USE OF DYNAMIC SIMULATION IN THE DESIGN OF ETHYLENE PLANTS

Vinod Patel
Chief Technical Advisor
KBR
Jeffrey Feng
Technical Advisor
KBR
Surajit Dasgupta
Senior Technology Manager
KBR
Jack Kramer
Chief Technology Engineer
KBR

Abstract: An ethylene plant experiences startup, shutdown, restart, feed variation and other
non-steady state operations over its life cycle. Proper design and protection of process
equipment units during these operations are key to the safe and sustained operation of the
entire facility. Dynamic simulation is routinely used to determine plant conditions during such
transients and ensure that equipment design has covered the entire range of operation. This
article reviews some of the relevant application areas of dynamic simulation, and discusses
several case studies based on past projects. The case studies cover several critical equipment
units relevant to ethylene plants, including compressors, steam turbine/motor drivers and
furnaces. One example shows how dynamic simulation was used to evaluate the operation of
process gas and refrigerant compressors and the steam turbine/motor drivers. Other examples
show how dynamic simulation was used in the validation of steam letdown philosophy and
evaluation of furnace control.

97
 Introduction

An ethylene plant experiences startup, shutdown, restart, feed variation and a multitude
of other non-steady state operations over its life cycle. Proper design, control and protection of
process equipment units during these operations are key to the safe and sustained operation
of the entire facility. Dynamic simulation is routinely used to determine plant conditions during
such transients and ensure that equipment design covers the entire range of operation (Ref 1,
2). An additional issue that has become important of late is plot size constraints – plants of
increasingly larger capacities are being designed within very tight plot spaces. Optimum sizing
of process equipment and piping is critical to the feasibility and constructability of such plants
under this scenario.
This article reviews some of the relevant application areas of dynamic simulation and
discusses several case studies based on recent projects. The case studies cover several critical
equipment units relevant to ethylene plants, including compressors, steam turbine/motor
drivers and furnaces. Examples of uses of a few other advanced simulation technologies such
as Operator Training Simulators and Computational Fluid Dynamics are also discussed.

Dynamic Simulation Application Areas

Main sections of a representative ethylene plant are shown in the simplified Block Flow
Diagram (BFD) in Figure 1 below.

Figure 1: Simplified BFD Showing Process/Utility Interactions

There are many areas of application for dynamic simulation for an ethylene plant. In
this article, we have selected a few areas where dynamic simulation has been used routinely in
the industry. The examples below cover process equipment such as the ethylene cracking
furnace and process gas and refrigerant compressors and drivers, and utility systems in an
ethylene plant such as steam, fuels gas and flare. Examples from each of these systems are
given to highlight typical applications and benefits.

98
Compressor Dynamic Simulation
An ethylene plant has three major compression systems: Process Gas Compressor
(PGC), Ethylene Refrigerant Compressor and Propylene Refrigerant Compressor. These three
compressor systems represent a significant capital investment and operating expense of an
ethylene plant, and are essential components in sustaining the revenue stream of the plant.
Therefore, operation and protection of the compressors are critical issues in the plant design
and operation.
These compressors can experience a wide range of operating conditions over their life
cycle, including startup, shutdown, restart and speed/load variation. The operating conditions
should remain within the flow, pressure and speed boundaries as defined by the compressor
and driver manufacturers. Figure 2 shows a typical compressor head vs flow map. The
minimum stable flow limit (surge limit) is shown on the left and maximum flow limit
(stonewall) is shown on the right.
Each compressor is equipped with an anti-surge control system to maintain adequate
stable flow through the compressor. The refrigerant compressors generally have multiple
compression stages, and each stage requires an anti-surge valve. Figure 3 shows the general
configuration of the anti-surge controller. In this example, the compressor is driven by a steam
turbine (ST). The system shown uses liquid quench to cool the recycle stream. An alternative
is to use a recycle cooler.
)
LL
(S
ne
Li
it
m
Li
e
rg
Su
Head

Figure 2: Typical Compressor Performance Map Figure 3: Anti-Surge Controller Configuration

Anti-surge systems typically become active during an upset or other non-steady state
condition. Therefore, dynamic simulation has been used routinely in the industry to address
issues associated with anti-surge system, such as the number of recycle loops, sizing of anti-
surge valves and stability of overall anti-surge control system (Ref 3, 4). A few examples of
compressor dynamic studies are presented below.

99
Startup
Compressors in an ethylene plant are generally driven by variable speed drivers such as
electric motor or steam turbine. Using a dynamic simulation model, we can predict the
operating point of the compressor during the entire startup process and evaluate the risk of
compressor operating outside of the boundaries. Figure 4 shows the simulation results for the
startup of a motor-driven propylene refrigerant compressor. The operating points during the
startup are plotted against the speed and flow/head boundary.
There are several transitions during the startup, as identified on the graph as steps A,
B, C and D:
• Step A: the compressor is

L
SL
started up on total recycle with M

L
SC
ax
the anti-surge controller on im
um
manual and the recycle valves No
rm sp
ee
al d
fully open. D
sp
ee
C d
• Step B: the compressor reaches
the minimum speed, and the
anti-surge controller is switched B
to automatic, which partially Mi
nim
closes the recycle valve. The um
sp
operating points are near the ee
d
surge control line (SCL) that is
set at 10% away from the A
surge limit line (SLL).
• Step C: the performance Volumetric flowrate
(speed) controller starts to
Figure 4: Representative Compressor Startup Map
increase the speed of the
compressor until the suction pressure equalizes with the process exchanger pressure
and the discharge pressure equalizes with the condenser pressure.
• Step D: the remainder of the plant is started up to full production and the
refrigerant compressor is loaded. The compressor operating point moves away from
the surge control line and the recycle valves are closed.
Startup is one of the operating modes where the compressor operates outside of its
normal conditions. By using dynamic simulation, we can accurately predict the operating
points of the compressor through the entire transition, and identify potential design/operation
problems that need to be addressed on a priority basis. Eventually, the use of dynamic
simulation ensures that the compressor can be operated safely after it is installed.
Using the simulation results, we can also evaluate whether the equipment units
associated with the compressor system are properly sized. Figure 5 below shows the predicted
flowrate across one of the anti-surge valves and the cooling duty of the compressor discharge
cooler for the same startup scenario. The transition periods A to D are shown on each graph.

100
 C C

Recycle flowrate

Cooling duty
B D
B

A A
D

time time

Figure 5: Anti-Surge Flow and Cooling Duty Response

As demonstrated in the previous graphs, the recycle flowrate and cooler duty can vary
considerably during the startup. Therefore, proper valve and cooler sizing requires evaluation
of both steady state and transient operations.

Control strategy
The compressor speed is generally controlled by the suction pressure, which can be
affected by opening of recycle valves. As a result, there are interactions between the speed
control and anti-surge control. By using dynamic simulation, we can determine the best control
strategy for the system. In this example, illustrated
in Figure 6, we use a turndown-to-recycle operation

L
SL
to evaluate the stability of the control system. In

TL
R

L
the first simulation shown in the A curve, we use an

SC
anti-surge controller for the recycle valves and a
separate speed controller for the driver (i.e. there is
Head

no communication between the speed and anti-

surge controllers). In the second simulation shown
in the B curve, we use an integrated anti-surge and
speed controller that has the ability to decouple the
speed and recycle actions (i.e. reducing the
compressor speed when the control senses that the
operating points move towards the surge limit). By
comparing the two curves, we can see that the
operating points on the A curve are close to the
Figure 6: Compressor Turndown to Recycle Operation
operating boundary as defined by the SLL and
recycle-trip line (RTL) on the map, whereas those in the B curve are further away from the
boundary during the transition. Based on the simulation results and discussions with the
control system supplier, decisions were made to include performance control as part of the
anti-surge control specifications, which was proven to be beneficial in the field operation.

Shutdown
One of the concerns in the compressor system design is the possibility of reverse
rotation after a trip. Reverse rotation can happen if the discharge pressure is considerably

101
higher than the suction pressure at the
moment the compressor stops (i.e. P2 is
considerably greater than P1 as shown
in Figure 7). The risk is high when there
is a large discharge volume but the
recycle valve is not adequate to
depressure the discharge volume quickly
during coastdown.
Figure 7: Pressures Differential across a Compressor
In a recent project, dynamic
simulation was used to predict speed decay and the pressure differential across the
compressor. The results are shown in Figure 8 below.
Compressor speed

Pressures time
time

Figure 8: Compressor Speed Decay and Discharge-Suction Pressure Differential during Coastdown

In this study, the results showed that when the compressor stopped, the pressure
differential was small (less than 5 psi), so the risk of reverse rotation was low. To further
reduce the risk, an isolation valve was added to the compressor discharge (circled in the
sketch) with a control action to close the isolation valves on trip. Field operation confirmed the
effectiveness of the design. It is worth noting that the results from this type of modeling are
sensitive to the speed decay during coastdown. To ensure the validity of the results, the
predicted coastdown curve should be reviewed by the compressor supplier or validated by field
testing.

Compressor Simulation Summary

These examples show how dynamic simulation can be used to evaluate compressor
operation and protection over a wide range of operating conditions and identify design and
operation issues prior to commissioning. Industry experience shows that in many instances the
use of rigorous dynamic simulations was able to identify and resolve design issues that could
not be fully addressed by just steady state simulation (Ref 5, 6).

Rigorous Furnace Simulation

Ethylene cracking furnaces are highly complex equipment. Efficient operation of
furnaces requires a thorough understanding of the following issues:
a) pass-by-pass control of reactor outlet temperature and coking
b) control of cracking in the radiant coils and transfer lines

102
 c) accurate balancing of pre-heat needs with reaction duty needs
Studying these issues requires a rigorous dynamic simulation model with the following
components:
a) an accurate yield model that can predict the product distribution over a wide range
of operating conditions. In this example, the yield model was compiled in Fortran
and linked to the dynamic model as a .dll file.
b) calculation of pre-cracking in the cross-over
c) accurate thermodynamics for the characterization of a wide variety of feedstock
d) physical dimension of the furnace, including the coils, radiant tubes and fire box
e) two-phase vaporization, heat transfer and pressure drop
f) details of control scheme for feed, firing, coil outlet temperature (COT), excess
oxygen and ID fan
The example below (Figure 9) shows a typical lineup of a naphtha or gas type of
cracker. Typical scenarios that can be studied for such furnaces are transition from cracking to
decoking and then back to cracking, change of firing conditions, change of hydrocarbon/steam
ratio, trip of induced draft (ID) fan and pass-by-pass variations, such as excess coking in one
pass or decoking in one pass.

Figure 9: Gas or Naphtha Cracker Furnace Simplified Schematic

Recently, dynamic simulation was used to evaluate the stability of different furnace
control schemes. Figure 10 below shows the simulation results for one of simulation scenarios
– increase of firing at time 0. As a result of firing change, the coil outlet temperature (COT)
increases momentarily before the returning back to the set point after the control system

103
increases the feedrate. By using these kinds of results, we were able to confirm the stability of
the overall control system and tune the controllers prior to commissioning.
Coil outlet temperature

Feedrate
time time

Figure 10: Representative Simulation Results for Furnace COT Control

Another application of a rigorous furnace model was to study “what-if” cases for basic
design. All major components of the furnace were integrated into a single model:
hydrocarbon/steam in the coils, fluegas in the stack, firing and cracking. This kind of
integrated model can be used to quickly evaluate design alternatives such as changes in the
heat transfer areas of convective bundles, impact of fuel type, impact of changes in various
control set points, etc.

Steam System Simulation

A simplified representative schematic of an ethylene plant steam system is shown in
Figure 11 below for illustrative purposes. The four horizontal headers represent the low-
pressure (at the bottom), medium-pressure, high-pressure and high-high-pressure steam
headers. Ethylene plants are complex heat integrated systems where the energy from the
process (shown in Figure 11 as coming from Plants 1 and 2) is used to generate high-pressure
steam which is then used to power steam turbines, provide process heat to auxiliary systems,
supply the ‘dilution’ steam used in the process, and in some cases directly used for online
‘decoking’ of furnace tubes. The systems are further complicated by having multiple pressure
levels and auxiliary backup boilers in case of plant trips, all of which leads to complex control
system design for the steam letdown stations, the handling of process upsets or trips, and
sparing philosophies.

104
 Figure 11: Representative Steam System

The list of issues that lead to complexities in ethylene plant steam system design can be
summarized as follows:
• Multiple sources of steam – e.g. primary steam generated through heat recovery
from process and auxiliary boilers
• Large let down valves to provide emergency steam to lower pressure headers
• Large multi-stage turbines with extraction and admission steam
• Minimum pressure limits in each of the steam headers
• Very compact design with relatively small header volumes
• Steam and motor drives
Thus dynamic simulation has been frequently used in the industry to study the
following:
• Confirm adequacy of auxiliary boiler ramp up specifications
• Confirm optimum base loading for auxiliary boilers
• Confirm stability of the steam headers and control system with respect to critical
process variations and upsets
• Confirm adequacy of let-down valve sizing and controls
• Verify relief valve placement and size

105
 • Determine preliminary tuning constants for control loops
In cases where a single integrated steam system services multiple plants in a complex,
advanced feedforward controls and de-coupling strategies may also need to be designed and
tested.
The plots in Figure 12 below show examples of the complex dynamics that might result
in steam systems due to plant upsets and trips. The first figure shows the variation of steam
make from Heat Recovery Steam Generator (HRSG) due to a gas turbine trip, while the second
figure shows the ramp up dynamics of an auxiliary boiler due to process trip. As the results
show in the cases below, the control systems were able to manage the upset however it is the
quality of control that is the key criteria for design based on these dynamic results.
Steam generation rate

Controller output
Header pressure
Controller output
(output of auxiliary boiler)

Header pressure

time time

Figure 12: Representative Steam System Dynamics due to Process Upsets

The benefits generated from such studies can be summarized as: a) prevention of
under-sizing or over-sizing of let down and relief valves, b) validation of the sizing and
ramping specifications for auxiliary boilers, c) validation of basic control philosophy and tuning
constants prior to actual plant startup, and d) validation of alarm settings and trip points for
the overall system. The last benefit is critical in cases where steam system upsets may lead to
process trips caused by inadequate setting of process trip points.

Fuel Gas System Simulation

Fuel gas systems are another critical area where dynamic simulation is often used to
determine the adequacy of the system with respect to plant upsets and trips. An illustrative
sketch of a representative fuel gas system is shown in Figure 13 to highlight the practical
complexities of such a system. Note the potential multiplicities of fuel types, different user
levels, and the various backup and failover strategies. Typically the Low Pressure (LP) users
are process boilers, incinerators and such. In this example, the plant also has gas turbines that
operate with High Pressure (HP) fuel gas. The requirement is to maintain header pressures
and fuel gas quality parameters such as Heating Value and Wobbe Index. There may also
be restrictions on rate of change of these fuel gas parameters. Dynamic simulation of
postulated plant upsets is then the only useful way to validate the adequacy of the process
and control system design for the fuel systems.

106
 HP HDR
TIC
001

PIC Fuel User 1

001
FIC
001
Fuel User 2

Main Fuel Gas Supply Fuel User 3

Aux. Fuel Gas

Fuel User 4
Aux. Fuel Gas

FIC
003 FIC
PIC
002
002 LP HDR
Aux.
LPG
LP User 1
LPG

LIC
LP User 2
001
Steam
LP User 3
Vaporizer

Condensate LP User 4

Figure 13: Representative Fuel Gas System

Figure 14 below shows some typical examples of dynamic responses for fuel gas
systems. The first plot shows the pressure recovery and heating value response to a primary
fuel source trip. As the plot shows, pressure recovery (top curve on the first graph) is possible
in a few minutes; however the header may experience a low pressure trip in addition to a
decrease in heating value. The second plot shows the ramp up rate of the secondary fuel
sources following a trip.

50
45
Primary fuel gas
Header pressure

Heating value 40
Heating value

35
Fuel rate

30 Makeup fuel from feed

25
20
15
Header pressure 10 Makeup fuel from LPG
5
0

time time

Figure 14: Representative Fuel Gas System Dynamics due to Process Upsets

Decisions regarding the size of the equipment or specific control strategies that have to
be designed for a specific plant can be derived or validated based on these kinds of responses.

107
In this example, the size of the LPG makeup capacity was increased to minimize the
fluctuations in the fuel gas system on the loss of primary fuel gas.

Flare and Relief System Simulation

Flare and relief systems are another area where dynamic simulation can be beneficially
used for accurate analysis of the peak relief loads and the adequacy of the whole flare header
system. The figures below are illustrative examples of the scope of typical relief header
systems (Figure 15) and the complexity of even a single relieving source such as a tower
(Figure 16).

COOLER

PC
1

LC
3
FEED
TC REFLUX DRUM
1

LC
1
DISTILLATE

COLUMN

TC
1

LC
2

STEAM

REBOILER

BOTTOMS

With current advancements in software and hardware, detailed dynamic models can be
built so that all interactions in the plant, inclusive of controls, can be incorporated in these
studies. Study scope can vary from accurate analysis of relief from single towers, vessels or
reactors to conglomeration of process units to whole complex-wide simulation. Following are
some typical examples of use in our experience:
• Evaluation of relief reduction strategies for the largest relievers in a plant.
• Determination of the extent of relief staggering and its effect of relief load reduction.
• Evaluation of the optimum location of HIIPS.
• Adequacy of existing headers in case of plant debottlenecking, upgrades or revamp.
• Optimum location of a new plant to minimize effect of back pressure of flare loads.
• Sizing of emergency depressuring valves
The plots in Figure 17 show some typical results from past studies. The first plot
highlights the fact that the timing of actual reliefs may not be coincident but staggered in a
real plant – and that, in some cases, the relief incident may only exist for a short duration;
thus the total actual flare load may in fact be significantly less than the simple sum of peak
loads. These kinds of studies are used to optimize flare header sizing and determining the
adequacy of existing header systems in revamps, and are also performed routinely for the
entire complex if the ethylene plant is integrated with a petrochemical plant. The second plot
shows results from an emergency depressuring study that is typically performed for the

108
reactor loops in the chemical processing units integrated with an ethylene plant. These studies
are used to determine the sizing of emergency depressuring valves and ensure that upstream
equipment is adequately sized to avoid choking during depressuring events.

Pressure
Total flare load
Flare load

Pressure
from Unit 1

Release from Unit 2

from Unit 2
Release
from Unit 3 from Unit 1

time time

Figure 17: Representative Plots of Staggered Releases (left) and an Emergency Depressuring Event (right)

Operator Training Simulators

Operator Training Simulators (OTS) are an interesting and high value application of
dynamic simulation. In an OTS, a ‘virtual plant’ environment is created by connecting a whole
plant dynamic model to actual DCS hardware/software loaded with the actual plant control
database and operator displays. Figure 18 below shows a conceptual architecture of a typical
modern OTS system. The main components of an OTS system can be summarized as follows:
• Simulation PC which hosts the dynamic model and the main OTS software platform
• Instructor Console which is used as the primary interface for controlling the training
sessions
• DCS Engineering Station which hosts the DCS database and controls configuration
• DCS Field Controls Emulator (s) which hosts the actual control algorithms for the
specific DCS system
• DCS Operator Console which is the main point of interaction for the trainees, and
• Auxiliary components like Ethernet switch, printer etc.

109
 Figure 18: Typical Architecture of an Operator Training Simulator

The following are some of the key functionalities of OTS systems:

• Ability to initialize the system from any initial state such as normal design or cold
start – this allows many modes of training
• Ability to save and restore any intermediate operating state – this allows training to
occur over long durations without losing data and also to capture interesting case
studies for future review
• Ability to cause standard or generic equipment faults – this allows the operators to
train on wide variety of faults and upsets
• Ability to run pre-programmed scenarios – this allows operators to repeatedly train
on specific scenarios to gradually build up expertise
• Ability to track operator performance – this provides Instructors with an objective
measure of operator proficiency both as a way to gauge skills progress and as a
method for operator certification for specific job classifications.
The primary uses of these systems can be summarized as follows:
• comprehensive training of plant operators – while these systems are primarily used
for new operator training, they are also used for refresher training and for operator
certification
• validation and optimization of plant operating procedures, and
• control system refinement.
Figure 19 below shows a representative example of an OTS display for an ethylene
furnace to illustrate the degree of detail and rigor that is typically included. In general the

110
models consist of a whole plant dynamic model inclusive of all basic controls, all field
operations, all panel operations and relevant safety logic.

Figure 19: Typical Furnace Detail Display in an OTS

Figure 20 below shows the layout of a complete OTS system. The computer on the left
is one of the PC’s for the Simulation Components. The three sets of screens on the right are
the Operator Consoles.

111
 Figure 20: Actual Layout of an OTS

OTS systems are typically built at the later stages of grassroots projects or even after a
plant has been operated for a while. Increasingly, these systems are being developed in
parallel with detailed engineering to completely validate the process, the controls, the safety
systems design and the plant operating procedure prior to actual startup. There are also
increasing requests for OTS from plants that have operated stably for long periods, say 20-30
years; in these cases the main issue is the loss of highly experienced operators and the need
to quickly certify new operators for board operation.
Recently, high fidelity dynamic models have been used in troubleshooting field
problems. In this sense, OTS systems and dynamic models are forming the core component of
what can be called ‘life cycle modeling’ as illustrated in Figure 21.

112
 Figure 21: Life-Cycle Modeling Concept

CFD
Computational fluid dynamics (CFD) is a numerical tool used to simulate fluid flow,
thermal dynamics, and heat transfer related problems. Pertinent to ethylene plant, CFD has
been used to study burner flame distortion, coil layout, fuel gas mixing drum internals and
storage tank nozzle arrangement (Ref 7). The graphs in Figure 22 are based on recent studies.
The left figure shows the flame distortion inside the fire box as predicted by the CFD model.
The right figure shows how different type of gas (in different color) is mixed inside a fuel gas
mixing drum. Based on these simulations, decisions can be made to modify the design of
equipment to achieve optimum performance.

113
 Figure 22: Examples of Use of CFD for Design Validation

Conclusions and Recommendations

Dynamic simulation has become a powerful tool for leveraging simulation technology to
assist the design and operation of ethylene plants. With collaborations with equipment and
control system suppliers, high-fidelity dynamic simulation models can be developed as a value-
added tool, even at early phases of a project. As demonstrated in the case studies, the
dynamic model allows for studying the equipment performance under the full range of
operation and eventually improving plant efficiency, availability and protection.

Acknowledgements

The authors wish to express sincere thanks to Greg Berry, Dick Churns, Javier Vazquez
and Deyuan Wu, of KBR, for their technical insights, as well as many individuals within the
company whose contributions in various dynamic simulation projects made this paper possible.

References

1. Cassata, J. R., Dasgupta, S., and Gandhi, S. L., 1993, “Modeling of Tower Relief
Dynamics”, Hydrocarbon Processing, October, pp. 71-76.

2. Dukle, N. and Narayanan, K., “Validating Anti-Surge Control Systems,” Petroleum

Technology Quarterly, Summer 2003.

3. Staroselsky, N. and Ladin, L., “More Effective Control for Centrifugal Gas Compressors
Operating in Parallel,” Presentation at the International Gas Turbine Conference and Exhibit,
Germany, 1986.

114
4. Wilson, J. and Sheldon, A., 2006, “Matching Antisurge Control Valve Performance with
Integrated Turbomachinery Control System,” Hydrocarbon Processing, August, pp. 55-58.

5. Patel, V., Feng, J., Dasgupta, S., Ramdoss, P., and Wu, J., 2007, “Application of
Dynamic Simulation in the Design, Operation and Troubleshooting of Compressor Systems,”
Presented at the 13th Turbomachinery Symposium.

6. Wu, J., Feng, J., Dasgupta, S., and Keith, I., 2007, “A realistic dynamic modeling
approach to support LNG Plant Compressor Operation,” LNG Journal, October, pp. 27-30.

7. Barnett, D. and Deyuan, W., "Fluegas Circulation and Heat Distribution in Large Scale
Down-fired Reformer Furnaces,” Presentation at the AIChE, September 2000.

115