International Journal of Heat and Mass Transfer 84 (2015) 642–652

Contents lists available at ScienceDirect

International Journal of Heat and Mass Transfer

journal homepage: www.elsevier.com/locate/ijhmt

A 3D method to evaluate moisture losses in a low pressure steam

turbine: Application to a last stage
Xinggang Yu, Zhihuai Xiao, Danmei Xie ⇑, Chun Wang, Cong Wang
School of Power and Mechanical Engineering, Wuhan University, Wuhan, China

a r t i c l e i n f o a b s t r a c t

Article history: This paper presents a physically consistent 3D method to evaluate moisture losses and it is further
Received 7 July 2014 employed to estimate the moisture losses in the last stage of a 1000 MW fossil-fired steam turbine.
Received in revised form 10 January 2015 The 3D method is based on the information of the flow-field and droplets deposition. An inhomogeneous
Accepted 10 January 2015
multi-phase flow model is adopted to simulate the condensing flow in the last stage. Massive
computations of droplets deposition by both inertial and turbulent mechanisms on the stationary blade
and moving blade have been conducted for different droplet diameters at the inlet of the last stage. In our
Keywords:
work, the moisture losses are divided into six categories namely thermodynamic loss, drag loss of fog
Moisture losses
Steam turbine
droplets, drag loss of coarse droplets, impact loss, capturing loss and centrifuging loss. The effect of
3D method the fog droplet size at the inlet of the last stage on the moisture losses has been analyzed and the results
Fog droplet size indicate that the overall moisture losses rise and the relative fractions of each category of the moisture
Last stage losses vary against the increase of the fog droplet diameter.
Ó 2015 Elsevier Ltd. All rights reserved.

1. Introduction
The steam in low-pressure steam turbines of thermal power
plants is generally expanded across saturation line and the rear
stages operate in wet-steam conditions. The droplet nucleation
process and the subsequent droplet laden flow in steam turbine
cause additional energy dissipations which are collectively referred
to as moisture losses [1]. A full understanding of the moisture loss
mechanisms is necessary for an accurate prediction of turbine
performance. The moisture losses in steam turbine have been
examined by many researchers. Baumann [2] firstly attempted to
quantify the moisture losses and proposed a simple empirical
correlation which suggested that a 1% reduction in dry isentropic
efficiency for each 1% mean wet ness fraction. However, the
empirical approach offers no insight into the moisture loss
mechanism [3].
Needless to say, the physically based predict ion of the moisture
losses in steam turbine requires the knowledge of the condensing
flow-field and droplets deposition. A theoretical, physic ally consis-
tent one-dimensional (1D) approach to evaluate moisture losses in
steam turbine was proposed firstly by Gyarmathy [4]. Since then,
similar treatments were presented by Laali [5] and Moore [6].
These approaches did not consider the highly three-dimensional
⇑ Corresponding author.
E-mail address: whupwr@foxmail.com (D. Xie).
(3D) flow in low pressure stages, e.g. for large power steam
turbines. In recent years, the advances in computer hardware
and CFD technology have made it possible for a 3D calculation of
condensing flow-field in steam turbine and thus a 3D analysis of
the moisture losses. Starzmann et al. [7] made a tentative predic-
tion of the moisture losses in a three-stage model low pressure
steam turbine by a 3D two-phase CFD modeling, but only thermo-
dynamic loss, kinematic and braking losses were considered. Petr
and Kolovratnik [8] employed a statistical two-dimensional (2D)
evaluation method to compute the moisture losses based on the
3D computational flow-field in a 1000 MW nuclear and a
210 MW fossil-fired low pressure steam turbine.
The moisture losses are related to the distribution of fog and
coarse droplets in steam turbine.
The coarse droplets originate from the fog drop lets deposition.
Therefore, an understanding of the droplet deposition process is
necessary for the prediction of the moisture losses. In general,
the mechanisms of fog droplet deposition onto the turbine blades
can be divided into the inertial impaction and turbulent diffusion.
The deposition of small particles in turbulent pipe flow has been
the subject of a large number of investigations but a few papers
focused on fog droplets deposition in steam turbine, for example
the work by Gyarmathy [4], Crane [9], Yau [10], Young [11,12]
and Starzmann [13]. Yau and Young [10,11] extended the theories
of diffusional deposition proposed by Wood [14] to predict the
diffusional deposition of fog droplets on the steam turbine blades
and employed a quasi-3D Lagrangian particle track method to

http://dx.doi.org/10.1016/j.ijheatmasstransfer.2015.01.066
0017-9310/Ó 2015 Elsevier Ltd. All rights reserved.
 X. Yu et al. / International Journal of Heat and Mass Transfer 84 (2015) 642–652 643

Nomenclature

A area (m2) Greek symbols

CD drag coefficient / volume fraction
D droplet diameter (m) c vapor specific heat ratio
h static enthalpy cd volumetric mass concentration of droplets (kg/m3)
J nucleation rate (m3 s1) e Kantrowitz’s non-isothermal correction factor
Kn Knudsen number q density (kg/m3)
K Boltzmann’s constant (=1.38071023 J/K) r surface tension (N/m)
L latent heat (kJ/kg) s droplet relaxation time (s)
kc thermal conductivity of the vapor s+ dimensionless droplet relaxation time
m mass of one water molecule (kg) sw fluid wall shear stress (N/m2)
m_ condensing mass flow rate l dynamic viscosity (kg/(m s))
N number of droplets per unit volume m kinematic viscosity (m2/s)
p Pressure (N/m2) f ratio of droplet and fluid rms fluctuating velocity nor-
Q deposited droplets mass flow per unit area (kg/(m2 s1)) mal to the surface
Q_ heat flow rate (W) dd fractional deviation
R gas specific constant x angular velocity (rad/s)
ks blade roughness
ReD droplet Reynolds number Subscripts
T temperature (K) c vapor phase
u velocity (m/s) d liquid phase
u friction velocity (m/s) i,j tensor notation
r radius (m) D diffusional deposition
z loss coefficient I inertial deposition
t time (s) cw coarse water droplets
nd number of liquid phases tip tip of blade
SD,d source term u circumferential direction
DS entropy increase n normal direction
DP loss
qc condensation coefficient
Vþ dimensionless deposition velocity
x spatial dimension (m)
G Mass flow rate

determine the inertial deposition. Starzmann [13] implemented the current work, three liquid phases are defined to model the inlet
the diffusional theory of Yau and Young [10] into a full 3D droplets, droplets nucleated in the stationary blade domain and
multi-momentum two phase model to calculate the diffusional droplets nucleated in the moving blade domain respectively. The
deposition of fog droplets on the stationary blades of the last stage standard shear stress transport (SST) model is employed to model
of a model low-pressure steam turbine. the turbulence and mixing planes are used between the stationary
In this work, the deposition of liquid droplets by the turbulent and rotating parts to perform steady flow simulations.
diffusion and inertial impact on the stationary and moving blades
of the last stage of a 1000 MW fossil-fired steam turbine are calcu- 2.1. Inhomogeneous multi-fluid model
lated by the method of Yau and Young [10] and a number of full-3D
Euler–Lagrangian computations. Based on the results of droplets To accurately simulate the behavior of liquid phases, the slip
deposition and the 3D multi-phase flow solutions, a physically velocity between vapor phase (c) and water droplets (d) is consid-
consistent 3D method to evaluate the moisture losses is presented ered by the employment of the inhomogeneous multi-fluid model.
and six categories of the moisture losses are determined for the last The conservation equations of the momentum for both phases are
stage. In addition, the effect of fog droplets size at the inlet of the given by:
last stage on the magnitude and relative fractions of each category
of the moisture losses is analyzed. @ qc /c ui;c @ @ð/sij Þc @p X nd Xnd
þ ðq / uj;c ui;c Þ ¼  /c  _ i;d þ
mu SD;d
@t @xj c c @xj @xi d¼1 d¼1

2. Mathematical models
ð1Þ

@ qd /d ui;d @ @ð/sij Þd @p
The condensing two-phase flow calculations are carried out þ ðq / uj;d ui;d Þ ¼  /d _ i;d  SD;d
þ mu ð2Þ
with the commercial implicit solver ANSYS CFX 12.1. The solver @t @xj d d @xj @xi
uses an Euler–Euler method to predict the condensing steam flow. where / represents the volume fraction, m_ is the condensing mass
The coupling between conservation equations of mass, momentum flow rate due to nucleation and growth of existing droplets and sij is
and energy for the vapor phase and liquid phase is realized via a the viscous stress tensor. The source term SD,d equates to the
source term formulation [15]. An advantage of the multi phase drag force for a single spherical droplets times the number of
model is the possibility to realize a ‘‘source specific’’ droplet repre- droplets N.
sentation [16]. Users are allowed to define several liquid phases
1 p
and activate or deactivate each liquid phase to nucleate for the SD;d ¼ N  C D D2 qc jui;d  ui;c jðui;d  ui;c Þ ð3Þ
calculated domains separately. For the last stage consider ed in 2 4
644 X. Yu et al. / International Journal of Heat and Mass Transfer 84 (2015) 642–652
The non-dimensional drag coefficient CD is obtained according to
Schiller and Naumann model having the form as follows [17]:
24  
CD ¼ 1 þ 0:15Re0:687
D ð4Þ
ReD
q ju u jD
where ReD ¼ c i;dl i;c
c
Eq. (4) is applicable to droplets in the range of 0.1 < ReD < 1000.
When the ReD value is larger than 1000, CD is clipped to 0.44. More
information about other conservation equations is given in [18].
2.2. Non-equilibrium condensation model
In this work, only the homogeneous nucleation is considered.
The number of droplets formed during the condensing process is
obtained from the classical nucleation theory with the non-
isothermal correction of Kantrowitz [19]:
 12 2
! turbines, f can be related to s+ by the curve given in Fig. 3 of
qc 2r q2c 4pr  r
J¼ exp 
1þe pm3 qd 3KT c
where e is the Kantrowitz’s non-isothermal correction factor:
 
c1 L L 1
e¼2  ð6Þ
c þ 1 RT c RT c 2
where qc is the condensation coefficient (generally taken as 1), K is
the Boltzmann constant, m is the mass of one water molecule, c is
the specific heat ratio of vapor, L is the equilibrium latent heat,
and R is the gas specific constant. The surface tension r has signif-
icant effect on the nucleation rate J and the value used here is that
for bulk water with adjustment through NBTF (nucleation bulk
tension factor, r = rbulk  NBTF).
The droplet growth rate describes the variation of droplet sizes
during condensation process and is given as follows [20]:
 
dr kc Td  Tc
¼ ð7Þ
dt r qd ð1 þ cKnÞ hc  hd
where Kn is the Knudsen number defined by the ratio of molecular
mean free path and droplet diameter, kc is the thermal conductivity
of the vapor and c is an empirical factor set to 3.18 [21].
The equations of International Association of Properties of Water
and Steam (IAPWS-IF97) are utilized to model the thermodynamic
properties of water and steam and the surface tension of water.
IAPWS-IF97 can also be extrapolated into metastable regions with
reasonable accuracy and is suitable for non-equilibrium phase
change models.
2.3. Diffusional deposition model
The turbulent-diffusional deposition is the process by which
droplets entrained into the boundary layers migrating to the blade
surface under the influence of turbulent fluctuations in the flow. To
model the diffusional deposition, the method suggested by Yau and
Young [10] is employed and described briefly as following.
The diffusional deposition data is correlated in the non-dimen-
sional form:
V þ ¼ f ðsþ Þ
where V þ ¼ Q =ðcd u Þ is the dimensionless deposition velocity and
sþ ¼ su2 =mc is the dimension less droplet relaxation time. Q is the
mass transfer rate of droplets to the blade surfaces per unit area
and cd is the volumetric mass concent ration of droplets just outside
pffiffiffiffiffiffiffiffiffiffiffiffiffi
the boundary layer. u ¼ sw =qc represents the friction velocity,
whereas sw is the fluid wall shear stress. The kinematic relaxation
time s is derived from the Stokes drag law and can be represented
by the following form with good accuracy as shown by Gyarmathy
[4]:
qd D2
s¼ ð1 þ 2:7KnÞ ð9Þ
18qc tc
With the increase of the dimensional droplet relaxation time s+,
three different deposition regimes are usually identified, namely
turbulent particle-diffusion regime, eddy diffusion-impaction
regime and particle inertia-moderated regime (s+ > 10). The mathe-
matical formulations to calculate the deposition velocity V+ are
given by:
V þ ¼ ðIS þ IB Þ1 sþ < 10 ð10Þ
 
4:42
V þ ¼ 0:56ferfc sþ  10 ð11Þ
fsþ
where f is the ratio of droplet and fluid rms fluctuating velocity
normal to the surface. For the conditions encountered in steam
ð5Þ Ref. [10] and the curve is numerically approximated in this work.
For the method to calculate IS and IB, one can refer to Yau and
Young [10].
As suggested by Wood [14], the outer limit of boundary layer at
a normalized wall distance of y+ = 30 seems to be a good assump-
tion. Therefore, the turbulent deposited mass flow per unit area Q
can be determined by:
Q ¼ V þ u ðcd Þyþ ¼30 ð12Þ
Finally, the integration of Q over the whole blade surface gives the
total diffusional deposition of droplets.
The equations to determine the diffusional deposition of drop-
lets have been implemented within ANSYS CFD POST by the
authors with user-defined Expressions. The multi-phase flow field
computed by the inhomogeneous multi-fluid model (Euler–Euler
model) gives the required parameters to estimate diffusional depo-
sition of water droplets. This way to implement Yau and Young’s
method to a 3D-CFD follows a proposal of Starzmann et al. [13].
2.4. Inertial deposition model
The inertial deposition is a laminar flow phenomenon caused by
the slip velocity between vapor phase and liquid droplets. A
Lagrangian particle track model already implemented in ANSYS
CFX is employed to calculate the inertial deposition of droplets.
The calculations of the inertial deposition are realized by tracking
the path lines of a large number of droplets and identifying the
limiting trajectories impacting on the blade surfaces [22]. The
vapor flow-field is specified by a computation of the inhomoge-
neous multi-fluid condensing flow. The droplets are assumed to
be spherical particles and unaffected by the proximity of other
droplets. The effects of condensation and evaporation are
neglected and the diameter of a droplet is assumed to be constant
when flowing through the domains. Since the liquid density is
much larger than the vapor density, non-drag forces, i.e. virtual
mass force and pressure gradient force, which are significant when
the density of disperse phase is less than the continuous phase’s
density, are neglected and only the viscous drag force on droplets
ð8Þ is considered [11]. The viscous drag force is calculated by Schiller
and Naumann model having the form of Eq. (4). In computing
the deposition rate, it is assumed that droplets incident on
the blading adhering to the surface and there is no rebound of
droplets.
3. Results of droplets deposition
The numerical simulation of the condensing flow in the last two
stages of the 1000 MW fossil-fired steam turbine indicates that the
 X. Yu et al. / International Journal of Heat and Mass Transfer 84 (2015) 642–652
nucleation mainly takes place in the penultimate stage domain at
normal operating condition (Fig. 1). The predicted span-wise distri-
butions of mean droplet diameter D and wetness fraction by the
homogeneous condensation model at the inlet of the last stage
are shown in Fig. 2. It shows a peak value of D around span height
70%, which can be attributed to the low nucleation rate at this loca-
tion in the stator and rotor as shown in Fig. 1. The typical droplet
sizes ranges from 0.05 lm to 0.5 lm. Such small droplets are able
Fig. 1. Contour plot of nucleation rate in the last two stages of the 1000 MW steam
turbine. The surfaces for the contour plot are obtained by offsetting 2 mm from the
suction surfaces of blades.
Fig. 2. The predicted span-wise distributions of mean droplet diameter and liquid
mass fraction at the inlet of the last stage by the last two stages calculation through
Euler–Euler multiphase model.
645
to follow vapor streamlines well. However, for larger droplets, the
slip velocity between the liquid phase and vapor phase becomes
larger and the droplets are less able to follow the steam flow. In
order to investigate the droplet size effect, droplets larger than
those predicted by steady flow calculations are considered. As a
consequence, all the calculations about the droplet deposition
and the estimation of moisture losses are limited to the last stage.
The droplet diameter at the inlet of the last stage is increased by a
multiplication factor (such as 2, 4, 6, 8, 12, 16,. . .), whereas the
liquid mass fraction is kept constant.
The inlet conditions of other parameters are taken from the
results of the last two stages computations by the Euler–Euler
multi-phase model. For the penultimate stage, the inlet steam is
in superheated state and the mass flow averaged degree of super-
heat is 5.1 K. The mass flow rate for the full annular passage is
107.6 kg/s. The rotational speed of the rotor is 3000 rpm. At the
outlet of the last stage, the static pressure is 5.4 kPa and the calcu-
lated mass flow averaged wetness fraction is 10.2%. There are 58
stationary blades and 85 moving blades in the last stage and all
computations are undertaken for one stationary blade passage
and one moving blade passage by applying periodic boundaries.
No leakage paths are modeled. The computational domains are dis-
cretized by structural multi-block grids with roughly 2,120,000
nodes for the stationary blade passage and 1,980,000 nodes for
the moving blade passage. To better resolve boundary layers near
the wall and blade leading and trailing edges, the grids are densely
packed in these locations.
Three parameters related to the droplet deposition are defined
for a better representation of the deposition data. The first one is
fractional diffusional deposit ion rate FD which is defined as the
ratio of the deposited mass flow rate of droplet by the turbulent
diffusion to the inlet droplet mass flow rate. The second one is frac-
tional inertial deposition rate FI which is defined as the fraction of
the deposited droplets by the inertial impact. The last one is total
deposition rate FT which is the sum of FD and FI.
When calculating the inertial deposition by the Lagrangian
particle track model, to consider the span-wise distributions of
droplets diameter and liquid mass fraction, 20 injection regions
were uniformly distributed at the inlet of the last stage over the
span height. The diameter and the mass flow rate for each inject
ion are shown in Fig. 2 with solid square points. The corresponding
liquid mass fraction over the span height is given by the solid circle
points in Fig. 2(b). In addition, 9 injection regions were uniformly
distributed in circumferential direction. A sensitivity analysis of
the number of injection regions selected, which contains two cases,
is conducted and the results are shown in Table 1. In the first case,
5, 7, and 9 injection regions are placed in circumferential direction
and 20 in span-wise direction. For the second case, 9 injection
regions are placed in circumferential direction and 10, 20 and 40
in span-wise direction. The inlet fog droplet diameter for both
cases is 16 times the droplet diameter (D  16) shown in
Fig. 2(a). It can be seen that 20 injection regions in span-wise direc-
tion are sufficient for predicting the inertial deposition of droplets.
Values in Table 1 also indicate that 9 injection regions in circum-
ferential direction are sufficient because it can be expected that
the change in the predicted inertial deposition rate will get insig-
nificant if more than 9 injection regions are used.
Table 1
Sensitivity analysis of the number of injection regions selected.
Circumferential number 5 7 9
FI (%) 5.77 5.21 5.18
Span-wise number 10 20 40
FI (%) 6.18 5.18 5.19
646 X. Yu et al. / International Journal of Heat and Mass Transfer 84 (2015) 642–652
To characterize the droplet diameter distribution in Fig. 2(a),
the D43 mean droplet diameter is employed and is calculated as
follows:
P
Ni D4i
D43 ¼ Pn 3
n N i Di
The summation in Eq. (13) runs through the flow area at the inlet of
the last stage for all grid nodes.
The method to calculate the diffusional deposition employed
has the ability to consider the influence of blade surface roughness
in the turbulent particle-diffusion regime and the eddy diffusion-
impaction regime. Typical roughness measurements on polished
turbine blades give equivalent sand grain roughness heights ks
between 1 lm and 4 lm [10]. The stationary blade surface rough-
ness effect on the diffusional deposition for different inlet droplets
sizes is investigated and the calculated results are shown in Fig. 3.
The fractional deviation dd is defined as follows:
F D  F D;ks¼1 lm
dd ¼  100%
F D;ks¼1 lm
It is evident that increasing the surface roughness enhances the
deposition of small droplets but has little effect on large droplets.
However, for small droplets, the deposition rate is so low as to be
negligible according to the results of Yau and Young [10] and
Starzmann et al. [13]. It can be deemed that the diffusional deposi-
tion on the polished blades is unaffected by the blade surface
roughness. Thus, for all the calculations about the diffusional depo-
sition in the present work, the blade surface roughness ks of the
stationary blade and moving blade is assumed to be equal to 2 lm.
Fig. 4 depicts the fractional deposition of fog droplets on the
stationary blade, as well as a comparison with the results of Yau
and Young [10] and Starzmann et al. [13]. The droplets deposition
rate of the current work and Starzmann et al. are plotted against
the mean droplet diameter D43 at the inlet of the last stage. The
upper curves and lower curves represent the total deposition rate
and fractional diffusional deposition rate respectively. The results
of Yau and Young provided the deposition rates at three stream-
surfaces referred to as hub, mid-height and shroud and were com-
puted by a quasi-3D method. The maximum and minimum values
of FT and FD at the three stream-surfaces form the shadow regions
in Fig. 4. The results of Starzmann et al. give the deposition of fog
droplets on the stationary blade of a model turbine and are
restricted to the main flow regions between 30% and 70% span
height. One may observe that the trends and magnitudes of the
fractional diffusional deposition rate of the three results are
roughly similar as expected since the method used to calculate
Fig. 3. Effect of the blade surface roughness on the diffusional deposition of
droplets on the stationary blade.
ð13Þ
Fig. 4. Fractional deposition rate of fog droplets on the stationary blade and a
comparison with other CFD studies, the upper curves are total deposition rate FT
and the lower curves represent fractional diffusional deposition rate FD. The
abscissa is the mean droplet diameter D43 at the inlet of the last stage.
ð14Þ
the diffusional deposition is same. The similar trends and magni-
tudes of the total deposition rate for the three results indicate that
the Lagrangian particle track model employed in the current work
and in the work of Young and Yau [11] to calculate the inertial
deposition can obtain equivalent results as the method based on
the inhomogeneous multi-fluid model utilized by Starzmann. The
three results show that, when the droplet size is small, the
diffusional deposition dominates. With the increase of droplet
diameter, the deposition by the inertial effect becomes significant.
However, there are also some differences among the three
results. The main reason for the differences may be attributed to
the different blades designs studied which affect the magnitudes
and distributions of droplets sizes and the flow conditions at the
inlet of the blade passage. The methods to calculate the inertial
deposition of droplets are also different. For the Lagrangian
method employed in this work, the growth of droplets is neglected
which may underestimate the inertial deposition rate. But accord-
ing to Young and Yau [12], the Lagrangian method can obtain the
consistent results as measurements. The method to calculate the
inertial deposition employed by Starzmann et al. [13] is based on
the inhomogeneous multi-fluid model in which the particle–wall
interaction is neglected. The resulting high droplet concentration
near the blades due to the accumulation of droplets may
overestimate the inertial deposition rate. In addition, the different
considered span heights among the three results may also bring
some differences.
Fig. 5 shows the span-wise distributions of relative deposition
rate for different inlet droplets diameters. The relative deposition
rate is defined as the deposited mass flow rate over a specified area
divided by overall deposited mass flow rate on the blade. There is
no overlap between the white and gray bars. The sum of the white
and gray bars represents the total fractional deposition rate at the
specified areas. In the regions around the span height 70%, the
deposition rates are higher than that of other locations due to
the maximum inlet droplet diameter shown in Fig. 2(a). If the
droplets’ size is increased, the span-wise distributions of deposi-
tion by the turbulent diffusion become more uniform since the dif-
fusional deposition of larger droplets follows into the inertia-
moderated regime and the effect of the droplet size differences
over span height on the diffusional deposition is small. For small
droplet size (D  1 and D  4), the inertial deposition only occurs
around 15% span height due to the large circumferential velocity
while for large droplets (D  8 and D  16) it mainly happens
around 70% span height owing to the strengthened inertial effect
of droplets.
 X. Yu et al. / International Journal of Heat and Mass Transfer 84 (2015) 642–652 647

Fig. 5. Relative deposition rate on the stationary blade over span height for different inlet droplets diameters.

The fog droplets deposited on the stationary blade coagulate
into films and rivulets and are drawn toward the trailing edges
of blades. Detachment of these films and rivulets at the trailing
edge forms coarse water droplets consisting of larger droplets in
the 10–100 lm size range [23]. The evaluation of droplets deposi-
tion on the moving blade should take these coarse droplets into
account. For this purpose, additional computations are conducted
for the domain downstream of the trailing edge of the stationary
blade together with the moving blade domain as shown in Fig. 6.
These additional computations include the simulations by the
Euler–Euler multiphase model and the Lagrangian particle track
model for different droplet sizes at the inlet of the last stage. In
Fig. 6, the coarse water inlet corresponds to the location of the
trailing edge of stationary blade. The span-wise distributions of
mass fraction of coarse droplets are given based on the results of
fog droplet deposition on the stationary blade shown in Fig. 5.
The size of coarse droplets can be roughly evaluated from the crit-
ical Weber number which relates to the forces acting on a droplet
to the surface forces. For the present case, the value of the critical
Weber number is taken to be 22 as recommended by [6].
1 22r
davg ¼  ð15Þ
2 u2i;c qc

Fig. 7. Predicted span-wise distributions of coarse droplets diameter according to

Fig. 6. The domain downstream of the trailing edge of the stationary blade together Eq. (15) for case D  1 and the boundary of coarse droplets for the Lagrangian
with the moving blade domain. calculation.
648 X. Yu et al. / International Journal of Heat and Mass Transfer 84 (2015) 642–652
The average diameter of coarse droplets estimated by Eq. (15)
ranges from 35 lm to 65 lm over span height as shown in Fig. 7.
It is worth pointing out that the differences among the predicted
diameter of coarse droplets for all cases are small. The initial veloc-
ity of the coarse droplets is taken to be 10% of vapor velocity accord-
ing to the experimental measurements of Du et al. [24]. When
calculating the droplets deposition on the moving blade, the diffu-
sional deposition of coarse droplets is neglected and a similar
method for the statio nary blade to calculate diffusional deposition
Fig. 8. The total deposition rate of fog and coarse droplets and the diffusional
deposition rate of fog droplets on the moving blade. The abscissa is the mean
droplet diameter D43 at the inlet of the last stage.
Fig. 9. Relative deposition rate on the moving blade over span height for different stage inlet droplets sizes.
of fog droplets is employed. The span-wise distributions of the fog
droplets at the inlet of the domain shown in Fig. 6 are extracted
from the results of inhomogeneous multiphase calculations for
the last stage. When computing the inertial deposition of coarse
droplets, additional 20 injection regions are uniformly defined over
span height in the coarse water inlet as depicted in Fig. 6. The mass
flow rate of coarse droplets for each injection region is determined
by the span-wise distributions of deposited mass flow rate of fog
droplets on stationary blade. The definition of injection regions
for fog droplets is similar to that for the stationary blade.
The total deposition rate of fog and coarse droplets, the total
deposition rate of fog droplets and the diffusional deposition of
fog droplets on the moving blade for different inlet droplet diame-
ters are compared in Fig. 8. The deposition rates are plotted against
the D43 at the inlet of stationary blade. The fractional deposition
rates for each case are defined as the deposited mass flow rate
divided by overall mass flow rate at the inlet of the moving blade.
Fig. 8 shows that, with the increase of inlet droplets diameter, the
contribution of coarse droplets deposition to the total deposition
enhances since the deposition rate of fog droplets on the stationary
blade rises.
Fig. 9 depicts the span-wise distributions of the relative inertial
deposition rate of fog and coarse droplets and relative diffusional
rate deposition of fog droplets on the moving blade. The meanings
of white and gray bars are same as that for Fig. 5. When the drop-
lets size at the inlet of the stage is small, the diffusional deposition
mainly concentrates on the regions around 70% span height. The
span-wise distributions of diffusional deposition become uniform
with the increase of the fog droplets size. These phenomena are
similar to that for the stationary blade. The main region of the
 X. Yu et al. / International Journal of Heat and Mass Transfer 84 (2015) 642–652
inertial deposition for small inlet droplets locates close to the hub,
while for large inlet droplets it has been shifted close to the shroud.
4. Results of moisture loss calculations
In this part, a physically consistent 3D method to evaluate the
moisture losses is presented. The method is based on the 3D
inhomogeneous multi-phase flow simulations and the results of
droplets deposition on the stationary blade and moving blade.
Six categories of loss mechanisms [25] are considered and
presented in detail in the following sections.
4.1. Thermodynamic loss
Thermodynamic loss is produced due to the heat and mass
transfer between the vapor phase and liquid droplets at a different
temperature during the droplet nucleation and growth processes
[26]. A principle to determine the entropy production rate is
employed to estimate the thermodynamic loss. According to the
second law of thermodynamics, the entropy increase due to the
heat transfer between the vapor with a temperature Tc and
droplets with a temperature Td can be expressed by [27]:
Q_ Q_
D S1 ¼  ð16Þ
Tc Td
where Q_ is the latent heat release along with the condensation of
vapor and equates to m _  L. m
_ is the mass transfer rate between
vapor and droplets and L is the latent heat. Multiplying the entropy
production with a mean temperature Tm between the vapor and
liquid droplets gives the energy dissipation for each control volume.
Thermodynamic loss is obtained through the integration of the
energy dissipation over the whole fluid domain.
Z  
1 1
DP 1 ¼ _
mL  T m dV ð17Þ
vol Tc Td
A non-dimensional loss coefficient z is defined through dividing the
loss by the work output of the last stage.
The computed thermodynamic loss of the last stage with differ-
ent inlet diameters is shown in Fig. 10. For small inlet fog droplets,
a fast increase of thermodynamic loss is predicted while for large
droplets, the loss coefficient remains approximately constant at
about 4.6%. Fig. 11 shows the nucleation rate for the case D  4
in the last stage. The nucleation mainly occurs around span height
70% in the stator and rotor and near the hub in the stator.
Compared to Fig. 1, the secondary nucleation rate for the case
D  4 is much higher than that for the case D  1. It can be
concluded that larger droplets have strengthened the secondary
Fig. 10. The thermodynamic loss for different droplet diameters at the inlet of the
stage. The abscissa is the mean droplet diameter D43 at the inlet of the last stage.
649
Fig. 11. Contour plot of nucleation rate in the last stage for the case D  4. The
surfaces for the contour plot are obtained by offsetting 2 mm from the suction
surfaces of blades. The red line denotes the position of span height 70%. (For
interpretation of the references to color in this figure legend, the reader is referred
to the web version of this article.)
nucleation in the last stage, causing the temperature of droplets
much higher than that of the surrounding steam. The resulting
temperature difference is responsible for the increase of thermody-
namic loss. The results in Fig. 10 indicate that when the droplets
size reaches a threshold value, i.e. 1 lm for the current last stage
studied, a further increase of droplet size has little effect on the
secondary nucleation.
4.2. Drag loss of fog and coarse droplets
The inability of droplets to follow exactly the curved vapor
streamlines within blade passage leads to energy dissipation which
is referred to as drag loss. Eq. (3) gives the expression of overall
drag force acting on droplets and the entropy production rate
resulting from the drag force is:
SD;d
DS2 ¼ ðui;d  ui;c Þ ð18Þ
Tm
The integration of Eq. (18) multiplying the average temperature
over the whole fluid volume determines the drag loss:
Z
DP 2 ¼ DS2 T m dV ð19Þ
vol
The computed drag losses of fog and coarse droplets are shown in
Fig. 12. Large droplets are less able to follow the steam path result-
ing in significant slip velocity. Consequently, the drag loss of fog
droplets rises with the increase of droplet diameter. The drag loss
of coarse droplets is estimated by employing the domains shown
in Fig. 6. In Fig. 12(b), the x-axis of the curve with solid square
points is the average diameter D43 at the inlet of the stage while
that for the curve with solid circle points is the deposition rate of
fog droplets on the stationary blade. The sizes of coarse droplets
are calculated by Eq. (15) and are approximately identical for each
case. A linearized dependency of the drag loss of coarse droplets on
the deposition rate of fog droplets on the stationary blade is
observed in Fig. 12(b). It denotes that the magnitudes of the drag
loss of coarse droplets are mainly affected by the deposition rate
of fog droplets on the stationary blade. In addition, the drag loss
of fog droplets is much larger than that of coarse droplets.
4.3. Impact loss
The coarse droplets re-entrained into the flow with much lower
velocity than that of the steam and impinged on the leading edge
of the moving blade. The impinging of coarse droplets on the mov-
ing blade causes a reduction of blade work, which is usually called
impact loss. The impact loss can be determined from the momen-
tum acting on the blade surfaces [7]. This momentum is evaluated
650 X. Yu et al. / International Journal of Heat and Mass Transfer 84 (2015) 642–652

Fig. 13. The impact loss by coarse droplets for different fog droplets diameters at
the inlet of the stage.

4.4. Capturing loss

The deposition of water droplets on the stationary and moving

blades by inertial impact and turbulent diffusion causes a reduc-
tion of droplets’ kinematic energy. This energy dissipation is
defined as capturing loss and is calculated by the summation of
droplets kinematic energy over all grid nodes on the blade surface:

1X k
DP 5 ¼ ðQ þ Q D ÞAi u2d;i ð22Þ
2 i¼1 I

where QI and QD refer to the deposited mass flow rate per area due
to the inertial and diffusional effect. ud,i corresponds to the absolute
velocity of droplets deposited on the blade surfaces and is taken at
the first grid node away from the blade surface.
Fig. 12. The drag losses of fog and coarse droplets for different fog droplets In the current work, the liquid film formation and movement
diameters at the inlet of the stage.
are not considered and it is assumed that droplets have a zero
velocity after impinging on the blade surface. The results of total
for each grid point i with the associated surface area and the capturing loss as well as capturing loss due to droplets deposited
summation over all grid nodes on the blade surface gives the on the moving blade for different fog droplet sizes at the inlet of
impact loss as follows: the last stage are shown in Fig. 14. The x-axis of the curve with tri-
angular points is the deposition rate of fog droplets on the moving
X
k blade. It can be seen that for all fog droplet sizes considered, cap-
DP4 ¼ x Gcw;i ucw;u;i r i ð20Þ turing loss is mainly due to the droplets deposited on the moving
i¼1 blade. The capturing loss rises with the increase of droplet diame-
ter. Moreover, a nearly linearized variation of capturing loss due to
where x, ri and ucw,u,i represent the angular velocity, the local
radius and the local circumferential velocity of coarse droplets,
respectively. Gcw,i is the local mass flow rate of coarse droplet
toward the blade and is estimated by:

Gcw;i ¼ ucw;n;i Ai ðccw Þyþ ¼30 ð21Þ

The normal coarse droplets velocity ucw,n,i is evaluated at the first

grid node away from the blade surface. The volumetric mass
concentration of coarse droplets is taken at a distance of y+ = 30
away from the wall as recommended by Starzmann et al. [7].
Fig. 13 depicts the calculated impact loss for different sizes of
fog droplets at the inlet of the last stage. The abscissa of the curve
with solid circle points is the deposition rate of fog droplets on the
stationary blade. Similar to the drag loss of coarse droplets, the
main effect affecting the impact loss is the mass flow rate of coarse
droplets and a linearized variation of impact loss to the deposition
rate of droplets on the stationary blade is observed. However, the
magnitude of impact loss by coarse droplets is slightly larger than Fig. 14. The capturing loss for different fog droplet diameters at the inlet of last
that of the drag loss by coarse droplets. stage.
 X. Yu et al. / International Journal of Heat and Mass Transfer 84 (2015) 642–652 651

droplets deposited on the moving blade to the deposit ion rate of

fog droplets on the moving blade is observed. It indicates that
the capturing loss is mainly due to the deposition of fog droplets
on the moving blade because the velocity of coarse droplets is
quite small.

4.5. Centrifuging loss

On the moving blades, the deposited droplets move radially

toward the blade tip by centrifugal force which causes energy dis-
sipations and reduces the blade work. Such energy reduction is
referred to as centrifuging loss which is calculated by:

X
k
DP 6 ¼ ðQ I þ Q D ÞAi x2 ðr2tip  r 2i Þ ð23Þ
i¼1
Fig. 16. Overall moisture loss in the last stage for different fog droplet diameters at
the inlet of the stage.
where rtip is the tip radius of blades and the summation is over all
the grid nodes on the blade surface.
In Fig. 15, the curve with solid square points shows the results
of centrifuging loss for different inlet fog droplets diameters. The
centrifuging loss rises with the increase of fog droplet sizes at
the inlet of the last stage. Both the droplet deposition rate and
the deposited location on the blade surface affect the centrifuging
loss. The curve with solid circle points shows the variation of cen-
trifuging loss to the total deposition rate of fog and coarse droplets
on the moving blade and a nearly linearized variation is observed.
This may indicate that for the present last stage studied, the main
factor influencing centrifuging loss is the droplet deposition rate.
The overall moisture losses in the last stage for different inlet
sizes of fog droplets are given in Fig. 16. A fast increase of the mois-
ture losses with droplets diameter D43 is observed when the drop-
lets sizes are small. It is mainly due to the fast increase of the
thermodynamic loss. If the droplet diameter D43 is larger than Fig. 17. Relative fractions of the six categories of the moisture losses for different
0.5 lm, the thermodynamic loss remains approximately constant fog droplet diameters at the inlet of the last stage.

and the increase of other five categories of the moisture losses

attributes to the increase of the moisture losses. This results in a
transition occurring when the droplet diameter D43 is 0.5 lm and
a nearly linearized variation of the moisture losses against the
droplet diameter D43 at the inlet of the last stage.
Fig. 17 depicts the relative fractions of the six categories of the
moisture losses for different droplets sizes at the inlet of the last
stage. For all cases considered, the thermodynamic loss dominates.
However, if the fog droplet size is increased, the relative fraction of
the thermodynamic loss decreases and the relative fractions of
other five categories of the moisture losses rise. For example, for
case D  1, the relative fraction of moisture is 93.8%, while for case
D  16, it decreases to 72.2%.
5. Discussions
The classical method to predict the moisture losses is the Bau-
mann rule [2] which relates the wet isentropic total-static effi-
ciency to the efficiency of a dry expansion. However, Kreimeier
et al. [28] argued that this representation of the real expansion
based on a dry reference process is misleading concerning the
deduction of the moisture losses. The reason is that even without
moisture losses, the difference still exists between the isentropic
efficiency of dry and wet expansion processes due to the difference
of isentropic exponent between wet steam and dry steam. Kreime-
ier et al. proposed a modified Baumann rule which requires the
examination of an equilibrium total-static isentropic efficiency
and is expressed as following equation.

gis ¼ gis;eq ð1  aeq ym;eq Þ ð24Þ

The equilibrium expansion process is defined using the identical

Fig. 15. The centrifuging loss for different fog droplets diameters at the inlet of the
stage.
inlet conditions, outlet pressure and rotational speed as the real
expansion, assuming that the non-equilibrium condensation
process is suppressed. Based on the analysis of the experimental
data of an outdated five-stage LP impulse turbine, Kreimeier et al.
suggest that the real moisture losses can be deduced by an almost
constant equilibrium Baumann factor of order 0.6.
Applying the modified Baumann rule proposed by Kreimeier
et al. to the studied last stage, the predicted moisture loss coeffi-
cient is about 3.09% which is quite consistent with the value of
652 X. Yu et al. / International Journal of Heat and Mass Transfer 84 (2015) 642–652
moisture loss coefficient of the case D  1 (3.01%) predicted by the
3D method employed in this work. However, the modified
Baumann rule cannot consider the effect of droplet size. If there
are large droplets, for example the coarse droplets due to the depo-
sition of fog droplets on the blades of the penultimate stage (which
is not considered in the current work), at the inlet of the last stage,
the modified Baumann rule will underestimate the moisture
losses. As shown in Fig. 16, for case D  16, the moisture loss is
about two times as larger as that for case D  1.
6. Conclusions
In the current work, a prediction of the diffusional and inertial
deposition of droplets on the stationary blade and moving blade
for different fog droplet diameters at the inlet of the last stage of
a 1000 MW fossil-fired steam turbine has been realized. For diffu-
sional deposition, a method proposed by Yau and Young was
employed and implemented within the results of inhomogeneous
multi-phase calculations by user-defined expressions. A Lagrang-
ian particle track model was used to predict the inertial deposition
and the span-wise distributions of droplet diameter and mass frac-
tion were considered by settings of injection regions. For the deter-
mination of the droplets deposition on the moving blade, separate
calculations were conducted on the domain downstream the trail-
ing edge of the stationary blade together with the moving blade
domain.
In the context of 3D multi-phase flow solutions and the droplets
deposition predictions on the blades, a physically consistent 3D
method to evaluate the moisture losses has been presented. Within
the method, the moisture losses are divided into six different cat-
egories namely thermodynamic loss, drag loss of fog droplets, drag
loss of coarse droplets, impact loss, capturing loss and centrifuging
loss. All six categories as well as the total moisture losses in the last
stage are determined for different fog droplet diameters at the inlet
of the last stage. Results demonstrate that inlet droplet diameter
has significant effect on the moisture losses and larger inlet drop-
lets results in higher moisture losses. In addition, for all cases con-
sidered, the thermodynamic loss dominates. With the increase of
fog droplet diameter, the relative fraction of thermodynamic loss
decreases and the relative fractions of other five categories of the
moisture losses rise.
For the current work, the analysis is based on the steady calcu-
lations. As indicated by Bakhtar [29], the wake chopping effects
due to the inherent unsteadiness in turbine may have impact on
the nucleation process and droplet sizes. It is the ambition of the
authors to consider the unsteadiness effect when estimating
moisture losses in the future.
Conflict of interest
None declared.
Acknowledgements
The financial support for the present study was provided by the
National Natural Science Foundation of China (Grant No.
51376140) and DongFang Steam Turbine Works (DFSTW).
References
[1] J.A. Hesketh, P.J. Walker, Effects of wetness in steam turbines, J. Mech. Eng. Sci.
219 (2005) 1301–1314.
[2] K. Baumannn, Some recent developments in large steam turbine practice, J.
Inst. Electr. Eng. 59 (1921) 565–623.
[3] T. Guo, W.J. Sumner, D.C. Hofer, Development of highly efficient nuclear HP
steam turbines using physics based moisture loss models. ASME Turbo Expo,
2007, GT2007-27960.
[4] G. Gyarmathy, Bases of a theory for wet steam turbines (Ph.D. thesis), Federal
Technical University of Zurich, Swiss, 1962.
[5] A.R. Laali, A new approach for assessment of the wetness in steam turbines, in:
Institution of Mechanical Engineers Conference: Turbomachinery, 1991,
pp.155–166.
[6] M.J. Moore, Two-Phase Steam Flow in Turbines and Separators: Theory,
Instrumentation, Engineering, Hemisphere Publishing Corporation, 1976 (pp.
115–135).
[7] J. Starzmann, Michael M. Casey, J.F. Mayer, Frank Sieverding, Wetness loss
prediction for a low-pressure steam turbine using computational fluid
dynamics, J. Power Energy 228 (2013) 1–16.
[8] Vaclav Petr, Michal Kolovratnik, Wet steam energy loss and related Baumann
rule in low-pressure steam turbines, J. Power Energy 228 (2013) 206–215.
[9] R.I. Crane, Deposition of fog drops on low pressure steam turbine blades, Int. J.
Mech. Sci. 15 (1973) 613–631.
[10] K.K. Yau, J.B. Young, The deposition of fog droplets on steam turbine blades by
turbulent diffusion, J. Turbomach. 109 (1987) 429–435.
[11] J.B. Young, K.K. Yau, The inertial deposition of fog droplets on steam turbine
blades, J. Turbomach. 110 (1988) 155–162.
[12] J.B. Young, K.K. Yau, P.T. Walters, Fog droplet deposition and coarse water
formation in low-pressure steam turbines: a combined experimental and
theoretical analysis, J. Turbomach 110 (1988) 163–172.
[13] J. Starzmann, P. Kaluza, M.V. Casey, F. Sieverding, On kinematic relaxation and
deposition of water droplets in the last stage of low pressure steam turbines, J.
Turbomach. 136 (2014) 1–10.
[14] N.B. Wood, A simple method for the calculation of turbulent deposition to
smooth and rough surfaces, J. Aerosol Sci. 12 (1981) 275–290.
[15] A.G. Gerber, M.J. Kermani, A pressure based Eulerian–Eulerian multi-phase
model for non-equilibrium condensation in transonic steam flow, Int. J. Heat
Mass Transfer 47 (2004) 2217–2231.
[16] A.G. Gerber, R. Sigg, L. Volker, M.V. Casey, N. Surken, Predictions of non-
equilibrium phase transition in a model low-pressure steam turbine, J. Power
Energy 221 (2007) 825–835.
[17] A.G. Gerber, Inhomogeneous Multi-fluid Model for prediction of non-
equilibrium phase transition and droplet dynamics, J. Fluids Eng. 130 (2008)
1–11.
[18] Xinggang Yu, Danmei Xie, Cong Wang, Chun Wang, Chu Nie, Numerical
investigation of condensing flow in the last stage of low-pressure nuclear
steam turbine, Nucl. Eng. Des. 275 (2014) 197–204.
[19] F. Bakhtar, J.B. Young, A.J. White, D.A. Simpson, Classical nucleation theory and
its application to condensing steam flow calculations, J. Mech. Eng. Sci. 219
(2005) 1315–1333.
[20] G. Gyarmathy, Condensation in flowing steam, in: M.J. Moore, C.H. Sieverding
(Eds.), A von Karman Institute Book on Two-Phase Steam Flow in Turbines and
Separators, Hemisphere, London, 1976, pp. 127–189.
[21] M. Grüebel, J. Stazmann, M. Schatz, et al., Two-phase flow modeling and
measurements in low-pressure turbines – Part 1: numerical validation of wet
steam models and turbine modeling. ASME Turbo Expo, 2014, GT2014-25244.
[22] L. Li, J.-D. Yang, W. You, et al., Investigation of the vapor–liquid two-phase flow
in the low-pressure cylinder of a 1000 MW nuclear power steam turbine, J.
Power Energy 228 (2014) 178–185.
[23] R.I. Crane, Droplet deposition in steam turbines, J. Mech. Eng. Sci. 218 (2004)
859–870.
[24] Zhan-bo Du, Jing-ru Mao, Bi Sun, Measurements of droplets sizes at mid-
height of stationary turbine blades and analysis of the critical Webber number,
Power Eng. 25 (2005) 643–646.
[25] Hiroyuki Kawagishi, Akihiro Onoda, Naoki Shibukawa, Yoshiki Niizeki,
Development of moisture loss models in steam turbines, Heat Transfer-Asian
Res. 42 (2013) 651–664.
[26] M.J. Kermani, A.G. Gerber, A general formula for the evaluation of
thermodynamic and aerodynamic losses in nucleating steam flow, Int. J.
Heat Mass Transfer 46 (2003) 3265–3278.
[27] J. Starzmann, Michael Casey. Non-equilibrium condensation effects on the
flow field and the performance of a low pressure steam turbine. ASME Turbo
Expo, 2010, GT2010-22467.
[28] F. Kreimeier, R. Greim, F. Congiu, J. Faelling, Experimental and numerical
analyses of relaxation processes in LP steam turbines, J. Mech. Eng. Sci. 219
(2005) 1411–1436.
[29] F. Bakhtar, A.V. Heaton, Effects of wake chopping on droplet sizes in steam
turbines, J. Mech. Eng. Sci. 219 (2005) 1357–1367.