulation- The only function requiring further explanation is the
design practice seems to favor the calculation of volume fourth, linkage control, which differs mathematically from
(and reactor cost) in the economic phase of the process the others. An example would occur in the energy man-
with the extent of conversion design parameter as an ex- agement between a fractionator reboiler and a process
ternally controlled variable. By and large, the reactor stream used as a heat medium. If the two activities are
module is the most frequently encountered instance of a separated, that is, fractionator reboiler duty obtained by a
user-supplied module for specific process chemistries. d e s i p method and energy abstraction from a process
Heat exchange modules are provided in multiples, some stream by simulation, then the linkage block transfers the
with additional multiple options. This is another area heat duty from the fractionator module to the stream
where design-based concepts encroach substantially into cnJllin~~mndiileas soon as it becomes available via the

AlChE Journal (Vol. 21, NO. 3)

Page 424 May, 1975
 TABLE
5. TYPICAL BLOCKFUNCTIONS
CONTF~OL (C.B.M., 1973) system, and the Uhde (1973) system.
The achievement of a highly reliable and industrially
Application Function useful TP package is a difficult task. As in all active com-
puter projects, however, these packages are not stationary,
1. Degree of vaporization Produce two streams with speci- they become more comprehensive, more reliable, and more
fied degree of vaporization of efficient from month to month.
the input There are still many gaps in our ability to predict the
2. Ratio or flow control Adjust the mixing of two streams liquid phase fugacity of supercritical gases, retrograde
to produce a specified com- phenomena, liquid activities, immiscibility or partial mis-
ponent ratio or total molal cibility, and various phase equilibria both with and with-
flow. Excess input becomes a out concurrent chemical equilibrium at both low and high
separate output or make-up re- reduced temperatures. Our impression is that the recent
1 uirement is obtained from
ummy feed stream
Adjust one or more arbitrary
thrust of physical property research is aimed at integrated,
consistent estimation systems covering the entire be-
3. Feedback control
equipment parameters over a havioral domain, (Cavett, 1972; Chueh and Stein, 1972;
span of one or more units to Hankinson and Cantwell, 1972) in addition to the more
achieve a desired component conventional research on individual property estimation.
ratio, flow, temperature, or
distillate fraction. Structure of TP Systems
4. Linkage control Connect equipment parameters of TP are generated from correlation equations or estima-
two units so that the updating tion equations such as an equation of state, a vapor pres-
of a parameter of a design- sure or activity coefficient equation, etc. Thus, during the
based module reflects on the calculation phase of the simulation only the estimation
specification of a simulation parameters of the methods employed for the specific set
module during computation. of chemical components appearing in the process need be
known to the TP package.
A TP package is a multiple entry, multiple option sub-
equipment parameter vectors. system available to any equipment module through a set
The ultimate purpose of a simulation program with of standard subroutine calls. Equipment modules can be
many design-based features is still simulation, but with developed without regard to the particular TP estimation
rational constraints. procedures or components to be used since the preproc-
User Defined Modules essing phase of the simulation run will condition the TP
package to use the correct options and components when
All major simulation programs provide facilities for the TP generation is invoked. Such an environment is re-
user’s own module. Once the stream vector information ferred to as autogeneration, and it frees the user from any
structure is known and the standard module interface is coding of property-related data once a few simple options
defined, the user can code a module in, for example, (one or two constants) and the component list have been
FORTRAN, assigning equipment parameter information
identified.
structure as he chooses. He may even call system unit
modules from his own module provided the called modules T P packages contain two classes of options. The first
are co-resident in main memory at calculation time and relates to the use of alternate equations to predict the same
property and the second to the road-map option which
provided the calling module manages the standard module
interface vectors for streams and equipment parameters. modifies the estimation sequence to compensate for vari-
Some systems have POL input services which assist the ous combinations of given or available and missing basic
user in defining a module. Another approach is that of physical constants. In either case the options are resolved
Soylemez and Seider (1972) who provide a translator during the preprocessing phase, the first by user specifica-
which converts any stand-alone FORrRAN program into tion and the second by the TP preprocessor setting up the
a unit module compatible with PACER 245. In any event, missing constants for direct use during autogeneration. A
the flexibility of a simulation program is greatly enhanced useful adjunct to simulation programs is a version of the
by user-defined modules since it is recognized that one TP package which generates physical property reports
such as Mollier charts and tabulated properties over
can only generalize perhaps 90% of any large simulation
job. limited ranges of temperature, pressure, and composition
to al?ow user inspection and validation by independent
means. Most simulation systems available do not provide
THERMODYNAMIC AND PHYSICAL PROPERTIES
this feature explicitly. An exception, however, is the
The most critical requirement imposed on simulation FLOWTRAN system.
programs by industrial users is the accuracy of the thermo- The most important and critical modules in any simula-
dynamic and physical property (TP) estimation or cor- tion program are the phase determination routines. These
relation methods employed. Academic users tend to be are not considered formally part of the TP package but
somewhat less demanding provided the program is to be there is an intimate relation, particularly in obtaining
used primarily for instruction or classroom projects rather phase equilibrium ratios which are composition dependent
than research. or in computing enthalpy or entropy for mutliphase
We will not discuss the merits of particular TP esti- streams. Most simulators will handle two phase vapor-
mation methods or packages nor survey recent research re- liquid equilibrium calculations. FLOWTRAN, in addi-
sults in this area. A useful critical review of many avail- tion, considers three phase equilibria with partially mis-
able T P systems, both stand-alone and embedded in simu- cible liquids. Experience, however, in these systems has
lation programs is the textbook for an AIChE Short been confined to mixtures well below their critical region.
Course (Reid and Evans, 1970). Hughson and Steymann Little or no experience has been reported in the retro-
(1971, 1973a, b ) have compiled an extensive list. Only grade region.
four major systems have appeared since: the British In- Three phase vapor-liquid-liquid equilibria may be very
stitution of Chemical Engineers PPDS (1971) package, sensitive numerically and often depend on reasonably good
the CONCEPT (Mark 111, 1973) system, the Solvay initial composition estimates. Although they have been

AlChE Journal (Vol. 21, No. 3) May, 1975 Page 425

generally approached from an activity coefficient point of information. Since many current estimation methods in-
view (Henley and Rosen, 1969), recent efforts have at- clude binary pair interaction coefficients (Wilson, 1964,
tempted to utilize an equation of state (Heidemann, 1966), the preprocessing phase may also include the re-
1974). duction of user-supplied x-y data or other vapor-liquid
One of the values of process simulation with a built-in equilibrium data on binary pairs from which are generated
TP package in industry is the property consistency ob- Wilson parameters, van-Laar activity coefficients, and
tained within an entire project or within classes of similar other binary coefficients.
projects. Where the system is widely used in the company, The TP package functions as a semi-independent execu-
there will usually be a supporting group of several people tive system with standard requests for any property from
responsible for supervising the accuracy and reliability of any module being serviced according to options identified
the estimation procedures used in the T P system. In the during the preprocessing phase both as to method and
past, such a group would have been responsible for the road-map. User supplied data have been put in compatible
compilation of company data handbooks and responding format so that the combination of phase determination
to requests for property information on new compounds. modules and TP package can work just as efficiently as
In the computer-based TP system environment, the best possible. The more rigor the user requires of TP, the
current knowledge and generalizations are coded and im- more time is spent in the TP system, often up to two-thirds
mediately accessible both to minor and major projects. of the entire job time.
The handbooks may still exist as a resource for checking
and validation and may themselves be generated on the TP Initialization
computer under careful control and supervision by TP During the preprocessing phase, the simulation pro-
experts. gram scans the list of feed streams and computes their
Nevertheless, the most difficult aspect of an effective T P phase fraction or enthalpy. It is thus convenient to enter
package is integration and consistency. Estimation meth- feeds at their bubble or dew point and have an enthalpy
ods are not universal, Good K-value correlations do not internally consistent with intermediate process streams
bridge the entire domain of conditions and compositions with which to begin the simulation.
from the critical temperature (or convergence pressure)
to cryogenic temperatures. Hence, the user may have to
COMPUTATIONAL EFFICIENCY
choose an option which is more valid in the dominant
temperature region of this process. Ideally, one should be The most time-consuming calculations in process simu-
able to specify the technique to be used in evaluating lation are the recycle calculations since they require itera-
phase behavior properties for each specific equipment tive methods. A large number of articles which have pro-
module. Generally, errors in K-values (and relative volatil- posed different methods for reducing the effort required
ities) tend to be more serious than errors in enthalpy. Cur- for recycle calculations have been published in recent
rent technology can usually predict K and enthalpy values years. In fact, according to Briddell (197413) : “The prob-
over reasonable ranges of temperature and pressure ade- lems of equation ordering and convergence acceleration
quate for process design. Errors in K-values may lead to tend to dominate the literature of simulation.” Neverthe-
serious design errors, especially when dealing with large less, there is plenty of need for improvement in this area.
recycles and or closely boiling components. Enthalpy er- The problem of recycle calculation may be best ex-
rors can be handled more easily in the equipment de- plained with reference to Figure 1. The composition and
sign (for example by adding more tubes to a heat ex- temperature of stream No. 3 may be easily calculated if
changer) than K-value errors (where for example more all the data about the feed and the design variables of
trays and a larger column design may be required). equipments (1) and ( 2 ) are known. But the temperature
Enthalpy estimation tends to be a patchwork quilt of of stream 4 cannot be calculated since there is no in-
methods. One may choose to rely on a conventional equa- foimation about stream No. 8. The same problem exists
tion of state coupled with ideal gas heat capacity for the for streams No. 11 and No. 7. The subsystem consisting
vapor enthalpy and a three-parameter corresponding states of equipment Nos. 3, 4,5, 6, 7, and 8 which are connected
correlation for the liquid phase enthalpy via the heat of by recycle streams is called the recycle system or rnaxi-
vaporization with the corresponding states approach re- ma1 cydic net. To calculate the streams of this recycle
served for subcooled liquids. In any event, matching pre- system, values for some of the streams must be assumed
dictions without a single universal, thermodynamically and the entire recycle system must be computed. From
consistent method may lead to some grief for the user, the calculated values of these streams, a better estimate
particularly around the critical region where the phase may be obtained. This procedure is continued iteratively
equilibrium correlation may predict a two-phase mixture until the difference between the estimated and calculated
and the equation of state and/or the corresponding states values of the assumed streams become smaller than a
correlation predict a supercritical system. It may be noted desired tolerance.
that in most commercial systems many options for phase The process of identifying separated maximal cyclic
equilibrium calculations are offered, but no options for nets and of ordering of the nodes that are not included in
stream enthalpy are indicated. This suggests that a single such nets is called partitioning. Those streams in the re-
correlation for enthalpy is adequate industrially for a cycle system which must be assumed are the tearing
broad spectrum of simulations. streams, and the optimal tearing streams may be found
by tearing or decomposition procedures. The convergence
Interfaces with Simulation Programs rate of the iterative process for solution of the recycle sys-
Every modern simulation program maintains an archival tem may be increased by convergence acceleration meth-
data base of T P estimation coefficients for one to several ods. The iterative process will be stopped by the con-
hundred standard components. This data base is scanned vergence test system.
during the property preprocessing phase to collect coef- This chapter will discuss the different approaches to
ficients for those species used in a particular job. partitioning, tearing, convergence acceleration, and test-
The preprocessing phase will also provide user service ing.
routines to compute estimation coefficients for nonstan-
dard species identified by the user from boiling-point, Partitioning and Tearing
molecular weight, and other parachoric or molecular type For the purposes of partitioning and tearing the in-

Page 426 May, 1975 AlChE Journal (Vol. 21, No. 3)

formation flow sheet may be represented by a directed A more recent method based on list processing opera-
graph or digraph, the chemical equipment represented tions performed on the signal flowgraph of the flow sheet
by nodes (vertices), and the streams by directed edges. was deve!oped by Barkley and Motard (1972) and ex-
A digraph of the information flow sheet (Figure 3 ) is tended to consider weighting criteria on the streams by
shown in Figure 4. The digraph may be stored in the Pho and Lapidus (1973). The signal flowgraph of the
computer in the form of an adjacency matrix. This matrix flow sheet in Figure 1 is given in Figure 6. It is also a
is a Boolean square matrix, where the number of rows and digraph. In this representation, the streams become the
columns of this matrix is equal to the number of the nodes nodes and the edges indicate dependency. A directed edge
in the digraph. If there exists an edge from node i to node from node (stream) i to node (stream) i indicates that
j , then the element ij of the adjacency matrix is unity; stream j is a physical output from an equipment for which
otherwise it is zero. The adjacency matrix of the digraph stream i was the physical input. This is a very efficient
in Figure 4 is shown in Figure 5. algorithm for finding minimum tears and partitioning
Methods for partitioning have been developed based on flow sheet calculations and is well suited to manual use.
path searching and on Boolean operations performed on The largest problems require only a few seconds of com-
adjacency matrixes. A detailed description of these meth- puter time and less than thirty minutes of normal calcula-
ods may be found in Kehat and Shacham (1973b). Some tion.
improvement of the adjacency matrix method has been The aim of the different tearing algorithms is to find an
done by Ledet and Himmelblau (1970), Jain and Eak- optimal computation order for a recycle system so that
man (1971), the Kehat and Shacham (197313). The aim this computational order will lead to solution of the re-
of these improvements was to reduce the storage and the cycle system with minimal computer time and storage re-
computer time requirements of the adjacency matrix quirements. There are no good optimum criteria for tear-
method. ing to minimize the computation time.
Batstone et al. (1971) compared the efficiency of the There are three different accepted criteria for tearing:
pathsearching and the basic adjacency matrix method for
a particular problem and found the path searching al- 1. Requirement of minimum tearing streams
gorithm clearly superior. 2. Requirement of minimum tearing parameters
3. Requirement of minimum of weighting factors for
the tearing streams.
For the first criterion only the number of streams are
taken into consideration; for the second, the number of
variables in every stream are also taken into consideration.
For the third criterion, it is possible to add weighting
factors to the different streams or to different sets of
streams according to the estimated sensitivity of the cal-
culation process to those streams.
None of these tearing criteria have proved to be op-
timal for every type of convergence acceleration method.
Fig. 4. The digraph of the flow sheet in Figure 1. In fact, Shacham and Motard (1974) have shown that
the minimum number of tearing streams or parameters will
not necessarily lead to the more rapid convergence of the
direct iteration method.
UNIT NO.
The sensitivity of different tearing streams to con-
I vergence acceleration which is required for the third type
2 of tearing criterion is not absolutely definitive. One re-
3 ported experience in finding less sensitive tearing streams
4 by an adaptive procedure with quasi-Newton acceleration
has been done by Genna and Motard (1973). In this pre-
5 liminary experiment, the procedure for locating the best
6 tearing streams may be more time consuming than the
solution of the whole recycle system since it requires an
7 estimation of the maximum eigenvalue of the Jacobian for
8 several alternate tearing sets. However, Upadhye and
Fig. 5. The adjacency matrix o f the digraph in Grens (1975) have fairly conclusively shown that the op-
Figure 4. timum tear set, that is, that with the minimal maximum
eigenvalue can be identified for the direct iteration method
by an a priori criterion. The optimum set belongs to a
family of tear sets characterized as nonredundant in rela-
tion to a simple but powerful rule called the replacement
theorem.
New tearing algorithms for recycle systems which have
been published in the last four years are summarized in
STREAM PRECURSORS STREAM PRECURSORS Table 6. Generally speaking, ail the tearing algorithms
I - 8 6 contain two different phases. The first phase is the recycle
2 I 9 4,7,II simplification, the second one is the search for optimum
3 2 10 9 tearing point on the simplified digraph.
4 39% II 10
5 4711 12 10 In the first phase, the edges which cannot be used as
6 '5 tearing edges are reduced, and the streams which must be
7 6 torn in any case are found. For the execution of the sec-
Fig. 6. Signal flow graph of the flow sheet in Figure 1 (stream 7 ond phase there are two approaches:
precursor omitted for clarity). 1. Exact methods which used dynamic programming or
AlChE Journal (Vol. 21, No. 3) May, 1975 Page 427
 TABLE6. CLASSIFICATION
OF TEARING
ALGORITHMS

Maximal system
Authors 5Pe Programmed solved

Upadhye and Grens (1975) Probability No information 6 nodes, 9 edges

Pho and Lapidus (1973) Exact No information Sulfuric acid plant
42 nodes, 68 edges
Kehat and Shacham (1973b) Probability Yes 42 nodes, 68 edges
Upadhye and Grens (1972) Exact No information 6 nodes, 9 edges
Barkley and Motard (1972) Probability Yes Sulfuric acid plant
Batstone and Prince (1970) Probability* Yes 27 nodes, 49 edges

Obtained by simplification of an exact method.

branch and bound methods,

2. Probability methods, where the tearing streams are
located according to some assumption that gives large
probability that the tearing point will be chosen such JI .. z . . J.
that one tear will cut a maximum number of internal
cycles. Many practical problems do not require the second
-
phase.
While the probability methods do not always ensure
optimal solution for every problem, they are easier to
program and required much less manipulation.
Feeds ’ PROCESS
products ’
A characteristic of many of the tearing algorithms is
that they may be applied manually relatively easily and
quickly. However, they are difficult to program and some-
times computer execution is exceedingly slow. The list Kehat and Shacham ( 1 9 7 3 ~ ) have reviewed these
processing method is much more efficient in computer methods. They have introduced ten basic methods, most
utilization. An example from Batstone and Prince (1970) of them standard methods for the solution of systems of
solved on the GE 225 computer in 2 minutes 53 seconds nonlinear algebraic equations. Only three of the methods
was solved manually by us in less than 15 minutes. have been used for process recycle systems. Some are
I n Table 6 we summarize whether the tearing algorithm never employed for this purpose because it is either ob-
has been programmed, and the maximal dimension of vious that they are not efficient enough for such recycle
recycle system reportedly solved. calculations or they have been tried and found deficient.
Research in the field of tearing algorithms must be di- The methods which are in use today are the Bounded
rected to the development of optimum tearing criteria. Wegstein method (Kliesch, 1967), the Dominant Eigen-
Convergence Acceleration value method (Orbach and Crowe, 1971), and the Quasi-
After the tearing and partitioning process has been Newton method (Broyden, 1965).
completed, every maximal cyclical net of the process is a The Wegstein and the Dominant Eigenvalue methods
function of its tear streams. Typically a net in the system accelerate the convergence by reducing the effective
would be viewed as shown in Figure 7 where = esti- magnitude of the maximal eigenvalue of the iteration ma-
mated recycle values of all variables in all tear streams; trix.
In both of these methods a better estimation for zi+l is
7 = calculated recycle values for the same variables. The obtained from (5.2)
streams in the net are computed iteratively, with im-
-
proved estimates of 7 obtained from one of the accelera- = 4.i
~ i + l + (1- 4) - ri (5.2)
tion methods applied to the difference between and 7
where q = acceleration parameter
The simplest iterative method is the Direct Iteration or
The difference between the two methods lies in the
Successive Substitution method. In this method, the new
calculation of 9. In the Wegstein method, q is calculated
z
estimates chosen for are the last values of
separately for every element of ;from Equation ( 5 . 3 ) .
Thus
- - ri,j - ri-l,j
zi+l= ri (5.1) a=
Shacham and Motard (1974) have shown in a recent and ( 5.3)
article that the Direct Iteration method must converge lor
every physically realizable recycle system from an initial q=-
estimate which is close enough to the solution. However, a-1
the convergence rate of the direct iteration method may
where zi,j (or ri,j) is some element number j of vector
be slow. This rate is a function
- of the maximum eigenvalue
(or 7) in iteration number i. In the Dominant Eigen-
of the iteration matrix dr/az. This eigenvalue is usually not value method q is a scalar and calculated from the norm
known in advance, but it may be estimated from the con- ratio:
vergence rate of the direct iteration method (Orbach
and Crowe, 1971; Shacham and Motard, 1974). If the and
maximal eigenvalue is close to unity in absolute value, x
there is a need for using convergence acceleration meth- q=-
ods. A-1

Page 428 May, 1975 AlChE Journal (Vol. 21, No. 3)

where h is an estimate of the maximal eigenvalue of the TABLE
7. THEu 9 E OF CONVERGENCE
ACCELERATIONM G m O D s
iteration matrix. PROGRAMS
FOR SIMULATION

When using these two methods, the process may be-

come unstable. Shacham and Motard (1974) have found Method Simulation program Examples
that the reason for this instability is that some values of
the acceleration parameter q may cause the eigenvalues Bounded CAPES (Maejima and Batstone et al.
of the modified iteration matrix to increase in magnitude. Wegstein Shindo) (1973) ( 1971 )
They have also demonstrated that there is an optimal CONCEPT (Peters and
value for the acceleration parameter, but there is as yet Barker) (1974)
no convenient method for computing this optimal value. FLOWTRAN (Seaderet Seader et al.
al.) ( 1974) (1974)
One may overcome the instability problem by bounding CHESS ( Worley and
the value of the acceleration parameter (Worley and Mo- Motard) (1972)
tard, 1972) or by multiplying it by a relaxation factor Dominant CAPES (Maejima and Batstone et al.
between zero and one (Orbach and Crowe, 1971). Eigenvalue Shindo) (1973) ( 1971 )
There are a number of Quasi-Newton methods. The Orbach and
most popular for chemical engineers is Broyden’s method Crowe (1971)
(Broyden, 1965). The use of this method is based on the Quasi-Newton CAPES Batstone et al.
similarity between a recycle system and a system of non- (1971)
linear equations. The rec cle system problem may be ex- CHESS Genna and
pressed by Equation (5.5r Motard (1973)

A better estimate for 2 according to the Quasi-Newton where E is a small tolerance which must be supplied by
method will be the user.
- -
- + tZif( Some authors have found (Orbach and Crowe, 1971;
zi+l = zi Zi) (5.6) Friedman and Pinder, 1972; Shacham and Kehat, 1973)
- that none of these tests are satisfactory. The user gen-
where t is a scalar damping factor and fl an approxima-
- erally knows what accuracy he needs for engineering pur-
tion to the negative inverse of the Jacobian matrix $/az. - -
poses, but he cannot specify this - as -the error tolerance
The iteration process is usually begun with an identity because the value of f (zi)= ~i - zi is a function not
matrix for?? and t = 1, that is, the direct iteration method only of the real error but also of the rate of convergence
- of the iterative method. The penalty for specifying inac-
(5.1). In every subsequent iteration, the H matrix is curate values of the error tolerance is either excessive
corrected according to 5.7 computation time or inaccurate answers.
- - - - -
- -
Hi+l = Hi - (ti% + -- -s T -H ~ -/s --
~ i y i ) ~ ~ l i y i(5.7)
Klaus and Moore (1972) and Shacham and Kehat
(1973) have found ways of bounding the solution and to
where measure the exact error independent of the solution
- method, but only for problems with one variable or one
yi = f ( z i + l ) -TCZi)
- =-
Pi Zi+l
-
- Zi
tear stream. Orbach and Crowe (1971) have used an
estimation of the maximal eigenvalue of the iteration ma-
trix for improving the accuracy of the error test.
and S T is an arbitrary vector. More research is needed to ensure that the user will
Broyden (1965) has chosensT = TTbut other choices obtain answers with exactly the required degree of ac-
are possible (Kehat and Shacham, 1 9 7 3 ~ ) .The ,Quasi- curacy without excessive computation time.
Newton method may be very efficient for the solution of
recycle systems (Genna and Motard, 1973; Batstone et al., EXPERIENCE WITH SIMULATION PROGRAMS
1971) but there are also counter examples (Batstone et al.,
Probably the most important experience gained in the
-
1971). The main drawback of this method is the large
storage requirement for the matrix H . This method is also
last few years is that process simulation is not an end in
itself. It is merely a new tool for the chemical engineer
more di5cult to program than the other two. since it cannot replace engineering judgement. Some ef-
Table 7 summarizes the different simulation programs fort is required to make the best use of this tool.
and their method of convergence acceleration. This table So far, the degree of acceptance and application of proc-
also shows where one may find examples using the different ess simulation in industry has varied widely. Industrial
methods. users are generally stubborn, uninterested in partial solu-
More research is needed to improve the Successive tions, and require systems that are flexible and reliable
Over Relaxation (SOR) type methods (Wegstein, Domi- using familiar techniques. Further, not all process engi-
nant Eigenvalue) . The Quasi-Newton methods must be neering jobs require a simulation program-some merely
require refurbishing an old solution manually.
carefully investigated and require further refinement be-
fore they can justifiably replace the SOR-type methods. In education, the major barriers to wide acceptance of
simulation seems to be the availability of reasonable data
Convergence Testa and cost.
Accepted methods for testing the convergence of the Industry
recycle system are the absolute error test [Equation The development and acceptance of process simulation
(5.8)1, and the relative error test [Equation (5.9)] by the chemical industry is a continuous process which
started in the middle 1950’s. Di Bella (1974) divides the
(5.8) history of process simulation into five-year periods.
The period 1955-1959 was characterized by the de-
(5.9) velopment of computer programs for individual unit op-
erations. The developer of these programs was required

AlChE Journal (Vol. 21, No. 3) May, 1975 Page 429

to be an expert in chemical engineering, mathematics, and changing the operating conditions. The simulation is help-
computers including machine language. Usually the de- ful to find the best strategy for raising the production, to
veloper was the only one who could use these programs enhance the efficiency of operation through better energy
for solution of actual problems. Successes were relatively integration, to adapt the existing plant to a different raw
few, but when there were successes, they generated huge material, or to a requirement for products with different
savings. These successes encouraged the management to compositions.
fund limited process simulator development projects in The major benefit of process simulation has been that it
the period of 1960-1964. produces better process designs with lower capital and
The simulators which survived from this era and which operating costs in less time than hand methods alone or
enjoy wide acceptance today had the following develop- hand methods in combination with stand alone computer
mental guidelines: programs. The Monsanto Company (Monsanto, 1974)
1. Coding would be in a high level language (for ex- estimates the saving obtained by using the FLOWTRAN
ample, FORTRAN) and would be highly modular. simulation program at 28 man-months in one project, and
2. The physical property correlations would be as rigor- a potential yearly net gain of $750,000 in increased pro-
ous and as accurate as possible. duction in another one. However, there are several other
3. The system would be easy to use with little or no in- intangible, yet highly important, benefits that are per-
tervention by an expert required. haps less well understood. These are the transferability of
process simulations and the improved communication
The period 1965-1969 was characterized by the refine- via the use of a process simulation. The different engineer-
ment of industrial simulation programs. As simulators ing groups which deal with the same plant in different
were released within their respective companies for gen- stages may use the simulator as a basis for faster and im-
eral use, serious problems were often found with their proved communication.
reliability and generality. This may have resulted from Briddell (1974a, b ) summarizes the most severe pitfalls
inadequate resources or testing being put into the effort. in the use of simulation. Among these are carrying out a
In addition, proprietary restrictions by industry may have simulation for its own sake, not having clear objectives,
tended to restrict development. Sometimes the simulators being overly rigorous, not knowing how much to assume,
were able to solve little more than the test problems that and not working with the people who have the problem.
had been used in developing them. Many of the innova- For proper use of simulation, Briddell proposes to judge
tive first users became disenchanted with the concept of the necessity of computer simulation and the degree of
process simulation and would not apply it for several years sophistication that must be employed against a realistic
after an early bad experience. cost estimate.
However, by the end of this period the concept of proc- The development, documentation, and maintenance of
ess simulation was firmly established and accepted with a commercial simulation program requires a large invest-
only a sprinkling of ho!douts. The reliability of the proc- ment and a large engineering team (Seider, 1972). Thus,
ess simulators had increased as had the speed and re- only large companies can afford the initial development of
liability of computers on which the simulation ran. commercial programs. The price of a commercial program
In the recent period (from 1970), the simulator has may be easily estimated. PACER 245 contains about
become a legitimate tool in the process engineer’s kit. 40,000 source cards. The cost of development of a good
But it is used mainly in the companies which have devel- computer library is about $15 to $20 per source card ac-
oped the simulator. The reasons for this will be discussed cording to Cody (1974). So, the cost of a program like
later. PACER 245 may be about $600,000 to $800,000. Re-
Theoretically, process simulation can be useful at all the producing FLOWTRAN (Harris, 1972) might cost twice
stages of the development of a chemical project. Davies as much.
(1971) and Davies and Perris (1971) have discussed in The main use of simulation programs is within the com-
detail the possible uses of the simulation programs and pany developing the program since the team which de-
Klumpar (1970a, b ) has given a detailed example of the veloped the program can use it most easily. New users
use of the PROVES simulation program in different stages may face difficulties in training the people who will use
of project development. the systems. In addition, problems may arise in communi-
At different stages of a project, there are needs for simu- cation with the developers when there are difficulties in
lation programs, with different levels of sophistication. the operation of the simulation program or interpreta-
Process simulators may be used at the following stages: tion of results.
1. Research and Development Stage. A simple simula- The extension of the study of process simulation in the
tion program which requires a minimum of data may be universities in recent years may accelerate the use of
used for testing the economic feasibility of different proc- process simulation in the chemical industry in succeeding
esses. years. Recently Monsanto Co. has made available its
2. Critical Examination Stage. Once a financially at- FLOWTRAN system to aid universities in such train-
tractive process has been found, different alternatives 6f ing (Institute News, 1974).
plant size layout and operating conditions must be tested For continuing research in process simulation, it is im-
for optimality. portant to remember the main types of industrial uses and
3. Pilot Plant Stage (Davies and Perris, 1971). The the most critical problems of the industrial user. Despite
use of the most sophisticated process simulator may help the fact that industrial process simulators have the ca-
to obtain good estimates of the operating conditions in pability of handling a large set of interacting units, much
the full-scale plant from relatively few results of the pilot use is made of them on comparatively simple processes.
plant. Often, simply a material balance (and not an energy bal-
4. Design Stage. The process simulator gives the engi- ance) has been adequate for the application. Using the
neer all of the process data required for the detailed de- physical property capabilities of the system for simple
sign of different equipments. In this way the safety fac- condensation curves has been another frequent use of
tors due to uncertainty in equipment design may be re- simulators.
duced. Although considerable attention has been given to the
5. Simulation of Existing Plants. Simulation of an exist- problem academically, recycle calculations, in terms of
ing plant may be very useful when there is a need for which stream to tear, have rarely caused much of a prob-

Page 430 May, 1975 AlChE Journal (Vol. 21, No. 3)

lem. The engineer, using his process knowledge, can gen- The simulation was not always successful; in some cases,
erally select suitable tear streams. Convergence accelera- they did not achieve a converged simulation. The McMas-
tion, although always desirable, has not been of major ter simulation team concluded that such a project is very
concern since the bounded Wegstein method has proven fruitful for both students and faculty because of the ex-
satisfactory in a broad range of industrial simulations. cellent industrial experience that it offers.
Major concern in industrial process simulation has been But yet there are a number of difficulties according to
with the area of physical properties (especially at cryo- Woods et al. (1973): “This approach requires an ex-
genic temperatures). Developing and using these proper- tremely large amount of time and effort for both staff and
ties over large ranges of temperature and pressure has students. There must be a plant engineer who is willing
often been the major stumbling block to a successful simu- to devote considerable time to the projects.
lation. “For the students, this detracts from other courses and
Faster and more reliable unit calculation procedures is particularly difficult to account for in courses outside
(especially in nonideal distillation and absorber calcula- their own department.
tions) also continue to be a problem area in process simu- “There is a disproportionate amount of time spent on
lation. Shortcut techniques with their limitations ’ often coding the models and ‘fighting’ with the computer with
are not adequate for use and rigorous calculations must a relatively small amount of time available to use the simu-
be carried out. Three-phase flash calculations can also lation imaginatively.”
pose vexing problems due to inadequately fitting data to We also agree with the conclusion of Woods et al. that;
suitable nonideal vapor-liquid equilibria models. “From our experience it is very important to have a proj-
ect that the students feel is important and not just an aca-
demic exercise. . . . Timely, realistic problems where non-
Education
trival contributions can be made are difficult to find.”
In the field of education one may distinguish between We think that the development of realistic problems
four different uses of process simulation: for simulation is important not only for educational pur-
1. Teaching of process simulation poses, but also for testing new methods and algorithms of
2. Use of simulation programs as an aid in teaching simulation.
There is also a need for development of simulation
other subjects
programs which can solve non-trivial problems but simple
3. Design and simulation projects
4. Research on techniques enough to be mastered easily by the studnets.
Peters and Barker (1974) have compared three differ-
Different aspects of the teaching of process simulation are ent simulation programs (PACER, GEMCS and
discussed by King et al. ( 1973), Gaddy ( 1974), and Rosen CONCEPT) according to their value to graduate and un-
( 1974). dergraduate students. They have found GEMCS (John-
The objective of the first part of a simulation course at son and Toong, 1968) superior to the other two programs
the undergraduate level is to give theoretical background for these purposes.
for the use of general purpose simulation programs. The We think that it is important to continue to work on
students gain limited experience in the use of a general these comparisons and to improve existing programs. This
simulation program. The topics covered in such a course may be a better approach than to develop new simula-
are: introduction to simulation, unit module subprograms, tion programs.
thermodynamic and physical property estimation, model-
ing, and computation efficiency.
The aim of the second or graduate level, simulation FUTURE DEVELOPMENT
course is to make the student familiar with the newest The chemical engineer involved with process simula-
developments. It covers topics like dynamic simulation, tion or other computer applications is working in a very
design and optimization problems, and process synthesis. dynamic environment. The rapid development of com-
Simulation programs may be used as a teaching aid in puter hardware and software offers new possible uses of
elementary material and energy balance courses (Henley the computer and suggests new types of problems to be
and Rosen, 1969) in thermodynamic courses (Poznanovic solved. Today’s difficult problems may be tractable tomor-
and Seider, 1972), and in the dynamics of processes and row with the development of a larger computer or a more
control (Nuttal and Himmelblau, 1973). efficient computer language.
Nuttal and Himmelblau have also tested the student Mah (1974) and Sargent (1972) have discussed some
responses to this type of teaching. Most of the students aspects of future development of process simulation. In
were uncertain about the superiority of computer based this chapter we shall discuss the different subjects of proc-
simulation over traditional instruction. ess simu!ation which are at present under development:
The senior design and simulation project seems to be 1. Equation-oriented techniques for process simulation
the most important educational use of process simulation. and design.
If the project is well designed, the student must use most 2. Process optimization and dynamic simulation.
of his fundamental knowledge and must translate this We shall also investigate the possible uses of the latest
knowledge into a practical and economically feasible de- developments in computer hardware and software for
sign. process simulation purposes.
There are some descriptions of design and simulation
projects in the literature (Gaddy, 1974; Worley and Mo- Equation Oriented Techniques for Process Sirnulotion and
tard, 1972; Crowe et a]., 1971; and Johnson, 1972b). Design
The most detailed discussion of this subject has been done A complete chemical process may also be represented
by the McMaster University simulation team (Woods et by a large system of nonlinear equations. Although a large
al., 1973). Six different simulation projects have been done amount of research has been done on the properties of
in seven years. All of these projects have been based on such a system, there is very little practical experience in
data from existing chemical plants. The number of stu- the use of this approach. We prefer to introduce this sub-
dents who worked on each of these projects was between ject as a future development.
five and twelve, in addition to up to seven graduate as- One of the most important properties of a nonlinear
sistants and between three and seven faculty members. equation system representing a chemical process is its

AlChE Journal (Vol. 21, No. 3) May, 1975 Page 431

sparseness. Thus, there are many variables and equations, the manipulated variables may not be practically useful
but only a few variables appear in each equation. It is because chemical equipment is manufactured in discrete
very inefficient to solve such systems wth the usual meth- sizes or more precise control of the operating variables
ods for nonlinear equations. is needed than can be achieved in the plant. For instance,
Recent methods approach the solution in four major Friedman and Pinder (1972) have found that for optimal
steps: operation, one separator must divide a stream, into four
1. Separation of the system into irreducible subsets
streams in the following flow fractions: 0.016, 0.333, ...
Obviously, there is no controller which can keep the flow
(partitioning) fractions exactly in those ratios.
2. Selection of the output variable for each equation Process optimization may be done with a general proc-
3. Identification of the iteration variables (tearing) ess simulation program or by constructing a complete
4. Iterative solution of the subsets equation-oriented model of the process. Crowe et al.
The partitioning methods for systems of equations are (1971) and Seader and Dallin (1972) executed process
similar to those for process flow sheets, and their descrip- optimization by a simulation program with the design
tion may be found in Ledet and Himmelblau (1970) and parameters changed manually. Noit ( 1971) included a
Mah (1972). Once partitioned, the larger system yields subroutine which optimizes a simulation by the gradient
several smaller systems which may be solved separately. search method.
Since these subsystems are also sparse, it is possible to se- Friedman and Pinder (1972) used the CHESS simula-
lect a very few iteration variables and to obtain new values tion program with additional optimization subroutines to
of all variables by intr.oducing the estimates of the itera- design a catalytic polymerization plant. They also com-
tion variables into the equations and solving each equa- pared three different methods for their adaptability to
tion for one output variable. process optimization.
Once an output set has been computed, the solution of A large amount of research effort has been done in the
an irreducible subset of equations becomes similar to the field of equation-oriented process optimization. The prac-
solution of process recycle systems. For the same equa- tical use of these methods are now beginning to appear in
tion subsystem, it is possible to select alternate output systems that generate linearized equation sets to be solved
sets. The best output set is chosen so that the resultant iteratively by modern sparse matrix techniques. The
system will be easiest to solve, and different approaches to earliest applications date from the early 1960’s in the
output set selection and tearing of systems of equations CHEVRON (Ravicz and Norman, 1964) system, how-
have been described by Christensen ( 1970), Westerberg ever, more recently there is some renewal of interest; see,
and Edie (1971a, b), Kevorkian and Snoek (1973), for instance, the TPG system in Mah and Rafal (1971)
Stadtherr et al. (1974), Mah (1972), and Soylemez and and the basis for the MULTICOL system in the CON-
Seider (1973). The iterative solution of the resultant sys- CEPT simulator in Hutchison and Shewchuk (1974). One
tem may be done by one of the regular methods for non- should also note that equation solving is the dominant
linear equation solution (Kehat and Shacham, 1 9 7 3 ~ ) . method used in electrical engineering in contrast to the
In design problems, there are usually more variables ascendancy of block-structured systems in the chemical
than equations. The values of some variables can be ar- process field.
bitrarily assumed (design variables), and the system must Some of the more recent developments in this area are
be solved for the remaining variables. One criterion for summarized by Debrosse and Westerberg ( 1973), Wester-
selecting design variables may be the ultimate degree of berg and Debrosse ( 1973), Stephanopoulos and Wester-
difficulty of solution of the resultant system as discussed berg ( 1973), Westerberg ( 1973), and Christensen
recently by Edie and Westerberg (1971), Rainirez and ( 1970).
Vestal ( 1972), and Friedman and Ramirez ( 1973). Dynamic simulation could be an important tool for the
study of start-up shutdown of chemical plants, and for
Process Optimization and Dynamic Simulation the design of process control systems. A brief survey ap-
The only similarity between process optimization and pears in Perry and Chilton (1973). Table 8 shows the
dynamic simulation is their degree of difficulty. If a regu- general dynamic simulation programs which have been
lar simulation program is used for optimization or dy- developed in recent years. The dynamic model of a chem-
namic simulation, this program will be executed many ical plant is similar to a large system of nonlinear dif-
times. Thus, process optimization and dynamic simula- ferential and algebraic equations. In obtaining the time-
tion at the moment appear to be academic problems with varying operation of the plant, this system must be inte-
limited practical use. grated numerically. To keep the integration error small,
Mah (1974) has discussed some problems of process the entire simulation model must be solved several times
optimization. The foremost question is: For which types per unit of the real time. Therefore, we may expect that
of problems should process optimization be pursued? One the computer time required for the dynamic simulation of
seldom uses process optimization for preliminary economic a process will be more than the time required for steady
evaluation because only approximate answers are required state simulation by several orders of magnitude. Thus,
at this stage. Process optimization requires more informa- most of the dynamic simulation programs have been
tion and is too expensive when a few parametric case used only for the simulation of simple processes. John-
studies may give adequate answers. By contrast, the com- son (1972a) and Ponton et al. (197.2) executed a dy-
plete configuration of a multistage evaporation unit may namic simulation of slightly more complicated processes.
only be obtained by process optimization techniques. Be- Table 8 indicates the types of problems which have been
tween these two extremes the decision to optimize is not solved by each author.
so easily made. In all cases, the uncertainty of the data and The computer time for dynamic simulation may be sig-
the required accuracy of the answers must be considered nificantly reduced by a proper choice of a numerical in-
to justify its use. tegration method. In most of the existing programs the
Having decided to optimize, one must choose to ma- user is permitted to choose between several simple classi-
nipulate only those few design parameters which have a cal methods as shown in Table 8. Recently, more sophis-
relatively significant influence on the economic objective ticated methods have become available. The description
function. Identifying these parameters requires a pre- and comparison of some of these methods may be found
liminary sensitivity analysis. Finally, the optimal values of in a paper by Hull et al. (1972).

Page 432 May, 1975 AlChE Journal (Vol. 21, No. 3)

 TABLE8. CHARACTERIZATION
OF DYNAMIC PROGRAMS
SIMULATION

Simulation
program Reference Maximal system solved Integration method

REMUS Ham (1971) Butane isomerization process Runge-Kutta"

DYNSYS Bobrow et al. (1971) Start up of butadiene ex- Adams-Moulton-Shell
traction plant
Ponton et al. (1972) Chlorine handling plant
Johnson ( 1972a)
PRODYC Ingels (1970 ) Runge-Kutta"
DYFLO Franks ( 1972) Controlled CSTR reactor Runge-Kutta'
Nuttal and Himmelblau ( 1973)
ACME Loibl et al. (1973) Six mixer units
DYSCO Briggs et al. (1974) Start up of SO2 adsorp- Runge-Kutta'
tion unit Adams-Moulton'

* It is possible to use less sophisticated integration methods (Euler, etc.).

We would expect substantial developments in the area computer programs for equipment design will continue to
of dynamic simulation in the near future using the more be independent. However, the transition to using one set
sophisticated integration methods and faster digital com- of programs to the other will be made simpler.
puters to reduce the cost of dynamic simulation. The most important directions in hardware development
will continue to be increased execution speed of the com-
Computer Software and Hardware Developments
puter and a reduction in cost of different components.
One aim of computer software development is to solve The first of these developments will expand the areas of
more complicated problems with maximal efficiency, with application of process simulation, the second will increase
minimal effort and knowledge from the user. For this the number of users. In particular, the reduction in cost
reason, higher level languages (FORTRAN, PL/1) were of main memory will probably obsolete the current em-
developed and are much easier to use than Assembly phasis on virtual storage technology for large applica-
Languages. Subroutine packages and problem oriented tion programs and in the future the user will have avail-
languages (POL) have been developed for efficient solu- able a very large memory to handle large simulations
tion of specific problems. The user of CSMP (IBM, 1969) efficiently. The reduction of the price of terminals and of
need not know either programming or numerical analysis graphic units makes these tools available to more and
for the solution of complicated systems of differential more engineers. Direct-view storage tube graphics is re-
equations. volutionizing the cost-effectiveness benefits of interactive
The development of process simulation programs has graphics, although its use in process simulation has yet
just begun to move in the same direction. So far, two to be demonstrated.
POL based simulation programs have been developed There are several other developments in computer
(Nott, 1971; Hanyak and Klaus, 1973). A program for hardware whose influence on process simulation have not
converting material and energy balance equations into yet been investigated. Such development; are in parallel
a FORTRAN program is available (hlah and Rafal, 1971; processing CPU units, minicomputer and microcomputer
Soylemez and Seider, 1972). The FLOWTRAN system enhancement, microprogrammed machines, and intelligent
includes a precompiler which translates the process struc- terminals. Other developments like distributed process-
ture supplied by the user into a FORTRAN program. ing, network computing, and certified scientific software
In the future, the user of simulation programs will have are bound to have an impact, but more research is needed
to know less about computing methods, languages, and to exploit these areas.
the structure of the simulation programs. A brief train- We may expect that many of the traditional duties of the
ing period will be sufficient to provide the chemical en- engineer will be fu!filled by the computer, and develop-
gineer with the ability to use process simulation. The ment in this direction has already begun (Leesley, 1973;
development and maintenance team will have to be much Tsubaki, 1973). Growing out of the present state of proc-
more expert in numerical and optimization methods, in ess simulation and design is the complete automation of
programming and job control languages, in addition to the the process engineering effort. The movement into design
classical chemical engineering subjects (thermodynamics, automation for the production of process hardware specifi-
unit operation). In this way, the simulation program will cations, bills of material, plant layout, and final engineering
remain up to date with other developments in the compu- construction drawings is gaining considerable momentum
tational field. in the chemical engineering industry in contrast to the
I n the industrial process simulation field, a major effort aerospace and computer or electronic engineering indus-
will probably concentrate on extending correlations for the tries where it has already had considerable impact.
physical properties handled by the simulator. Present
simulators handle nonpolar fluid components adequately. ACKNOWLEDGMENT
Future developments will include the ability to handle
rigorously solids processing (including polymers), solids Preparation was supported in part by National Science
handling, and polar components. Such developments in Foundation Grant GK-36525X.
physical property correlations will of course spawn nu-
merous developments in adding unit operations modules, NOTATION
for example, solids handling, etc. a = parameter
Although process simulators will continue to add at f = function
least rudimentary equipment design functions, the con- H = estimate of the negative inverse Jacobian matrix
sensus is that computer programs for process design and i = subscript index

May, 1975 Page 433

AlChE Journal (Vol. 21, No. 3)
f = subscript index Di Bella, C. W., Monsanto Co., private communication ( 1974).
P = difference between the estimated recycle values Dluzniewski, J. H., and S. B. Adler, “Calculation of Complex
in two successive iterations Reaction and (or) Phase Equilibria Problems,” in Decision.
4 = acceleration parameter Design and the Computer, 5th European Symp. Ser. No. 35.
Inst. of Chem. Engrs., London (1972).
r = calculated recycle value Edie, F. C., and A. W. Westerberg, “‘Computer Aided Design
s = arbitrary vector Part 3, Decision Variable Selection to Avoid Hidden Singu
t = scalar damping factor larities in Resulting Recycle Calculation,” Chem. Eng. J., 2,
y = difference between the function values in two 114 (1971).
successive iterations Flower, J. R., and B. D. Whitehead, “Computer-Aided De-
z = estimated recycle value sign: A Survey of Flowsheeting Programs. Part 1,” Chem.
Eng. (London), 272,208 (1973a).
Greek Letten , Computer Aided Design: A Survey of Flowsheeting
Programs, Part 2,” ibid., 273, 271 (197313).
c = error tolerance FLOWPACK-User’s Manual, Runcorn, I.C.I. Ltd, CIRL
A = estimate of the maximal eigenvalue of the itera- (1970).
tion matrix Franks, R. G. E., Modeling and Simulation in Chemical Engi-
neering, Interscience, New York ( 1972).
Friedman, F., and W. F. Ramirez, “Convergence Properties of
LITERATURE CITED
Systems of Algebraic Equations-Explicit Loops,” AZChE J.,
Barkley, R. W., and R. L. Motard, “Decomposition of Nets,” 1 9 ( 3 ) , 566 (1973).
Chem. Eng. J., 3,265 ( 1972). Friedman, P., and K. L. Pinder, “Optimization of a Simulation
Batstone, D. B., and R. G. H. Prince, “‘Computer Aided De- Model of a Chemical Plant,” Znd. Eng. Chem. Process De-
sign and its Application to the Planning Steam Systems in sign Deoelop., l l , 512 (1972).
Sugar Factories,” Chemeca ’70, Aug., Sess. 7, p. 107, (1970). Gaddv, T. L., “The Use of Flowsheet Simulation Proerams in
Batsone, D. B., G. Fenton, and R. G. H. Prince, “The Steady Teaching Chemical Engineering Design,” Chem. E%g. Ed.,
State Digital Sumulation of Chemical Plant of Arbitrary 8 ( 3 ) .124 (1974).
Configuration,” paper presented at IFAC Symp. of Digital Genna:’P. L.; and R. L. Llotard, “Optimal Decomposition of
Simulation of Continuous Process, GYOR, Hungary ( 1971). Process Networks,” paper presented at AIChE 66th Ann.
Bjornbom, P. H., “Failure of the Element-By-Species Method to Meeting, Philadelphia ( 1973).
Calculate the Number of Independent Reactions,” AZChE J., Ham, P. G., “The Transient Analysis of Integrated Chemical
20,1027 ( 1974 ) . Processes,” Ph.D. dissertation, Univ. Pennsylvania, Philadel-
Bobrow, S., J. W. Ponton, and A. I. Johnson, “Simulation of phia (1971).
the Transient Behavior of Complex Chemical Plant Using a Hankinson, R. W., and K. R. Cantwell, “The Phillips Petro-
Modular Approach,” Can. J . Chem. Eng., 49, 391 ( 1971). leum Company Computer Stored Physical Data System,”
Briggs, D. E., B. Carnahan, and L. A. Lopez, “DYSCO: An Chem. Eng. Computing, Vol. 2, AIChE Workshop Series
Interactive Executive Program for Dynamic Simulation and (1972).
Control of Chemical Processes,” paper presented at the Hanyak, Ll. E., and R. L. Klaus, “Use of Translator Writing
AIChE 78th Nat. Meeting, Salt Lake City (1974). Systems in the Development of Problem Oriented Lan-
Briddell, E. T., “‘Process Design by Computer, Part I,” Chem. guages,” paper presented at AIChE 66th Ann. kfeeting,
Eng., 8 1 ( 3 ) 60 (1974a). Philadelphia ( 1973).
“Process Design by Computer, Part 2,” ibid., 81(5), Harris, R. E., “Distillation Design Using Flowtran,” Chern. Eng.
113 (1974b). Progr., 68( l o ) , 56 (1972).
- , “Process Design by Computer, Part 3,” ibid., 81(7),
77 ( 1 9 7 4 ~ ) .
Heideniann, R. A., “Three-Phase Equilibria Using Equations
of State,” AlChE J., 2 0 ( 5 ) , 847 (1974).
Broyden, C. G., “A Class of Methods for Solving Nonlinear Henley, E. J., and E. M. Rosen, Material and Energy Balance
Simultaneous Equations,” Math. Comp., 19, 577 ( 1965). Computations, Wiley, New York ( 1969).
CACHE-Computer Programs #or Chemical Engineering, AZ- Hughson, R. V., and E. H. Ste)mann, “Computer Programs for
TEC, Austin, Texas ( 1973). Chemical Engineers,” Chem. Eng., 78( 14 ), 66 ( 1971 ).
Cavett, R. H., “Monsanto Physical Data System,” paper pre- “Computer Programs for Chemical Engineers, 1973,
sented at AIChE 65th Ann. Meeting, New York ( 1972). Parti,” ibid., 80( 19), 121 1973a).
C. B. M.-‘Calcul Des Bilans de Matiered-System, Solvay & , “Computer Programs for Chemical Engineers, 1973,
CIE, S. A., Brussels, Belgium ( 1973). Part 2,” ibid., 80( 21), 127 ( 1973b).
Christensen, J. H., “The Structuring of Process Optimization,” Hull, T. E., W. H. Enright, B. Sf. Fellen, and A. E. Sedgwick,
AZChE I.. 16(2). 177 (1970). “Comparing Numerical Methods for Ordinary Differential
Chueh, CI F., and T. W. Stein, “A Comprehensive Thermal Equations,” SZAM. J. Numer. Anal., 9 ( 4 ) , 603 (1972).
and Physical Properties Information System-Halcon Physical Hutchison, H. P., and C. F. Shewchuk, “Computational Method
Properties System,” in Chem. Eng. Computing, Vol. 2, for Multiple Distillation Towers,” Trans. Instn. Chem. Eng.,
AIChE Workshop Series (1972). 5 2 ( 4 ) , 323 (1974).
Cody, W. J., “‘The Construction of Numerical Subroutine Li- IBM-S>stem/360 Continuous System Modeling Program-
braries,” S I A M Retiiew, 16( l ) ,36 ( 1974). Users Manual H20-0367-3 ( 1969).
CONCEPT-Mark 111, Computer Aided Design Centre, Cam- Institute News “Monsanto Aids CAChE,” Chern. Eng. Progr.,
bridge ( 1973). 70(8), 73 (1974).
Crowe, C. M., A. E. Hamielec, T. W. Hoffman, A. I. Johnson, Ingels, D. Sl., “A System for Simulating Chemical Process
P. T. Shannon, and D. R. Woods, Chemical Plant Simulation, Dynamics and Control,” Ph.D. dissertation, Univ. Houston,
Prentice Hall, Englewood Cliffs, N. J. (1971). Texas ( 1970).
Davies, C., “Application of Systems Engineering Techniques to Jain, Y. V. S., and J. 11. Eakman, “Identification of Process
Projects in Process Industry,” Chem. Eng. ( London ), 248, Flow Networks,” Chem. Eng. Computing Vol. 2, AIChE
149 (1971). Workshop Series (1972).
Davies, C., and F. A. Perris, “Experience in the Industrial Johnson, A. I., “The hlodular Approach Applied to the Un-
Application of Generalized and Special-Purpose Computer steady State Behavior of Systems,” Brit. Chem. Eng., Proc.
Programs for the Steady-State Design and Simulation of Tech., 1 7 ( 3 ) ,217 (1972a).
Complete Chemical Processes,” paper presented at IFAC , “Computer Aided Process Analjsis and Design-A
Symp. of Digital Simulation of Continuous Processes, GYOR, 3Iodular Approach,”ibid, 1 7 ( l ) , 28 (1972b).
Hungary ( 1971). , and T. Toong, “The klodufar Approach to System
Debrosse, c.I., and A. w. Westerberg, “A Feasible Point AI- Analysis and Design,” (GEMCS Manual), LIcSIaster Univ.,
gorithm for Structured Design Systems in Chemical E”,+ (196‘8).
neering,” AIChE I., 19(2), 251 (1973).
Kehat, E., and M. Shacham, “Chemical Process Simrilation
Page 434 May, 7975 AlChE Journal (Vol. 27, No. 3)
 Programs-1,’’ Process Technot., 18( 1/2), 35 (1973a). Pho, T. K., and L. Lapidus, “Topics in Computer Aided De-
, “Chemical Process Simulation Programs-2, Partition- sign; Part 1 An Optimum Tearing Algorithm for Recycle
ing and Tearing System Flowsheets,” ibid., 1 8 ( 3 ) , 115 Systems,” AIChE J., 19, 1170 (1973).
( 1973b). Ponton, J. W., A. I . Johnson, and P. Browne, “Application of
, “Chemical Process Simulation Programs-3, Solution the Modular Computer System to the Control of Reactor
System of Nonlinear Equations,” ibid., 18( 4/5), 181 Train,” Can. J. Chem. Eng., 50, 275 (1972).
(1973~). Poznanovic, D. S., and W. D. Seider, “A Physical Property
Kevorkian, A. K., and J. Snoek, “Decomposition of Large Information System for Undergraduate Education,” Chem.
Scale Systems, Theory and Applications in Solving Large Eng. Computing, Vol. 1,AIChE Workshop Series (1972).
Sets of Non-Linear Simulations Equations,” in D. M. Him- PPDS-Physical Property Data System, The Inst. of Chem.
melblau, (ed. ), Decompostition of Large Scale Problems, Eng., London (1971).
North Holland/American Elsevier, Amsterdam ( 1973). Ravicz, A. E., and R. L. Norman, “Heat and Mass Balancing
King, C. J., A. S. Foss, E . A. Grens, S. Lynn, and D. F. Rudd, on a Digital Computer,” Chem. Eng. Progr., 6 0 ( 5 ) , 71
“Chemical Process Design and Engineering,” Chem. Eng. (1964).
Ed., 7 (2 ) , 72 (1973). Ramirez, W. F., and C. R. Vestal, “Algorithms for Structuring
Klaus, R. L., and W. S. Moore, “A New Interval Reduction Design Calculations,” Chem. Eng. Sci., 27, 2243 ( 1972).
Method for the Solution of Nonlinear Algebraic Equations,” Reid, R. C., and L. B. Evans, “Design Data for Industry,
paper presented at AIChE 73rd Nat. Meeting, Minneapolis, Property Preduction with Computer System,” AIChE Today
Minn. ( 1972). Series, New York ( 1970).
Kliesch, H. C., “Analysis of Steady State Process Simulation: Rosen, E. M., “A Junior Course in Chemical Engineering Com-
Formulation and Convergence,” Ph.D. dissertation, Tulane putation,” Chem. Eng. Educ., 8 ( l ) ,48 (1974).
Univ., New Orleans (1967). Sanderson, R. V., and H. H. Y. Chien, “Simultaneous Chemical
Klumpar, I. V., “Process Economics by Computer,” Chem. and Phase Equilibrium Calculation,” Ind. Eng. Chem. Proc-
Eng., 77( l ) , 107 (1970a). ess Design Decelop., 12, 81 ( 1973).
“Project Evaluation by Computer,” ibid., 77( 14), Sargent, R. W. H., “Forecasts and Trends in System Engineer-
76 ( i970b). ing,” Chem. Eng. (London), 262,226 (1972).
Ledet, W . D., and D. M. Himmelblau, ‘“Decomposition Pro- Seader, J. D., and D. E. Dallin, “Interactive Graphical System-
cedures for Solving a Large Scale Systems,” Ado. Chem. Computer Design of a Toluene Dealkylation Process,” Chem.
Eng., 8,186 ( 1970). Eng. Computing, Vol. 1, AIChE Woikshop Series ( 1972).
Leesley, M. E., “Process Plant Design by Computer,” Process Seader, J. D., W. D. Seider, and A. C. Pauls, FLOWTRAN
Technol., 18( l ) ,403 (1973). Simulation An Introduction-CA.CHE Committee, Ulrichs
Loibl, J. M., D. T. Camp, G. S. Wilkins, “The Dynamic Anal- Bookstore, Ann Arbor, Mich. ( 1974).
lysis of Chemical Processes with an User Oriented Executive Seider. W. D. ( M O D ) . “Comuuter Aided Analvsis and Design
Program,” Proc. 1973 Summer Computer Simulation Conf., Packages,” Chem. Eng. Co;puting, Vol. 2, AiChE Workshgp
Montreal, Canada ( 1973). Series ( 1972).
Maejima, T., and A. Shindo, “‘System Structure of a Computer Shacham, M., and R. L. Motard, “Applications of the Theory
Aided System for Process Engineering,” Proc. 1973 Summer of Linear Recycle Systems,” paper presented at the AIChE
Computer Simulation Conf., Montreal, Canada ( 1973). 78th National Meeting, Salt Lake City (1974).
Mah, R. S. H., “Recent Development in Process Design,” Shacham, M. and E. Kehat, “Converging Interval Methods for
Symp. on Basic Questions of Design Theory, Columbia the Iterative Solution of a Nonlinear Equation,” Chem. Eng.
Univ., New York ( 1974). Sci., 28( 1 2 ) , 2187 (1973).
, “Structure Decomposition in Chemical Engi- Soylemez, S., and W. D. Seider, “The Chemical Engineering-
neering Computation,” paper presented at AIChE 72nd Na- Process Analysis and Design Interface,” Chem. Eng. Com-
tional Meeting, St. Louis, Mo. (1972). puting, Vol. 2, AIChE Workshop Series (1972).
, and M, Rafal, ‘“Automatic Program Generation in , “A New Technique for Precedence Ordering Chemi-
Chemical Engineering Computation,” Trans. Instn. Chem. cal Process Equation Sets,” AIChE J., 1 9 ( 5 ) , 934 (1973).
Eng., 49,407 ( 1971 ) . Stadtherr, M. A,, W . A. Gifford, and L. E. Scriven, “Efficient
Menzies, hl. A., and A. I. Johnson, “Modeling and Parametric Solution of Sparse Set of Design Equations,” Chem. Eng.
Studies of a Vapor Recompression System,” Can. J. Chem. Sci., 29, 1025 ( 1974).
Eng., 49,407 ( 1971 ). Stephanopoulos, G., and A. W. Westerberg, “Overcoming
Monsanto Corporate Engineering Dept., FLOWTRAN ( 1974). Deficiencies of the Two Level Method for System Optimiza-
Motard, R. L., and H. M. Lee, CHESS, Chemical Engirieering tion,” AlChE J., 19(6 ) , 1269 ( 1973).
User’s Guide, 3rd Edit., Univ. Houston, Texas (1971). Tsubaki, M., “Computer Aided Process Analysis and Design,”
Nott, H. D., “Sled: Simplified Language for Engineering De- Chem. Eng. Progr., 69( 9 ) , 78 ( 1973).
sign.-A Computerized System for the Design of Chemical Uhde Thermophysical Properties Pro ram Package, Friedrich
Processes,” Ph.D. dissertation, Univ. Michigan, Ann Arbor Uhde Gh4BH, Dortrnund, Germany f1973).
(1971). Upadhye, R. S., and E . A. Grens, “An Efficient Algorithm for
Nuttal, H. E., Jr., and D. M. Himmelblau, Interactive Reactor Optimum Decomposition of Recycle Systems,” AIChE J.,
Simulation Proc. 1973 Summer Computer Simulation Conf., 18,533 ( 1972 ) .
Montreal, Canada ( 1973). , “Selection of Decompositions for Process Simulation,”
Orbach, O., C. M. Crowe, and A. I. Johnson, “‘Multicomponent AlChE J., 21, 136 ( 1975).
, I

Separation Calculations by the Modified Form of Newton’s Wells, G. L., and P. ‘M. Robson, Computation for Process En-
Method,” Chem. Eng. J., 3, 176 (1972). gineers, Wiley, New York ( 1973).
Orbach, O., and C. X I . Crowe, “Convergence Promotion in the Westerberg, A. W., “Decomposition Methods for Solving
Simulation of Chemical Processes with Recycle-The Dom- Steady State Process Design Problems,” p 379 in D. M.
inant Eigenvalue Method,” Can. J. Chem. Eng., 49, 509 Himmelblau ( e d . ) , Decomposition of Large Scale Problems,
(1971). North Holland/American Elsevier, Amsterdam ( 1973).
PACER 245 User Manual, Digital Systems Corp., Hanover, , and C. J. Debrosse, “An Optimization Algorithm for
N. H. (1971). Structured Design Systems,” AIChE J., 19, 355 (1973).
Parmelee, R. P., T. I. Peterson, C. C. Tillman, and D. J. Hat- , and F. C. Edie, “Computer Aided Design, Part 1,
field, “Virtual Storage and Virtual Machine Concepts,” I B M Enhancing Convergence Properties by the Choice of Output
Systems I., 11, No. 2, p. 99 (1972). Variable Assignments in the Solution of Sparse Equation
Perry, R. H., and C. H. Chilton, Chemical Engineers’ Hand- Sets,” Chem. Eng. J., 2, 9 (1971a).
book, 5th Edit., pp. 2-85, McGraw-Hill, New York ( 1973). , “Computer Aided Design, Part 2, An Approach to
Peters, N., and P. E . Barker, “An Appraisal of the Use of Converqence and Tearing in the Solution of Sparse Equation
PACER, GESICS and CONCEPT for Chemical Plant Sim- Sets,” ibid., 2, 17 (1971b).
ulation and Design,” Chem. Eng., (London), 283, 149 Wilson, G. M., “Calculation of Enthalpy Data from a Modified
(1974). Redlich-Kwong Equation of State,” Ado. Cryogenic Engi-

AlChE Journal (Vol. 21, No. 31 May, 1975 Page 435

 neering, 11, 392 (1966). Systems Simulation Laboratory. He is spending the 1974-1975
, ‘Yapor Liquid Equilibria, Correlation by Means of academic year at the Computer Aided Design Centre, Cam-
Modified Redlich Kwong Equation of State,” ibid., 9, 168 bridge, England.
(1964). Mordecai Shacham received his B.Sc. and DSc. in chemical
Woods, D. R., T. W. Hoffmann, and A. I. Johnson, “Teaching
Experience with Design and Simulation Project,” Chem. Eng. engineering from the Technion, Israel Institute of Technology.
Educ., 7 ( 2 ) , 96 (1973). He spent the 1973-1974 academic year on a post-doctorate
Worley, F. L., Jr., and R. L. Motard, ‘‘Information Systems in fellowship at the University of Houston and is currently with
Chemical Engineering Design,” Chem. Eng. Computation, Ben-Gurion University, Beersheva, Israel.
Vol. 2, AIChE Workshop Series (1972). Edward M. Rosen received his B.S. and M.S. degrees from
the Illinois Institute of Technology and his Ph.D. from the
THE AUTHORS University of Illinois. Currently he is a Science Fellow and
Manager, Systems Technology at the Monsanto Company
Rodolphe L. Motard received his B.S. in chemical engineer-
ing from Queen’s University, Canada, and his M.S. and D.Sc. where he has been since 1959, except for a postdoctoral aca-
in chemical engineering from Carnegie Mellon University. demic leave at Stanford University in 1962-1963. With E. J.
From 1951-1957 he worked for Shell Oil Company and sub- Henley, he is co-author of the text Material and Energy B d -
sequently joined the faculty of the University of Houston, ance Computations ( 1969). His hobbies include violin and
where he is Professor of Chemical Engineering and Director, foreign language study.

Quantitative Overdesign of Chemical

Processes
R. A. FREEMAN
Overdesign is applied to chemical processes to account for expected
variations in the design data or design conditions. These variables must Monsanto Company
be treated quantitatively in order to design a process that is certain to St. Louis, Missouri 6 3 1 6 6
perform adequately. Process dependability is introduced as a means of and
quantifying variable process behavior. Overdesign using the dependability J. L. GADDY
criterion is a stochastic optimization problem. An example problem with
three stochastic variables and two design variables is presented to illustrate Deportment of Chemical Engineering
University of Missouri
these procedures.
Rollo, Missouri 6 5 4 0 1

SCOPE
I n recent years, developments such as sophisticated The expected value is not always suitable for evaluating
optimization methods and process synthesis techniques chemical processes since adverse deviations from the
have significantly advanced the procedures of process de- average might mean unacceptable or even hazardous per-
sign. However, the final stage of design, that of specifica- formance. Also, these methods have only been applied
tion of overdesign or safety factors, is still largely a sub- to a few processes where the statistical behavior of a
jective procedure based upon experience. process performance criterion is known or can be obtained
analytically.
I n general, safety factors are increased as the uncer-
The purpose of this study is to demonstrate how over-
tainty of the process variables increases; a procedure that
design can be achieved by the use of a process depend-
can be expensive if safety factors are specified too small ability criterion and stochastic simulation procedures. A
or too large. Recent studies to quantify overdesign repre- simple process involving a reactor and separator was
sent the uncertain parameters as random variables and chosen for study. This process has both variable conditions
compute the overdesign based upon the expected value and uncertain data, with two design variables, reactor
criterion. volume and number of separator stages.

CONCLUSIONS AND SIGNIFICANCE

The use of process dependability, the fraction of time procedure was applied to the solution of this stochastic
the process must meet certain performance criteria, was optimization problem.
found to be an improved method of designing and opti- The optimum design of this process was found to be
mizing a chemical process. Since there are many com- surprisingly different from the design based upon the
binations of the design variables that produce the same usual methods. An overdesign factor of 350% is optimal
process dependability, the selection of the proper design for the reactor while less than 10% is required for the
becomes an optimization problem. An exhaustive search column.

Page 436 May, 1975 AlChE Journal (Vol. 21, No. 3)