On the influence of the Kaplan turbine runner blade

thickness on its stress parameters

Zoran Markov Predrag Popovski

Faculty of Mechanical Engineering-Skopje Faculty of Mechanical Engineering-Skopje
PO Box 464 PO Box 464
Skopje Skopje
Macedonia Macedonia

Andrej Lipej Vesko Djelic

Turboinstitute Turboinstitute
Rovsnikova 7 Rovsnikova 7
Ljubljana Ljubljana
Slovenia Slovenia

Introduction
The current trend in the design and refurbishment procedures of the hydraulic turbines focuses on both the
hydraulic, as well as the stress analysis of the operation parameters. Of course, very useful data during the process of
validation of the numerical models is the previous operational data during the lifetime of the unit. In many cases,
high stressed areas are damaged. Lack of the measurements makes the solution of the problems and verification of
the numerical results very difficult. This work represents an integrated approach in solving hydraulic and structural
problems in design (re-design) stage or optimization of an axial hydro turbine. Computational Fluid Dynamics
(CFD) approach is implemented in solving the flow through an axial flow turbine, taking into account all the
necessary factors influencing the real unsteady flow. The results from the CFD calculations are used as an input for
the performed Finite Element Analysis (FEA) modelling and structural analysis of the blade. Possibilities for
modifications of the blade thickness are analyzed influencing the hydraulic and stress parameters of the turbine. The
runner efficiency and strength parameters' changes due to thickness modifications are analyzed.

1. Background
The prediction of the hydraulic, as well as the stru tural performances of a hydraulic turbine is one of the most
important steps in the development procedure. However, for re-design of an old machine, a fast and reliable flow
analyses tool is necessary for allocation of the geometry that can be improved. This work describes a computational
analysis approach, which takes into account the hydraulic loads on the runner during operation and the influence on
the stress performances of the runner itself. A Kaplan turbine was calculated, as an example of the applicability of
this methodology.

The simulation of three-dimensional turbulent flow through the runner, as a part of the complete turbine calculation
is presented. The operational regimes were calculated based on the Computational Fluid Dynamics - CFD
Methodology (Navier-Stokes equations and the k-ω SST turbulent model) and the hydraulic loads were applied to
the runner using FEA (Finite Element Analysis) software.

The mesh discretization, boundary conditions and calculated results are presented and they shall be a useful example
for development, optimisation, refurbishment or rehabilitation projects, since the runner behaviour can be a major
criterion for increasing the performances of the turbines. The complete flow field consists of spiral case, a channel
between stator vanes, a channel between guide vanes, a channel between rotor vanes and the draft tube cone and
elbow. The calculations are performed in one step for all geometry.

The numerical analyses were performed using the Ansys CFX and Design Space software packages.
The work presented in this paper coincides with the current trend of improvement in the CFD field, as a faster and
less expensive way of testing and development of hydraulic machinery. The results are promising, but a lot of
experiments are still necessary for the completion of the validation process.

2. CFD Simulation
The first step in the flow calculation is the geometry modelling. In this case, the calculated flow area was meshed by
means of volume element method. The number of grids for total flow passage is over two million elements.

Figure 1 shows the geometry model and the discretized domain of the calculated turbine.

Figure 1. Numerical model of the calculated Kaplan turbine.

After the completion of the geometry modelling of the turbine, the next step is to discretize the flow domain
(produce the mesh). Figure 1 shows the complete calculation grid (mesh) of the flow field. The optimal number of
calculation elements is hard to set, having in mind the time needed for the calculation, as well as the accuracy that is
required.

Next step in the calculation procedure is creating the boundary conditions. The main boundary conditions are the
Inlet into the spiral case and the Outlet from the draft tube. Flow rate (discharge) is set at the inlet and the average
pressure over the outlet. Additional boundary conditions come from the positions of the guide vanes and the runner
blades, which correspond to each operation point, including the rotational speed of the runner.

Interaction rotating was simulated by means of Frozen rotor model as a simplification of the unsteady model i.e. the
steady state of rotating and stationary regions is predicted with rotor solved in rotating frame and the stator is solved
in stationary frame in fixed relative positions. Transient model, as a full unsteady simulation, was employed only for
regimes, where a stress analysis was conducted. The calculation was performed by means of Navier-Stokes
equations and the k-ω SST turbulent model. The k-ω two-equation models have been built into most general-
purpose computational fluid dynamic software packages.

The number of calculation elements varies for the different flow regimes, starting with 1.2 million as the coarsest
grid, but the most significant points have over 2.3 million. It is clear that a model of this size had to be calculated on
a multi-processor cluster, property of Turboinstitute - Ljubljana. The full steady simulation usually lasts for approx.
24 hours if 16 processors are engaged.

The CFD calculation always predicts lower efficiency of the turbine model. The factors that influence such results
can be mesh quality, completely accurate geometry model of the blade surface, tip clearance effects that also require
finest possible grid etc. The significance of the mesh quality has been investigated, and the results showed that the
increase in the number of grid elements in the draft tube alone from 140,000 up to 670,000 improved the prediction
accuracy for almost 2 percent. This trend didn't continue with the further increase in this part. The same grid
refinement could be implemented for all parts of the turbine, but the calculation cost would increase significantly, as
well as the time needed for the final results.
In the next step, the data from the turbine model measurements have to be scaled to calculate the prototype
characteristics, so results from the CFD calculations could be verified. IEC 60995 regulation is implemented for the
scaling process. In this manner, the efficiency scale-up correction Δη=3.1 percent was calculated.

The acqired values show that the efficiency correction Δη from the simulations doesn't even reach 2 percent,
compared to the calculated 3.1 percent. The causes for this occurance should be located in the grid density (same
number od elements was used for both the model and the prototype-approx. 2.3 million), but also zonal modelling as
an flow simulation approach becomes interesting, because the volume of boundary layer flow (where viscosity
effects are significant) is far greater in the homologous turbine model, so calculation grid could be finer in these
regions only.

The results of the CFD calculation are important because they give the pressure distribution on the pressure and
suction side of the turbine runner blade. This pressure contours are presented in figure 2.

Figure 2. Pressure distribution on a) pressure side; b) suction side.

The next step in the co-simulation process is to effectively transfer the calculated data regarding the pressure on the
runner blade to FEA software package. Ansys Workbench is used for this purpose. A procedure has been created for
data transfer using macro in CFX Post processing module. The values of the pressure loads are then applied to the
solid model prepared for Workbench.

3. FEA simulation
The first action that has to be taken is the geometry modelling of the complete blade assembly, for mounting on the
turbine shaft. Then, a discretization is again needed for the FEA simulation. A mesh of approximately 23,000 nodes
is used for this specific case (figure 3a). General verification of the entered geometry and the software capabilities
can be performed by calculating the natural frequencies of the turbine in air environment and comparing with the
relevant experimental data given in Albijanic [3]. A graphical presentation of mode 2 is given in figure 3b.

Figure 3. Turbine blade: a) Mesh, b) Presentation of mode f2.

Next step is to simulate water environment in the FEA simulation. There are two possibilities. The first one is to
simulate additional mass (pressure) onto the blade surface, thus creating lesser values of the natural frequencies.
This method showed some unreliability for the cases studied, but showed good overall trend. The other is an
analytical method developed by Blevins [3], which suggests the following formula for taking into account the water
environment influence:
f fluid 1
= 1/ 2
f vacuum & A#
$1 + !
% m"
where: A-added mass of the fluid (from the table given in Blevins [3] for different solids), m-mass of the analysed
body (blade). This model shows more significant error for modes higher than mode five, because the correction
coefficient is the same for all modes, and the experiments prove otherwise. However, the first four calculated modes
have very good compatibility with the measurements, and it is also clear that this is the area where most of the
resonance problems can occur.

After this preliminary verification of the model capabilities, the next step, as it is already mentioned, is to transfer
the hydraulic loads into the FEA model of the blade. Figure 4 shows the calculation results of the equivalent stresses
in Workbench.

Figure 4. Calculated equivalent stresses a) on the entire blade; b) in the root area.

The calculated results show good compatibility with the values of measured stresses [1] on the blade. The other
important fact is that the stress distribution over the blade surface also corresponds to the actual stress field.

Other very important aspect of the stress analysis is the calculated value of the strength coefficient nσ, which
represents the influence of the cyclic loads during long-term turbine oparation. It depends on the maximum stress,
blade thickness, number of cycles in operation etc. This value, according to Aronson [2] should be between 1.5 and
P
1.6. Figure 5 shows this coefficient at operation point at relative turbine output of P= =0.71, taking into
Pmax
account the calculated maximum stress value from the FEA simulation (110.16 MPa) and maximum measured value
of 78.64 MPa. There is a significant difference between the two curves resulting from the different maximal stresses
that are used.

This is most probably as a results of the geometry imperfections of the numerical model, especially in the blade root
area (figure 4b). Similar calculation, presented by Gajic et al. [4] has given a maximum stress value of 83 MPa.
However, the calculation puts the designer on the safe side in regard to the danger of possible fatigue damage to the
blade. In this case this curve (with max. stress of 110 MPa), the value of 1.5 for nσ is acheved in 27 years of
operation.
 Figure 5. Strength coefficient.

The lack of measured data for the total deformation values prevents the further verification and calibration of the
numerical FEA model in this case and of course, more measured data are necessary to make this method a more
reliable tool in the design and refurbishment processes.

4. Blade Modifications
There are several possible approaches to the blade geometry modifications in order to obtain improved hydraulic
and stress performances. Those are: blade thickness, inlet angle, different profile (chord) and different number of
runner blades. Some of them, as is the inlet angle, must be carefully analysed because of the influence of the flow
regimes, i.e. already known inlet velocity angles. Modifications in the blade thickness will be analysed in this work.

The proposed change is approximately 10 percent of the blade thickness. So, two modifications will be analysed:
modification 1:90 percent of the blade thickness and modification 2:110 percent of the blade thickness.

The proposed modifications influence the stress parameters of the turbine even more significantly than the hydraulic
performances. The geometry change obviously leads to changes of the natural frequencies (figure 6). This graph
shows that the increased thickness of the blade leades to their increased values. This figure shows that the changes
are not linear, and additionally, even though the thickness modifications are approx. 10 percent in both cases, there
are different influences to the original values: increase of approx. 15 percent for the thicker blade and decrease of
almost 5 percent for the thiner modification.

Figure 6. Changes in the natural frequencies as a result of the geometry modifications.

Table 1 presents the comparison of the relative (to the original geometry) values for the stress and deformation
conditions of the three blade designs.
Geometry Relative Relative Relative Minimal
maximal normal stress maximal safety factor
equivalent at measuring deformation in regard to
stress point MT6 the equivalent
stress
Original runner 1 1 1 2.16
Modification 1 (90%) 1.146 1.145 1.398 1.8856
Modification 2 (110%) 0.809 0.846 0.792 2.67

Table 1. Comparative analysis of the relative stress perfomaces for the proposed geometry modifications of the runner blade

The final parameter that needs to be examined is the strength coefficient nσ, which according to the maximal values
of the equivalent stress is calculated and the results are presented in figure 7. The calculated values are low in
general (according the Aronson recommendations) and only modification 2 has a significant period of approx. 17
years in operation over the referenced 1.5. But, the analysis produced above from figure 5, must be once again taken
into account considering the values of the equivalent stress used in the calculation, which should be lower than the
calculated maximum.

Figure 7. Changes of the strength coeffcient using the modified blade geometries.
Finally, one of the more significant criteria that influence on the "quality" of the possible solution is the mass of the
runner. Table 2 shows how the blade and the runner mass differ with regard to the proposed geometry modifications.
It is clear that the changes in the runner mass means cheaper or more expensive product, but they also influence the
total loads on the turbine bearings and structure in general.

Geometry Change in the total weight of

the turbine runner
Modification 1 (90%) - 7,43 %

Modification 2 (110%) + 8,23 %

Table 2. Changes in the runner mass

Turbine produced with 7 percent less expensive stainless steel used in modification 1 is clearly the better solution
using this criterion. The other positive aspects, such are the static and dynamic loads decreases are additional gains
that are achieved.
5. Conclusions
The work presented in this paper coincides with the current trend in application of the CFD and FEA tools, as a
faster and less expensive way in the development of the hydraulic machinery. The results are promising, but a lot of
additional measured data are still necessary for the completion of the validation process.

The numerical analysis of the flow field in the turbine shows the velocity vectors, pressure distributions and the
streamline behaviour. Also, the over-all performances of the machine are calculated. Using these results,
conclusions regarding the design processes and operational performances can be achieved.

The differences between the calculated and measured values, both for the hydraulic, and for the structural
characteristics are not significant having in mind the main goal - verification process of the turbine performances.
Further improvements in the flow modelling are still necessary (geometry and boundary modelling, turbulence
modelling, secondary flow effects, complex draft tube flow, as well as fatigue analysis etc.)

Most important conclusions are given as follows:

• Numerical simulations of the Kaplan turbine model flow provide the runner and the total efficiencies, as
well as the losses in each particular part of the flow field. The acquired results are lower in general than the
measured values, from two percent for the operation regimes closest to the optimal point, up to 5 percent in
some cases.
• These discrepancies in the values are reducing with the increase of the number of calculation grid elements,
which leads to meshes with over 2.3 million elements, i.e. more costly and time consuming simulations.
Furthermore, the most accurate results, which can be used as a reliable assistance in the turbine design and
analysis process, are achieved using transient flow calculations demanding additional CPU resources.
• The trend of the CFD calculated efficiency curves is compatible with the measured data, which is very
useful in determining the "better" solution, when several geometries or operation points are analysed.
• The results of the runner geometry modifications show that "modification 1" (the thinner blade) has
marginally higher runner efficiency and also a marginally lower total efficiency, but in general, these
figures are within the calculation accuracy. So, it means that using this modification won't significantly
influence on the efficiency performances, but will, on the other hand, improve the cavitation safety and,
maybe most important, lead to material saving of more than 7 percent which results in lighter design less
demanding on the whole structure. The downsides of this solution are the worsened stress and deformation
conditions (for approx. 14 percent), which are still allowable for the static stress and the worsened fatigue
strength factors.
• The second analysed modification (the thicker blade) offer significant improvement of the stress condition
of the runner, but showed worsening of the hydraulic performances such are the total efficiency and the
cavitation safety, as well as the price and the weight of the assembly.
• The implemented multi-criteria approach analysis takes into account all relevant parameters and shows that
the thinner blade has clear advantages from hydraulic and financial aspect and puts less weight on the
whole structure, but further analysis is needed for the fatigue influence having in mind the significant peak
stresses calculated in the root of the blade, which are difficult to verify, having in mind that the static stress
loads are within the allowed limits.
• The overall research shows that the analysis and the selection of the optimal form of the turbine runner
blade is a very complex problem. In that regard, a lot of criteria influence the decision making process.
Nevertheless, the justification of this approach lies in the fact that the existing turbine designs from twenty
years or so ago have a "reserve" regarding the stress parameters, because it wasn't a dominant factor in the
design process, as it is nowadays.

References
1. Albijanic R. et al., "Structural Dynamic Identification of Kaplan turbine blades", Proceedings, International Conference
Classics and Fashion in Fluid Machinery, pp. 265-276, Belgrade, 2002.
2. Aronson A., Babacenko V.E., Zaicev G.Z., "Issues in Hydro turbine Runner Strength Calculations Regarding Cyclic
Loads" (in Russian), pp 127-136, Izdatelstvo Masinostroenie, Leningrad, (Gidroturbinostroenie, editor Segolev G.S.), 1969.
3. Blevins R. D., "Formulas for Natural Frequency and Mode Shape", Krieger Publishing Company, Florida, 1995.
4. Gajic A. et al., "Stresses of Kaplan Turbine Runner Blade During Transients", Proceedings, XXI IAHR Symposium on
Hydraulic Machinery Systems, Lausanne, 2002.
The Authors
Z. Markov graduated at the Faculty of Mechanical Engineering of the University in Skopje in 1998, got his master degree in
2001 and his Ph.D. degree at the same University in 2007. He worked as a research assistant at the Faculty of Mechanical
Engineering - Skopje from 1999-2007 and afterwards as an assistant professor at same Faculty. He involved in teaching of CFD
methods in Hydraulic Turbomachines; Design of Turbopumps and Hydropower plants, and is actively involved as a member of
the team from the Faculty of Mechanical Engineering as a an expert in several projects for large and small hydropower plants.

P. Popovski graduated at the Faculty of Mechanical Engineering of the University in Belgrade in 1971, got his master degree in
1974 and his Ph.D. degree at the same University in 1976. He also completed post-doctoral research training at the Pennsylvania
State University, Department of Aerospace Engineering in 1983 as a Fulbright scholar. He worked as a research Assistant at the
Faculty of Mechanical Engineering - Belgrade from 1971-1974 and afterwards as a research assistant and assistant professor at
the Faculty of Mechanical Engineering, University “St Cyril and Methodius” -Skopje until 1982. In 1984, he was a visiting
assistant professor at the Department of Aerospace Engineering at the Pennsylvania State University, USA. From 1985, he is
employed as a professor at the Faculty of Mechanical Engineering, University “St. Cyril and Methodius” - Skopje. He teaches
Fundamentals of Energy; Hydraulic Turbomachines; Design of Turbopumps and Hydropower plants, and is a head of the team of
the Faculty of Mechanical Engineering involved in several projects for large and small hydropower plants for design and
consultant services.

A. Lipej graduated at the Faculty of Mathematics of the University in Ljubljana in 1984, got his master degree in 1990 at the
same faculty and his Ph.D. degree at the Faculty of Mechanical Engineering of the University of Ljubljana in 1999. He worked as
a researcher in the CFD Division at the Turboinstitute - Ljubljana from 1984-1994 and afterwards as a head of the CFD division.
He is now manager of basic research and education at the Turboinstitute. He was employed as an assistant at the Faculty of
Mechanical Engineering, University of Ljubljana from 1994 to 2005 and since is employed as an assistant professor at the same
Faculty. His experience is mainly in the field of numerical methods, research and development of water turbines, especially
Kaplan turbines and using software for flow analysis and development of Francis, Kaplan and reversible pump-turbines.

V. Djelic graduated at the Faculty of Mechanic Engineering of the University in Belgrade in 1984 and got his master degree in
1991 at the Faculty of Mechanical Engineering of the University of Ljubljana. He worked as a researcher in the Turbine Division
at the Turboinstitute - Ljubljana from 1988-1994 and afterwards as a head of the Turbine development and tests division until
1998. He is now manager of turbine research and development at the Turboinstitute. His experience is mainly in the field of
research and development of water turbines. He led more than 30 model development and acceptance tests in accordance with
IEC 60193 Code and led more than 35 site measurements of various types of turbines and is involved in development of model
and site measurements methods and procedures.