TSINGHUA SCIENCE AND TECHNOLOGY

ISSN  1007-0214  02/25  pp587-592

Volume 13, Number 5, October 2008

Dynamic Stresses in a Francis Turbine Runner Based on

Fluid-Structure Interaction Analysis
XIAO Ruofu ()1,2, WANG Zhengwei ()2,**, LUO Yongyao ()2
1. College of Water Conservancy and Civil Engineering, China Agriculture University, Beijing 100083, China;
2. State Key Laboratory of Hydroscience and Engineering, Tsinghua University, Beijing 100084, China
Abstract: Fatigue and cracks have occurred in many large hydraulic turbines after they were put into production. The cracks are thought to be due to dynamic stresses in the runner caused by hydraulic forces.
Computational fluid dynamics (CFD) simulations that included the spiral case, stay vane, guide vane, runner
vane, and draft tube were run at various operating points to analyze the pressure distribution on the runner
surface and the stress characteristics in the runner due to the fluid-structure interactions (FSI). The dynamic
stresses in the Francis turbine runner at the most dangerous operating point were then analyzed. The results
show that the dynamic stresses caused by the hydraulic forces during off-design operating points are one of
the main reasons for the fatigue and cracks in the runner blade. The results can be used to optimize the
runner and to analyze other critical components in the hydraulic turbine.
Key words: Francis turbine; computational fluid dynamics (CFD); fluid-structure interaction (FSI); stress
characteristics

Introduction
In recent years, many large hydraulic turbines used in
China have experienced cracks in the runner blade,
which constitute a threat to safe operation of the hydraulic turbine units. Traditionally, stability evaluations
of hydraulic turbines have primarily depended on vibration and water pressure pulsation. However, the
present stress problems in the hydraulic turbines have
revealed that these two factors are not sufficient for
analyzing the fatigue and crack probabilities in the
runners. Many studies have demonstrated that runner
fatigue cracks result from static and dynamic stresses
on the runner starting from micro cracks that grow due
to the stresses[1,2].
Recently, many researchers have engaged in improving calculation and measurement methods for the
static and dynamic stresses in runner blades caused by
Received: 2006-09-01; revised: 2006-11-13

** To whom correspondence should be addressed.

E-mail: wzw@mail.tsinghua.edu.cn; Tel: 86-10-62791262

hydraulic forces. The static stresses in the Francis turbine runner at various operating points were calculated
using the sequential fluid-structure interaction (FSI)
analysis method[3]. The dynamic stress variations during load changes were also calculated using the sequential FSI analysis method[4]. However, these two
analyses did not account for runner pressure pulsations
caused by the spiral case, stay vane, guide vane, and
draft tube. Most recently, the dynamic stress characteristics on the runner blades in a large hydraulic turbine
have been measured at various operating points[5].
In this paper, a computational fluid dynamics (CFD)
simulation of the whole flow passage of the Francis
turbine was performed to obtain more accurate predictions of the pressure distributions and pressure pulsations on the runner blade. Then the FSI method was
used to calculate the stresses in the Francis turbine
runner to analyze the static stress characteristics in the
Francis turbine runner at various operating points.
Next, a full analysis was made to examine the dynamic
stresses at the most dangerous operating point of high

Tsinghua Science and Technology, October 2008, 13(5): 587-592

588

head and low load to identity the factors causing the

fatigue cracks in the runner blade.

Computational Method

1.1

Flow field simulation

where M is the mass matrix, C is the damping matrix,

{u} is the node velocities, and {u} is the node accel-

The first step is to determine which turbulence model

most accurately predicts the frequency and amplitude
characteristics of the water pressure pulsation. The
Reynolds stress and k- turbulence model which are
based on the Reynolds-averaged Navier-Stokes (RANS)
method are frequently used. However, some studies
have shown that the standard k- turbulence model
cannot accurately predict the water pressure pulsation
in the draft tube and that other k- models, such as the
realizable k- model, tend to have excessive dissipation
which attenuates the pressure pulsation. Others have
argued that the Reynolds stress model and large eddy
simulation (LES) give improved result[6].
For a real unit, it is impractical to employ the LES
method since it requires elements smaller than the energy but larger than the dissipation scale. Current
computers are not able to analyze such a large number
of elements. The RANS model, as a statistical turbulence model, requires much less computational resource than the LES model, and thus, the shear stress
transport (SST) k- turbulent model was used in the
present analyses[7].
1.2

The static stresses can be calculated using

K {u} {Fs }  {Ft }
V DB{u}

(1)
(2)

where K is the stiffness matrix, D is the elastic stiffness

matrix, B is the strain-displacement matrix, {u} is the
node displacement, {Fs} is the forces on the interface
between the fluid and the solid of the runner, and {Ft}
is the inertia force caused by the runners rotation and
gravity. The von Mises or equivalent stress, e, was
calculated by using the fourth strength theory,

Ve

1
[(V 1  V 2 )2  (V 2  V 3 )2  (V 3  V 1 )2  6(W xy2  W yz2  W zx2 )]
2
(3)
Table 1

Number of elements

erations. Equation (4) was solved using Newmarks

algorithm[8]. Equation (2) was solved to obtain the
stresses at each node at each moment. Then the von
Mises stress at each moment was calculated using the
fourth strength theory.
1.3

Spiral case
fluid domain
80 000

Sequential FSI solution process

There are two strategies for the solution of the fluidstructure interaction. One is the fully coupled method,
in which the fluid and structure are solved within the
same code and same grid. However, the calculation
based on the fully coupled fluid-structure interaction is
CPU-expensive due to the intensive iterations between
the fluid and structure system, especially for the large
Francis turbine runner. The other is sequential coupled
fluid-structure interaction, which assumes that the influence of the runner deformation on the flow field is
negligible in the calculations. That is to say, there is no
feedback of the runner blade motion on the flow. This
hypothesis is reasonable since self-excited oscillations
are beyond the scope of the present paper.

Stress analysis

Domain

A linear structural transient power balance equation

was used to calculate the dynamic stresses[8],
M {u}  C{u}  K {u} {Fs }  {Ft }
(4)

Computational Model and

Boundary Conditions

The model was based on a Francis turbine installed in

Shanxi Province, China. The reference runner diameter
was Dref =6 m. Three-dimensional geometric models
were built for the flow field and structural field in order to calculate the flow field in the Francis turbine
and in the structural field of the runner. Tetrahedral
meshes were used for both the flow field and structure
simulation. The meshes of the runner for the solid domain and the fluid domain were created together to
ensure the location load correspond at the fluid-solid
interface for accurate transmission of the water pressure load. The numbers of elements in the whole domain are listed in Table 1.

Number of elements in each domain

Fluid domain around the

stay vane and guide vane
100 000

Fluid domain around

the runner vane
150 000

Solid domain of the

runner vane
120 000

Draft tube fluid

domain
200 000

XIAO Ruofu ( )et alDynamic Stresses in a Francis Turbine Runner Based on Fluid-Structure

Therefore, the runner rotated once during 100 time

steps. The output data was saved at each step. To acquire a long period of irregular flow fluctuation in the
runner and draft tube, 2100 time steps were computed
corresponding to 21 rotations of the runner. To reduce
the computational complexity of the runner stress
analysis, only the result for the last 4 rotations were
used in the dynamic stress analysis.
The dynamic stresses in the runner of the Francis
turbine were then analyzed using the mesh shown in
Fig. 2. Since stress concentrations often occur at the
blade root, the fillet of the runner blade with the runner
crown and the runner band were modeled accurately.
These sensitive areas were more accurately meshed to
avoid stress concentrations due to the mesh. The constraint for the stress calculation was the fixed junction
between the runner crown and the main shaft. The
loads on the runner included the inertia force and the
surface forces. The inertia force included the runners
own weight and the rotational inertia force. The surface force was due to the water pressure and pressure
pulsation on the fluid-solid interface caused by flow,
with the water pressure obtained from the whole flow
passage simulation.

The computational mesh for the flow field simulation is shown in Fig. 1. The fluid domain for the
Francis turbine included the spiral case, the stay vane,
the guide vane, the runner vane, and the draft tube. The
total pressure and the velocity direction at the spiral
case inlet were used as the inlet conditions. The outlet
of the draft tube average static pressure was assumed
to be zero. Analysis of the flow in the whole flow passage made the internal flow field calculations more
accurate, providing a more accurate surface pressure
load prediction on the blade for the stress analysis.

Fig. 1 Computational mesh for the flow field calculation for the Francis turbine

The time step for the unsteady flow simulation was

equal to rotating the runner by 3.6 degrees per step.

Leading edge

Runner crown

Leading edge

Runner band
(a)

Fig. 2

589

(b)

(c)

Stress simulation mesh for the Francis turbine runner

Operating Points

To get a full understanding of the Francis turbine internal flow and runner stress characteristics, 11 operating points including the highest head 79 m, the rated
head 68 m, and the lowest head 51 m were chosen for
the simulations. These operation points are shown in
Fig. 3.

Fig. 3

Operating points for the simulations

Tsinghua Science and Technology, October 2008, 13(5): 587-592

590

Static Stress Characteristics

The steady-state flow and stresses were calculated for

the 11 operating points to analyze the static stress
characteristics in the Francis turbines runner for a variety of operating points. The results indicate that the
maximum stresses often concentrated on the link between the leading edge and the runner band as well as

on the trailing edge close to the runner crown. At the

low heads and low load operating points #1, #2, and #3,
the maximum von Mises static stresses on the runner
blade occurred on the link between the leading edge
and the runner band. At the other operating points, the
maximum von Mises static stresses occurred on the
trailing edge close to the runner crown, as shown in
Fig. 4.

Node 2983

(a) Operating point #1: Head = 51 m, P = 50 MW

Fig. 4

Node 323

(b) Operating point #4: Head = 51 m, P = 110 MW

Static stress distribution on the runner for different operating points

Figure 5 shows the variation of the von Mises static

stresses on nodes 2983 and 323 in the hydraulic turbine
at various operating points. Node 2983 was close to the
link between the leading edge and the runner band,
where node 323 was on the trailing edge close to the
runner crown, as shown in Fig. 4. Figure 5 indicates
that the von Mises static stresses on node 323 changed
linearly with the load, which means that the static
stresses on the trailing edge close to the runner crown
were mainly determined by the hydraulic torque on the
whole blade surface, while the pressure distribution on
the runner blade had little influence. For node 2983,
the static stresses on the link between the leading edge
and the runner band were almost linearly related to the
hydraulic turbine load at higher head. But at lower
heads, the variation was not linear. The pressure distribution on the runner vane surface had a considerable
influence, especially at the operating point #1. For the
Francis runner blade, the maximum stresses on the
runner blade in Fig. 5 show that at most operating
points the maximum static stresses are in general related with the turbine power for both low and high
heads. At a few operating points, especially low heads
and low loads, the pressure distribution on the runner
blade has a considerable influence on the stress distribution. Whether this regularity applies to other Francis
turbines needs further research.

Fig. 5 Variation of the maximum stress on the runner

blade in a Francis turbine

Dynamic Stress Characteristics

As has been noted, the operating point with the largest

XIAO Ruofu ( )et alDynamic Stresses in a Francis Turbine Runner Based on Fluid-Structure

dynamic stresses occurred at high heads and low loads.

To analyze the dynamic stress characteristics of the
Francis turbine runner blade, the unsteady flow was
first simulated at operation point #9 with high head and
low load. The pressure distribution on the surface runner vane was then obtained at various times to calculate the transient dynamic stress distribution at this
operating point.
The pressure pulses in the hydraulic turbine at operating point #9 were first related to the pressure pulses
on the draft tube access door, which are shown in
Fig. 6. The water pressure pulse amplitude on the draft
tube access door is 37 kPa with a dominant frequency
of 0.23fr (fr is the runner rotation frequency). The calculated results have been compared with measurements at various operating points[9]. The results revealed that for openings from 65% to 83%, the CFD
results for the pressure pulses amplitudes and frequency corresponded well to the measurements.
Therefore, the unsteady flow simulation is a reliable
prediction for the actual performance.
The helical vortex ropes in the Francis runner draft
tube at 5 different times are shown in Fig. 7. The low

(a) t =10.6 s

(b) t =11.1 s

Fig. 7

591

pressure helical vortex rope in the draft tube originated

near the runner cone and moved down in the draft tube
with time. The movement changed the water pressure
inside the draft tube, and eventually produced the water pressure pulses in the draft tube and at the runner
vane[10].

Fig. 6

(c) t =11.6 s

Pressure fluctuation at the draft able

(e) t =12.1 s

(f) t =12.6 s

Vortex rope motion law in the draft tube

Figure 8 shows the dynamic stresses on nodes 323

and 2983 where the static stresses were the highest. On
node 323, the dynamic stresses were very high with
amplitudes reaching 15 MPa. Note that fatigue cracks
were found at the location of node 323. The dominant
dynamic stress frequency was 0.77fr on nodes 323 and
2983. The sum of the dominant frequency of the dynamic stresses and the dominant frequency of the low
pressure vortex rope in the draft tube is equal to the
runner rotation frequency, which corresponds with the
measured dynamic stress frequency characteristics[5].
In this Francis turbine, whose runner blade was
made of ZG0Cr16Ni5Mo, the allowable stress was
196 MPa, with the ultimate failure stress of 780 MPa.

Although those values will be reduced when the runner

vane is submerged in water, the static stress analyses at
the various operating points showed that even for the
hydraulic turbine runner operating at maximum power,
the maximum von Mises static stress was far less than
the materials failure stress. Therefore, the runner vane
cracks were possibly not the static stresses. However,
the combined analysis for the static stresses with the
dynamic stresses on the runner vane showed that the
runner vane fatigue crack was produced by the combined force of the residual stresses, static stresses, and
dynamic stresses on the blade.
Through the simulations of the flow in the Francis
turbine and the stresses in the runner vane at various

592

operating points, we could accurately model the static

and dynamic stress characteristics in the runner vane.
The operating points with the highest static and dynamic stress or water pressure fluctuations could then
be found. The hydraulic turbine should then not be
operated at these dangerous load conditions to extend
the runner service life. The flow field and stress characteristics can also be employed in turbine designs to
optimize the runner vane structure and reduce the static
and dynamic stresses in the runner caused by the hydraulic effect to extend the runner blade life time.

Tsinghua Science and Technology, October 2008, 13(5): 587-592

entire flow passage to accurately model the runner

vane surface pressure distribution so that stresses in the
runner vane can be accurately calculated;
(2) The analysis of the runners static stresses field
for various operating points indicates that the maximum von Mises static stress is far less than the materials ultimate stress, so the runner blade cracks are not
caused by excessive static stresses;
(3) The analysis of the dynamic stresses on the runner indicates that the runner blades micro-cracks are
produced due to the combined residual stresses, static
stresses, and dynamic stresses on the runner blade.
References
[1] Xue W, Chen Z Y. Study on blade cracks of hydraulic turbine runner and prevention methods. Large Electric Machine and Hydraulic Turbine, 2002, 25(5): 42-45.
(in Chinese)
[2] Rodriguez C G, Egusquiza E, Escaler X, et al. Experimental investigation of added mass effects on a Francis turbine
runner in still water. Journal of Fluids and Structures,
2006, 22(2): 699-712.
[3] Luo Y Y, Wang Z W, Lian Q W. Stress of Francis turbine
runners under fluctuating work conditions. J. of Tsinghua
Univ. (Sci & Tech), 2005, 45(2): 235-237. (in Chinese)
[4] Wang Z W, Luo Y Y, Fang Y. FSI computation of a Francis
runner during load adjustment. Journal of Hydroelectric
Engineering, 2005, 24(4): 58-61.
[5] Pang L P. Study on runner stress measurement of
large-capacity turbine. Beijing: Tsinghua University, 2004.
(in Chinese)
[6] Vu T C, Shyy W. Navier-Stokes flow analysis for hydraulic
turbine draft tubes. Journal of Fluids Engineering, 1990,
12(2): 199-204.
[7] Menter F R. Two-equation eddy-viscosity turbulence models for engineering application. Journal of AIAA, 1994,
32(2): 1598-1605.
[8] Yu M H. Engineering Strength Theory. Beijing: Higher

Fig. 8 Dynamic stress analysis of the Francis turbine

runner vane

Conclusions

The fluid-structure interaction was analyzed to accurately calculate the flow field and static and dynamic
stresses in a Francis turbine runner vane. The results
show that:
(1) The flow field simulation needs to include the

Education Press, 1999. (in Chinese)

[9] Wang Z W, Zhou L J, He C L. Simulation and measurement of pressure oscillations caused by vortex rope in a
prototype draft tube. J. of Tsinghua Univ. (Sci & Tech),
2005, 45(8): 1138-1141. (in Chinese)
[10] Wang Z W, Zhou L J. Simulations and measurements of
pressure oscillations caused by vortex ropes. Journal of
Fluids Engineering, 2006, 128(4): 649-655.