Failure Analysis of Runner
Failure Analysis of Runner
Failure Analysis of Runner
net/publication/272101789
CITATIONS READS
11 1,074
5 authors, including:
S.A. Rodríguez
Universidad del Valle (Colombia)
48 PUBLICATIONS 300 CITATIONS
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
Integral Design Of Strategies For The Control Of Erosive Wear In Turbines View project
All content following this page was uploaded by S.A. Rodríguez on 31 October 2017.
a r t i c l e i n f o a b s t r a c t
Article history: This paper presents the analysis of a 16-bucket Pelton impeller from a hydroelectric plant
Received 1 May 2014 in Colombia, a plant with two turbo generators with a nominal capacity of 2.33 MW each.
Received in revised form 23 August 2014 Multiple cracks were detected one year subsequent to geometry reconstruction welding
Accepted 28 August 2014
repair. Metallographic analyses were performed on the impeller zones near the cracks,
Available online 16 September 2014
including an analysis of the fracture surface, a computational simulation of the fluid
dynamics using finite volume software that permitted the establishment of the impeller’s
Keywords:
load state and a simulation via finite element analysis to determine the state of the impel-
Pelton impeller
Fatigue
ler’s nominal stress. It was concluded that the material presented multiple defects such as
Cavitation non-metallic inclusions and microcracks, and trapped slag was found in the filler material.
Finite element analysis The computational fluid dynamics analysis permitted the identification of low-pressure
Computational fluid dynamics zones caused by an inadequate geometry for the bucket profile where cavitation is present.
The finite element analysis permitted the identification of critical points on the bucket
zone neck, coinciding with the crack found. This zone is under tensile stress due to the
effects of centrifugal force and compression when the bucket comes in contact with the
jet, causing fatigue.
Ó 2015 Published by Elsevier Ltd.
1. Introduction
The Pelton turbine is one of the most efficient types of hydraulic turbines; it is a cross-sectional flow turbo motor machine
with partial admission and action. It consists of a wheel (impeller or rotor) outfitted with buckets in its periphery, which are
specially designed to exploit large low-flow hydraulic jumps. Hydroelectric plants outfitted with these types of turbines gen-
erally have a large pipe called a pressure gallery to transport the fluid from great heights, often more than 200 m. At the end
of the pressure gallery, water is provided to the turbine through one or several needle valves, called injectors, which are noz-
zle shaped to increase the speed of the flow impact upon the buckets [1]. The nozzle or injector launches the water jet
directly against a series of bucket-shaped paddles mounted around the outer border of the runner. The water provides a driv-
ing force on the buckets, exchanging kinetic energy with the wheel by virtue of its change in the amount of movement.
One of the principal failures and perhaps the most catastrophic failure that could occur in a Pelton turbine is the sepa-
ration of the bucket from its base at the root, where the bending stress product of the jet force is at its maximum [2]. These
types of failures may occur during operation because of cracks not detected during maintenance that could cause the dis-
ablement of the impeller, destruction of the housing, and – in the worst case scenario – injury of operators who are near
the machine. There are no reports of failure analyses of Pelton turbine impellers in operation in the literature. The failures
reported in literature are attributed to fracture because of manufacturing and material defects after heat treatment [3].
⇑ Corresponding author.
E-mail address: [email protected] (S.A. Rodríguez).
http://dx.doi.org/10.1016/j.engfailanal.2014.08.012
1350-6307/Ó 2015 Published by Elsevier Ltd.
298 J.C. Chávez et al. / Engineering Failure Analysis 48 (2015) 297–307
The forms of wear usually found in these impellers are erosion [4,5] and cavitation [6,7]. Wear due to erosion is caused by
the impact of fluid or solid particles. The wear rate caused by erosion is mainly a function of the shape, size, hardness, par-
ticle concentration, impact speed, and incidence angle [4,8]. Cavitation is wear attributed to vapour bubbles that interact
with the bucket surfaces and are produced in low-pressure zones because of the detachment of the limit layer by flow sep-
aration where liquid vapour pressure is reached; when the vapour bubbles collapse, they generate shock waves that progres-
sively degrade the material. The low-pressure zones are generally created by changes in the hydraulic profile product of
impeller defective designs [6,7]. The principal consequences of wear are the loss of efficiency and the generation of noise
and vibrations; hence, impellers are recovered by welding to generate the desired geometry [9].
To conduct a failure analysis for an impeller in service, it is necessary to establish the load state to which it is subjected;
the use of computational fluid dynamics (CFD) has permitted the design of higher-efficiency turbines [10,11] and established
the profile of operating pressures that can later be used in models of finite elements for mechanical design [12,13]. This study
analysed materials and fracture surfaces and used computational tools to establish the viability of recovery for a Pelton
impeller with cracks and with evidence of cavitation and erosion.
The average power generated by the Pelton turbine impeller during the time it was in service (after recovery by welding)
was obtained from the daily record of operation submitted by the plant. From analysing these data, it was concluded that the
impeller was in service for 359 days. The values of power generated were recorded each hour during that period. In total,
8616 values were registered; to determine the average power, zero-generation values were discarded. The results obtained
are summarised in Table 1.
By knowing the average value of the power generated, the water jet theoretical output speed from the injector was cal-
culated. In addition to the values from Table 1, it was necessary to use results from the efficiency tests conducted in the
plant. The results and parameters of these tests are summarised in Tables 2 and 3, respectively.
The net height (Hn) is defined as the load at the disposition of the turbine. In other words, it is the maximum load value
that could be transferred to the turbine if no load losses existed within the turbine [14].
X
2
Hn ¼ Z HrExt;Turbine ð1Þ
1
P
To determine load losses external to the turbine ð 21 HrExt;Turbine Þ, we must know the pipe’s length, internal diameter, and
material or have knowledge regarding the turbine’s total efficiency.
Using the following ratios amongst volume, net height, and hydraulic power available for the turbine (Pn) and efficiency,
Pn ¼ cQHn ð2Þ
Pa
gT ¼ ð3Þ
Pn
where c is the specific weight of the fluid and knowing the value of gT (Table 2), the values of Pn and Hn were determined
(Table 4) using the average Pa value (Table 1).
The fluid output speed from the injector (C1) is related to Hn in the following manner:
pffiffiffiffiffiffiffiffiffiffiffi
C1 ¼ Cv 2gHn ð4Þ
where g is the gravity acceleration and Cv is the injector or nozzle design coefficient; this value is between 0.96 and 0.98 for
good designs [15], and our analysis used an intermediate value between these values (0.97).
To know the properties of the impeller’s base material and the welding applied for its recovery, samples were extracted
from the tip of the bucket and from the neck close to the main crack. For the initial polishing, sandpaper between 220 and
600 grades was used and to bring the samples to their final, mirror-type polish, alumina particles (1 lm and 0.3 lm) were
used on a rotating disk. The samples from the base metal and thermally affected zone were attacked by immersion with Nital
at 2% for 4 s. The welding sample was attacked by immersion with aqua regia (30-ml distilled H2O, 17.5-ml HNO3, 52.2-ml
HCl) at 70% for 2 min. To observe the microstructures from each sample, an Olympus optical microscope was used.
Table 1
Values of power generated by the Pelton turbine impeller during 359 days of service.
Average generation Pa Maximum generation Minimum generation Generation mean Standard deviation of the generation
(kW) (kW) (kW) (kW) (kW)
2073.51 2800.00 500.00 2100.00 558.07
J.C. Chávez et al. / Engineering Failure Analysis 48 (2015) 297–307 299
Table 2
Results of the efficiency test.
Table 3
Necessary parameters to perform the efficiency test.
Table 4
Operating conditions for analysis.
Pn (kW) 2657.32
Hn (m) 160.73
C1 (m/s) 54.46
Two different simulations were performed to calculate the load state and the actual stress–strain state at each bucket in
the Pelton wheel. First, a two-phase flow CFD simulation was performed to obtain the respective forces acting on each Pelton
bucket. Second, a structural simulation was developed to calculate the stress–strain state present on each bucket.
The CFD simulation consisted of a steady state evaluation for a water jet coming out from the nozzle, impinging on a pair
of Pelton buckets at three different angular positions with respect to the nozzle. These locations are shown in Fig. 1. The rea-
son why two buckets were simulated is because only two of them are always in contact with the water jet at the same time,
as is shown in Fig. 1b. Table 5 shows the general data used to develop the CFD simulation.
The CFD simulation also allows us to identify the zones of low- and high-static pressure to determine possible causes of
cavitation. The second simulation was the structural one, and it took the force exerted by the water jet on the Pelton buckets
to generate an input on the respective bucket, which was also subjected to a centrifugal load. The result was a stress–strain
state on the entire bucket. The most critical case is shown in this paper.
Fig. 1. Locations of two Pelton buckets with respect to the nozzle; the Pelton wheel rolls counterclockwise. (a) The water jet is impinging the first bucket.
(b) The water jet is impinging both buckets. (c) The water jet is impinging the second bucket.
300 J.C. Chávez et al. / Engineering Failure Analysis 48 (2015) 297–307
Table 5
Data used to perform a CFD simulation of a water jet impinging two Pelton buckets.
Phases Air–water
Water inlet velocity (m/s) at the nozzle’s exit 54.47
Gauge pressure outlet (Pa) 0
Reference temperature (K) 298
Reference pressure – atmospheric pressure (Pa) 101,325
Surface tension between air and water (N/m) 0.078
Mesh (grid points) 1,392,039
Turbulence models SST, k–e
3. Results
Several forms of damage were identified on the buckets on the impeller surface: erosion wear, cavitation, corrosion, and
neck cracks. Fig. 2a shows the interior zone of one of the impeller buckets, indicating the scratches caused by abrasive par-
ticles, which upon impacting the surface remove the material. Fig. 2b shows the frontal zone of one of the buckets, which
presents cavitation, and the amount of material removed as a consequence of this mechanism can also be observed. The
impeller has cracks on the impeller neck in a zone close to the zone where recovery welding was applied (Fig. 2c).
Fig. 3a presents the base metal microstructure, constituted by ferrite grains (white) and fine pearlite (dark). This micro-
structure is typical of steels with low carbon content (AISI 1026 steel). The microphotography in Fig. 3b shows the detail of
the ferrite grains, pearlite and some manganese sulphide inclusions.
The microphotography in Fig. 4a shows the inclusions present along the base material. The most common inclusions are
from manganese sulphide and oxides. A high concentration of these inclusions favours the appearance of micro-cracks; the
figure shows the micro-crack formed. The micrograph in Fig. 4b displays the molten zone, constituted by various deposits of
stainless steel with the presence of slag trapped during the welding process. The microstructure is made up preferentially by
austenite and ferrite. The presence of slag trapped within the welding passes indicates that careful cleaning of the slag was
not performed when applying shielded metal arc welding. These defects contribute to the nucleation of cracks.
The left part of Fig. 5a shows the molten zone of a low-carbon steel deposit constituted preferentially by ferrite. It is evi-
dent that low-carbon steel deposits were applied prior to the stainless steel deposits. The right part shows an acicular grain
Fig. 2. (a) Erosion wear from recovery welding present in the impeller; (b) zone of the bucket where cavitation is present; (c) crack on the impeller neck.
J.C. Chávez et al. / Engineering Failure Analysis 48 (2015) 297–307 301
Fig. 3. Microstructure of the base metal (a) 100 and (b) 500.
Fig. 4. (a) Microphotography of the base material without attack, 500; (b) microphotography of welding showing trapped slag, 100.
Fig. 5. (a) Microphotograph of a zone affected by heat with the presence of defects where cracks nucleate. (b) Micrograph of a thermally affected zone,
100 (a) 200.
zone with cracks originating from the defect. The microstructure in Fig. 5b exhibits a section of the zone affected by heat or a
zone partially affected, constituted by spheroid-shaped ferrite and pearlite grains.
The neck fracture surface was analysed via scanning electron microscopy. The topography was analysed by using second-
ary electrons; to differentiate the composition of the materials, backscattered electrons were used. Fig. 6a shows the fracture
surface close to the impeller surface; the fracture presents a flat surface typical of fatigue failure, and beach marks cannot be
identified. In Fig. 6b, unclear striations can be observed due to fatigue crack growth. This patron of unclear striations appears
in many fatigue failure surfaces, all metals do not form distinct striations under cyclic loads, and those that do may not form
striations under all stress ranges and atmospheres [16].
The origin of the fracture cannot be determined, but it can be nucleated in a defect similar to those found because of effect
of the residual stress product on recovery through welding or because of the effect of operation stress. Fig. 8c shows second-
302 J.C. Chávez et al. / Engineering Failure Analysis 48 (2015) 297–307
Fig. 6. (a) and (b) Fracture surface using secondary electrons and (c) using backscattered electrons.
ary cracks, one of which crosses through a manganese sulphide inclusion, confirmed by EDS, of approximately 30 lm, which
provides evidence that they favoured nucleation and the propagation of cracks.
Theoretically, on a Pelton turbine, all the energy available in the flow becomes kinetic energy at atmospheric pressure
through a nozzle before the fluid comes in contact with the moving blades.
Fig. 7 shows that at the bucket entry, absolute C1 and tangential u speeds have the same direction and sense; thereby, we
can write:
x1 ¼ C 1 u ð5Þ
where x1 is the relative speed of entry between the jet speed C1 and the impeller tangential speed u. At the control volume
(CV) output, the direction of the relative speed x2 is defined by the output angle h, and we have:
x2 ¼ K m x1 ð6Þ
where Km is the denominated bucket coefficient that depends on the thickness of the water layer, surface roughness of the
bucket, and type of material. Its value varies between 0.88 and 0.92 [17]; the analysis uses a more conservative value of 0.92.
J.C. Chávez et al. / Engineering Failure Analysis 48 (2015) 297–307 303
Fig. 8. Determination of the angle for water flow output using the bucket geometry obtained.
According to the principle of conservation for the amount of movement, the jet force on the x component is given by [18]:
Q Q
Fx ¼ q x2x þ x2x Q x1x ¼ qQ ðx2 cos h þ x1 Þ ¼ qQ x1 ðK m cos h þ 1Þ ð7Þ
2 2
where q is the water density and is equal to 998 kg/m3 at 20 °C. Finally, an expression is obtained for the force of the jet on
the bucket:
pffiffiffiffiffiffiffiffiffiffiffi
F jet ¼ qQ ðC v 2gHn uÞðK m cos h þ 1Þ ð8Þ
Establishing the most critical position as observed in Fig. 1b in which the interaction of the jet on the bucket is maxi-
mised, the h angle is defined, which determines the direction of the water flow output after interacting with the bucket,
through a normal plane to it from the central point where the jet impacts (Fig. 8).
Hence, discarding the flow losses due to jet dispersion, analytically, the force exerted by the jet upon the bucket using Eq.
(8) is equal to 167.81 N for the impeller in the halted position.
Based on the computer fluid dynamics analysis, we obtained a field of speeds and pressures on the fluid and on the
bucket. Fig. 9 illustrates the speed contour of the water jet over a medium cut plane through the buckets.
This contour indicates the deviation effect the buckets have over the water flow, provoking recirculation zones because of
the bucket’s hydrodynamic profile, where low-pressure areas occur on the reverse side of the bucket (Fig. 10a). Because of
the turbulent nature of the jet water volume from the injector, a small interaction is observed on the speed contour between
the jet and the first bucket; this interaction is not considered in the analysis and represents a negative loss of the total force
of the jet over the second bucket.
The low-pressure zone in the first bucket (Fig. 10a) presents a high probability of cavitation. Given that the water has a
saturation pressure of approximately 3 kPa (absolute pressure) at a mean temperature of 25 °C, it may be deduced that in
Fig. 10. (a) Low-pressure zones and (b) distribution of total pressure.
these low-pressure zones, a fluid phase change occurs, provoking cavitation [1]. It is worth highlighting that the signs of
wear due to cavitation found in the impeller buckets are found precisely in these places (Fig. 2b). Fig. 10b shows the pressure
distribution on the buckets as a consequence of the water volume collision against these buckets.
With this pressure distribution, we established that the resulting force of the water jet acting on the surface of the buckets
in the x direction is 127,020 N and represents the resulting hydrodynamic force over the buckets available for power gen-
eration. In addition, a resulting force was found on the y and z coordinates because of the flow deviation, which is not con-
sidered analytically. This characteristic is one of the causes of the 23% difference found between the theoretical
hydrodynamic force of the jet on the buckets and the finite volume model in addition to the interaction of the fluid with
the second bucket.
The stress distribution obtained for the first load states as a consequence of rotation can be seen in Fig. 11.
Fig. 11 shows the effect of the centrifugal force on the buckets; stress is higher in the zone for the union between the base
and the buckets. Fig. 12 shows the lower zone of the union for the bucket and the impeller base; the maximum stress to
tension occurs in this zone with an average value of 44.17 MPa. It is important to highlight that the main crack was found
in this zone (Fig. 2c).
Fig. 11. Stress distribution versus the effect of rotation on the impeller.
J.C. Chávez et al. / Engineering Failure Analysis 48 (2015) 297–307 305
The stress distribution obtained by superimposing the effect of the impeller rotation and the distribution of pressures on
the bucket can be observed in Fig. 13. In addition, the deflection caused by the jet force can be observed (exaggerated for
illustration purposes).
The maximum value present for this state of stress occurred in the same zone of Fig. 12 but with a maximum value of
179.98 MPa at compression (Fig. 14).
To corroborate that the finite element model used to determine the stress present was adequately conducted, the torque
value generated by the pressure distribution was determined. The torque value on the shaft was: Tnumerical model = 53338.2 -
N m in the x direction. The value of the real torque (Treal) was calculated from data obtained during efficiency tests as
55001.6 N m. Comparing the torque value obtained through simulation of finite elements with respect to the real torque
revealed an error of 3.02%. This result indicates that the CFD model determined the correct field of pressure and, thereby,
that the results obtained are reliable.
The magnitude of the von Mises effective maximum stress obtained from simulation through finite elements was
179.98 MPa. Thus, the safety factor (S.F.) through yield for this load state is 1.37, considering a theoretical stress yield of
248.21 MPa for AISI 1026 steel.
To calculate the fatigue safety factor (Nf), we considered the two states of stress to which any of the necks of the impeller
under operation is subjected; the first (44.17 MPa) due to impeller rotation and the second (179.98 MPa) when the bucket
comes in contact with the jet due to the superposition of the impeller rotation and application of the pressure distribution.
From the aforementioned information, the mean and alternating stresses were obtained at 67.905 MPa and 112.075 MPa,
respectively. These stress values already include the intensification of the stress caused by the neck geometry but does not
Fig. 13. Distribution of stress due to impeller rotation and application of field of pressures.
306 J.C. Chávez et al. / Engineering Failure Analysis 48 (2015) 297–307
include additional stress concentration caused by micro-structural defects. The fatigue resistance limit without correcting
(S0e ) for steel with Sut < 400 MPa is estimated as half the ultimate stress under tension; hence, for the impeller material, it
was estimated at 200 MPa. The corrected fatigue resistance limit Se was calculated and is equal to 114.82 MPa through cor-
rection for surface finish (corrosion in water) and 90% reliability. A fatigue safety factor equal to 1.02 was calculated by using
the modified Goodman diagram in the scenario in which the ra/rm ratio varies constantly and bearing in mind that the mean
stress is negative, the failure line was horizontal at the level of the point Se (it was considered that the negative mean stress
does not increase the fatigue resistance). If the material in this impeller were homogeneous without imperfections, we
would expect that cracks would not nucleate because of fatigue; however, the defects found, micro cracks, non-metallic
inclusions, and trapped slag, are stress concentrators from which multiple cracks can propagate and cause fracture. This state
of stress explains the crack propagated based on the fatigue found. Additionally, for this calculation of safety factors, residual
stresses are not considered for which the magnitude and distribution are unknown. Given the level of uncertainty on the
state of residual stress, the safety factors found are quite low. A recovery procedure for this impeller is ruled out because
of the density of defects found and the low safety factors.
4. Conclusions
This article presents a failure analysis of a Pelton-type turbine impeller. The analysis performed permitted the following
conclusions.
The material of which the impeller is constructed presents multiple defects, such as non-metallic inclusions and micro
cracks, that have the potential to nucleate and facilitate the propagation of cracks on the impeller. Slag was found trapped
in the filler material, which could likewise nucleate and facilitate the propagation of cracks on the impeller.
The crack found on the neck propagated due to fatigue from the surface, near the zone affected by welding, in the pres-
ence of residual stresses that were not alleviated with heat treatment.
The computational fluid dynamics (CFD) analysis permitted the identification of low-pressure zones caused by an inad-
equate geometry for the bucket profile where cavitation is present. Additionally, from this analysis, we could establish the
field of pressures generated by the interaction between the fluid and the bucket; the field of pressures permitted the calcu-
lation of the load state with a 3% margin of error in the torque value with respect to the experimental value.
The finite element analysis permitted the establishment of the state of the stress to which the impeller is subjected during
operation, and the critical points in the neck of the buckets’ zone were identified, coinciding with the crack found. This zone
is under tensile stress due to the effect of the centrifugal force and compression when the bucket comes into contact with the
jet, causing fatigue. Safety factors were calculated at 1.37 and 1.02 for static load and fatigue, respectively.
Acknowledgements
The authors thank to Eng. Francisco Larrahondo at the Empresa de Energía del Pacifico, EPSA.
References
[1] Streeter VL, Wylie EB. Mecánica de Fluidos Octava edición. USA: McGraw-Hill; 1988.
[2] Veselý J, Varner M. A case study of upgrading of 62.5 MW pelton turbine. In: Proceedings of international conference. IAHR; 2001.
J.C. Chávez et al. / Engineering Failure Analysis 48 (2015) 297–307 307
[3] Ferreño D, Álvarez JA, Ruiz E, Méndez D, Rodríguez L, Hernández D. Failure analysis of a Pelton turbine manufactured in soft martensitic stainless steel
casting. Eng Fail Anal 2011;18:256–70.
[4] Padhy MK, Saini RP. Effect of size and concentration of silt particles on erosion of Pelton turbine buckets. Energy 2009;34:1477–83.
[5] Padhy MK, Saini RP. Study of silt erosion on performance of a Pelton turbine. Energy 2011;36:141–7.
[6] Kumar Pardeep, Saini RP. Study of cavitation in hydro turbines – a review. Renew Sustain Energy Rev 2010;14:374–83.
[7] Escaler X, Egusquiza E, Farhat M, Avellan F, Coussirat M. Detection of cavitation in hydraulic turbines. Mech Syst Signal Process 2006;20:983–1007.
[8] Hutchings IM. Tribology: friction and wear of engineering materials, Ed. London: Edward Arnold; 1992.
[9] Santa J, Blanco JA, Giraldo JE, Toro A. Cavitation erosion of martensitic and austenitic stainless steel welded coatings. Wear 2011;271:1445–53.
[10] Keck H, Sick M. Thirty years of numerical flow simulation in hydraulic turbomachines. Acta Mech 2008;201:211–29.
[11] Vesely J, Varner M. A case study of upgrading of 62 5 MW pelton turbine. In: Proc Int Conf. IAHR; 2001.
[12] Angehrn R. Safety engineering for the 423 MW-pelton-runners at bieudron. In: 20th IAHR symposium August 6–9. N.C., USA: Charlotte; 2000.
[13] Schmied J, Weiss T, Angehrn R. Detuning of pelton runners. In: 7th IFToMM-conference on rotor dynamics. Austria: Vienna; 2006.
[14] Mataix C. Mecánica de fluidos y máquinas hidráulicas. Segunda edición. Alfaomega Grupo Editor, S.A de C.V. Capítulo 19. Turbomáquinas Hidráulicas –
Generalidades; 2005.
[15] Perrig A. Hydrodynamics of the free surface flow in pelton turbine buckets. École polytechnique fédérale de lausanne. Thèse n. 3715. 2007. The Pelton
turbine. P. 47 [Chapter 4].
[16] Metals Handbook, vol. 11. Failure Analysis and Prevention; 1992.
[17] Robert L. Mott, Mecánica de Fluidos Aplicada, Cuarta Edición. Berkeley: Prentice Hall; 1996.
[18] White FM. Fluid mechanics. 4th Ed. McGraw-Hill; 1997.