AN INNOVATIVE TWO-PHASE FLOW PUMP AND SEPARATOR SOLUTION

AN INNOVATIVE TWO-PHASE FLOW PUMP AND SEPARATOR SOLUTION

François Gruselle Université Libre de Bruxelles Aero-Thermo-Mechanics Department

and

Fonds pour la formation à la Recherche dans l’Industrie et dans l’Agriculture (FRIA)

Brussels, Belgium [email protected]

Johan Steimes Université Libre de Bruxelles Aero-Thermo-Mechanics Department

Brussels, Belgium [email protected]

Patrick Hendrick Université Libre de Bruxelles Aero-Thermo-Mechanics Department

Brussels, Belgium [email protected]

ABSTRACT

The Aero-Thermo-Mechanics (ATM) department of Université Libre de Bruxelles (ULB) develops a new system to simultaneously pump and separate a two-phase flow, in particular oil/air mixtures. Two-phase flows are encountered in many applications (oil extraction, flow in nuclear power plant pumps, pulp and paper processing) but the study is mainly focused on aeroengine lubrication systems. The main objective is to obtain a compact and efficient system that can both extract the gas of a two-phase flow and increase the pressure of the liquid phase. Particular care is given to the liquid flow rate lost at the gas outlet of the system. A large range of gas/liquid volume ratio has been studied, leading to different two-phase flow regimes at the inlet of the system (slug, churn or annular flow).

After successful tests with water-air prototypes, which have allowed to identify the key design and working parameters, the technology has been implemented for a hot oil- air mixture. This paper presents the test results of the first oil/air prototype under real in-flight operating conditions. The tests with oil/air mixtures were performed on the aeroengine lubrication system test bench of the ATM department. The identification and implementation of appropriate two-phase flow rate measurement systems is an essential contribution to the project. Two attractive measurement systems have been considered: a Coriolis density meter for the volume fraction at the liquid outlet and radio-tracing elements for the measurement of the oil consumption at the air outlet.

In parallel, the flow field in the pump and separator system has been studied with commercial CFD (Computational Fluid Dynamics) software packages. The choice of the two-phase flow model is highly dependent on the two-phase flow regime.

But different regimes can simultaneously exist in the pump and separator system. So, the Eulerian two-phase flow model, the most complex and general model, seems to be the most appropriate. A coupling of this model with a dispersed phase model is under investigation to take all two-phase flow phenomena into account.

Key words: two-phase flow, separation, pumping, experiments, multiphase CFD simulations.

1. INTRODUCTION 1.1 Motivation

The ATM department of ULB develops an integrated device to simultaneously pump and separate a two-phase flow, in particular oil/air mixtures. Many natural and man-made applications deal with two-phase flows (oil extraction, flow in nuclear power plant pumps [1-3], pulp and paper processing) but this study is mainly focused on the aeroengine lubrication systems [4, 5]. This paper presents the test results of the first oil/air prototype under real in-flight operating conditions.

In aircraft gas turbine engines, the lubrication and cooling of gears and shaft bearings located in the engine sumps and gearboxes are performed using oil injection into sealed bearing chambers. Aeroengine bearing chambers are often sealed by Proceedings of ASME Turbo Expo 2011

GT2011 June 6-10, 2011, Vancouver, British Columbia, Canada

GT2011-46917

air-blown seals pressurized by compressor bleed air. The sealing air escapes the bearing chamber through a dedicated vent line. The oil flows out of the chamber through the bottom of the sump and is sent back to the oil tank using scavenge pumps. The two outlets of the chamber (the vent and the offtake at the bottom of the sump) have to deal with an oil-air mixture, not a single-phase flow. A prior concern is the oil loss through the vent line which must be minimized to limit the environmental impact and because the in-flight oil quantity is limited by the tank size.

The scavenge circuit of an aeroengine lubrication system is usually composed of scavenge pumps and a cyclone separator located at the inlet of the oil tank [6]. This separator aims to separate the air-oil mixture collected from the sumps. The shaft can also be used for this de-aerating function. A typical aeroengine lubrication system is shown on Fig. 1. The two- phase flow pump and separator system in development could be used in the scavenging lubrication system of classical aircraft gas turbine engines.

FIGURE 1: CLASSICAL LUBRICATION SYSTEM [6]

1.2 Previous work

Prototype design. The first steps of the development of this system are described in a previous paper [7] which aims to understand the relevant parameters affecting the efficiency of this kind of device. Several prototypes with a pump and a separator that are integrated into a unique device were developed and tested. The first few prototypes were using a water/air mixture instead of a hot oil/air flow. When the technology was mature enough, it has been implemented for oil/air mixtures.

In this integrated pump and separator system, the two- phase flow separation is a forced centrifugal separation (unlike cyclone separators for which the tangential inlet causes a natural swirling flow [8]). This solution takes advantage of the rotary motion of the pump to separate variable gas/liquid ratios and flow rates in a compact and efficient way.

The design of the pump and separator systems is similar to an axial-centrifugal pump (Fig. 2). The axial part is used to separate the two phases of the flow in order to collect the liquid phase in the centrifugal part, after the separation has been performed. Indeed, the liquid is pumped more efficiently when the gas is removed before [1]. The liquid phase, heavier, is thrown against the outer wall (stator or carter), while the gas is recovered in the center of the axial part (inside the rotor). The liquid phase is moved towards the centrifugal part where it gets pressure from the impeller before going out of the system passing through a volute. The gas going out through the center of the axial part is released into the atmosphere by a dedicated outlet. So, the gas exhaust pressure is not regulated and the pressure in the core of the axial part is simply the sum of the atmospheric pressure and the pressure losses in the exhaust pipe. The centrifugal forces generate a liquid film on the outer wall; its thickness depends on the pressure at the liquid outlet and the rotating speed.

FIGURE 2: WORKING PRINCIPLE OF AXIAL-CENTRIFUGAL PROTOTYPES

Theoretical approach. The qualitative description of the prototype working principle may be supplemented with a theoretical formulation [7]. A schematic of the working principle of the gas-liquid separator is shown in Fig. 3. The following assumptions are made:

 The gas-liquid separation is perfect.

 The position of the interface is one-dimensional.

 The friction losses inside the system are negligible.

 The axial and radial velocities of the liquid are not taken into account in the momentum equation because they are negligible relative to the tangential velocity.

With these assumptions, the momentum equation inside the impeller enables the calculation of the position of the gas-to- liquid interface R, as a function of the pressure at the impeller outlet (Pout) and the angular velocity ω:

FIGURE 3: WORKING PRINCIPLE OF AXIAL-CENTRIFUGAL PROTOTYPES: LIQUID FILM FORMATION

Centrifugal part Axial part

P

− P

=

ρ

ω

(R

− 𝑅

)

and the outer radius of the impeller R₂.

This equation proves that the position of the gas/liquid interface is regulated by two main parameters: the back- pressure P_out and the angular velocity ω (P_in, ρ_liquid and R₂ are supposed to be constant). All the other geometric and flow parameters (flow rates, slip coefficient, viscosity, friction coefficients, geometric characteristics as blade angles, etc.) are secondary effects and have a limited impact on the position of the interface with respect to P_out and ω. So, a low back-pressure at the volute outlet means a thin liquid film and a high back- pressure means a thick liquid film. There is a kind of self- regulation of the liquid film thickness according to the needed back-pressure.

Experimental results. Previous experiments [7] have shown that the gas volume fraction at the liquid outlet decreases with an increasing thickness of the liquid film because gas bubbles have to cross a thicker liquid ring.

Otherwise, if the position of the gas-liquid interface goes under the radial position of the gas outlet, liquid flows through the gas outlet. But, no liquid could be observed at the gas outlet under usual operation conditions.

In the range of operating conditions that could be tested, the separation efficiency of the axial-centrifugal prototype was very high (over 99% in volume fraction). These results, obtained with simple measurement systems, have been confirmed by visual observations of the two-phase flow in the different prototypes and at their two outlets. But, as the air quantity at the liquid outlet and the liquid quantity at the air outlet are very small, new volume fraction measurement systems were needed to evaluate more precisely the quality of the separation. Some technologies have been selected and are being developed for this project.

1.3 New prototype, tests and operating conditions After successful tests with water-air prototypes, the technology has been implemented for a hot oil-air mixture. The tests with oil-air mixtures were performed on the aeroengine lubrication system test bench of the ATM Department. This test bench is being equipped with more accurate separation measurement systems which allow to have a better characterization of the prototype efficiency. Two attractive measurement systems have been considered: a Coriolis density meter for the volume fraction at the liquid outlet and radio- tracing elements for the measurement of the oil consumption at the air outlet.

The first oil/air prototype has been developed and designed for real in-flight operating conditions. In particular, the impeller diameter, the rotating speed and the air flow rate have been increased in comparison with the previous water/air prototypes.

Also, these new operating conditions are more critical and cause, in a general way, an increase of the liquid quantity at the gas outlet. In practical terms, the oil consumption (which is a prior issue) has increased in comparison with the water/air prototypes [7].

This paper presents the test results of the first oil/air prototype under real in-flight operating conditions. In particular, the main factors impacting the oil consumption will be discussed. As the oil is present as drops at the air outlet, the dynamics of these oil droplets (creation, trajectory, escape) are also investigated in this study. Indeed, the new operating conditions lead to different two-phase flow regime at the inlet of the system. The two-phase flow regime in the inlet pipe is, now, mainly an annular flow, characterized by the presence of a continuous liquid film flowing on the channel wall and surrounding a central gas core laden with entrained liquid droplets [9]. Only the smallest oil droplets should be able to follow the air stream and to go out via the air outlet.

1.4 Numerical simulations

CFD simulations are also being performed with the following long-term objectives:

 Study the separation performance and internal phenomena with numerical simulations,

 Validate a numerical tool in order to quickly study other separator designs,

 Use numerical simulations for design optimization.

The choice of the two-phase flow model is highly depending on the two-phase flow regime. But different regimes can simultaneously exist in the pump and separator system:

 Annular flow at the inlet of the prototype,

 The two phases may eventually be mixed when attacking the blades,

 Film deposition on outer walls (gas-liquid interface),

 Main air flow laden with entrained liquid droplets (dispersed phases),

 Small air bubbles in the oil flow at the liquid outlet (volute outlet).

Unfortunately, there is currently no multiphase model that can simulate all these two-phase flow regimes simultaneously. Even the Eulerian two-phase flow model, which is the most complex and general model, is not well suited to track the motion of oil droplets because droplet cloud is treated as a continuous medium.

So, the actual objective is to study the capabilities of two multiphase models, the Eulerian model and the dispersed phase model, and to use them to simulate two major two-phase flow mechanisms of the integrated pump and separator system: the development of the liquid film and the behavior of oil droplets in the air flow (oil consumption is a prior concern).

Previous simulations have already shown that the Volume- of-Fluid (VoF) model is able to simulate the development of the

liquid film on the outer wall of the system [7]. But this model is quite limited because only a pure phase can exist at any inlet or outlet. So, the VoF model cannot simulate, for example, the presence of air bubbles at the liquid outlet.

2. EXPERIMENTAL STUDY 2.1 Design of the prototypes

The pump and separator system developed in this study has been designed for real in-flight operating conditions, considering a medium-to-small civil aeroengine. Therefore, the oil flow rate is typically between 1000 l/h and 4000 l/h; the air flow rate between 10 g/s and 70 g/s; the rotating speed between 3000 RPM and 10000 RPM.

The design of the oil/air pump and separator prototype is similar to the one represented in Fig. 2, except that the inlet is a double tangential inlet (two inlets placed at opposite sides). As explained before, the centrifugal impeller aims to pressurize the liquid phase and is followed by a volute chamber. The volute outlet is, therefore, named the oil outlet. The second outlet, for the air flow, is axial.

The main dimensions of the prototype are calculated with Eq. 1 (depending on the requested outlet pressure and the rotating speed) and using good practice rules of maximum air flow velocity in pipes. The mean outer diameter of the axial blades is typically 150 mm. The centrifugal blades outer diameter is around 160mm. The width of the impeller (centrifugal part) is around 5mm. The diameter of the air outlet is typically 40 mm (it depends on the air flow rate to avoid important head losses). For most of the prototypes, the blades of the axial part are radial. Indeed, for this study, the pumping efficiency is not essential with respect to the separation efficiency. Thus, the shapes of the axial and centrifugal blades have not been optimized.

2.2 Test bench

The ATM department has developed a test bench for components of aeroengine lubrication systems [10]. It is mainly used to test mature oil-air pump and separator devices. This bench has been modified to test mature oil/air pump and separator systems in real in-flight operating conditions. Figure 4 shows a schematic of this test rig.

The two prototype inlets receive an oil/air mixture in variable proportions. As the prototype has two inlets, two mixing chambers are used in parallel and are fed on one side with air from a compressor and on the other side with oil from a tank. The oil flow can be heated up to 150°C by an electric resistance heater (28 kW) placed upstream of the mixing chambers. Oil and air flow rates are measured by flowmeters located in the inlet pipes (flowmeters accuracy: 1 to 2% of measured value). The inlet pressure is taken as the sump pressure, with calibrated pressure transducers (accuracy: 0.5%

full scale for the water bench, 0.2% for the oil bench). PT100

sensors are used on the whole oil bench. The mixing chamber outlets are then connected to the prototype inlets. Pressure transducers and PT100 are also used at the outlets. All the acquisition is made with a 4-20mA National Instruments acquisition card and is post-processed using LabView.

FIGURE 4: LUBRICATION TEST BENCH SCHEMATIC As the two-phase separation is not perfect, one separation efficiency measurement system is needed at each outlet. The void fraction at the oil outlet is measured thanks to an Emerson Coriolis flow meter, which can perform a continuous measurement of the fluid density (ρ_fluid) of multiphase flows [11]. The oil volume fraction (α) is then calculated with Eq. 2.

α =

oil−ρ_air

where ρ_air is the air density which depends on the pressure and the temperature; ρ_oil is the oil density (pure oil) which is a known property (depending on the temperature).

This system is widely used in the industry to manage process where air bubbles can appear. Its accuracy for density measurement is 0,2 g/m³. Visualization methods [12], tomography [13], wire-mesh sensors [14] and optic fiber techniques [15] were also envisaged, but are more costly or inadequate. The results given by the Coriolis flow/density meter have been confirmed by visualization of the two-phase flow through a borosilicate tubular sight glass placed just downstream of the Coriolis meter (Fig. 5).

At the gas outlet, the oil flow rate can be measured by two methods. The first one is simply a weighting method. The oil droplets are collected on a homemade filter/blotting paper system (named oil collector in Fig. 4). The blotting paper is simply weighed before and after the test. This measurement system is being validated by using a new oil consumption measurement method based on the use of radiotracers

compounds that are mixed to the oil [16]. As a first step, oil is labeled by adding a radiotracer and a monitoring system is installed near the exhaust line (air outlet) where the marker will be trapped in a filter. The measurement consists of monitoring, in real-time, the amount of tracer trapped in this filter.

Amplitude of the detected signal is proportional to the oil consumption. This radiotracer method has the advantage that the results are very accurate (this system can typically measure oil flow rate of 10 g/h) and can be analyzed on-line. Also, it is not limited by the temperature. In addition, the quantity of marker added to the oil (< 1ml) is such that oil properties remain unchanged.

FIGURE 5: VISUALIZATION OF THE OIL/AIR FLOW DOWNSTREAM OF THE CORIOLIS DENSITY METER

Research is still going on to get a better knowledge of the two-phase flow at the air outlet. It goes into the same direction as [4], i.e. to use laser Doppler measurement system to detect the micro-bubbles. This technique represents a well-established method for determining droplet size distributions of polydisperse sprays. The objective is to obtain an in depth understanding of the key parameters which drive the oil consumption in order to reduce it.

2.3 Results

Several tests have been performed to study the variables which impact the separation efficiency at the two outlets of the pump and separator system. Different rotating speeds, liquid and gas flow rates have also been used to evaluate their effect on the separation efficiency, but, currently, the oil is kept at ambient temperature. Tests at rotating speeds higher than 5000 RPM could not be performed because the maximum power of the electrical motor was reached (3,2 kW). But rotating speed can be increased up to 10 000 RPM if the oil flow rate is decreased.

Separation efficiency at oil outlet. Tests have been performed to study the parameters impacting the quality of the oil outlet. This quality is given by the oil volume fraction measured by the Coriolis density meter, α. Figure 6 shows, for

different rotating speeds, the oil volume fraction at oil outlet in function of the liquid film thickness which is estimated using Eq. 1 (air flow rate = 0; oil flow rate = 2500l/h).

FIGURE 6: SEPARATION EFFICIENCY AT THE OIL OUTLET;

FILM THICKNESS INFLUENCE

This figure clearly shows that the oil outlet separation efficiency highly depends on the oil film thickness (which only depends on P_out and ω). Indeed, according to Eq.1, the oil film thickness (R₂-R) which is generated on the outer walls increases with Pout. The air quantity at the oil outlet decreases when the thickness of the liquid film increases because air bubbles have to cross a wider liquid ring. But if the outlet pressure is too low, large air bubbles are sucked into the impeller and go out through the volute outlet. Figure 6 also shows that the separation efficiency at the volute outlet does not (directly) depend on the rotating speed.

Figure 7 shows the influence of the air flow rate on the oil outlet separation efficiency for different rotating speeds and back-pressures (or pressure coefficient c_p defined in Eq. 3). It shows that the air flow rate has a minor influence on the separation efficiency at the air outlet. If the air flow rate increases, the air quantity at the volute outlet increases slightly (in agreement with [7]). This phenomenon can be explained by a decrease of the oil film thickness when the air flow rate is decreased. Indeed, when air flow rate increases, the core pressure of the prototype increases (P_in in Eq.1). But, as the volute outlet pressure (P_out) is kept constant, the film thickness decreases which leads to have more air at the oil outlet.

c

=

2ρ_oil(ωR₂)²

In conclusion, the film thickness is the key parameter for the separation efficiency at the oil outlet. If the film thickness is superior to a few millimeters, the separation at the oil outlet is efficient.

In Out

Air bubbles

Pure oil

FIGURE 7: SEPARATION EFFICIENCY AT THE OIL OUTLET;

AIR FLOW RATE INLFUENCE

Separation efficiency at air outlet. Tests have also been performed to study the parameters impacting the quality of the air outlet. It is given by oil quantity collected at the air outlet (also named the oil consumption). The measurement is done by the blotting paper method. The results are also being validated by the radiotracer measurement system.

Figure 8 shows the influence of the air flow rate on the separation efficiency at the oil outlet. For these tests, the oil flow rate is equal to 2500 l/h. Two different rotating speeds and back-pressures have been tested.

FIGURE 8: SEPARATION EFFICIENCY AT THE AIR OUTLET;

AIR FLOW RATE INLFUENCE

This figure clearly demonstrates a strong influence of the air flow rate on the oil consumption. This result is very important because oil consumption of the pump and separator prototype is an essential concern for engine manufacturers. So, to reduce the oil consumption without reducing the air flow rate, one solution is to increase the flow section in order to decrease the air velocity. A trade-off between size and performance has to be found.

Figure 8 also shows that, for high air flow rates, the oil consumption is slightly better at 4500 RPM than at 3500 RPM.

Indeed, the centrifugal forces which generate the oil/air separation are proportional to ω². This figure also shows that the back-pressure P_out (or pressure coefficient c_p) has not a clear influence on the oil consumption. However, the oil will directly flow out by the air outlet if the back-pressure exceeds Pout-max:

P_out−max= P_in+¹

2ρ_liquidω²(R²₂− R²_air−outlet) (4) with Rair-outlet the radius of the air outlet section.

The oil liquid flow rate has also an influence on the separation efficiency at the air outlet, but not as marked as the air flow rate. In a general way, if the oil flow rate is divided by 2, the oil consumption will be divided by 3/2. This phenomenon will be investigated more thoroughly in future tests.

2.4 Discussion on oil droplets mechanisms

The results obtained with this new oil/air prototype are quite different from those obtained with the previous water/air prototypes, especially in terms of oil consumption. This oil is present at the air outlet as droplets, not as a continuous liquid flow. Figure 8 shows that the oil consumption is virtually equal to zero if the air flow rate (or air velocity) is low. This result explains why no liquid could be observed at the air outlet of previous water/air prototypes which had been tested with air flow rate inferior to 20 Nm³/h.

This difference in the operating conditions between the water/air and oil/air prototypes can be expressed in terms of two-phase flow regime at the prototype inlet. This latter is represented in Fig. 9; the operating conditions of oil/air prototype are included in the red zone and those of the water/air prototype in the blue zone. The original map was prepared by Willenborg et al. [4] to characterize the dominating flow patterns from the global air and oil flow rates and the physical properties like viscosity, density and surface tension of air and oil. The corresponding flow patterns are depicted in Fig. 10.

Figure 9 shows that the two-phase flow regime is quite different for both cases. The two-phase flow regime of the oil/air prototype is mainly an annular flow, characterized by the presence of a continuous liquid film flowing on the channel wall and surrounding a central gas core laden with entrained liquid droplets [17]. For water/air prototype, the flow regime is either a churn flow or a slug flow. The latter is characterized by large Taylor bubbles separated from one another by slugs of liquid, which may include small bubbles. Increasing the velocity, the slug flow becomes a churn flow and the structure of the flow becomes unstable.

The important point concerning the inlet two-phase flow regime is the presence of droplets for the oil/air prototype, not for the water/air prototypes. The droplet diameter is not constant; it can vary from a few microns to several hundreds of

microns. Only the smallest oil droplets should be able to follow the air stream and to go out via the air outlet.

FIGURE 9: TWO-PHASE FLOW MAP FOR INLET SECTION FLOW, VERTICAL UPWARD [4]

FIGURE 10: TWO-PHASE FLOW PATTERN FOR VERTICAL UPFLOW [17]

In annular flows, the distribution of droplet diameters depends on air and oil velocities, pipe diameter and orientation, pressure, oil viscosity, etc. Azzopardi et al. [18] shows that an increase of the gas velocity causes a reduction in the mean drop size. Therefore, the more the air flow rate increases the more the droplet diameter decreases and the more the centrifugal force decreases but, at the same time, the drag force increases as the air velocity increases. This explains why the air flow rate has such an impact on the oil consumption (Fig 8). On the other hand, the drop size rises when increasing the liquid flow rate (except for low gas velocity cases) or when increasing the pipe diameter.

The trajectory of each droplet entering into the system will depend on the centrifugal forces and the air drag force [19]. It is obtained by integrating the force balance on the particle (Eq. 5):

du_i^p

dt

= F

(u

− u

) +

p

+

p

u_i^pis the i^th component of the particle velocity,

u_i is the i^th component of the continuous phase (air) velocity, gi is the i^th component of the acceleration due to gravity, ρp is the density of the particle,

ρ is the density of the continuous phase (air),

F_i contains the additional forces (pressure gradient, rotating reference frame, Brownian motion, Saffman lift, etc.).

The term FD(ui− u_i^p) is the drag force per unit particle mass. For small oil droplets and high air velocity (ui), this term is the dominant term of the Eq. 5 and will force the particles to go out via the air outlet.

F

=

pd_p² C_DRe_p

24

Re

=

μ is the dynamic viscosity of the continuous phase (air), dp is the diameter of the particle,

CD is the drag coefficient (can be estimated from literature), Rep is the relative Reynolds number of the particle.

Currently, one of the most precise ways to perform droplet trajectory calculations, in such a complex device, is to use Lagrangian CFD simulations (section 3.2). Those simulations are being performed to determine the operating conditions which force some particles to go out via the gas outlet. The study of the oil droplet distribution in the inlet section is important to define correctly the inlet boundary conditions of the CFD simulations.

3. NUMERICAL STUDY

CFD simulations are also being performed with Ansys Fluent 6.3 in order to simulate two major two-phase flow mechanisms of the integrated pump and separator system: the development of the liquid film and the behavior of oil droplets in the air. For these simulations, the rotary motion of the pump and separator system is modeled using the “Rotating Reference Frame” method (a similar CFD study of multiphase separation can be found in [20]). The mesh represents a typical geometry of oil/air prototypes. It includes around one million hexahedral cells (the mesh size was limited by the memory capacity of available computers). The volute chamber has not been modelled to simplify the calculations but has been simply replaced by a disk to improve the stability of calculations (Fig.

11). The diameter of the two inlets is 40mm. The axial air outlet diameter is 52mm. The radial oil outlet diameter is 300mm. The diameter of the axial part is 160mm. The external diameter of the blades is 174mm. The model includes six radial blades. The flow was modeled as isothermal and incompressible. Oil Oil/air prototype

Water/air prototype

density is 950 kg/m³ and its dynamic viscosity is 0.005 kg/ms.

Air density is 1.225 kg/m³ and its dynamic viscosity is 1.7894e- 5 kg/ms. The turbulence model was the k - ε model, RNG, swirl dominated flow.

FIGURE 11: SCHEMATIC OF THE CFD GEOMETRY

FIGURE 12: CROSS SECTION VIEW OF THE MESH The appropriate multiphase numerical model depends on the two-phase flow regime. There exist two general approaches for the numerical calculation of multiphase flows [19]: the Euler-Lagrange approach and the Euler-Euler approach. In the former, the main phase is treated as a continuum by solving the time-averaged Navier-Stokes equations and the dispersed phase is solved by tracking a large number of particles (bubbles or droplets) through the main phase. The fundamental assumption with the Lagrange approach is that the dispersed phase has a low volume fraction. In the Euler-Euler approach, the different phases are treated mathematically as interpenetrating continua.

The concept of volume fraction occupied by each phase is introduced. The volume fractions of the different phases are assumed to be continuous functions of space and time and their sum is equal to one.

3.1 Eulerian two-phase flow simulations

For these Eulerian simulations, “soft” operating conditions (low rotating speed and low air flow rate) have been voluntarily used to avoid converge problems: the rotation speed is 1000 RPM and the boundary conditions are the followings:

 Inlet air flow rate is 25 Nm³/h;

 Inlet oil flow rate is 2750 l/h;

 Inlet volume fraction is 0.5;

 Oil outlet pressure is 15 kPa;

 Atmospheric pressure has been imposed at the air outlet.

The simulation begins with steady state single-phase calculations (pure oil). As the oil outlet pressure is low, an oil back-flow appears at the “air” outlet. When the convergence is reached, the Eulerian model is activated and an unsteady formulation is used. Then, an air back-flow appears at the air outlet. The calculation is continued until the air back flow stops and the position of oil/air interface becomes stable. Then, air is also injected and, therefore, the two inlets receive an oil/air mixture. The calculation is continued until the monitored values (outlet flow rates, interface position, etc.) are stabilized.

Figure 13 shows the contours of the gas volume fraction when the back-pressure at the centrifugal oil outlet is 15 kPa (relative pressure). The most important result of this figure is the formation of the oil/air interface on the outer walls of the axial part. As the back-pressure is quite low, the liquid film has a small thickness and some air penetrate into the centrifugal part. Figure 13 also shows that the natural centrifugal separation in the inlet part of the system is less efficient than the forced centrifugal separation in the rotor part. Indeed, an oil film appears on the outer wall of the inlet part but the air flow is still laden with oil unlike in the rotor part where the air flow is pure.

FIGURE 13: CONTOUR OF AIR VOLUME FRACTION WITH POUT = 15 KPA (CROSS SECTION VIEW)

Air outlet Inlet

Blade Oil outlet

Internal wall

Figure 14 shows the contours of the gas volume fraction when the back-pressure at the centrifugal oil outlet is 20 kPa (relative pressure) and when the inlet air flow rate is 50 Nm³/h (everything else being constant). Compared to Fig. 13, the liquid film is thicker, air in the centrifugal part has disappeared and centrifugation in the inlet part is better (because inlet air velocity is higher).

FIGURE 14: CONTOUR OF AIR VOLUME FRACTION WITH POUT = 20 KPA (CROSS SECTION VIEW)

3.2 Lagrangian two-phase flow simulations

Steady state Lagrangian CFD simulations are also being performed with the Discreted Phase Model (DPM) of Ansys Fluent 6.3. These simulations aim to study the motion of oil droplets in the pump and separator system. The long term objective is to reduce the oil consumption using DPM simulations to optimize the design of the system. Currently, DPM simulations are mainly used to determine the droplets which are collected by centrifugation and those which go out via the air outlet. The diameter of the largest droplet which is present at the air outlet is defined as the critical droplet diameter. Therefore, the more the separation is efficient the more the critical droplet diameter is low.

The operating conditions of these DPM simulations are:

 Rotation speed is 4500 RPM;

 Inlet air velocity is 20 m/s (around 180 m³/h);

 Inlet droplet velocity is 10 m/s;

 Inlet oil flow rate (droplets) is 250 l/h;

 Oil outlet is defined as a wall;

 Atmospheric pressure has been imposed at the air outlet.

Droplets are injected in the two inlet sections (25 injection points, 5 x 5 grid, at each inlet). As explained in section 2.4, the real droplet diameter distribution and the entrained oil flow rate depend on air and oil velocities, pipe diameter and orientation, pressure, oil viscosity, etc. So, to simplify this preliminary

study, an arbitrary diameter distribution has been used. This distribution has been chosen to be able to identify the critical droplet diameter. So, the parameters of this Rosin–Rammler distribution are:

 Mean droplet diameter is 5 µm;

 Minimum droplet diameter is 1µm;

 Maximum droplet diameter is 20 µm;

 Spread factor is 3.5;

 Number of diameters is 10.

500 particle trajectories are computed. The interaction between continuous phase and particles is a two-ways interaction.

Figure 15 shows the particle trajectories in the pump and separator system. Most of particle streams are not entrained through the air outlet: on 500 particle streams, 30 particles go out by the air outlet. This first DPM simulation gives a critical droplet diameter equal to 15µm. This result cannot be validated for the moment. Furthermore, this study has to be extended to different rotating speed, air velocity and droplet distribution. In a few months, a laser Doppler measurement system will be used to characterize the droplet distribution at the air outlet and, therefore, to validate the DPM simulations.

FIGURE 15: INFLUENCE OF PARTICLE DIAMETER ON SEPARATION AT AIR OUTLET

4. CONCLUSIONS AND PERSPECTIVES

Different prototypes of pump and gas/liquid separator systems have been successfully developed. The tests with oil/air mixtures were performed on the aeroengine lubrication system test bench of the ATM Department. Two attractive measurement systems have also been implemented: a Coriolis density meter for the volume fraction at the liquid outlet and radio-tracing tools for the measurement of the oil consumption at the air outlet.

The two key parameters of the separation process are the back-pressure at the volute outlet and the rotation speed. They define the liquid film thickness. The air quantity at the liquid outlet mainly depends on this liquid film thickness. On the

other hand, the air flow rate is also essential for the oil consumption. If the air velocity is high, oil droplets are entrained by the air flow. The smallest droplets will follow the air stream and go out through the air outlet.

In parallel, the two-phase flow field has been simulated using Ansys Fluent. The development of the liquid film was successfully simulated with the Eulerian model even if this model is quite diffusive for the calculation of the interface position. New simulations with more realistic operating conditions will be performed soon. DPM simulations are also being performed to obtain a better knowledge of oil droplets trajectories. First results are encouraging but this study has to be continued and validated by comparison with experiments.

NOMENCLATURE

Acceleration due to gravity g

Additional forces F

Angular velocity of the impeller ω

Density of air

ρ

Density of the continuous phase ρ Density of fluid (oil/air mixture)

ρ

Density of liquid phase

ρ

Density of oil

ρ

Density of particle ρ_p

Drag coefficient CD

Dynamic viscosity of the continuous phase μ

Oil volume fraction α

Outer radius of the impeller R₂

Particle diameter d_p

Particle velocity u^p

Pressure at the impeller outlet P_out Pressure at the inlet of the system P_in

Pressure coefficient

𝑐

Radial position of the gas-liquid interface R Radius of the air outlet section R_air-outlet Relative particle Reynolds number Rep

Velocity of the continuous phase u

REFERENCES

[1] Chan, A., Kawaji, M., Nakamura, H., and Kukita, Y, 1999,

“Experimental study of two-phase pump performance using a full size nuclear reactor pump,” Nuclear Engineering and Design, 193, pp.

159–172.

[2] Furuya, O., 1985, “An analytical model for prediction of two-phase (non-condensable) flow pump performance,” ASME Journal of Fluids Engineering, 107 (1), pp. 139–147.

[3] Noghrehkar, G., Kawaji, M., Chan, A., Nakamura, H., and Kukita, Y, 1995, “Investigation of centrifugal pump performance under two- phase flow conditions,” ASME Journal of Fluids Engineering, 117, pp.

129–137.

[4] Willenborg, K., Klingsporn, M., Tebby, S., Ratcliffe, T., and al, 2008, “Experimental analysis of air/oil separator performance,”

Journal of Engineering for Gas Turbines and Power, 130.

[5] Eastwick, C. N., Simmons, K., Wang, Y., and Hibberd, S., 2006,

“Study of aero-engine oil-air separators,” Proc. IMechE Part A: J.

Power and Energy, 220 (7), pp. 707-717.

[6] Linke-Diesinger, A., 2008, Systems of Commercial Turbofan Engines: An Introduction to Systems Functions, Springer Berlin Heidelberg, Berlin, chap. 3.

[7] Gruselle, F., Steimes, J., and Hendrick, P., 2010, “Study of a Two- Phase Flow Pump and Separator System”, Proc of ASME Turbo Expo 2010: Power for Land, Sea and Air., Glasgow.

[8] Hoffmann, A.C. and Stein, L.E., 2008, Gas Cyclones and Swirl Tubes, Springer, 2nd ed., New York.

[9] Cioncolini, A., Thome, J. R., 2010, “Prediction of the entrained liquid fraction in vertical annular gas–liquid two-phase flow”, International Journal of Multiphase Flow, 36, pp. 293-302.

[10] Hendrick, P., Berten, O., Gruselle, F., Raimarckers, N., and Saive, G., 2008, “Test benches of innovative systems and components for aircraft gas turbine engines,” in Lubmat Conference, San Sebastian.

[11] Emerson, Company web site: www.emerson.com

[12] Klasinc, R., Hocevar, M., Baicar, T., and Sirok, B., 2007,

“Visualization method for volume void fraction measurements in gas- liquid two-phase flows of a water turbine outlet channel,”

Computational Methods in Multiphase Flow IV, 56, pp. 249-259.

[13] Takenakaa, N., Asanoa, H., Fujiia, T., and Matsubayashib, M., 1998, “Three-dimensional visualization of void fraction distribution in steady two-phase flow by thermal neutron radiography,” Nuclear Engineering and Design, 184, pp. 203–212.

[14] Prasser, H.-M., Böttger, A., and Zschau, J., 2001, “Bubble size measurement using wire-mesh sensors,” Flow Measurement and Instrumentation, 12, pp. 299–312.

[15] Kiambi, S.L., Duquenne, A.M., Bascol, A. and Delmas, H., 2001,

“Measurements of local interfacial area: application of bi-optical fiber technique,” Chemical Engineering Science, 56, pp. 6447-6453.

[16] DSI, Company web site: www.deltabeam.net

[17] Thome, J. R., 2010, Heat Transfer Engineering Data Book III, Wolverine Tube Inc., www.wlv.com/products, chap. 12.

[18] Azzopardi, B. J., Piearcey, A. and Jepson D.M., 1991, “Drop size measurements for annular two-phase flow in a 20 mm diameter vertical tube”, Experiments in Fluids, 11, pp. 191-197.

[19] Crowe, C.T., 2006, Multiphase Flow Handbook, CRC Press Taylor & Francis Group, USA.

[20] Chochua G. and Maier W., 2007, “Computational and Analytical Study of Multiphase Rotary Separator Turbine Line Outlet”, Proc of ASME Turbo Expo 2007, Montreal, Canada.