NUMERICAL AND EXPERIMENTAL ANALYSIS OF THERMAL VAPOR

COMPRESSOR (TVC) PERFORMANCE UNDER OFF-DESIGN CONDITIONS

Muhammed Adnan Saroosh, Charbel Siklawi, Hyunchul Chung, Gabjin Jun

DOOSAN R&D Center, Al Turki Business Park 9th Floor, Al Khobar 31952, Eastern Province,
Kingdom of Saudi Arabia
muhammedadnan.saroosh@doosan.com

Summary

Thermal Vapor Compression (TVC) performance is very crucial in MED-TVC seawater desalination
units (MED: Multi Effect Distillation) in order to get designed distillate production and guaranteed GOR
(Gain Output Ratio = distillate production [kg] / steam consumption [kg]). TVCs are basically single
point designed and are appraised for their stability and robustness when operating under designed
conditions. However, the performance of TVC at off-design motive steam condition is also of interest to
understand the stability of the designed TVC and to predict the TVC performance under partial load
operation of the desalination plant. To better clarify this issue, DOOSAN has conducted experimental
and numerical studies on typical TVCs since 2004 and on high temperature (HT) TVCs since 2010. This
work is concerned with the numerical and experimental investigation of TVC performance under
various load conditions. 2D-axisymmetric and 3D Computational Fluid Dynamics (CFD) simulations
are performed for a fixed TVC geometry using Fluent 6.3 with varying motive pressures. In this
modeling the k-ω SST turbulence model is employed.
The experimental studies are performed on the 0.1 MIGD (Million Imperial Gallon per Day, 1 MIGD =
4,546 m3/day water production) HT MED-TVC pilot plant installed by DOOSAN in Changwon, South
Korea. The pilot plant is huge in size as it is constructed to ensure that experimental data correspond to
industrial MED units. Results show that numerical values are in close agreement with experimental
values for designed and near designed operating conditions. Comparison between 2D and 3D
simulations reveals slight deviation in the pressure curves however the entrainment ratio showed higher
values with the 2D simulations.

I. INTRODUCTION

Thermo-compressors are compressible fluid ejector devices, where the phase condition for both
motive stream and suction load are gas phases (steam and vapor). The thermo-compressor is a static heat
pump which, when used in MED desalination plants, it recycles part of the vapor produced in one of the
intermediate effects by mixing it with live steam (2.5 to 50 bar absolute) into higher pressure vapor to be
used as heating media for the first effect. The steam (vapor) densities change as pressure and
temperatures change within the thermo-compressor. Typically, the velocity of motive steam emerging
from the nozzle is supersonic, and the velocity remains supersonic after mixing with suction stream, thus
compression shock occurs in the diffuser.
Since the early nineties it has been shown by many researchers that it is necessary to improve the
performance of ejectors in order to make them economically more attractive. Hence, a deep
comprehension of their operation is essential. Several experimental investigations have been carried out
to analyze the effect of ejector geometry on its performance, such as nozzle exit location [1-3]; mixing
chamber/nozzle area ratio [4-6], nozzle design [6, 7], etc. On the other hand, optimal design is not

The International Desalination Association World Congress on Desalination and Water Reuse 2013 / Tianjin, China
REF: IDAWC/ TIAN13-284
simple, due to the complex nature of the fluid flow mechanisms and its sensitivity to operational
conditions. A deviation from design conditions can affect its performance very negatively. A review on
ejector theory and applications can be found in [8].
Most of the theoretical works carried out used extensions of the 1D constant pressure mixing
ejector theory, first developed by Keenan et al. [9]. This method is based on a number of assumptions
such as ideal gas behaviour, negligible supply, downstream velocities, etc. Moreover, it was assumed
that mixing between primary and secondary flows occurs at constant pressure starting from the point
where the secondary fluid gets chocked at a certain distance from the nozzle exit. Thermodynamic
irreversibilities were accounted by introducing constant efficiencies (eg. primary nozzle efficiency noz)
at various points along the ejector [10, 11]. In some cases, these constants were selected arbitrarily [12],
in others taken from literature [11] or selected such that experimentally measured performance data give
the best fitting with the predicted model values [4].
CFD tools have been proved to be valuable tools for understanding and analysing complex fluid
flow problems (eg. the entrainment and mixing processes in ejectors). Bartosiewicz et al. [13] compared
the pressure distribution by using different turbulence models for the simulation of an ejector with
experimental data. It was concluded that, for certain conditions, simulated results were in excellent
agreement with measured data. The effect of ejector geometry on the flow field was investigated. It was
pointed out that according to CFD model results, 1D model assumptions were not met under test
conditions. Sriveerakul et al. [14] simulated a steam ejector under different operating conditions. It was
concluded that the effective area, where the secondary fluid gets chocked, exists as assumed in the 1D
model; however it is very difficult to locate. Pianthong et al. [15] analysed the performance of a steam
ejector for different operating conditions and some geometrical factors (nozzle exit position and throat
length). It was demonstrated that a computationally demanding 3D model resulted in similar pressure
distribution compared to a less demanding axi-symmetrical case.
In this study a constant pressure type of ejector is employed, since it is known that the constant-
pressure ejector has a better performance than the constant-area ejector and is thus widely used. In early
2000s, DOOSAN developed TVC design program based on the 1D theory of Huang et al. 1998 [4] and
adjusted the empirical coefficients from pilot experiments and CFD simulations since 2004.
Thermo-compressors are basically single point designed; the design and operation are best at one
specified design condition of suction pressure and load, motive-steam pressure, and vapor rates. If the
operating pressure and therefore motive-steam flow are varying, the performance of TVC will vary and
this will affect the overall performance and reliability of MED desalination plant. Therefore, DOOSAN
has conducted CFD simulation to predict the performance at off-design conditions, e.g. during design
phase of Yanbu 15MIGD MED-TVC project in Saudi Arabia. Since 2008, DOOSAN has investigated
its effort on the study of higher TBT MED-TVC system, and higher compression ratio TVC design and
performance validation has been of interest. In this study, some recent results of both numerical (CFD)
and experimental works by DOOSAN, in order to gain better understanding for TVC behavior including
off-design conditions, are presented.

II. Ejector operation and performances

A typical ejector is presented in Figure 1. A boiler supplies high-pressure vapour at the ejector
primary inlet (e). Expansion to supersonic conditions of this primary stream, takes place in a
convergent–divergent nozzle, creating a depression and drawing in the vapour from the evaporator,
which forms the secondary stream (e). Primary and secondary streams come into contact in the mixing
chamber. Complete mixing is assumed to occur in this chamber and the resulting flow is generally
supersonic, such that on entering the constant cross-section zone, a normal shock wave occurs,
The International Desalination Association (IDA)
World Congress on Desalination and Water Reuse
REF: IDAWC/ TIAN13-284

-- 2 --
accompanied by a significant pressure rise. The location of the mixing section depends on operating
conditions and ejector geometry, but should ideally take place somewhere in the constant area section
(m) or in the beginning of the diffuser. In the diffuser the mixed fluid is then decelerated and
recompressed.

M ix in g c h a m b e r C o n v e rg in g C o n s ta n t- a re a s e c tio n D iffu s e r
d uct ( M ix in g tu b e )
L cd L mt Ld
S e co n d a ry  cd
flu id
Motive Discharge
(g) Fluid D sc D no D m t (m) D d (c)
Mixed Flow
D nt

(e)

Fluid
From
Evaporator

Figure 1: Ejector schematics

A detailed description of the mixing process is not simple, since the motive fluid flow is
characterised by a series of oblique/normal shock waves called the shock train [13], after expanding
through the primary nozzle. During this process its static pressure tends to increase until it levels with
the pressure of the secondary fluid. After the mixing process is completed, the resulting flow undergoes
a final shock resulting in subsonic conditions. The pressure is then further increased in the subsonic
diffuser. The exit pressure (c) depends on the conditions in the first effect. 1D models are not capable of
predicting the shock train phenomena, although it affects the performance of an ejector significantly.

Ejector performances is often measured by the entrainment ratio defined as:

md
E/R  (1)
mm

where md and mm are respectively the mass flow of the discharge steam and the motive steam.
The efficiency of TVC ejector can be defined also by two other parameters such as Compression Ratio
and Expansion Ratio defined as:

P
CR  d (2)
Pm

P
ER  s (3)
Pm

where Pd is the discharge pressure, Pm the motive pressure and Ps the suction pressure. Several
methods are available in the literature to evaluate entrainment ratios; most of these methods need
The International Desalination Association (IDA)
World Congress on Desalination and Water Reuse
REF: IDAWC/ TIAN13-284

-- 3 --
lengthy computation procedures and use many correction factors. Two main simple methods are used to
evaluate this ratio:
- Power's graphical data method (Graham Curves, etc.),
- Semi-empirical model [16]

Although Power's method is a straightforward and the entrainment ratio can be extracted
directly from the graph in terms of compression ratio and expansion ratio, it is too difficult to use in such
simulation models.

III. Experimental Setup

In 2004, DOOSAN installed 0.1MIGD MED-TVC pilot plant (Figure 2) and a separate TVC
experimental facility in Changwon, Korea. Various experimental studies for design verification and
TVC performance tests have been conducted. In this study, the 0.1MIGD MED-TVC pilot plant is
employed to understand overall plant performance under off-design conditions. This pilot plant has five
Effects, TVC, condenser, three pre-heaters, two ejector condensers, and plate-type heat exchangers for
winter use. TBT (Top Brine Temperature) is designed as 85°C where BBT (Bottom Brine Temperature)
is controlled as the set value of around 67°C, after the rehabilitation in 2010. 5 and 6 bara motive steam
TVCs have been used for the tests. Steam is supplied from a 3 ton/hr boiler system, and TVC is placed
between the last and the first Effect. The experimental data are collected from pressure gauges reading
placed at different positions along the TVC wall as illustrated in Figure 3. Readings are logged when
steady state condition is reached for each operating condition.

Figure 2: DOOSAN 0.1MIGD MED-TVC Pilot Plant

The International Desalination Association (IDA)

World Congress on Desalination and Water Reuse
REF: IDAWC/ TIAN13-284

-- 4 --
 Figure 3: Pressure gauges positions along TVC wall

IV. CFD Model

Fluid flow in the ejector is typically compressible and turbulent. The functional relationship
between the three major unknown variables temperature (T), pressure (p) and velocity vector (v)
describing compressible flow of isotropic Newtonian fluids, is given by the conservation of energy,
momentum and continuity equations, in the form of a set of partial differential equations (PDE):

Continuity:
D
 .div ( v )  0 (4)
Dt
Momentum:
Dv
  . g  p  ij (5)
Dt
Energy:
Dh Dp v
   div( kT )  ij i (6)
Dt Dt x j
In Equations (5) and (6) the term ij can be written as:

 v v 2 
ij   i  i  Idiv(v)  (7)
 x j xi 3 
 

The turbulent behaviour was treated using k-ω SST turbulence model. Where ω is the turbulent
kinetic energy dissipation rate. The realisable version of the k- ω SST turbulence model was chosen in
this work, since it has been applied by other authors for ejector simulation [13, 14]. It was reported to
predict more accurately the spreading rate of jet flows and provides better performance for separation
and recirculation. The actual form of the resulting transport equations for the turbulent kinetic energy (k)
and the turbulent kinetic energy dissipation rate can be found in the Fluent 6.3 User’s Guide. In order to
solve Equations (5) and (6), proper boundary conditions must be applied. Pressure boundary conditions
were applied on the inlets and outlet according to saturation conditions depending on the temperatures in
the generator, evaporator and condenser of the ejector cooling cycle. Heat transfer through the walls was
neglected (zero heat flux). In this work, a commercial package, Fluent 6.3, was used to simulate fluid
flow in the ejector. In Fluent, the space domain is subdivided into a number of small control volumes
called finite volumes. For each finite volume each PDE is transformed into a set of algebraic equations
and then integrated using numerical techniques. The unknown quantities were calculated for each cell
The International Desalination Association (IDA)
World Congress on Desalination and Water Reuse
REF: IDAWC/ TIAN13-284

-- 5 --
centre using a combination of a segregated and a coupled algorithm. The simulated ejector was designed
using a 3D and 2D models.
In the 2D model axi-symmetry assumption was used [15]. In order to optimise the CFD model,
several mesh densities were examined from the finer to the coarser.

III. Results and discussion

TVC simulation is a crucial part in designing MED-TVC desalination plants offering the
possibility to check the performance of TVC while avoiding the time and expenses incurred from
experimental testing. To this end, axi-symmetrical (2D) TVC simulation were performed to study the
performance of the 0.1 MIGD MED-TVC pilot plant. In order to validate the TVC simulation, the 2D
simulations results are compared to partial load and full load experimental data. Moreover, 2D
simulation results are evaluated by using 3D simulations data to check the accuracy of 2D model.

3.1 Validation of the 2D Model

The numerical simulation is performed using Fluent 6.3 with k-ω SST turbulence model. The 2D
axi-symmetric simulation results are validated against the experimental data in Figure 4. This figure
shows that the calculated TVC wall pressure curves are in close agreement with the measured values
under typical designed conditions.

TVC exp. results of April 21st comapred to numerical results

1

0.9 Num. sim. with logged data @ 1:25am

Num. sim. with logged data @ 3:30am

0.8 exp. 21st april @ 1:25 am

exp. 21st april @ 3:30 am

0.7

0.6
[Bar]
pressure

0.5
pressure

0.4

0.3

0.2

0.1

0
0 1 2 3 4
X
x [m]
Figure 4: Comparison between wall pressure: experimental data (dot) and simulation profile (line)

The corresponding pressure contour from 2D axi-symmetric simulation is shown in Figure 5

showing the variation of the pressure along the ejector.

The International Desalination Association (IDA)

World Congress on Desalination and Water Reuse
REF: IDAWC/ TIAN13-284

-- 6 --
 Figure 5: Pressure contour from 2D axi-symmetric simulation

3.2 Partial Load Performances

When partial load performance of desalination plant is required the consumption of steam is
usually decreased. In MED-TVCTVC configuration, however, it’s not that easily expected because the TVC
performance should be estimated first. In here, TVC performancess are simulated for the same
configuration withth reduced motive steam quality (lower pressure). Concerning the Partial LoadL TVC
experimental tests, wall pressure profile is determined based on the pressure gauges located as
mentioned in Figure 3. The obtained data are used as boundary conditions in the CFD simulations (using
k- SST turbulence model).. Figure 6 shows a comparison between the simulated results and the
experimental readings for a partial load corresponding to 4 bara (28.55% performance).
performance) In this study the
performance is related only to TVC and not to the whole MED plant. For some positions a good
agreement between experimental and numerical data are highlighted. On the other hand, few points
show some gaps (Figure 6). The difference between the results is attributed to the high level of
fluctuation in the experimental data which in turn is due to partial load conditions. Periodic oscillations
were observed on TVC wall pressure gages, indicating the overall vapor flow in TVC is unsteady under
partial load condition, while CFD simulation is based on steady flow assumption.

The International Desalination Associat

Association (IDA)
World Congress on Desalination and Water Reuse
REF: IDAWC/ TIAN13-284

-- 7 --
 pressure [Bar]

x [m]

Figure 6: Comparison for partial load experimental data and simulation profile

Figure 7 shows the variation of the pressure along TVC ejector at different motive steam
pressures. The Entrainment ratio was determined for all the simulated cases and it was shown that this
parameter is very sensitive to the motive steam pressure. In fact, a variation of 2 bara in motive steam
pressure induces to a variation of more than 70% in Entrainment Ratio.

Figure 7: Pressure values along the TVC axis for different motive pressure.

After CFD simulation, we iterated heat and mass balance to determine the corresponding plant
load operation condition; the data are summarized in Figure 8.

The International Desalination Association (IDA)

World Congress on Desalination and Water Reuse
REF: IDAWC/ TIAN13-284

-- 8 --
 100

75

Partial Load Operation

50

25

0
4 4.2 4.4 4.6 4.8 5 5.2 5.4 5.6
Motive Steam Pressure

Figure 8: Partial Load Operation versus Motive Steam Pressure

3.3 Comparison between 2D and 3D Models

2D Model allows reducing time calculation; 23 hours are required to perform the 2D simulation
(against 1975 hours for 3D simulation). Axi-symmetrical simulations neglect the third dimensionality in
calculating the results. In some cases this affects the end results; in fact in many researches 2D
turbulence models are unable to give a good prediction of the real life flow results. In order to validate
our choice of doing axi-symmetrical 2D simulations, a 3D TVC simulation is conducted with very fine
mesh to remove any ambiguity related to meshing discrepancies. The same mesh used for 2D simulation
is employed for the 3D simulation after it was convoluted over a 90 degree angle to depict one quarter of
the full geometry (the TVC having 2 symmetrical surfaces). Both simulations used the k-w SST
turbulence models and same boundary conditions. The validation of 2D model was done by comparing
the calculated axial pressure values for 3D and 2D simulation (see figure 9).

Comparison of Axial Pressure profiles

1

3 D simulation
2 D simulation

0.8

0.6
Pressure [Bar]

0.4

0.2

0
0 1 2 3 4
X [m ]

Figure 9: Axial pressure profiles comparison between 2D and 3D simulations

The International Desalination Association (IDA)
World Congress on Desalination and Water Reuse
REF: IDAWC/ TIAN13-284

-- 9 --
This Figure shows an excellent agreement between the both results (the difference can be neglected) and
2D simulations can be used for faster calculation. Comparison between 2D simulation and 3D
simulations highlights the following:
- 2D results show lower values in the mixing region (Figure 9). The 3D results have the peaks
ahead of the 2D results and this can be attributed to the 3D effect.
- In the wall pressure profiles (Figure 10), the 2D results show somehow lower values of wall
pressure along the mixing region of the TVC and it is overlapping with 3D results in all the
other sections.
- The axial Mach number profiles for 2D and 3D simulations (Figure 11), highlights that 2D
simulation shows higher value of velocities.

In the performed 3D model, we considered quarter of TVC configuration with symmetric

manner. This was done for the comparison purpose related to 3D effect especially on turbulence model
with same mesh resolution. Hence, some differences with experimental data, in terms of suction inflow
direction and duct shape for our 3D simulation, were highlighted.

Com pariso n of W all P ressu re P rofiles

1
3 D sim ula tion

2 D sim u la tion
0.8

0.6
Pressure [Bar]

0.4

0.2

0
0 1 2 3 4
X [m ]

Figure 10: Axial pressure profiles comparison between 2D and 3D simulations

The International Desalination Association (IDA)

World Congress on Desalination and Water Reuse
REF: IDAWC/ TIAN13-284

-- 10 --
 Comparison of Axial Mach Number Profiles
4

3D simulation
3
2D simulation

Mach Number M
mach-number
2

0
0 1 2 3 4
X [m]

Figure 11:: Mach number profiles comparison between 2D and 3D simulation

From the previous profiles it is shown that the flow field for thehe 3D simulation evolved faster
than the 2D simulation and this is approved by the contour plots (Figures 12-16).. Figure 121 presents the
static pressure contour plots showing that the 3D simulation contour lines are ahead in distance
distan of the
2D results. Figure 13 shows the velocity contour plots for both simulations with 2D maximum velocity
value of 1060 m/s higher than 3D simulation result of 1043 m/s. Figure 114 shows the Mach number
contour plots and both simulation look similar in trend. It is shown in Figure 15 the turbulent viscosity
contour plots for 2D and 3D simulations. Finally
inally the specific dissipation rate contour plots are shown in
Figure 16.

Figure 12:: Static pressure contour plots [Pa] for 2D and 3D simulations.

The International Desalination Associat

Association (IDA)
World Congress on Desalination and Water Reuse
REF: IDAWC/ TIAN13-284

-- 11 --
Figure 13:: Velocity contour plots [m/s] for 2D and 3D simulations

Figure 14:: Mach Number contour plots for 2D and 3D simulations

The International Desalination Associat

Association (IDA)
World Congress on Desalination and Water Reuse
REF: IDAWC/ TIAN13-284

-- 12 --
 Figure 15:: Contours of turbulent viscosity [kg/m.s] for 2D and 3D simulations

Figure 16:: Contour plots of specific dissipation rate, ω [1/s] for 2D and 3D simulations

All the obtained results show a maximum difference of about 4.5 % between results of 3D and
2D simulation showing again the reliability of 2D Model
Model.

IV. CONCLUSIONS

TVC operation performance is of high importance to ensure reliable and optimum production of
MED-TVCTVC seawater desalination unit. In this study, the numerical investigation of TVC is validated
against the experimental results under design and off-design conditions. The main contribution of this
investigation is to show the validity of 2D Model compared to 3D and experimental data. The T 3D
simulation showed that three dimensional effects have small yet safely
afely negligible consequence on the
final results given the
he longer time needed for running the simulations. For future cases, 2D TVC
simulation is more convenient and reliable to be conducted whenever possible, however when
The International Desalination Associat
Association (IDA)
World Congress on Desalination and Water Reuse
REF: IDAWC/ TIAN13-284

-- 13 --
conducting simulation for multi-nozzle TVC the 3D aspect should be used to model the geometry
correctly therefore a mesh optimization is needed to leverage between the time consumption and
reliability. Finally, this investigation shows that with numerical and experimental studies, the
performance of commercial huge size MED-TVC system can be predicted more reliably.

V. REFERENCES

1. Aphornratana, S., Eames, I.W., A small capacity steam-ejector refrigerator: experimental

investigation of a system using ejector with movable primary nozzle. Int. J. Refrig 20 (5), 1997,
352–358
2. Chunnanond, K., Aphornratana, S., An experimental investigation of a steam ejector refrigerator: the
analysis of the preessure profile along the ejector. Appl. Therm. Eng 24, 2004, 311–322.
3. Eames, I.W., Ablwaifa, A.E., Petrenko, V., Results of an experimental study of an advanced jet-
pump refrigerator operating with R245fa. Appl. Therm. Eng 27, 2007, 2833–2840.
4. Huang B.J., Chang J.M., Wang C.P., Petrenko V.A., A 1-D analysis of ejector performance,
International Journal of Refrigeration, 22, 1999, 354-364.
5. Sankarlal, T., Mani, A., 2007. Experimental investigations on ejector refrigeration system with
ammonia. Renew. Energy 32, 2007, 1403–1413.
6. Chang, Y.-J., Chen, Y.-M,. Enhancement of a steam-jet refrigerator using a novel application of the
petal nozzle. Exp. Therm. Fluid Sci. 22, 2000, 203–211.
7. Eames, I.W., A new prescription for the design of supersonic jet pumps: the constant rate of
momentum change method. Appl. Therm. Eng 22, 2002, 121–131.
8. Chunnanond, K., Aphornratana, S., Ejectors: applications in refrigeration technology. Renew.
Sustain. Energy Rev. 8, 2004, 129–155.
9. Keenan, J.H., Neuman, E.P., Lustwerk, F., An investigation of ejector design by analysis and
experiment. ASME J Appl. Mech 72, 1950, 299–309.
10. Cizungu, K., Mani, A., Groll, M., Performance comparison of vapour jet refrigeration system with
environment friendly working fluids. Appl. Therm. Eng. 21, 2001, 585–598.
11. Cizungu, K., Groll, M., Ling, Z.G., Modelling and optimization of two-phase ejectors for cooling
systems. Appl. Therm. Eng. 25, 2005, 1979–1994.
12. Tyagi, K.P., Murty, K.N., Ejector-compression systems for cooling: utilising low grade waste heat.
Heat Recovery Syst 5, 1985, 545–550.
13. Bartosiewicz, Y., Aidoun, Z., Desevaux, P., Mercadier, Y., Numerical and experimental
investigations on supersonic ejectors. Int. J. Heat Fluid Flow 26, 2005, 56–70.
14. Sriveerakul, T., Aphornratana, S., Chunnanond, K., Performance prediction of steam ejector using
computational fluid dynamics: Part 2. Flow structure of a steam ejector influenced by operating
pressures and geometries. Int. J. Thermal Sci. 46, 2007, 823–833.
15. Pianthong, K., Sheehanam, W., Behnia, M., Sriveerakul, T., Aphornratana, S., Investigation and
improvement of ejector refrigeration system using computational fluid dynamics technique. Energy
Conv. Manage 48, 2007, 2556–2564.
16. El-Dessouky, H., Ettouney, H., Alatiqi, I., Al-Nuwaibit G. Evaluation of steam jet ejectors,
Chemical Engineering and Processing, 41, 2002, 551–561.

The International Desalination Association (IDA)

World Congress on Desalination and Water Reuse
REF: IDAWC/ TIAN13-284

-- 14 --