Process Simulation

Integrate
Process Simulation
and Process Synthesis
Dominic Chwan Yee Foo
The design of a new process using
Universiti Teknologi Malaysia computer simulations to analyze a
Zainuddin Abdul Manan
Universiti Teknologi Malaysia
process flowsheet based on the onion
Murugan Selvan model can save time, effort
WinSim, Inc.
Michael Lynn McGuire
and, ultimately, money.
WinSim, Inc.

Process synthesis models

T
he design of a chemical process involves synthesis and
analysis. Process synthesis is the overall development of Two important process synthesis models are the hierarchi-
a process flowsheet by combining individual steps cal approach outlined in Table 1 (1, 2) and the “onion model”
(equipment and operating conditions) into an optimal arrange- illustrated in Figure 1 (3, 4).
ment. Process analysis breaks down the flowsheet to evaluate The first attempt at combining process synthesis and analy-
the performance of each individual element as well as how the sis during the development of a new process employed the
overall process would perform, typically by a process simulator. hierarchical approach to synthesizing a hydrodealkylation
Process analysis is often performed after the synthesis task
has been completed. The major disadvantage of this approach
is that if the flowsheet is found to be infeasible during the
analysis stage, the synthesis task must be repeated before the
next analysis step can take place. This rework can be avoided
if the flowsheet is synthesized with the use of a process syn-
thesis model and simulation tool.
Reactor
This article shows how these tools can be used hand-in-
hand to generate a reasonably good process flowsheet. This is
particularly useful for evaluating a new process path or gener-
ating alternatives for new process development. Separation and
Recycle
Table 1. Hierarchical approach to process design (1, 2).
Heat Exchange
1. Batch vs. continuous Network
2. Input-output structure of the flowsheet
3. Recycle structure of the flowsheet
4. General structure of the separation system Utilities
Vapor recovery system
Liquid recovery system
5. Heat exchanger network
■ Figure 1. The onion model of process design (4, 5).

CEP www.cepmagazine.org October 2005 25

Process Simulation

process in an equation-based process simulator (5). The main Catalyst modeling can usually be omitted from reactor
drawback of this approach is that equation-based process sim- modeling, provided mass and energy balances are the only
ulators often require much time to write the software code targets of the simulation. However, if a catalyst used in the
before the simulation can be performed. Modular-based process involves a phase change, including it in the analysis
process simulators provide an effective means of handling the will result in a more complete analysis of the reactor system’s
process simulation tasks. heating requirements.
The onion model is an alternative way to present the hier-
archical approach to process design. Process design begins at Layer 2 — separation and recycle
the center of the onion, with the reactor, and proceeds out- After the reactor system synthesis and modeling is finished,
ward. The reactor design influences the separation and recycle the focus shifts to the second layer of the onion model. Products
structures (the second layer of the onion), which are designed and any byproducts formed in the reactor need to be separated
next. The reactor, recycle and separator structures dictate the from unconverted reactant for further purification, while the
overall heat-recovery requirements, so the heat-recovery net- unconverted raw material is recycled back to the reactor.
work design comes next. Finally, the process utility systems Separation system. Separation systems can be broadly
are designed to provide additional heating and cooling classified as liquid or vapor separations. When a reactor
requirements that cannot be satisfied through heat recovery. effluent contains a mixture of liquid and vapor, a phase sepa-
This model emphasizes the sequential and hierarchical nature rator such as a flash column is normally used to separate the
of process flowsheet synthesis. phases before they enter into their respective separation sys-
However, the onion model of process synthesis requires tems (Figure 2).
the use of a process analysis tool as well. Synthesis decisions If a flash column with two degrees of freedom is
made at each layer of the onion model may require a detailed employed, the designer must specify two process variables for
analysis — and this is the role of a process simulator. the modeling. For simplicity, temperature and pressure (or
Simulation is performed at each individual layer after new pressure drop) are frequently used.
units are added or new decisions are made. This ensures that a Vapor separation systems include condensers, flash tanks,
feasible process flowsheet (in terms of mass and energy bal- absorbers, adsorbers, and gas separation membranes. These
ances, operating conditions, etc.) is developed at each layer of unit operations are normally used to purify a vapor recycle
the onion. Optimization may also be performed on each stream before it re-enters the process. A purge stream is
newly added unit to identify the optimum design variables always employed to avoid undesired contaminant build-up.
(hardware optimization) as well as its operating conditions Liquid separation systems include distillation (including
(parametric optimization). extractive distillation), solvent extraction, stripping, filtration
(including membrane separation), centrifugation, and so on.
The heart of the process — the reactor The selection of the appropriate separation process has been
Synthesis of a new process flowsheet should start at the discussed extensively in design textbooks (e.g., 2, 6, 7) and
heart of the chemical process, i.e., the reactor system. This is will not be covered here. Because distillation is so widely used,
where raw materials are converted into valuable products. distillation column modeling will be discussed in detail here.
When synthesizing and modeling a reactor system, the
process designer must consider the following questions:
• What is the right reactor model (continuous stirred-
Vapor Purge
tank reactor, plug-flow reactor, etc.), and what are its Separation
operating conditions (isothermal, adiabatic, constant outlet System
temperature, vacuum, etc.)?
• How should the product conversion and yield be Feed
determined? Reactor
Feed Flash
• Is a catalyst needed in the reactor system modeling? System
Answers to the first two questions can be found in the
literature for a wide variety of reactions. When multiple sets
Liquid Products
of operating conditions exist, process simulators can serve
Separation
as a tool for comparing the viable options. Simulation pro- System Byproducts
vides more information than the available literature in terms
of heating or cooling requirements, operating conditions,
■ Figure 2. The overall separation scheme consists of vapor,
and so on. liquid and flash separations (1, 2).

26 www.cepmagazine.org October 2005 CEP

 Specifications required for the modeling of a distilla- Tear Recycle Stream
tion column typically include: R1 R2
• number of theoretical trays. A good initial estimate can //
Recycle Stream
often be obtained using short-cut methods such as the
Fenske equation (2) or a simplified separation model avail- A B C D E F
able in a simulator.
• column top and bottom temperatures. Estimates obtained
Unit Operation
from short-cut modeling will often lead to faster convergence in Simulator
during column simulation.
• column pressure. Either the top or bottom column pres- ■ Figure 3. The tear stream concept is used in recycle simulation (8).
sure is normally set by the column designer, or a column pres-
sure drop is specified based on the column top pressure. Previous CEP articles (9, 10) provided some good sugges-
• feed tray location(s). Feed tray locations are selected tions to aid the recycle simulation. Here are a few more:
based on such considerations as energy conservation (both • Maintaining product specification remains the highest pri-
feed stream and feed tray have the same temperature) and ority of the process.
required product purity (a higher feed tray location might • Take note of the changes in feed temperature and pressure.
affect the top product stream composition), among others. • Beware of the accumulation of unwanted pollutants in the
• estimated product flowrates. Some simulators require an process loop. A purge stream is important to ensure that the
estimate of the top and bottom product flowrates for the ini- recycle system does not trap unwanted material.
tialization of the column convergence calculations. (This value An additional tip to speed up the recycling loop conver-
will be different from the desired product flowrate specified as gence is to increase the convergence tolerance at the initial
the convergence criterion during the simulation.) stage of the recycling simulation. When the flowsheet has
Most column modeling for non-complex mixture separa- converged at this larger tolerance, the convergence tolerance is
tions will converge without much difficulty. Occasionally, col- then reduced. This will enable the flowsheet to converge faster
umn modeling fails to converge. The following steps can be than if a tight convergence tolerance is specified at the initial
taken to aid the convergence of a column: stage. One can also explore various optimization options with-
1. Evaluate thermodynamic choices, especially K-values. in the recycle system associated with the reaction and separa-
2. Generate initial guesses using a short-cut method. tion systems of the process (4).
3. Look for unachievable and impossible specifications (for
example, reboiler duty that vaporizes the entire feed, product Layer 3 — the heat exchange network
specifications that violate the column material balance, etc.). The process heating and cooling loads are determined after
4. Simplify choices for heat and material balance specifica- the process structure within the two inner layers of the onion
tions. Avoid complex approaches that set specifications (e.g., model (i.e., the reactor, and separation and recycle systems)
component recoveries, reflux ratios, and reboiler ratios) for the has been finalized. It is now time to design and model the heat
top and bottom streams that might be in conflict with each other. exchange network (HEN). This is usually done using the well-
Recycle system. Recycling is the tricky part of flowsheet established tool of process integration, which divides the HEN
modeling. A good start (for beginners) in modeling a recycle design procedure into two stages — utility targeting, and net-
loop is to use the concept of a “tear stream” (8). work design. The details of this are beyond the scope of this
As shown in Figure 3, the recycle stream after unit F is article, and readers are referred to Refs. 2, 3, 6 and 7.
considered as two separate tear streams, R1 and R2. After unit After a preliminary network has been synthesized, the
A and B are solved, the simulation moves to unit C. Some ini- process flowsheet will normally undergo a complete re-simu-
tial guesses for the tear stream R1 are made so the simulation lation to verify the energy balances. Often, more recycle loops
can proceed to units D, E and F. After unit F converges, the will be involved as the process streams that were used for
resulting flowrate of stream R2 is compared to the initial guess process-to-process heat exchange are now interconnected. The
for R1. If the values agree to within a specified tolerance, it is tear stream concept is also useful at this stage.
likely that the simulation model has converged. The calculated
value of R2 is then used in place of R1 in unit C and the simu- Layer 4 — utilities
lation is rerun. After the heat exchange network has been synthesized, the
If tear streams R1 and R2 do not agree to within the speci- outermost layer of the onion model — i.e., the utility system
fied tolerance, the initial guess for R1 is revised and the simula- — is addressed. The selection of hot and cold utilities is
tion is rerun (without connecting the recycle stream to unit C). another well-established application of process integration (3,

CEP www.cepmagazine.org October 2005 27

Process Simulation

Table 2. Molar feed flowrate for each component

in the production of n-octane (example).
6, 7). Other options to be explored include the placement of
the heat pump and heat engine.
Flowrate, A process simulator is a useful tool to evaluate the selected
Components kg-mol/h Specification utilities. Often, a simulated process flowsheet provides a good
Nitrogen, N2 0.1 picture of how well a process is likely to perform after start-up.
Ethylene, C2H4 20 T = 30ºC
n-Butane, C4H10 0.5 P = 20 psia Example: production of n-octane
i-Butane, C4H10 10 The onion model synthesis and simulation technique will
be used to develop a process flowsheet for n-octane (C8H18)
production from ethylene (C2H4) and i-butane (C4H10).
Purge
Component flowrates (with some impurities) and stream spec-
Unconverted ifications for the fresh feed are given in Table 2.
Reactant Chiller
Compressor
DESIGN II for Windows is used as the simulation tool. This
modular-based software has 886 components in its databank and
Distillation uses the sequential-modular approach to perform its calculations.
Various thermodynamic models can be loaded into the simulation.
Fresh
Feed Flowsheet development. Ethylene and i-butane react isother-
Mixer
mally in a stoichiometric isothermal reactor at 93ºC to produce
Heater Flash
Reactor
n-Octane n-octane. The key component that limits the reaction conversion
n-Octane
is taken as ethylene, with an overall conversion of 98%. The pres-
sure drop across the reactor is specified at 5 psi. The reaction is:
■ Figure 4. The preliminary flowsheet for the production of
n-octane after completion of onion model layers 1 and 2. 2C2H4 + C4H10 → C8H18
Fresh
73.57°C Feed After the reactor simulation has converged, the synthesis and
93.3°C
30°C analysis task focuses on the separation and recycle systems. A
101.4°C Purge flash column is added to the reactor effluent to separate the
82.23°C Compressor unconverted raw materials from the desired product. A pressure
Unconverted
Reactant

Heater drop of 2 psi is introduced, while the operating temperature is

93.3°C maintained the same as that of the reactor. The more-volatile
Distillation compounds (ethylene, i-butane, and other impurities) are
flashed to the top product stream together with a small portion
Mixer Flash of the heavier product, n-octane, while the remaining n-octane
Reactor
n-Octane leaves at the bottom. An additional separation unit is needed to
n-Octane
recover the n-octane product from the top stream.
■ Figure 5. The complete flowsheet with a heat-integrated Distillation is then added to the flash column’s top product
distillation column. stream to recover n-octane. The short-cut design method deter-
mines that this column has 10 theoretical trays and operates at
Fresh Feed 30°C 15 psia. The remaining n-octane component is recovered at the
93.3°C Purge column bottom while the volatile components leave from the
101.4°C column top. Since the n-octane separation involves both the
91.4°C Compressor
flash and distillation models, parametric optimization is per-
Unconverted
Reactant

Heater formed to determine the best combination of operating parame-

93.3°C ters in these models for optimal n-octane recovery.
Distillation
The unconverted raw material leaving at the distillation top
stream is now pure enough for recycle. A purge stream is
added before the stream is recompressed, reheated and sent
Mixer Flash
Reactor n-Octane back to the reactor. The tear stream concept is utilized to facil-
n-Octane itate convergence of the recycle stream. Figure 4 is a prelimi-
nary process flowsheet based on the synthesis and simulation
■ Figure 6. The complete flowsheet with a stand-alone
conducted this far.
distillation column.

28 www.cepmagazine.org October 2005 CEP

 Next, the design of the heat exchange network and utility black box. Rather, it is a well-developed set of rules, correla-
system will be handled simultaneously. Stream enthalpy data tions and relationships that can execute complex calculations
needed for the analysis is extracted from the converged flow- very quickly without violating first principles.
sheet in Figure 4. After the HEN is designed using process Simulation does not replace that most useful of all tools of
integration techniques, the simulation is re-run to verify the a chemical engineer — common sense (14). Always use engi-
overall mass and energy balances for the heat-integrated flow- neering judgment to evaluate simulation errors or suspicious
sheet (Figure 5). The tear stream concept is utilized at this results to find their source. Computing efforts are nothing but
stage since the integrated process streams are considered as speedy number crunchers that have logical clues, fingerprints
recycle streams in the sequential modular approach (e.g., the and reasons. It is just a matter of tracking them down with less
raw material recycle stream in Figure 5). time using fundamentally sound principles. CEP

Alternatively, if distillation is not preferred for the heat

integration scheme because of controllability reasons, the
alternative flowsheet is that shown in Figure 6. This gives the
Literature Cited
process designer another option for comparison (e.g., energy,
1. Douglas, J. M., “A Hierarchical Decision Procedure for
controllability, complexity, etc) during process development. Process Synthesis,” AIChE Journal, 31 (3), pp. 353–362
(Mar. 1985).
Closing thoughts 2. Douglas, J. M., “Conceptual Design of Chemical Processes,”
McGraw Hill, New York (1988).
The final advice to simulation users is this: Check the sim- 3. Linnhoff, B., et al., “A User Guide on Process Integration for
ulation results and don’t accept that everything is as it appears. the Efficient Use of Energy,” IChemE, Rugby, U.K. (1982).
4. Smith, R., and B. Linnhoff, “The Design of Separators in
The garbage-in, garbage-out (GIGO) principle applies to all the Context of Overall Processes,” Chem. Eng. Res. Des., 66,
computer models (11). They are not smart enough to identify pp. 195–228 (May 1988).
5. Lott, D. H., “Simulation Software as an Aid to Process
wrong information provided by the user, and in turn, they pro- Synthesis,” in “Understanding Process Integration II,” Crump,
duce poor results without the user’s awareness (12, 13). P. R., et al., eds., IChemE, Rugby, U.K., pp. 1–22. (1988).
Also, the simulator’s physical property system is not a 6. Smith, R., “Chemical Process Design,” McGraw Hill, New
York (1995).
7. Seider, W. D., et al., “Product and Process Design
Principles: Synthesis, Analysis and Evaluation,” Wiley,
Hoboken, NJ (2003).
DOMINIC CHWAN YEE FOO is a research associate at the Chemical Engineering
Pilot Plant (CEPP) at the Universiti Teknologi Malaysia (Phone: +60-7-553-1662;
8. Turton, R., et al., “Analysis, Synthesis and Design of Chemical
Fax: +60-7-556-9706; E-mail: [email protected]). His main research interests Processes,” Prentice Hall, Upper Saddle River, NJ (1998).
include process synthesis and design as well as process integration via pinch 9. Schad, R. C., “Don’t Let Recycle Streams Stymie Your Sim-
analysis. He has extensive training in using process simulators for process ulation,” Chem. Eng. Progress, 90 (12), pp. 68–76 (Dec. 1994).
modeling and debottlenecking of continuous and batch processes. He obtained 10. Schad, R. C., “Make the Most of Your Process Simulation,”
his BEng and MEng degrees in chemical engineering from the Universiti Chem. Eng. Progress, 94 (1), pp. 21–27 (Jan. 1998).
Teknologi Malaysia. He is a member of the Institute of Chemical Engineers 11. Petrides, D. P., et al., “The Roles of Process Simulation in
Malaysia (IChEM) and the Institution of Chemical Engineers (IChemE). Pharmaceutical Process Development and Product
ZAINUDDIN A. MANAN is an associate professor and head of the Chemical Commercialization,” Pharmaceutical Engineering, 22 (1),
Engineering Dept. at the Universiti Teknologi Malaysia (Phone: +60-7-553- pp. 1–8 (Jan./Feb. 2002).
5512; Fax: +60-7-558-1463; E-mail: zain@[email protected]). For 15 years, he has 12. Agarwal, R., et al., “Uncovering the Realities of
been extensively involved as a researcher, consultant and trainer for the Simulation,” Parts 1 and 2, Chem. Eng. Progress, 97 (5), pp.
chemical process industries in the area of process systems design and process 42–52 and (6), pp. 64–72 (May, June 2001).
improvement, with an emphasis on efficient energy utilization (pinch analysis) 13. Le, N. D., et al., “Doublecheck Your Process Simulations,”
and waste minimization. He received a BSc in chemical engineering from the Chem. Eng. Progress, 96 (5), pp. 51–52 (May 2000).
Univ. of Houston, an MSc in process integration from the Centre for Process 14. Carlson, E. C. “Don’t Gamble with Physical Properties for Sim-
Integration at the Univ. of Manchester Institute of Science and Technology, and ulations,” Chem. Eng. Progress, 92 (10), pp. 35–46, (Oct. 1996).
a PhD in chemical engineering from the Univ. of Edinburgh. He is a member of
the IChEM and IChemE. Further Reading
MURUGAN S. SELVAN is a support engineer at WinSim, Inc. (Sugar Land, TX; Gupta, V. S., and N. D. Messey, “Solving Unsteady State Problems
Phone: (281) 565-6700 x106; Fax: (281) 565-7593; E-mail:
Using ChemCAD, a Steady-State Process Simulator,” AIChE
[email protected]). He is responsible for providing technical guidance,
quality and training in the areas of physical properties and chemical process
National Spring Meeting, Houston, TX (1995).
industries modeling needs of process simulation users. He received a BS from Horwitz, B. A., “Avoid Nausea when Solving Dynamic Problems,”
Annamali University - India and an MS and PhD from the Univ. of Alabama, all Chem. Eng. Progress, 92 (3), pp. 41–51 (Mar. 1996).
in chemical engineering. He is a member of AIChE. Horwitz, B. A., “Make the Most of Your Process Design
Software,” Chem. Eng. Progress, 90 (1), pp. 54–57 (Jan. 1994).
MICHAEL LYNN McGUIRE is president of WinSim, Inc. (Phone: (281) 565-6700 Horwitz, B. A., “Are You ‘Scotomatized’ by Your Simulation
x102; Fax: (281) 565-7593; E-mail: [email protected]). He has been responsible
Software?,” Chem. Eng. Progress, 92 (9), pp. 68–71 (Sept. 1996).
for implementing and developing process simulation software for over 20
years. He earned his BS in mechanical engineering from Texas A&M Univ. and
Horwitz, B. A., “Hardware, Software, Nowhere,” Chem. Eng.
is a member of AIChE. Progress, 94 (9), pp. 69–74 (Sept, 1998).
Kister, H. Z., “Can We Believe the Simulation Results?,” Chem.
Eng. Progress, 98 (10), pp. 52–58 (Oct. 2002).
Acknowledgement Westerberg, A. W., et al., “Process Flowsheeting,” Cambridge
The authors acknowledge Sivakumar Kumaresan for reviewing this article. University Press, Cambridge, U.K., (1979).

CEP www.cepmagazine.org October 2005 29