43 Turbomachinery & 30 Pump Users Symposia (Pump & Turbo 2014) September 23-25, 2014 - Houston, TX - Pumpturbo - Tamu.edu

View metadata, citation and similar papers at core.ac.
uk brought to you by CORE

provided by Texas A&M University
43rd Turbomachinery & 30th Pump Users Symposia (Pump & Turbo 2014)
September 23-25, 2014 | Houston, TX | pumpturbo.tamu.edu
OVERVIEW OF IMPORTANT CONSIDERATIONS IN WET GAS COMPRESSION TESTING AND ANALYSIS
Grant O. Musgrove Melissa A. Poerner

Research Engineer Senior Research Engineer
Southwest Research Institute Southwest Research Institute
San Antonio, TX, USA San Antonio, TX, USA
Massimiliano Cirri Matteo Bertoneri

GE Global Research – Conceptual Design Engineer
Europe GE Oil & Gas – Nuovo Pignone
Germany Florence, Italy
.
Grant O. Musgrove is a Research compressors for LNG and refinery jobs, and in Rotordynamic
Engineer in the Machinery Program at team, working on high speed rotor balancing. He has also
Southwest Research Institute. He currently worked as Test Engineer, responsible of Mechanical and
conducts applied research for Performance Tests of Centrifugal Compressors,
turbomachinery applications in the Oil & Turboexpanders and Steam Turbines. Mr. Bertoneri received
Gas and power generation industries. His a B.S. and M.S. degree (Mechanical Engineering, 2010) from
active research areas are wet gas University of Pisa.
compression, supercritical CO2, and
turbomachinery design. Mr. Musgrove’s ABSTRACT
responsibilities range from technical analysis to project During upstream production of natural gas fields, it is
management for both experimental and computational common that a gas-liquid mixture of product is brought to the
activities. Mr. Musgrove graduated from Oklahoma State surface. The mixture, termed wet gas, is generally made up of
University in 2007 and from The Pennsylvania State mostly gas with a small amount of liquid, typically up to 5%
University in 2009 with B.S. and M.S. degrees in Mechanical by volume of the mixture. Because of the difficulties of
Engineering, respectively. compressing wet gas, the practical approach has been to
separate the liquid and gas phases before compression.
Melissa A. Poerner is a Senior Research However, large separation equipment is unfavorable for
Engineer and Test Program Coordinator in subsea installations because of the cost to place machinery on
the Machinery Program at Southwest the sea floor. Instead, a compressor designed for wet gas
Research Institute. Her background operation is preferred because it eliminates the need for large
includes work related to analysis and testing separation equipment leading to plant simplification and cost
of compressors and other large machinery reduction. To address this design need, researchers have been
in adverse conditions such as wet gas or active in addressing the challenges with wet gas compression.
corrosive gas. In the past, she has been As result, experimental work has been conducted to study the
directly involved with design, operation, and project effects of wet gas on compressor aerodynamic and mechanical
management or wet gas compression test programs. Ms. performance. This experimental research has presented many
Poerner’s work experience is supported by a Bachelor’s of challenges in recreating wet gas conditions and quantifying
Science in Mechanical Engineering from Texas A&M the effect of the liquid on the compressor performance. The
University and a Master’s of Science in Mechanical results from this testing have helped to characterize the
Engineering from Georgia Institute of Technology. performance effects. But so far each work has focused on a
range of test variables without identifying those that have the
Matteo Bertoneri is a Conceptual Design largest effect on compressor performance. This paper aims to
Engineer within GE Oil & Gas, in provide the reader with an overview of the completed wet gas
Florence, Italy. His current duties involve research, the challenges associated with doing the
research, development and prototype- experimental work, and a discussion of the resulting trends
testing activities on mechanical and observed in most of the wet gas research. This will include an
thermodynamics conceptual design of new in-depth review of relevant literature on wet gas compression
Centrifugal Compressors. His previous testing and performance, a discussion of the important
tasks were in Requisition office, designing
Copyright© 2014 by Turbomachinery Laboratory, Texas A&M Engineering Experiment Station

research topics in wet gas compression, and a description on As the gas field becomes depleted, the reservoir
how wet gas experiments are set-up, performed, and the pressure and flow rate decrease. Therefore, the suction
challenges associated with that testing. Also, this paper pressure at the compressor inlet reduces over time along with
reviews the available test data using a multiple regression the compressor delta pressure and flow rate. The change in
analysis to identify the important test variables and their effect flow conditions directly affect the liquid volume fraction
on compressor power and pressure ratio. Some of the test entering the compressor; whereby a maximum of 5% LVF is
parameters that are discussed are inlet pressure and estimated as an upper limit for wet gas compression. Higher
temperature, gas-liquid temperature difference, liquid volume LVF than 5%, however, may be experienced in the field.
fraction, and speed. The results of the analysis are useful for Depending on the required flow rate and number of units at
establishing variables for a future test program to focus on the field, compression power may vary from 5 to 12.5 MW. It
operating conditions with the largest effect on performance. is important to note that all three of these subsea compression
Additionally, the effects of wet gas on machinery performance projects are not currently operational in the field. The Gullfaks
are discussed relative to machine vibration and seals. Using and Åsgard projects are planned for field installation in 2015
the observed trends in test data and the knowledge from while the Ormen Lange in 2016 [9, 10, 11, 6]. All of these
previous wet gas research, conclusions are presented to guide projects are currently working on research programs to
future analytical and experimental work in the area of wet gas understand how their compression systems will perform and
compression. how reliable they will be when they are installed. The past and
current research to support the planned installations of subsea
INTRODUCTION compression systems is discussed in more detail below.
The flow stream from offshore upstream production of Current Research
wellheads will often have a mixture of both gas and liquid There have been concentrated efforts to understand
with up to 5% liquid volume fraction. The liquid is often wet gas compression from both theoretical and experimental
removed from the flow stream to avoid the negative influences perspectives. While wet gas research is ongoing, the analytical
of the liquid on the compression process. As production and experimental research that has been completed to date in
equipment is moved to the ocean floor, it becomes desirable to the area of wet gas compression is discussed below
reduce the equipment footprint by eliminating the liquid
separation equipment. Without liquid separation, however, the Experimental Research
gas compressor must operate with two phase flow that can The primary organizations that are currently involved in
have significant effects on the compression process including experimental wet gas research are the Norwegian University
both performance and reliability [1, 2, 3]. Therefore, it is of Science and Technology (NTNU), Kårstø laboratory (K-
important that compression systems be designed or modified lab), Framo Engineering, Dresser Rand (DR), General Electric
to account for the liquid influences and maintain the same (GE), and Southwest Research Institute (SwRI).
efficiency and reliability as in dry compression. To improve NTNU represents the academic research area in wet gas
compressor designs, there have been several experimental and compression. NTNU has a single-stage centrifugal impeller
analytical research programs focused on understanding and test facility dedicated to wet gas compression research. The
improving wet gas compression. This paper aims to give the facility operates at ambient pressure and is used to
reader a comprehensive understanding of the state-of-the-art measurement compressor performance and aerodynamic
of wet gas compression, what is currently known about wet stability, and surge detection in wet gas compression. Much of
gas compression, the research being done to advance the this work is complimented with analytical research which will
technology, and the future planned research and trends for wet be discussed later [12, 13].
gas compression. Also in Norway is the Kårstø laboratory which is
commonly known as K-lab. K-lab is a metering and
PAST AND CURRENT RESEARCH technology laboratory that is integrated with the Statoil Kårstø
processing plant that serves gas and condensates from the
Field Conditions Norwegian continental shelf. In recent years, K-lab has been
Before looking into the past and current research, it is used to test wet gas machinery at actual field operating
important to have a clear picture of the expected operating conditions [14].
conditions for a wet gas compression system. As is often the In 2008, a full scale wet gas compression test facility was
case, the research conditions are not actual conditions due to constructed at K-lab, where two compressors from different
restrictions of time, money, and resources. The most well- vendors were tested. These compressors were subjected to
known subsea compression applications are the Gullfaks, flow streams up to 2-3% LVF with a mix of natural gas, water
Ormen Lange, and Åsgard. A review of these production field and condensate [15, 16, 17, 18]. In 2013 K-lab commissioned
conditions indicate that the typical compression system is as a facility to test compressors for wet gas applications while the
outlined below [4, 5, 6, 7, 8]. compressors were submerged under water.
• Suction Pressure: ranges from 20 to 140 bara over Statoil has partnered with Dresser-Rand in 2003 to
lifetime of field complete testing of a single stage compressor under wet gas
• Compressor delta pressure: 30 to 60 bara conditions at the Dresser-Rand laboratory. The single-stage
• Compressor flow rate: 2 to 30 SMM^3/day testing was completed at suction pressures of 30 and 70 bara
• Liquid Volume Fraction (LVF): 0.25 to 5% (which is representative of the actual conditions) and with a
• Compressor power rating: 5 to 12.5 MW hydrocarbon/condensate mixture [3].

Framo Engineering, also based on Norway, is testing the compression have typically been based on thermodynamic
compression system that is planned for installation in the effects of the liquid phase on compressor performance. With
Gullfaks field. Framo have built a loop to test an axial the presence of liquid, the heat of compression is dissipated
compressor in wet gas conditions, including submerged into the liquid which in turn evaporates throughout the
compressor tests. The testing at Framo Engineering includes compression process. Most of the open literature focuses on
evaluation of compressor performance, mechanical the thermodynamic effect of wet compression such that the
performance, and system endurance. Testing will be aerodynamic effect is only accounted through the gas
completed with an idealized fluid made up of nitrogen and properties. Much of the work for predicting wet gas
Exxsol D80 and also a real fluid comprised of hydrocarbons compression performance is founded in one-dimensional
[4]. models of the compression path that are coupled to thermal
General Electric (GE) has been heavily involved in models of liquid droplets that include evaporation. However,
evaluating and designing for wet gas compression. In 2008 three-dimensional CFD methods have been utilized for axial
they conducted performance testing of a single stage compressor modeling. This section will provide a brief
compressor at ambient pressures at their Munich Research overview of the analytical models that have been applied for
Facility [19]. In 2010, they partnered with Southwest Research wet gas compression.
Institute® (SwRI®) and completed testing on a two stage
Direct Integration Approach
compressor in wet gas conditions at an elevated suction
Characterizing the performance of wet gas compression is
pressure of 20 bara with air and water [1,2]. In 2013, they
difficult when defining a polytropic efficiency using the
completed another set of tests with air and water at SwRI but
widely-accepted equations from ASME PTC-10 [20]. The
with a single stage rotor. This testing also included more
equations of PTC-10 are based on the method presented by
detailed aerodynamic measurements and a strong focus on
Schultz [21] to characterize compressor polytropic
rotordynamics as compared to the testing completed in 2010.
performance using real gases instead of the ideal gas
SwRI has been actively involved in wet gas compression
assumption. In the method, the inlet and exit conditions of the
research with General Electric in the past 4 years. In addition,
compressor are used along with gas properties taken at the
SwRI has conducted independent research to quantify
average conditions between the inlet and exit. In wet gas
aerodynamic performance in wet gas conditions using a low-
compression, however, the compression process is expected to
speed atmospheric wind tunnel. The advantage of the wind
have a significant effect on the fluid properties due to phase
tunnel is that the flow is essentially isothermal and
change of the liquid. Therefore, a direct integration method
incompressible, resulting in test conditions that emphasize
has been used to quantify wet gas compression by iterating
aerodynamic effects instead of the compressibility and thermal
along incremental pressure steps through the compression
effects present in actual compressor testing. The results of this
process. In this method, the polytropic efficiency is estimated
internal work are currently unpublished.
for the calculation and held constant for all iteration steps.
The general trends observed in the wet gas compression
From the method, the polytropic head and efficiency for the
testing that has been documented in the public domain are
compressor are determined when the calculated discharge
listed below.
temperature matches the actual value. The direct-integration
• The compressors dry performance does not
method has been shown by Huntington [22] to be more
significantly change after it has been subjected to wet
accurate than the Schultz method, and is compared among
conditions
different equations of state by Hundseid [23]. A major
• There is a significant power consumption increase advantage of the direct-integration method over the Schultz
when liquid is introduced into the flow stream method is that the fluid properties are calculated for each step
• The pressure ratio across the compressor generally in the calculation instead of an average. Therefore, the method
increases when liquid is present is better suited for high pressure ratio impellers or applications
• The temperature ratio across the compressor with significant property variation. The direct-integration
decreases when liquid is present method has been applied to wet gas compression data by
• The volume flow rate through the compressor is Hundseid [24] to illustrate the large difference between the
reduced from the dry condition when liquid is in the Schultz and direct-integration methods for wet gas
flow stream compression. The drawback to the direct-integration method,
• Liquid droplet size or flow pattern has no effect on however, is that both fluid phases are assumed to be in
compressor performance when injected far enough equilibrium. This leads to the fact that the prediction method is
upstream of the compressor to allow a natural two- based on knowing the discharge temperature of the
phase flow regime to develop. Liquid droplet size or compressor, which is difficult to establish whether the
flow pattern has a noticeable effect on compressor measured value at the discharge flange is either the liquid or
performance when the liquid is injected at the gas temperature. Nevertheless, the direct-integration method
compressor inlet guide vanes provides a promising approach to characterizing wet gas
There is a significant amount of analytical research that is performance
being done in conjunction with the above experimental
research. This research is described in more detail below. Evaporation Performance Models
To predict the effect of wet gas on compressor
Analytical Methods performance, the most popular approach in the literature is to
The analytical methods that have been applied to wet gas couple a one-dimensional model of the gas compression path

to a thermal model of the liquid phase. The coupling of the between the discrete phase and multiphase models; however,
two models is done through equations of enthalpy and entropy the multiphase models present a clearer picture of the liquid
that include phase change of the liquid as the gas temperature distribution near the compressor blades. In other words, the
and pressure changes. The liquid phase in the model is usually concentration of liquid film on the pressure side of the blades
assumed to be uniformly distributed in the flow in the form of and leaving the blade trailing edge is explicitly seen in the
very small, spherical droplets. The liquid is treated as a multiphase results
discrete phase such that the presence of the liquid is assumed
to not affect the gas flow and droplet-droplet interactions are IMPORTANT CURRENT RESEARCH TOPICS
neglected. When individual liquid droplets are not considered
in the analysis, bulk treatment of the liquid phase change is Compressor Aerodynamic Performance
handled by calculating vapor fraction and enthalpy of The primary application where wet gas compression is
vaporization. Considering individual liquid droplets, however, being considered is for subsea gas extraction. In this
the evaporation is calculated using an estimate of the heat application and many others, the important operating
transfer coefficient around the droplet that is calculated from a parameters for a compressor are pressure head and flow rate.
slip velocity. Operators desire to run the compressor in such a way to obtain
The method with and without individual treatment of the enough pressure head to transfer the gas from the production
droplets is explained well by Härtel and Pfeiffer [25] for an field to shore or an off-shore platform at a desired flow rate
axial compressor who compare the droplet model to using an and delivery pressure. If the compressor cannot generate
ideal model that does not account for droplet size. Between the enough pressure head, then the flow rate will be reduced or the
droplet model and an ideal model, the droplet model includes delivery pressure will be lowered or both. Both of these
droplet diameter effects of evaporation that can be important parameters will impact the production of the gas field.
for predicting droplet penetration in multi-stage compressors. Another factor to consider is that over time the gas field
Their work showed that the ideal model was equivalent to the depletes to lower compressor suction pressure. As a result, the
droplet model when using 1micron droplet sizes. Other delivery pressure and mass flow rates will be lower for both
analyses using the evaporation method on axial compressor dry and wet applications. However, if the performance of the
performance have been done [26, 27], while many other works compressor in wet conditions is influenced by the suction
have been neglected here for brevity. Beyond the one- pressure, then the variation must be well understood for
dimensional model, White and Meacock [28] conducted a operation of the compressor.
pseudo two-dimensional analysis of wet gas compression by When compressors are selected for gas field production
calculating multiple mean lines through the flow path. From operation, they are designed and selected to have an operating
the droplet trajectory predictions, they estimated the amount of map that will cover the various operating conditions that are
droplet impact on the blades as well as liquid film migration expected. If the performance of the compressor significantly
and evaporation. changes due to the presence of liquid, then this will cause
The evaporation model has been applied to centrifugal limitations in the operation of the compressor. Figure 1 shows
compressors using the same assumptions as axial compressor a graph of the performance of a two stage centrifugal
analyses. Abdelwahab [29] coupled the droplet model to a compressor in dry and wet conditions. The dry condition is
mean line prediction for a centrifugal compressor to estimate show as black lines and the wet is individual data points. The
the performance effects. Fabrizzi et al. [19] attempted to use wet condition data points clearly show when the compressor is
the droplet model to better understand the thermodynamic subjected to wet conditions, the performance changes.
effects of the liquid droplets when comparing to test data, and Therefore, the compressor in a gas field may not operate as
obtained an estimate of polytropic efficiency. needed if exposed to wet conditions.
To combat this issue, compressor manufacturers, users,
CFD Modeling and research institutions have been actively trying to find
Three-dimensional CFD methods have been applied for appropriate methods to predict compressor performance in wet
wet compression in axial compressors. The work of Zheng et gas conditions. At the core of this issue is the change in the
al. [30] and Luo et al. [31] have used a multi-stage axial aerodynamic performance of the compressor. When liquid is
compressor model to study the thermodynamic effects of introduced into the flow stream, the temperature ratio across
evaporating liquid droplets on aerodynamic performance. the compressor drops, the pressure ratio changes, the power
These models, similar to the one-dimensional methods, only consumption increases, and the flow rate is limited. Some of
account for the thermodynamic effect of the liquid. Khan [32] these influences are related to thermodynamic type issues
conducted a two-dimensional CFD study to compare the (temperature ratio decreases because you have a higher mass
predictions between using the discrete phase model or a of fluid with a higher heat capacity, which requires more
multiphase model. The difference between the two models is energy to have a temperature change) while other influences
in the calculation of the second phase. The discrete phase are related to aerodynamics. For instance, the blade shape
model follows the method of previous studies that calculate essentially changes as a liquid film flows across the
individual droplet trajectories using a force-balance on the compressor blades and can affect the flow near the blade
droplet while the multiphase method treats the liquid as a surface. A study by Grüner et al. [33] qualitatively showed
continuum and solves the momentum and energy equations. through observations that wet gas flow around a single
For both comparisons, a small amount of liquid mass is compressor blade results in a liquid film flowing along the
injected into the domain equivalent to 10µm droplets for the blade. They concluded that the liquid film contributes to early
discrete phase simulations. The results were nearly identical separation on the airfoil to result in reduced aerodynamic

performance. To illustrate the effect of wet gas on compressor little influence of the wet conditions on the radial vibrations.
power requirement, Gilarranz et al. [34] presented a simple Ransom, et al. [1] also indicated little influence of wet gas
correlation of the relationship between wet and dry power conditions on radial vibration. However, they did find that the
requirement based on the mass flow of the liquid and gas. axial vibration was influenced by the wet conditions. With the
Converting the correlation to the common wet gas parameter limited published literature on mechanical stability with wet
of LMF, the ratio of wet-to-dry power is found to agree well gas conditions, it is difficult to draw general conclusions.
with experimental data (Figure 2) when LMF is less than 50%. More experimental research on the influence of wet gas
It should be noted, however, that this correlation assumes that conditions on the stiffness and damping of internal
compressor efficiency is not affected by the liquid in the components is necessary.
compressor.
1 Erosion
𝑊𝑒𝑡 𝑃𝑜𝑤𝑒𝑟 = 𝐷𝑟𝑦 𝑃𝑜𝑤𝑒𝑟 � � In wet gas conditions, there is a potential for erosion
1 − 𝐿𝑀𝐹 of internal compressor components. There are many studies on
erosion of metallic surfaces by a gas/liquid flow stream. One
example is the erosion of steam turbine blades [35, 36, 37].
Pacheco et al. [38] evaluated several steam turbine erosion
models and adapted one for wet gas compressions.
Little or no experimental data has been published
specifically on erosion in the wet gas conditions described
above. However, validated models developed in other
applications can be applied to the wet gas applications given
that important parameters such as material hardness, droplet
size, and droplet velocity are matched to the wet gas
application
System Performance
Much of the focus of research has been on quantifying/
predicting the performance of the compressor with wet gas
Figure 1. Two Stage Compressor Map Showing Variation conditions. It is important to note that the piping, valves and
between Dry and Wet Performance other items connected to the compressor also influence the
compressor operation. One area that has lacked significant
3.00 development is modeling of the full system exposed to wet gas
8,000 rpm Wet conditions. There are several system performance
2.80 9,500 rpm Wet
11,000 rpm Wet
considerations that need to be taken into account when
2.60 selecting a compressor and designing the piping system
1/(1-LMF)
2.40 connected to the compressor such as system resistance,
2.20
temperature effects, slugging issues, and two-phase flow
Wet Power
modeling. For example, liquid slugging upstream of the
2.00
Dry Power compressor can cause the compressor to surge by momentarily
1.80 starving the compressor of gas flow
1.60 Liquid Separation
1.40 Liquid separation is an important topic for wet gas
1.20
compression research because the outlet of the separator is
typically the inlet of the compressor. Separator footprint is
1.00
0% 20% 40% 60% 80% 100%
being minimized for subsea systems which can result in
Liquid Mass Fraction (LMF) decreased separation efficiency. Therefore, it is important that
the separation efficiency be known for selection of the
Figure 2. Comparison of Correlation from Gilarranz et al. compressor to estimate how much liquid will enter the
[34] to Test Fata compressor.
Compressor Rotordynamics Liquid Flow Profiles
Another important part of compressor operation is As mentioned above, it has been seen in testing that
mechanical stability; whereby operating outside of the normal varying droplet size does not cause variation in compressor
operating regime can result in significant physical damage to performance if the liquid flows through a length of suction
the machine. Mechanical stability is even more critical for pipe. As the flow stream travels along the suction piping, the
subsea applications due to the fact that the compressor is liquid distribution will develop into a natural two-phase flow
essentially inaccessible for repair on the ocean floor. Most that is determined by the pipe geometry and flow conditions.
subsea compressors are being designed for 5 years continuous For example, liquid injected as droplets may coalesce into
running before an overhaul is required. larger droplets or a river at the bottom of the pipe, resulting in
The influence of the wet gas conditions on mechanical annular or stratified flow. The development of the two-phase
stability is still unclear. Brenne, et al. [3] found that there was flow regime results in a somewhat consistent flow profile for

the liquid at the inlet of the compressor. This is especially true
for any suction piping that has elbows or sharp turns that the
flow experiences.
Currently, there is no research that identifies a desired
flow profile for the gas/liquid mixture at the inlet. For
instance, it is not known if small droplets or a liquid film
would allow for higher compression efficiency. With the
effect of the liquid distribution on compressor performance,
future compressor inlets could be designed to achieve the
desired flow profile.
Slugging
Depending on the layout of the compressor piping and
valves, it is possible to generate liquid slugs upstream of the
compressor. Slugs can be generated from vertical piping, tight
flow restrictions, or very low gas velocity. Liquid slugs
entering the compressor will cause the compressor to surge
and should be avoided. Liquid slugging is primarily a concern Figure 3. Dry Compressor Test Loop
for wet gas compressor installations without an upstream Figure 4 shows the dry compressor test loop with the
liquid separator. Compressor installations that utilize an added equipment necessary for wet performance testing
upstream liquid separator have low risk of slugging. highlighted in blue. Some of the key components include:
• Liquid storage tank – Needed to have volume of fluid
WET GAS COMPRESSOR TEST SET-UP at pump suction and to generate enough NPSH
There are many challenges that are presented when • Liquid pump – Used to pressurize fluid for injection
constructing and performing a compressor performance test in into the gas loop
wet conditions. In the laboratory setting, matching the actual • Liquid flow meter – Measures the flow rate of liquid
conditions is not always possible or desired. Often a smaller
into the gas loop to control LVF and LMF
scale test at less severe conditions is convenient for gaining
understanding into basic phenomena. This provides some • Liquid Injection system – Introduces liquid into flow
simplification to the testing, but there are still many systems loop in desired pattern (droplets, film flow)
that must be operated and tuned to work in the desired • Liquid/gas separator – Removes the liquid from the
conditions. gas before gas returns to suction of compressor.
A typical wet gas test set-up in a laboratory with air and
Liquid is flowed back to tank.
water is described below. This includes a review of various
systems needed for a wet gas test and the challenges • Liquid cooler – Removes heat from liquid absorbed
associated with operation of these systems. during compression process
• Control valves – Used to control flow of liquid
General Arrangement
In a laboratory setting, wet conditions are simulated for through pump, out of liquid/gas separator, and
testing by injecting liquid somewhere upstream of the through cooler
compressor. When the liquid conditions are created, it is
important to have tight control on the amount of gas and the
amount of liquid flowing into the system such that specific
Liquid Volume Fractions (LVF) and Liquid Mass Fractions
(LMF) can be generated. In addition to generating the
appropriate LVF and LMF, the loop must be capable of
reaching steady operating conditions and measuring the
compressor performance.
Figure 3 shows a typical dry compressor closed test loop.
The test compressor is labeled C-101. The piping connected to
the compressor has a control valve, a gas cooler, and
instrumentation to measure pressure, temperature, and flow
(orifice meter run). A few other features include the sealing air
system, the motor and gearbox, and the loop blow-off silencer. Figure 4. Wet Compressor Test Loop
Note that in dry compressor performance tests, pressure and
temperature are typically adequate for measuring the
compressor performance. In wet compressor performance
tests, however, additional equipment and measurements are
needed for the liquid.

Liquid Injection average droplet diameter, almost every LVF can be achieved.
The liquid injection system introduces the liquid into the
Table 1. Liquid Flow Rate and LVF vs. Number of
gas flow. This is comprised of the tank, pump, control valve,
Nozzles Open for Fixed Average Droplet Diameter
instrumentation, and injectors. In a laboratory setting,
researchers typically seek to achieve a specific flow profile No. of Liquid flow rate LVF
(droplets or film flow) at the injection location. If injecting Nozzles Open (gpm) (500 ACMH gas flow)
water droplets, there is typically a desired mean droplet 1 6.5 0.3%
diameter. 2 13 0.6%
Typical injection systems have a fixed number of 3 19.5 0.9%
injection nozzles and allow for flow through all or a subset of 4 26 1.2%
those nozzles. Injection nozzles have a relationship between 5 32.5 1.5%
the average droplet diameter, pressure differential across the
nozzle, and the flow rate through the nozzle. The pressure 6 39 1.7%
differential usually drives the droplet size and the flow rate 7 45.5 2.0%
through the nozzle. When injecting liquid into the flow stream, 8 52 2.3%
the average droplet diameter and the flow rate (LVF) are
fixed. Therefore, a fixed pressure differential must be 9 58.5 2.6%
maintained across the injector nozzles. As a result, the flow 10 65 2.9%
rate can only be varied incrementally by opening or closing
individual injector nozzles. Figure 5 shows the relationship of
pressure differential to droplet diameter and flow rate for a
typical nozzle where the solid circle indicates the desired
average droplet size of 400 μm. To achieve this average
droplet diameter, the pressure differential must be 3.2 bar with
a provided flow rate of 6.5 gpm. Table 1 shows the various
flow rates and LVF that could be achieved with up to 10
nozzles using a fixed pressure differential of 3.2 bar. There is
a large range of flows available based on nozzle configuration;
however, experimental testing is usually requires tests at
specific target values. In wet gas testing, typical target LVF
values are 0.5, 1, 1.5, 2, 2.5, and 3%. With the available flow
rates listed below, the 0.5, 1, and 2.5% targets could not be
met while providing a consistent mean droplet size.
Figure 6. Injector Nozzle Droplet Diameter, Flow Rate, and

Differential Pressure Curves for Range of Average Droplet
Diameter
Table 2. Liquid Flow Rate and LVF vs. Number of
Nozzles Open for Range of Average Droplet Diameter
No. of Water flow LVF
nozzles rate (gpm) (500 ACMH gas flow)
0pen Min Max Min Max
1 5.3 8 0.2% 0.4%
2 10.6 16 0.5% 0.7%
3 15.9 24 0.7% 1.1%
4 21.2 32 1.0% 1.4%
Figure 5. Injector Nozzle Droplet Diameter, Flow Rate, and 5 26.5 40 1.2% 1.8%
Differential Pressure Curves for Fixed Average Droplet
Diameter 6 31.8 48 1.4% 2.1%
7 37.1 56 1.7% 2.5%
If there is some leniency on the average droplet diameter, 8 42.4 64 1.9% 2.8%
then the pressure differential across the injector nozzles can be
varied to allow other target LVF values to be met. Figure 6 9 47.7 72 2.1% 3.2%
shows that if the average droplet size can be varied from 350 10 53 80 2.4% 3.5%
to 450 µm then this allows for a variation in flow from 5.3 to 8
gpm. The range of LVF that can be achieved with a different To be able to achieve the control of the flow rate,
number of nozzles is outlined in Table 2. With leniency on droplet size, and pressure differential as discussed above,

provisions must be in place for control of the pump flow rate sufficient to remove enough liquid to have the desired
and pump discharge pressure. The configuration shown in accuracy for LVF or LMF. Note that there is no instrument
Figure 4 allows for pump control using a pump bypass loop placed after the separator to detect how much if any liquid is
with a fixed-speed pump. Pump discharge pressure is present. Therefore, if the separator does not have sufficient
maintained at or near a fixed value during all testing. The flow liquid separation efficiency, the measured LVF or LMF can be
rate from the pump exit is also fixed, but a portion of the flow skewed to an unknown value.
is diverted back to the tank using the bypass loop. An actuated During testing at wet conditions, the liquid must be
control valve on the bypass loop is used to control the liquid removed from the separator to prevent the separator from
flow rate to the injection system. overflowing. Also, in a closed loop system the pressure in the
The control valve used for adjusting the flow rate to the loop will increase as the separator level increases.
injection system should have the capability to make small Alternatively, the loop pressure decreases with decreasing
incremental changes (on the order of 1% change in valve liquid level in the separator. It is recommended that a control
position), which is important to reach specific LVF values. valve be placed on the liquid exit of the separator to control
Also, the valve should be sized such that the full flow from the the liquid flow leaving the separator within 1% opening
pump can flow through the control valve to the tank without increments. With this control valve, the flow rate of the liquid
exceeding the manufacturer recommended velocity limits. out of the separator can be matched to the flow rate into the
During start-up of the pump, it is important that the control separator. This will allow the separator level to remain
valve be wide open and allow flow only back to the tank. This constant and the pressure in the test loop to be steady. Figure 7
will avoid introducing any unexpected liquid into the gas flow shows an example of a system where a control valve was not
stream that could flood the compressor. After startup, the used to regulate the liquid flow out. Instead, a solenoid valve
control valve should slowly be closed to start introduction of was used and the liquid was flowed out of the separator
liquid into the flow stream. Moving too quickly to introduce a intermittently. This resulted in significant pressure fluctuations
large amount of liquid flow to the compressor could cause in the process (0.5 bara pressure fluctuations).
unexpected damage to the compressor.
The liquid injection system also requires pressure
measurement, temperature measurement, and flow
measurement. The pressure measurement should be placed as
close to the injection location as possible. If the pressure
measurement is placed far upstream of the injection system,
then the actual differential pressure across the injector nozzles
will be unknown due to line losses. The temperature
measurement should also be placed as close to the injection
location as possible. This measurement allows the test
operator to determine the temperature difference between the
gas and liquid temperatures at the injection location.
Researchers are interested in both the identical and varying
temperature between the gas and liquid. The gas and liquid at
Figure 7. Pressure Fluctuations in a Test Loop Due to
the same temperature is representative of a gas/liquid flow at
Intermittent Opening and Closing of a Solenoid Valve on
the suction of the compressor. If the gas temperature is higher
Separator Liquid Exit
than the liquid temperature, this could be representative of a
compressor with a side stream flow. Sealing on the Compressor
The liquid flow measurement should be made with a flow Compressors in dry operation can have a variety of
meter that has sufficient accuracy to measure the lowest liquid different seals including labyrinth seals, brush seals, and dry
flow rate (lowest LVF on left hand side of map near surge at gas seals. These sealing systems are designed to operate with
lowest speed) but still be able to handle the highest liquid flow dry gas present in the seal. In wet gas operation, both gas and
rate (highest LVF on right hand side of map near stonewall at liquid are present in the flow stream which means that there is
highest speed). Since the liquid flow rate is dependent on the high likelihood that liquid can enter the sealing area. The
desired LVF and the flow rate of the compressor (varies across authors of this paper primarily have experience with using
the map) it can vary significantly (over 10:1 ratio). For labyrinth seals in wet conditions.
example, the dry compressor map shown in Figure 1 has a When using labyrinth seals, purging gas at a pressure
minimum flow (at the lowest speed) of 320 ACMH and a higher than the compressor suction pressure is required to
maximum flow (at highest speed) of 1450 ACMH. If the keep liquid from entering the seals. It is important to note that
minimum and maximum LVFs were 0.5% and 3%, changes between dry and wet conditions, speed, and operation
respectively, this means that the liquid flow rates would need changes can lead to pressure fluctuations at the suction of the
to be 7 and 197 gpm, resulting in a 27:1 ratio. compressor. Some of these changes will cause the suction
pressure to increase, which could exceed the purge gas
Gas/Liquid Separation
pressure if not monitored closely. If the suction pressure
A gas/liquid separator is used in the test loop to remove
exceeds the purge gas pressure, a mixture of liquid and gas
the liquid from the flow stream after it has traveled through
can enter the labyrinth seals. It should be noted that the
the compressor. This allows control of LVF and LMF at the
authors have found that liquid enters labyrinth seals even with
suction of the compressor. The separator efficiency should be

a pressure differential between the purge gas and compressor to increase the pressure at a specified rate. This is especially
suction. The movement of the liquid from compressor suction true for any static measurement devices such as pressure
to the labyrinth seals may be due to secondary effects in the scanners that may not be able to allow for fast pressure
sealing system, and is currently being investigated. transients. This same consideration is true for depressurizing
The condition and clearance of the labyrinth seals is the system.
also important. The seals cannot have large clearances due to There are various methods for measuring pressure in a
wear, which will not keep the liquid inside the compressor wet gas system. In many compressor performance testing
loops, pressure measurements are made via pressure scanners.
Pressure, Temperature, and Flow Rate Measurement and
In wet systems, pressure scanners are difficult to use, because
Control
the pressure lines fill with liquid during testing. It is necessary
With any research-grade testing, it is important to control
to purge these lines intermittently to remove the liquid and
flow rate, pressure, and temperature. The flow rate of the gas
also protect the pressure scanner (if a dry-only scanner is
and liquid are easily controlled with actuated valves. The
used). Purging systems can be elaborate and costly. However,
liquid flow rate control has been previously discussed. The gas
these systems are often necessary to obtain pressure
flow rate is controlled by varying the opening of an actuated
measurements inside the compressor or to acquire a large
control valve on the discharge of the compressor. Flow
number of pressure measurements in a cost effective manner.
measurement of the gas flow rate can be done with
For static pressure measurements on piping, direct mounted
conventional gas measurements techniques. However, it is
probes are recommended. This eliminates the pressure tubing
important that the gas flow rate be measured at a location that
and the purging system. Note, that the orientation of the probe
has “dry” gas only. When testing in a closed loop, the gas will
is important. If a probe is oriented on the bottom of the pipe,
likely be saturated, but no liquid droplets should be present
the sensing face will almost always be flooded with liquid.
(unless there is significant carryover from the separator).
Temperature control is important to be able to meet
Figure 1 shows that the gas flow measurement as an orifice
desired conditions and maintain consistent test conditions for
meter run which is placed after gas/liquid separation and the
comparison at a later time. In wet gas testing, the three
gas cooler. Two orifice meters were used for this loop to
primary temperatures that must be controlled are the gas inlet
improve the accuracy of flow measurement in both the lower
temperature, liquid inlet temperature, and combined gas/liquid
and upper flow ranges.
inlet temperature. The gas/liquid inlet temperature is a
The pressure in the loop is controlled by two mechanisms.
function of the gas/liquid mixing upstream of the compressor
The first mechanism, the separator liquid level, has already
and thus is controlled by the flow in the upstream piping.
been discussed. Note that it is desirable to have a somewhat
During a wet test, the liquid is heated as it flows through
consistent separator level, but this level can be used to control
the compressor. In a closed system, the heat gained by the
the pressure inside the loop if needed. The second pressure
liquid must be removed before it re-injected. A separate cooler
control mechanism is the source for pressurizing the closed
for the gas and liquid flows is recommended for precise
loop. In the test loop being discussed, the loop is pressured
control. In past testing, a single cooler for the mixed flow
through the seal purging system. The seal purging system
stream has been used, but it was found that the liquid was not
continuously has a flow of gas into the seal area to create a
adequately cooled in this configuration and the liquid
seal buffer that sustains the pressure in the loop. Seal purging
temperature slowly increased overtime. Therefore, a steady
systems can be set up with automatic or manual valves to
liquid inlet temperature could not be maintained. With
control the flow and pressure of the gas going into the seal
separator coolers, a steady liquid inlet temperature can be
chambers. The authors of this paper, however, have found that
maintained.
a manual pressure regulator is sufficient for controlling the
Since a separate cooler is needed for both gas and liquid,
seal purging pressure. The primary issue with the seal purging
this means that cooling occurs after separation. When testing a
system is that it cannot automatically allow the seal purge
wet system, the gas is fully saturated with liquid at all points
pressure to increase (in case of a transient type of event that
in the test loop. Therefore, after the gas has cooled, there
causes the suction pressure to increase, as discussed above).
could be some liquid drop out. If the cooler is placed after the
It is important to note that since there is more than one
separator, then this liquid will travel to the suction line where
component used to control sealing pressure; these two devices
injection occurs and the liquid will not be accounted for in the
will interact with each other. For example, if the manual
LVF or LMF measurement. However, the liquid drop out can
pressure regulator on the seal purging system is set at 20 bar,
be estimated. Past experience testing with two separate coolers
but the current test is being conducted at 18 bar, the separator
has shown that there is an insignificant amount of liquid drop
level liquid will have to be continually decreasing in order to
out after the gas cooler, within the measurement uncertainty of
maintain a constant 18 bar pressure. This is because the seal
the liquid flow rate. The primary situation where this could be
purging system is allowing mass to flow into the compressor
a concern is for control of liquid flow at very low LVFs.
(instead of just creating a buffer). The liquid level in the
Temperature measurements can be made on a wet system
separator must continually decrease to keep adding volume to
with conventional temperature sensors. The primary thing to
the test loop to allow space for the mass of the purge gas being
note about temperature measurement is that the location of the
added to the loop. This explanation may seem cumbersome,
sensing element can influence the temperature reading. For
but this control methodology was used several times in past
example, the liquid may be stratified flow along the bottom of
testing to maintain a fixed pressure for a testing.
the pipe; whereby, the temperature sensing element placed at
Pressure control is also important for start-up and
the bottom of the pipe will be reading liquid temperature while
shutdown. During start-up of the test loop, it may be necessary

other temperature sensors may be reading gas temperature at neglected in this paper because the no-interaction models were
the top of the pipe. sufficient to predict the output variables.
GENERAL TRENDS OBSERVED IN WET GAS First-Order Regression Model:

PERFORMANCE TESTING 𝑦 = 𝛽0 + 𝛽1 𝑥1 + 𝛽2 𝑥2 + ⋯
From the available wet gas compressor performance data
presented in the literature, there are consistent trends showing Second-Order Regression Model:
that as more liquid is added to the compressor, the compressor 𝑦 = 𝛽0 + 𝛽1 𝑥1 + 𝛽2 𝑥1 2 + 𝛽3 𝑥2 + 𝛽4 𝑥2 2 + ⋯
power increases (reduced efficiency), pressure ratio increases,
and gas volume flow rate decreases. It is important to note, Using the performance data described in the previous
however, that the effect of wet gas on pressure ratio is section, first- and second-order regression models are
dependent on the operating condition because the pressure presented to determine which test parameters most affect
ratio may not be affected at high volume flow rates. The test compressor performance. The results of this analysis can be
data used to illustrate the performance trends are from a past used to guide future test programs. The test parameters
experimental program completed at SwRI [1, 2], as illustrated considered for the regression models are listed in Table 3 with
in Figure 8. their associated order of magnitude. Because of the large
difference in orders of magnitude among the test parameters,
the variable coefficients of a regression model cannot be used
to compare the relative importance of the variable. In other
words, the regression model built for compressor pressure
ratio may have a very low variable coefficient (β) for
compressor speed to counter the large order of magnitude of
the speed values. As an alternative, the test parameter values
can be expressed as a value in terms of standard deviation
above or below the mean value. The result is a set of test
parameters that have similar order of magnitude – which
translates to improved accuracy of the regression model, and a
set of variable coefficients that can be compared to show
relative importance. The model outputs are also standardized
𝑥1 − 𝑥̅1 2
𝑥́ 1 = � �
𝑠1
Figure 8. Wet Gas Test Data Set [2] In presenting the regression analysis, the adequacy of the
regression model fit is first verified for both dry and wet
Effect of wet gas operating parameters on performance
compressor test data. Next, the group of test parameters
When defining a test matrix for wet gas compression
tests, it is easy to get lost in the test variables that may or may included in the model is investigated to determine if some
not influence the compressor performance. Much of the past parameters have no effect on the model fit. Finally, the
testing in the literature focus on density ratio, LVF, gas variable coefficients are presented to show the relative
volume flow, suction pressure, and suction temperature. importance of the test parameters used to generate the final
regression model. Using the complete set of standardized test
However, it is possible that other variables may have a
significant influence on the compressor operation, such as parameters, both first-order and second-order regression
those that were thought to be insignificant when developing models are fit to the data to determine which regression form
the test matrix. For example, test variables of LMF and air- is required for the model fit. A regression model is generated
water temperature difference can sometimes take a backseat to for each test output; compressor pressure ratio, isentropic
efficiency, shaft torque, and axial thrust. The initial check of
variables of LVF and suction air temperature. To determine
the regression model adequacy is done by comparing the
the influence of test variables on compressor performance, a
multiple regression model is used here to describe which predicted output values to the test data, as shown in Figure 9
variables should or should not be neglected. and Figure 10. For a perfect model fit, all plotted points would
A multiple regression model is a method to estimate a set lie on the line with slope of 1. From Figure 9, it can be seen
of outputs for a given set of inputs. In the simplest form, a that the first-order regression model is sufficient to predict all
the test output variables except isentropic efficiency, where a
regression model can be thought of as a linear curve fit.
large amount of scatter is seen from the line of perfect fit. In
𝑦 = 𝛽0 + 𝛽1 𝑥1
comparison, the second-order regression model plotted in
In this simple form, the input is x1 and the output is y. The
Figure 10 shows even better agreement than the first-order
magnitude coefficient (β0) and the variable coefficient (β1) are
model for all output variables. With the second-order model,
results of the regression analysis. When many inputs are
the pressure ratio, torque, and axial thrust variables are much
monitored for a single output, a multiple regression model is
better predicted than the first-order model, and the compressor
expanded to include variable coefficients for each input.
efficiency is predicted within an acceptable amount of scatter.
Generally, a multiple regression model can take the form of a
The fit of the regression models is quantitatively compared
first-order or second-order equation. The effect of variable
using the R2 error estimate in Table 4, adjusted for the number
interaction can be included in the regression model, but is
of input parameters. An R2 value of 1.0 indicates a perfect fit

of the model to the test data, while an R2 value of 0 indicates a suited for wet data than dry data may be due to the influence
poor fit. It can be seen from the comparison of regression of the liquid parameters damping the effects of the dry
models that the second-order model fits both dry and wet parameters. In any case, the second-order model is determined
performance data very well, while the first-order model was be sufficient for this analysis and outperforms the fit of the
more successful to predict wet performance data than dry first-order model which was insufficient to predict the
performance data. The reason the first-order model was better isentropic efficiency of the dry data.
Table 3.
Test Parameters and Order of Magnitude
Test Variable Units Order of Range of Values
Magnitude
Speed [rpm] 103 8,000 – 11,000
Suction Pressure [bar] 101 19 – 20
Suction Air Temperature [°C] 101 42 – 64
Air-Water T [°C] 100 -8 – 10
Air Volume Flow [ACMH] 102 - 103 460 – 1160
Water Volume Flow [gpm] 101 - 102 0.6 – 5.8
Water-Air Density Ratio [-] 101 45 – 50
LVF [-] 10-2 0.3 - 3.0%
LMF [-] 10-1 12 – 60%
2.5 2.5
Perfect Fit Perfect Fit
2 Dry Regression Model 2 Dry Regression Model
Wet Regression Model Wet Regression Model
1.5 Measured Isentropic Efficiency 1.5
Measured Compressor PR
1 1
0.5 0.5
0 0
-0.5 -0.5
-1 -1
-1.5 -1.5
-2 -2
-2.5 -2.5
-2.5 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 2.5 -2.5 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 2.5
Predicted Compressor PR Predicted Isentropic Efficency
2.5 2.5
1.5 1.5
1 1
Measured Shaft Torque
Measured Axial Thrust
0.5 0.5
0 0
-0.5 -0.5
-1 -1
-1.5 -1.5
-2 -2
-2.5 -2.5
-2.5 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 2.5 -2.5 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 2.5
Predicted Shaft Torque Predicted Axial Thrust
Figure 9. Comparison of First Order Regression Models to Measured Data

(Regression Variables: N, Psuc, Tsuc, Qair, ∆Tair-water, LMFnom, LVFnom)

2.5 2.5
1.5 1.5
Measured Isentropic Efficiency

Measured Compressor PR
1 1
0.5 0.5
0 0
-0.5 -0.5
-1 -1
-1.5 -1.5
-2 -2
-2.5 -2.5
-2.5 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 2.5 -2.5 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 2.5
Predicted Compressor PR Predicted Isentropic Efficency
2.5 2.5
1.5 1.5
1 1
Measured Shaft Torque
Measured Axial Thrust

0.5 0.5
0 0
-0.5 -0.5
-1 -1
-1.5 -1.5
-2 -2
-2.5 -2.5
-2.5 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 2.5 -2.5 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 2.5
Predicted Shaft Torque Predicted Axial Thrust
Figure 10. Comparison of 2nd Order Regression Models to Measured Data

(Regression Variables: N, Psuc, Tsuc, Qair, ∆Tair-water, LMFnom, LVFnom)
Table 4. Regression Model R2 Goodness of Fit
Axial Shaft Compressor Isentropic
Thrust Torque PR Efficiency
1st Order R2 Fit 0.95 0.89 0.89 0.35
Dry Regression Model nd 2
2 Order R Fit 1.00 1.00 1.00 0.96
st 2
1 Order R Fit 0.92 0.98 0.98 0.92
Wet Regression Model nd 2
2 Order R Fit 0.98 1.00 1.00 0.96
Using the second-order regression model, the set of volume flow of the wet operating point. Nominal values of
inputs was investigated to determine which of them were not LMF and LVF are calculated using the gas volume flow of the
required to adequately predict compressor performance. A corresponding dry operating point. The purpose of
total of 23 variations of the input set were investigated to investigating nominal values lies in the simplicity of selecting
determine the minimum set required. The adequacy of each operating conditions for future wet gas testing. The second-
variable set is determined using the R2 fit of the model to the order model presented above is shown in Figure 11 as the first
measured test data, as shown in Figure 11. A primary goal of variable set that includes both nominal and actual values of
defining the variable set is to determine the importance of LMF and LVF. Moving from left to right on Figure 11 each
LVF and LMF on the model output. The motivation for this variable set is reduced by one variable at a time. The influence
exercise is that much of the literature uses LVF as a primary of using the actual and nominal values of LMF and LVF is
test variable for elevated pressure tests, and LMF as a primary seen to affect only the isentropic efficiency and axial thrust by
test variable for atmospheric pressure tests. The goal is to as much as one percent of the R2 fit. The air-water density
determine if one is more important than the other, or if both ratio was found to not affect the regression model fit, likely
should be considered. The horizontal axis of Figure 11 is because the air-water density ratio is included in the
generated to show the set of variables that were combined calculation of LMF. Removing the water volume flow rate
with the noted LVF or LMF variables. For this data, both from the variable set had a small effect on the model fit;
nominal and actual LMF and LVF values are compared. however, including LMF and LVF together results in a
Actual values of LMF and LVF are calculated using the gas regression model with similar fit to larger variable sets.

Further reduction of the variable set to remove the air-water • Speed • Air Volume Flow Rate
temperature difference and air suction temperature results in a • LMF • Air-Water Temperature Difference
significant reduction of the model fit to the output variables. • LVF • Suction Pressure
Therefore, the smallest set of input variables found to produce
• Suction Temperature
a regression model consistent with larger variable sets is to
include the following variables:
1.00
0.99
0.98
0.97
Regression 0.96
Model2
R Fit 0.95
R2 Fit
0.94
Compressor PR
0.93
Shaft Torque
0.92
Axial Thrust
0.91 Isentropic Efficiency
0.90
Regression Model Iteration (2nd Order)
Speed Speed Speed Speed

Suction Pressure Suction Pressure Suction Pressure Suction Pressure
Suction Temperature Suction Temperature Suction Temperature Suction Temperature
Air Volume Flow Rate Air Volume Flow Rate Air Volume Flow Rate Air Volume Flow Rate
Air-Water ∆T Air-Water ∆T Air-Water ∆T
Speed
Water Volume Flow Rate Water Volume Flow Rate
Suction Pressure
Air-Water Density Ratio
Air Volume Flow Rate
Figure 11. Sensitivity of 2nd Order Wet Regression Model to Selection of Variables
As expected, the water flow rate and air-water density ratio are 20
not required to define an input variable set as long as LMF and 18
LVF are included to account for the density ratio and water
16
flow rate. The unexpected result, however, is that the
temperature difference between the air and water has a major 14
effect on the model fit. This suggests that air and water 12
temperature differences are important for wet gas Bin
10
Frequency
compression. Considering the thermodynamics of 8
compression, it would be expected that the inlet temperature
6
difference would be important. However, the temperature
difference has not been thoroughly studied in the literature, 4
and the amount of temperature difference required to 2
significantly affect compressor performance has not been 0
quantified. From the test data set, the maximum temperature
10
-10
0
1
2
3
4
5
6
7
8
9
-9
-8
-7
-6
-5
-4
-3
-2
-1
difference was less than 10°C; whereby most data points had a Air-Water Temperature Difference [°C]
temperature difference less than 2°C, as shown in Figure 12.
Because only a few degrees of temperature difference were Figure 12. Histogram of Air-Water Temperature Difference
measured for most data points, the air-water temperature for all Wet Gas Data Points
difference was not expected to be significant for the regression
model.

The second-order regression model obtained through the more affected by temperature than pressure change, within the
study of variable sets is used to illustrate the relative range of conditions of the test data set. Similar to pressure
importance of test variables. The variable coefficients of the ratio, axial thrust is most influenced by speed and air volume
regression model are normalized by the maximum variable flow. Axial thrust is similar to pressure ratio because the axial
coefficient and shown in Figure 13 and Figure 14 for dry and thrust of the machine is strongly tied to pressure ratio across
wet data sets, respectively. While a regression model for dry the impeller. Lastly, isentropic efficiency is affected most by
gas compressor performance is not of significant importance suction temperature and volume flow, but is shown to have
because dry gas compressor performance is well understood, it strong secondary effects of suction pressure and speed.
is a good check of the regression model results to ensure that Isentropic efficiency has strong influences from most of the
the important parameters make sense. From the dry gas test variables because the efficiency calculation incorporates
regression model, the results make sense that the largest values of pressure ratio and shaft torque, making it the most
variable coefficients correspond to the expected important test complex of the output variables. It is also noteworthy that out
variables. The large variable coefficients for speed and air of the second-order terms in the dry regression model, speed
volume flow are shown to be the most significant for and volume flow are the most significant likely because of
predicting compressor pressure ratio, as expected. Shaft torque their second-order relationship compressor operation that is
is most influenced by suction temperature and speed. Suction evident in typical compressor performance maps
temperature is important to torque because the air density
1.0
0.9 Compressor PR
0.8 Shaft Torque
0.7 Axial Thrust
Isentropic Efficiency
Normalized 0.6
Regression 0.5
Coefficient 0.4
0.3
0.2
0.1
0.0
Figure 13. Normalized Regression Coefficients for 2nd Order Dry Model
1.0
0.9 Compressor PR
0.8 Shaft Torque
0.7 Axial Thrust
Normalized 0.6
Regression 0.5
Coefficient 0.4
0.3
0.2
0.1
0.0
Figure 14. Normalized Regression Coefficients for 2nd Order Wet Model

From the comparison of R2 fit results, it was found that gas model, however, isentropic efficiency is most strongly
using either the actual or nominal values of LMF and LVF affected by the LMF and LVF nearly equally. The other
were acceptable to generate a regression model as long as both operating variables of the compressor do not have as strong an
LMF and LVF were included in the model. Therefore, the influence on efficiency as the LMF and LVF. It is important to
regression model using actual values of LMF and LVF is realize that variables such as suction pressure and suction
presented in Figure 14 to be consistent with the available wet temperature may have an effect on isentropic efficiency, but
gas literature. Similar to the dry gas regression model, wet gas the effects are not shown because the range of test data from
pressure ratio is strongly influenced by speed, and air volume which the regression model is created does not have a large
flow rate. However, pressure ratio is also most strongly enough variable range to show the influence.To summarize,
affected by the LMF and LVF, which is evident in test data the sensitivity study of test parameters on wet gas compressor
showing the increase of compressor pressure ratio as liquid is performance uses a multiple regression model to predict the
added to the suction flow. Shaft torque was found to have compressor performance using a set of test data. The
influential variables of speed and air volume rate similar to the regression model fit to the measured data was checked using
dry regression, but a strong influence from LMF. The variable an R2 fit and sequentially reducing the number of test
LVF was found to only have a little more influence on shaft variables to find the smallest data set possible without
torque than the influence of the air-water temperature significantly reducing the model accuracy. The influential
difference. The small effect of LVF on shaft torque suggests parameters determined from the regression models are shown
the strong effect of density ratio on compressor power, which in Table 5 for both dry gas and wet gas performance. Overall,
has been documented in test data from the literature. In fact, it was found that the most influential test variables for wet gas
the dependence of LMF on power is demonstrated in performance are the compressor speed, air volume flow rate,
atmospheric test data that measure a significant effect of wet LMF, and LVF. It was found that including both LMF and
gas on compressor power when the LVF values are nearly LVF provided better accuracy of the regression model, which
zero due to the density ratio of 1,000 to 1. As with the dry gas indicates that future wet gas testing should consider both LMF
model, the parameters influencing axial thrust are similar to and LVF ranges in the test matrix. It is important to emphasis
those influencing pressure ratio; however, axial thrust is not that the sensitivity analysis presented here is specific to the
found to be influenced by LVF. The lack of influence of LVF compressor and range of test variables from the data set. For
on axial thrust may suggest that axial thrust effects are example, all measurements were conducted at nearly constant
strongly governed by the density ratio between the gas and suction pressure; therefore, the effect of suction pressure is not
liquid affecting the secondary flow through the compressor. seen as a driving factor in this study. Future regression models
The isentropic efficiency was found for the dry gas model to that include a range of suction pressures will likely show it as
be influenced by all the operating parameters, most notably an influential parameter.
the suction temperature and air volume flow rate. For the wet
Table 5. Summary of Influential Test Variables
Regression Model Output Influential Variables for Influential Variables for
Dry Gas Compression Wet Gas Compression
Speed Speed
Air Volume Flow Rate Air Volume Flow Rate
Pressure Ratio
LMF actual
LVF actual
Speed Speed
Shaft Torque Suction Temperature Air Volume Flow Rate
LMF actual
Speed Speed
Axial Thrust Air Volume Flow Rate Air Volume Flow Rate
LMF actual
Speed LMF actual
Air Volume Flow Rate LVF actual
Suction Pressure
Suction Temperature
that has been done to characterize the thermodynamic effect of

FUTURE WET GAS WORK wet gas, it is obvious that the aerodynamic effect has still not
Analytical been quantified. The major assumption of the thermodynamic
Analytical methods to predict wet gas performance models is that the liquid is in the form of disperse, spherical
continue on the work of others to define wet gas performance droplets that follow the gas flow through the compressor. This
using thermodynamic models or CFD. In either case, assumption is valid for axial compressors in power generation
analytical work continues to be based on flange-to-flange applications because the liquid is injected as a fog with droplet
compressor performance measurements. Considering the work diameters on the order of 10µm. In Oil & Gas applications,

however, the liquid enters the compressor with a distribution droplets – additional corrections are required for analytical
that is determined by the natural two-phase flow regime in the models to account for heat transfer between the gas and liquid
suction piping upstream of the compressor. The liquid entering phases, as well as the effect of the liquid on the gas flow
the compressor is made up of droplets, ligaments, and films through the compressor. For example, the liquid film on the
covering the flow path surfaces. An example of the liquid impeller blade surface will affect the boundary layer
distribution for wet gas flow over a blade is illustrated in development to potentially reduce aerodynamic performance.
Figure 15. Because the liquid is not dispersed, spherical
Figure 15. Potential Influence of Wet Gas: (left) Liquid Droplets Leading to Premature Flow Separation or (Right) Liquid Film
Significantly Increasing Surface Roughness [39]
The interaction between the gas and liquid for non-ideal compression, much of the research is dis-jointed among a
liquid conditions is not a simple task, and this is a growing range of compressor designs and operating conditions such
area of interest to improve wet gas predictions. Boundary that a clear picture of wet gas effects is not possible. As wet
layer analysis is quite complex to estimate the flow near the gas compressor prototypes are being designed for sub-sea
air and water interface [40] when trying to model the natural service, more experimental work will be needed to determine
physics. Computational methods are more attractive, but still the effect of flow path design features on compressor
rely on empirical models for the interaction of the phases. To performance. To that end, a harmonized test campaign
resolve the complex interactions between gas and liquid evaluating the performance and durability of multiple designs
phases, computational cost is high due to the number of in the same test facility can provide directly comparable test
equations to solve the momentum, energy, and turbulence data. While aerodynamic performance data is still needed for
equations of both gas and liquid phases. Additional equations future compressor designs, future experimental work will
are included to account for the interfacial communication likely begin investigating rotor-dynamic, and durability of the
between the fluids. Furthermore, resolving the fluid inter- machine.
phase requires a very fine mesh in regions where the gas and
liquid interface is to occur. At this time, CFD solutions for wet CONCLUSIONS
gas compression are being done, but information is limited on Test experience from the authors and a review of relevant
their accuracy and computational efficiency. Because of the literature on the subject of wet gas compression has been
complexity of modeling the multiphase flow while accounting presented for Oil & Gas applications with centrifugal
for the natural physics of the gas-liquid interface, SwRI has compressors. Considerations for experimental testing have
recently been investigating the use of an alternative been presented to discuss a typical test setup in detail and
computational flow solver using a lattice-Boltzmann (LB) provide insight for selection of loop hardware and control. An
method. Effectively, the LB method arrives at the Navier overview of analytical work in wet gas compression was
Stokes equations while utilizing a mesoscopic method for presented to show the current state-of-the-art in the literature.
fluid predictions. Advantages of the LB method are that it is A regression analysis was presented to show the significant
implicitly transient, does not require a discretized mesh, and test variables observed during a wet gas test campaign.
can account for multiphase flows with a few additional Important test variables were found to be dependent on the
equations. Through internal research work at SwRI, the LB measurement being made; however, liquid mass fraction was
method shows promise, but requires more work to accurately found to be important for measuring compressor pressure
predict wet gas aerodynamics. ratio, efficiency, torque, and thrust. Liquid volume fraction
Experimental was found to be important for measuring pressure ratio and
As part of sub-sea technology development, compressor efficiency, which suggests the importance of future tests to use
performance and operation at sub-sea conditions are being fluids that allow for gas-liquid density ratios similar to actual
actively researched at a handful of locations throughout the operation. From the regression analysis and the overview of
world. Although wet gas compressor test results have been experimental and analytical, suggestions for wet gas work
published in the literature for nearly a decade, the Oil & Gas were presented to call attention to areas of wet gas research
industry still does not have enough information to confidently that require further study.
predict wet gas performance, mechanical operation, or
machine durability for all conditions. While the available wet NOMENCLATURE
gas testing has revealed important information on wet gas LMF = Liquid mass fraction, ṁℓ/(ṁℓ+ṁg) (-)

LVF = Liquid volume fraction, Qℓ/(Qℓ+Qg) (-) safeguard Asgard flow Assurance,”Offshore, August
N = Rotor speed (rpm) 2008.
PR = Pressure ratio (-) [18] “Gas Compression from Production thru
Psuc = Suction pressure (bara) Transmission,” Presented at Gas/Electric Partnership,
Tsuc = Suction temperature (°C) Houston, TX February 2010.
Qair = Volume flow rate or air (m3/hr) [19] Fabbrizzi, M., et. al., “An Experimental Investigation
∆Tair-water = Air-water temperature difference (°C) of a Single Stage Wet Gas Centrifugal Compressor,”
Proceedings of the TurboExpo Power for Land, Sea
REFERENCES and Air, GT2009-59548, 2009.
[1] Ransom, D., Podestà, L., Camatti, M., Wilcox, M., [20] ASME, “Performance Test Code on Compressors and
Bertoneri, M., Bigi, M., “Mechanical Performance of Exhausters,”Standard PTC 10-1997.
a Two Stage Centrifugal Compressor under Wet Gas [21] Schultz, J.M., “The Polytropic Analysis of
Conditions,” Proceedings of the 40th Turbomachinery Centrifugal Compressors,” J. Engineering Power, pp.
Symposium, pp. 121-128, 2011. 69-82, 1962.
[2] Bertoneri, M., Duni, S., Ransom, D., Podestà, L., [22] Huntington, R.A., “Evaluation of Polytropic
Camatti, M., Bigi, M., Wilcox, M., “Measured Calculation Methods for Turbomachinery
Performance of Two-Stage Centrifugal Compressor Performance,” ASME Turbo Expo, 85-GT-13.
under Wet Gas Conditions,” Proceedings of ASME [23] Hundseid, O., Bakken, L.E., Helde, T., “A Revised
Turbo Expo, GT2012-69819, 2012. Compressor Polytropic Performance Analysis,”
[3] Brenne, L., et. al., “Performance Evaluation of a ASME Turbo Expo, GT2006-91033
Centrifugal Compressor Operating Under Wet Gas [24] Hundseid, O., Bakken, L.E., “Wet Gas Performance
Conditions,” Proceedings of the Thirty-Fourth Analysis,” ASME Turbo Expo, GT2006-91035.
Turbomachinery Symposium, Houston, TX, 2005. [25] Härtel, C., Pfeiffer, P., “Model Analysis of High-
[4] Knudsen, T., Solvik, N., “World First Submerged Fogging Effects on the Work of Compression,”
Testing of Subsea Wet Gas Compressor,” ASME Turbo Expo, GT2003-38117.
Proceedings of the Offshore Technology Conference, [26] Horlock, J.H., “Compressor Performance with Water
Houston, TX, 2011. Injection,” ASME Turbo Expo, GT2001-0343.
[5] Camatti, M., “Design and Technology Qualification [27] White, A.J., Meacock, A.J., “An Evaluation of the
of “BlueC” Subsea Gas Compressor for Ormen Effects of Water Injection on Compressor
Lange,” Presentation at the Underwater Technology Performance,” ASME Turbo Expo, GT2003-38237.
Conference, Bergen, Norway, 2011. [28] White, A.J., Meacock, A.J., “Wet Compression
[6] Aguilera, L.C.P., “Subsea Wet Gas Compressor Analysis Including Velocity Slip Effects,” J.
Dynamics,” Masters Thesis, Norwegian University of Engineering Gas Turbines Power, 133, 2011.
Science and Technology, June 2013. [29] Abdelwahab, A., “An Investigation of the Use of Wet
[7] Torpe, H., “Separation and Related Aspects for Compression in Industrial Centrifugal Compressors,”
Asgard Subsea Compression,” Presentation at the ASME Turbo Expo, GT2006-90695.
Separation Technology Conference, Stavanger, [30] Zheng, Q., Shao, Y., Zhang, Y., “Numerical
Norway, 2013. Simulation of Aerodynamic Performances of Wet
[8] Bjorge, T., Brenne, L., “Subsea gas compression; Compression Compressor Cascade,” ASME Turbo
Available technology and future needs,” The Expo, GT2006-91125.
Research Council of Norway, 2010. [31] Luo, M., Zheng, Q., Sun, L., Deng, Q., Yang, J.,
[9] “Press Kit Asgard Subsea Gas Compression,” Statoil “Numerical Simulation of an Eight-Stage Axial
website, March 9, 2014 Subsonic Compressor with Wet Compression,”
[10] Arntzen, R., Winther, L., “Ormen Lange, Subsea ASME Turbo Expo, GT2013-94486.
Compression Project,” Shell, 2012. [32] Khan, J.R., “Comparison between Discrete Phase
[11] “Enhanced Recovery through Subsea Compression Model and Multiphase Model for Wet Compression,”
at Gullfaks,” Statoil News Release, May 22, 2012. ASME Turbo Expo, GT2013-96022.
[12] Amundsen, S., “Wet Gas Compression, Impeller [33] Grüner, T.G., Bakken, L.E., Brenne, L., Bjorge, T.,
Rig,” Masters Thesis, Norwegian University of “An Experimental Investigation of Airfoil
Science and Technology, July 2009. Performance in Wet Gas Flow,” ASME Turbo Expo,
[13] Jellum, M., “Wet Gas Compressor Surge Detection,” GT2008-50483.
Masters Thesis, Norwegian University of Science and [34] Gilarranz, J.L, Kidd, H.A., Chochua, G., Maier,
Technology, June 2013. W.C., “An Approach to Compact, Wet Gas
[14] Vintersto, T., “Subsea Gas Compression Now and in Compression,” ASME Turbo Expo, GT2010-23447.
the Future,” Energy Claims Conference, 2013. [35] Krzyzanowski, K., “The Correlation Between
[15] Rogno, H., “Statoils subsea relevante FoU,” NC Droplet Steam Structure and Steam Turbine Blading
Subsea, Bergen, 2013. Erosion,” Journal of Engineering Gas Turbines
[16] “Compressor Testing Kicks Off at StatoilHydro’s Power, vol. 96, pp. 256-266, July 1974.
Karsto Laboratory,” OilVoice, April 21, 2008.
[17] Terdre, N., “Subsea compression under review to

[36] Luo, S., et. al. “Theoretical Model for Drop and
Bubble Breakup in Turbulent Dispersions,” AIChE
Journal, vol. 45, pp. 1225-1233, 1996.
[37] Joliffe, K., “The Development of Erosion Damage in
Metals by Repeated Liquid Droplet Impacts,”
Proceedings of the Royal Society of London – Series
A, Mathematical and Physical Sciences, vol. 303, pp.
193-205, February 1968.
[38] Pacheco, J., et. al., “Assessing Liquid Droplet
Erosion Potential in Centrifugal Compressor
Impellers,” Proceedings of the Thirty-Eighth
Turbomachinery Symposium, Houston, TX, 2009.
[39] Musgrove, G.O, Ransom, D.L., “Removal of Liquid
from Airfoil of Equipment Have Gas-Liquid Flows,”
US Patent Application 14/092976, 2013.
[40] Nelson, J.J., Alving, A.E., Joseph, D.D., “Boundary
Layer Flow of Air over Water on a Flat Plate,” J.
Fluid Mech, 284, pp. 159-169, 1995.
ACKNOWLEDGEMENTS
The authors wish to thank GE Oil & Gas for their support
and allowing the use of test data in this paper. The authors
also thank David Ransom at Southwest Research Institute for
supporting this paper.

43 Turbomachinery & 30 Pump Users Symposia (Pump & Turbo 2014) September 23-25, 2014 - Houston, TX - Pumpturbo - Tamu.edu

Uploaded by

Document Informationclick to expand document informationgas

Document Informationclick to expand document information

Copyright:

Available Formats

43 Turbomachinery & 30 Pump Users Symposia (Pump & Turbo 2014) September 23-25, 2014 - Houston, TX - Pumpturbo - Tamu.edu

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

43 Turbomachinery & 30 Pump Users Symposia (Pump & Turbo 2014) September 23-25, 2014 - Houston, TX - Pumpturbo - Tamu.edu

Uploaded by

Copyright:

Available Formats

View metadata, citation and similar papers at core.ac.

uk brought to you by CORE

OVERVIEW OF IMPORTANT CONSIDERATIONS IN WET GAS COMPRESSION TESTING AND ANALYSIS

Grant O. Musgrove Melissa A. Poerner

Massimiliano Cirri Matteo Bertoneri

Copyright© 2014 by Turbomachinery Laboratory, Texas A&M Engineering Experiment Station

Copyright© 2014 by Turbomachinery Laboratory, Texas A&M Engineering Experiment Station

Copyright© 2014 by Turbomachinery Laboratory, Texas A&M Engineering Experiment Station

Copyright© 2014 by Turbomachinery Laboratory, Texas A&M Engineering Experiment Station

Copyright© 2014 by Turbomachinery Laboratory, Texas A&M Engineering Experiment Station

Copyright© 2014 by Turbomachinery Laboratory, Texas A&M Engineering Experiment Station

Figure 6. Injector Nozzle Droplet Diameter, Flow Rate, and

Copyright© 2014 by Turbomachinery Laboratory, Texas A&M Engineering Experiment Station

Copyright© 2014 by Turbomachinery Laboratory, Texas A&M Engineering Experiment Station

Copyright© 2014 by Turbomachinery Laboratory, Texas A&M Engineering Experiment Station

GENERAL TRENDS OBSERVED IN WET GAS First-Order Regression Model:

Copyright© 2014 by Turbomachinery Laboratory, Texas A&M Engineering Experiment Station

Measured Axial Thrust

Figure 9. Comparison of First Order Regression Models to Measured Data

Copyright© 2014 by Turbomachinery Laboratory, Texas A&M Engineering Experiment Station

Measured Isentropic Efficiency

Measured Axial Thrust

Figure 10. Comparison of 2nd Order Regression Models to Measured Data

Copyright© 2014 by Turbomachinery Laboratory, Texas A&M Engineering Experiment Station

Regression Model Iteration (2nd Order)

Speed Speed Speed Speed

Copyright© 2014 by Turbomachinery Laboratory, Texas A&M Engineering Experiment Station

Copyright© 2014 by Turbomachinery Laboratory, Texas A&M Engineering Experiment Station

that has been done to characterize the thermodynamic effect of

Copyright© 2014 by Turbomachinery Laboratory, Texas A&M Engineering Experiment Station

Copyright© 2014 by Turbomachinery Laboratory, Texas A&M Engineering Experiment Station

Copyright© 2014 by Turbomachinery Laboratory, Texas A&M Engineering Experiment Station

Copyright© 2014 by Turbomachinery Laboratory, Texas A&M Engineering Experiment Station

You might also like