Professional Documents
Culture Documents
a r t i c l e i n f o a b s t r a c t
Article history: In this paper the hydraulic stresses induced in a Francis turbine runner blade by steady fluid flow were
Received 2 November 2011 investigated. Based on the one-way coupled simulation, the approach consists of a fluid flow analysis
Received in revised form 15 May 2012 which provides the distribution of the fluid pressure on the blade, followed by the structural finite ele-
Accepted 31 May 2012
ment analysis. The three dimensional turbulent flow in both distributor and runner of Francis turbines
Available online 27 June 2012
were computed. The computational domains correspond to interblade channels for the Francis turbine
distributor and runner, respectively. In order to couple the steady absolute distributor flow field with
Keywords:
the runner steady relative flow, a mixing interface technique is used on the conical distributor–runner
Hydraulic turbine blade
Fluid flow analysis
interface. The hydrodynamic field is computed in seven operating points at constant head from part load
Structural stress analysis to full load conditions. The pressure coefficient distribution on the blade is plotted in order to evaluate
the blade loading and region with cavitational risk. Further, the stress distributions were obtained by a
structural finite element analysis performed for the steady loading in order to determine the areas with
highest stress values.
Ó 2012 Elsevier B.V. All rights reserved.
1. Introduction Several studies were reported in the last years for the computa-
tion and experimental measurement of the static and dynamic
Hydropower is the largest source of renewable energy and it is stresses induced in Francis turbine blades by the steady and the
the most efficient way to generate electricity. Besides being an unsteady loading, respectively. The static stresses in the blades of
environmentally clean and renewable source of energy, the variable a Francis turbine runner at different operating points were investi-
demand on the energy market, as well as the limited energy storage gated by Saeed [3] by using the finite element method with inlet
capabilities, requires a high flexibility in operating hydraulic tur- boundary conditions imposed from analytical computation. More-
bines. As a result, hydraulic turbines are frequently operated with over, in this case the three-dimensional flow generated by guide
a large number of start–stop cycles and over an extended range of vanes is not taken into account. A good agreement between the fi-
regimes quite far from the best efficiency point (BEP), often at part nite element analysis and the strain gauge experimental measure-
load conditions (PL) with high pressure fluctuations generated by ments was found by Bjørndal et al. [4] for the flow induced static
vortex rope in the draft tube. Therefore, strong vibrations are in- stresses in a medium head Francis runner. Based on a sequential
duced which can produce fatigue failure on the mechanical compo- coupled fluid–structure interaction, a similar approach was used
nents [1], especially on the turbine runner blades [2]. by Sadowski and Golewski [5] in the multidisciplinary analysis of
The loads acting on the Francis turbine runner can be steady or the thermal heat transfer of the turbine blades in combustion
unsteady. The first type includes the fluid pressure and the centrif- engines.
ugal force. The second type is composed of the high frequency The purpose of this paper is the analysis of stress field induced
pressure fluctuations due to stator–rotor interaction as well as vor- in the runner blades of a medium specific speed Francis turbine by
tex rope phenomenon. Under these loading conditions, the devel- the steady loading at seven operating regimes at constant head.
opment of fatigue cracks is the major concern with regards to The following analysis is based on the one-way coupled simulation
the structural integrity of the runners and its represent a threat approach. Section 2 consists of a computational fluid dynamics
to safe operation of the hydraulic turbines. (CFD) simulation which provides the fluid pressure distribution
on the blade. It is followed by the structural finite element analysis
(FEA) in Section 3. The conclusions are drawn in last section.
⇑ Corresponding author. Address: Politehnica University of Timisoara, Depart-
ment of Mechanics and Strength of Materials, Bv. Mihai Viteazu 1, 300222
2. Fluid flow analysis
Timisoara, Romania. Tel.: +40 256 40 35 75, fax: +40 256 40 35 23.
E-mail addresses: blackradu@yahoo.com, radun@mec.upt.ro (R. Negru), seby@
acad-tim.tm.edu.ro (S. Muntean), msvina@mec.upt.ro (L. Marsavina), resiga@mh. The case corresponds to a medium specific speed Francis turbine
mec.upt.ro (R. Susan-Resiga), nicu_pasca21@yahoo.com (N. Pasca). with parameters presented in Table 1, where: Q [m3/s] – discharge,
0927-0256/$ - see front matter Ó 2012 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.commatsci.2012.05.073
254 R. Negru et al. / Computational Materials Science 64 (2012) 253–259
(1) velocity field is prescribed on the inflow section for both spiral casing outlet, and the mixed velocity profile on the
distributor and runner domains. The velocity profile pre- distributor outlet section is imposed on the runner inlet
scribed on the distributor inlet section corresponds to the section;
(2) turbulence quantities (turbulence intensity and turbulence
length scale) are prescribed on the inflow section for both
distributor and runner domains. Also, the turbulence quanti-
Table 1 ties are imposed on the runner inlet section for each itera-
Parameters of the Francis turbine with medium specific speed. tion. It is well known, the Reynolds number gives a
measure of the ratio of inertial forces to viscous one. The
Parameters Value Equations according to IEC [6]
Reynolds number associated to the actual Francis turbine
Characteristic speed 70 nQ0.5H0.75
operating at BEP is 2.35 107. Consequently, the inviscid
Discharge coefficient 0.28 Q ðpxR32e Þ1
computation can be performed in order to evaluate with
Energy coefficient 1.264 2EðxR2e Þ2
enough accuracy the blade loading for operating conditions
Dimensionless specific speed 0.444 u0.5w0.75
near to BEP [8].
R. Negru et al. / Computational Materials Science 64 (2012) 253–259 255
Fig. 5. The pressure distribution on the st1, st2 and st3 sections, at BEP.
Fig. 3. The numerical results against experimental data of the dimensionless shaft
power vs. dimensionless discharge at constant head.
Fig. 6. The pressure distribution on the st1, st2 and st3 sections, at PL3.
(b) compute the distributor flow, keeping the inlet velocity con- can see a good agreement between the shaft power values com-
stant and using the outlet pressure distribution obtained at puted based on the fluid flow simulation (with line) and experi-
the runner inlet; mental data measured in situ (with points). As a result, it is
(c) he stopping criterion considered in this procedure was that performed a global validation of the blade loading further used
the pressure distribution on the distributor–runner interface in Section 3 for structural analysis.
remains practically unchanged from one iteration to The numerical results obtained in the fluid flow analysis are
another. presented below for two operating points, BEP and PL3, along to
three streamlines (st1 near the crown, st2 middle and st3 near to
In order to compute the velocity and pressure fields a segre- the band, Fig. 4).
gated solver was used within the FLUENT code, [9]. The pressure coefficient cp distribution vs. normalized blade
Moreover, the numerical results obtained with described meth- length s is plotted in Fig. 5 at BEP. The pressure coefficient is com-
odology were validated against experimental data on the Francis puted as follows:
turbine model, [7].
The turbine power shaft is computed based on fluid flow simu- cp ¼ ðp pref Þ=ðqEÞ ½ ð5Þ
lation using next equation:
P ¼ T x ½W ð3Þ where p [Pa] is the static pressure, pref [Pa] the reference static pres-
sure on the cone wall, q [kg/m3] the water density and E [J/kg] the
where x = 39.27 [rad/s] is the angular speed and T is the runner tor- specific energy. One can observe the minimum pressure was ob-
que computed as follows: tained on the suction side (SS) near to the band. It is well known
Z that this region is associated with cavitational risk. The pressure
T¼N ~r s n dS ~
~ iz ½Nm ð4Þ distribution near to the crown (solid line) presents an unwanted
S
shape, see Fig. 5, due to the straight leading edge selected in the
where N = 14 is the runner blade number, S represents the surface old design of the Francis runner. The flow impact is moved on the
comprising the rotating parts of the runner, s is the stress tensor pressure side (PS) when the Francis runner operates at part load
(including the pressure and the viscosity stresses), ~
n is the unit vec- conditions. As a result, the minimum pressure coefficient cp arises
r is the position vector and ~
tor normal to the surface, ~ iz is the unit on suction side near the leading edge (LE) at the junction with the
vector along to the axis of rotation. band. This statement is promoted by pressure coefficient cp distri-
The power of the turbine shaft computed for investigated oper- bution along to the streamline near to the band at part load operat-
ating points are plotted in Fig. 3 against experimental data. One ing point PL3, see Fig. 6.
Fig. 7. The von Mises equivalent stress distribution on pressure (left) and suction sides (right) at BEP.
R. Negru et al. / Computational Materials Science 64 (2012) 253–259 257
3. Structural analysis Poisson’s coefficient equal to 7850 [kg/m3], 2.05 [MPa] and respec-
tively 0.3.
The structural analysis response is based on the one-way cou- After the generation of the finite element mesh, the following
pled simulation approach, in which the fluid pressure obtained boundary conditions were applied on the model: (i) loads due to
from CFD is applied as a load condition. Due to high rigidity of the centrifugal force induced by rotation at 375 [rpm], to the fluid
the Francis turbine runner, the influence of the runner blade defor- pressure distribution imported from fluid flow analysis and to the
mation is not taken into account in the fluid flow analysis. The runner blade own weight (the fluid volume weight was neglected);
structural numerical analyses were performed using Ansys Work- (ii) on the two ends of the blade, representing the transition areas
bench software [10]. to crown and to band, zero displacements were assigned (fixed
Considering the periodical symmetry of the runner, in the struc- supports).
tural stress analysis the geometrical model was reduced to one sin- The three-dimensional stress analysis of the turbine runner
gle runner blade of which the crown and band were removed. This blade was performed at the operating points presented in Table 2.
simplified model neglects the geometrical details such as the Characteristic stress fields are presented at BEP in terms of von
rounded junction between blade and crown, as well as blade and Mises equivalent stress Seqv in Fig. 7, and maximum principal
band, respectively. Therefore, this model will underestimate the le- stress S1 in Fig. 8, for pressure and suction sides of the blade.
vel of maximum stresses, which controls the fatigue mechanism. The highest stress values occur at the junction between blade
This deficiency can be compensated by applying to the maximum and crown, near the trailing edge, on the pressure side of the blade.
stresses obtained a stress concentration factor which correspond In this case the maximum principal stress S1 is 38.5 [MPa], and the
to the rounded junction mentioned above. von Mises equivalent stress Seqv is 44.18 [MPa]. Toward band,
To ensure the accurate transfer of the fluid pressure field ob- along the trailing edge, these stresses decrease gradually to 3.17
tained in CFD the finite element solid mesh of the runner blade [MPa] for S1, and to 7.93 [MPa] for Seqv. For the leading edge, the
and the finite volume mesh of the fluid domain were generated to- highest stress values occur at the junction between blade and
gether. Using an eight nodes homogeneous structural solid ele- band, on the pressure side. The stress distributions are plotted
ment SOLID 185 [10] with three degrees of freedom at each vs. normalized blade surface length s in Fig. 9. In this paper the
node, a solid mesh containing 61,140 elements and 38,186 nodes stress distributions are plotted along to the junctions between
was obtained. the blade-crown (denoted j1 in Fig. 4) and blade-band (marked
The material considered in the structural linear-elastic analysis with j2 in Fig. 4), respectively.
was a martensitic–ferritic–austenitic stainless steel [2], with the Similar stress distributions were found for the two over load
following mechanical properties: density, Young’s modulus and operating points (OL1 and OL2), but with higher values for stress
Fig. 8. The maximum principal stress distribution on pressure (left) and suction sides (right) at BEP.
258 R. Negru et al. / Computational Materials Science 64 (2012) 253–259
Fig. 9. The stress distribution along to the junction j1 (up) and to the junction j2 Fig. 10. The stress distribution along to the junction j1 (up) and to the junction j2
(down), at BEP. (down), at PL4.
level peaks. For the part load operating points investigated compa-
rable levels of stress peaks occur on both trailing and leading edge,
as can be seen in Fig. 10 at PL4.
The values of the maximum principal stress S1 and equivalent
stress Seqv vs. dimensionless discharge Q/QBEP, on junction j1 at
trailing edge, are plotted in Fig. 11. As can be seen, the static stress
values change nonlinearly with the dimensionless discharge, in
terms of the operating conditions. One can observe a quasi-
constant stress value around to the BEP (from PL2 to OL1) due to
the small changes of the fluid flow stream. The maximum value
of static stress is obtained at maximum discharge, while the min-
imum values are associated to the part load conditions (under
PL2). However, the structural analysis performed in this investiga-
tion takes into account only the steady loading. The further work
have to take into account the unsteady phenomena (e.g. vortex
rope and interblade vertices) associated with the operation at part
load conditions.
4. Conclusions
and runner; (2) the maximum static stresses occur at the transition References
between the blade and the crown, near the trailing edge, at BEP
and over load (OL1, OL2) operating conditions. Contrary, at part [1] F. Casanova, Engineering Failure Analysis 16 (2009) 2202–2208.
[2] D. Frunzaverde, S. Muntean, G. Marginean, V. Campian, L. Marsavina, R. Terzi,
load operating conditions (PL4, PL3, PL2 and PL1) comparable lev- V. Serban, Failure analysis of a Francis turbine runner, in: IOP Conf. Series,
els of stress peaks occur on both trailing and leading edge; (3) the Earth and Environment Science, vol. 12, 2010, pp. 012115 1–10.
static stress values change nonlinearly with the dimensionless dis- [3] R.A. Saeed, A.N. Galybin, V. Popov, Advances in Engineering Software 41 (2010)
1245–1255.
charge, in terms of the operating conditions. [4] H. Bjørndal, T. Moltubakk, H. Aunemo, Flow induced stresses in a medium head
Francis runner – Strain gauge measurements in an operating plant and
Acknowledgements comparison with Finite Element Analysis, in: 10th International Meeting of the
IAHR Work Group on the Behaviour of Hydraulic Machinery Under Steady
Oscillatory Conditions, Trondheim, Norway, 2001.
This paper was supported by the project ‘‘Development and sup- [5] T. Sadowski, P. Golewski, Computational Materials Science 50 (2011) 1326–
port for multidisciplinary postdoctoral programmes in major tech- 1335.
nical areas of national strategy for Research-Development- [6] International Electrotechnique Commission, IEC 60913 Standard Hydraulic
Turbines, Storage Pumps and Pump-Turbines-Model Acceptance Tests,
Innovation’’ 4D-POSTDOC, Contract No. POSDRU/89/1.5/S/52603, International Electrotechnical Commission, Geneva, 1999.
project co-funded by the European Social Fund through Sectorial [7] S. Muntean, R. Resiga, I. Anton, Mixing interface algorithm for 3D turbulent
Operational Programme Human Resources Development 2007– flow of the GAMM Francis turbine, in: J. Vad, T. Lajos, R. Schilling (Eds.),
Modelling Fluid Flow – The State of the Art, Springer-Verlag, 2004, pp. 359–
2013. N. Pasca was partially supported by the strategic grant 372.
POSDRU ID77265 (2010), co-financed by the European Social Fund [8] G. Sottas, I.L. Ryhming (Eds.), 3D-computation of incompressible internal
– Investing in People, within the Sectorial Operational Programme flows, in: Proceedings of GAMM Workshop, NNMF 39, Vieweg Verlag,
Braunschweig, 1993.
Human Resources Development 2007–2013. Dr. S. Muntean and [9] Fluent Inc., FLUENT 6.3, User’s Guide, Lebanon, New Hampshire, 2006.
Prof. R. Susan-Resiga were supported by the Romanian Academy [10] ANSYS Inc., ANSYS Workbench User’s Guide 12.1, Canonsburg, Pennsylvania,
program. 2009.