Materials Transactions, Vol. 49, No. 7 (2008) pp.

1606 to 1615
#2008 The Japan Institute of Metals

Water Drop Erosion on Turbine Blades: Numerical Framework and Applications

Qulan Zhou1;2; * , Na Li1;2 , Xi Chen2 , Akio Yonezu3; *, Tongmo Xu1 , Shien Hui1 and Di Zhang1
1

The State Key Laboratory of Power Engineering Multiphase Flow, Xian Jiaotong University, Xian 710049, P. R. China
Department of Civil Engineering and Engineering Mechanics, Columbia University, New York, NY 10027, USA
3
Department of Mechanical Engineering, Osaka University, Suita 565-0871, Japan
2

When small droplets are formed in the wet steam stage of a steam turbine, they may impact the blade surface at a high velocity
and repetitive impacts cause water drop erosion, which emerges as one of the primary reliability concerns of the turbine. We propose
an eective numerical framework that couples uid mechanics with solid mechanics. The movements of water drops in a blade channel are
analyzed based on the solution of the ow eld of water steam in turbine, and impact statistics such as impact frequency, velocity, and position
are obtained as the working condition and particle size are varied. A nonlinear wave model is established for high velocity liquid-solid
impact, from which the characteristic impact pressure in liquid and peak impact stress in solid are obtained; the solutions are then superimposed
with the pathways of water particles, and a fatigue analysis is carried out to elucidate the mechanisms of water drop erosion. The lifetime map
on a blade surface with two dierent materials (1Cr13 and Ti-6Al-4V) under typical working conditions are obtained, in terms of operation
hours, and the most dangerous water drop erosion regions and operating conditions of the steam turbine are deduced.
[doi:10.2320/matertrans.MRA2008025]
(Received January 17, 2008; Accepted April 15, 2008; Published June 11, 2008)
Keywords: water drop erosion, liquid-solid impact, numerical simulation

1.

Introduction

Near the exhaust of a steam turbine engine (also known as

the penult stage or nal stage), the low-pressure working
condition favors phase transformation, and small water drops
are produced by the condensation of steam. Once formed, the
water drops move with the ow and some of them may
impact the blade surface with a velocity over 200 m/s
upon consecutive impact on the turbine blades, the surface
material may spall o and this is known as the water drop
erosion which severely aects the system reliability (Fig. 1).
According to experimental observations, the most favorable
impact locations include the leading or trailing edges of back
arc and inner arc, and the erosion rate is around 0.1 micron
per hour.1) Therefore, water drop erosion emerges as one
of the primary reliability concerns in a turbine and its
mechanism must be suciently understood.
The water drop erosion is also a very complex problem that
couples multiphase uid ow with impact mechanics and
fatigue analysis, which requires a seamless coupling between
uid and solid mechanics. First, it is important to solve for
the trajectories and velocities of water drops, as well as the
distribution of mass and size of the particles. While some
water drops may follow the steam ow and exit the turbine,
others may impact the blade surface and key information
needs to be collected, including the impact frequency,
location, velocity, which depend on the aforementioned
variables. Finally, the liquid-solid impact problem imposes a
tremendous challenge and an appropriate model is needed to
solve for the distribution and magnitude of transient impact
stresses in the solid. When the solution of the fundamental
liquid-solid impact problem is superimposed with the impact
statistics, an fatigue analysis may be carried out to evaluate
the lifetime, most dangerous areas and working conditions. In
this paper, we establish multidisciplinary computational and
*Corresponding

authors, E-mail: qz2129@columbia.edu (Q. Zhou),

akio yonezu831@hotmail.com (A. Yonezu)

Mixing plane
Rotating direction
Condensation of
steam to water drops

Inner arc

Back arc
Stator (nozzles row)

Moving traces of
drops, some drops
impact the blade,
others move through
the blade channel.

Rotor (blades row)

Fig. 1 Schematic of the water erosion problem. The turbine stage, where
the uid ows through nozzle and blade channels, connected by the
mixing plane. The stator consists of nozzles, while the rotor (which rotates
to export mechanical work) incorporates blades. As the stator is static and
the rotor is rotating, there is a narrow gap between them (mixing plane).
The water drops (from condensation of wet steam) carried by the ow tend
to impact on the blade and induce water drop erosion.

mathematical models toward such an objective. The exible

framework may also be extended to similar problems
encountered in practice, such as rain drop erosion on civil
structures or airplane engines.
2.

Flow Field in a Wet Steam Turbine Stage

2.1 Model and computation method

For a long blade turbine stage, the Ma number varies from
low at the inlet to high (
supersonic). The conventional
approach2,3) cannot be directly applied to simulate the ow in
a turbine stage that involves both static nozzles and rotating
blades. In order to reduce computational cost, in this section,
the compressible pressure correction method (SIMPLE) is
combined with the mixing plane model;4,5) the static and
moving regions are treated separately and connected
through the mixing plane. A series of techniques are
proposed to take into account the various boundary con-

Water Drop Erosion on Turbine Blades: Numerical Framework and Applications

ditions that are related with dierent speed ranges in the

nozzle and blade.
The traditional K  " turbulence model2,6) is employed
in numerical study, which takes the general form of N-S
equation:
@
@
@

u
v
w
@x
@y
@z









@
@
@
@
@
@

S0 S1 1
@x
@x
@y
@y
@z
@
where u, v, w are the velocity components in x, y, and z
directions, respectively;
is the density of uid,
is the
generalized viscosity factor, S0 is the source term caused by
turbulence model; S1 is the source term caused by other
factors, and  is the eld variable, which can be a velocity
component (u, v, w), temperature (T), turbulence energy (K),
and turbulence energy dissipation ("), etc., see Table 1. The
rst column is the eld variable of the standard K  "
turbulence model; the second column is the generalized
viscosity factor; and the third column is the source term
caused by turbulence model. In Table 1, c 0:09, c1
1:44, c2 1:92, k 1:0, " 1:3, T 1:0. Pr is the
Prandtl number, p is pressure,  is the molecular viscosity
factor, and t is the turbulence viscosity factor:6)
2

t c
K =":
eff  t
( "


 #
@u 2
@v 2
@w 2
G t 2

@x
@y
@z


2



 )
@u @v
@w @u 2
@v @w 2

:
@y @x
@x
@z
@z
@x

2
3

Since there are 8 variables in (1),
; p; u; v; w; K; "; T, the

state equation is needed to close the equation set. For wet
steam, the second Virial coecient equation is used:
p RT=1 B=;

where  is the kinematic-viscosity coecient, and R100800

is the
34900T .
gas constant, and B 2:0624  103  2:61204

10
T
In this work, the coordinate system is xed on the rotating
blade. Consider a uid block in a rotary system with angular
velocity !, both the centrifugal force ( fe ) and the Coriolis
force ( fc ) are acting on the block:
2

fe x 0; fe y !2 y  y0 ; fe z !2 z  z0
fc x 0; fc y 2!w; fc z 2!v

 t

 t

 t

T
K
"


t

Pr T
t

K


t
"


@p
@
@u
eff


@x
@x
@x



@p
@
@u
eff


@y
@x
@y



@p
@
@u
eff


@z
@x
@z

S0



@
@v
eff

@y
@x



@
@v
eff

@y
@y



@
@v
eff

@y
@z

Su 0; Sv fc y fe y; Sw fc z fe z:

2.2 Flow eld of wet steam ow in a turbine stage

The numerical framework established above is employed
to simulate the ow eld in a whole wet steam turbine stage,
which is a penult stage of a 125 MW turbine (Dongfang
Steam Turbine Company, China). The geometry and dimension are given in Fig. 2, where the height of nozzle,
l1 450 mm and the height of blade is l2 460 mm. The
outlet of nozzle channel and the inlet of blade channel are
aligned with a plane in the radial direction, so as to apply the
mixing plane model. Two operating conditions (Table 2) are
simulated, under full designed and reduced (75%) loads,
respectively. For simplicity, we consider only the ux of the
working uid (i.e. the inlet Ma number) is varied and all other
parameters are xed between these two working conditions.
Note that the numerical protocol can be readily extended
to other turbines with dierent dimensions and operating
conditions.
The ow elds under the two operating conditions are
shown in Fig. 3(a) and (b), respectively, for the top section of
the stage. The color bar and arrow length denote the
Top

Top


@
@w
eff
@z
@x



@
@w
eff
@z
@y



@
@w
eff
@z
@z

l1=450m

(a)

l2=460

Bottom

"
c1 G  c2
"
K

The inlet conditions include the temperature and velocity

of the uid. If the speed at the outlet approaches supersonic,
the static inlet pressure must be given. The distribution of
static pressure is given on the outlet, and the outlet velocity
is assumed to be unidirectional;6) other outlet variables are
obtained by extrapolation. The uid ow near the wall
follows the wall function.6) The static and dynamic regions
are solved separately based on time-averaged parameters,
and interacted through a mixing plane model.7) In the
physical space, the mixing plane (Fig. 1) is the narrow gap
between the nozzle and blade, where the wet steam exits
the nozzles and ows into the blade channels.

0
G 
"

6
7

where the direction of angular velocity vector is aligned

~ !; 0; 0. The coordinates of the
with the x-axis, i.e. 
rotary axis is y0 ; z0 , thus, the polar radius vector is
r~ 0; y  y0 ; z  z0 . The source terms of the momentum
equations become

Table 1 The meanings of terms in eq. (1) when the standard K  "
turbulence model is used.


1607

Bottom

(b)

Fig. 2 The prole and dimension of the nozzle (a) and blade (b) used in the
simulation.

1608

Q. Zhou et al.
Table 2

The inlet parameters of the turbine stage.

Operating condition

Ma number

Temperature
T (K)

Static pressure
p (MPa)

No. 1 (the rated load)

0.216

356

0.03

No. 2 (75% load)

0.16

356

0.03

(a)

(b)

Fig. 3 The ow eld on the top section of the stage, for both nozzle (left)
and blade (right) under (a) operating condition No. 1 (full designed load),
(b) operating condition No. 2 (75% reduced load).

magnitude of velocity vectors. While the patterns of ow

eld in the nozzle channel are fairly close for the two
working conditions, the ow elds in the blade channel are
distinct. When the load of turbine varies, the ow speed
at the inlet/outlet of the nozzle changes yet the angular
velocity of blade is invariant. When these two factors are
combined, both the magnitude and direction of the ow eld
in the blade channel become very sensitive to the load.
Subsequently at reduced load, the inlet velocity vectors
become opposite to the direction of rotation, which cause
the inlet steam ow toward the back arc, and a backow
region is formed on the top section (Fig. 3(b)).
3.

where ! is the angular velocity of the coordinates (same

as that of the blade), and rp is the radial coordinate of the
water drop, t 
p dp2 =18 is the relaxation time of
particle
8
1
Re < 1
>
<
1 2=3
f Re 1 Re
10
1  Re  1000
>
6
:
0:01854Re Re > 1000
Re
g ug  up dp =;
11
where
p is the particle density, dp is the particle diameter,
 is the dynamic-viscosity coecient of gas. up , vp , wp
are the velocity components of the particle, ug , vg , wg are
the velocity components of the uid, and ug0 , vg0 , wg0 are
the pulsant velocities of the turbulent ow. Assuming the
turbulence is isotropic,
p 2
p 2
ug0  u0 2 K; vg0  v0 2 K;
3
3
12
p 2
0
2
0

wg  w K
3
where  is a random number with normal distribution, and
it remains constant during a time step which tracks the
movement of the particle. The turbulence energy K in the
K  " turbulence model can be solved from the steam ow
eld (see Section 2). The velocities up ; vp ; wp of the particle
are obtained from (9); we then integrate the velocities
up ; vp ; wp to obtain the position coordinates xp ; yp ; zp of
the particle.
3.2 Trace of water drops in blade channel
With the turbine stage and working conditions specied in
Section 2.2, sample traces are shown in Figs. 4 and 5 at the
bottom and top of blade channel, respectively. Combinations

Motion of Water Drops in Blade Channel

3.1 Model and computation method

The diameter distribution of particles entering the blade
channel is given by Shi.8) The particle trace model in the
Lagrangian coordinates9) is employed in this study to obtain
the statistics of movement of water drops in the blade
channel.
Since the typical humidity of wet steam is lower than
10%1,10) and the volume fraction of water drops in steam is
under 0.01%, it is reasonable to assume that the motion of
water drop would not counter-aect the uid ow. We further neglect the crash and coalescence of water drops since
the transient stress develops very quickly in the solid upon
impact, and it is the most critical for erosion (see below).
We use a semi-random trace model which incorporates the
correction of turbulence dissipation and with moderate
computational cost. In the rotary coordinate system attached
to the blade, the motion of particles is described as:
8
dup
1
>
>
ug ug0  up f Re
>
>
>

dt
>
<
dvp
1
9
vg vg0  vp f Re 2!wp
>

dt
>
>
>
>
dwp
1
>
:
wg wg0  wp f Re  2!vp !2 rp

dt

(a)

(b)

(c)

(d)

Fig. 4 The traces of water drops at the bottom of blade channel (view in x-y
plane): (a) 5 mm particles with operating condition No. 1; (b) 100 mm
particles with operating condition No. 1; (c) 5 mm particles with operating
condition No. 2; (d) 100 mm particles with operating condition No. 2.

(a)

(b)

(c)

(d)

Fig. 5 The traces of water drops at the top of blade channel (view in x-y
plane): (a) 5 mm particles with operating condition No. 1; (b) 100 mm
particles with operating condition No. 1; (c) 5 mm particles with operating
condition No. 2; (d) 100 mm particles with operating condition No. 2.

Water Drop Erosion on Turbine Blades: Numerical Framework and Applications

1.820

0.4400

2.080
0.2

0.2

Bottom

(a)

3.150

Bottom

0.0
0.0 0.5 1.0

Bottom

(b)

(c)

42.00

42.00

56.00

56.00

45.00

0.4

98.00

120.0

56.00
70.00
84.00

0.4

98.00

98.00
112.0

112.0

126.0

0.2

42.00

84.00

112.0

108.0

0.2

28.00

0.6

Z
0.4

84.00

14.00

70.00

84.00

96.00

126.0

0.2

140.0

2.160
2.400

Bottom

0.0
0.0 0.5 1.0

Bottom

0.0
0.0 0.5 1.0

0.0
0.0 0.5 1.0

Bottom

(a)

126.0

0.2

140.0

140.0

(d)

13

where N is the number of drops collected by the element

area, A; N0 is the total number of drops entering the
channel, and A0 is the area of the entry plane of the channel.
Figure 6 gives the maps of s of 100 mm on the inner and
back arcs of the blade, and under both operating conditions.
The x and z coordinates are normalized by the total length of
blade in the x direction (xb ) and height in the z direction (zb ),
respectively. The maximum impact rate with the reduced
load is 2.6, which is greater than under full load. Even though
the number of drops launched on the entry plane is same, the
impact frequency on the head of back arc with operating
condition No. 2 (Fig. 6(b)) is almost 5 times of that of No. 1
(Fig. 6(a)). In practice, the number of drops on the entry
plane under the low load operating condition should be more
than that of the full load condition, because the steam under
the low load operating condition has lower enthalpy and more
water drops will be produced. Thus, the head of back arc of
blade is the most dangerous region that will be eroded by
drops under the low load operating conditions. For the inner
arc, the maximum impact frequency with full load is larger
than that of reduced load. Under full load, the most critical
erosion region is close to the diagonal line of the blade,
whereas the region of concern upon reduced load is at the
bottom of the blade.
Figures 7 and 8 give the maps of normal impact velocity
distributions on the back arc/inner arc under the operating

Bottom

0.0
0.0 0.5 1.0

0.0
0.0 0.5 1.0

Bottom

(c)

Bottom

(d)

(e)

Fig. 7 The distribution of normal impact velocity (m/s) on the back arc
under the operating condition No. 2 (view in x-z plane), with the particle
diameter: (a) 5 mm, (b) 10 mm, (c) 50 mm, (d) 150 mm, (e) 250 mm.

1.0

Top

1.0

Top

16.00

36.00

32.00

24.00

48.00

54.00

48.00

64.00
80.00

Z
56.00

0.4

112.0

(a)

126.0

0.4

112.0

128.0

144.0

128.0

72.00

144.0

162.0

144.0

0.2

160.0

0.0
0.0 0.5 1.0

Bottom

(b)

0.2

180.0

0.0
0.0 0.5 1.0

Bottom

(c)

64.00
80.00

96.00

64.00
0.2

48.00
0.6

80.00

108.0
0.4

32.00

64.00

0.6

90.00

96.00

80.00

Bottom

72.00

0.6

16.00
0.8

32.00

0.6

1.0
0

0.8

0.8

16.00

48.00

0.0
0.0 0.5 1.0

Top
0

18.00

40.00

0.2

1.0

16.00

32.00

0.4

Top

8.000
0.8

0.6

1.0

of dierent particle size and working conditions are explored. Most 5 mm particles move along with the steam ow
consistently, and only very few particles impact the inner
arc of blade with small normal impact velocities. Because
the inlet angles of ow are dierent, more 5 mm particles tend
to impact the blade under the full load than that under
reduced load (although the percentage of particles colliding
with the blade is fairly small in both cases). By contrast, with
the dominance of their inertia, the 100 mm particles can be
easily separated from steam ow and impact the blade
surface at a high normal velocity; in particular, under the
reduced load, the head of back arc appears suering to
high-possibility of water drop impacts. We introduce a
dimensionless impact rate s :

0.0
0.0 0.5 1.0

(b)

0.8

N  A0
A  N0

Z
0.4

0.6

70.00

72.00

50.00

Top

A!0

0.6

60.00

0.2

Fig. 6 The distribution of s of 100 mm particles under operating condition

(a) the back arc of blade of No. 1; (b) the back arc of blade of No. 2; (c) the
inner arc of blade of No. 1; (d) the inner arc of blade of No. 2 (view in x-z
plane).

s lim

48.00

0.6

1.920

3.500

0.0
0.0 0.5 1.0

36.00

1.0

0.8

0.8

15.00

1.440
1.680

14.00
28.00

35.00

0.2

Top
0

28.00

40.00

0.4

1.0

14.00
0.8

30.00
0.4

Top
0

24.00

25.00

1.200

2.800
0.2

0.6

1.0

12.00

10.00

20.00

0.9600

2.450

2.600

0.5500
0.0
0.0 0.5 1.0

2.100

0.4

2.340

0.4950

1.750

Top
0
0.8

0.3850

0.7200

0.6

1.560

0.4

0.2400

1.400

1.300

1.0

1.040

0.2750
0.3300

1.050

Top

0.8

0.4800

0.6

0.2200

0.4

0.8

0.7000

0.7800

0.6

1.0

5.000

0.3500

0.5200

0.1650

0.6

0.8

Top

0.2600

0.1100

Trail

0.8

0.05500

Head 1.0
Top

Trail

0
0.8

Head1.0
Top

Trail

Head 1.0
Top

Trail

Head 1.0
Top

1609

96.00
0.4

112.0
128.0
144.0

0.2

160.0

0.0
0.0 0.5 1.0

Bottom

(d)

160.0

0.0
0.0 0.5 1.0

Bottom

(e)

Fig. 8 The distribution of normal impact velocity (m/s) on the inner arc
under the operating condition No. 2 (view in x-z plane), with the particle
diameter: (a) 5 mm, (b) 10 mm, (c) 50 mm, (d) 150 mm, (e) 250 mm.

conditions No. 2, as the particle size is varied. In all cases,

the impact velocity patterns of the larger particles are
essentially the same when the diameter of drops is bigger
than 50 mm. Since the larger particles are more critical for
water drop erosion, once the diameter of drop exceeds
50 mm, the impact frequency and velocity distributions can
be regarded as invariant so as to reduce the computational
cost.
Therefore, for larger particles the distributions of impact
velocity are independent of water drop diameters, and since
they are also more critical, we mainly focus on the impact by
larger particles in the following sections. At reduced load,
the maximum impact velocity (Fig. 7) appears at the head
of blade top, consistent with the highest impact frequency
(Fig. 6(b)). Thus, this region is quite likely to subject to
the most severe erosion; this agrees with.10) The maximum
impact velocity also increases signicantly with larger
particle size, which may be as high as 160 m/s for large
particles on the inner arc (Fig. 8). Such critical region which
spans from the front bottom to the middle height of trailing
edge, is partially overlapped with that of highest frequency
(Fig. 6(d)), which may be regarded as the second possible
erosion region.
The impact velocity and frequency maps are only partly
responsible for erosion mechanism. The blade fails by
transient stress during impact, and the liquid-solid impact
problem must be solved so as to evaluate the blade lifetime
in a more reliable manner. The liquid-solid impact stress eld
is analyzed in the next section, followed by erosion life
analysis in Section 5.

1610

4.

Q. Zhou et al.

Liquid-Solid Impact

4.1 Fundamental considerations of modeling

There are very limited experimental studies on high
velocity liquid-solid impact.1113) Moreover, these experiments focused on the pressure eld in the uid, however the
stress eld in the solid is the most critical for causing water
drop erosion. Since the surface material spalls o upon
consecutive impact, the transient peak stress causing such
damage due to impact must be below the surface. Therefore,
eective theoretical model and numerical simulation are
necessary for exploring the stress eld and related erosion
mechanisms. Due to the diculty of coupling between liquid
and solid phases, the study of liquid-solid impact is far less
comparing with solid-solid impact counterparts.14,15)
Along previous analytical attempts, many researchers
established simple models upon liquid-rigid surface interaction (many of them are 1D) and obtained the steady-state
pressure,1619) and then applied that pressure to obtain stress
characteristics in an elastic half-space.16,17,2023) Apparently,
in these works, the pressure eld in liquid was not simultaneously coupled with the stress wave in solid, the solutions
were not valid for the realistic 3D condition, and the transient
peak pressure in liquid and peak stress in solid, which are
much more severe than the steady-state values,15) were not
pursued.
In this section, we develop a fully-coupled, 3D dynamic
solid-liquid model. On the interface between liquid and solid,
all transient parameters, include pressure, force, mass point
position and movement, are exchanged between the two
regions and solved in situ. When a spherical water drop
impacts on a solid plane, the shock wave forms inside the
water drop due to the compressibility of liquid.24) The water
drop thus imposes a pressure distribution on the surface of
solid which varies with time and space, inducing stress waves
which transmit through the solid, with possible formation
and propagation of microcracks.
The Navier-Stokes (N-S) equation is the basic mathematical model for uid:
@

r 
V 0;
@t

@
V
r  T pI D 0 14
@t

where V is the velocity vector of mass point in liquid; the rst

equation of (14) is the continuity equation and the second
equation is the momentum equation. Here, T is the
momentum tensor, pI is the pressure tensor, and D is the
viscosity force tensor. I is the Kronecker delta. Due to the
complexity of the liquid-solid impact, a few basic assumptions are necessary for our model:
(1) We assume normal impact is much more critical than
tangential impact, and thus we focus only on the normal
component of the impact velocity (since the tangential
component does not produce impact stress). Thus we
only focus on normal axisymmetric problem (Fig. 9)
which simplies formulation.
(2) Due to the short characteristic impact duration (
ns) on
turbine blade, the deformation in the metal is elastic for
high-velocity impacts due to the strain rate eect.15)
This greatly simplies the current analysis and linear
elasticity theory with small deformation can be adopted

Spherical surface
of water drop

r
Solid region

Liquid-solid
Contact point e

ve

Edge cell E
e
Solution region

Contact angle

Liquid region

Impact speed v0
o B

Fig. 9 The coordinates system of axisymmetric liquid-solid impact and

coupling and. schematic of the moment when the shock wave breaks o
the liquid.

for the solid, and the coupling between the Euler

coordinate system in liquid and Lagrange coordinate
system in solid becomes easier.
(3) Since the maximum pressure appears before the shock
wave leaves the body of water drop,25,26) which will be
validated in this paper, we do not consider subsequent
events after the shock wave leaves the water drop. If the
sonic speed cE at the point e (Fig. 9, inside the edge cell
E) is larger than the moving speed of the point (ve ), that
implies the disturbing pressure wave (i.e. the shock
wave front) is moving faster than the edge of liquid, and
the shock wave will break o the liquid edge, thus
determining tSB , the critical moment when the shock
wave breaks o.
(4) From Section 3, at a high impact velocity up to 300 m/s,
the contact angle (Fig. 9) of the shock wave leaving
the water drop is about 10 .26) Moreover, the water
drop retains its spherical shape before the shock wave
leaves;24) later we will also show that the sonic wave is
conned to a very small region before it leaves the
water drop. Therefore, both the shape and volume
variations of water sphere are negligible before the
shock wave leaves the liquid drop,25,26) and thus the
impact is an acoustic procedure with negligible viscosity eect. We may thus ignore the viscosity of liquid
drop during high velocity impact (while retaining the
compressibility of uid), and reduce the N-S equation
to the nonlinear wave equation.
4.2 3D Nonlinear wave model for liquid-solid impact
The coordinates in this problem are xed at the interface
of the undisturbed liquid, and thus we consider the problem
where the solid is impacting the liquid (Fig. 9). The
undisturbed uid is static, V 0 0. The local density,
pressure, and velocity in the liquid can be expressed as the
summation of the values of undisturbed liquid (with subscript
0) and the relevant perturbation, which are all functions of
time and space coordinates:


0 
;

p p0 p;

V V 0 V

15

Substitute eq. (15) into eq. (14), note that @
0 =@t 0,
r
0 0, r p0 0. r  T  0 after neglecting the higher
order terms. After ignored the viscosity D, eq. (14)
becomes

Water Drop Erosion on Turbine Blades: Numerical Framework and Applications

@
=@t r 
V 0
@
V =@t r  pI 0

16a
16b

For a uid, 
can be written as27)


@
=@pp p=c2

17

where c is the sonic speed in the uid. Subtract the time

derivative of eq. (16a) from the divergence of eq. (16b), and
also subtract the time derivative of eq. (16b) from c2 times
the divergence of eq. (16a), and use eq. (17):
r2 p

1 @2 p
;
c2 @t2

r2 V

1 @2 V
c2 @t2

18

These are the wave equations. Introduce the velocity

potential function  with V r and p 
 @=@t,
the wave equations of p and V are merged to one wave
equation
r2 

1 @2 
c2 @t2

19

Since the present mathematical model of liquid-solid impact

is based on the wave equation, it is also termed as the wave
model of liquid-solid impact, or the wave model. Since
the sonic speed c is a variable of the density (and thus
pressure), the wave equation is a nonlinear second order
PDE. According to the Tait state equation of water28)
p A

 B
where A 1:0147663  10

19

20
8

, B 2:858987  10 Pa, and

8
1 @2 
>
>
< r2  2 2 ; jt0 0
c @t
Liquid
>
@
@Us
>
:
jx0 v0 
jx0
@x
@t

Solid

1611

7:15. c can be calculated from (17) as:

s s
@p
pB

c
@

21

Combine eqs. (19)(21), the governing equation set of the

nonlinear wave model is derived, which dominates the
physical procedure in the liquid region of the liquid-solid
impact:
8
1 @2 
@
>
2
>
p 

<r  2 2
c @t
@t
r
22
>
p
B
>
: p A

 B
c

There are 4 equations and 4 variables, , c, p,
. The
equation set is closed and can be solved. The compressibility
of liquid is embedded in the sonic wave speed c. When the
wave model for liquid is coupled with elasticity, the stress
eld in the solid can also be solved.
Specied for axisymmetric impact (Fig. 9), the origin O is
located at the point where the liquid sphere and solid plane
just contacts each other at the beginning of impact. The x axis
is normal to solid plane and also along the impact direction,
and the r axis is radial direction. Due to elastic deformation,
the solid wall is moving with an impact velocity v0 . In the
solid region, the displacement solution of elastodynamics
(the Lame equations) is used so as to ensure consistent
movement of mass points along the interface. The governing
equations and coupling conditions of axisymmetric liquidsolid impact are:
8
>
<

@2 U~s
@t2
pjx0

 rr  U~s r2 U~s
s

>
: ~
Us jt0 0; xr jx0 0; x jx0

23

Here, U~s is the elastic displacement vector of solid mass point, with components Us and Vs in x and r directions, respectively.
The velocity in hoop ( ) direction is zero due to axisymmetry. Denote the solid Youngs modulus as Es and Poissons ratio as
Es
s Es
s ,  1s12
and  21
are the Lame elastic coecients of solid. x jx0 and xr jx0 are the normal and shear stress
s
s
components at the solid surface.
The in situ coupling between liquid and solid are realized through the last equation in liquid region (via velocity coupling)
and last equation in solid region (via stress-pressure coupling) inside the contact area; outside the contact area the solid normal
stress is zero at the interface. In each time step, the geometry of contact area is updated using iteration to ensure consistency of
the above equations. The equations are solved using the nite dierence method, and the boundary conditions are exchanged
on the interface.
Once the transient displacement eld inside the solid is obtained, the strain eld is readily obtained, followed by the stress
eld. A useful measure of the magnitude of the multiaxial stress eld at a material point is the equivalent stress:
q
2 2 2
e x  r 2 =2 r   2 =2   x 2 =2 3xr
24
r
x
In what follows, the equivalent stress in solid and pressure in
liquid are both normalized by the water hammer force
0 c0 v0
(with
0 and c0 the undisturbed density and sonic speed
in liquid, Section 3.3), and the dimensionless stress and
pressure are:
 e =
0 c0 v0

and

P p=
0 c0 v0

25

It will be shown that the maximum pressure and maximum

equivalent stress occur within the liquid and solid, respectively. The dimensionless pressure and stress are expressed as

the dimensionless coordinates X and R, and the dimensionless time T:

X x=r0 ;

R r=r0 ;

T c0 t=r0

26

where r0 is the initial radius of liquid drop.

4.3 Simulations of water drop-1Cr 13/Ti-6Al-4V impact
Although the exible model and numerical framework
developed in this study can be readily extended to any liquid
and elastic solid, we use two representative blade materials,

1612

Q. Zhou et al.
The physical properties of 1Cr13 and Ti-6Al-4V.

4620

1000

Sonic speed (m/s)

5096

5010

1430

Acoustic impedance (kg/m2 s)

3:924  107

4:255  107

1:430  106

2  10
468

11

1:16  10
586

0.8790

1.172

1.465

1.758

2.051

-0.2
0.05
0.04
0.03
0.02
0.01
0.00

0.0

Pressure field in liquid

Water drop surface

0.00

2.344

2.637

0.2
0.05
0.04
0.03
0.02
0.01
0.00

2.930

0
0.2360
0.4720

0.00

Interface

0.7080

0.05

0.05

0.10

0.10

0.15

0.15

0.9440
1.180
1.416
1.652
1.888

Stress field in solid

0.20
-0.20

2.124
0.20

-0.15

-0.10

-0.05

0.00

0.05

0.10

0.15

2.360

0.20

(a)

0.2680

0.5360

0.8040

1.072

1.340

1.608

1.876

2.144

2.412

2.680

R
-0.2
0.05
0.04
0.03
0.02
0.01
0.00

0.0

Pressure field in liquid

Water drop surface

0.00

0.2
0.05
0.04
0.03
0.02
0.01
0.00

0
0.2240
0.4480

0.00

Interface

0.6720

0.05

0.05

0.10

0.10

0.15

0.15

0.8960

1.120
1.344
1.568
1.792

Stress field in solid

0.20
-0.20

2.016
0.20

-0.15

-0.10

-0.05

0.00

0.05

0.10

0.15

2.240

0.20

(b)

Dimensionless depth, X

Fig. 10 The dimensionless pressure in liquid and equivalent stress in solid

(for r0 1 mm, v0 100 m/s, and t 9 ns) with dierent solid materials:
(a) 1Cr13; (b) Ti-6Al-4V.

0.04
0.03
0.02
0.01
0.00
0.00

0.01

0.02

0.03

0.04

0.05

0.06

0.07

0.05

0.06

0.07

Dimensionless radius, R

(a)
Dimensionless depth, X

the high strength stainless steel 1Cr13 (ASTM #S41000) after

quenching and tempering, and Ti-6A1-4V, to demonstrate
the results of the liquid-solid impact. The main material
properties (measured from experiments) are shown in
Table 3, along with that of water.
As the particle size is varied, the dimensionless parameters, including the dimensionless time when shock wave
breaks o the solid, maximum interface pressure and the time
of its appearance, maximum stress and the time/position of
its occurrence, are found to be invariant and they are only
functions of impact velocity. This nding is supported by
experimental data in,24) where the water drops with dierent
diameters have the same contact angle of lateral jetting. Thus,
in the following examples, we only use one water drop size
(1 mm) during simulation and all results are expressed in
dimensionless terms.
We rst x the impact velocity at a representative v0
100 m/s. Figure 10 shows a snap shot of the pressure
distribution inside water drop and the stress eld inside
elastic solid, at 9 ns after impact. The pressure in water drop
increases away from the origin, and within the contact area,
the highest pressure is found at the edge. Semicircular-shaped
pressure waves are formed in the disturbed liquid region
because of the moving disturbing source (contact point), and
the cap-like shock wave front is the envelope of the pressure
wave. The stress distribution in solid exhibits an interference
eld of two symmetrical wave sources. There are two
prominent high stress regions: the rst one is near the contact
edge, where the peak impact pressure is also found which acts
as the primary wave source; the second one (which is
important for erosion) is near the axis and below the surface,
caused by the superposition of stress waves. For these two
solid materials, the patterns of pressure and stress are slightly
dierent, because their sonic speeds are relatively close. The
stress in Ti-6Al-4V is a little bit lower than 1Cr13 because of
its smaller modulus (see Table 3).
After the stress eld of every time step is solved, the
maximum equivalent stress point can be found and within
each time step, this point should be regarded as the most
dangerous point. If we plot the trace of such critical
dangerous point for every time step, then the inuence range
of the critical impact stress can be known. Figure 11 shows
the moving traces of the maximum equivalent stress point at
each time step, with an interval t 0:05 ns. The basic
length against the dimensionless depth X and dimensionless
radius R is the diameter of water drop. The inuence zone of
impact in the solid mainly develops in the redial direction and
near to the surface. The two materials have similar traces and
inuence zone because of their close sonic speed.
The solution of liquid-solid impact depends strongly on the
impact speed v0 . With c0 the sonic speed in the undisturbed

0.5860

Youngs modulus (Pa)

Fatigue limit/threshold (MPa)

11

0.2930

7700

Water

Ti-6Al-4V

Density (m3 /kg)

1Cr13

Table 3

0.04
0.03
0.02
0.01
0.00
0.0

0.01

0.02

0.03

0.04

Dimensionless radius, R

(b)
Fig. 11 The moving trace of the maximum equivalent stress points
(for r0 1 mm, v0 100 m/s): (a) 1Cr13; (b) Ti-6Al-4V.

liquid, the impact Mach number is Ma0 v0 =c0 . We dene 3

important dimensionless parameters to characterize the main
features of impact:
(i) The maximum dimensionless stress in solid, inf , is the
maximum dimensionless stress in solid (at all time steps
before the shock wave breaks o), inf e,max =

Dimensionless stress, inf

6
5
4

0.4188

inf=9.89132Ma0

0.40202

inf =9.12974Ma 0

3
2
1Cr13
Ti6Al4V

1
0
0.0

0.1
0.2
0.3
Dimensionless Mach number, Ma0

0.4

Dimensionless influence radius and depth, Rinf , Linf

Water Drop Erosion on Turbine Blades: Numerical Framework and Applications

Ti-6Al-4V: Linf 1:10598Ma0  8:414Ma20

0.20
Fitting formula R inf =0.37957Ma0.6904

0.18
0.16

19:03864Ma30

Numerical simulation results of R inf

0.14
0.12
Numerical simulation
results of L inf

0.10
0.08
0.06
0.04

1Cr13
Ti6Al4V

0.02
0.00
0.0

0.1
0.2
0.3
Dimensionless Mach number, Ma0

(a)

0.4

(b)

Fig. 12 Important parameters in the solid with respect to the impact Mach
number Ma0 : (a) the maximum dimensionless stress; (b) inuence radius
and depth.

0 c0 v0 , where e,max is the maximum equivalent stress

in solid at all time steps.
(ii) The dimensionless inuence radius, Rinf , is the maximum dimensionless radius reachable by the maximum
stress (before the shock wave breaks o), Rinf
rmax =r0 .
(iii) The dimensionless inuence depth, Linf , is the maximum depth at which the maximum stress may achieve
before the shock wave breaks o, Linf lmax =r0 .
Figure 12(a) shows the maximum stress in the solid, which
varies almost monotonically with the impact Mach number.
Ti-6Al-4V has lower stress than 1Cr13, because it has lower
Youngs modulus. The data may be tted as a power-law
relationship:
1Cr13: inf 9:89132Ma0:4188
0
Ti-6Al-4V: inf

1613

9:12974Ma0:40202
0

27
28

The dimensionless inuence radius and depth are given in

Fig. 12(b). The two materials have almost same curves
because of their similar sonic speeds. The inuence radius
basically depends on the movement of liquid-solid contact
area rather than the properties of solid material, thus dierent
solid materials should have almost the same inuence radius.
Whereas, the inuence depth primarily depends on the sonic
speed of solid, and the inuence depths of 1Cr13 and Ti-6Al4V are somewhat dierent.
When the impact velocity is low (Ma0 < 0:07), the
impact-aected region in solid develops in both radial and
axial (depth) directions. Although the maximum stress can
reach a depth up to about 0.08, the magnitude of such stress is
low which may not yet be sucient to cause damage. When
the impact velocity is moderate (Ma0 0:07
0:2), the
inuenced region mainly develops in the redial direction and
near the solid surface. When the impact velocity is very
high (Ma0 > 0:2), the aected zone again expands on both
directions the deep depth achievable plus the large
magnitude of resulting stress make the high velocity impact
most dangerous, and repetitive impact may cause the material
near the surface to spall o (with a thickness several percent
of the water drop radius); this is water drop erosion. A
qualitative tting can be obtained by:
1Cr13 and Ti-6Al-4V: Rinf 0:37957Ma0:6904
0
1Cr13: Linf 1:21536Ma0  9:24615Ma20
19:82268Ma30

29
30

31

In this Section, we established a nonlinear wave model to

solve the fully coupled liquid-solid impact problem, and
detailed information of the stress eld in solid and pressure
eld in liquid are obtained. The main features of the
solutions, eqs. (27)
(31), which characterizes the peak
stresses occurred during water drop-1Cr13 and Ti-6Al-V
impact, are expressed in dimensionless terms. When analyzing the fatigue limit, such explicit information can be used to
calculate the dimensional values of the impact stress and
inuence range, and then substituted into a fatigue model to
evaluate the erosion process in a typical blade channel made
by a typical material. If one uses a dierent material for the
blade, similar results may be generated by following the
framework in this paper.
5.

Water Drop Erosion Analysis in a Typical Blade

Channel

5.1 A fatigue model

Upon repetitive and high-frequency impact, the erosion
process of a typical turbine blade are similar to fatigue29,30)
and thus water drop erosion can be explored via fatigue
analysis.31) For water drop-1Cr13 and water drop-Ti-6Al-V
impacts, the maximum impact stress and inuence zone size
(Section 4) can be superimposed with the statistics of impact
(Section 3). Since the stress eld is multiaxial and relatively
complex, we use the equivalent stress as a caliber to analyze
the fatigue of solid. Although the magnitude of maximum
equivalent impact stress is found to be irrelevant of the
diameter of water drop, the position of critical stress point
scales with the diameter, which is also related with fatigue
analysis. Since the sizes of water drop as well as its inuence
zone are much smaller than the actual blade dimension, for
simplicity, we ignore the dierence of the inuence zone size
of dierent particle sizes, and assume each particle would
only cause damage directly beneath it. Thus, the problem of
water drop erosion is transformed to the problem of fatigue
cracking of the solid material under the inuence of impact
stress.
A simple fatigue model is established, from which the
lifetime of the specimen is32)

2
2
2

 1n
1n
Ni Ci eqv  eqv th
32
Here, Ni is the fatigue lifetime; Ci is the crack resistance
coecient; eqv th is the stress range of fatigue crack
threshold, which is related with the fatigue limit of material.
When eqv  eqv th , Ni ! 1.
5.2 Erosion lifetime map
The material and fatigue parameters are obtained from our
experiment. For 1Cr13: eqv th 468 MPa, Ci 24:3 
1014 , and n 0:085. For Ti-6Al-V, eqv th 586 MPa,
Ci 24:3  1014 , and n 0:043. Using 1Cr13 as an example, from normal impact tests, it is found that when the speed
of incident water drop is below vds 150 m/s for 1Cr13, no
apparent erosion could be observed experimentally. Accord-

Q. Zhou et al.
1.0

Top

33

8.000

8.000

12.00

12.00

12.00

16.00

0.4

28.00

0.2

34

32.00

32.00

36.00

36.00

36.00

0.0

37

Since the number ux (N0 ) and diameter of water drops are

dicult to be measured accurately at the entrance, for
convenience, an empirical coecient Cn is introduced and
the lifetime counted by operating hours is (Nf is the impact
times to failure)
t Nf =N Cn Nf =s

(hour)

38

The value of Cn can be obtained from engineering practice

or experimental data. We estimate the value of Cn of the
target turbine blade is about 10  105 hour (under operating
condition No. 1). Due to the value of enthalpy of steam of
reduced operating condition No. 2, the number of water drops
N0 is estimated to be 2 times of that of No. 1.
Figure 13 shows lifetime distributions on back/inner arcs
upon dierent operating conditions of 1Cr13. On both arcs
with full load (Fig. 13(a), (b)), because the impact speed
is less than the threshold speed, the minimum lifetime of
whole surface is more than 40  105 hours. However under
the operating condition No. 2 (Fig. 13(c), (d)), due to high
impact speed and high impact frequency, the minimum
lifetime of the surface is about 1:6  105 hours. For Ti-6Al4V (Fig. 14), the corresponding value is 2:2  105 hours
which is 37% longer than 1Cr13. For both materials, the
reduced operating condition is always more dangerous, and

0.5

16.00
20.00
24.00

0.4

28.00
32.00
36.00

0.2

40.00

40.00

0.0
0.0

1.0

Bottom

(b)

0.0

0.5

0.0

1.0

Bottom

(c)

0.5

1.0

(d)

(Unit: 10 5 operating hours Water drop diameter 50m)

Fig. 13 Lifetime map of metal surface in terms of operating hours, for

water-1Cr13 impact: (a) back arc upon operating condition No. 1; (b) inner
arc upon operating condition No. 1; (c) back arc upon operating condition
No. 2; (d) inner arc upon operating condition No. 2.

Top

1.0

Top

1.0

1.0

12.00

12.00

12.00

16.00

0.6

20.00

28.00

36.00

36.00

0.2

0.0
0.0

0.0
0.0

0.5

0.4

28.00

1.0

Bottom

36.00

0.2

(a)

1.0

(b)

20.00
24.00
0.4

28.00
32.00
36.00

0.2

40.00

0.0

0.5

16.00

32.00

40.00

40.00

12.00
0.6

24.00

32.00

32.00

Bottom

20.00

Z
0.4

8.000

16.00

0.6

24.00

28.00

0.2

0.8
8.000

24.00
0.4

4.000

0.8

8.000

20.00

4.000

0.8

8.000

16.00

1.0

4.000

4.000
0.8

0.6

Top

Top
0

36

(1/hour)

Bottom

(a)

The value predicted from our model is 9% lower than the

experimental value, which should be regarded as a good
agreement given uncertainties (e.g. surface roughness, water
lm on blade surface, stress concentration, etc.) encountered
during the experiment, and thus validated that our multidisciplinary models and numerical framework could eectively capture the most dominant aspects of water drop
erosion. After same procedure, the threshold impact speed vds
of Ti-6Al-4V can be obtained as 159 m/s. It is higher than
1Cr13 because of its higher fatigue limit value.
By employing the impact velocity/distribution of water
drops on the blade surface (Section 3), the erosion lifetime
map turbine blade in the whole wet steam turbine stage is
obtained as a function of impacted times. The result can be
further translated to that represented by operating hours. With
respect to eq. (13), if the entrance ux of water drop (N0 =A0 )
is given, the impact times per hour on an area of interest (A)
can be calculated by:
N s N0 A=A0

0.0
0.0

1.0

35

Finally, the threshold impact speed vds of 1Cr13 can be

calculated:
vds 137 m/s

inf 0:674738v1:4188
MPa
0

0.5

0.2

40.00

0.0

Bottom

We rewrite eq. (27) in dimensional form:

12.00

28.00

32.00
0.2

8.000

0.6

24.00
0.4

28.00

40.00

n ds 726:85 MPa

20.00

24.00

24.00

And thus the threshold of the nominal stress range are

obtained from the fatigue model:

4.000

16.00

0.6

20.00

20.00

0
0.8

0.8
8.000

0.6

1.0

4.000

0.8

16.00

0.4

Top
0

4.000

4.000
0.8

0.6

1.0

Top
0

eqv th 468 MPa

1.0

Top

ing to eq. (32), if the water erosion fatigue limit is set as

Ni 107 , the threshold value of the equivalent stress range
can be calculated as:

1614

0.0

Bottom

40.00

0.0
0.5

1.0

0.0

Bottom

(c)

0.5

1.0

(d)

(Unit: 105 operating hours Water drop diameter 50m)

Fig. 14 Lifetime map of metal surface in terms of operating hours, for

water-Ti-6Al-4V impact: (a) back arc upon operating condition No. 1; (b)
inner arc upon operating condition No. 1; (c) back arc upon operating
condition No. 2; (d) inner arc upon operating condition No. 2.

the inner arc suers to more severe damage than the back arc.
Finally, the most dangerous water drop erosion region and
operating condition can be deduced: Under reduced load
condition, the head of back arc of the blade and the band
that spans from the bottom front to the middle of trailing
edge are likely to suer the most severe water drop erosion,
with the second region more critical.
6.

Conclusion

In this study, we developed multidisciplinary models and

numerical tools to study the water drop erosion in turbine
engines. Computational uid dynamics and particle model
are employed to simulate the trajectories of water drops in
a ow of wet steam in the blade channel, and the most
important statistics of impact are obtained. A nonlinear wave
model and relevant analytical and numerical algorithms are
established to explore the fundamental aspects of liquid dropsolid impact, which leads to the erosion analysis based on a
fatigue model. Results are specied for water drop normal
impact on two representative blade alloys, 1Cr13 and Ti-6Al4V, and the lifetime distribution on blade surface under
typical working conditions are obtained. Important conclusions include:
(1) Once the initial diameter of water drops gets larger
than 50 mm, the distribution maps of impact frequency

Water Drop Erosion on Turbine Blades: Numerical Framework and Applications

and velocity on the blade surface are essentially the

same. In other words, the statistics of impact frequency and velocity of large water particles may be
replaced by that of a xed particle size (e.g. 50 mm),
which may save computational time considerably.
Regions where high impact velocity and frequency
coexist may represent potentially dangerous areas for
water drop erosion.
(2) The nonlinear wave model could capture the most
essential characteristics of high-velocity liquid-solid
impact, by ignoring the viscosity of uid. The 3D
model fully couples liquid pressure/compressibility
and solid stress/deformation, and the solution focuses
on the transient eects which dominate the highvelocity impact process. The transmission speed of
shock wave in the radial direction is faster than that
in the axial direction in the water drop. Two high
transient stress regions are identied in the solid:
one near the edge of the contact area and the other
one near the normal axis (and several mm below the
surface).
(3) The dimensionless parameters dened in this paper can
well describe the impact procedure for particles of
all sizes. The characteristic dimensionless scale (e.g.
inuence zone radius and depth), dimensionless time
(e.g. time when shock wave breaks o the liquid body),
dimensionless pressure (e.g. the maximum pressure at
contact surface), and dimensionless stress (e.g. the
maximum transient equivalent stress inside solid) are
nonlinear functions of the impact velocity. Comparisons between impacts on 1Cr13 and Ti-6Al-4V are
carried out, whose dierence mainly arises from their
dierent moduli. The stress patterns are close in these
two materials, because their sonic speeds are close.
The maximum dimensionless stress in the solid varies
nonlinearly with the impact Mach number, in a form
close to power-law, inf 9:89132Ma0:4188
for water0
1Cr13 impact and inf 9:12974Ma0:40202
for
water0
Ti-6Al-4V impact. For other materials similar functions
may be tted by following the same procedures using
the numerical framework.
(4) With the development of a fatigue model, the most
dangerous water drop erosion regions are deduced for a
representative steam turbine upon typical operating
conditions. The maps of erosion lifetime in terms of
operations hours are obtained. Under a reduced load,
the head of blade top (back arc) and the band that spans
from the bottom front to the middle part of trailing edge
are most critical regions for water drop erosion, which
agrees with previous experimental observations. Thus,
in order to alleviate water erosion, the blade should
avoid to be operated under reduced load. The erosion
resistance is material-dependent: although having similar sonic speed, the life time of Ti-6Al-4V is 37%
longer than 1Cr13.
The analytical and numerical framework established in
this paper is quite exible and it may be extended to other
liquid-solid impact/erosion problems, such as rain drop
erosion.

1615

Acknowledgement
The work is supported in part by Hitachi Ltd., in part
by the National Natural Science Foundation of China
(50276051), the National Basic Research Program of
China (2005CB221206), U.S. National Science Foundation
CMS-0407743 and CAREER-CMMI-0643726, and in part
by the Department of Civil Engineering and Engineering
Mechanics, Columbia University.
REFERENCES
1) B. Stanisa and V. Ivusic: Wear 186187 (1995) 395400.
2) S. V. Patankar: Numerical Heat Transfer and Fluid Flow. (1980,
Hemisphere Publishing Corporation).
3) I. Demirzic, Z. Lilek and M. Peric: International Journal for Numerical
Methods in Fluid 16 (1993) 10291050.
4) T. Art: Trans of ASME, Journal of Gas and Power 107 (1992) 286.
5) J. D. Denton: Trans of ASME, Journal of Turbomachinery 114 (1992)
18.
6) W. Tao: Numerical Heat Transfer. (1988, Xian, China: Xian Jiaotong
University Publishing Company).
7) Y. Liu: Calculation Methods of Inner 3-D Flow Field under Stage of
turbomachine. 1996, Xian Jiaotong University: Xian.
8) H. Shi: The thoeretical investingation and experimental research on the
water drop impact destroy of blades in the wet steam turbine stage.
1989, Xian Jiaotong University: Xian, China.
9) I. G. Chatzilamprou, M. W. Youds, M. J. Tierney and B. Armstrong:
Applied Mathematical Modelling 30 (2006) 11801195.
10) B. Stanisa, Z. Schauperl and K. Grilec: Wear 254 (2003) 735741.
11) F. Pang: Nuclear Power Engineering 13 (1992) 3.
12) H.-Y. Kim, S.-Y. Park and K. Min: Rev. Sci. Instrum. 74 (2003)
49304937.
13) H.-H. Shi, K. Takayama and N. Nagayasu: Wear 186187 (1995)
352359.
14) X. Chen and J. W. Hutchinson: Journal of the Mechanics and Physics of
Solids 50 (2002) 26692690.
15) X. Chen, M. Y. He, I. Spitsberg, N. A. Fleck, J. W. Hutchinson and
A. G. Evans: Wear 256 (2004) 735746.
16) S. S. Cook: Proceeding of the Royal Society A 119 (1928) 481488.
17) O. G. Engel: J. Res. Nat. Bur. Stand. 54 (1955) 281.
18) R. M. Blowers: Int. J. Inst. Maths Applications 5 (1969) 167193.
19) M. M. Grant and P. A. Lush: Journal of Fluid Mechanics 176 (1987)
237252.
20) M.-K. Lee, W.-W. Kim, C.-K. Rhee and W.-J. Lee: Nuclear Engineering and Design 214 (2002) 183193.
21) O. G. Engel: J. Appl. Phys. 44 (1973) 692704.
22) F. J. Heymann: Trans of ASME Journal of Basic Engineering 90 (1968)
400402.
23) H.-S. Kim, J.-S. Kim, H.-J. Kang and S.-R. Kim: Trans of ASME
Journal of Applied Mechanics 68 (2001) 346348.
24) J. E. Field, J. P. Dear and J. E. Ogren: J. Appl. Phys. 65 (1989)
533540.
25) J. E. Field: Wear 233235 (1999) 112.
26) F. J. Heymann: J. Appl. Phys. 40 (1969) 51135122.
27) E. J. Finnemore and J. B. Franzini: Fluid Mechanics With Engineering
Applications ((10th edition). 2001, McGraw-Hill).
28) P. G. Tait: Physical Chemistry 2 (1888) 117.
29) J. H. Brunton and M. C. Rochester: Treatise on Materials Science and
Technology, ed. by C. M. Preece, (Academic Press, New York, 1976)
p. 220.
30) F. J. Heymann: Proc. 2nd. Meersburg Conf. on Rain Erosion and Allied
Phenomena, Royal Aircraft Establishment. 1967. Farnborough, England.
31) G. S. Springer: Erosion by Liquid Impact. (1976, Washington D.C.,
Scripta Publishing Company, John Wiley & Sons).
32) The detailed derivation is at http://www.civil.columbia.edu/xichen/
mater trans supplement.pdf