You are on page 1of 8

Engineering Failure Analysis 17 (2010) 192–199

Contents lists available at ScienceDirect

Engineering Failure Analysis


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

Fatigue of piston rod caused by unsteady, unbalanced, unsynchronized


blade torques in a Kaplan turbine
Luo Yongyao a, Wang Zhengwei a,*, Zeng Jidi b, Lin Jiayang b
a
Department of Thermal Engineering, Tsinghua University, Beijing 100084, PR China
b
Fujian Shuikou Hydropower Station, Fuzhou 350800, PR China

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

Article history: The stresses in Kaplan turbine piston rod were analyzed to predict failure conditions based
Received 7 May 2009 on the 3D unsteady flow through the Kaplan turbine. The results show that the predicted
Accepted 4 June 2009 stress concentration position in the rod agrees well with the actual fracture position. The
Available online 11 June 2009
mean stress and the dynamic stress in the rod varied greatly with operating conditions.
The dynamic stress reached 46.0 MPa at the high head low output condition, which caused
Keywords: the final rod failure. The mean of the dynamic stresses in the rod increase with increasing
Kaplan turbine
blade angle error, but the amplitudes decrease.
Piston rod
Fatigue fracture
Ó 2009 Elsevier Ltd. All rights reserved.
Unsynchronized
Dynamic stress

1. Introduction

Kaplan turbines are widely used in low head hydropower stations since they can operate over a wide range with high average
efficiency. To achieve this, the Kaplan turbine has movable blades and a very complex blade-control system inside a hollow hub
and axis, the main parts of which are shown in Fig. 1. The turbine was put into operation in November, 1995. On February 10,
2000, the turbine was found to be vibrating severely. The oil level in the oil collection groove then rapidly dropped. Upon open-
ing the housing, the central axis of the piston rod was found to have broken at the joint of the M540 nut with the crosshead fallen
into the cone as shown in Fig. 2a. The nut structure was then changed to a retainer ring structure [1]. However, 5 years later, on
March 16, 2006, the retainer ring structure also broke at the same location as shown in Fig. 2b. A comprehensive investigation
was then performed including dynamic analyses of the piston rod to identify the design problem.
A detailed inspection of the two cases showed that the broken sections showed characteristics of fatigue failure due to
alternating loads. The most probable reason was the unsteady pressure loads on the blades. The blades experience high
hydraulic pressures as they convert fluid energy to mechanical energy. In the blade-control system, the hydraulic pressure
on the blades creates torque on the blade pivot which is transferred to the other parts of the system. In addition, the blade
torques on the six blades are unbalanced and unsynchronized. A dynamic analysis of the design not only requires stress com-
putations but also analysis of the flow path through the entire turbine.

2. Hydraulic torque on the piston rod

The flow path is from the spiral casing inlet to the draft tube outlet. The whole path was discretized with an unstructured
hybrid mesh of hexahedral and tetrahedral elements. The final mesh had about 730,000 nodes as shown in Fig. 3. The water

* Corresponding author. Tel./fax: +86 10 62791262.


E-mail address: wzw@tsinghua.edu.cn (Z. Wang).

1350-6307/$ - see front matter Ó 2009 Elsevier Ltd. All rights reserved.
doi:10.1016/j.engfailanal.2009.06.003
Y. Luo et al. / Engineering Failure Analysis 17 (2010) 192–199 193

Fig. 1. Blade control system for one blade.

Fig. 2. Fractured piston rods.

Fig. 3. Calculation domain for the flow field simulation in the Kaplan turbine.

was assumed to be incompressible. A time-dependent Reynolds average Navier–Stokes (RANS) simulation was performed
with the SST k–x turbulence model [2] used to calculate the flow field through the Kaplan turbine.
The calculations were performed for 13 operating conditions. For all cases, after several rotational periods, the frequencies
and amplitudes of the pressure oscillations on the monitor nodes changed very little. The six blades torques, Tb, were then
recorded for three periods. The blade numbers are shown in Fig. 4. The computed torques for three typical operating condi-
tions are shown in Fig. 5. The operating parameters of the three selected conditions are listed in Table 1. For conditions near
the optimum operating point and those with large blade angles and guide vane openings, the torques on the blade pivot fluc-
194 Y. Luo et al. / Engineering Failure Analysis 17 (2010) 192–199

Fig. 4. Blade numbers.

tuated with the runner rotation frequency, fn, with a phase angles between the torque fluctuations on adjacent blades of 60°,
as shown in Fig. 5a and b for conditions GK8 and GK11. However, for small blade angles and small guide vane openings, the
blade torque fluctuations were accompanied by increased turbulence with 120° phase angles between the oscillations so
that the phases of #1 and #4 are the same, the phases of #2 and #5 are the same, and the phases of #3 and #6 are the same
as shown in Fig. 5c for condition GK3, due to a vortex in the flow tube.
Fig. 5 also shows that the mean values and amplitudes of the torque fluctuations were small for conditions near the opti-
mum operating point (GK8), were larger for large blade angles and large guide vane openings (GK11) and were even larger
for small blade angles and small guide vane openings (GK3), which was verified by the power station’s operating recordings.
When the turbine was operating near the optimum (for example, GK8) the operating oil pressure inside the piston chamber
was normal, but when operating near condition GK3 the oil pressure inside the chamber was extremely high.

3. Dynamic stresses in the piston rod

The transient dynamic equilibrium equation for the stresses in a linear structure is:
€ g þ ½Cfug
½Mfu _ þ ½Kfug ¼ fF t g ð1Þ
_
where [M] is the mass matrix, [C] is the damping matrix, [K] is the stiffness matrix, {u} is the nodal displacement vector, fug
€ g is the nodal acceleration vector and {Ft} is the load vector on the nodes.
is the nodal velocity vector, fu
Eq. (1) was discretized using the FEM method and solved using the Newmark method. The iterative equation was [3]:
     
1 d 1 1 1
½M þ ½C þ ½K funþ1 g ¼ fF t g þ ½M fun g þ f _
u n g þ  1 fu€ n g
aDt2 a Dt a Dt 2 aDt 2a
     
d d Dt d
þ ½C fun g þ  1 fu_ n g þ  2 fu€n g ð2Þ
aDt a 2 a
where a and d are the Newmark integration parameters and Dt = tn+1  tn.
Eq. (2) shows that the effects of the velocity and acceleration terms are considered in the transient displacement compu-
tation. {u} was obtained from Eq. (2) with {r} calculated at every time step from:
frg ¼ ½D½Bfug ð3Þ
where [D] is the elastic matrix based on Young’s modulus and Poisson’s ratio for the material and [B] is the strain–displace-
ment matrix based on the element shape functions.
The dynamic stresses in the piston rod were computed based on the fourth strength theory. The von Mises or equivalent
stress, re, was computed using [4]:
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1h i
re ¼ ðrx  ry Þ2 þ ðry  rz Þ2 þ ðrz  rx Þ2 þ 6ðs2xy þ s2yz þ s2xz Þ ð4Þ
2
The piston rod material was hammered steel 20SiMn with the properties listed in Table 2.
The piston rod, retainer ring, crosshead and key were connected with contact elements to form a multi-body system. The
system was discretized using 10 node tetrahedral elements with the mesh shown in Fig. 6.
Y. Luo et al. / Engineering Failure Analysis 17 (2010) 192–199 195

Fig. 5. Torques on all six blades for three typical conditions.

4. Results

The displacement and stress distributions in the piston rod are shown in Fig. 7 for condition GK11, where the maximum
static stress for all conditions is at node 22,289. The dynamic stresses at node 22,289 for the three conditions are shown in
196 Y. Luo et al. / Engineering Failure Analysis 17 (2010) 192–199

Table 1
Operating parameters for three typical operating conditions.

Operating condition H (m) Power (MW) Opening (%) Blade angle (°)
GK3 61 100 36.2 5.6
GK8 53 150 59.0 11.7
GK11 43.8 215 90.5 23.3

Table 2
Piston rod physical properties.

Density (kg/m3) Young’s modulus (Pa) Poisson’s ratio


7.8  103 2.1  1011 0.3

Fig. 6. System mesh.

Fig. 7. Calculational results for the piston rod for case GK11.
Y. Luo et al. / Engineering Failure Analysis 17 (2010) 192–199 197

Fig. 8. The mean stresses and the amplitudes (with 95% probability) of the dynamic stresses at node 22,289 are summarized
in Table 3.
Comparison of Figs. 2 and 7 shows that the predicted position of the stress concentration agreed well with the actual frac-
ture position. Fig. 8 and Table 3 show that the mean stresses and the dynamic stresses at node 22,289 were quite large and
varied greatly with the operating condition. The mean stresses and the dynamic stresses in the piston rod were much higher
at half the rated output, especially for case GK3. The amplitude of the dynamic stresses reached 46.0 MPa with a stress ratio

Fig. 8. Dynamic stresses for node 22,289.

Table 3
Maximum mean and dynamic stresses for the two structures.

Operating condition Mean stress Dynamic stress Stress ratio (%)


GK3 134.5 46.0 34.2
GK8 112.5 6.8 6.0
GK11 121.4 22.8 18.8
198 Y. Luo et al. / Engineering Failure Analysis 17 (2010) 192–199

(the dynamic stress divided by the mean stress) of 34.2%, which is very high for the piston rod. The water turbine was run-
ning in (Automatic Generation Control) AGC mode, so the load was always at about 100 MW (near GK3) and the mean and
dynamic stresses in the piston rod would have been very high during operation resulting in fatigue in the piston rod. The
crack first appeared at the location of the highest stress concentration and then grew due to the effects of the high dynamic
stresses from the shoulder and to the center. Finally, a shell line appeared at the fracture. Therefore, the high dynamic stres-
ses were the main reason for the fracture of the retainer ring structure piston rod and the rod broke because of high cycle
fatigue.

5. Effect of blade angle error on the dynamic stresses

According to the power station’s operating recordings, blade angle errors occur due to machining errors during adjust-
ment of the blades. The dynamic stresses in the piston rod were calculated for two hypothetical conditions to analyze the
effect of blade angle error. The two hypothetical conditions were based on the GK11 conditions with the #1 blade angle in-
creased by 0.5° and the #4 blade angle decreased by 0.5°, with this condition referred to as GK11-0.5. The second case had
the #1 blade angle increased by 1.0° and the #4 blade angle decreased by 1.0°, with this condition referred to as GK11-1.0.
The blades torques for the two conditions are shown in Fig. 9, with the torques for three operating conditions compared in
Table 4. Figs. 5c, 9 and Table 4 show that the amplitudes and phases of the torques varied little with the increased blade
angle error, but the mean value varied greatly, with the torque on the #1 blade increased the most and that on the #4 blade
decreased the most.
The dynamic stresses in the piston rod for the two blade angle error conditions were then calculated. The dynamic stres-
ses for node 22,289 shown in Fig. 10 show that the mean stress increased but the amplitude decreased with increasing blade
angle error.

Fig. 9. Torques on the six blades for the two blade angle error conditions.
Y. Luo et al. / Engineering Failure Analysis 17 (2010) 192–199 199

Table 4
Comparison of blades torques for three operating conditions (unit: T m).

GK11 GK11-0.5 GK11-1.0


Mean Amplitude Mean Amplitude Mean Amplitude
#1 76.6 6.5 80.5 6.7 84.5 6.9
#2 76.1 6.5 78.0 6.5 79.4 6.6
#3 75.9 6.5 79.8 6.6 83.6 6.5
#4 75.8 6.4 71.1 6.3 66.6 6.1
#5 76.2 6.4 74.7 6.6 73.4 6.6
#6 76.6 6.5 72.8 6.6 68.2 6.7

Fig. 10. Comparison of dynamic stresses for node 22,289.

6. Conclusions

The dynamic stresses in a piston rod of a Kaplan turbine were analyzed using CFD analyses of the flow coupled with stress
analyses of the rod subjected to unsteady, unbalanced and unsynchronized blade torques. The predicted position of the max-
imum stress concentration agrees well with the actual fracture position. In the AGC mode, the excessive dynamic stresses at
half rated output caused the crack to grow leading to fracture. Increases of the blade angle error increased the mean dynamic
stresses in the rod but reduced the amplitudes.

Acknowledgements

Special thanks are due to the National Natural Science Foundation of China (Nos. 50879036 and 50879088) and National
High Technology Research and Development Program of China (863 Program No. 2009AA05Z424) for supporting the present
work.

References

[1] Zhong Su. Failure analysis and design strength study of piston rod for the Shuikou hydroelectric power unit 6. Large Electr Machine Hydraul Turb
2001;7:7–11.
[2] Menter FR. Two-equation eddy-viscosity turbulence models for engineering applications. AIAA J 1994;32:1598–605.
[3] Hallquist John O. LS-DYNA theoretical manual. Livermore Software Technology Corporation; 1998.
[4] Yu MH. Engineering strength theory. Peking: Higher Education Press; 1999.

You might also like