Proceedings of ASME Turbo Expo 2012

GT2012
June 11-15, 2012, Copenhagen, Denmark

GT2012-70072

DYNAMIC PIPELINE SYSTEM SIMULATION OF MULTI-STAGE COMPRESSOR

TRAINS

Augusto Garcia-Hernandez Theodore Sean Tavares Melissa Wilcox

Southwest Research Institute® Southwest Research Institute® Southwest Research Institute®
San Antonio, Texas, USA San Antonio, Texas, USA San Antonio, Texas, USA
Phone: 210-522-5693 Phone: 210-522-6544 Phone: 210-522-6046
Email: agarciahernandez@swri.org Email: t.tavares@swri.org Email: melissa.wilcox@swri.org

ABSTRACT The results of this analysis provide recommendations to

An oil and gas company was facing process and
mechanical related problems on the multiple-stage compressor
trains at two important booster installations. The frequency of
these problems has increased lately, and this has led to frequent
trips and shut downs. These interruptions affect the operation of
the plant leading to a loss in production and consequences of
lost revenue for the company. The two platforms each contain
one compressor train comprising a four-stage compressor with
a gas turbine driver. Each train is fitted with an integrated
turbine compressor control panel.
Thus, a detailed dynamic pipeline system simulation of the
subject compressor trains was performed in order to provide a
series of recommendations that would improve the safe
operation and increase the reliability of the compression
systems. The analysis included a review of the existing
compression systems including all the equipment and hardware
related with the compression anti-surge system. In addition, a
site visit was performed to review and understand the existing
anti-sure control system at each facility. A detailed dynamic
model of the multi-stage compression system was built for each
train. These models included compressor performance maps,
gas compositions for each stage and train, piping yard, recycle,
isolation, check and blowdown valves, scrubbers, separators,
and coolers.
Several simulation cases were conducted for both the
platform systems. These cases evaluated the effect of the delay
and travel times of the existing anti-surge valves, delay the
coast down action, failure of the non-return valves (NRVs),
action of a blowdown valve on the emergency shutdown (ESD)
sequences, recycle valve bypasses, check valve arrays, and
process upset conditions. In addition, parametric studies were
conducted for each of the most important parameters of the
system to quantify their effect of any possible modification.
solve some of the existing issues while understanding more of
the dynamics of the system. It was found that any propose
recommendation or change in the sequence or timing of one
stage will affect the surrounding stages since they are not only
connected through the piping as they are driven by the same gas
turbine shaft. Therefore, a very comprehensive analysis was
conducted for each train to provide recommendations that
would be feasible for implementation while reducing the
constant risk of mechanical failure and surge events. Thus,
results of the analysis and some of the recommendations
obtained are presented in this paper.
INTRODUCTION
Multi-stage compressors are being used very frequently
due to their high performance characteristics. Currently, a vast
variety of applications are present in the industry for these
machines. In many cases, the different stages are located in one
solid body structure, while in other applications the multiple
stages are separated in various bodies or barrels. Another key
feature of the multi-stage units is the driver and connection
used to power them which can vary from gas turbines to
internal combustion engines to electric motors. Moreover, the
flexibility to have one driver that can power all the stages
provides some critical space-power-cost advantages. In any
system, inter-stage cooling is required to improve the process
efficiency and maintain the operating condition within
acceptable limits; thus, this typically is accomplished with
inline or back-to-back coolers.
For the case presented in this paper, a four-stage
centrifugal compressor train barrel type is driven by a gas
turbine; thus, all the stages are connected through a common
shaft. Mechanical couplings are installed between the GT driver
and the multi-stage compressor. Each stage has its own
dedicated anti-surge loop and cooling system. The trains are fed
with gas from different production fields and the design flow

1 Copyright © 2012 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/conferences/asmep/74824/ on 03/11/2017 Terms of Use: http://www.asme.org/abo

 and compression ratios are different for each stage. Moreover, NOMENCLATURE
some stages are fed through more than one stream at a time. Non-return valve NRV
The compressor trains have low pressure (LP), medium Emergency shutdown ESD
pressure (MP) and high pressure (HP) stages and each one of Low pressure LP
them includes a suction scrubber, cooler, recycle loop, and Medium pressure MP
blowdown line. In addition, each train is fitted with an High pressure HP
integrated turbine compressor control panel. Gas turbine GT
Surge Control Line SCL
Those two compressor trains are used to boost the pressure
Boost Line BL
of natural gas coming from off-shore production fields to an on-
Actual inlet flow (am3/hr) ܳ௔
land processing facility. Thus, they are key components of the
Surge flow (m3/hr) ܳ௦
production process; moreover, any unexpected shutdown or
Compressor speed (rpm) N
failure can originate upset conditions that will affect the
Surge flow at operating head ܳ௦ு
production process and the performance of the producing wells.
Compressor head H
In the past, those units have presented different types of
Flow coefficient Cv
mechanical and control failures. Therefore, a detailed analysis
Power (kW) P
of the system was required to mitigate the process and control
Inertia (kg-m2) J
issues. Thus, the two multi-stage compressor trains were
Speed (rpm) N
dynamically analyzed to determine if their anti-surge logic and
Time (s) t
controls were operating properly. In addition, it was very
Compressor Torque (N-m) T
critical to understand and deeply evaluate the interaction of the
Surge Margin SM
different stages with their surrounding stages and components,
Operating Point OP
since minor changes in one stage will affect the transient
Turndown TD
behavior of the neighboring stages.
Anti-Surge Valve ASV
The analysis of each compression train presented in this
paper was carried out using commercially available pipeline/ COMPUTATIONAL MODEL AND METHODOLOGY
process simulation software [1] coupled with detailed models The quantitative analysis of each compression train is
for the system components tailored for the particular based on fluid simulation utilizing pipeline/process simulation
application by the authors. software. The software used is able to accurately model rapid
trips of centrifugal compressors. Moreover, this software has
The techniques used in this work are based upon a large
been validated against real data and it has been used to model
body of previous work devoted to understanding and modeling
very complex and large oil/gas systems. The Benedict-Webb-
the complex dynamic interaction between compressor
Rubin-Starling (BWRS) equation of state was used for the fluid
equipment and associated piping. The state of the art in this
properties calculations considering the actual gases
area is well represented in various publications [2, 3, 4]. These
compositions that feed the system. Two separate computational
references also cite an extensive body of work containing both
models were built for the compression systems. Those models
the conceptual ideas upon which these techniques are based,
accounted for the operation of recycle valves, their actuators,
their numerical implementation, and their experimental
vent valves, compressor coast down speed, and control system
validation using both laboratory and field data. Two of the
responses, while including the effects of upstream and
earlier underlying works include that of Wylie et al. [5]
downstream piping, scrubbers, after-coolers, series compressors
involving use of the method of characteristics in numerical
(multiple stages), and other features of the compressor
calculations of transient flow in pipelines and that of Sparks [6]
installation.
describing transient behavior of centrifugal compressors
coupled to piping systems. The computational model solves non-linear systems of
differential equations at each time step using the method of
The focus of the current study was to identify changes to
characteristics [1]. Transient analysis consists of a linearized
piping and valves that would avoid damaging surge events
solution of partial differential equations. In addition, an
during emergency shutdown (ESD) events. Therefore, the
optimum time step for the computations is automatically
criterion was to keep the operating point of the compressor
determined by the software without sacrificing accuracy [1].
away from the surge line at least until speed had bled off to
For the transient cases simulated in this study, the maximum
where surge energy would be low enough to avoid equipment
time step taken was one millisecond. This value was found to
damage. For this reason, there was no need to model the
be more than sufficient to accurately capture all the transient
compressor at flows below the defined surge line.
behavior of the system.
Configurations for which the surge line was reached during
ESD, even in a transient manner, were rejected as valid design The process of creating a system model consists of
candidates. reviewing available data for the system, creating a functional
process block diagram, developing a system model in the

2 Copyright © 2012 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/conferences/asmep/74824/ on 03/11/2017 Terms of Use: http://www.asme.org/abo

 pipeline software, running the simulation based on given initial simulation of normal and off-design operating conditions. The
and boundary operating conditions, validating the model, and following boundary conditions were utilized; a flow boundary
post-processing and interpreting the results of the condition was taken at the suction header while a pressure
computations. These computational models allow for the boundary condition was considered at the discharge header. The
evaluation and comparison of various conditions for a given inlet flow boundary is adjusted by the flow control valves
system, and also for the variation of characteristics of system located in the inlet multiphase separators that are part of the
components such as valve responses and compressor operating upstream process. In addition, inlet and outlet flow boundary
conditions. Thus, parametric studies of the main control conditions were incorporated, including different separators,
variable and elements such as anti-surge valve opening, signal inputs, and outputs. All flow resistance elements were
delays, and ESD sequences were conducted to determine in considered to ensure the modeled system impedance was as
which direction the possible system changes should be going accurate as possible. The model for each compressor was
and how they will affect the entire train and compression validated against steady-state data provided. To validate the
system [7, 8]. Figure 1 presents a typical diagram of the pipeline models of each compressor, steady-state flows,
compression trains and its main components. temperatures, and pressures collected from each compressor, a
MULTISTAGE CENTRIFUGAL COMPRESSOR TRAIN
specified configuration was compared to that same
Anti-Surge Valve
configuration replicated in the simulation. For the validation to
Out Flow Discharge
Cooler Header be complete, several different operating points needed to be
evaluated and data collected for each unique operating
Suction
Header
Check
Valve
condition. A baseline operating condition was used to tune the
Scrubber
Flare hydraulic model and to account for losses, which are not
accounted for directly in the model. The simulation was run at
other operating points and compared to the reference data. This
Gas Turbine
Low Pressure
Stage (LP)
Low Pressure
Stage (LP)
Medium Pressure
Stage (MP)
High Pressure
Stage (HP)
analysis provides quantitative results as to the predictive
Driver
capability of the developed models.
Figure 1. Multi-stage Compressor Train Diagram Table 1. Comparison between the Model Predictions and the
Provided Data
A large amount of detailed data was required in order to
develop an accurate and complete model of the centrifugal
STAGE  I ‐LP ‐ Point #2
compressors and their controls. The complete piping system
PARAMETERS / COMPRESSOR 
with lengths, diameters, and all branches or connections were STONER ‐  Relative 
defined and input into the model builder. Details such as STAGE Reported 
Calculated  Difference 
volumes, lengths, and heat transfer surface areas were provided Values ‐ LP
Values (%)
for coolers, heaters, scrubbers, or filters. The rotational inertia Inlet Conditions
of the compressor trains and their drivers, along with the torque Pressure (bara) 1.02 1.02 0.000
characteristics of the drivers were needed so the rate of coast Temperature (ºC) 42.78 42.78 0.000
down or start up could be simulated. In this case, the train Molecular Weight (kg/kmol) 45.21 45.21 0.000
inertia was back calculated using a coast down data provided Specific Gravity (‐) 1.561 1.561 0.008
from unit trip events. Thus, the rate of deceleration is Compressibility (Z1) 0.985 0.985 0.000
determined by the inertia of the train J (referenced to Inlet Actual Volume (m3/h) 5381 5466 1.572
Standard Flow (SCMH) 5025.4 5004.0 0.426
compressor speed (N) for geared trains), any residual power
Density (kg/m3) 1.782 1.778 0.245
generated by the gas turbine (PGT,resid), the friction losses in Mass Flow Rate (kg/s) 2.66 2.70 1.323
the system, and the power absorption of the compressor [3].

dN Discharge Conditions
P  Pcompr  Pfrict  PGT, resid  T  N  J  N  (1) Pressure (bara) 4.453 4.453 0.000
dt Temperature (ºC) 112.7 110.587 1.875
Compressibility (Z2) 0.967 0.967 0.000
Solving Eq. 1 for steady-state power and speed at the
initial operating point and assuming a constant Q/N Eq. 1 yields Reported Compressor Operating Conditions
the following expression for the speed decay: Polytropic Head (KJ/Kg) 92.8 92.8 0.045
Polytropic Efficiency (%) 73 72.7 0.411
3 Speed (RPM) 10739.1 10739.1 0.000
J No
N (t )  (2) Power (Kw) 338.7 344.3 1.662
J N o  Po t
2
Average Relative Difference ‐ All Parameters (%) 0.562
The normal and upset condition pressures, temperatures,
flows, and gas compositions were included. A complete
compressor map was prepared and input to the model for the

3 Copyright © 2012 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/conferences/asmep/74824/ on 03/11/2017 Terms of Use: http://www.asme.org/abo

 The computed percentage differences between the supplied
data and quantities computed by the models are seen to be quite
small and serve to confirm the accuracy with which the system
behavior was modeled. The differences between the developed
model and the data are between 0.9% and 1.8% for both
systems. For each point, a relative difference was calculated for
each parameter provided, as shown in Table 1, while the total
uncertainty of an operating point was defined as the average
summation of all the parameters’ uncertainty, Eq. 3. Thus, the
model uncertainty was computed as the average of multiple
operating points’ uncertainties. An example of the comparison
between the model and the data is presented in Table 1. This
process was repeated for at least five different operating points
at steady-state conditions, to determine the average total
difference or computational uncertainty for each model.
n U flow  U pressure  Uhead  Uspeed  Utemp  ........Un
Up  
1 n
DYNAMIC SIMULATION ANALYSIS
Simulations of the compressor were focused on various
operational conditions identified to be of interest for the
performance improvement study. Transient sequences were
initiated from design, minimum, and other normal flow
conditions, since they represented the most critical scenarios for
a compressor shutdown. Unit emergency shutdown (ESD)
sequences were simulated to determine if the units go through
surge and also how stable they were while they were
undergoing coast down. In addition, anti-surge control logic
function and parameters were programmed into the model
logic. Thus, typical process upset conditions such as sudden
loss of a separator or flow and pressure changes were
simulated. A very critical situation occurs when centrifugal
compressors go through surge, which is the flow reversal within
the compressor that originates high fluctuating loads and
vibrations that could damage the compressor [9, 10, and 11]. So
the main objective of the simulations is to find out how the
compressors behave during critical upset conditions such as
ESD. In addition, it is very important to verify that the
compressors’ surge margins during normal operation are
sufficient to avoid any surge phenomenon. However, if the
surge margin reaches a stipulated value of 10%, it is expected
that the anti-surge control opens the recycle valve; thus, the
flow through the compressor increases and the pressure ratio
decreases.
Several emergency shutdown simulations were completed
on the model of the multi-stage compressor and piping systems.
These were simulated at the provided operating conditions. The
same starting operating conditions were used for all
simulations. The simulations investigated the sensitivity of
various parameters in the surge control systems, such as recycle
valve opening times, recycle valve sizes, delay of opening of
recycle valves, and the effects of the opening time of the
blowdown valve. Thus, the anti-surge system control
parameters were built into the simulation as well as the ESD
Downloaded From: http://proceedings.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/conferences/asmep/74824/ on 03/11/2017 Terms of Use: http://www.asme.org/abo
sequences and coast down speed of the compressor. The values
of control parameters such as Surge Control Line (SCL) and
Boost Line (BL) were based on the existing compressor
configuration.
The different analyses were classified based on parametric
studies carried out for different variables such as ASV opening
time, signal delay time, and flow coefficients. The general
concept is to identify areas where small changes in the existing
system could drive the compressor to a safer operation while
avoiding surge events. For each simulation, run control
parameters (such as Surge Margin (SM, Eq. 4), Operating Point
(OP, Eq. 5), and Turndown (TD, Eq. 6) were monitored and
used to quantify the benefits of the evaluated changes.
ொೌ ିொೞ
ܵ‫ ܯ‬ൌ │ேୀ௖௢௡௦௧ (4)
(3) ொೌ
ொೌ
ܱܲ ൌ │ேୀ௖௢௡௦௧ (5)
ொೞ
ொೌ ିொೄಹ
ܶ‫ ܦ‬ൌ │ுୀ௖௢௡௦௧ (6)
ொೞ
Different levels of ESD events were simulated for both of
the compressor train systems since different valves delay times
(to begin opening and closing) existed for the various types of
ESDs. The main differences between the ESD sequences were
in the delay to open a specific recycle valve in the entire train
and the depressurization of a particular stage or all stages. In
addition, four other variations in the ESD sequence were
considered in order to find solutions to the surging of the
compression system during an ESD event: 1) Varying the valve
delays and opening times, 2) delaying the coast down of the
compressors, 3) adding an additional line and valve on the
second and third stage recycle lines (referred to as hot bypass),
and 4) combining the coast down delay and hot bypass on the
second stage.
CASE RESULTS
Since it was very critical to evaluate, understand, and
quantify the existing system’s issues and their effect on the
process, a series of simulation analyses were conducted first for
the existing configuration. In addition, it was vital to
understand how the different stages and their diverse
components were acting as a whole system, since any minor
change in operation of one stage anti-surge valve (such as
opening time, or relocation of the check valve) would affect the
behavior of the entire train. The system dynamics was very
sensitive to changes; for example a minor modification of
approximately 5% in the flow coefficient of the third stage
(MP) anti-surge valve would cause the fourth stage (HP) to
surge during an ESD. Similarly, increasing the flow coefficient
of the anti-surge valve would allow more flow through its
recycle line during the initial few milliseconds of opening,
which would cause a sudden decrease in the suction flow of the
4 Copyright © 2012 by ASME
 last stage (HP) until the actuation of the check valve; thus, the Surge Margin for each Compressor Stage during ESD Event
60
rate of head change versus flow (dH/dQ) of the HP stage would
decrease drastically making the compressor move much faster
toward the surge line. In addition, any change in the impedance 50

of the system will affect the initial reaction of any stage, since
the pressure waves that originate when the machine is 40

Surge Margin (%)
shutdown travel in the system based on its impedance. Thus,
the system’s impedance can affect the initial dH/dQ behavior of 30
the compressor until the system flow resistance starts
dominating the movement of the gas flow [2, 12]. 20

Existing System
SM_S1:VAL
10
An ESD event was simulated for the existing configuration SM_S2:VAL
SM_S3:VAL
in which the compressor started to shutdown (speed ramped SM_S4:VAL
0
down) at 0 seconds. The valves opened and closed according to 0 0.25 0.5 0.75 1 1.25 1.5 1.75 2 2.25 2.5 2.75 3 3.25 3.5
their specified timing. The anti-surge valve opening events are Time (s)
shown in Figure 2. The surging of the compressor occurs within
the first 1 to 2 seconds after shutdown, and the blow down and Figure 3. Surge Margin Values during an ESD Event for the
Existing Configuration
isolation valves do not start to actuate until after this time.
Thus, it was found that the first few milliseconds were very Figure 3 shows the surge margin of the four stages of the
decisive for the existing anti-surge configuration. compressor during the ESD. In this figure, it is clear that both
the second and third stages surged. The ESD is shown to have
Anti‐Surge Valves Opening versus Time during an ESD
little effect on the operating point of the first stage. This is most
110
likely due to two factors: 1) the recycle valve of the first stage
100
was already opened to a fraction of 31% at the beginning of the
90 ESD event, and 2) the first stage recycle valve is the first valve
80 that opens during an ESD event and has the quickest opening
time. All of these factors reduce the effect of the compressor
Valve Opening (%)

70

60
shutdown on the surging of the first stage. The first stage had a
minimum surge margin of 18.4% during the transient
50
simulation, which is considered acceptable.
40

30
The second and third compressor stages experienced a
surge event during the simulated ESD for the existing anti-
20 ASV‐Stage 1
ASV‐Stage 2
surge configuration. Those two stages presented the slowest
10 ASV‐Stage 3 opening time and the longest signal delay. Thus, they did not
ASV‐Stage 4
0
0 0.25 0.5 0.75 1 1.25 1.5 1.75 2 2.25 2.5 2.75 3 3.25 3.5
have sufficient flow to recover from the rapid speed reduction
Time (s) that originated during the ESD. Moreover, as soon as the fourth
stage started drawing more flow it affected the dH/dQ behavior
Figure 2. Anti-Surge Valves Opening during an ESD Event for the of the second and third stages, since the dQ component
Existing Configuration
decreased rapidly. Thus, the compressors move to the surge
condition quickly.
Contrary to the other stages, the flow and head in the
second and third stages reduced rapidly bringing the
compressors to a surge condition. It was observed that the anti-
surge valves of both units were not able to divert sufficient flow
to the suction of the compressor stage during the transient.
Moreover, the second stage valve has the slowest travel time
and the third stage valve has the longest delay time. Thus, those
conditions affect the behavior of the system, since the high
pressure energy is maintained in the discharge of both units for
a longer period of time during the coast-down of the unit.
Therefore, the lack of enough flow in the suction side and the
high energy present at the discharge (head) during the transient
event cause a harsh surge condition in both compressor stages,
which are not acceptable even for a small period of time.

5 Copyright © 2012 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/conferences/asmep/74824/ on 03/11/2017 Terms of Use: http://www.asme.org/abo

 During a simulated rapid shutdown, the speed of the opening times on the second and third stages to avoid surge; all
compressor, the head across the compressor, and the flow stages avoided surge with the proposed timing. The surge
through the compressor decreased. If the surge line is crossed at avoidance on the second and third stages was marginal. A
a relatively low head, the surge or flow upset at the compressor second modification in the anti-surge logic introduced a coast-
will be mild and the surge control will be considered down delay of the compressor unit; thus, all stages avoided
successful. On the other hand, if the surge line is crossed at a surge when the compressor shutdown was delayed by
relatively high head, then the surge or fluid driven forces acting 1.15 seconds. The surge avoidance on the second stage was
on the compressor will be significant, and the surge avoidance marginal. Another option evaluated was to implement hot
will not be considered satisfactory. bypasses in some of the stages. In this option, the original valve
timing was used and additional recycle valves were added in
Therefore, the current anti-surge design was not sufficient
parallel to the recycle valves on the second and third stages
to protect the compressor train when it experienced an ESD
(referred to as hot bypass valves). Several cases were presented
event from a normal operating condition. Thus, a new system
where surge was avoided. However, in the majority of these
design was required.
cases the surge avoidance was considered marginal. One case
Parametric Study and Modification of ESD Logic was presented where the surge margin was considered
acceptable. This included the use of a 6 and 3-inch “quick”
In order to propose feasible solutions or modifications of
opening valve on the second and third stages, respectively.
the anti-surge system, several variables and scenarios were
analyzed. The first step in the analysis was to conduct a A final option blended two of the previously evaluated
parametric study of the opening and delay times of the anti- options in one; a compressor coast-down delay and hot bypass.
surge valves, delay of the gas turbine fuel cut-off signal, and This sequence took advantage of both the coast down delay and
timing of the blowdown valves. Figure 4 shows typical a hot bypass valve on the second stage. Surge was avoided with
parametric study results. In this case, the second stage anti- a coast down delay of 0.5 seconds and a second stage hot
surge valve opening time was varied up to 200% (3 seconds) bypass valve with a flow coefficient (Cv) equivalent to a 4-inch
from its current value and then various ESD simulations were valve. The surge with this delay and valve was considered
conducted. The results of this parametric study indicated that a marginal. The surge avoidance was found to be better with a
slower stroke second stage valve, combined with a coast down coast down delay of 0.75 seconds and a second stage hot
delay of 100 milliseconds, would reduce the surge margin of bypass valve of 4-inch as presented in Figure 5. Typical inlet
the second stage by almost 50% during ESD, while slightly and outlet temperature and pressure conditions used as the
increasing the surge margin of the third and first stages by initial steady-state for the transient events are presented in
approximately 2% and 1%, respectively. Table 1 of the appendix section.
Second Stage Anti‐Surge Valve Opening Time Parametric Study ‐ Surge Margin for each Compressor Stage during ESD Event with 
Compressor Shutdown Delay of 100 Milliseconds the Coast‐Down Delay and Hot‐Bypass
25.0 60
Minimum  Surge Margin Reached during an ESD (%)

22.5
50
20.0

17.5
40
Surge Margin (%)

15.0

12.5 30

10.0
20
7.5

5.0 10 SM_S1:VAL
LP‐S1 SM_S2:VAL
LP‐S2 SM_S3:VAL
2.5
MP‐S3 SM_S4:VAL
HP‐S4 0
0.0 0 0.5 1 1.5 2 2.5 3 3.5 4
1.50 1.75 2.00 2.25 2.50 2.75 3.00 3.25 Time (s)
2nd Stage Anti‐Surge Valve Opening  Time (s)

Figure 4. Parametric Study for the Second Stage Anti-Surge Valve Figure 5. Coastdown Delay and Hot Bypass Option Results
Opening Time including a Coastdown delay of 100 Milliseconds
Results from the different parametric studies indicated the
most critical scenarios to be analyzed for the entire train, as
well as the best direction for any proposed change or
modification. Therefore, modified sequences and alternatives
were simulated and analyzed for the compressor trains. The
first modified sequence involved changing valve delay and

6 Copyright © 2012 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/conferences/asmep/74824/ on 03/11/2017 Terms of Use: http://www.asme.org/abo

 Compressor Train 2nd Stage Map with Different Sequences  Blowdown and Check Valves Effect
90
Several simulations were completed in order to determine
80
the effect of the main blowdown valve opening time on the
70 surge margin of the compressor stages. The first alternate case
was used to determine the effect of a short valve delay on the
Polytropic Head (kJ/kg)

60
surge margins of each compressor stage. The second alternate
50 case investigated the effects of a short valve opening time on
40
the surging of the compressor stages. In all cases simulated, the
ESD sequence with a 0.75 second coast down delay and
30
inclusion of a hot bypass valve on the second stage was used.
Existing ESD Sequence
20
Modified Valve Timing The results of these simulations indicate that the first and
Coastdown Delay
10 Hot Bypass
second stages were not affected by the change of timing on the
Coastdown + Hot Bypass main blowdown valve. This was expected since the valve is
0
4000 6000 8000 10000 12000 14000 16000 18000 20000 22000 located on the blow down line at the suction of the fourth stage
Inlet Flow (am3/hr) compressor. The third stage was slightly affected by the delay
Figure 6. 2nd Stage Compressor Map with Decel Paths for
opening time of the valve. When the valve had a 0.5 second
Different Sequences delay, the surge characteristics of the third stage compressor
slightly improved. The minimum surge margin point increased
Figure 6 and Figure 7 show each of the five sequences to 17.2% from 15.4%. The third stage was not affected by the
discussed above on the compressor maps for the second and short opening time of the valve simulated in the second case.
third stages. Based on the results of these studies, it is The minimum operating point on the fourth stage slightly
recommended that either the hot bypass sequence (use of 6 and decreased when the main blowdown valve opened with a 0.5
3-inch quick opening valves with 1.2 and 1.5 seconds on second delay. However, the decrease experienced (4.3%) was
second and third stages, respectively) or a combination of a not considered significant since the fourth stage has an
coast down delay and hot bypass on the second stage adequate surge margin. The fourth stage was not affected by the
compressor be implemented. In addition, it was observed that short opening time of the valve as simulated in the second case.
the implementation of the hot bypass will affect the neighbor
stages, since it will produce a quick drop of the discharge Additional scenarios were also analyzed for critical process
pressure of one stage as well as the suction pressure of the upset conditions and failure of check and isolation valves. The
following stage as observed in Figure 6 where the coast down existing anti-surge system responded well to mild process
delay and the hot bypass combined effect is in the middle. In upsets, such as lower suction pressure; however, for critical
this case, the hot bypass option of the 2nd stage is diminishing upset conditions, such as a sudden drop of the inlet flow due to
the positive effect of the coast down delay. a loss of a process separator, its response was very marginal
due to the slow response of the second and third stages recycle
Compressor Train 3rd Stage Map with Different Sequences  valves. Thus, this confirmed again that the anti-surge logic or
140
sequence for those stages should be adjusted to provide
120
sufficient protection for the entire train.
The main objective of the check valves in the gas system is
100
to stop reverse flow through the compressor stages. A high
Polytropic Head (kJ/kg)

amount of energy is accumulated in the discharge of each

80
compressor in form of pressure – volume relationships [13].
60
Thus, if this energy is not controlled or diverged in the
appropriate manner, it will affect the behavior of the
40 compressor during an ESD or process control situation.
Existing ESD Sequence
Modified Valve Timing
Moreover, when an ESD sequence is activated the flow
20 Coastdown Delay distribution within the system is very critical, since the gas will
Hot Bypass
Coastdown + Hot Bypass
try to move to a lower energy condition, meaning the suction
0
2000 3000 4000 5000 6000 7000 8000 9000 10000 11000
side of the compressor when the ASV opens. Thus, in order to
Inlet Flow (am3/hr) avoid an excessive amount of flow from an upcoming stage,
non-return valves are used to check the mass transfer between
Figure 7. 3rd Stage Compressor Map with Decel Paths for Different the stages. Therefore, the flow recirculation remains local and
Sequences is sufficient to reduce the energy level in each stage to a safe
condition.

7 Copyright © 2012 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/conferences/asmep/74824/ on 03/11/2017 Terms of Use: http://www.asme.org/abo

 A faster reduction of the head through the compressor will
cause the inter-stage check valves to close, isolating each stage
from the surrounding stages. Moreover, an inter-stage flow
distribution takes place when the compressor is shutdown or
unloaded, affecting the amount of flow that is available for
recirculation within stages, since one stage could be taking part
of the flow that is required by the other stage to avoid a surge
condition. Therefore, the appropriate closure of the check
valves is vital in an ESD event to avoid a surge condition.
Surge Margin for each Compressor Stage during a Failure of the  should not be applied, since they could increase the risk of
Interstage Check Valves
60
50
40
Surge Margin (%)
30
20
10
0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5
Time (s)
Time (sec)
Figure 8. Surge Margin during an ESD with a Failure of the
Interstage Check Valves.
Thus, a failure of the check valves during an ESD event
with a 0.75 second coast down delay and second stage hot
bypass valve sequence was simulated to determine how this
would affect the protection of the entire train system. Figure 8
shows the surge margin of the compressor stages with time.
During the simulation three compressor stages surged: second,
third, and fourth. The second and third stages surged first. In all
other simulations discussed above, the fourth stage has
maintained a very high surge margin and has not surged. In the
simulation with all check valves remaining open, the fourth
stage compressor surged after approximately 6.5 seconds. The
first stage of the compressor did not surge, but the surge margin
was affected by the check valves staying open. The minimum
surge margin for the first stage was 13.4%. This minimum
surge margin is 5.75% below the minimum surge margin in the
case where all check valves were operating correctly.
Therefore, it was proved that the action and placement of the
check valves play a critical role in the anti-surge control system
and help to prevent critical conditions during an ESD.
SUMMARY AND CONCLUSIONS
A dynamic pipeline system simulation of a multi-stage
compressor train is a complex task that should include both
very detailed system data and a good analysis methodology.
The dynamics of the multi-stage system is affected by the
interaction of each stage and its components, since they are
critically linked and related. This paper presented and
summarized a methodology for analyzing a multi-stage system
Downloaded From: http://proceedings.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/conferences/asmep/74824/ on 03/11/2017 Terms of Use: http://www.asme.org/abo
through parametric studies of the main variables and
components of the anti-surge system and its logic. In addition,
the case study results illustrated typical steps that should be
performed to properly analyze a multi-stage compressor train.
It was found that traditional solutions and changes from a
single compression stage can be extrapolated to multi-stage
systems. However, special considerations and analysis should
be conducted to determine the correct level of application for
those changes. Thus, typical rules of thumb or quick fixes
unintentionally introducing adverse system behavior due to the
SM_S1:VAL
SM_S2:VAL
complex nature of the multi-stage compression system.
SM_S3:VAL
SM_S4:VAL Parametric studies of the main components and their
interaction with the system should include installation of hot
bypasses; placement or relocation of the check valves; anti-
surge valve timing (stroke rate); different type and size of
recycle valves; and shutdown delays.
The dynamic interaction between the stages and their anti-
surge loop can be modified by properly adjusting the anti-surge
logic and sequences. Moreover, the initial dH/dQ (pressure
wave) behavior of each compressor stage is affected by the
changes made in the entire train. The system dynamics is very
5 5.5 6 6.5 sensitive to changes. In addition, any change in the impedance
of the system will affect the initial reaction of the pressure
waves originated when the machine is shutdown.
The case study presented a methodology to dynamically
analyze multi-stage compressor trains and to determine if their
anti-surge logic and controls operate properly. In addition, it is
very critical to understand and thoroughly evaluate the
interaction of the different stages with their surrounding stages
and components since minor changes in one stage will affect
the transient behavior of the neighboring stages.
REFERENCES
[1] Advantica, Inc. “Stoner Pipeline Simulation (SPS)
Reference Manual and Software”
[2] Botros, K. K., “Single Versus Dual Recycle System
Dynamics of High Pressure Ratio, Low Intertia Centrifugal
Compressor Stations, ASME Journal of Engineering for
Gas Turbines and Power, 133, 122402, 2011.
[3] Brun, K., and Kurz, R., “ Analysis of the Effects of
Pulsations on the Operational Stability of Centrifugal
Compressors in Mixed Reciprocating and Centrifugal
Compressor Stations, ASME Journal of Engineering for
Gas Turbines and Power, 132, 072402, 2010.
[4] Morini, M., Pinelli, M., and Venturini, M., “Development
of a One-Dimensional Modular Dynamic Model for the
Simulation of Surge in Compression Systems”, ASME
Journal of Turbomachinery, 129, pp. 437-447, 2007.
[5] Wylie, E.B., Stoner, M.A., and Streeter, V.L., “Network
System Transient Calculations by Implicit Methods”,
Society of Petroleum Engineers Journal, 1971, pp. 356-
362.
[6] Sparks, C.R., “On the Transient Interaction of Centrifugal
8 Copyright © 2012 by ASME
 Compressors and Their Piping Systems”, ASME Paper 83-
GT-236, 1983.
[7] Brun, K., Nored, M., 2007, “Application Guideline for
Centrifugal Compressor Surge Control Systems,” Gas
Machinery Research Council.
[8] McKee, R., Garcia, A., 2007, “Simulation of Centrifugal
Compressor Trips for Surge Avoidance System Design,”
Pipeline Simulation Interest Group, Calgary, Canada.
[9] Moore, J. J., Kurz, R., Garcia, A., and Brun, K., 2009,
“Experimental Evaluation of the Transient Behavior of a
Compressor Station during Emergency Shutdowns,”
ASME GT2009-59064, Proceedings of ASME Turbo Expo
2009: Power for Land, Sea, and Air.
[10] White, R. C., Kurz, R., 2006, “Surge Avoidance in
Compressor Systems,” Turbomachinery Symposium,
Houston, TX.
[11] Kurz, R., White, R.C., 2004, “Surge Avoidance in Gas
Compression Systems,” ASME Turbo Vol. 126 No. 4.
[12] Botros, K. K., “Single vs. Dual Recycle System
Requirement in the Design of High Pressure Ratio, Low
Inertia Centrifugal Compressor Stations,” Proceedings of
ASME Turbo Expo 2011: Turbine Technical Conference &
Exposition.
[13] Botros, K. K., “Selection Criteria of Spring Stiffness of
Nozzle Type Check Valves for Compressor Station
Applications,” Proceedings of ASME Turbo Expo 2011:
Turbine Technical Conference & Exposition.

9 Copyright © 2012 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/conferences/asmep/74824/ on 03/11/2017 Terms of Use: http://www.asme.org/abo

 ANNEX A MULTI-STAGE COMPRESSOR TRAIN DIAGRAM

MULTISTAGE CENTRIFUGAL COMPRESSOR TRAIN

Anti-Surge Valve
Out Flow Discharge
Cooler Header

Suction Check
Header Valve
Scrubber
Flare

Low Pressure Low Pressure Medium Pressure High Pressure

Gas Turbine Stage (LP) Stage (LP) Stage (MP) Stage (HP)
Driver

TABLE 1. COMPARISON BETWEEN DATA AND COMPUTATIONAL MODEL

STAGE  I ‐LP ‐ Point #2 STAGE  II ‐LP ‐ Point #1 STAGE  II ‐LP ‐ Point #2 STAGE  III ‐MP ‐ Point #1 STAGE  III ‐MP ‐ Point #2 STAGE  IV ‐ HP ‐ Point #1 STAGE  IV ‐ HP ‐ Point #2

PARAMETERS / COMPRESSOR 
STONER ‐  Relative  STONER ‐  Relative  STONER ‐  Relative  Reported  STONER ‐  Relative  Reported  STONER ‐  Relative  Reported  STONER ‐  Relative  Reported  STONER ‐  Relative 
STAGE Reported  Reported  Reported 
Calculated  Difference  Calculated  Difference  Calculated  Difference  Values ‐  Calculated  Difference  Values ‐  Calculated  Difference  Values ‐  Calculated  Difference  Values ‐  Calculated  Difference 
Values ‐ LP Values ‐ LP Values ‐ LP
Values (%) Values (%) Values (%) MP Values (%) MP Values (%) HP Values (%) HP Values (%)
Inlet Conditions

Pressure (bara) 1.02 1.02 0.000 4.164 4.164 0.000 4.263 4.263 0.000 9.85 9.85 0.000 9.93 9.93 0.000 32.25 32.25 0.000 31.88 31.88 0.000

42.78 42.78 0.000 39.87 39.87 0.000 29.89 29.89 0.000 39.93 39.93 0.000 29.94 29.94 0.000 39.93 39.93 0.000 29.94 29.94 0.000
Temperature (ºC)
45.21 45.21 0.000 30.73 30.73 0.000 29.24 29.24 0.000 30.39 30.39 0.000 28.95 28.95 0.000 28.11 28.11 0.000 26.91 26.91 0.000
Molecular Weight (kg/kmol)

Specific Gravity (‐) 1.561 1.561 0.008 1.061 1.061 0.000 1.010 1.010 0.000 1.049 1.048 0.116 1.000 0.999 0.050 0.971 0.970 0.052 0.929 0.929 0.008

0.985 0.985 0.000 0.975 0.975 0.000 0.975 0.985 1.026 0.943 0.943 0.000 0.944 0.944 0.000 0.84 0.84 0.000 0.841 0.841 0.000
Compressibility (Z1)
5381 5466 1.572 16241 16281 0.245 15647 15907 1.659 9588 9560 0.292 9328 9443 1.229 2105 2076 1.368 1981 2004 1.181
Inlet Actual Volume (m3/h)

Inlet Actual Volume (ft3/min) 3167 3217 1.572 9559 9583 0.245 9209 9362 1.659 5643 5627 0.292 5490 5558 1.229 1239 1222 1.368 1166 1180 1.181

5025.4 5004.0 0.426 63136.9 62680.0 0.724 64324.8 64875.0 0.855 91145.2 89860.0 1.410 92242.5 92570.0 0.355 73550.2 71430.0 2.883 70594.7 70590.0 0.007
Standard Flow (SCMH)
4.259 4.241 0.426 53.512 53.125 0.724 54.519 54.985 0.855 77.250 76.161 1.410 78.180 78.458 0.355 62.338 60.541 2.883 59.833 59.829 0.007
Standard Flow (MMSCFD)

Density (kg/m3) 1.782 1.778 0.245 5.043 5.026 0.337 5.074 5.065 0.184 12.195 12.130 0.535 12.085 12.049 0.297 41.462 41.034 1.031 40.481 40.191 0.717

0.111 0.111 0.245 0.315 0.314 0.337 0.317 0.316 0.184 0.761 0.757 0.535 0.754 0.752 0.297 2.588 2.562 1.031 2.527 2.509 0.717
Density (lbm/ft3)

Mass Flow Rate (kg/s) 2.66 2.70 1.323 22.75 22.73 0.092 22.05 22.38 1.472 32.48 32.21 0.825 31.31 31.60 0.928 24.24 23.67 2.385 22.28 22.38 0.456

Discharge Conditions
4.453 4.453 0.000 10.39 10.39 0.000 10.41 10.41 0.000 32.55 32.548 0.006 32.15 32.15 0.000 102.5 102.499 0.001 102 101.993 0.007
Pressure (bara)

Temperature (ºC) 112.7 110.587 1.875 94.4 96.261 1.971 85 83.253 2.055 122.8 125.549 2.239 113.2 110.657 2.246 146.7 150.28 2.440 139.1 136.36 1.970

Compressibility (Z2) 0.967 0.967 0.000 0.965 0.964 0.104 0.966 0.966 0.000 0.92 0.92 0.000 0.923 0.923 0.000 0.871 0.871 0.000 0.877 0.877 0.000

Reported Compressor Operating Conditions
92.8 92.8 0.045 81.6 81.9 0.357 81.4 81.6 0.185 107.7 108.1 0.401 108.2 108.5 0.252 105.6 107.8 2.095 108.7 110.6 1.728
Polytropic Head (KJ/Kg)
73 72.7 0.411 83.1 83.2 0.120 83 83 0.000 74.7 74.4 0.402 75.3 74.9 0.531 58 58.2 0.345 58.5 58.7 0.342
Polytropic Efficiency (%)

Speed (RPM) 10739.1 10739.1 0.000 10802.3 10802.3 0.000 10739.1 10739.1 0.000 10802.3 10802.0 0.003 10739.0 10739.0 0.000 10802.3 10802.0 0.003 10739.1 10739.0 0.001

Power (Kw) 338.7 344.3 1.662 2234.0 2236.7 0.118 2163.0 2199.1 1.672 4682.9 4824.5 3.025 4499.5 4717.0 4.834 4414.0 4518.0 2.356 4136.0 4343.0 5.005

Average Relative Difference ‐ All Parameters (%) 0.562 0.281 0.594 0.690 0.801 1.130 0.861

10 Copyright © 2012 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/conferences/asmep/74824/ on 03/11/2017 Terms of Use: http://www.asme.org/abo