Version of Record: https://www.sciencedirect.

com/science/article/pii/S0896844620303235
Manuscript_06382d4256dee44fe691cbcc8e09478c

Volumetric efficiency and pressure pulsations of a triplex reciprocating

supercritical CO2 pump

Shaun R. Aakre*, Mark H. Anderson, Paul W. Brooks

Department of Mechanical Engineering. University of Wisconsin – Madison, 1500 Engineering Drive, Madison, WI 53705

Abstract

A reciprocating triplex pump has been integrated into a closed-loop supercritical carbon dioxide test facility. The

pump is designed to compress liquid or supercritical carbon dioxide and has a maximum volumetric flow rate of 51 L/min. A

sweep across the pump’s speed range of 100-1000 rev/min was performed with inlet conditions between 60 and 160 bar. The

volumetric efficiency was found to increase as pump speed increased from 10 to 40% before it plateaued at higher pump speeds.

The mechanical efficiency increased linearly with pump speed and was not affected by system pressure. Pressure pulsations

caused by the plungers and valves were also measured at each speed. A 35-liter buffer volume, connected near the pump’s

discharge was isolated at each speed to determine its effectiveness as a pulsation dampener. Overall the reciprocating pump

was shown to perform exceptionally with liquid or supercritical inlet conditions applicable to many supercritical fluid

applications.

© 2020 published by Elsevier. This manuscript is made available under the Elsevier user license
https://www.elsevier.com/open-access/userlicense/1.0/
Nomenclature

, Volumetric or mechanical efficiency uncertainty

Total error in measurement of variable X

Discharge density [kg/m3]

Mechanical efficiency

Volumetric efficiency

Standard deviation of signal for variable X

Δ Measurement uncertainty for variable X

Δ Total pump pressure differential [bar]

Speed of sound [m/s]

ℓ Effective length of resonator nozzle [m]

Mass flow rate [kg/s]

Measured pump speed [rev/min]

Area of resonator inlet nozzle [m2]

Τ Torque input [N-m]

Ideal volumetric displacement [m3/rev]

Volume [m]

2
1. Introduction

The use of supercritical carbon dioxide (sCO2) as a solvent and working fluid has spread into many industries. Demand

for more environmentally-friendly processes has supported its use in chemical extraction, dry-cleaning, refrigeration, and in

the food and beverage industry [1–3]. The properties of carbon dioxide near its critical pressure of 71 bar (1080 psi) enable

enhanced extraction and separation processes compared to other solvents. The petrochemical industry has employed sCO2 for

enhanced oil and natural gas recovery because of its abundance and its ability to penetrate porous geological layers. More

recently, efforts to employ supercritical CO2 as a working fluid for electricity generation have entered the demonstration phase

of development [4,5]. This power cycle has the potential for higher conversion efficiency using smaller components than equal-

capacity steam power plants because of high fluid density and volumetric heat capacity. Finally, growing concerns over the

environmental impacts of increasing atmospheric CO2 concentrations will demand reliable equipment to capture, compress,

and store anthropogenic CO2 emissions.

All of these applications require highly efficient and reliable compression machinery to become economically feasible.

Inlet and differential pressure, mass flow rate, efficiency, and leakage requirements will guide the selection of compressor or

pump configuration. Commercially-available options for supercritical CO2 are limited to centrifugal compressors, diaphragm,

and reciprocating plunger pumps [6]. Reciprocating pumps can provide high differential pressures and operate at pressures

greater than 75 MPa (10,900 psi). These characteristics have led to the widespread use of reciprocating pumps for high

performance liquid chromatography (HPLC) and commercial-scale enhanced oil recovery operations. Most HPLC pumps can

provide relatively small CO2 flow rates (10-30 kg/hr), while reciprocating pumps capable of up to 40,000 kg/hr have been

requested from commercial vendors [7]. Some of the largest have been constructed for oil recovery operations but most details

regarding these designs remain proprietary to protect the manufacturers’ and users’ competitive advantage.

Reciprocating pumps typically operate at lower rotational speeds than centrifugal compressors, limiting their flow rates to

small applications such as batch extractions or development-scale CO2 component testing. Furthermore, compressing CO2 near

its critical pressure (74 bar) with a reciprocating machine requires low inlet temperatures (< 30°C) to increase fluid density and

decrease compressibility. Despite their commercial utility, there is little data available in the literature on the performance of

reciprocating pumps operating with supercritical CO2 at variable speed and over a range of inlet pressures. Designers of

supercritical CO2 systems in chemical extraction plants or power plant concepts require information regarding the volumetric

and mechanical efficiencies and operating characteristics (e.g., pressure pulsations, and valve dynamics) associated with

compression machinery to maximize the performance of their facilities.

3
 The pump described in this paper was purchased by the Thermal Hydraulics Laboratory at the University of Wisconsin-

Madison to study the hydraulic and thermal performance of compact heat exchangers. These heat exchangers represent a

technical challenge and opportunity for electricity-generation power cycles using supercritical water or CO2. The supercritical

CO2 (sCO2) Brayton cycle has gained the interest of natural gas, solar, and nuclear energy developers by offering the potential

for higher temperature operation and smaller turbomachinery compared to traditional open-air or closed-steam power cycles

[4]. The construction of the test loop at Wisconsin was sponsored by the United States Department of Energy to investigate

unique heat exchanger designs which could couple a supercritical CO2 power cycle to an advanced nuclear reactor. Additional

information on the thermal and mechanical performance of these components is required to quantify and improve their

performance under high-temperature (>500°C) conditions. It is also necessary to inform fabrication and inspection codes, such

as the ASME Boiler and Pressure Vessel Code (BPVC), with material strength data and design methodologies that are suitable

for these heat exchangers. More background on the Integrated Research Project encompassing this work and investigation of

the thermal-mechanical performance of several sCO2 heat exchanger is provided elsewhere [8-11].

Since the performance of the pump will directly impact the scope of heat exchanger testing and potentially the results, the

authors measured the pump’s performance to supplement the manufacturer’s efficiency data which was measured using room-

temperature water. The volumetric efficiency measurements were performed in conjunction with pressure loss measurements

across two diffusion-bonded heat exchangers. Since no tuned pulsation dampener is employed on the discharge piping, pressure

pulsations in the pump’s suction and discharge manifolds were recorded. These measurements served to determine the

effectiveness of the system’s buffer volume to act as a discharge dampener. The magnitude of these pulsations is recorded at

each operating condition to examine their effect on pressure drop measurements across the system’s heat exchangers as well.

2. Experimental sCO2 Loop

The Wisconsin supercritical CO2 (WisCO2) loop consists of a fixed-displacement, three-plunger (triplex) reciprocating

pump and three diffusion-bonded heat exchangers (DBHE). A line schematic of the WisCO2 system is shown in Figure 1. The

main goal of the loop is to quantify the pressure drop and heat transfer performance of the Primary HX and Recuperator shown

in Fig. 1. In the configuration shown, the Primary HX simulates the heat exchanger which could couple the sCO2 power cycle

to an advanced nuclear reactor or concentrated solar power plant. In this case, electrical heaters are the heat source to 350 kg

of nitrate salt which acts as the primary energy carrier. All components in Fig. 1 were present for pump testing, to characterize

4
the pump’s performance in this specific configuration. It should be noted that the salt loop was empty and all components were

near room temperature during the flow testing discussed here.

Fig. 1. Schematic of WisCO2 loop

2.1. Pump Description

The triplex pump was manufactured by Speck-Triplex of Germany and supplied by the North American distributor,

AquaBlast (Model CLP124C). The plunged volume is calculated using the manufacturer’s nominal dimensions: 24 mm bore

and 40 mm stroke. With each of the three plungers sweeping this volume once per revolution, the pump’s theoretical

displacement is 54.29 cm3 per revolution. The manufacturer is highly confident in the accuracy of the plunged volume

calculation due to extremely tight tolerances on the bore diameter and the crankshaft’s diameter and eccentricity. The clearance

volume or dead space for each plunger is 8.65 cm3. The maximum operating speed is 1000 rpm, suggesting a maximum

volumetric flow rate in excess of 50 L/min (13.2 gpm). The manufacturer’s maximum operating pressure is 300 bar (4350 psi).

The total pump head, or maximum pressure difference is very high due to the incompressibility of liquid or cool supercritical

CO2. The pump shaft is directly coupled to a 37 kW three-phase induction motor. The motor is powered by a variable frequency

drive (Danfoss VLT FC 102), which is controlled by a LabVIEW control and data acquisition interface.

The pump is equipped with several water channels to remove heat due to CO2 compression and seal friction. There are ten

circuits in total, six of which cool the exterior of the low- and high-pressure Teflon seal packings. Three of these circuits cool

large plugs threaded into ports in the suction manifold, while the final circuit runs perpendicular to the plungers and removes

heat from the valve casing near the discharge manifold. Chilled water is continuously supplied to these channels during

operation of the CO2 pump. The manufacturer specifies a minimum pressure drop of 1.5 bar (22 psi) across these channels

5
when connected in the parallel and the pump is operating at full speed. The chilled water is provided at approximately 10°C

and 60 psi. Usage of the chilled water system by adjacent laboratories will occasionally affect the supply pressure. To ensure

there is sufficient pressure drop across the channels, the supply pressure to the channels is increased by a high-head centrifugal

water pump (AMT Model 4901-98). This pump also forces flow through the Chiller, shown in Fig. 1. The CO2 stream in

counter-flow on the opposite side of this diffusion-bonded heat exchanger is cooled to improve the efficiency and longevity of

the pump. During heated tests, the chiller will reduce the CO2 temperature from around 100°C to less than 30°C where the

density is 2-3 times greater depending on system fill pressure. For the tests discussed here, only heat from the surroundings is

absorbed by the CO2 stream through the uninsulated system piping.

A simplified schematic of the reciprocating pump is shown in Figure 2. Spring-loaded valves control the flow of CO2 into

the plunger space and into the discharge port. These valves automatically actuate in response to the pressure differential between

the manifolds and the plunger space. The inlet, or suction pressure must be greater than the saturation pressure to prevent

cavitation in the suction valves or plunger space. This work was limited to an inlet pressure of 60 bar to maintain a conservative

margin of error (~10°C) between the inlet temperature and saturation temperature.

Fig. 2. Simplified Cross-Section of reciprocating triplex pump

Fig. 2 also shows the two sets of Teflon seal packing surrounding the plunger. The space between the low- and high-

pressure seals is maintained at an intermediate pressure to reduce leakage and to provide additional seal cooling. This

intermediate space is connected to a vessel independent of the main flow loop. This vessel is labeled Seal Cooling Inventory in

Fig. 1. The vessel is externally cooled with chilled water which flows within a shell surrounding the 18 L vessel. An Omega

PX119 pressure transducer and K-type thermocouple within the vessel are used in conjunction with an equation of state to

determine CO2 density within in the vessel [12]. The mass change in this vessel provides an indication of the condition of the

high-pressure seals. The bottom of the vessel connects to ports in the bottom of the plunger housing, while the top of the vessel

is connected to ports at the top of the plunger housing. This configuration is used to encourage natural circulation. The pump

manufacturer suggested that the pressure of this volume be maintained between 55 and 82 bar at all times to maintain high CO2

density for cooling purposes and to balance the pressure differential between the high-pressure and low-pressure seal sets.

6
 The crossheads for each plunger are lubricated by a spring-loaded automatic grease applicator with Multis Complex SHD

220 lithium complex grease. The crankshaft main bearings and journal bearing are all grease-lubricated roller bearings. These

are supplied with synthetic roller bearing grease manually at 4-hour intervals by the operator. The manufacturer offers an oil-

cooled configuration as well. At the time the pump was purchased, the oil-lubricated design was known to have some oil

leakage problems which were later resolved by including a labyrinth seal to prevent high-velocity CO2 leakage from impinging

on the oil seals in pumps with worn CO2 seals. The University of Wisconsin – Madison has purchased an oil-lubricated pump

to serve as direct replacement for the WisCO2 loop when then existing pump requires seal, valve or plunger replacement.

2.2. Heat Exchangers and Inventory Vessels

The chiller, recuperator, primary heat exchanger, and all interconnecting piping have been designed for supercritical

pressures. During high-temperature tests with molten salt flowing through the primary heat exchanger, the recuperator serves

to transfer heat between the streams running to and from the primary heat exchanger. Preheating of the CO2 stream above

220°C is required to prevent freezing of the molten salt. Although the maximum pressure of this work was 160 bar (2320 psi)

due to pressure limitations of one of the flow meters, new instrumentation will permit heat exchanger testing at the prototypical

pressure of 250 bar (3625 psi) frequently proposed for sCO2 Brayton cycles [4,8]. The behavior of CO2 encourages operation

at higher pressures for two reasons. First, is the non-linear pressure dependence of CO2 density. Figure 3 shows the density of

carbon dioxide at various pressures over a range of potential pump inlet temperatures. Increased inlet pressure will increase the

mass flow rate for a constant pump speed, while also permitting slightly elevated inlet temperatures for a constant inlet density.

Therefore during elevated-temperature testing, higher pressures will loosen the heat removal requirement of the CO2 Chiller.

Second is the effect of fluid compressibility. Figure 4 shows the isothermal compressibility of CO2 which quantifies the relative

volume reduction with an incremental increase in pressure at constant temperature. The isothermal compressibility is discussed

since the CO2 is actively cooled during compression and nearly mimics isothermal compression. For the tests described here,

the pump inlet was maintained at a nearly constant temperature of 11ºC +/- 1.5°C to maintain high density, while the outlet

temperature never exceeded 14°C.

7
 Fig. 3. Supercritical CO2 density at various pressures [11]

Fig. 4. Isothermal compressibility of CO2 versus temperature [11]

The buffer volume is a 1.37 m long section of 8 inch NPS Schedule 160 pipe (196 mm ID x 219 mm OD) with

hemispherical end caps. The internal volume is approximately 35 L, or nearly 10 times the volume of the balance of the

pressurized system, which includes the flow volume of all the heat exchangers and piping. The buffer volume serves two

functions. First is for regulation of the system pressure. During start-up, the tank is evacuated and filled with a bottle of

industrial grade CO2. A full CO2 bottle can be used to fill the buffer tank and loop, independently or together, to near the

saturation pressure at room temperature, which is near 60 bar (870 psi). Introducing additional CO2 inventory into the loop

requires the use of a Thar Laboratories Model P200A high-performance liquid chromatography (HPLC) pump. Under the best

conditions, the HPLC pump can supply 0.2 kg/min of CO2 to the buffer volume or experimental loop. With the HPLC pump

8
alone, it would require several hours to achieve the appropriate pressure. Therefore, tape heaters and Pryogel HPE insulation

are wrapped around the buffer volume to externally heat the vessel. Since there is minimal mass transfer between the buffer

volume and the loop, it was possible to maintain elevated temperatures in the buffer volume without insulating the rest of the

loop. The density of CO2 at ambient temperature and 60 bar is near 90% the density of water and its bulk modulus is high,

therefore a small temperature increase results in a significant pressure rise. Heating the CO2 inside the tank to 90°C will increase

the system pressure to near 250 bar. During high-temperature operation, as the recuperator and primary heat exchanger change

temperature due to flow rate or heat load adjustments, the buffer volume will help to limit pressure drift as well.

The second function of the buffer volume is to dampen pressure pulsations in the discharge piping. Because the physical

properties of the CO2 vary widely during start-up and steady-state testing, the tank was not acoustically designed to dampen

specific frequencies. The tank’s inlet port, a 19.3 mm hole in one end cap, is located directly above the pump’s discharge. The

line between the end cap and the discharge line is effectively 318 mm long and has inner diameter of 19.3 mm. A 1 inch ball

valve in this connecting line allows the buffer to be isolated from the loop at any time. The opening in the valve (Swagelok SS-

AFSS16) is 12 mm in diameter. These dimensions are used to estimate resonant frequencies of the buffer tank. Smaller ports

enable filling and venting of the buffer volume independent of the main loop. These auxiliary ports remained closed during

testing to close off any alternative flow paths between the buffer volume and the experimental loop.

3. Performance Measurements

3.1. Volumetric Efficiency Measurement

The volumetric efficiency of the triplex pump was measured over its speed range at three pressures. The pressure assigned

to each dataset is approximately the static pressure of the loop when the pump is stopped. The pump is equipped with two ports

at either end of both the suction and discharge manifolds. One of the suction and discharge ports are connected to the system

piping while the other was fitted with a custom plug with instrument feedthroughs. These plugs allowed temperature and

pressure measurement in both manifolds during operation. The suction or inlet manifold is a 48 mm diameter cylindrical cavity

running perpendicular to the plungers. The discharge manifold has a more complicated internal structure but also runs

perpendicular to the plungers. K-type thermocouples inserted into both manifolds measure the CO2 temperatures before and

after compression. To minimize the wall cooling effects, the thermocouples were located within 5 mm of the manifold

centerline and at a depth that locates the tip of the thermocouple directly above or below the nearest valve. Meanwhile, the

9
nominal suction and discharge pressures were measured by a pair Rosemont 3051T pressure transducers with a damping time

of 0.5 seconds.

To measure the flow rate produced by the pump ( ), a pair of Coriolis flow meters were used. Lower flow rates were

measured solely with a Siemens MASSFLO 2100F (DI6) Coriolis flow meter. This meter is calibrated to up 1000 kg/hr. Above

this flow rate, a ball valve located upstream of the second Coriolis meter, an Endress-Hauser CubeMass DCI (6 mm), is opened

to split the flow. This second meter is also calibrated to 1000 kg/hr. During split-flow operation the two meters reported flow

rates within 10% of each other under all conditions. This unequal split was the result of the difference in flow resistance of the

two flow meters. The pump speed was controlled by the variable-frequency drive to allow operation at increments of 50 rev/min

for performance testing. The operator entered the desired speed into the control interface which also provided a speed calculated

from the inverter’s feedback signal. The feedback signal was remarkably stable (+/- 0.05%) during all tests but its value was

offset by 1-2% of the true speed. The true rotational speed of the pump and motor shaft was measured by a strobe tachometer

(Nova-Strobe BAX) to improve the accuracy of the volumetric efficiency calculation.

Using the bore and stroke (24 x 40 mm) provided by the manufacturer, the ideal volumetric flow rate was calculated as the

product of the measured pump rotational speed (Np) and the ideal volumetric displacement per revolution (Vi = 54.3 cm3/rev).

Pressure and temperature measurements, taken in the suction and discharge were averaged over a 2-minute period of steady

operation. These average values were used in conjunction with the Span-Wagner equation of state to determine the time-

averaged densities within the suction and discharge manifolds [12]. The real volumetric flow rate output by the pump is

calculated as the measured mass flow rate ( ) divided by the discharge density ( ). Equation 1 shows the volumetric

efficiency defined as the calculated volumetric flow rate divided by the ideal flow rate suggested by swept volume and rotational

speed. The propagated uncertainty in this calculation is discussed below.

=
!"
[1]

Figure 5 shows the volumetric efficiency and observed mass flow rate versus pump speed for the three operating pressures.

The lowest pressure test, at 80 bar, was terminated at 800 rev/min as the pump inlet pressure dropped below 60 bar to avoid

operating near the saturation pressure. The highest pressure test was conducted at 160 bar due to the pressure limitation of the

second Coriolis flow meter (CubeMass). In the 120 and 160 bar test series, it was possible to exceed the calibrated flow rates

of both flow meters operating in split-flow (>0.56 kg/s). This occurs at speeds greater than 750 rev/min. These meters could be

10
calibrated to a combined rate of 3500 kg/hr, however the lower calibration range reduces uncertainty for heat exchanger testing,

which will typically use less than 40% of the pump’s flow capacity. Despite operating outside the calibrated ranges, data was

collected up to the maximum pump speed. The relative uncertainty of the flow meter reading to assumed to remain constant up

to the maximum pump speed. The mass flow rate plotted below the dashed line in Fig. 5 is remarkably linear due to the

incompressibility of the CO2 at these cool (11ºC +/- 1.5°C) conditions, deviating only slightly at low flow rates due to the

rapidly improving volumetric efficiency. The combined mass flow error for both meters is 0.3% of the measured value. These

error bars were not included in Fig. 5 for clarity, however the size data markers was selected to represent the maximum spread

of the error bars. The volumetric efficiency data, plotted above the dashed line is observed to be lowest at low speeds and

plateaus at about half of the rated speed near 500 rev/min. Volumetric efficiency dependence on system pressure appears most

clearly at mid-range speeds. Two sets of 80 bar data were collected due to the presence of undesirable valve behavior in the

data series denoted by **. This behavior is discussed in more detail in the next section.

Fig. 5. Volumetric efficiency and mass flow rate versus pump speed

The manufacturer provided a volumetric efficiency curve which was measured with water. The graphical curves provide

seven data points between 300-1800 rev/min for outlet pressures of 100, 200, and 400 bar [13]. The plot has not been reproduced

here due to a lack of numerical values accurate enough for comparison with this dataset. Two trends are present from the

11
manufacturer’s data. First, is the overall reduction in the volumetric efficiency curve as inlet pressure increases, approximately

a 3% reduction between each of the three curves. Secondly, all data points below the rated CO2 speed of 1000 rev/min are

greater than 91% before dropping at higher speeds, likely due to excessive losses in the valves.

In the CO2 tests conducted here, the relatively narrow pressure range makes it difficult to draw a conclusion regarding inlet

pressure. 120 bar inlet shows higher efficiency than the 160 bar dataset. Meanwhile the 80 bar data tends to drop below the

other data series. This is thought to be due to the CO2 properties at the inlet. All 80 bar data points above 300 rev/min have

sub-critical inlet conditions, meaning the pump compresses through the critical pressure and the CO2 exiting the pump has

physical properties that are highly variable with pressure and temperature which may contribute to valve losses.

3.2. System details affecting measured efficiency

It is important to note that the flow path of the experimental loop includes four passes through microchannel heat

exchangers, which generate significant pressure losses. Since there is no needle valve or interchangeable orifice plate installed

in the loop, it has a fixed flow restriction and does not allow the creation of a traditional flow versus pump head performance

curve. The pump tests were performed after the installation of the diffusion-bonded heat exchangers, because the triplex pump

in the WisCO2 loop will always be used in a configuration similar to that shown in Fig. 1. The pressure difference between the

pump suction and discharge is due to the pressure losses in the flow meters, heat exchangers, and inter-connected piping. Figure

6 shows the pressure differential across the pump for each operating point. Overall, the pressure losses in the loop are weakly

affected by the initial fill pressure. Higher fluid density and viscosity at higher pressures drive the increase in observed pressure

loss across the heat exchangers and piping for a given volumetric flow rate. The divergence observed below 450 rev/min is due

to the single flow meter versus split-flow configuration. Once the loop is switched to split-flow this lowers the overall flow

resistance of the loop. Since all 160 bar data was collected with both flow meters open, it exhibits a lower pressure than the

other data low-speed data collected with just one Coriolis meter measuring the flow.

12
 Fig. 6. Pump pressure differential versus pump speed

During some of the volumetric efficiency tests, a pressure phenomenon thought to be induced by the valves was observed

to affect the pump’s efficiency. These audible pressure pulsations were only present during the 80 bar tests reported here.

However, they were also observed at 130 bar during earlier shakedown testing. The occurrence of this behavior is intermittent

and is more likely to occur when the pump is cold (i.e. immediately after start-up).

Two sets of 80 bar data have been included in Fig. 5 to display the effect of this behavior on the volumetric efficiency.

The first 80 bar dataset was collected without observing this behavior, by collecting the data points at 350 and 400 rev/min at

the end of the test series, when the pump casing was warm. The second 80 bar dataset, denoted by (**) was collected after the

120 and 160 bar datasets in ascending order of speed. No data was collected below 300 rpm due to the severity of the valve

behavior at low speeds and to reduce wear on the suspect components. The phenomenon occurs more frequently as the pump

is operated at low fill pressure. The three lowest data points of the second dataset (350 - 450 rev/min) were collected over a

reduced collection period of 1 minute. Audible noise, unique vibration in the discharge piping, and high frequency (~560 Hz)

pressure pulsations were observed continuously during the collection of these data points. These pulsations are discussed further

in the following section. The ‘80 bar **’ data points between 350-450 rev/min exhibit a significant reduction in pump efficiency

compared to the first set, before the data sets converge at 500 rev/min as the high frequency vibrations dissipated.

All data shown in Fig. 5 was collected with the buffer volume isolation valve open. Since the buffer volume was isolated

at each operating point for collection of pressure pulsation data, efficiency data was computed for an isolated loop as well. For

all but two operating points, the calculated efficiencies for an open versus closed buffer isolation valve are within the propagated

measurement uncertainty. The maximum difference in measured flow rate of the two outliers was 1.3%, which is approximately

equal to the difference in calculated volumetric efficiency. The occurrence of these discrepancies coincides with large pressure

pulsation reductions observed by opening the buffer volume. In general, the volumetric efficiencies tended to be higher but

within the uncertainty when the buffer volume was open to loop. Since the maximum observed differences also exhibit this

trend, this provides strong evidence that the buffer volume serves to enhance the pump’s performance by dampening the

pressure pulsations caused by the reciprocating plungers.

3.3. Mechanical Efficiency

The mechanical efficiency of the pump was also calculated using the outputs from the Danfoss VLT FC102 variable-

frequency drive. During constant speed operation the calculated torque (#) from the inverter was recorded. The torque estimate

is calculated by the inverter based on voltage measurements between the motor coils. Danfoss reports an uncertainty of 5% on

13
these torque estimates. The mechanical efficiency ( ) is defined at the flow power divided by the mechanical power input to

the pump as shown in equation 2.

$%
= &' ( ) !"
[2]

The mechanical efficiency was calculated at each operating point and is shown in Figure 7. The efficiency is observed to

increase in a linear fashion over the operating range. Similar to the volumetric efficiency, no significant impact from changes

in absolute pressure are observed. The spread of the data between 400-500 rev/min is due to the use of a split-flow configuration.

The overall differential pressure (Δ ) used in eq. 2 comes from the total pressure drop shown in Fig. 6.

Fig. 7. Mechanical Efficiency versus pump speed

At high speeds the differential pressure across the pump approached 35 bar. If the loop losses were higher than shown in

Fig. 6, it is likely this would tend to shift the trend in Fig. 7 upward since differential pressure directly impacts the calculation

of mechanical efficiency. Higher differential pressure, increases the flow work or product of flow rate and pressure rise across

the pump. The 80 bar and 120 bar dataset show this for speeds less than 500 rev/min when only one flow meter was used to

measure flow. In other words, the pump is able to produce a larger pressure rise without a proportional increase in torque input.

The 80 bar data set in Fig. 7, which had subcritical inlet conditions, matches the higher pressure dataset where inlet conditions

remained supercritical at all speeds. This shows the utility of this reciprocating pump for applications with high differential

pressure or pump head requirements and liquid inlet conditions above or below the critical pressure.

3.4. Uncertainty Propagation in Efficiency Calculations

Each data point shown in Fig. 5 was calculated using the average value of the instrument outputs from 2-minute acquisition

at a 1 Hz record rate. For clarity, only the error bars for the 120 bar dataset are shown in Fig. 5 and 7. Although the signal

14
variance from all instruments over the acquisition period was much less than the manufacturer’s uncertainty specification of

each instrument, signal variance is included in the error calculation. The error of each independent variable (*) is defined as

the square-root sum of the manufacturer’s uncertainty specification (∆x) from Table 1 and the standard deviation of the 2-

minute acquisition (σx), as shown in eq. 3.

+ = ,(∆*)& + +
& [3]

Table 1
Uncertainty of instruments for efficiency calculations
Parameter Instrument Uncertainty
Siemens MASSFlo ± 0.15% of
Mass Flow 2100F measured value
Rate Endress-Hauser ± 0.15% of
CubeMass DCI measured value
Rosemount 3051S
Manifold ± 0.15% of set span
Absolute Pressure
Pressure (4000 psi)
Transducer
Outlet fluid Omega K-type
± 1.1 K
temperature Thermocouple
Rotational Nova-Strobe BAX
± 0.5 rev/min
speed Strobe tachometer
± 5% of measured
Motor Torque Danfoss FC 102 Inverter
value

Using the combined uncertainty for each variable determined by eq. 3, the overall uncertainty in volumetric efficiency

( ) is calculated using the error propagation formula for independent variables, shown in eq. 4. The three terms correspond

to the uncertainty introduced by the mass flow measurement, calculated discharge density, and pump rotational speed,

respectively.

&
234 & &
= 01
23 236
5 & + 1 45 (
& + !
&
2 2( 2!"
[4]

The variation in discharge density results from the variation in measured temperature and pressure in the discharge

manifold. Due to the numerical complexities of analytically propagating these errors through the Equation of State for CO2,

the computation of density variation ( ( ) and the propagation of this uncertainty was performed numerically in Engineering

Equation Solver (EES) [14]. Figure 8 shows the error in the volumetric efficiency measurement relative to the calculated

efficiency for each data point. Peak uncertainty is measured at low pump speeds where the error in rotational speed is the most

significant contributor. Overall, the 80 bar datasets show the greatest error. Steep variation of the fluid density with temperature

and pressure results in larger uncertainty in the calculation of observed volumetric flow rate in the 80 bar data series.

The uncertainty of the mechanical efficiency was calculated using the same numerical functionality in EES and is shown

in Figure 9. In addition to the independent variables in eq. 4, error due to torque estimate and the differential pressure
15
measurement are included as well. Below 200 rev/min the low differential pressure causes the relative uncertainty to exceed

100%. As pump speed, flow rate, and pressure drop increase, the relative error is dominated by the 5% error associated with

the inverter’s torque estimate.

Fig. 8. Relative error in calculated volumetric efficiency

Fig. 9. Relative error in calculated mechanical efficiency

3.5. Pressure Oscillations

A common characteristic of all reciprocating pumps are their pressure and flow rate oscillations. Triplex pumps, which

have three plungers 120° out of phase, exhibit much smaller pulsations than simplex or duplex pump due to the overlapping

discharge cycles of the plungers. However, theoretical calculation of a triplex pump’s mass flow rate for an incompressible

fluid indicates an oscillation of up to 23% about the average value is possible [15]. If these pulsations are unacceptable for the

application, a pulsation dampener can be installed near the pump’s suction and discharge piping. These dampeners are usually

tuned to absorb the pressure pulsations and stabilize the mass flow rate at a specific range of speeds. For large supercritical

pressure systems, dampeners can become expensive and difficult to maintain. Furthermore, compatibility issues with

diaphragm materials further restricts the options for supercritical CO2 pump users. Due to high cost, challenges associated with

variable operating pressure, and added system complexity, pulsation dampeners were not installed on the WisCO2 loop. In lieu

of dampeners, the inlet port for the buffer volume was located close to the pump’s discharge port. The intent was to use the 35
16
liter volume as a dampener whenever possible. Having no flexible diaphragm like traditional pulsation dampeners, the buffer

volume is only expected to act as a dampener at some speeds, while it may tend to act as a resonator at others. The goal of these

tests was to determine the influence of the buffer volume on the magnitude of pressure pulsations. The pressure pulsations

provide an indication of the mass flow variability that might be experienced by the system’s heat exchangers under test. Future

testing will measure the pressure pulsation near the heat exchangers where fluid properties are drastically different. If severe

enough, these pulsations could change the pressure loss or heat transfer behavior within the heat exchangers.

The triplex pump is equipped with a supply and discharge port on either end of the valve housing. In the WisCO2 loop, the

discharge piping exits the end opposite to the suction piping. To measure the magnitude of pressure pulsations directly, high-

speed pressure transducers were installed into the instrumented plugs in the second port of each manifold. The PCB Model

112A05 piezoelectric pressure transducers develop a charge that is proportional to the applied pressure. The charge is measured

by a PCB Model 462A charge amplifier, which outputs a 0-10 VDC signal to an oscilloscope. The Agilent DSO1024A

oscilloscope records the voltage waveform over a period of 0.5 – 1.5 seconds at 250 or 4500 Hz. The oscilloscope recorded

data with a resolution of 2 mV, which corresponds to a pressure resolution of 82 mbar. Linearity of these pressure transducers

is 0.15% of the full scale of 0 – 5000 psi.

Figure 10 shows an example of the pressure pulsations in the suction and discharge manifold at 198 rev/min when the

buffer volume was isolated from the discharge by means of a manually-actuated ball valve.

Fig. 10. Pressure variation in discharge (blue) and inlet (orange) headers

The pressure variation plotted in Fig. 10 is the instantaneous pressure reading from the transducer minus the average

pressure reading over the acquisition period. Since these transducers output a charge signal, charge leakage over time and

average pressure changes induced by pump speed will shift the voltage output from the signal converter-amplifier. Charge-

17
mode pressure transducers are designed to provide a rapid linear response to pressure fluctuations, not long-term pressure

measurements.

Since the pump speed was varied between 100-1000 rev/min this suggests pressure pulsation frequencies between 5 – 50

Hz. A set of three peaks represents one revolution of the pump. Variation in the pulsation amplitude over one pump revolution

is due to the placement of the sensors in the end of the valve casing, and thus being in closer proximity to the outermost plunger.

For example, the highest discharge peak is followed by the lowest segment of suction pressure in Fig. 10. As Plunger 1, nearest

the discharge transducer completes its stroke, Plunger 3 nearest the suction transducer begins its suction stroke.

Pulsation waveform measurements, like shown in Fig. 10 were collected with the buffer volume isolation valve open and

closed for each speed. The difference between these measurements was the primary indication of the effectiveness of the buffer

volume as a dampener. The waveform results are simplified by determining the maximum pressure pulsation in the discharge

header, by subtracting the average pressure in each dataset from the maximum instantaneous pressure signal from the

transducer. All files which exhibited valve phenomenon were omitted from this portion of the analysis. Figure 11 shows the

maximum discharge pulsation with the buffer volume isolated (Closed) and (Open) to the loop, for the three static fill pressures

associated with the efficiency measurements discussed in the previous sections.

Fig. 11(a). Maximum pressure pulsation at 80 bar

Fig. 11(b). Maximum pressure pulsation at 120 bar

18
 Fig. 11(c). Maximum pressure pulsation at 160 bar

Fig. 11 shows that the buffer volume is typically observed to reduce the maximum pulsation between speeds of 200 - 450

rev/min. It is less effective at mid-range speeds (500-650 rev/min) and may actually make the pressure pulsations worse at the

highest pump speeds. To provide insight into the varying ability of the buffer volume to reduce the pressure pulsation, two

simple resonance equations were used [16]. Equation 5 is used to determine the fundamental frequency (78 ) of a Helmholtz

resonator volume ( ), with an opening of area ( ), and corrected nozzle length (ℓ), filled with a fluid whose sonic velocity is

( ). The fundamental frequency for a pipe (7% ) of internal length (9) closed at one end was also calculated using Equation 6.

;
: <
78 =
&' ℓ
[5]

:
7% =
=>
[6]

Figure 12 plots the sonic velocity for various pressures. The boxes overlaid on the plot bound the temperature conditions

of the buffer volume during pump testing at each pressure. Since heating the buffer volume was the primary means of pressure

control during these tests, its internal fluid temperature changed between the dataset as shown, thus changing the expected

resonances. Using the average temperature and pressure conditions in the buffer volume to determine the speed of sound, the

two resonant fundamental frequencies are calculated for the tank in Table 2 among with the corresponding pump speed.

Fig. 12. Speed of sound in carbon dioxide under supercritical conditions [12]

19
 The results for a pipe closed at one end neatly correspond to high pump speeds where the buffer volume exacerbated the

pressure pulsations as seen in Fig. 11. Meanwhile, the fundamental Helmholtz resonance is below the minimum pump speed

for the 80 bar dataset while the predicted frequencies may correspond to the poor performance observed at low speeds for the

120 and 160 bar datasets. Using a methodology proposed in [17] for the second resonant mode of a cylindrical Helmholtz

resonator was calculated, but this frequency exceeded that associated with the maximum pump speed (1000 rpm : 50Hz) in all

datasets. Furthermore, the resonant frequency of the open piping between the pump and buffer volume was also significantly

greater than 50 Hz. Overall, the pressure pulsation magnitude is observed to increase in a linear fashion with pump speed and

the buffer tank is typically helpful in reducing the magnitude of these pulsations.

Table 2

?A [Hz]
Calculated resonant frequencies for buffer volume
P [bar] T [°C] a [m/s]
?@ [Hz]
(rpm) (rpm)
80 35 184 4.71 (94) 29.8 (596)
120 52 265 6.76 (135) 42.9 (858)
160 68 301 7.68 (154) 48.7 (974)

3.6. Valve Flutter

During the tests discussed in this paper, high-frequency pressure pulsations were observed between 300 and 450 rev/min.

These pulsations were occasionally present when accelerating through this speed range to higher speeds as well. After operating

at high speeds, decreasing pump speed to into the affected range allows for a few minutes of smooth operation before the

behavior resumes. The audible phenomenon induces significant vibration in the discharge piping and generated a high-

frequency pressure oscillation which was captured by the manifold pressure transducers. Figure 13 shows an example of the

behavior. Using the FFT algorithm in MATLAB, the peak frequencies were observed to be between 550-565 Hz.

Fig. 13. Example of high-frequency pressure pulsation due to valve flutter

20
 Using the discharge valve plate’s mass of 7.6 g and spring stiffness of 12,300 N/m, the valve’s natural frequency calculated

to be 1272 Hz. Since this is far from the pressure oscillation frequency observed at several different pump speeds, this behavior

is likely a forced vibration caused by flow through the valve port. This phenomenon has been studied extensively as

reciprocating compressors have developed over the last few decades [18]. It is clear that the temperature of the valve housing,

and thus the discharged CO2 has a direct impact on the valve behavior. It is possible to operate in the affected speed range for

a few minutes after operating at higher speeds. Review of the fluid conditions during this phenomenon does not reveal any

clear relation to the fluid properties. In addition to the tests presented here, this behavior has been observed at 101, 115, and

129 bar inlet pressure, with discharge temperatures measured outside the pump manifold between 11-13°C. The behavior has

not been observed below 200 or above 550 rev/min. The reduction in volumetric efficiency and occurrence of this behavior in

a repeatable speed range, supports the idea that this flutter phenomenon is a fluid-structure interaction. Due to lack of detailed

information about the valve geometry, further analysis of this behavior can not been performed at this time.

During future testing, the occurrence of valve flutter will be recorded and avoided whenever possible. This effectively

means, the pump will not be run in the mid-speed ranges. The introduction of a bypass circuit allows some of the pump’s flow

to bypass the experimental section and immediately pass through the chiller before returning to the pump. This bypass valve

will allow the operator to choose a pump speed where operation is smooth and tune the experimental section flow rate

accordingly using the bypass. Designers investigating the use of a reciprocating pump should keep this in mind if variable flow

rates are required.

4. Conclusions

The performance of a supercritical CO2 triplex pump was examined in a closed-loop test facility. The volumetric and

mechanical efficiencies are observed to increase with pump speed and show weak dependence on system fill pressures between

80 and 160 bar. Fluid compressibility remained low since inlet temperature hovered near 11°C for all tests, thus its effect on

volumetric efficiency was not observed. The volumetric efficiency was degraded by the presence of valve flutter at between

about 300-500 rev/min. The magnitude of pressure oscillations in the discharge and suction manifold were measured over the

entire speed range. The magnitude of the pulsations showed an increasing trend with pump speed and showed weak dependence

on system pressure. The buffer volume used for pressure control in the WisCO2 loop was observed to reduce pressure pulsations

at low pump speeds. Its effectiveness as a dampener reduces in the mid-range and may amplify the pulsations at high speeds.

The results of this test campaign served to identify the best operating conditions for heat exchanger performance testing. The

layout of the loop has been presented to provide background on how the pump and loop configuration may affect the results or

21
limit the scope of future heat exchanger performance investigations. Operation at higher pressures, up to 25 MPa, will expand

on the volumetric efficiency and pressure pulsation trends presented here.

Acknowledgements

This research has been performed using funding received from the U.S. Department of Energy Office of Nuclear Energy's

under an Integrated Research Program entitled “Advancements toward ASME Nuclear code case for compact heat exchangers

(IRP-17-14227)” with award number DE-NE0008714.

The authors would like to extend special thanks to Speck Triplex Pumps of Germany and Aquablast Incorporated for

providing the pump and providing information essential to this paper and to Vacuum Process Engineering and CompRex for

providing the heat exchangers incorporated into the test facility.

22
References

[1] E.I. Koytsoumpa, C. Bergins, E. Kakaras, The CO2 economy: Review of CO2 capture and reuse technologies, J.
Supercrit. Fluids. 132 (2018) 3–16. https://doi.org/10.1016/j.supflu.2017.07.029.
[2] E.J. Beckman, Supercritical and near-critical CO2 in green chemical synthesis and processing, J. Supercrit. Fluids. 28
(2004) 121–191. https://doi.org/10.1016/S0896-8446(03)00029-9.
[3] R.U. Rony, H. Yang, S. Krishnan, J. Song, Recent Advances in Transcritical CO2 (R744) Heat Pump System: A
Review, Energies. 12 (2019) 457. https://doi.org/10.3390/en12030457.
[4] K. Brun, P. Friedman, R. Dennis, (Eds.), Fundamentals and Applications of Supercritical Carbon Dioxide (sCO2) Based
Power Cycles, Woodhead Publishing, Cambridge, MA, 2017.
[5] M.-J. Li, H.-H. Zhu, J.-Q. Guo, K. Wang, W.-Q. Tao, The development technology and applications of supercritical
CO2 power cycle in nuclear energy, solar energy and other energy industries, Appl. Therm. Eng. 126 (2017) 255–275.
https://doi.org/10.1016/j.applthermaleng.2017.07.173.
[6] C. Gorecki, Plains CO2 Reduction Partnership Phase III Final Report, 2019. https://doi.org/10.2172/1580755.
[7] B. Wadas, GE Oil and Gas: CO2 Capture and Compression Technologies, Proceedings of the CO2 Summit: Technology
and Opportunity. Vail, Colorado, June 6-10, 2010.
[8] S.R. Aakre, I.W. Jentz, M.H. Anderson, Nuclear code case development of printed-circuit heat exchangers with thermal
and mechanical performance testing, 6th sCO2 Symposium. Pittsburg, Pennsylvania, March 27-29, 2018.
[9] R. Keating, J. Nestell, S. McKillop, Potential ASME Code Case for Compact Heat Exchangers in High Temperature
Reactors: ICONE27-1138, 27th International Conference on Nuclear Engineering, Tsukuba, Japan, May 19-24, 2019.
[10] H. P. Mahajan, I. W. Jentz, T. Hassan, Allowable Stress Development of Diffusion Bonded alloy 800H for Section III:
PVP2020-21499, ASME 2020 Pressure Vessels & Piping Conference, Minneapolis, Minnesota, July 19 – 24, 2020.
[11] G. Selby, S. Aakre, Z. Fan, Non-destructive Examination of Diffusion-Bonded Heat Exchangers: PVP2020-21293,
ASME 2020 Pressure Vessels & Piping Conference, Minneapolis, Minnesota, July 19 – 24, 2020.
[12] R. Span, W. Wagner, A New Equation of State for Carbon Dioxide Covering the Fluid Region from the Triple‐Point
Temperature to 1100 K at Pressures up to 800 MPa, Journal of Physical and Chemical Reference Data. 25 (1996) 1509–
1596. https://doi.org/10.1063/1.555991.
[13] Speck Triplex Pumps, Technical Specifications for P52/51-300CZ Circulation CO2 Pump. https://www.speck-triplex.de/
(accessed November 2017).
[14] F-Chart Software, EES – Engineering Equation Solver, Academic Professional Version 10.644,
http://fchartsoftware.com, June 2019 (accessed April 2020).
[15] J.E. Miller, The Reciprocating Pump: Theory, Design, and Use, second ed., Kreiger Publishing Company, 1995.
[16] L.E. Kinsler, A.R. Frey, A.B. Coppens, J.V. Sanders, Fundamentals of Acoustics, second ed., Wiley, New York, 2000.
[17] R.L. Panton, J.M. Miller, Resonant frequencies of cylindrical Helmholtz resonators, The Journal of the Acoustical
Society of America. 57 (1975) 1533–1535. https://doi.org/10.1121/1.380596
[18] G. Wang, L. Zhong, X. He, Dynamic Behavior of Reciprocating Plunger Pump Discharge Valve Based on Fluid
Structure Interaction and Experimental Analysis, PLOS One 10 (2015). doi:10.1371/journal.pone.0140396

23