You are on page 1of 11

Proceedings of ASME Turbo Expo 2009: Power for Land, Sea and Air GT2009 June 8-12, 2009,

Orlando, Florida, USA


Proceedings of ASME Turbo Expo 2009: Power for Land, Sea and Air 2009 June 9-12, 2009, Orlando, Florida, USA

GT2009-59548
GT2009-59548
AN EXPERIMENTAL INVESTIGATION OF A SINGLE STAGE WET GAS CENTRIFUGAL COMPRESSOR
Michelangelo Fabbrizzi GE Oil & Gas Florence, Italy michelangelo.fabbrizzi@ge.com Francesco Del Medico GE Oil & Gas Florence, Italy francesco.delmedico@ge.com ABSTRACT Wet Gas Compression technology for centrifugal compressors is without doubts one of the topics that have enjoyed a most rapidly growing interest in recent years [1]. This is certainly due to the new opportunities for exploiting aging gas fields, which are rendered possible by this technology. However, despite the numerous recent works on this topic, the fundamental understanding of wet gas compression processes remains challenging in the sense that no unified standards for testing and performance exist, and that the understanding of the basic physical phenomena is not fully complete. The present work is concerned with such issues and contributes experimental results and performance analysis for a low pressure-ratio centrifugal compressor stage in heavily wet gas streams. An open loop test rig was built for a single scale model centrifugal compressor to be run at typical design operating range. The impeller was a standard impeller with no design adjustments for wet streams. The performance was assessed at liquid mass fractions (LMF) ranging between 0% and 50% and at average droplet diameters between 50m and 75m, as well as non-atomized liquid streams. The results clearly demonstrate that wet gas strongly influences compressor performance, considering adsorbed power, efficiency and compressor characteristics. The results also show that, although the impeller design did not account for heavily wet streams, the overall compression characteristics remain good and the machine was able to process the two-phase mixture without failures, in some cases with improved pumping. The compressor exit temperatures of the gas phase in heavily wet conditions are key to assessing machine performance, in particular efficiency, for wet gas compression. Ciro Cerretelli GE Global Research Europe Munich, Germany ciro.cerretelli@ge.com Maria D'Orazio GE Oil & Gas Florence, Italy maria.dorazio@ge.com Although an accurate technology approach for measuring such temperatures has not yet been developed, in the present work a droplet model is proposed. This model is then calibrated with experimental data in order to develop a methodology providing reliable efficiency values. NOMENCLATURE ADroplets cp D D32 Fr GMF GVF h LMF LVF M m n Nu p Q Re T VF VDroplets W X = = = = = = = = = = = = = = = = = = = = = = = = Droplet Mean Surface [m2] Constant Pressure Heat Capacity [kJ/(kg K)] Droplet Diameter [m] Sauter Mean Droplet Diameter [m] Frssling number [-] Gas Mass Fraction [-] Gas Volume Fraction [-] Specific Enthalpy [kJ/kg] Liquid Mass Fraction [-] Liquid Volume Fraction [-] Mach Number [-] Mass Flow Rate [kg/s] Rotational Speed [rpm] Nusselt Number [-] Pressure [Pa] Volumetric Flow Rate [m3/s] Reynolds Number [-] Temperature [C] Vapor Fraction [-] Droplet Mean Volume [m3] Power [kW] Molar Concentration [-] Static to Static Pressure Ratio [-] Total to Total Pressure Ratio [-]

SS TT

Copyright 2009 by ASME

Downloaded 15 Mar 2010 to 190.177.4.3. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

= = = =

Flow Coefficient [-] Efficiency [-] Constant Pressure to Volume Heat Capacity [-] Rotational Speed [rad/s]

Subscripts 0 = 1 = 2 = ev = g = l = s = surf = t = =

Stagnation Quantity Compressor Inlet Compressor Discharge Evaporation Gas Liquid (water in this work) Steam Droplet Surface Total (Gas+Liquid) Ambient Condition

Superscripts * = Normalized Quantity INTRODUCTION In the oil and gas industry it is customary that large, multistage compressors are used to process gas in upstream and downstream applications. In recent years, due to the large demand for fossil fuels, most gas fields have been exploited to the limits that current compression technologies allow to reach. Further production from such gas fields will hinge on two alternatives: either large scale sub-sea separation with traditional gravity scrubbers, or direct compression of the unprocessed well stream. The latter is often referred to as wet compression, and it constitutes the subject of the present investigation. Wet compression for industrial oil & gas applications strongly differs from the more established techniques for power augmentation in large multi-stage axial machines. Typical liquid mass fractions of the unprocessed stream are not limited to 2-3% but can range up to 50%. In order to evaluate the impact of such large wet streams on the compression work, a simple non-equilibrium thermodynamical model has been developed, based on droplet evaporation and heat transfer. This droplet model follows the wet compression on the compressor mean-line. However such a model is unable to assess the degradation of the aerodynamic efficiencies associated with wet gas compression, especially at high liquid content, and is thus unable to characterize the machine performance. In order to perform this characterization, a single stage compressor test rig operating in open loop at ambient conditions has been designed and built at the GE Global Research Center in Munich. SCOPE AND PREVIOUS WORK Most of the previous work on wet gas is associated with axial compressors, particularly in applications such as inlet or inter-stage fogging. Such applications are aimed at augmenting the turbine output power by reducing compressor power consumption. As a consequence, most performance evaluations are related to water/air mixtures for industrial installations (see Sanjeev et al [2, 3], Bhargava et al [4, 5] and Zachary et al.[6]). While these works have addressed primarily the first order effects of compression power reduction and the consequent

increase of the turbine mass flow and heat capacity due to water injection, Hrtel & Pfeiffer [7] and Zheng et al [8] conducted research on the influence of the several fogging parameters such as the droplet diameter, relative humidity and gas volume fraction on the compressor performance. In the past, attempts to employ wet compression technologies in industrial compressors have been unsuccessful, mainly because the fogging equipment technology was insufficient for application in such machines. Recent advances in water injection equipment and analytical tools have generated renewed interest, as demonstrated by the work of Abdelwahab [9] and Shibata et al. [10]. With the possibility of employing droplet radii < 5m, sensible performance improvements should become increasing possible. The scope of the present work, however, is not related to the applications described above, but focuses on the oil and gas industry. As previously discuss, such applications are typically characterized by much larger liquid mass fractions and, if subsea, by very high gas density and consequently much smaller density difference in the two-phase mixture. The most relevant published work to date is certainly that of Brenne et al. [11] and Hundseid, Bakken et al [12], which relate to the Statoil wet compression test program. In 2003/2004 Statoil tested a single-stage centrifugal compressor in a wet gas mixture with hydrocarbon fluids typical of those encountered in North sea gas fields, and evaluated correction methods to analyze the machine polytropic performance. They found that the application of centrifugal compression technology is a viable option for single-stage, two-phase compression at GVF at or above 0.97, corresponding to GMF as low as 0.52 with suction pressure ranging from 30-70 bar. The pressure ratio increased relative to dry gas conditions when the GVF was decreased. The polytropic efficiency decreased, and the overall specific power consumption was found to be higher than separating the two phases and boosting them individually. In order to assess methodologies that would be extremely valuable in the design phase, it would be interesting to see whether these results are reproducible in a laboratory environment, and therefore to test the performance of a singlestage impeller at ambient conditions in open loop operation. This is particularly true when the extremely high costs of a testing program reproducing sub-sea conditions are taken into account [1]. In this work, the goal was to assess this methodology and to investigate the behavior of a single-stage centrifugal compressor processing an air-water mixture in open loop at ambient conditions. The LMF ranged from 0-0.6, and the average D32 ranged from 50-75 m, which are typical of a wet gas application. Working with air/water mixtures at ambient conditions means dealing with extremely high density ratios (about 900) and thus substantially different mass and energy transfer with respect to subsea conditions.

EXPERIMENTAL SETUP The test vehicle used for this work was a single stage centrifugal compressor. This compressor, equipped with an unshrouded impeller, a vaneless diffuser and a volute, is not a commercial GE product; rather, it was designed specifically for this research program. It has an impeller diameter of 0.15 m

Copyright 2009 by ASME

Downloaded 15 Mar 2010 to 190.177.4.3. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

and a design flow coefficient of 0.0748 m. The compressor was originally designed to handle inlet flows at ambient conditions and to provide discharge pressure ratios up to 1.64 at a speed of 35,000 rpm. The original design of this machine was for standard dry gas applications, therefore its impeller and the internal vane geometries are in no way optimized for heavily wet flows. An electric motor drives the compressor through two transmission stages consisting of a belt-pulley system and an internal oil-lubricated gearbox with an overall gear ratio of 11.67. A variable speed drive allowed for a variation of the rotational speed to the desired value up to a maximum of 35,000 rpm. Figure 1 shows a schematic diagram of the test rig, whose main constituents are the compressor with the related oil lubrication circuit, the spray system for injecting water, the water pump, the separator setup, the mass flow measuring setup and the water disposal apparatus. Figure 2 shows the entire assembly. The compressor draws the airflow at ambient conditions through a bellmouth where water is injected as droplets of different diameters. The injector is a FlowMax air atomizing nozzle from Spraying Systems. The injector was specifically calibrated for varying the D32 between 50-75 m since droplets in this size range are typical after a scrubber in oil & gas applications. After the single-stage centrifugal compressor, the processed air-water mixture passes through a horizontal separator, which was used to separate the water flow from the air stream in order to permit accurate measurement of the mass flow of each stream (liquid and gas). This separation is also necessary in order to allow the reduction valve to operate on dry air and thus provide for finer control along the compressor operating speedline. The gas stream was measured with a calibrated orifice plate, while the liquid stream was measured using a calibrated liquid level measurement system in the separator vertical collector tank. The liquid could be introduced into the gas in two different ways; as a uniformly distributed droplet mist injected directly onto the compressor inducer, or as a liquid film uniformly coating the wetted surface of the inlet pipe. The latter was preferred because it allowed for precise measurements of the inlet pressure and temperature. However, some interesting results with the uniformly distributed droplet mist will also be presented. The liquid and gas phases were injected at the same temperatures in the inlet pipe, and the mixture can be considered in equilibrium, neglecting second order effects.

The temperature and static pressure of the gas stream entering and exiting the compressor have been measured using calibrated thermocouples and pressure transducers. These instruments have been flush mounted on the wall in order to avoid water related problems in the transmission lines. The accuracy of those measurements had been proven to be less than 0.5% even at wet conditions, except for the temperatures at the compressor discharge in wet streams. This temperature measurement represents a fairly major issue that, to our knowledge, has not been solved to date. In fact, due to the inherent non-equilibrium nature of the mixture at the compressor discharge, the temperatures of the liquid and gas phases strongly differ, as the air temperature is presumably substantially higher than the water temperature. While it is relatively easy to measure the liquid temperature at such conditions, the heavy wetness of the mixture prevents any available measuring device from accurately measuring the gas phase temperature, since water deposition on the sensing elements strongly affects the measurements. This problem will be discussed in details in a separate section. WET GAS DEFINITIONS Wet gas flows are characterized by high gas volume fractions. The parameters traditionally used to define the main features of such mixtures, when processed by a compressor, are defined below:

Figure 2: Wet gas test rig facility at GRC in Munich.

GVF =

Qg Q g + Ql

(1)

Figure 1 Schematic of the Wet Gas Test Rig.

Copyright 2009 by ASME

Downloaded 15 Mar 2010 to 190.177.4.3. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

LVF =

Ql Qg + Ql
mg m g + ml

(2)

of the dry lines. It is now interesting to see how the picture changes when the performance with respect to the volumetric flow coefficient is considered. D32 (m) 50 n 20000 25000 30000 20000 25000 30000 20000 25000 30000 max LMF [%] 60 50 40 50 50 40 60 40 30

GMF =

(3)

LMF =

ml m g + ml

(4)

75
The Sauter droplet mean diameter is

D32 =

6VDroplets ADroplets

(5)

air-off

whereas the flow coefficient is

Table 1: Summary of test conditions.

8QINLET
3 DOUTLET

(6)

For a centrifugal compressor these primary variables are also complemented by the pressure ratio, the required power and the temperature ratio (efficiency). Compressed two-phase mixtures are usually considered to be wet gas compression when the liquid content is below 5% in volume, that is GVF > 0.95; otherwise they are considered to be multiphase pumping. To orient the reader, for an air-water mixture, an LVF = 0.05 translates to about LMF= 0.50 at typical subsea conditions (p =70 bar), while at ambient conditions this mass fraction becomes LMF=0.97! This is, of course, due to the very high density ratio as atmospheric pressure is approached. Therefore, for a wet compression test in open loop such as that discussed in the present work, the volumetric water flow is always extremely small so that GVF >> 0.99. In such cases, it is more effective to refer to the mass fractions. Due to the high velocities and the low volumetric liquid content in the inlet pipe of our experiment, the multiphase flow can be classified as annular flow, where the liquid phase is present as a liquid film at the wall and as a dense droplet flow in the core.

LMF [%] 0 5 10 20 30 40 50 60

GVF [%] 100.000 99.994 99.987 99.971 99.952 99.925 99.890 99.836

Table 2: Gas Volume Fractions and Liquid Mass Fraction at ambient inlet conditions for an air-water mixture.

LMF = 40% 1.07 1.00 0.93 * [- ] 0.87 0.80


30000* Dry D32=50 D32=75

WET GAS PERFORMANCE Table 1 displays the key parameters and the range of variation during the testing, while in Table 2 the correspondence between the gas mass fraction and volumetric content is illustrated. Clearly, the liquid volumetric content is almost negligible even for large mass fractions at our test conditions. The compressor performance at heavily wet conditions can be first illustrated by a representative case with LMF=0.40. Defining the total mass flow rate (mTOT*) as * mTOT = m * + m * (7) l g

0.73
20000

25000

0.67 0.17 0.33 0.50 m* [-] 0.67 0.83 1.00

Figure 3. Total Mass Flow Rate vs. Static Pressure Ratio for Different Droplet Diameters (LMF=40%).

and plotting this against the static pressure ratio (*) for a dry case and two different droplet diameters (D32*) (see Figure 3), it appears clear that the drop in pressure ratio is compensated for by a higher mass flow rate elaborated by the impeller, so one could say that the wet speedlines are just a "continuation"

Figure 4, Figure 5 and Figure 6 illustrate the performance of the compressor, in terms of pressure ratio when processing wet gas mixtures. The quantities , and P are shown as normalized with the corresponding reference value. The first noteworthy result is that the impeller continued to work well, being able to process the mixture up to LMF=60. The experiments indicate that such limit could in principle be pushed to even higher values, however, it was not possible to

Copyright 2009 by ASME

Downloaded 15 Mar 2010 to 190.177.4.3. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

perform such testing with the current system. At such an extremely high liquid content, the gearbox, designed for normal dry gas operations, would have been overstressed and at risk of failure because of the high axial thrust and torque. The impeller was able to operate for sustained cycles even when the air atomization was turned off and therefore, only a liquid water stream was injected (Figure 6). As far as the performance is concerned, the key results are: a) The pressure ratios are generally lower than those for dry conditions except at LMF 10% in the low flow coefficient range b) The slopes of the characteristic curves are higher, particularly in the high flow coefficient range. The results showing that the machine produces lower pumping characteristics are not in agreement with previous data by Brenne et al [11] and Hundseid et al. [12], who attributed the higher pressure ratios to the increased density of the processed fluid. At their test conditions, however, the inlet pressure varied from 30-70 bar, thus the density difference between dry and wet conditions was much smaller than in the present campaign, whereas the density difference between the two phases is about 820-900. It is therefore possible that, when the phase differences are smaller, the fluid thermodynamics in the impeller couples better with a machine that is originally designed for dry gas operation, allowing for better performance. In the future, closed loop testing at similar inlet conditions could reveal whether this large performance decay is due to the higher density difference or to the particular geometry of the present setup. A more thorough analysis on the impact of the liquid slip at the compressor inlet is needed and will be implemented in the future. The influence of the impeller size and the measurement error in the overall power balance between the driver and the compressor also need to be better assessed. The difference in slopes is also shown by Brenne et al (Figure 8 in [11]), and is probably due to the inlet distortion and consequent pressure losses induced by the liquid phase on the gas phase. Although the volumetric liquid content is basically negligible (Table 2), the spray pattern has a strong influence at the injection point, due to the very high slip factor between the droplets and the air stream, which presumably creates high distortion. These losses are higher at high flow rates where the slip factor increases. Employing an isokinetic spray injection mechanism would minimize such effects, however, this was not possible in the present configuration. A very interesting data subset is shown by the injection with LMF=5-10% where the pumping characteristics are worse than those of the dry gas at high flows. However, for low flow coefficients, the results show higher pressure ratios. This fact can be explained as the combination of the benefits given by the density increase and the aerodynamic distortion caused by the presence of droplets. For low LMF and the benefits due to increased density are more important than the aerodynamic distortion, thus producing a net benefit. Another benefic contribution is given by the so called intercooling effect yielding a reduction of gas temperature due to the presence of the liquid. However these conclusions must be further investigated and better quantified.

From Figure 4-Figure 6, it is interesting to note that there is no major performance variation for different droplet sizes, although smaller droplets do result in slightly better performance. This observation is quantified in Figure 7, where the speedlines are plotted for a LMF=0.4 at different droplet diameters. This almost negligible difference in performance is due to the inlet geometry. Because of the rather long inlet pipe, most of the injected the droplets create a film of liquid water on the pipe surface. Different droplet sizes have only a minimal effect on this film, changing only the core flow. Surprisingly, when air atomization is turned off, the performance becomes slightly better. However, since this flow regime consists of a water jet in the core with no liquid film at the wall, this latter case is completely different and cannot be directly compared to the previous two.

1.2 1.15 1.1 1.05


*
Dry 5% 10% 20% 30% 40%

1 0.95 0.9 0.85 0.8 0.4

0.6

0.8

1.2

1.4

1.6

Figure 4. Wet Compression Map for D32=50m.

1.2 1.15 1.1 1.05


*
Dry 5% 10% 20% 30% 40%

1 0.95 0.9 0.85 0.8 0.4

0.6

0.8

1.2

1.4

1.6

Figure 5. Wet Compression Map for D32=75m.

Copyright 2009 by ASME

Downloaded 15 Mar 2010 to 190.177.4.3. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

1.2 1.15 1.1 1.05


*
Dry 5% 10% 20% 30% 40%

1 0.95 0.9 0.85 0.8 0.4

0.6

0.8

1.2

1.4

1.6

Figure 6. Wet Compression Map for non-atomized liquid injection.

In order to better investigate the effect of droplet size, experiments were performed with a different inlet geometry in which the spray nozzle has been placed in front of the inducer so that the sprayed particles were directly injected into the compressor. Photos of the inlet areas for the two different configurations are shown in Figure 8. The latter setup could not be used for detailed performance characterization because of the strong inlet distortion that is induced by the nozzle head, and because of the lack of a viable way of measuring the inlet pressure. It was however useful in order to assess the influence of particle size. The results for the new setup are shown in Figure 9. It can be seen that the performance variation is larger, the smaller droplets allowing for a considerably higher pumping characteristic. For a real application it would therefore be important to carefully design the inlet geometry for a stronger atomization if needed. No evidence of liquid erosion has been detected by visual inspection of the machine internals after the test. The test results also showed that the vibration of the machine was not significantly affected by liquid ingestion for the different injection patterns that have been tested.

Figure 8: Inlet area and injection setups. Setup with inlet pipe is shown above, while direct injection without the inlet pipe is shown below.

.
1.2 1.15
Dry 50m 75m

1.2 1.15 1.1


*
Dry 50m 75m No air

1.1 1.05 1 0.95 0.9 0.85 0.8 0.4

1.05
*

1 0.95 0.9 0.85 0.8 0.4

0.6

0.8

1.2

1.4

1.6

1.8

0.6 0.8 1 1.2 1.4 1.6 1.8 2

Figure 7: Wet Compression Map for LMF=0.4, with inlet pipe.

Figure 9: Wet Compression Map for LMF=0.4, without inlet pipe.

Copyright 2009 by ASME

Downloaded 15 Mar 2010 to 190.177.4.3. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

POWER AND TEMPERATURE MEASUREMENTS In order to determine the work reduction achieved by wet compression, the compression work has to be calculated. This work can be computed knowing the temperatures of the mixture between any successive states. Using the first law of thermodynamics for an adiabatic open process and applying it to the air-water vapor-liquid water-mixture, following the derivation by Hrtel and Pfeiffer [7] yields:

suspended in a gas phase undergoing compression. For the mathematical description of the gas-droplet system, standard models of dispersed gas-liquid two-phase flow shave been adopted, as described e.g., by Young [13], Berlemont et al. [14], or Faeth [15], and the approach of Hrtel and Pfeiffer [7] have been employed using the following simplifying assumptions: 1. 2. 3. 4. 5. 6. Spherical droplets Absence of droplet-droplet interactions (i.e., breakup) Infinite k within the droplet Homogeneous properties within phases Air saturated at droplet surface Thin vapor layer ("reference state") surrounds the droplet

& & W = ma (ha 2 ha1 ) + ml [(hs 2 hs1 ) +

(VF2 VF1 ) hev1 + (1 VF2 )(hev1 hev 2 )]

(8)

This expression is quite general and allows for evaporation, and for the liquid (droplet) temperature to differ from both the wet and the dry bulb gas. This is of paramount importance in any wet compression problem. At the inlet, it is assumed that the mixture enters the compressor at equilibrium, and therefore the liquid and gas phase temperatures are the same. However, as the mixture is processed, the gas phase temperature increases because of the compression and therefore a heat flux ensues from the gas to the liquid phase, generating non-equilibrium thermodynamic processes resulting in a cooling effect for the gas phase, and heating and evaporation for the liquid. In our applications, given the large values of the water mass fraction and the relatively low pressure ratios, the evaporation and the steam contribution terms are negligible, since such values were determined as lower than 1% if compared to the others. Therefore, neglecting the losses in the pump, the overall power becomes:
1 2 & W = m g C g T2 g + M 2 T1 + Qliq ( p 2 p1 ) 2 & + ml Cl (T2 L T1 )

During compression the temperature of the gas rises leading to a heat flux from the gas to the water droplet, which increases the water temperature and causes evaporation of the liquid. During evaporation, a thin vapor layer surrounds the droplet and the difference in relative humidity between this vapor layer and the ambient gas is the driving force for the evaporation process. The following equation is obtained for the steam mass rate of transfer from the droplet to the air, where indices "" and "surf" denote, respectively, conditions in the ambient gas and at the droplet surface: D 1 X s , & (10) m s = 4 ref ref ln 1 X s , surf 2 The molar concentrations Xs, and Xs,surf are defined as the ratio of the vapor pressure to the mixture pressure: p surf p X s , = X s , = (11) p p where psurf is the saturation pressure at the droplet surface and can be computed by means of the Antoine equation [16], and p can be easily derived from the known steam content and Dalton's law. In this model the droplet is treated as having infinite conductivity, implying that the temperature inside the droplet and at the droplet surface is uniform. The index "ref" denotes a vapor reference state surrounding the droplet. This reference state is computed here with the aid of the widely used 1/3-rule (initially proposed by Sparrow and Gregg [7]), which we applied to the temperature and vapor mass fraction. The Nusselt number is thus given by the formula:
& m s cp s ,ref Nu = 2 ref D 2 & exp m s cp s ,ref 1 4 ref D 2 (12)

(9)

It becomes clear that in order to properly assess the power and the efficiency of a wet compressor, accurate pressure, flow rate and temperature measurements are necessary. As previously pointed out, pressure measurements and liquid phase temperature measurements can be carried out very accurately. The same holds for the flow rate of both the liquid and gas phases. However, the gas temperature measurements at the compressor discharge (T2g) cannot be carried on accurately in such conditions because of the heavily wet environment and the inherent non-equilibrium nature of the mixture, which is characterized by a significant heat flux between the air and water. In fact, droplets hit the sensing end of the thermocouple, and thus profoundly alter the measured temperature from that of the true gas phase temperature. Furthermore, any attempt at isokinetic sampling in order to separate the two phases for measurements is not successful in this rapidly changing, nonequilibrium environment. Therefore, in order to assess the gas phase exit temperatures, a droplet model has been developed.
DROPLET MODEL As previously noted, to investigate how the work reduction in wet compression is influenced by thermodynamic nonequilibrium, a model must be employed which allows taking into account the finite relaxation time of evaporating droplets

The following differential equations can then be derived for the time evolution of the droplet diameter and the droplet temperature:

Copyright 2009 by ASME

Downloaded 15 Mar 2010 to 190.177.4.3. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

X s , 1 4 ref ref dD = ( Fr ) ln X s,surf 1 dt l D


dTl 6 Fr Nu = Nu ref (T Tl ) dt l D 2 Cpl 2hev ref ref X s , 1 ln X s ,surf 1

(13)

This calculation procedure is then repeated until the compressor exit conditions are reached. If the droplet size reaches D 2 10 7 the liquid phase is considered to be completely evaporated, and normal, dry polytropic compression follows up to the compressor discharge. However, this condition does not occur given our very large inlet LMF.

(14)

In the previous formula, the gas and liquid properties (, , , Cp) are computed from curve fits derived from correlations given in [17]. The equations above can be solved numerically. The inlet conditions must of course be specified in terms of the temperatures of the gas and liquid phases Tg and Tl (equal at the inlet since the mixture is assumed to be at equilibrium), the droplet size D, and the flow velocity V. In order to adapt to a compressor simulation, the average pressure increase p (x ) and

axial velocity distribution u (x ) must be specified along the compressor mean line. The axial position with time x(t ) is assumed to be related to the velocity distribution along the mean line. The numerical solution can then proceed in incremental time steps related to the incremental pressure increase as specified by the given pressure mean line profile p(x(t )) . The incremental compression steps are governed by polytropic compression laws, which also account for losses through a polytropic index n and efficiency pc defined as in [18]. The differential equations (14) and (15) were solved for constant ambient air conditions during one time step. As the pressure, and thus the temperature, increase during one time step, the mean value of the pressure and the temperature of this time step without evaporation was taken to evaluate the thermodynamic quantities of the differential equations (14) and (15). The pressure after one time step was taken from the compressor mean line profile. Thus the temperature at the end of this time step can be obtained from the simple polytropic laws. This temperature is the temperature without feedback from cooling. Knowing the temperature and the pressure, the thermodynamic quantities in (13), (14) and (15) can be determined and the differential equations can be integrated for one time step. At the end of this time step, the feedback from cooling occurs in an infinitesimal time at constant pressure. Applying the constant enthalpy condition:

COMPRESSOR WET EFFICIENCY CALCULATIONS The droplet model illustrated in the previous chapter takes into account the geometrical properties of a given compressor along with the pressure ratio related to different compression paths. The effects related to the presence of water droplets is taken into account through combination of standard thermodynamic laws yielding to a non-equilibrium state; this condition represents the best description of the two-phase non homogeneous flow at the outlet highlighting the difference between gas and water temperatures. Given the geometrical properties of the compressor, the developed model necessitates three main external inputs: 1. A pressure mean line profile p(x(t)) 2. A polytropic efficiency pc 3. The Frssling number (Fr), The static pressure profile p(x) can be measured quite accurately in the wet gas tests, and it therefore can be directly input into the droplet numerical calculation. The axial position with time x(t) can be also calculated from the mass flow data and the compressor geometry. The use of a given compression ratio yields precise determination of the pressure profile through direct integration within the model. The polytropic efficiency is defined as the ratio of the equivalent polytropic head to the effective compression power computed as described above. The Frssling number is a correction factor for convection and is defined as:

Fr = 1 + 0.276 Re Pr1 / 3

(16)

where the Reynolds number is calculated with the average slip velocity between the gas and the droplets. In most previous works related to wet compression technology for power augmentation [7, 8, 9] convection effects were considered negligible and therefore Fr was assumed to be 1. But this assumption would certainly not hold in the present work, because of the large liquid mass fractions and the relatively big droplet sizes. Therefore, convection effect will be accounted for when applying the droplet model to the present experiments. It can be noticed that the two parameters pc and (Fr) cannot be directly measured but they can be obtained by calibrating the droplet model to the experimental data. The main concept of such methodology consists in the determination of efficiency and Fr yielding a perfect matching between the measured power (W) and discharge liquid temperature (Tl) with those computed through the model. For such reason, the problem can be approached as the research of zeros for a non-linear constrained system in the form:

& & 0 = ma (ha 2 ha1 ) + ml [(hs 2 hs1 ) +

(VF2 VF1 ) hev1 + (1 VF2 )(hev1 hev2 )]

(15)

The temperature difference of the liquid phase and the amount of evaporated water are available from the solution of the differential equations of the foregoing time step so that the temperature of the gas phase with feedback from cooling can be computed. The temperature value thus obtained is the new temperature at the end of the time step.

Copyright 2009 by ASME

Downloaded 15 Mar 2010 to 190.177.4.3. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

Tl ,mod Tl ,meas ( pc , Fr ) 0 = Wmod Wmeas ( pc , Fr ) 0

(18)

Given the thermo-dynamical constraints, the problem can be solved using Jacobian method starting from a good guess solution. The convergence to the exact solution is reached on the entire domain and the stability of such methodology has been verified varying the starting iteration point. The outputs of this calibration process are reported in the next Figure 10 and Figure 11, where the efficiency curves for two representative sets of data for D32=75 m. Curves for the other data follow the same trends.
1.5
Dry 5% 10% 20% 30% 40%

1.2

consequent overheating of the liquid phase, the latter being of course utterly useless for any performance needs. Another interesting result of the matching operation is reported in Figure 12 where Fr number for different LVF is plotted as a function of flow coefficient; the general behavior of such curves highlights the effects the growth of convection effect with increasing gas mass flow. This can be explained considering that the relative velocity between the two phases tends to grow along with the gas mass flow because the increase of gas phase discharge velocity is higher than that related to liquid velocity. Given the higher density of the water, the frictional forces acting on the droplets yields an increase in the water velocity that is always lower than that associated to the gas flow with lower density. In addition, it must be noticed that the Fr numbers computed are around 10, enforcing the initial hypothesis around the strong influence of heat convection effect between the two phases. This clearly results in a lower gas discharge gas temperature (Tg) and a higher liquid temperature (Tl).
20
5% 10% 20% 30% 40%

0.9

0.6
16

0.3
12

0 0.4

0.6

0.8

1.2

1.4

1.6
8

Figure 10: Wet Compression Efficiencies for D32=75m, 30000 RPM.


1.5
Dry 5% 10% 20% 30% 40%

Fr
4 0 0.4

0.6

0.8

1.2

1.4

1.6

1.2

Figure 12: Frossling Number for D32=75m, 30000 RPM.

0.9

0.6

0.3

0.4

0.6

0.8

1.2

1.4

1.6

Figure 11 Wet Compression Efficiencies for D32=75m, 25000 RPM.

These figures show that the efficiency drops with respect to dry gas values, and they also clearly demonstrate that when the amount of liquid is increased this drop becomes more pronounced. The reduction in the machine efficiency as the mass fraction of liquid increased is due to larger internal losses in the compressor. These large internal losses result in a increased power necessary to compress the gas phase, and in a

Figure 12 and Figure 13 illustrate the measured power breakdown for two representative cases at high mass fraction (LMF=0.40) and low (LMF=0.05), including the losses in the transmission. The measured curves in wet gas conditions are compared to dry gas interpolated data for the same volumetric flow, in order to carry a direct comparison. While the power dissipated in the evaporation process and to pump the liquid phase is negligible, the power that is wasted in heating the liquid phase is substantial, especially at high LMF where it amounts to almost half of the compression power. However, even at the lowest tested LMF the compression power is higher than the equivalent dry case, thus showing that for these experiments the benefits from the inter-cooling effects are never large enough to offset the loss in aerodynamic efficiency.

Copyright 2009 by ASME

Downloaded 15 Mar 2010 to 190.177.4.3. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

1.2

Compr. Power Compr. Power Dry Heating Power Dissipated Power Transmission Power Evap. Power

Power *

0.8

0.6

0.4

0.2

0 0.4

0.6

0.8

1.2

1.4

1.6

Figure 13 Measured power breakdown for D32=75m, LMF=0.05, 30000 RPM.


Compr. Power Compr. Power Dry Heating Power Dissipated Power Transmission Power Evap. Power

1.2

Power *

0.8

0.6

0.4

0.2

0 0.4

0.6

0.8

1.2

1.4

1.6

Figure 14 Measured power breakdown for D32=75m, LMF=0.40, 30000 RPM.

A good understanding of the three-dimensional unsteady multi-phase fluid dynamics that develops in the compressor internal vanes is necessary to understand the physics of the loss mechanism. Since the test vehicle was not instrumented along the internal flow path, so the details of such physical mechanism could not be identified. Plans to perform additional testing with internal instrumentations are underway, as well as detailed CFD computations, in order to discover more insight about the two-phase internal flows. Comparing these data with the ones previously obtained by Brenne et al [11], one can see that the drop in efficiency is more pronounced for our experiments. This is in agreement with the findings in the above works, as the efficiency drops at higher LMF is more pronounced at lower pressures. Such effect is due to the increasing density difference between the gas and the condensate when the suction pressure is reduced. The increasing density difference leads to a considerable increase in the mass fraction of liquid entering the compressor, even at our very low liquid volumetric contents. This is in line with what already shown by the pressure curves.

CONCLUSIONS This paper presented the results of tests performed on a centrifugal compressor operating under wet gas conditions, showing the effects of the liquid mass fraction (gas-volume fraction), impeller speed, droplet size and injection pattern on the performance of the machine. These wet gas tests confirmed the possibility of compressing a gas-liquid mixture with very high liquid content, thus also confirming that the application of centrifugal compressor technology is a viable option for singlestage, two-phase compression at a liquid mass fraction of up to 0.60 for atmospheric inlet conditions. The results of the tests showed that, relative to a dry gas compressor, the compressor pressure ratio initially increased for small amounts of injected liquid (liquid mass fractions equal to or less than 10% at low flow coefficients) but then decreased when the liquid mass fraction was increased within the range of values covered in this program. While the increase in pressure ratio is attributed to the higher density of the fluid that was being handled by the compressor when liquids were injected, the major decrease at higher LMFs is probably due to the extremely large density differences at the ambient inlet conditions and the consequent inability of the impeller, designed for dry gases, to handle a large portion of liquid mass flow. Due to the specific coupling between the atomizing spray nozzle and the inlet pipe geometry, most of the injected water has been introduced into the impeller in the form of a liquid film on the pipe wall, and thus the influence of the droplet size on the machine performance was negligible. The effect of the film generation should be considered in future developments of the analytical model. When the straight inlet pipe was removed and the liquid was injected directly onto the inducer, a stronger relationship between droplet size and pressure ratio has been measured, with smaller droplets giving substantially better performance. This result suggests the need for fine atomization for such type of compressor inlet. The present work also dealt with the problem of accurate measurements of the gas temperature at the compressor exit, which are required to measure the polytropic efficiency. This problem is of difficult solution because of the heavily wet conditions of the flow streams, and because of the strong temperature difference between the gas and the liquid phases that are typically associated with the wet compression process, and to the inherently non-equilibrium nature of the wet compression thermodynamics. Since it is impossible to accurately measure the gas phase temperature under such conditions, a droplet model has been developed, where timedependent conservation equations for mass and energy were solved in order to take into account the transient time for the exchange of heat between the evaporating droplets and the surrounding gas phase. The heat transfer model accounted for both conduction and convective effects related to the slip velocity between the accelerating compressing gas stream and the liquid droplets. This droplet model has then been applied to our experimental setup and has been solved for the appropriate initial boundary conditions. By calibrating the model with experimental data for the liquid temperature at the compressor

10

Copyright 2009 by ASME

Downloaded 15 Mar 2010 to 190.177.4.3. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

exit and the absorbed compression power, reliable measurements of the polytropic efficiency have been obtained. The efficiency drops with respect to dry gas values, and as the amount of liquid is increased this drop becomes more pronounced. The reduction in the machine efficiency as the mass fraction of liquid increased was due to larger internal losses in the compressor. These large internal losses result in increased power required to compress the gas phase, and in a consequent overheating of the liquid phase. A good understanding of the three-dimensional unsteady multi-phase fluid dynamics that develops in the compressor internal vanes is necessary to understand the physics of the loss mechanism. Plans to perform additional testing with internal instrumentation are underway, as are detailed CFD computations in order to provide more insight on the two-phase internal flows. Future tests at real subsea conditions are planned in order to establish the relationship between wet compression performance at ambient conditions in open loop and the performance of the same machine in closed loop at typical subsea inlet pressures (p=70-150 bar). Multi-stage wet compression testing is of course also planned.

11. Brenne, L., Bjrge, T., Gilarranz, J. L, Koch, J.M., Miller, H., Performance of a Centrifugal Compressor Operating Under Wet Gas Conditions, 34th Turbomachinery Symposium, Houston, USA 12. Hundseid, O., Bakken, L.E., Grner, T.G., Brenne, L. and Bjrge, T., Wet Gas Performance of a Single Srage Centrifugal Compressor, ASME Paper No. GT-2008-51156 13. Young, J. B.: The fundamental equations of gas-droplet multiphase flow. International Journal of Multiphase Flow 21, 175191 (1995). 14. Berlemont, A.; Grancher, M.-S.; Gouesbet, G.: On the Lagrangian simulation of turbulence influence on droplet evaporation. International Journal of Heat and Mass Transfer 34, 2805-2812 (1991). 15. Faeth, G. M.: Evaporation and combustion of sprays. Progress in Energy and Combustion Sciences 9, 1-76 (1983). 16. Dean, J. A.: Langes handbook of chemistry, 9th edition.McGraw-Hill, New York (1956). 17. VDI.: VDI Heat Atlas, 6th edition. VDI Verlag, Dsseldorf (1991). 18. ASME: "Performance Test Code on Compressorsand Exhausters", Standard PTC 10-1997.

ACKNOWLEDGMENTS The financial support of the General Electric Company is gratefully acknowledged. We are very grateful to our colleagues Massimo Camatti, Michael Schmitz, Vittorio Michelassi and Sergio Boris. We have gained invaluable insight from discussion with them.

REFERENCES
1. Brenne, L., Bjrge, Bakken and Hundseid, O., " Prospect for Sub Sea Wet Gas Compression", ASME Paper No. GT-2008-51158. 2. Sanjeev Jolly, Wet Compression A Powerful Means of Enhancing Combustion Turbine Capacity, Power-Gen International, Orlando, FL, 2002. 3. Sanjeev Jolly, Performance Enhancement of GT 24 with Wet Compression, Power-Gen International, Las Vegas, NV, 2003. 4. Bhargava, R. and Meher-Homji, C.B., Parametric Analysis of Existing Gas Turbine with Inlet evaporative and Overspray Fogging, ASME Paper No. GT2002-30560. 5. Bhargava, R., Bianchi, M., Melino, F., and Peretto, A, Parametric Analysis of Combined Cycles Equiped with Inlet Fogging, ASME Paper No. GT-2003-38187. 6. Zachary, E., and Hudson, D., Method and Apparatus for Achieving Power Augmentation in Gas Turbine via Wet Compression, US Patent 5,867,977, The Dow Chemical Company, 1999. 7. Hartel, C. and Pfeiffer, P., Model Analysis of High-Fogging Effects on the work of Compression, ASME Paper No. Gt-200338117. 8. Zheng, Q., Sun, Y., Li, S., and Wang, Y., Thermodynamic Analysis of Wet Compression Process in The Compressor of a Gas Turbine, ASME Paper No. GT-2002-30590. 9. Abdelwahab, A., "An Investigation Of The Use Of Wet Compression In Industrial Centrifugal Compressors", ASME Paper No. GT-2006-90965. 10. Shibata, T., Takahashi, Y., and Hatamiya, S., "Inlet Air Cooling with Overspray applied to a Two-Stage Centrifugal Compressor", ASME Paper No. GT-2008-50893.

11

Copyright 2009 by ASME

Downloaded 15 Mar 2010 to 190.177.4.3. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

You might also like