Engineering Failure Analysis 90 (2018) 1–13

Contents lists available at ScienceDirect

Engineering Failure Analysis

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

Fatigue failure of the bolts connecting a Francis turbine with the

T
shaft
⁎
Fernando Casanovaa, , Carlos Mantillab
a
Escuela de Ingeniería Mecánica, Universidad del Valle, Calle 13 # 100-00, Cali, Colombia
b
CELSIA, Calle 15 # 29B-30, Autopista Cali Yumbo, Colombia

A R T IC LE I N F O ABS TRA CT

Keywords: This paper describes the failure analysis of bolts used to connect a turbine to the shaft of a
Hydropower plant hydroelectric power generator. Three of ten bolts were found broken during a disassembly
Fatigue failure process to replace the turbine. The methodology included visual inspection of the thread to detect
Bolts eventual stress concentrations. Using loads measured on the turbine shaft for different power
Preload
levels and the actual power generation history for a period of one year, a stress history on the
Francis turbine
Actual stress history
bolts was obtained. With this stress history, damage was calculated for three preload values.
The results showed some scratches on the root of the thread produced during the machining
process and by subsequent corrosion during operation, which likely significantly affected the
fatigue life. A preload that was smaller than the recommended one was found in the maintenance
report, which put the bolts at risk of loosening. A fatigue life of only 16.4 years was estimated if
loosening occurs on the bolts.
Better preload control during the assembly process and a change in the bolt material to
stainless steel are the primary recommendations from this analysis.

1. Introduction

Hydroelectric power plants are a source of electrical energy and are widely used around the world because of the low operational
costs and the low environmental costs compared with fuel-driven plants. In a hydropower plant, a turbine is rotated by the potential
and kinetic energy of water to transform it to mechanical work, which is transferred by a shaft to a generator where the mechanical
energy is transformed into electrical energy (Fig. 1).
The interactions of the water with the turbine and the electrical network with the generator, under some conditions, may produce
vibrations on the structural elements of the machine. Vibrations may produce important fluctuating stresses, which may produce
fatigue failures [1,2]. In large hydroelectric power units, fatigue failure may be catastrophic, resulting in human casualties, high costs
of repair or replacement of elements, and long repair times. High levels of vibrations are produced in Francis turbines, specifically
when operating at partial load conditions due to von Carman Vortices that form at the vanes and Vortex Ropes at the draft tube [3,4].
These vibrations have been found to be sufficient to produce fatigue failures in the structural elements [5]. This paper describes a
failure analysis on the connecting bolts of a Francis turbine belonging to an 18 MW hydro generator. During disassembling of the
machine for maintenance, three of the 10 bolts were found to have broken.

⁎
Corresponding author.
E-mail address: gonzalo.casanova@correounivalle.edu.co (F. Casanova).

https://doi.org/10.1016/j.engfailanal.2018.03.015
Received 15 November 2016; Accepted 13 March 2018
Available online 14 March 2018
1350-6307/ © 2018 Elsevier Ltd. All rights reserved.
F. Casanova, C. Mantilla Engineering Failure Analysis 90 (2018) 1–13

Fig. 1. Schematic of a hydroelectric power machine.

2. Description of the failure

The hydroelectric power unit that is the object of this study has been in operation during the last 50 years. The unit is powered by
a Francis-type turbine, which is 1.3-m high and has a major diameter of 2.03 m. The turbine is connected to a 0.46-m diameter shaft
using 10 bolt reference DIN 259 R2 ¼″, with a material that, according to the manufacturer, has a tensile strength between 784 and
931 MPa. The turbine disassembled approximately once per year for maintenance. By the year 2014, the company decided to replace
the old turbine with a new one. To characterize the machine before and after turbine replacement, the company decided to measure
the normal and shear stresses on the turbine shaft at several power generation levels. During disassembly of the old turbine, three of
the 10 bolts used to connect the turbine to the shaft were found to have broken. The three broken bolts exhibited a fatigue pro-
pagation region and a final fracture region. Fig. 2 shows the shaft with the bolts including the three bolts that fractured. Fig. 3 shows
the fracture surface of a bolt where the fatigue propagation region is clearly observed. The three bolts were found to have broken at
the threaded region in the first fillet in contact with the nut, as observed in Fig. 3.

3. Methodology

A fatigue life analysis was performed on the bolts to establish the probable cause of the failure. The steps used in the analysis are
explained in detail in this section. Visual inspection of the notched section of the thread was performed to evaluate the presence of
stress concentrators. A model of the shaft was generated to calculate the forces at the joint using the stresses measured at the shaft in

Fig. 2. Turbine shaft with the bolts after turbine removal.

2
F. Casanova, C. Mantilla Engineering Failure Analysis 90 (2018) 1–13

Fig. 3. A piece of fractured bolt inside the nut showing the fatigue propagation and final fracture regions.

the field tests. A power generation history was obtained from the company, and the measured stresses were compared to the unit
generation history to obtain a stress history. Using the stress history, the damage was calculated for every cycle of the stress history
and the total damage was calculated using Miner's rule. Tension tests were performed on specimens taken from the bolts to obtain the
mechanical properties of the material. Additionally, the applied preload on the bolts after maintenance was determined using the
maintenance reports. From interviews with the company staff, it was determined that maintenance operations did not include bolt
replacement. Instead, the same bolts have been used during the last 50 years. Therefore, the damage produced by the stress cycling
during the preload was also evaluated.

3.1. Visual inspection of the root of the threaded region

The root of the thread was observed with a stereoscope to evaluate the roughness of the surface and the presence of any corrosion.

3.2. Material mechanical properties

Tension tests according to the ASTM-E8 standard were performed on five specimens with a 6-mm diameter machined from one
bolt. A Tinius Olsen H50KS universal test machine with a capacity of 50 kN and error of 0.108% was used for the tests. The diameter
of specimens was measured with a ± 0.01-mm-resolution micrometer.

3.3. Strain measurement

Strain gauges DEA-06-250US-350 (Micro measurement, Malvern, PA) with a 10 K and a 9000 Torque Track system (Biensfeld
Engineering, Maple City, MI) were used to measure normal and shear strains on the turbine shaft. The signals from the Torque Track
systems were registered at a sampling frequency of 50 Hz using a 12 bit data acquisition system (model PMD 1280-LS, Measurement
Computing, Norton, MA). The normal strain was measured using four strain gauges connected in a full Wheatstone bridge config-
uration, with two of them aligned to the longitudinal direction of the shaft and the other two at 90 degrees. The shear strain was
measured using four strain gauges rotated 45 degrees with respect to the longitudinal direction of the shaft so that any normal strain
in the longitudinal direction was cancelled and only the principal strain produced by the torque was registered. The strain mea-
surements with the old turbine were performed during February 2015, with the machine operating at 2, 4, 6, 8, 10, 12, 14, and
18 MW. The new turbine was assembled in March 2016 during a dry season, meaning that the tests were performed with some
restrictions on the power output. Therefore, with the new turbine, strains were registered with the unit operating at 2, 4, 6, 8, 10,
12.5, and 15.2 MW. The reservoir level was not high enough to achieve a power greater than 15.2 MW. These power levels were
visually read from an analog indicator with a 0.5-MW resolution.
The measured normal strain was assumed to be produced by a hydraulic axial thrust and a bending moment applied on the shaft.
The strain produced by the bending moment was calculated from the 5 Hz component of the normal strain signal since the shaft
rotates with that frequency and it was assumed a constant orientation of the bending moment. The strain due to axial thrust was
calculated from the difference between the total and the 5 Hz component. Using the strains, the stresses were calculated from the
plane stress equations. The respective loads, axial thrust (Fa) and bending moment (M) were calculated using the equations:
Fa = Eεa A = σa A (1)

Eεb I σb I
M= =
r cos θ r cos θ (2)

where E is the Young's modulus of steel, A is the shaft cross-sectional area, I is the shaft inertial moment, r is the shaft radius, ε and σ
are the strains and stresses, respectively, where the index a corresponds to axial and sub index b corresponds to bending, and θ is the
angular position with respect to the point where εb is maximum (Fig. 4). Because of the cosine function in the denominator of Eq. (2),
π
inexact values are expected for θ~ 2 i for i = 1, 3, 5, …. Therefore, to compare between the old and new turbine and power levels, the

3
F. Casanova, C. Mantilla Engineering Failure Analysis 90 (2018) 1–13

Fig. 4. Cross section of the shaft and bolts showing the positions of the strain gauges (SG) used to measure the normal strain (r is the radius of the shaft, R is the radial
distance of the center of the bolts to the center of the shaft, θ is the angular position of the strain gauges with respect to the point of the maximum stress produced by
the bending moment, and di is the projected distance of each bolt to the center of the shaft).

σb I
RMS value of was used.
r

3.4. Bolt load conditions

The measured axial force and bending moment on the shaft were used to calculate the stresses on the bolts. The model for the
shaft-turbine assembly is presented in Fig. 5, which displays the relative location of the bearings, the bolts and the point where axial
strain was measured. Radial and axial restrictions were applied on the top side (the generator side) where a radial and an axial
bearing hold the shaft at that position. A radial restriction was considered for the lower side (turbine side) corresponding to the radial
bearing at that location. The axial force Pa is the weight of the turbine plus the measured hydraulic axial thrust. We assumed that a
radial force P in the middle of the runner produces the bending. With the dimensions presented in Fig. 5, the radial force P was
calculated as a function of the measured bending moment (M):

Fig. 5. Free body diagram of the shaft-runner assembly: P is the radial force at the midpoint of the turbine, Pa is the axial load produced by the weight of the runner
plus the measured axial thrust, a = 5.042 m, b = 1.417 m, c = 3.99 m, and e = 0.62 m.

4
F. Casanova, C. Mantilla Engineering Failure Analysis 90 (2018) 1–13

aM
P=
bc (3)
The bending moment (Mu) at the location of the lower bolts was calculated as:
eaM
Mu =
bc (4)
The axial forces acting on the bolts at the failure plane are produced by the preload force (Fp), the axial force produced by the
bending moment (Fm), the hydraulic axial thrust of the water on the runner and the weight of the runner (Pa).
The force due to the bending moment is:
Mu R cos θ
Fm = n
∑1 di2 (5)
Using Eqs. (2) and (4):
eaσb IR
Fm = n
bcr ∑1 di2 (6)
The total axial load on the bolts can be calculated using [6]:
Pa
F = Fp + C ⎛Fm + ⎞
⎝ n ⎠ (7)
where n is the number of bolts and C is a constant of the joint given as a function of the stiffness of the bolt (kb) and the stiffness of the
joint (km) by [6]:
kb
C=
kb + k m (8)

At Ad E
kb =
At ld + Ad lt (9)

0.577πEd
km =
( 0.577l + 0.5d
2 ln 5 0.577l + 2.5d ) (10)

l = ld + lt + 2t (11)
where At is the cross-sectional area of the threaded portion (0.00302 m2), lt is the length of the threaded portion under tension
(0.015 m), Ad is the area of the non-threaded portion (0.003848 m2), ld is the length of the non-threaded portion (0.155 m), E is the
elastic modulus of the steel (200 MPa), t is the thickness of the washer (5 mm), and d is the nominal diameter of the bolt (75.4 mm).
Finally, the stress may be calculated as:
F
σ=
At (12)

3.5. Fatigue life calculations

The power generation history of the unit was obtained from the operation data during 2014 as the hourly generation data (Active
power) with 1 MW resolution. The power generation history and measured stresses were linked to determine the stress history. The
measured stresses at each power level, where found to consist of an average value and a small variation of stresses produced by
vibrations. The average load values were linearly interpolated and fit to the power history plus one cycle of stress due to vibrations, to
determine a stress history corresponding to every change in the power level (CPL) of the unit. The rainflow counting method was
implemented to determine every stress reversal from the stress history, and the damage was calculated for every reversal. Separately,
the rainflow counting method and damage calculations were performed on the vibration stress history obtained for the machine
operating in steady state at every constant power level. In this case, the damage was calculated per hour. The damage due to small
stress variations corresponding to powers where the loads were not measured was assumed to be equal to the damage of the measured
power immediately higher than this (i.e., the damage due to a small stress variation at 3 WM was assumed to be same as the damage
at 4 MW). The damage due to small stress variations at every power level during the entire year was obtained as the product of the
damage per hour and the number of hours spent in operation at the corresponding power level.
Following an approach that is similar to the one used by Joyce and Lambrecht [7], the total damage during one year of normal
operation was calculated as the sum of the damage due to the CPL plus the damage from every small stress variation. The total
damage (D) was calculated using Miner's rule:
n
1
D= ∑N
i=1 fi (13)

5
F. Casanova, C. Mantilla Engineering Failure Analysis 90 (2018) 1–13

where n is the number of cycles and Nfi is the fatigue life for the i_th cycle. The fatigue life for each cycle was calculated using the S-N
uniaxial approach. Assuming a log-log straight line S-N relationship between a stress of 90% of the tensile strength (0.9Sut) and the
endurance limit (Se), the fatigue life (Nf) is:

σeqv ⎞1/ m
Nf = 106 ⎛ ⎜ ⎟

⎝ Se ⎠ (14)

where m is the slope of the S-N line and is given as:

log(Se ) − log(0.9Sut )
m=
3 (15)

The Goodman criterion was used to calculate σeqv, which was used to take into account the effect of the mean stress:
Sut σa
σeq =
Sut − σm (16)

where σa is the alternating stress and σm is the mean stress.

The endurance limit was estimated as:
Kb K c
Se = 0.5Sut
Kf (17)

where Kb is the size correction factor, Kc is the load correction factor, and Kf is the stress concentration corrector factor. The
endurance limit and the modification factors were calculated according to Shigley and Mischke [6]. For this case, the modification
factors were Kb = 1, Kc = 0.85, and Kf = 2.8 for machined thread, where the surface modification factor is included.

4. Results

4.1. Visual inspection of the thread

Although some regions of the thread were found to have a good surface finish (Fig. 6A), scratches that were likely produced

Fig. 6. Images showing the surface of the thread: A) Root of thread with a relatively good surface finish. B) Scratches at the root probably produced during the
machining. C) Scratches at the flank produced by the machining process. D) Cavity at the root produced by corrosion.

6
F. Casanova, C. Mantilla Engineering Failure Analysis 90 (2018) 1–13

Table 1
Mechanical properties of the bolt material: Sy is the yield strength, Sut is the tensile strength, σf is the true fracture strength, and εf is the true fracture strain.

Specimen Sy (MPa) Sut (MPa) σf (MPa) εf

1 589.9 794.8 1187.4 0.32

2 599.2 794.8 1241.2 0.34
3 587.7 791.3 1149.7 0.31
4 584.7 784.9 1138.4 0.32
5 560.2 763.3 1205.4 0.36
Average 584.3 785.8 1184.4 0.33

during the machining process were also found at the root (Fig. 6B) and at the flank of the thread (Fig. 6C). However, the most
worrying situation was the presence of groves produced by corrosion on the surface. The presence of corrosion may significantly
reduce the fatigue strength of the material.

4.2. Mechanical properties

The mechanical properties obtained from the tension tests are presented in Table 1. The average tensile strength (785.8 MPa) was
used to estimate the endurance limit.

4.3. Stress and loads on the shaft

The axial stresses on the shaft obtained for the new and old turbines are shown in Fig. 7. Clearly, greater stresses were obtained for
the old turbine at the evaluated power generation levels.
The average torque values as a function of power are shown in Fig. 8A. With the old turbine, the average torque was slightly
smaller, which may be due to an error in the power value read visually from the indicator. The standard deviation of the torque is
included in Fig. 8B as an indicator of the torque fluctuations around the average value. With the new turbine, the standard deviation
was smaller in general, meaning smaller torque fluctuation.
The average axial forces obtained with the new and old turbines are presented in Fig. 9A. Clearly with the old turbine, the axial
force is significantly greater. Fig. 9B shows the standard deviation of the axial force. With both turbines, the axial load fluctuations
were greater for powers between 6 and 10 MW but were slightly smaller with the new turbine.
The RMS value of the bending moment is shown in Fig. 10. The new turbine bending moment was approximately constant over
the entire power range, whereas with the old turbine, peaks were found for powers between 6 and 8 MW and for the full load.
Interestingly, for powers under 6 MW, the bending moment using the old turbine was smaller.

4.4. Loads on the bolts

Further investigation into the maintenance reports showed that a tightening torque (Tp) of 2500 lbf-ft (3389.5 N-m) was used in
the assembly process of the turbine during 2011. The preload force (Fp) was calculated using the equation [6]:
Tp = KdFp (18)
where K is a constant that depends on the joint. Taking K = 0.3 as recommended in [6], the obtained preload was Fp = 0.175 MN.
This value was compared with the value recommended by [6] given by:
Fp = 0.75Sp At (19)
where Sp is the proof strength, which can be approximated by 0.85Sy = 496.4 MPa. Then, the recommended preload value was
1.124 MN, meaning that the applied preload was only 15.5% of the recommended preload. With this low preload, low stresses on the

Fig. 7. Axial stress on the turbine shaft as a function of the generated power measured the old and new turbines.

7
F. Casanova, C. Mantilla Engineering Failure Analysis 90 (2018) 1–13

Fig. 8. Torque on the turbine shaft as a function of power: A) Average value and B) Standard deviation.

Fig. 9. Axial force on the turbine shaft as a function of power: A) Average value and B) Standard deviation.

bolts were obtained, as shown in Fig. 11, where the stress on the bolt during 0.4 s for the evaluated power level is shown together
with the stress generated by the preload. However, the problem with the low preload is that there is a possibility of separation in the
joint if the external load (Fext) is high enough. If separation occurs, the entire external load must be supported by the bolt, which
significantly increases the stresses on the bolt and can result in its failure [8]. Separation occurs if the portion of the external load
supported by the joint (Fg) becomes greater than Fp. The load supported by the joint is given by [8]:
Pa
Fg = (1 − C ) Fext = (1 − C ) ⎛Fm + ⎞
⎝ n ⎠ (20)
The separation condition was checked by calculating Fg and comparing with Fp. Fig. 12 shows Fg obtained for 2 and 18 MW
together with Fp, where it can be seen that for low power, Fg was considerably smaller than the preload. However, for 18 MW, even
though Fg was smaller than the preload, it was too close to be considered a safe condition, given the importance of this joint. With this

8
F. Casanova, C. Mantilla Engineering Failure Analysis 90 (2018) 1–13

Fig. 10. Bending moment RMS value as a function of power.

Fig. 11. Load on the bolts for different generated power levels. The number on the label of each curve corresponds to the generated power in MW. Fp corresponds to
the stress generated by the preload.

Fig. 12. Comparison of the force on the joint at 2 and 18 MW with the preload. The number on the label of each curve corresponds to the generated power in MW. Fp
corresponds to the preload.

low preload, vibrations in the machine may produce bolt loosening, which significantly increases the loads on the bolts and generates
fatigue failure. To evaluate the important influence of preload, fatigue analysis was performed with three preload values: the re-
commended preload (Fp = 1.124 MN), the applied preload (Fp = 0.175 MN), and the bolts loosening condition (Fp = 0).

4.5. Damage and fatigue life results

The power generation history during one year is shown in Fig. 13. A histogram of the generation history is shown in Fig. 14. The

9
F. Casanova, C. Mantilla Engineering Failure Analysis 90 (2018) 1–13

Fig. 13. Power generated during one year of unit operation.

Fig. 14. Histogram of the power generated during one year.

history shows that the machine operated at 15 MW most of the time.

The stress history during the entire year due to the CPL with a preload 0.175 MN is show in Fig. 15A, and a detailed view is shown
in Fig. 15B, where the change in the average stress due to the CPL and small oscillations due to the first cycle of vibrations can be
observed. A minimum value of 58.16 MPa is obtained when the machine is stopped. The stresses produced by the vibrations during
steady state operation at every power level are shown in Fig. 11.
The damage due to the CPL was 9.82e-6, 8.58e-7, and 2.33e-6 for preloads of 0, 0.175, 1.155 MN, respectively.
The damage per hour produced by vibration as a function of the power level is shown in Fig. 16. The loosening condition
produced damage that was significantly greater than the other two preload values. With respect to the power, the maximum damage
per hour was at 8 MW, which was the power where the bending moment was maximum. However, the maximum damage during the
entire year occurred at a power level of 15 MW (Fig. 17), as the machine was working at 15 MW most of the time.
Table 2 summarizes the damage produced by the CPL, the total damage produced by vibrations when calculated as the sum of the
damage at every power, the damage produced by the application of the preload during assembly every year, the total damage, and
the calculated fatigue life for every preload.

5. Discussion

Failure analysis of the turbine connecting bolts performed in this paper included mechanical properties evaluation, surface
inspection and fatigue analysis using loads measured on the shaft and actual generation history.
The mechanical tests showed that the material corresponds to the specifications from the manufacturer. However, the low quality
surface produced during machining is a significant stress concentrator. In addition, the corrosion detected on the root of the thread
most likely decreased the fatigue life of the pieces.
The normal stresses with the old turbine were greater than those obtained with the new one. This higher stress should have played
an important role in the failure of the bolts, especially those stresses produced by the bending moment, as shown by the results from
the damage per hour calculations. This is because the bending moment produces stresses with high fluctuations, which change from a

10
F. Casanova, C. Mantilla Engineering Failure Analysis 90 (2018) 1–13

Fig. 15. Stress history produced by the CPL: A) Stress history during the entire year and B) Detailed view of the stress history for a 30-s period.

Fig. 16. Damage per hour in log scale produced by vibrations during steady state operation at each power level.

positive to a negative value due to the rotation of the shaft.

The damage-preload analysis showed that loosening is the most critical condition for the bolts with a life of only 16.4 years. This
life is smaller than the actual life of the bolts (~50 years). This discrepancy may have several explanations: 1) loads on the shaft and
bolts during the entire life of the machine are not necessarily the same as was measured in this study. With time, wear of the seals of
the turbine allows water to move into the upper part of the turbine, increasing the axial load. This may be the reason why with the
new turbine axial load was significantly smaller. 2) The preload from other assembly processes may have been greater than the
preload reported for the 2011 assembly. 3) The loosening condition was not necessarily reached for all of the bolts at the same time
because the load produced by the bending moment fluctuates around zero. Even though with a preload greater than zero, separation
may occur at a bolt for the angular position of the maximum load, this is not necessarily the case for the entire cycle of the bolt
around the rotation axes.

11
F. Casanova, C. Mantilla Engineering Failure Analysis 90 (2018) 1–13

Fig. 17. Total damage during the entire year in log scale produced by vibrations during steady state operation at every power level.

Table 2
Damage due to the CPL, damage produced by vibrations, damage produced by application of the preload during assembly of the
machine, total damage, and the fatigue life for every evaluated preload.

Preload 0 MN 0.175 MN 1.124 MN

Vibration 0.060946 0.00521 0.014225

CPL 9.82E-06 8.58E-07 2.33E-06
Assembly 0 1.05E-07 3.62E-06
Total 0.060956 0.005211 0.014231
Life (years) 16.40538 191.9079 70.26912

The calculated life obtained with the recommended preload was smaller than that with the reported preload (0.175 MN). At first
glance, the reported preload produces better working conditions for the shaft. The problem with this low preload is that the bolts
have a high probability of loosening. From the authors' point of view, separation of the joint due to the low preload or loosening of
the bolts was the likely reason for the failure, despite the fact that physical evidence of such a separation was not found. The reason
for this may be that the corrosion of the surfaces may have masked any trace of the relative motion between the surfaces.
Additionally, as mentioned before, the separation condition does not have to occur at the same time for all of the bolts. In this case,
the bolts with any remaining preload may restrict the relative motion between the surfaces.
All of the equivalent stresses detected by the rainflow counting method were included in the damage calculation. However,
several of those equivalent stresses were smaller than the endurance limit of the material. It is believed that those small equivalent
stresses do not produce damage and should be excluded from the damage calculation. At this point, it is important to consider the fact
that the bolts are in contact with water, which produced corrosion. In the presence of corrosion, the endurance limit does not exist
and fatigue cracks propagate, even with low stresses. For this reason, the calculated damage is not necessarily overestimated. More
research is needed to evaluate the actual effect of corrosion on the S-N curve of the material.

6. Final comments

• Corrosion was detected on the root of the thread, which may have played an important role in fatigue. Stainless steel should be
considered as a material to replace the current material.
• Greater normal stresses with greater fluctuation were obtained with the old turbine. However, as the damage was mainly due to
the bending moment, bolts with the new turbine are approximately at the same risk of damage because the bending moment with
both turbines at those powers where the machine was working most of the time was comparable.
• A preload smaller than the recommended one, which may have produced separation of the joint, is a probable reason for the
failure. Better control of preload should be implemented in the protocols of machine assembly.
• Even though the calculated life with the recommended preload was smaller than the life with the reported preload, the life with
the recommended preload was higher than the expected life of the turbine (50 years). The most important thing is that with the
recommended preload, the risk of separation is smaller.

12
F. Casanova, C. Mantilla Engineering Failure Analysis 90 (2018) 1–13

Acknowledgements

This work was supported by the Vicerrectory of research from the Universidad del Valle, Cali, Colombia, grant number CI 806.
The authors also thank CELSIA-EPSA (Empresa de Energia del Pacifico) for supporting this research.

References

[1] D. Momcilovic, Z. Odanovic, R. Mitrovic, I. Atanasovska, T. Vuherer, Failure analysis of hydraulic turbine shaft, Eng. Fail. Anal. 20 (2012) 54–66, http://dx.doi.
org/10.1016/j.engfailanal.2011.10.006.
[2] Z. Zhang, Z. Yin, T. Han, A. Tan, Fracture analysis of wind turbine main shaft, Eng. Fail. Anal. 34 (2013) 129–139, http://dx.doi.org/10.1016/j.engfailanal.2013.
07.014.
[3] Z. Qian, J. Yang, W. Huai, Numerical simulation and analysis of pressure pulsation in Francis hydraulic turbine with air admission, J. Hydrodyn. Ser. B 19 (2007)
467–472, http://dx.doi.org/10.1016/S1001-6058(07)60141-3.
[4] C. Nicolet, J. Arpe, F. Avellan, Identification and modelling of pressure fluctuations of a Francis turbine scale model at partial load operation, 22nd IAHR
Symposium on Hydraulic Machinery and Systems, Secretariat 22th IAHR Symposium, Stockholm, 2004, pp. 1–14.
[5] F. Casanova, Failure analysis of the draft tube connecting bolts of a Francis-type hydroelectric power plant, Eng. Fail. Anal. 16 (2009) 2202–2208, http://dx.doi.
org/10.1016/j.engfailanal.2009.03.003.
[6] J.E. Shigley, C.R. Mischke, Design in Mechanical Engineer, 6th ed., McGRAW-HILL, Mexico, 2002.
[7] J.S. Joyce, D. Lambrecth, Status of evaluating the fatigue of large steam turbine-generator caused by electrical disturbances, IEEE Transactions on Power
Apparatus and Systems Vol. PAS-99, No. 1, 1980, pp. 111–119, , http://dx.doi.org/10.1109/TPAS.1980.319616.
[8] S.H. Molaei, R. Alizadeh, M. Attarian, Y. Jeferian, A failure analysis study on the fractured connecting bolts of a filter press, Case Stud. Eng. Fail. Anal. 4 (2015)
26–38, http://dx.doi.org/10.1016/j.csefa.2015.07.003.

13