Eirik Lødemel

Fatigue loads in a Francis turbine

runner
Master’s thesis

Master’s thesis in Department of Mechanical Engineering and

Industrial Design
Supervisor: Roy Johnsen
Academic Supervisor: Ole Gunnar Dahlhaug
January 2019
NTNU
Norwegian University of Science and Technology
Faculty of Engineering
Department of Mechanical and Industrial Engineering
Eirik Lødemel

Fatigue loads in a Francis turbine runner

Master’s thesis in Department of Mechanical Engineering and

Industrial Design
Supervisor: Roy Johnsen
Academic Supervisor: Ole Gunnar Dahlhaug
January 2019

Norwegian University of Science and Technology

Faculty of Engineering
Department of Mechanical and Industrial Engineering
Norwegian University Department of Energy and
of Science and Technology Process Engineering

EPT-M-2018-
MASTER THESIS
for
Eirik Lødemel

Autumn 2018

Fatigue loads in a Francis turbine runner

Utmattingslaster i et høytrykks Francis løpehjul

Background
The average age of Norwegian hydropower plants is 45 years, and many show sign of fatigue
and needs to be constantly maintained or refurbished. Additionally, some power plants in
Norway has experienced failures on new Francis runners. The main problem is the formation
of cracks in the turbine runner.
A classic problem in the start-up phase of a Francis turbine is the synchronization with the grid:
The turbine is sped up to synchronous speed, and spins at spin-no-load for a while until it is in
stable phase with the grid. Due to the large pressure pulsations at this operating regime, the
turbine will not in fact be rotating at a fixed RPM, making the synchronization time longer,
further exposing the runner to the stresses of the operating regime. This causes reduced lifetime
of the runner. With a variable-speed turbine, there is no need for this synchronization, and
speed-control can be employed at lower-than-synchronous speeds, where the pressure
pulsations are less aggressive. The challenges in this work is to investigate the fatigue loads of
the turbine runner and estimate crack growth in extreme flexible operating conditions.

Objective
Evaluate how variable speed operation of Francis turbines be utilized to minimize dynamic
loads at start-stop scenarios.

The following tasks are to be considered:

1. Literature study
a. Pressure pulsations of a Francis turbine
b. Fatigue loads on a Francis turbine
2. Software knowledge
a. Labview will be used for the measurements
b. Matlab will be used for the evaluation of the measurements
3. Laboratory preparations
a. Dynamic and static calibration of the instruments used in the test rig.
4. Measurements in the Waterpower laboratory:
a. Complete hill diagram for a Francis-99 turbine.
b. Stress measurements in the runner at runaway speed, start-up and shut down

i
 -- ” --

Within 14 days of receiving the written text on the master thesis, the candidate shall submit a
research plan for his project to the department.

When the thesis is evaluated, emphasis is put on processing of the results, and that they are
presented in tabular and/or graphic form in a clear manner, and that they are analyzed carefully.

The thesis should be formulated as a research report with summary both in English and
Norwegian, conclusion, literature references, table of contents etc. During the preparation of
the text, the candidate should make an effort to produce a well-structured and easily readable
report. In order to ease the evaluation of the thesis, it is important that the cross-references are
correct. In the making of the report, strong emphasis should be placed on both a thorough
discussion of the results and an orderly presentation.

The candidate is requested to initiate and keep close contact with his/her academic supervisor(s)
throughout the working period. The candidate must follow the rules and regulations of NTNU
as well as passive directions given by the Department of Energy and Process Engineering.

Risk assessment of the candidate's work shall be carried out according to the department's
procedures. The risk assessment must be documented and included as part of the final report.
Events related to the candidate's work adversely affecting the health, safety or security, must
be documented and included as part of the final report. If the documentation on risk assessment
represents a large number of pages, the full version is to be submitted electronically to the
supervisor and an excerpt is included in the report.

Pursuant to “Regulations concerning the supplementary provisions to the technology study

program/Master of Science” at NTNU §20, the Department reserves the permission to utilize
all the results and data for teaching and research purposes as well as in future publications.

The final report is to be submitted digitally in DAIM. An executive summary of the thesis
including title, student’s name, supervisor's name, year, department name, and NTNU's logo
and name, shall be submitted to the department as a separate pdf file. Based on an agreement
with the supervisor, the final report and other material and documents may be given to the
supervisor in digital format.

Work to be done in the Waterpower laboratory

Field work

Department of Energy and Process Engineering, 13. August 2018

________________________________

Ole Gunnar Dahlhaug

Academic Supervisor

Roy Johnsen
Supervisor

ii
Co-Supervisors:
 Chirag Trivedi
 Igor Iliev
 Einar Agnalt

iii
Preface
This master’s thesis was conducted, and written, at the Waterpower Laboratory, Department
of Energy and Process Engineering at Norwegian University of Science and Technology
during the autumn of 2018. My background as a student of Product Development and
Material Engineering, so that is why this thesis is under the department of Mechanical
Engineering and Industrial Design. Which is why I have two professors as supervisors, as
well as three co-supervisors. It should be noted that Ole Gunnar Dahlhaug is the supervisor I
have worked with and who has provided guidance during this thesis.

The work has been demanding, challenging and overall- fun. The work consisted of a large
amount of practical laboratory measurements. I would like to thank Joar Grimstad and PhD
candidate Einar Agnalt for helping me both understand and conduct the practical experiments.

As a student with no background in hydropower, there was quite a lot to learn about turbines
during this semester. Professor Ole Gunnar Dahlhaug and PhD candidate Igor Iliev for
helping me understand the dynamics of the Francis turbines.

I would also like to thank my last co-supervisor PhD candidate Chiraq Trivedi, as well as all
the students and employees at the Waterpower Laboratory for creating a welcoming and
supportive atmosphere during my time there.

Eirik Lødemel
Trondheim, January 7, 2019

iv
v
Abstract

The expansion of intermittent renewable energy sources in the European energy market is
leading to an increasing demand of regulatory energy sources to stabilize the energy grid.
Hydropower can act as a regulatory energy supply, but that is requiring a more flexible day-
to-day operation of turbines. This leads to turbines having to operate to a larger degree in
unfavorable operating conditions when it comes to efficiency and fatigue, and an increase in
start-stop cycles.

Variable speed turbines do not have to operate at a synchronous speed, and therefore has
greater flexibility when it comes to operation. This thesis aims to investigate the effect
variable speed operation can have on the fatigue life of the runner. To accomplish this,
pressure measurements for a scaled model of a low-specific-speed Francis turbine has been
conducted.

These pressure measurements are used to map the pressure pulsations and calculate stresses
and fatigue over the operating range for start-stop cycles and for steady-state operation.
The frequencies of the pressure pulsations were also analyzed. But couldn’t be used for the
fatigue analyzes since they were not measured for the runner blade.
The results show that the fatigue loads can be significantly reduced by reducing the runner
speed with a variable speed turbine. The most significant reductions in fatigue loads are from
start-stop cycles, where a slight reduction in runner speed at part load gave above 80 percent
reduction in fatigue loads.

Because the fatigue results were based solely on pressure measurements, several assumptions
had to be made in order to produce results. The uncertainty these assumptions bring are large
enough that the results should not be used directly, but rather as indications on the relative
differences on the operating range.

vi
 Sammendrag
Utvidelsen av fornybare energikilder i det europeiske energimarkedet fører til en økende
etterspørsel etter regulerbare energikilder som kan stabilisere nettfrekvensen. Vannkraft kan
fungere som en regulerbar energikilde, men det stiller krav om en mer fleksibel daglig drift av
turbinene. Dette fører til at turbinene i større grad å operere under ugunstige driftsforhold når
det gjelder effektivitet og utmattelse, i tillegg til en økning i antall start-stop sykluser.

Turbiner med som kan operere med variabel turtallskjøring har større fleksibilitet når det
gjelder hva slags operasjonsparametere de kan driftes med. Dette oppgaven tar til sikte å
undersøke hvordan operasjon med variabelt turtallskjøring kan redusere utmattelsen i
løpehjulet til Francis turbiner. For å oppnå dette har det blitt utført trykkmålinger på en skalert
modellturbin med lav spesifikk hastighet.

Trykkmålingene har blitt brukt til å kartlegge trykkpulsasjoner og beregne spenninger og

utmattelse i løpeskovlen for hele driftsområdet. Frekvensene til trykkpulsasjonene har også
blitt undersøkt, men har ikke blitt brukt videre siden spenningene som oppstår på
løpehjulsskovlen har andre frekvenser.

Resultatene viser at utmattelsesbelastningen kan reduseres kraftig ved å bruke en turbin med
variable hastighet. De største reduseringene i utmattelsesbelastninger finner man for start-stop
sykluser, hvor en liten reduksjon i løpehjulshastighet førte til over 80 prosent mindre
utmattelsesbelastninger.

Siden utmattelsesberegningene kun var basert på trykkmålinger måtte en rekke antagelser tas
for å produsere resultater. Usikkerhetene disse beregningene skaper gjør at resultatene ikke
kan bruker som absoluttverdier, men heller som indikatorer på de relative forskjellene
innenfor driftsområdet.

vii
viii
 Symbols and abbreviations

Symbol Description Unit

𝜖 Strain -
𝜎 Stress Pa
𝑣 Poisson’s ratio -
𝑛 Runner Speed RPM
𝑆 Constant amplitude stress range Pa
𝑃 Pressure Pa
𝑁 Cycles to failure -
𝐶 Fraction of lifetime consumed -
T Torque Nm
QED Dimensionless Discharge Factor -
Q Charge V
Q Minimum change in value -
NED Dimensionless Rotational Speed Factor -
M Number of bits -
m Constant for inverse slope of S-N curve -
Kt Stress concentration factor -
K1 Constant for S-N curve. -
g Gravity m/s2
E Modulus of Elasticity Pa
𝜎𝑚 Mean stress Pa
𝜎𝑒𝑓𝑓 Effective stress Pa
𝜎𝑎 Stress amplitude Pa
𝜎𝑉𝑀 Von Mises stress Pa
𝑛ℎ Cycles per hour -
𝑍𝑠𝑏 Number of Splitter Blades -
𝑍𝑟 Number of Runner Blades -
𝑍𝑔𝑣 Number of Guide Vanes -
𝑃𝜎 Stress to pressure factor -
𝑁𝑂 Cycles to failure for operating point -
𝐹𝑟𝑠𝑏 Splitter Blade Passing Frequency Hz
𝐹𝑟𝑏 Runner Blade Frequency Hz
𝐹𝑟 Rheingans Frequency Hz
𝐹𝑛 Runner Frequency Hz
𝐹𝑔𝑣 Guide Vane Frequency Hz
𝐸𝐹𝑆𝑅 Range of analog signal -

ix
Abbreviation Description

BEP Best Efficiency Point

CAL Constant Amplitude Loads
VAL Variable Amplitude Loads
cDAQ Compact Data Acquisition System
CFD Computer Fluid Dynamics
FEM Finite Element Method
FFT Fast Fourier Transform
H Head
HCF High Cycle Fatigue
Hz Frequency
LCF Low Cycle Fatigue
RPM Revolves Pr. Minute
RSI Rotor-Stator Interaction
S-N Stress-Cycles to failure
VAL Variable Amplitude Loads
NTNU Norwegian University of Science and Technology
LSB Least Significant Bit
FE Finite Element
DAQ Data Acquisition
cDAQ Compact Data Acquisition
NI National Instruments
FSI Fluid Structure Interaction
UTS Ultimate Tensile Strength
SNL Speed-No-Load

x
 Contents

1. Theory ............................................................................................................................................. 1
1.1. Pressure pulsations ...................................................................................................... 1
1.2. Material science............................................................................................................... 3
1.2.1. Fatigue ...................................................................................................................... 3
1.2.2. Goodman’s Method ................................................................................................. 5
1.2.3. Cumulative Damage ................................................................................................. 6
1.3. Recent studies done on Francis turbines ...................................................................... 6
1.3.1. Pressure pulsations and fatigue ............................................................................ 6
1.3.2. Strain gauge measurements ................................................................................ 10
1.3.3. Flexible turbines ................................................................................................. 10
1.4. Analysis method. ....................................................................................................... 11
2. Hypothesis .................................................................................................................................... 16
2.1. Formulated Hypothesis .............................................................................................. 16
3. Laboratory setup .......................................................................................................................... 18
3.1. The Francis Rig ......................................................................................................... 18
3.1.1. Components ........................................................................................................ 18
3.1.2. Open loop configuration..................................................................................... 18
3.2. Instrumentation .......................................................................................................... 19
3.3. Calibration and uncertainty ....................................................................................... 21
3.3.1. Pressure tranducers ............................................................................................. 21
3.3.2. Strain gauges ...................................................................................................... 22
3.4. Data acquisition and processing ................................................................................ 25
3.4.1. Data acquisition. ................................................................................................. 25
3.4.2. Sample rate and data processing ........................................................................ 25
4. Results........................................................................................................................................... 27
4.1. Pressure pulsation diagrams ...................................................................................... 28
4.1.1. GV4 .................................................................................................................... 28
4.1.2. GV5 .................................................................................................................... 28
4.1.3. GV6 .................................................................................................................... 29
4.2. Pressure pulsations along constant guide vane angles ............................................... 30
4.3. Frequencies in different operating regimes ............................................................... 35
4.3.1. BEP..................................................................................................................... 35
4.3.2. 2 degrees’ guide vane opening ........................................................................... 35

xi
 4.3.3. 5 degrees’ guide vane opening ........................................................................... 37
4.3.4. 7 degrees’ guide vane opening ........................................................................... 38
4.4. Speed-no-load ............................................................................................................ 39
4.5. Stress levels ............................................................................................................... 41
4.5.1. Stress levels at Speed-no-Load .......................................................................... 44
4.6. Fatigue ....................................................................................................................... 45
4.6.1. High Cycle Fatigue ............................................................................................. 46
4.6.2. Low Cycle Fatigue ............................................................................................. 47
5. Discussion ..................................................................................................................................... 49
5.1. Reducing pressure pulsations with variable speed turbine ........................................ 49
5.2. Frequency analysis .................................................................................................... 51
5.3. Strain gauges.............................................................................................................. 51
5.4. Fatigue assessment. ................................................................................................... 52
5.4.1. High Cycle Fatigue ............................................................................................. 52
5.4.2. Low Cycle Fatigue ............................................................................................. 53
5.4.3. Combining start-stop cycles and operating hours. ............................................. 54
5.4.4. Impact of variable speed turbines....................................................................... 55
5.4.5. Scaling to a larger turbine .................................................................................. 56
5.5. Known errors and assumptions made. ....................................................................... 56
5.5.1. Stress levels ........................................................................................................ 56
5.5.2. Stress frequencies ............................................................................................... 57
5.5.3. Start-stop cycles. ................................................................................................ 57
6. Conclusion .................................................................................................................................... 59
7. Further Work ................................................................................................................................ 61
8. Bibliography .................................................................................................................................. 62
Appendix A – Risk Assesment ................................................................................................................ 65
Appendix B – Matlab scripts .................................................................................................................. 75
Hill chart and pressure diagrams .......................................................................................... 75
Appendix C – Calibration reports .......................................................................................................... 96
Calibration report for GV4 – Meas XP5 .......................................................................... 96
Calibration report for GV5 – Kulite XTE ........................................................................ 98
Calibration report for GV6 – Kulite XTE ...................................................................... 100
Calibration report for Friction Torque – TW T2 ............................................................ 102
Calibration report for Generator Torque – WT T2 ........................................................ 104

xii
xiii
 1.Theory

1.1. Pressure pulsations

One of the most common operational difficulties for Francis turbines is pressure pulsations.
Pressure pulsations creates vibrations which can cause fatigue and material defects. Because
of the design of Francis turbines and the nature of flow, pressure pulsations will occur in all
operational regimes.
Due to the expansion in intermittent renewable energy in Europe the European energy market
to an increasing degree is in need of regulatory energy supplies. Hydropower has the ability to
meet these demands, but that requires the turbines to operate at a higher number of start-stop
cycles and to a larger degree operate on part-load. [11]
Operating on part load can be severely damaging to Francis turbines as a vortex rope will
appear in the draft tube at some operation points, usually between 50% and 85% load. [8]

Pressure pulsations with particularly high amplitudes are mainly cause by four different
frequencies. Guide vane frequency, Runner blade frequency, Rheingans frequency, Runner
frequency. [6]

Runner frequency:
The runner frequency will usually not have large enough amplitudes to cause significant
problems if the runner is undamaged and balanced. The frequency is determined by the runner
speed. [5]
𝑛
𝐹𝑛 = 60 (Hz) [1]

Where 𝑛 is the runner speed denoted in rpm.

Two different frequencies occur between the runner blades and the guide vanes. One for every
time a particular runner blade passes a guide vane and one for every time a particular guide
vane is passed by one of the runner blades. This is known as Rotor-Stator Interaction. [7]

1
 Figure 1.1: Flow field distortion between guide vanes and runner blades. Figure from One-dimensional modeling
of rotor stator interaction in Francis pump-turbine. [9]

Runner blade frequency:

𝐹𝑟𝑏 = 𝑍𝑟 ∗ 𝐹𝑛 (Hz) [2]

Where 𝑍𝑟 is the number of runner blades on the runner.

Every time a runner blade approaches a guide vane there will be an increase in local pressure,
creating a pressure pulsation. Because this pressure pulsation is created every time a runner
blade passes the same guide vane its frequency is determined by the number of runner blades
and the runner frequency. The amplitude is greatly affected by the distance between the
runner blades and the guide vanes, where an increase in the distance causes a formidable
reduction. [7]

Guide vane frequency:

𝐹𝑔𝑣 = 𝑍𝑔𝑣 ∗ 𝐹𝑛 (Hz) [3]
𝑍𝑔𝑣 is the number of guide vanes.
The pressure difference between the pressure-and suction side between two guide vanes
creates a non-uniform pressure and velocity field that the runner blades passes every time they
move past a guide vane. There are created one such non-uniform pressure and velocity field
for each guide vane, which means that the frequency is decided by the number of guide vanes
and the runner frequency. [5,7]

Rheingans frequency:

Due to the rotating component of the absolute velocity at the outlet from the runner, a
cavitated vortex rope appears in the draft tube. This creates low frequency pressure
amplitudes. The direction of the vortex is decided by the operating regime of the turbine. At
full load, the rotating component of the absolute velocity will move in the opposite direction
of the runner and at part load the same direction as the runner. This is because there is a
rotational component of the absolute velocity when the water exits the runner when it is
operating outside of the Best Efficiency Point. The rotational velocity of the water exiting the
runner will be larger the further away from BEP the runner is operating.

2
 𝐹𝑛 𝐹𝑛
< 𝐹𝑟 < (Hz) [4]
3,6 3

1.2. Material science

𝐹𝑜𝑟𝑐𝑒 𝐹
Stress 𝜎 = 𝐶𝑟𝑜𝑠𝑠−𝑠𝑒𝑐𝑡𝑖𝑜𝑛𝑎𝑙 𝐴𝑟𝑒𝑎 = 𝐴 (Pa) [5]

𝐸𝑙𝑜𝑛𝑔𝑎𝑡𝑖𝑜𝑛 ∆𝐿
Strain 𝜖 = 𝑂𝑟𝑖𝑔𝑖𝑛𝑎𝑙 𝑙𝑒𝑛𝑔𝑡ℎ = (-) [6]
𝐿

Strain can be separated into nominal tensile strain and nominal lateral strain. Where tensile
strain is the strain that occurs parallel to the stress applied and lateral strain is the strain that
occurs normal to the stress applied.

The relationship between the nominal lateral strain and the nominal tensile strain is called
Poisson’s ratio and is defined by
𝜀
𝑣= 𝑙 (-) [7]
𝜀𝑡

For elastic deformation, the relationship between stress and strain is described by Hooks law.

𝜎 =𝐸∗𝜀 (Pa) [8]

Where 𝜎 is the stress applied, 𝜀 is the strain and E is Young’s modulus. Which is the stiffness
value of the material. This relationship only applies when the stress is relatively small, after a
certain threshold, the yield limit 𝜎𝑌 , plastic deformation will occur. If the stress keeps
increasing this will eventually lead to fracture. Fracture can also occur as a fatigue or fast
fracture. Fast fracture is the result of a flaw in the material, often a crack, expands quickly due
to applied stress so that the material fractures.

1.2.1. Fatigue
During the lifetime of most materials they are subjugated to both constant amplitude fatigue
loads and variable amplitude loads. The runner blades in a Francis turbine are exposed to
both. These low level stress cycles will at some point form cracks in the material which will
grow until the material fractures.
To give a lifetime assessment of the runner blades, measuring the impact of both amplitude
loads are performed and then the number of cycles the given material can withstand is plotted
in an S-N curve. The S-N curve describes the relationship between amplitude size of cyclic
loads and the number cycles to failure. Both in a logarithmic scale. An S-N curve can usually
be divided into three regions. The plastic region, elastic region and the infinite life region.
Stress amplitudes that leads to plastic deformation leads to a low number of cycles before
failure, this region describes what is called low cycle fatigue. While stress amplitudes that
3
leads to elastic deformation either leads to a medium number of cycles or if they are low
enough over large number of cycles, often referred to as infinite life. These are called high
cycle fatigue. [33]

Figure 1.2: Stresses. Showing a cycle of a one mean stress and several stress amplitudes
occurring on top of it.

For Francis turbine runners, high cycle fatigue is the result of fluctuating stresses during
operation at a specific operating point, noted as stress amplitude in the picture above. Because
of the high rotational speed of the turbine these amplitudes occur thousands of times per
minute. The larger stress amplitudes that creates low cycle fatigue occur when changing
operating conditions, the largest being a change from stationary condition to operating
condition, which occurs during start-stop cycles.

When the stress amplitudes are low enough they will reach the endurance limit, and one can
assume that they will never lead to failure. Since infinite number of cycles can’t be tested on a
material this is commonly assumed at over 107 cycles. However, for Francis turbines infinite
life cannot be assumed at 107 cycles, as there are pressure fluctuations occurring at very high
frequencies during operation. RSI from the guide vane passing frequency for a turbine with 20
guide vanes occurs 12000 times every minute if operating at 600 RPM, which would result in
over 107 cycles in under 14 operating hours.

4
Figure 1.3: S-N curve from ISO 19902[16]. Showing two different S-N curves based on
structure parameters.

In this thesis an S-N Curve model from ISO 19902 is used. [16]
ISO 19902 describes the relationship between the number of cycles to failure 𝑁 and the
constant amplitude stress range 𝑆 with the following equation.

𝑙𝑜𝑔10 𝑁 = 𝑙𝑜𝑔10 𝐾1 − 𝑚 ∙ 𝑙𝑜𝑔10 𝑆 (Cycles) [9]

Where 𝐾1 is a constant and 𝑚 is the inverse slope for the S-N curve. Their values are values
are determined by the size of N and the construction details of the material at the spot which
are being considered. Important to notice that the constant amplitude stress range is measured
in MPa in this equation.

1.2.2. Goodman’s Method

It’s not only the stress amplitude that contributes to the fatigue of a material. A stress
amplitude during already high levels of stress contributes more than a stress amplitude in a
lower state of stress. To account for the mean stress an effective stress parameter is calculated.
A commonly used method to account for this is Goodman’s method. [32] [14]
𝑈𝑇𝑆
𝜎𝑒𝑓𝑓 = 𝜎𝑎 ∗ (𝑈𝑇𝑆−𝜎 ) (Pa) [10]
𝑚

𝜎 −𝜎
Where 𝜎𝑎 is the stress amplitude: 𝜎𝑎 = 𝑚𝑎𝑥 2 𝑚𝑖𝑛 (Pa) [11]
UTS is the ultimate tensile strength of the material and 𝜎𝑚 is the mean stress
𝜎 +𝜎
𝜎𝑚 = 𝑚𝑎𝑥 2 𝑚𝑖𝑛 (Pa) [12]

5
 1.2.3. Cumulative Damage
When there is variable amplitude loads the lifetime of the structure is described by Miner’s
rule.
𝑛
𝐶 = ∑𝑘𝑖=1 𝑁𝑖 (-) [13]
𝑖

Where k is the number of different levels of stress, 𝑛 is the number of cycles for stress level i
and 𝑁 is the number of cycles before failure for stress level i. The result, C, is the fraction of
lifetime consumed by the stresses. [8]

1.3. Recent studies done on Francis turbines

In the recent years there have been done a lot of different studies on fatigue on Francis
runners. Both by doing experiments and simulations, often a combination.

1.3.1. Pressure pulsations and fatigue

Fatigue analysis of the Prototype Francis runner based on site measurements and
simulation. [3]

In Fatigue analysis of the Prototype Francis runner based on site measurements and
simulation, Huang, Chamberland-Lauzon, Oram, Klopfer and Ruchonnet presents fatigue
analysis on prototype Francis runners based strain gauge site measurements and numerical
simulations. The paper also discusses the damage factors at different operating points, arguing
that due to the increasing energy production from intermittent renewable energy sources
hydropower plants will have to operate more outside of BEP.

Strain gauges were installed on the blades of several Francis runners. They used stress
calculations for different operating conditions to place the strain gauges at stress hot spots.
The results from the strain gauge measurements provided them with the static and dynamic
stresses at the different operating conditions.
They obtained the corresponding numerical static stress with CFD simulation and FEM
calculation. X Huang et al validated the static stress calculation using strain gauge
measurements on several turbines.

Fast Fourier transform analysis were used to evaluate the dynamic stresses caused by rotor-
stator interaction. CFD calculation for dynamic pressure and the harmonic response of the
6
runner is standard analysis in the industry. They created a finite element model of a runner
with boundary conditions for dynamic stress analysis.

The stochastic loads regime occurs during start-up, speed-no-load and transient events.
From a group of blade deformations procured by standard calculations, the one matching best
with the strain gauge measurements was selected. Geometrical extrapolation from the strain
gauges to the peak stress location was used to find the peak stress amplitude.

5 turbines were studied in regards of fatigue and life time assessment, with varying design
types and power categories.
Stochastic stresses dominate from speed-no-load to 40 percent of maximum power. As the
power increases above 40%, the fatigue load from the rotor-stator interaction gave the
dominating fatigue component until best efficiency point was reached.
Low load operation can have a huge impact on lifetime if the runner is operated in it over
extended periods of time and the runner hasn’t been specifically designed for it.
Each start-stop can equal the fatigue load of extended periods of operation on speed-no-load
or low part load and several years of operation on full load. Because of the high fatigue load
from start-stop there is a case for running the turbine on speed-no-load instead of stopping it
to prevent damage.

Static and Dynamic stress analyses of the prototype high head Francis runner based on
site measurement [2]

In 2014, X Huang, together with Oram and Sick takes a closer look at static and dynamic
stress in Static and Dynamic stress analyses of the prototype high head Francis runner based
on site measurement
Due to the increased demand of hydropower to operate outside of BEP there is a need to
analyze life time expectancy based on dynamic behavior analysis of Francis turbine runners.
There have been plenty of such analysis during the past years, but most of them have not been
validated by site measurement. Therefore, both site measurements and numerical analysis on
a high head Francis turbine were performed. The prototype had 28 guide vanes, 17 runner
blades, a head of 377 meters and the nominal rotational speed was 375 RPM. They placed
both pressure transducers and strain gauges on both the suction and pressure side of the runner
blades. The strain gauges were placed both in radial and tangential directions.

The sampling frequency used when recording the measurement data was 1613 Hz. They split
the measurements into 12 operating points.

Figure 1.4: Table of selected operating points used in Static and Dynamic stress analyses of the prototype high head
Francis runner based on site measurement. [2]

They found that the radially directed static stresses decreased with increasing power while the
tangentially directed slightly increased. Those located close to the band “showed an opposite
behavior.”

7
For dynamic stresses, rotor-stator interaction is the main contributor. The guide vane passing
frequency for this runner at 375 RPM is 175 Hz. They graphed the dynamic stresses at the
various operating conditions. Being relatively low until a steep ascent from OP4 to OP5 and
then a gradual decline from OP6 to overload.

Numerical simulations:
Stress analysis were conducted with the finite element method. They used computational fluid
dynamics to create pressure distributions for the different load cases. Then they compared the
measurement results with their simulation for static stresses on the strain gauge locations and
used linear regression to validate that their model was sufficiently accurate. The R2 values
were all above 0.890.

Thereafter locations and maximum static von Mises stresses were identified and graphed
versus power output.

A numerical modal analysis was carried out, finding the important natural frequencies of the
runner. One natural frequency that was very close to the guide vane passing frequency was
found. Because of this the dynamic stresses are very sensitive on the damping. The damping
values of OP5 and OP11 were adjusted so that the calculated stresses caused by RSI matched
the measured ones.

Dynamic loads in Francis runners and their impact on fatigue life [11]

In Dynamic loads in Francis runners and their impact on fatigue life, Seidel, Mende, Hubner,
Weber and Otto compares the damage contributions to a medium high-head Francis turbine
for two different operational modes. Base load and grid stabilization.

Operational Start up Speed no Low part Part load Around High load
mode [Cycles/day] load [%] load [%] [%] BEP [%] [%]
Base load 1 1 0 25 49 25
Grid 10 4 24 24 24 24
stabilization
Table 1.1: Assumed load universes for different operational modes of a medium high head Francis used in Dynamic loads in
Francis Runners and their impact on fatigue life[11]

The paper looks at the different flow patterns that characterizes the different operational
conditions. At runaway you have cavitation in the runner and draft tube as well as strong
secondary flow effects.

At speed-no-load there was a large backflow both in the draft tube and the runner creating
cavitation channel vortices. This results in high amplitude pressure fluctuations.

8
Base load

Grid
stabilization

Figure 1.5: Relative damage contributions of the load universes. From Francis used in Dynamic loads in
Francis Runners and their impact on fatigue life.[11]

The paper gives an insight in the difference in damage contributing factors which has to be
expected for a future where hydropower to a larger degree will be used for grid stabilization.

Mechanical robustness of Francis runners, requirements to reduce the risk of cracks in

blades [1]

In Mechanical robustness of Francis runners, requirements to reduce the risk of cracks in

blades, Bjørndal, Reynaud and Holo measured runner stresses for both low head and high
head prototype Francis runners with strain gauges as well as doing Finite Elements method
(FEM) analysis on a research model runner. The objective of this research was to improve
turbine contracts to “ensure the mechanical robustness of new runners”.

For the low head runner, operation at part load was the major source of large stress variations.
The conference paper assumes that this for the most part is related to hydraulic inter-blade
vortexes in the runner. This was solved for this turbine by restricting operation at part load to
only uploading and downloading.

For the high head runners, the guide vane passing frequency was the main contributor to
stress variations. The amplitudes of the guide vane passing frequency were found to be the
most significant between 50 and 60% opening. For fatigue loading, operating at maximum
load gave high stress variations and high mean tensile stress, which is unfavorable.

Start-up scenarios was also specifically investigated. While start-up gives large variation in
turbine stresses, the problem is usually negligible when evaluating runner fatigue lifetime
because of the limited amount of cycles.

With strain gauge measurements they found out that the startup procedure can significantly
affect the runner life expectancy. A rainflow analysis of the startup procedure gave 75 load
cycles with amplitudes over 100 MPa. Which results in 1.4*106 load cycles over 50 years
with one start per day. Reducing the guide vane opening during the startup procedure reduced
the dynamic loads on the runner.

FEM analysis found that thicker runner blade outlet will increase the runner fatigue lifetime
and reduce the risk of blade cracking.

9
Bjørndal et al propose a temporary solution to ensure the mechanical strength of new runners.
That the manufacturer should “document the expected dynamic load on the runner based on
advanced studies and measurements”.

1.3.2. Strain gauge measurements

[13] In 2016 Einar Agnalt did pressure measurements inside the Francis turbine runner in the
waterpower laboratory at the Norwegian University for Science and Technology.

The experiments were done with a splitter blade runner design with 15+15 blades and 28
guide vanes. During the experiments semiconducturbased strain gauges were used on one
runner blade as well as five pressure sensors mounted in the hub. The pressure sensors were
Kulite XTE-190(m) with a range of 0-3.5 bar and the strain gauge was Kulite S/UDP-350-
175. The strain gauge was set up in a quarter Wheatstone bridge configuration, which gives
no temperature compensation. There was a zero point drift during the experiment, so the
results from the strain gauge mostly only provided indications of correlation between strain in
the blade and pressure pulsations, but not precise and reliable strain data.

1.3.3. Flexible turbines

[4] In Variable-speed Operation and Pressure Pulsations in a Francis Turbine and a Pump-
Turbine(2018), Iliev, Trivedi, Agnalt and Dahlhaug experimented with variable-speed
operation of the runner blades to reduce the amplitudes of pressure pulsations. They
conducted the experiments with two different turbine runners. One splitter-bladed Francis
turbine and one reversible-pump turbine.
The experiments were conducted with three scenarios, all assuming constant net head in the
entire range of the discharge. The three scenarios were synchronous-speed operation at
optimal head, +10% of the optimal head and -10% of the optimal head.
For the Francis turbine, variable-speed operation achieved 0,5% gain in maximum hydraulic
efficiency compared to synchronous-speed operation at optimal head and up to 1,2% at 10%
of the optimal head. Pressure pulsations were also affected by the variable-speed operation,
with a reduction in the vaneless space and a none-to-slight increase in the draft tube cone.

10
 1.4. Analysis method.
Histogram – peak to peak values

When measuring pressure pulsations, the important information is the size and frequency of
the pulsations. The size is determined by the peak to peak value of each measured series. A
histogram can be applied to remove statistically unlikely values by only using the data within
a confidence interval, usually between 95 and 99 percent. The standard IEC60193
recommends a confidence interval of 97 percent. The pressure pulsations are measured
together with the operating conditions. This makes it possible to accurately map the pressure
pulsations in different operating points.

Figure 1.6: Measurement series shown in the time domain. With time in seconds along the x-axis and volts recorded along
the y-axis

Figure 1.7: The same measurement series as shown in figure 1.6 displayed as a histogram. With volts along the x-axis and
number of measured values along the y-axis. The red lines show the limits of a 97% confidence interval.

Figure 1.8: Measurement series shown in the time domain. With time in seconds along the x-axis and volts recorded along
the y-axis The red lines show the limits of a 97% confidence interval.

11
Interpolation and number of measurements

The accuracy of hill diagrams for hydraulic efficiency and pressure pulsations is determined
by the number of operating points measured, the uncertainties in measurement and the method
of interpolation and extrapolation. For hill diagrams a shape-preserving interpolation method
gives the most accurate results when interpolating along the guide vane angles.

Sampling rate

The data acquired during measurements is a number of analog values from an analog signal.
They are then converted to a digital signal. This is beneficial because it allows for storing a
vast amount of information and makes it easier to sort it and use it for practical purposes
afterwards. The analog signal has infinite resolution, however, when sampling the signal, it is
not possible to maintain an infinite resolution, as it would require an infinitely long number to
represent an infinite resolution. The resolution of the sampled signal is determined by the
number of discrete values that can be represented inside the range of the analog value.
That number is determined by the number of bits used to store the digital signal. [21]

The minimum change that results in a change in the digital input is formulated by the least
significant bit.
𝐸𝐹𝑆𝑅
𝑄= (V) [14]
2𝑀

Where Q is the minimum change, 𝐸𝐹𝑆𝑅 is the range of the analog signal and M is the number
of bits.

The sampling rate of the measurements is defined by the number of analog values stored per
second. Since the recorded values are discrete and taken from a continuous signal, it is
important to have a high enough sampling rate to imitate the continuous signal accurately.

The Nyquist-Shannon theorem [28] states that the sample rate should at least be twice the size
of the highest significant frequency. If the frequency isn’t high enough it can lead to aliasing.
Since the signals is digitally reconstructed as sinusoids, the frequency of each cycle of
sinusoids can be misrepresented if the sampling rate is too low, which is called aliasing.
Which is when higher frequencies are misidentified as lower frequencies in the digitally

12
reconstructed signal due to too low sampling rates.

Figure 1.9: Example of how a higher frequency sinusoid can be misrepresented as lower frequency signal if the sampling
rate is too low.

If the sample rate is too low, discrete values from the orange signal can be misrepresented as
2𝑇
the blue signal, since they overlap every 5 periods.

The analog values can always be represented as a higher frequency section of sinusoids, there
is no upper limit.

Oversampling can be used to minimize the effects of aliasing. By having a sampling rate far
above what’s required from the Nyquist-Shannon theorem. Oversampling can significantly
increase the size of the data stored, but is common to do because it also increases the
resolution and reduces the effect of noise in the measurements.

Fast Fourier transform is used to look at the impact of the different frequencies within the
signal. The fast Fourier transform breaks down the sinusoids within the signal by converting it
from the time domain to the frequency domain. An example of a fast Fourier transform of a
signal from the time domain to the frequency domain is shown below.

Figure 2.10:Measurement series shown in the time domain. With time in seconds along the x-axis and volts recorded along
the y-axis

13
Figure 1.11: Same measurement series as shown in figure 1.10, presented in the frequency domain after a Fast Fourier
Transform. Frequencies in Hz along the x-axis and signal values along the y-axis.

In this thesis the fast Fourier transform is done using Matlab R2018a. Matlab uses the
following functions for FFT. Where Y is the output signal, X is the input signal and n is the
length of the signal. [22]

(𝑗−1)(𝑘−1)
𝑌(𝑘) = ∑𝑛𝑗=𝑖 𝑋(𝑗) ∗ 𝑊𝑛 (-) [15]

−2𝜋∗𝑖
𝑊𝑛 = 𝑒 𝑛 (-) [16]

The frequency signal can be used to determine the sources of the pressure pulsations. It is
clear in the example above that a pressure pulsation with a bit less than 150 in frequency is
the main source. Because the frequency is slightly less than 150 and the runner speed is 297
RPM, we can determine that the main propagator of pressure pulsations at this operating point
297
is the Blade Passing Frequency: 𝐹𝑟𝑏 = 𝑍𝑟 ∗ 𝐹𝑛 = 30 ∗ 60 = 148.5

Welch method:

The Welch method separates the measurement series into several time-based segments. [30]
Each of these segments are modified by a window function. This is done because the FFT
assumes the signal to be periodic, and can be perfectly represented by an infinite number of
sinusoids. Because this assumption does not hold true for actual measurement series, it causes
errors in the transformation from time-based signal to frequency-based signal. These errors
are called spectral leakage, and results in an intensity distortion of the peaks of the frequency
signal.
The Welch method of time averaging over short modified periodograms aims to minimize
spectral leakage by making the signal periodic in the time domain. The modification of the
segmented windows does however remove information from the original signal. To counter
this the windows are overlapped.

IEC60193[16] recommends using Hann window when analyzing pressure fluctuations. In this
thesis a Hann window with 50 percent overlap is used. The function of the Hann window is
shown below as well as the frequency specter of the same measurement series as used before.
[29]

14
 1 𝑛
𝑤(𝑛) = 2 (1 − cos (2𝜋 𝑁)) , 0 ≤ 𝑛 ≤ 𝑁 (-) [17]

Figure 1.12: Same measurement series as shown in figure 1.10, presented in the frequency domain after a Fast Fourier
Transform with the Welch method. Hann windowing used. Frequencies in Hz along the x-axis and signal values along the y-
axis.

In the figure above the blade passing frequency is very easy to identify. Frequencies at around
75 Hz and 300 Hz can also be identified. The one around 75 Hz is because the runner blades
are splitter blades, which means two different sets of 15 blades, resulting in a splitter blade
passing frequency:
𝑅𝑃𝑀
𝐹𝑟𝑠𝑏 = 𝑍𝑠𝑏 ∗ 𝐹𝑛 = 15 ∗ 60 = 74.25 (Hz) [18]

Where 𝑍𝑠𝑏 is the number of splitter blades.

The frequency observed at around 300 Hz is due to harmonics. Which can be observed at
every integer of each high-amplitude frequency, usually with a rapidly declining amplitude.
An amplitude can also be seen closer to zero Hz. This is the Rheingans Frequency, which for
297 297
60 60
297 RPM would be between = 1.375 and = 1.65.
3,6 3

Because of the window length used is 5000 measurement values and the sample rate is 5000
Hz, a low frequency amplitude like the Rheingans Frequency will only be included once or
twice per window, resulting in lower precision and a smearing of the results in the frequency
spectrum.

15
2. Hypothesis
The objective of this thesis is to Evaluate how variable speed operation of Francis turbines
can be utilized to minimize dynamic loads at start-stop scenarios.

Research has shown that start/stop cycles on Francis turbines significantly damages the
runners and causes fatigue and eventually failure. [2][11] The main cause of these damages is
pressure pulsations caused by RSI. To evaluate how variable speed operation can minimize
dynamic loads, the amplitudes of the pressure pulsations occurring during different speed
start-ups and synchronization has to be mapped.

Ideally strain gauges would be used to evaluate the strain on the runner blade. However, due
to the complex surface of the runner blades, calibrating and getting absolute results from
strain gauges was not possible at the time of this thesis. Strain gauges could have been used
on the runner blade, but would only provide relative values and therefore would not give
information that evaluating the pressure regime with pressure transducers doesn’t already
provide.

Since calibrated strain gauges could give valuable information about fatigue on the runner
blades this thesis also attempts to calibrate strain gauges on a runner blade by comparing
strain results from an FE simulation with results from a strain gauge on a runner blade.

As Illiev, Trivedi, Agnalt and Dahlhaug has shown in Variable-speed operation and pressure
pulsations in a Francis turbine and a pump-turbine, it is possible to reduce pressure pulsations
by operating with variable speed operation. [4]

This thesis will mostly focus on measuring pressure and pressure pulsation in the vaneless
space between the guide vanes and the runner blades. Previous studies will be used to
correlate these pressures with stress on the runner blade and a fatigue analysis will be
conducted based on that.

2.1. Formulated Hypothesis

It is possible to significantly reduce the accumulated damage from start-stop cycles by
operating with a variable speed turbine.

16
17
3. Laboratory setup
The experiments done during this thesis were conducted at The Waterpower Laboratory at
NTNU. The facilities have test rigs for Francis, Pelton and pump turbines. In this thesis the
Francis rig was used, which enables model test according to the IEC 60193 and IEC 60041
standards. [16] [17]

3.1. The Francis Rig

The Waterpower Laboratory at NTNU has an operational Francis turbine test rig for model
testing. The system is set up so that it can be run in both an open loop and a closed loop
configuration by regulating which valves are open. With an open loop configuration, which is
what is used during this thesis, the Francis test rig has an available head of 16 meters. [18]

The Francis turbine is a scaled model of a 107.5 MW turbine used at Tokke power plant. It
has a splitter blade runner with 15 splitter blades and 15 full length blades. It has 28 guide
vanes, 14 stay vanes and 30 runner vanes. The inlet diameter is 0.63 meter and the outlet
diameter is 0.349 meter.

3.1.1. Components
The Francis test rig consists of the following components.
1. Basement reservoir
2. Pumps
3. Piping
4. Upper reservoir
5. Two over-head tanks
6. A high pressure tank
7. Generator
8. Turbine
9. Draft tube
10. Draft tube tank

3.1.2. Open loop configuration

In an open loop configuration, water is pumped from the basement reservoir to the upper
reservoir. From there is flows to the pressure tank and then to the turbine. After going through
the turbine, the water goes through a draft tube and into a draft tube tank, which simulates a
lower reservoir. After the draft tube tank, the water goes back into the basement reservoir.

18
Figure 3.1: Open loop configuration of the Francis Turbine in the waterpower laboratory at NTNU.

3.2. Instrumentation

Figure 3.2: Illustration of the Francis turbine setup in the Waterpower Laboratory. Figure showing from left to right: The
high pressure tank, inlet pipe to the turbine, turbine, generator, draft tube and draft tube tank.

19
 Operating condition sensors
Sensor placement Sensor name Description
Inlet pipe to turbine FTQ1 Measuring flow rate in the
inlet pipe to the turbine
Inlet pipe to turbine PT PIN Measuring pressure in the
inlet to the turbine
Inlet pipe to turbine and PT DP Measuring differential
draft tube pressure between inlet and
outlet of turbine
Generator hydrostatic WT T1 Measuring generator torque
bearing
Thrust bearing WT T2 Measuring friction torque
Thrust bearing WT T3 Measuring axial load
Guide vane shaft ZT 42 Measuring guide vane
position
Inlet pipe to turbine TT 41 Measuring temperature
Pressure transducers
Between guide vanes and GV4 Measuring pressure in the
runner, position 4. vaneless space
Between guide vanes and GV5 Measuring pressure in the
runner, position 5. vaneless space
Between guide vanes and GV6 Measuring pressure in the
runner, position 6. vaneless space
Table 3.1: Operating condition sensors and pressure transducers. Placement, name and description.

Figure 3.3: Placement of pressure transducers. Placed flush to the surface

between the guide vanes and the runner blades.

20
 3.3. Calibration and uncertainty
To ensure the validity of the results of the experiments the equipment has to be calibrated
first. Both the equipment used for direct measurements and the equipment measuring the
operating conditions during the experiment should be calibrated for accurate measurements.
Because some of the equipment used has been calibrated recently and to decrease the scope of
calibration work some of the equipment used was not calibrated right before the experiments.
However, all equipment used for operating the Francis Rig have been previously calibrated to
match the IEC 60193 Standard.

The uncertainty for operational sensors except for WT T1 and WT T2 are gathered from a
measurement report done by Einar Agnalt in 2018. [25]
Calibration reports for friction torque and generator torque can be found in appendix C.

Sensor Unit Expanded Expanded Expanded Total expanded

name calibrated long time measurement uncertainty
uncertainty stability repeatability In units Percentage
FTQ1 [m3/s] 0.3*10 -3
0.1*10 -4
1.1*10-3 0.001 0.56%
PT PIN [kPa] 0.2 0.5 1.2 1.32 0.96%
PT DP [kPa] 0.02 0.22 0.16 0.27 0.24%
ZT 42 [⸰] 0.1 0.1 0.14 1.41%
AT 41 [⸰C] 0.1 0.3 0.30 2.00%
Table 3.2: Uncertainties for operational sensors. Results from a measurement report done by Einar Agnalt[25].

Sensor Unit Expanded Total expanded

name calibrated uncertainty
uncertainty In units Percentage
WT T1 [Nm] 0.6862 0.6862 0.2486%
WT T2 [Nm] 0.0910 0.0910 *
Table 3.3: Uncertainties for operational sensors.

*WT T2 was calibrated to a zero value, so the percentage is infinite.

3.3.1. Pressure transducers

The static calibration of the pressure sensors have done in accordance to the procedure by the
German Calibration Service[12]. Which is a procedure designed to fulfill the requirements of
IEC 17025[19], which is the main IEC standard used by calibration laboratories.
This procedure was used instead of how it has previously been done at The Waterpower
Laboratory because it is designed to mitigate human errors. Unlike the previously used
calibration sequences it also places a larger significance on the whole measurement range,
which ensures a larger certainability that the calibration curve is linear.

21
The equipment used for calibration was a dead weight manometer and air as the medium.
The pressure sensors were connected through a cDAQ system to a computer running a
calibration program made with LabView. The cDAQ system consisted of a NI 9237
simultaneous bridge module connected to a NI cDAQ-9178 chassis. [23][24]

A minor drawback to using this procedure is that it takes longer to do the calibration, which
provided some problems because the atmospheric pressure in the locale where the calibration
was done was not constant. During one of the calibration processes the change in the
atmospheric pressure was recorded to be over 2 millibar. To counter this the atmospheric
pressure was checked before each measurement was taken in the final calibration series.

Calibration Model name Range Total expanded uncertainty

Equipment In units Percentage
Dead weight GE 30-2000 0.16 mbar 0,008 %
manometer deadweight mbar
tester, P3023-
6-P
Pressure sensors Total expanded uncertainty
Sensor Model Unit Expanded In units Percentage
name name calibrated
uncertainty
GV4 MEAS [kPa] 0.06630 0.06630 0.06383 %
XP5
GV5 Kulite [kPa] 0.02542 0.02542 0.02447 %
XTE
GV6 Kulite [kPa] 0.02385 0.02385 0.02296 %
XTE
Table 3.4: Uncertainties for pressure transducers.

3.3.2. Strain gauges

Due to difficulty with installing strain gauges onto the runner blades in the turbine and getting
precise and accurate absolute measurement results from the strain gauges this thesis instead
focuses on creating a procedure for calibration strain gauges using the finite element method.

This was done by first creating a model for measuring the strain on a runner blade with
ANSYS mechanical. The model is built up by a runner blade held up by two fixed supports on
at one end. The forces working on the runner blade consisting of the constant gravity as well
as a remote force close to the edge of the blade.

This was recreated with a physical model. The forces were applied by gluing a nut close to the
edge of the blade and then fastening weights to it.

The strain on the blade is measured with seven different sizes of remote forces. Ranging from
0.122 N to 85.0 N.

22
The weights added to the physical model ranged from 0.125 Kg to 8.66 Kg.

Due to the complex surface of the model, there were only two flat edges suitable for mounting
the blade unto a stable surface. This put a limit on the direction of the force added to the
blade.

Figure 3.4:Ansys mechanical model of a runner blade.

Two strain gauges were placed on the runner blade in a half bridge configuration. The
position of the strain gauge on the top side of the blade is illustrated in the picture above.

The results from the ANSYS mechanical model was used to calibrate the strain gauges. The
average normal elastic strain along the direction of the strain gauges from the areas where the
strain gauges were placed was graphed with the volt signals gained from the physical model.

Figure 3.5: ANSYS mechanical model of the runner blade. Results for the whole blade showing equivalent elastic strain.

23
 Calibrated strain curve - Linear
0,0003

0,000295 y = 4,6604x + 0,0003

R² = 0,9961
0,00029
Volt

0,000285

0,00028

0,000275

0,00027
0,00E+00 1,00E-06 2,00E-06 3,00E-06 4,00E-06 5,00E-06
Strain

Figure 3.6: Calibrated strain curve, With Strain along the x-axis and voltage along the y-axis. Fitted with a linear regression
curve.

The results show that the directional strain doesn’t follow a linear curve with the volt signal
from the strain gauges.

Calibrated strain curve - Second degree polynom

0,0003

0,000295 y = -219819x2 + 5,697x + 0,0003

R² = 0,9998
0,00029
Volt

0,000285

0,00028

0,000275

0,00027
0,00E+00 1,00E-06 2,00E-06 3,00E-06 4,00E-06 5,00E-06
Strain

Figure 3.7: Calibrated strain curve, With Strain along the x-axis and voltage along the y-axis. Fitted with a second degree
polynomial regression curve.

A second degree polynomial resulted in a much better fitting curve.

It should be noted that this procedure has several uncertainties and doesn’t calibrate the strain
gauges for the dynamic pressure fluctuations the runner blades experiences during operation.
It also does not take into account temperature changes.

24
 3.4. Data acquisition and processing

3.4.1. Data acquisition.

The pressure pulsations are measured with three pressure transducers connected to a cDAQ
system. The cDAQ system consists of a NI 9237 simultaneous bridge module connected to a
NI cDAQ-9178 chassis. The NI 9327 simultaneous bridge module has a resolution of 24 bits
and a sample rate of 50 kHz/s/Ch. [23][24]

3.4.2. Sample rate and data processing

The sample rate used for the measurements was 5000Hz. With the largest expected frequency
being the runner blade frequency:
𝑛𝑚𝑎𝑥 530
𝐹𝑟𝑏 = 𝑍𝑟 ∗ 𝐹𝑛 . With 𝐹𝑛𝑚𝑎𝑥 = = and 𝑍𝑟 = 30. 𝐹𝑟𝑏𝑚𝑎𝑥 = 265
60 60

The Nyquist-Shannon theorem requires the sample rate to be at least 530 Hz, so during the
measurements oversampling with a factor of 9.43 to the Nyquist sample rate was used. [28]

The measurement series were afterwards imported to Matlab together with the operating
conditions and analyzed using the histogram method to evaluate the pressure pulsations.
Spectral analysis of the signals was also conducted with Matlab, using Hann-windows.

The length of most measurement series was 40 seconds. A few were also 30 or 60 seconds.
The length of each window segment was 5000 measurement points and the overlap used was
50 percent.

To histogram method was used to measure the size of the pressure pulsations. A 97 percent
confidence interval vas used. The values presented in the results phase of this thesis is the
amplitude of the 97 percent confidence interval of the signal, which is the peak to peak value
divided by 2.

25
26
4. Results
The experimental results from the pressure pulsation measurements are presented in various
forms. 254 operating points were measured along 12 guide vane angles. In addition, 15
separate operation points were measured at speed no load. The results are presented below in
the form of pressure pulsation diagrams, pressure pulsations along constant guide vane angles
and frequency analysis of interesting operating conditions. The head of the model varied
slightly during the experiments, from 12,3 to 12,6 meters.

Along the constant guide vane angles, both a reduction and increase in speed has been looked
at to estimate how flexible generator operations can reduce pressure pulsations.

The Best efficiency point was found to be at 338,9 RPM, NED=0.1785 and QED=0.1526, with
a hydraulic efficiency of 93.90%. However, the flowmeter had not been calibrated prior to
doing the measurements, so the hydraulic efficiency is probably lower than what is found,
which would be more consistent with earlier efficiency testing done on the Francis turbine in
the Waterpower Laboratory.

Figure 4.1: Hill chart with efficiency-lines. X-axis NED from 0.12 to 0.28. Y-axis QED from 0.02 to 0.205.Shows hydraulic
efficiency at different operating points.

27
 4.1. Pressure pulsation diagrams
The pressure pulsation amplitudes at BEP for the pressure sensors placed at GV4, GV5 and
GV6 was 1,6313 KPa, 1,8297 KPa and 1.6660 KPa. The pressure pulsations diagrams
presented shows the pressure amplitude values relative to the pressure amplitude values found
at BEP.

The three pressure diagrams show varying consternation of pressure pulsation amplitudes at
different operating conditions. While all three sensors record high pressure pulsations when
approaching speed-no-load, the scenario between 0.12 and 0.22 NED is hugely variable.

4.1.1. GV4
GV4 has generally low amplitude pressure pulsations from NED 0.12 to 0.19 below 6 degrees’
guide vane opening. As well as a high amplitude at low speed at 8 and 9 degrees’ guide vane
opening.

Figure 4.2: Pressure pulsation diagram for GV4. X-axis Dimensionless Speed Factor NED. Y-axis Dimensionless Discharge
Factor QED. Shows relative pressure amplitudes compared to pressure amplitudes at BEP.

4.1.2. GV5
The pressure pulsation amplitudes for GV5 appears to be more dependent on runner speed
than guide vane opening, except at very low turbine speed where the pressure pulsation rise
rapidly from 1 degree to 3 degrees’ guide vane opening. GV5 has generally low pressure
pulsations around synchronous speed and along low guide vane angles.

28
 Figure 4.3: Pressure pulsation diagram for GV5. X-axis Dimensionless Speed Factor NED. Y-axis Dimensionless Discharge
Factor QED. Shows relative pressure amplitudes compared to pressure amplitudes at BEP.

4.1.3. GV6
GV6 has low pressure pulsation amplitudes between 0.14 and 0.16 NED up until guide vane
angle at 8 degrees. Rather high pressure pulsation amplitudes above 0.21 NED and below 0.13
NED.

Figure 4.4: Pressure pulsation diagram for GV6. X-axis NED. Y-axis QED. Show relative pressure amplitudes compared to
pressure amplitudes at BEP.

29
 4.2. Pressure pulsations along constant guide vane
angles

Pressure pulsation amplitudes at 2 degrees’ guide vane openings is generally low up until 380
RPM, when pressure pulsation for all three sensors starts rising rapidly. Very similar pressure
pulsations for all sensors in the vaneless space at this guide vane angle.

Figure 4.5: Relative pressure pulsations to BEP along constant guide vane at 2 degrees. From 228 to 440 RPM.

At 6 degree’s guide vane opening there are a lot more difference in pressure pulsations
recorded by the sensors. GV4 never goes above 1.6 in normalized pressure pulsations, and
rises up only a bit when the RPM is very low or high. GV5 is very sensitive to low speed,
going up to 2.2 at 228 RPM. Relatively low from 300 to 380 RPM and then rising slightly as
the RPM increases to 450. GV6 also experiences high pressure pulsations at really low
speeds, but goes down very fast until 280 RPM. Has a very low zone between 280 and 300
RPM before a gradual upwards incline to 370 RPM. After 370 RPM, the pressure pulsations
for GV6 rises quite fast from 1.1 in normalized frequency up to 2.3.

Figure 4.6: Relative pressure pulsations to BEP along constant guide vane at 6 degrees. From 228 to 450 RPM.

30
Tables showing the results along the constant guide vane angles are presented below. The
pressure pulsations amplitudes at synchronous speed ± 10%, 20% and 30% were compared to
the pressure pulsations at synchronous speed. This was done for all three sensors as and for
their average values combined. The largest reduction is presented as well, compared to the
pressure pulsation at synchronous speed and the actual pressure pulsations at BEP.

Pressure pulsations at 1 degree guide vane opening

Pressure pulsation amplitudes from 30% reduced speed to 30% Largest
Sensor increased speed [KPa] reduction as
to [%]
237.2 271.2 305.0 338,9 372.8 406.7 440.6 BEP 338,9
RPM RPM RPM RPM RPM RPM RPM RPM
Beyond Beyond 29,63 23,69
GV4 1,193 1,148 1,308 1,504 1,950 Runaway Runaway
Beyond Beyond 29,03 0,68
GV5 1,394 1,291 1,308 1,300 1,529 Runaway Runaway
Beyond Beyond 40,87 20,64
GV6 1,040 0,985 1,103 1,241 1,482 Runaway Runaway
Combined Beyond Beyond 33,08 15,36
3,627 3,424 3,719 4,046 4,961 Runaway Runaway
Table 4.1: Pressure pulsation at 1 degree guide vane opening.

Pressure pulsations at 2 degrees guide vane opening

Pressure pulsation amplitudes from 30% reduced speed to Largest
Sensor 30% increased speed [KPa] reduction as to
[%]
237.2 271.2 305.0 338,9 372.8 406.7 440.6 BEP 338,9
RPM RPM RPM RPM RPM RPM RPM RPM
Beyond 28,71 18,17
GV4 1,317 1,163 1,287 1,421 1,719 3,040 Runaway
Beyond 22,88 -1,95
GV5 1,877 1,695 1,524 1,403 1,431 2,391 Runaway
Beyond 36,69 16,69
GV6 1,373 1,074 1,055 1,266 1,406 2,281 Runaway
Combined Beyond 24,47 5,52
4,568 3,932 3,865 4,091 4,555 7,712 Runaway
Table 4.2: Pressure pulsation at 2 degrees’ guide vane opening.

31
 Pressure pulsations at 3 degrees guide vane opening
Pressure pulsation amplitudes from 30% reduced speed to Largest
Sensor 30% increased speed [KPa] reduction as to
[%]
237.2 271.2 305.0 338,9 372.8 406.7 440.6 BEP 338,9
RPM RPM RPM RPM RPM RPM RPM RPM
GV4 1,478 1,391 1,435 1,462 1,662 2,360 5,246 14,72 4,85
GV5 2,634 2,162 1,741 1,465 1,477 2,088 4,307 19,50 -0,85
GV6 1,838 1,142 1,061 1,260 1,427 2,257 4,271 36,31 15,78
Combined 5,950 4,695 4,237 4,187 5,322 6,705 13,825 18,18 -1,20
Table 4.3: Pressure pulsation at 3 degrees’ guide vane opening.

Pressure pulsations at 4 degrees guide vane opening

Pressure pulsation amplitudes from 30% reduced speed to 30% Largest
Sensor increased speed [KPa] reduction as
to [%]
237.2 271.2 305.0 338,9 372.8 406.7 440.6 BEP 338,9
RPM RPM RPM RPM RPM RPM RPM RPM
GV4 1,560 1,413 1,555 1,699 1,762 1,861 2,525 13,38 16,82
GV5 3,264 2,496 1,947 1,927 1,577 1,859 2,518 13,35 18,16
GV6 2,036 1,166 1,183 1,696 1,531 2,244 3,114 30,03 31,26
Combined 6,861 5,075 4,684 5,322 4,869 5,963 8,157 8,46 11,98
Table 4.4: Pressure pulsation at 4 degrees’ guide vane opening.

Pressure pulsations at 5 degrees guide vane opening

Pressure pulsation amplitudes from 30% reduced speed to 30% Largest
Sensor increased speed [KPa] reduction as
to [%]
237.2 271.2 305.0 338,9 372.8 406.7 440.6 BEP 338,9
RPM RPM RPM RPM RPM RPM RPM RPM
GV4 1,720 1,463 1,534 1,556 1,951 1,921 2,390 10,30 5,94
GV5 3,575 2,640 1,933 1,827 1,909 1,984 2,609 -0,41 -4,49
GV6 2,252 1,306 1,158 1,574 1,979 2,465 3,449 30,49 26,41
Combined 7,547 5,409 4,625 4,956 5,839 5,839 6,370 9,62 6,69
Table 4.5: Pressure pulsation at 5 degrees’ guide vane opening.

Pressure pulsations at 6 degrees guide vane opening

Pressure pulsation amplitudes from 30% reduced speed to 30% Largest
Sensor increased speed [KPa] reduction as
to [%]
237.2 271.2 305.0 338,9 372.8 406.7 440.6 BEP 338,9
RPM RPM RPM RPM RPM RPM RPM RPM
GV4 2,148 1,733 1,712 1,596 1,823 2,161 2,334 2,14 -7,22
GV5 3,695 2,627 1,905 1,753 1,661 2,243 2,757 8,73 5,24
GV6 2,497 1,414 1,122 1,468 1,858 2,857 3,718 32,65 23,57
Combined 8,340 5,774 4,738 4,817 5,342 7,261 8,809 7,40 1,63
Table 4.6: Pressure pulsation at 6 degrees’ guide vane opening.

32
 Pressure pulsations at 7 degrees guide vane opening
Pressure pulsation amplitudes from 30% reduced speed to Largest
Sensor 30% increased speed [KPa] reduction as to
[%]
237.2 271.2 305.0 338,9 372.8 406.7 440.6 BEP 338,9
RPM RPM RPM RPM RPM RPM RPM RPM
GV4 2,821 2,208 2,026 1,867 1,986 2,256 2,382 -14,44 -6,37
GV5 3,860 2,646 1,888 1,754 1,630 2,213 2,909 10,41 7,04
GV6 2,722 1,514 1,158 1,469 1,885 2,948 3,990 30,48 21,15
Combined 9,403 6,367 5,073 5,090 5,501 7,416 9,280 0,86 0,33
Table 4.7: Pressure pulsation at 7 degrees’ guide vane opening.

Pressure pulsations at 8 degrees guide vane opening

Sensor Pressure pulsation amplitudes from 30% reduced speed to Largest
30% increased speed [KPa] reduction as
to [%]
237.2 271.2 305.0 338,9 372.8 406.7 440.6 BEP 338,9
RPM RPM RPM RPM RPM RPM RPM RPM
GV4 3,465 2,649 2,313 2,073 2,068 2,216 2,536 -26,75 0,24
GV5 4,053 2,811 1,999 1,785 1,545 2,082 3,216 15,07 13,42
GV6 3,145 1,687 1,261 1,526 1,905 2,942 4,402 24,28 17,36
Combined 10,663 7,147 5,574 5,384 5,518 7,239 10,154 -5,22 -2,48
Table 4.8: Pressure pulsation at 8 degrees’ guide vane opening.

Pressure pulsations at 9 degrees guide vane opening

Sensor Pressure pulsation amplitudes from 30% reduced speed to Largest
30% increased speed [KPa] reduction as
to [%]
237.2 271.2 305.0 338,9 372.8 406.7 440.6 BEP 338,9
RPM RPM RPM RPM RPM RPM RPM RPM
GV4 3,235 2,636 2,327 2,083 2,052 2,159 2,596 -25,80 1,50
GV5 4,096 2,902 2,136 1,827 1,556 2,113 3,324 14,51 14,86
GV6 3,076 1,865 1,453 1,618 2,027 3,101 4,580 12,80 10,19
Combined 10,407 7,403 5,916 5,528 5,634 7,373 10,500 -8,03 -1,92
Table 4.9: Pressure pulsation at 9 degrees’ guide vane opening.

33
 Pressure pulsations at 10 degrees guide vane opening
Sensor Pressure pulsation amplitudes from 30% reduced speed to 30% Largest
increased speed [KPa] reduction as
to [%]
237.2 271.2 305.0 338,9 372.8 406.7 440.6 BEP 338,9
RPM RPM RPM RPM RPM RPM RPM RPM
GV4 2,364 1,922 1,800 1,631 1,775 2,021 2,687 0,00 -8,82
GV5 4,166 2,956 2,296 1,820 1,648 2,164 3,142 9,46 9,46
GV6 3,212 1,971 1,572 1,666 2,122 3,250 4,817 5,63 5,63
Combined 9,742 6,849 5,669 5,117 5,545 7,435 10,645 0,00 -8,36
Table 4.10: Pressure pulsation at 10 degrees’ guide vane opening.

Pressure pulsations at 11 degrees guide vane opening

Sensor Pressure pulsation amplitudes from 30% reduced speed to Largest
30% increased speed [KPa] reduction as
to [%]
237.2 271.2 305.0 338,9 372.8 406.7 440.6 BEP 338,9
RPM RPM RPM RPM RPM RPM RPM RPM
GV4 2,525 2,225 1,948 1,764 1,887 2,194 2,718 -6,982 -8,12
GV5 4,311 3,099 2,383 1,872 1,555 2,149 3,148 16,919 14,55
GV6 3,284 2,239 1,816 1,778 2,172 3,453 4,951 -2,102 -6,74
Combined 10,120 7,563 6,147 5,414 5,614 7,797 10,817 -5,80 -3,71
Table 4.11: Pressure pulsation at 11 degrees’ guide vane opening.

Pressure pulsations at 12 degrees guide vane opening

Sensor Pressure pulsation amplitudes from 30% reduced speed to 30% Largest
increased speed [KPa] reduction as
to [%]
237.2 271.2 305.0 338,9 372.8 406.7 440.6 BEP 338,9
RPM RPM RPM RPM RPM RPM RPM RPM
GV4 2,514 2,073 1,846 1,787 2,031 2,474 3,088 -3,34 -9,52
GV5 4,361 3,295 2,572 1,905 1,722 2,269 3,193 9,59 5,38
GV6 3,318 2,507 2,010 1,921 2,494 3,874 5,416 -4,62 -15,32
Combined 10,193 7,875 6,428 5,612 6,247 8,617 11,697 -9,68 -11,31

Table 4.12: Pressure pulsation at 12 degrees’ guide vane opening.

34
 4.3. Frequencies in different operating regimes
Observing which frequencies that impacts the different pressure gives valuable information
about what are the sources of the pressure pulsations. In the figures below frequency analysis
𝑓 𝑓
of different signals are shown with normalized frequency 𝑓 = 𝑛
𝑛 60
The size of the amplitudes is shown as pressure in KPa, it should be noted that these values
does not represent the amplitudes of the actual pressure pulsations occurring.

4.3.1. BEP
Noticeable normalized frequencies are the Rheingans Frequency at around 0.3, the splitter
blade passing frequency at 15 and the blade passing frequency at 30.

Figure 4.7: Frequencies observed at BEP. Showing values for GV4, GV5 and GV6 for normalized frequencies from 0 to 70.

4.3.2. 2 degrees’ guide vane opening

Noticeable normalized frequencies are Rheingans, splitter blade passing and blade passing
frequency. The main difference between the three sensors is the magnitude of the blade
passing frequency. The amplitudes seen at 60 are likely harmonics of the blade passing
frequency. There are also small inclines at around 3 and 11.

Figure 4.8: Frequencies observed at 2 degrees guide vane opening and 272,4 RPM. Showing values for GV4, GV5 and GV6
for normalized frequencies from 0 to 70.

35
As the runner speed increase the blade passing frequency grows more significant for GV4.
Alongside Rheingans, blade passing and the splitter blade passing frequency there is also a
small incline around 3 for all sensors.

Figure 4.9: Frequencies observed at 2 degrees guide vane opening and 339 RPM.. Showing values for GV4, GV5 and GV6
for normalized frequencies from 0 to 70.

More stochastic pulsation showing along the frequencies as the runner is nearing speed-no-
load. Rheingans, blade passing and splitter blade passing frequency still clearly most
significant.

Figure 4.10: Frequencies observed at 2 degrees guide vane opening and 408,1 RPM.. Showing values for GV4, GV5 and
GV6 for normalized frequencies from 0 to 70.

36
 4.3.3. 5 degrees’ guide vane opening

Blade passing frequency dominating for GV5. Far smaller for GV4 and GV6. Splitter blade
passing frequency can also be clearly seen for all sensors. The very low frequencies are too
distorted to determine frequencies. Harmonics of the blade passing frequency is observed at
60, as well as another unknown frequency at 65.

Figure 4.11: Frequencies observed at 5 degrees guide vane opening and 277,8 RPM. Showing values for GV4, GV5 and
GV6 for normalized frequencies from 0 to 70.

At 337,8 Rpm the Rheingans frequency is much more noticeable. Blade passing frequency is
still dominating and the splitter blade passing frequency is also easy to spot. Harmonics of the
blade passing frequency can be seen for GV5 and GV6. There also appears to be an
unidentified frequency at 53.

Figure 4.12: Frequencies observed at 5 degrees guide vane opening and 377,8 RPM.. Showing values for GV4, GV5 and
GV6 for normalized frequencies from 0 to 70.

37
 4.3.4. 7 degrees’ guide vane opening
At 337,6 RPM the blade passing frequency and the splitter blade passing frequency are
noticeable for all sensors, with the amplitude of the blade passing frequency being far the
largest. Rheingans frequency can also be seen, as well as some other low frequencies, which
are hard to pinpoint on exact frequencies due to spectral leakage. For GV5 and GV6
harmonics of the blade passing frequency can also be observed at 60 RPM, as well as a small
unknown frequency at 54 for GV5.

Figure 4.13: Frequencies observed at 7 degrees guide vane opening and 337,6 RPM.. Showing values for GV4, GV5 and
GV6 for normalized frequencies from 0 to 70.

At 374,2 RPM the amplitude of splitter blade passing frequency have grown considerably.
Rheingans and the blade passing frequency are also significant. Harmonics of the blade
passing frequency can be seen at all sensors at 60, as well as two small unknown amplitudes
at around 45 for GV5 and 58 for all sensors.

Figure 4.14: Frequencies observed at 7 degrees guide vane opening and 337,6 RPM.. Showing values for GV4, GV5 and
GV6 for normalized frequencies from 0 to 70.

38
 4.4. Speed-no-load
When running at speed-no-load, the pressure pulsations are relatively low when the speed is
low. The pressure pulsation regime is very similar for all three sensors.

Figure 4.15: Relative pressure pulsation amplitudes at speed-no-load. Pressure pulsations amplitudes at speed-no-load
compared to BEP at the y-axis. RPM along the x-axis.

Frequency analysis at speed-no load.

At 295 RPM Rheingans, splitter blade passing and blade passing frequencies are significant,
with the blade passing frequency at 30 dominating in magnitude.
The spectral leakage is quite low compared to other signal measurements at speed no load and
there are smaller noticeable frequencies at 3 and 60 for all sensors.

Figure 4.16: Frequencies observed at speed-no-load 295 RPM. Showing values for GV4, GV5 and GV6 for normalized
frequencies from 0 to 70.

39
At 340 RPM the blade passing frequency have similar amplitudes while Rheingans frequency
have increased a bit for all. The splitter blade passing frequency has also increased quite a bit
for GV4. Harmonics at 60 can only be observed for GV4.

Figure 4.17: Frequencies observed at speed-no-load 340 RPM.. Showing values for GV4, GV5 and GV6 for normalized
frequencies from 0 to 70.

At 386 RPM the blade passing frequency has increased in amplitude. The stochastic
fluctuations have grown considerably, making the previously noticeable frequency at around
3 disappear. Splitter blade passing frequency and Rheingans frequency are large enough to
still be clearly visible.

Figure 4.18: Frequencies observed at speed-no-load 386 RPM.. Showing values for GV4, GV5 and GV6 for normalized
frequencies from 0 to 70.

40
 At 441 RPM the distortion can be seen over the whole frequency spectrum. Rheingans,
splitter blade passing and blade passing frequencies are still noticeable. The blade passing
frequency amplitude having increased for GV4 and declined for GV5 and GV6 compared to
Figure 4.18.

Figure 4.19: Frequencies observed at speed-no-load 441 RPM.. Showing values for GV4, GV5 and GV6 for normalized
frequencies from 0 to 70.

4.5. Stress levels

To assess how much flexible speed turbines can reduce the damage contributors to the runner
blades, the correlation between the measured pressure pulsations in the vaneless space and the
fatigue on the runner blade has to be evaluated.

The relationship between pressure changes in the vaneless space has been mapped together
with the von-Mises equivalent stress on the runner blade. This was done with a combination
of CFD-analysis, FSI-analysis and experimental measurement on the Francis turbine in the
Waterpower Laboratory at NTNU by Valkvæ in 2016. [27] A drop of 10,5 KPa in the
vaneless space correlated with a decrease in the maximum equivalent von Mises stress on the
runner blade of 3,4 MPa. This maximum stress occurs along the trailing edge at the tip
towards the shroud. This provides a factor for stress to pressure 𝑃𝜎 = 328,8
If this is a linear relation, then: 𝜎𝑉𝑀 = 𝑃𝐺𝑉 ∗ 𝑃𝜎 + 𝑐 (Pa) [19]

Where 𝜎𝑉𝑀 is the maximum equivalent von Mises stress on the runner blade and 𝑃𝐺𝑉 is the
pressure in the vaneless space.

From the measurements conducted in this thesis the minimum mean pressure while operating
was 136 KPa, which if with an assumed linear relationship factor of 328,8 would mean a
41
stress of 44,72 MPa, which is obviously far high. If we deduct the atmospheric pressure of
98,5 KPa, it would still be 12 MPa.

From the measurements Valkvæ analyzed, the drop in pressure was from 68,5 KPa to 58 KPa
and the drop in Maximum equivalent von Mises Stress on the runner blade was from 8,4 MPa
to 5 MPa. Which mean that if a linear relationship over the whole pressure range is assumed,
then the lowest measurement from this thesis would still be far more than what Valkvæ found.
So assuming a linear relationship stress to pressure factor of 328,8 is not a valid assumption.

Since a linear relationship doesn’t fit, a Polynomial function of the second degree was made
based on the data gathered in Valkvæ’s thesis. An assumption of close to zero stress when
only experiencing atmospheric pressure was made.

𝜎𝑉𝑀 = 𝑃𝐺𝑉 2 ∗ 𝑃𝜎,1 + 𝑃𝐺𝑉 ∗ 𝑃𝜎,2 + 𝑐 (Pa) [20]

Where 𝑃𝜎,1 = 0,0015546, 𝑃𝜎,2 = 8,644 and 𝑐 = 66846

This relationship between pressure in the vaneless space and maximum equivalent von Mises
stress is used to calculate the stress mean stresses that occurs on the runner blade.
For the smaller pressure amplitudes occurring because of pressure pulsations, the stress to
pressure factor of 𝑃𝜎 = 328,8 will still be used. This assumption is made because that is the
factor Valkvæ got for the actual pressure change during measurement, and should based on
that correlate fairly well with small pressure changes.

This assumption makes it possible to estimate the amount of equivalent von Mises stress for
the whole operating range. Because Valkvæ doesn’t make it clear which sensors in the
vaneless space was used to measure the pressure in 2016, an average of the values from GV4,
GV5 and GV6 is used.

Valkvæ found that maximum Equivalent von Mises stress occurs at the tip of the trailing edge
of the runner blade towards the shroud. That will be the part of the runner blade that fatigue
analysis will look into.

Figure 4:20: Stress distribution on the runner blade.

Valkvæ 2016[27]

42
Figure 4.21: Maximum von Mises equivalent stress amplitude diagram. X-axis Dimensionless Speed Factor NED. Y-axis
Dimensionless Discharge Factor QED.

The Maximum equivalent von Mises stress amplitudes in the runner blade has low stress
amplitudes at low speed and low load. The stress amplitudes have the same ratio between
different operating points as the pressure pulsations described in 4.2 and 4.3.

Figure 4.22: Mean maximum equivalent von Mises stress diagram. X-axis Dimensionless Speed Factor NED. Y-axis
Dimensionless Discharge Factor QED.

The mean maximum equivalent von Mises stress diagram shows that speed is the most
dominating factor for the mean stresses at the runner blade

43
 4.5.1. Stress levels at Speed-no-Load
At speed-no-load the stress amplitudes are relatively similar to those at very low part load.
With a slow increase from 300 RPM to 365 RPM before increasing rapidly after that.

Figure 4.23: Maximum equivalent von Mises stress amplitudes at speed-no-load graph. X-axis RPM. Y-axis stress in MPa. .

The mean stresses at speed-no-load are generally quite a bit higher than those found at lower
part load. At synchronous speed it is 8,7 MPa, which is 64% higher than the mean stresses
found at synchronous speed at very low part load.

Figure 4.24: Mean maximum equivalent von Mises stress at speed-no-load graph. X-axis RPM. Y-axis stress in MPa.

44
 4.6. Fatigue

The Goodman method is used to obtain the effective stress. Each operating point has two
stress amplitudes and two mean stresses. Where one represents the stresses induced by
operating at that point and the other represents the stresses induced by starting the turbine,
moving to that operating point and then stopping.

The material used to investigate fatigue is a 17Cr-4Ni cast stainless steel, which Hans-Jörg
Huth looked into in Fatigue Design of Hydraulic Turbine Runners in 2005. [34]
The material has an ultimate tensile strength of 910 MPa and a yield strength of 661 MPa and
a stress concentration factor of 1.36. It should be noted that this yield limit is a bit higher than
what the S-N curves from ISO 19902 is based upon.

Figure 4.25: S-N curve of 17Cr-4Ni cast stainless steel.

From Fatigue Design of Hydraulic Turbine
Runners[34]

By using the calculated effective stresses from the Goodman method together with the S-N
curve from ISO 19902 the cycles to failure for each operating point is calculated.

𝑙𝑜𝑔10 𝑁 = 𝑙𝑜𝑔10 𝐾1 − 𝑚 ∗ 𝑙𝑜𝑔10 𝑆 (Cycles) [21]

𝑁 = 10𝑙𝑜𝑔10 𝐾1 −𝑚𝑙𝑜𝑔10 𝑆 (Cycles) [22]

The constant amplitude stress range can be described as

𝑆 = 2 ∗ 𝜎𝑎 ∗ 𝐾𝑡 [MPa] [23]

Since the maximum stress occurs close at the edge of the trailing edge right by the connection
the material parameters for the S-N curve must reflect that. This fatigue analysis assuming
that the blade is welded to the shroud.

45
 From ISO 19902[15], weld metal in load-carrying joints, gives the following 𝐾1 and 𝑚
values.

For N>107 For N<107

𝐾1 1013,62 1010,97
𝑚 5 3
Table 4.13: Constants for S-N curve.

For this thesis the values for N>107 will be used for High Cycle Fatigue and the values for
N<107 for Low Cycle Fatigue.

4.6.1. High Cycle Fatigue

The stress amplitudes for the range of operating conditions have been used to calculate the
number of cycles to failure when operating at that condition. The cycles to failure are high at
low load and low speed, and rapidly decreases when the speed reaches above 0.2 NED for all
levels of load.

Figure 4.26: High Cycle Fatigue. X-axis Dimensionless Speed Factor NED. Y-axis Dimensionless Discharge Factor QED.

Daniel Sannes conducted pressure measurements in 2018 and found that the dominating
pressure amplitude on-board the runner is the guide vane frequency. [7] For this thesis we will
assume that the stress amplitudes occurring during operation occurs with that frequency.
This is used to calculate the number of cycles per hour of operation.
𝑛
𝑛ℎ = 𝐹𝑔𝑣 ∗ 3600 = 𝑍𝑔𝑣 ∗ 𝐹𝑛 ∗ 3600 = 28 ∗ 60 ∗ 3600 = 1680 ∗ 𝑛 (-) [24]

Where 𝑛ℎ is the number of cycles per hour and 𝑛 is the number of runner revolutions per
minute. Miner’s rule has been used to calculate damage per hour of operating for all operating
conditions.

46
Figure 4.27: Accumulated Damage per Hour of Operation. X-axis Dimensionless Speed Factor NED. Y-axis Dimensionless
Discharge Factor QED.

4.6.2. Low Cycle Fatigue

Start-stop cycles have been looked at for LCF. Where each cycle represents going from
stationary position to an operating condition and then returning to a stationary position.
Due to having only conducted pressure measurements in the vaneless space, another
assumption has to be made in regards to calculating fatigue induced by start-stop cycles.
The assumption that each start-stop cycle only consists of only one pressure amplitude,
namely the difference in effective stress from stationary to the operating point, calculated by
using the Goodman method with the stress levels found in 4.5.

The diagram for start-stop cycles to failure shows that going to a reduction in speed reduces
the damage done per start-stop cycle significantly.

47
Figure: 4.28: Start-Stop Cycles to Failure. X-axis Dimensionless Speed Factor NED. Y-axis Dimensionless Discharge Factor QED.

48
5. Discussion

5.1. Reducing pressure pulsations with variable

speed turbine

Figure 5.1: Pressure pulsation diagram with values from GV4, GV5 and GV6 added together. With the Dimensionless Speed
Factor NED along the x-axis and the dimensionless discharge factor along the y-axis.

While the pressure pulsations for the individual sensors can be reduced quite drastically by
changing the runner speed within a 30 percent range of the synchronous speed, the reduction
for all three sensors combined is quite modest for most of the operating range. The exception
being low loads, where the reduction for the sensors combined can reach up to 15,36 percent
at 1 degree’ guide vane opening. At 4 degrees’ guide vane opening there is also a reduction of
11,98 percent. Above 6 degrees’ guide vane opening the maximum reduction for the
combined sensors is only 0,33 percent, achieved by reducing the runner speed 10 percent at 7
degrees’ guide vane opening. The largest reduction in an individual sensor above 6 degrees is
on the other hand 21,67 percent.

49
 Reduction in pressure pulsations at constant guide vane
20,00% angles
15,00%

10,00%

5,00%

0,00%
1 2 3 4 5 6 7 8 9 10 11 12
-5,00%

-10,00%

-15,00%
Figure 5.2: Maximum reduction in pressure pulsation along constant guide vane angles. Guide vane angles along the
x-axis and percentage decrease along the y-axis. Negative percentage mean that the pressure pulsations could not be
reduced and is showing the lowest increase from NED-BEP prto 10, 20 or 30 % reduced or increased runner speed.

A slightly larger reduction can be found at speed-no-load. Where a decrease from 340 RPM to
295 RPM decreased the combined pressure pulsation by 15,66 percent. However, while there
is a decrease in the pressure pulsation amplitudes, the frequency analysis of the two operating
points shows us that the high-amplitude blade passing frequency increases slightly. The
decrease is caused by a lower Rheingans and splitter blade passing amplitude. So from a
fatigue perspective, where each of those frequencies contributes to fatigue in the runner blade.
There might instead be an increase in sum of those fatigue load contributors despite the
reduction in pressure pulsations and the increase in the not-normalized frequencies.

Sensors Lowest pressure Lowest constant Lowest Lowest

pulsation with constant RPM up to 10 pressure pressure
RPM up to 12 degree’s degree’s pulsation pulsation
RPM Sum of RPM Sum of with flexible with flexible
pressure pressure RPM values RPM values
pulsations(one pulsations up to 12 up to 10
measurement degrees degrees
for each
whole guide
vane angle)
GV4 338,9 20,443 338,9 16,893 19,343 15,793
GV5 372,8 19,240 372,8 15,963 18,880 15,603
GV6 305,0 15,952 305,0 12,126 15,691 11,991
Combined 338,9 59,563 305,0 48,099 57,651 46,625
Table 5.2: Showing lowest summed values for guide vane openings.

For the combined sensors, the difference in pressure pulsations for their combined results is
quite modest.

50
 5.2. Frequency analysis
Frequency analysis on the measured shows that the blade passing frequency is by far the most
dominating factor for pressure pulsations over most operating point.

Only in 42 out of 762 measurement series recorded along the constant guide vane angles of 1
to 12 are other frequencies more dominant. 13 of these normalized frequencies are very low,
while 28 are of the is the splitter blade passing frequency at 15. The vast majority of these are
from GV4, with 27 measurement series with 15 in normalized frequency and 8 with below 1
in normalized frequency. As well as one anomaly with 19.9 in normalized frequency,
happening at 224,4 RPM at 2 degrees’ guide vane opening. The very low frequency results
are not valid for measuring the size of the pressure amplitudes because they appear a very
limited amount of times during each window segment when using the welch method, resulting
in distorted amplitude peaks.

Figure 5.3: Figure 4.18: Frequencies observed at 224,4 RPM with 2 degree’s guide vane opening. Showing values for GV4,
GV5 and GV6 for normalized frequencies from 0 to 70.

As shown in the 3D-plot, there is a noticeable normalized frequency at 10, 20 and 30. With 20
actually being the largest for GV4. Harmonics of the blade passing frequency can also be seen
at 60. This could be a result of the generator changing the torque slightly during the
measurement series.

5.3. Strain gauges

Strain gauges can be an effective tool to measure stresses occurring in the runner, but needs to
be calibrated properly to provide accurate strain values during the experiments.
In 3.3.2 a calibration of a strain gauge on a runner blade was conducted. But the calibration
lacked several key factors to give accurate enough calibration for the operating conditions
inside a turbine runner. A calibration procedure should instead be done with the strain gauge
submerged and subjected to known pressure and pressure pulsations.

51
 5.4. Fatigue assessment.
Because the pressures used to calculate the stress amplitudes and mean stresses were based on
pressure measurements on a scaled model turbine used in the Waterpower Laboratory the
damage per hour of operation is very low for almost all operating conditions measured. While
operating at BEP it would take 9.16*10^6 hours before failure occurs, which equals 1046
years of constant operation. The main reason for this is that the material used for analyzing
fatigue. With an Ultimate tensile strength of 910 MPa both the effective stress amplitudes and
the effective mean stresses are very low in comparison.

The results of the fatigue analysis are therefore only useful to show the relative damage
contribution.

5.4.1. High Cycle Fatigue

For operation at high load or around BEP, a change in runner speed will increase the damage
contribution. Lowering the flow rate generally reduces the damage contribution from
operation over the whole operating range. The relevant exception from this is when operating
at synchronous speed at part load, where a draft tube vortex creates pressure pulsations that
contributes to fatigue load. At part load around QED at 0.07, reducing the runner speed 11
percent can decrease the effect of this vortex and thereby decrease the damage contribution
from operation. This can result in a decrease of more than 50 percent.

Figure 5.6: Damage contribution relative to steady state operating at BEP. X-axis Dimensionless Speed Factor NED. Y-axis
Dimensionless Discharge Factor QED.

By comparing the fatigue load at synchronous speed for various flow rates with the fatigue
load with a reduction in runner speed the effectiveness of speed reduction becomes apparent.

52
 QED 0,175 0,16 QED-BEP 0,14 0,12 0,10 0,08 0,06 0,04
0,1526
NED - 0,1775 - - 0,160 0,160 0,160 0,1568 0,1487
reduced
Reduction 0 2,15 0 0 24,27 24,27 36,49 49,80 38,25
in damage
[%]
Table 5.2: Reduction in damage by reducing the Dimensionless Speed Factor while holding the Dimensionless Discharge
Factor constant. For steady-state operation.

As Table 5.2 demonstrates, reducing the speed at higher loads does not decrease the damage
factor of operating at that load. For lower loads the effect of reducing speed is large,
especially around 0,06 Dimensionless Discharge factor where a vortex ropes occur when
operating at synchronous speed.

5.4.2. Low Cycle Fatigue

For Francis Turbines Low Cycle Fatigue is mostly a result of starting and stopping the
turbine, which results in large changes in stresses on the runner blades. Large changes in
stresses resulting in Low Cycle Fatigue can also occur as a result of changing the operating
conditions, but this thesis focuses on start-stop cycles.

The results show that the damage from a start-stop cycle can be significantly reduced by
operating at lower runner speed. For a start-stop cycle to BEP, the number of cycles to failure
is 6,06*10^7. Applying Miner’s rule and scaling damage relative to a start-stop cycle to BEP
reveals the damage contribution per start-stop cycle relative to a start-stop cycle to BEP.

Figure 5.7: Damage contribution relative to start-stop cycle to BEP. X-axis Dimensionless Speed Factor NED. Y-axis Dimensionless
Discharge Factor QED.

The diagram shows that reducing the speed of the runner significantly reduces the damage per
start-stop cycle. Reducing the flow through the runner also reduces the damage per cycle,

53
except when being close to Speed-No-Load at high speeds. By reducing the Dimensionless
Speed Factor from NED at synchronous speed to 0,14, the following reductions are made if the
Dimensionless Discharge Factor is held constant.

QED 0,175 0,16 QED-BEP 0,14 0,12 0,10 0,08 0,06 0,04
0,1526
Reduction in 67,71 69,73 70,96 73,15 74,24 78,03 80,67 82,89 83,52
damage [%]

Table 5.3: Reduction in damage by reducing the Dimensionless Speed Factor while holder the Dimensionless Discharge
Factor constant. For start-stop cycles.

Which demonstrates clearly that fatigue loads from start-stop cycles can be significantly
reduced by reducing the operating speed. While reducing the speed is most efficient at lower
loads, it’s also very effective at higher loads, as a reduction by 67,71 percent damage for start
stop cycles at 0,175 QED demonstrates.

5.4.3. Combining start-stop cycles and operating hours.

To compare the damage contributions of both HCF and LCF on the runner blade the number
of operating hours it takes to do the same damage as one start-stop cycle to the same operating
point has been calculated with Miner’s rule.

ℎ∗𝑛ℎ 1 𝑂 𝑁
= 𝑁 𝑎𝑛𝑑 𝑛ℎ = 1680 ∗ 𝑛 → ℎ = 1680∗𝑛∗𝑁 Where 𝑛ℎ is cycles per hour, ℎ is hours, 𝑁𝑂
𝑁𝑂 𝑆 𝑆
is cycles to failure for steady-state operation at the operating point, 𝑁𝑆 is the cycles to failure
for start-stop cycles to that operating point and 𝑛 is runner revolutions per minute.

54
Figure 5.8: Hours of operation at steady state to equal the amount of damage from 1 start-stop cycle. X-axis Dimensionless Speed Factor
NED. Y-axis Dimensionless Discharge Factor QED

The results first and foremost shows that something is wrong with the assumptions made to
calculate fatigue. Even while operating inside of the zones with relatively low pressure
pulsations the fatigue from the pressure pulsations will in a matter of minutes accumulate
more damage than a start-stop cycle to the same operation point. That is most likely the result
of a start-stop cycle only being counted as one cycle, instead of several thousands.

The figure 5.8 does make a recommendation for what operation conditions should be chosen
to minimize damage based how long planned operation is. However, due to the invalid
calculation of the relation of the damage from pressure pulsations during steady-state
operation and start-stop cycles, it only serves to show the relative difference between
operating inside and outside of the low-pressure pulsation zones.

5.4.4. Impact of variable speed turbines

As shown in the previous sections of this chapter, a variable speed turbine can significantly
reduce the damage accumulated for the runner blades and increase the lifespan of the turbine.
Especially the damage from start-stop cycles can be reduced significantly by reducing the
speed. As table 5.3 demonstrates, with over 60 percent reduction in fatigue load for start-stop
cycles by reducing NED from synchronous speed to 0.14. The damage accumulated during
continuous operation can mostly be reduced at lower loads.

55
This fatigue calculated does not factor in the reduced time, and therefore reduced number of
stress amplitudes, start-up would take with a variable speed turbine, as it calculates only one
amplitude per start-stop. Changing the model to account for the actual start-stop process
would likely change the results drastically.

5.4.5. Scaling to a larger turbine

In a larger turbine, the stresses would be much higher and damage accumulated would be
significantly increased compared to the numbers presented here. While the stresses used to
calculate fatigue in this thesis was derived from experiments done on a smaller scaled model
turbine at the Waterpower Laboratory, the material used for analyzing fatigue loads had
material properties similar to those used in larger turbines.

5.5. Known errors and assumptions made.

5.5.1. Stress levels
The main assumption that the results presented here is based upon is the correlation between
Equivalent von Mises Stress on the runner blade and the pressure in the vaneless space. While
there is a correlation between the pressure pulsations in the vaneless space and the stresses
occurring in the runner blades, the factors determining this correlation is based upon only one
pressure drop and its corresponding Equivalent von Mises stress reduction analyzed by
Valkvæ in 2016. [27] Analyzing the results using the linear ratio between the stress and
pressure drop was attempted at first, but it became clear that a linear relationship between
them could not be assumed for the whole pressure range. A second degree polynomial
function based on that pressure drop and the assumed slope to a zero value for both stress and
pressure was used instead.

The only way to validate these results was to compare them to previous stress analysis done
one the runner blades of the scaled model turbine at the Waterpower Laboratory. Andreas
Nilssen Skorpen found the following Maximum mean equivalent Von Mises Stresses for the
runner blades in Impact from flexible operation on High head Francis turbines in 2018 [20]:
17.62, 14.92, 14.65 and 16.50 MPa for low part load, part load, BEP and High load
respectively, while the values used in this thesis for similar operating points was 5.9, 7.3, 8,0

56
and 8,34 MPa respectively. Which would imply that the factors used to calculate stresses in
this thesis is far too low.

However, comparing them to the stresses Daniel Sannes found stresses when conducting
measurements and simulations in 2018 paints another picture. [7] His results for mean
maximum equivalent von Mises stresses was 7,575 MPa for BEP, 5,138 for part load and
2,301 for minimum load. Which are similar to the values used in this thesis for part load and
BEP, while a lot lower for minimum load.

5.5.2. Stress frequencies

Another assumption is that all the pressure pulsation amplitudes originates from the Guide
vane passing frequency, which is not the case in reality. While the guide vane passing
frequency is the dominating frequency for the runner blades, as demonstrated in Daniel
Sannes’s thesis from 2018[7], stochastic pressure pulsations, harmonics and some lower
frequency pressure pulsations also occur. What is in fact several different pressure amplitudes
occurring with different frequencies are all counted as one larger amplitude, distorting the
calculated damage. As this increases the amplitude of the highest significant frequency, and
thereby increases the accumulated damage from steady-state operation. This in fact introduces
a safety factor to the calculations.

By applying a rainflow counting algorithm on the signal one could calculate the fatigue load
from the pressure pulsations correctly. But since this thesis only has signals from the pressure
pulsations in the vaneless space this was not possible.

5.5.3. Start-stop cycles.

With regards to start-stop cycles two assumptions were made, both significant.
First that a start-stop cycle only consists of one effective stress amplitude, which led to some
bizarre results as can be seen in figure 5.8. A study by Gagnon, Tahan, Bocher and Thibault in
2010 [31], shows that a startup consists of thousands of effective stress amplitudes of varying
magnitude. So even though the effective amplitude stress used for start-stop cycles in this
study is comparable with stresses found by Daniel Sannes in 2018[7], the lifetime assessment
based upon them get is of by a factor of over 1000.

The other assumption made in regard to start-stop cycles is that the effective amplitude stress
is simply a function of mean von Mises stress and the stress amplitude created by the pressure
pulsations at steady state operation. Both found with the assumption discussed in 5.5.1. While
the study from Gagnon et al [31] shows that there are a lot of different stress amplitudes
occurring during the cycle, with a large portion of them being high frequency amplitudes
occurring at speed-no-load before connecting the generator. This Thesis completely neglects
the reduction in fatigue loads that could be made by not having to subject the runner to the
stresses occurring at speed-no-load for large portion of the startup duration.

57
58
6. Conclusion
Pressure measurements has been conducted on a model turbine at the Waterpower Laboratory
at the Norwegian University of Science and Technology. The results of these measurements
show that pressure pulsations in the vaneless space can be significantly reduced by operating
with a flexible speed turbine. The largest reductions can be found when operating at part
loads.

A correlation between pressure pulsations in the vaneless space and stresses on the runner
blade found in a previous measurement at the Waterpower Laboratory was used to analyze the
Fatigue on the runner blades. [27] The fatigue analysis focused on how accumulated damage
to the runner blades could be reduced by operating with a flexible speed turbine.

The results of the fatigue analysis show that damage to the runner blades can be significantly
reduced by reducing the speed of the runner. A slight reduction in runner speed at part load
gave above 80 percent reduction in fatigue load for start-stop cycles, and can also
significantly be reduced at all other load levels, while damage accumulated during steady-
state operation could mostly be reduced at lower loads.

In order to do this fatigue analysis based only on pressure measurements from the vaneless
space between the guide vanes and the runner blades several key assumptions had to be made,
and the value of the results presented here can only be treated as indicative. The way fatigue
load from start-stop cycles were calculated was especially questionable, as no measurements
from actual start-stop scenarios were done.

59
60
7. Further Work
To properly calculate the impact of flexible speed turbines, the stresses occurring in the
runner blades has to be accurately measured during both steady-state and transient operation.
And then compared to the stresses occurring when operating the turbine as if it were not a
variable speed turbine, only using the generator when synchronous speed is reached and
stable enough.
This could be done with strain gages directly on the turbine at the Waterpower Laboratory.
The strain gages would have to be calibrated beforehand and checked for drift during the
experiments. With accurate results from strain gages the stresses and the frequencies they
occur at can be calculated accurately. Rainflow counting can then be used on the measured
stresses to calculate the fatigue over the whole operating range.

Because of the increased demand for Hydropower to act as a battery to balance out the grid,
start-stop cycles are the most important phase to reduce fatigue at in the future. Therefore, the
focus of the experiments should be the stresses occurring while going from a completely
stationary turbine to various operating conditions and then shutting it down. Doing such an
experiment and combining it with rainflow counting would make it possible to predict the
damage reductions from operating at runner speeds outside of the speed at the Best Efficiency
Point without having to make as many assumptions that invalidates the end-result.

61
8. Bibliography
[1] Bjørndal, Halvard & Reynaud, Andre & L. Holo, Anders. 2011. Mechanical robustness of
Francis runners, requirements to reduce the risk of cracks in blades.

[2] Huang, Oram, Sick. 2014. Static and Dynamic stress analyses of the prototype high head
Francis runner based on site measurement doi:10.1088/1755-1315/22/3/032052

[3] Huang, Chamberland-Lauzon, Oram, Klopfer, Runchonnet. 2014. Fatigue analyses of the
prototype Francis runners based on site measurements and simulations. Doi:10.1088/1755-
1315/22/1/012014

[4] Iliev, Trivedi, Agnalt, Dahlhaug. 2018. Variable-speed operation and pressure pulsations
in a Francis turbine and a pump-turbine.

[5] Tovsild. 2013. Trykkpulsasjoner i Francisturbiner - sammenligning av modell og

prototypmålinger. Norwegian University of Science and Technology.

[6] Haugan.2007. Trykkpulsasjoner i Francisturbiner. Norwegian University of Science and

Technology.

[7] Sannes.2018. Pressure Pulsation and Stresses in a Francis Turbine Operating at Variable
Speed. Norwegian University of Science and Technology.

[8] Dörfler, Sick, Coutu, 2013, Flow-Induced Pulsation and Vibration in Hydroelectric
Machinery. Springer London

[9] Ruchonnet N, Nicolet C, Avellan F. One-dimensional modeling of rotor stator interaction

in Francis pump-turbine. Proceedings of the 23rd IAHR Symposium on Hydraulic Machinery
and Systems 2006.

[10] Haga. 2014. Dynamic load on High Head Francis turbines during start/stop. Norwegian
University of Science and Technology.

[11] Seidel, Mende, Hubner, Weber, Otto. 2014. Dynamic loads in Francis runners and their
impact on fatigue life. doi:10.1088/1755-1315/22/3/032054

[12] 2014. Guideline DKD-R 6-1, Calibration of Pressure Gauges. German Calibration
Service.

[13] Agnalt. 2016. Pressure measurements inside a Francis turbine runner. Norwegian
University of Science and Technology.

[14] R. A. Gujar, S. V. Bhaskar.2013. Shaft Design under Fatigue Loading By Using

Modified Goodman Method. International Journal of Engineering Research and Applications.

62
[15] ISO 19902:2007 Petroleum and natural gas industries - Fixed steel offshore structures

[16] IEC60193 Hydraulic turbines, storage pumps and pump-turbines - Model acceptance
tests

[17] IEC 60041 Field acceptance tests to determine the hydraulic performance of hydraulic
turbines, storage pumps and pump-turbines

[18] https://www.ntnu.edu/nvks/francis-99

[19] IEC 17025 General requirements for the competence of testing and calibration
laboratories.

[20] Andreas Nilssen Skorpen. 2018. Impact from flexible operation on High head Francis
turbines. Norwegian University of Science and Technology.

[21] Scott Woodward.2009. MAE 334 – Introduction to computers and instrumentation.

Signal Characteristics. University at Buffalo.

[22] https://se.mathworks.com/help/matlab/ref/fft.html(timestamp 14.14 14.12.2018)

[23] Datasheet for NI-cDAQ 9178 http://www.ni.com/pdf/manuals/374046a.pdf (timestamp

11.12 5.12.2018)

[24] Datasheet for NI 9237 http://www.ni.com/pdf/manuals/374186a_02.pdf (timestamp

11.17 5.12.2018)

[25] Einar Agnalt. 2018. Measurement report 2018. HiFrancis WP 1.3***

[26] Arne Kjølle.2006. Hydropower in Norway, Mechanical Equipment.

[27] Valkvæ, I. 2016. Dynamic loads on Francis turbines. Norwegian University of Science
and Technology.

[28] Shannon.1949. Communication in the Presence of Noise*

[29] https://se.mathworks.com/help/signal/ref/hann.html (timestamp 10.50 06.01 2019)

[30] Welch. 1967. The Use of Fast Fourier Transform for the Estimation of Power Spectra. A
Method Based on Time Averaging Over Short, Modified Periodograms. IEEE Trans. Audio
and Electrocoust.

[31] Gagnon, Tahan, Bocher, Thibault. 2010. Impact of startup scheme on Francis runner life
expectancy. IOP Conf. Series: Earth and Environmental Science. doi:10.1088/1755-
1315/12/1/012107

[32] Michael Bak. 2016. Mean Stress Corrections in Fatigue. https://caeai.com/blog/mean-

stress-corrections-fatigue (timestamp 13.50 07.01 2019)

63
[33] DeLuca. Understanding Fatigue. https://files.asme.org/igti/knowledge/articles/13048.pdf
(timestamp 14.21 07.01 2019)

[34] Hans-Jorg Huth.2005. Fatigue Design of Hydraulic Turbine Runners. Norwegian

University of Science and Technology.

64
Appendix A – Risk Assesment

65
66
67
68
69
70
71
72
73
74
 Appendix B – Matlab scripts
Because of limited time the Matlab scripts have not been properly organized and might be
hard to understand. Matlab version R2018a has been used. All functions that aren’t part of the
standard Matlab functions are mentioned.

Hill chart and pressure diagrams

%script to extract values from excel and tdms files to create hill charts
and pressure data

%Trykkhilling is an excelfile containing operating parameters

[num,txt,raw]=xlsread('trykkhilling');
i=1;
for i=3:14798
hill{1,i-2}=raw{i,21};
hill{2,i-2}=raw{i,22};
hill{3,i-2}=raw{i,23};
hill{4,i-2}=raw{i,41};
hill{5,i-2}=raw{i,9};
end
i=1;
for i=10001:14798
hill2{1,i-10000}=raw{i,21};
hill2{2,i-10000}=raw{i,22};
hill2{3,i-10000}=raw{i,23};
hill2{4,i-10000}=raw{i,41};
hill2{5,i-10000}=raw{i,9};
end

i=1;

for i=1:14796
hill{1,i}=str2num(hill{1,i});
hill{2,i}=str2num(hill{2,i});
hill{3,i}=str2num(hill{3,i});
hill{4,i}=str2num(hill{4,i});
hill{5,i}=str2num(hill{5,i});
end
for i=1:length(hill)
if hill{1,i} > 2
hill{1,i}=hill{1,i};
elseif hill{1,i} ==0;
hill{1,i}=0;
hill{2,i}=0;
hill{3,i}=0;
hill{4,i}=0;
hill{5,i}=0;
else
hill{1,i}=0;
hill{2,i}=0;
hill{3,i}=0;
hill{4,i}=0;

75
hill{5,i}=0;
end
end

HILL=zeros(7,258);
a=0;
b=0;
n=1;
i=1;
Eff=0;
Ned=0;
Qed=0;
GV=0;
RPM=0;
NumOfMeasurements=0;
for n=1:255
Eff=0;
Ned=0;
Qed=0;
GV=0;
a=0;
NumOfMeasurements=0;
RPM=0;
extra=0;
if n==255;
extra=15;
end

for i=1:70-extra
b=b+1;
if hill{1,b} >0.1
%Skip point 127
if b==127
b=128;
end

Eff=Eff+hill{1,b};
Ned=Ned+hill{2,b};
Qed=Qed+hill{3,b};
GV=GV+hill{4,b};
RPM=RPM+hill{5,b};
NumOfMeasurements=NumOfMeasurements+1;
else
a=i;
b=b+1;
break;
end

end

if extra==15;
a=i+1;

76
end
HILL(1,n)=Eff/(a-1);
HILL(2,n)=Ned/(a-1);
HILL(3,n)=Qed/(a-1);
HILL(4,n)=GV/(a-1);
HILL(5,n)=RPM/(a-1);
HILL(6,n)=NumOfMeasurements;
HILL(7,n)=(a-1);

end
%Skip point 233
for n=1:232
HILLTO(:,n)=HILL(:,n);
end
for n=233:254
HILLTO(:,(n))=HILL(:,n+1);
end

HILL=zeros(7,254);
HILL=HILLTO;
%Should only be used if some sequences are missplaced, ignore if
%the data is fine.
%Sortering=zeros(5,5);
%i=0;
%for i=1:5
% Sortering(:,i)=HILL(:,i+62);
%end

%HILL(:,63)=Sortering(:,3);
%HILL(:,64)=Sortering(:,4);
%HILL(:,65)=Sortering(:,5);
%HILL(:,66)=Sortering(:,1);
%HILL(:,67)=Sortering(:,2);

i=0;
for i=1:254
RPM(1,i)=HILL(5,i);
Qed(1,i)=HILL(3,i);
Ned(1,i)=HILL(2,i);
Eff(1,i)=HILL(1,i);
end

clearvars raw HILLTO extra a b n i NumOfMeasurements num hill txt

%TDMS_readTDMSFile is a function written by Jim Hokanson

%trykkfredag and trykksenfredag are files containing sensor output from
%pressure measurements
tolvtilni =
TDMS_readTDMSFile('C:\Users\eirilo\Documents\MATLAB\trykkfredag.tdms');
attetilen=TDMS_readTDMSFile('C:\Users\eirilo\Documents\MATLAB\trykksenfreda
g.tdms');

for i=3:length(tolvtilni.data)
Trykk{1,i}=tolvtilni.data{1,i};
b=i;
end
for i=3:length(attetilen.data)
Trykk{1,b+i-1}=attetilen.data{1,i};

77
end
%constants for for the sensors to match the atmospheric pressure
Off1=16,44964607;
Off2=7,014733975;
Off3=-4,140070341;
a1=attetilen.propValues{1,3}{1,11};
a2=attetilen.propValues{1,4}{1,11};
a3=attetilen.propValues{1,5}{1,11};

i=0;
j=0;
l=1;
k=0;

%Extracting pressure amplitudes and mean values

for k=3:length(Trykk)

T=Trykk{1,k};
PressFreq=Trykk{1,k};

[N,e] = histcounts(T,1000);
Me = mean(T);
NumberOfPoints = sum(N);
temp=0;
conf=0.97;
for i=1:length(N)
temp=temp +N(i); value=temp/NumberOfPoints;
if value >=(1-conf)/2
lower=e(i);
break
end
end

for j=1:length(N)
temp=temp +N(j); value=temp/NumberOfPoints;
if value >=conf+(1-conf)/2
upper=e(j);
break
end
end

VoltAmp=(upper-lower)/2;

if l == 1
m = (k/4)+0.25;
if NumberOfPoints > 198000
GV4Volts{1,m}=VoltAmp;
GV4Stress(1,m)=Me*a1+Off1;
FrekA{1,m}=PressFreq;
end
l=10;
end
if l == 2

78
 m = (k/4);
if NumberOfPoints > 198000
LabVolts{1,m}=VoltAmp;
GV5Stress(1,m)=Me*a2+Off2;
FrekA{2,m}=PressFreq;
end
l=11;
end
if l == 3
m = (k/4)-0.25;
if NumberOfPoints > 198000
GV6Volts{1,m}=VoltAmp;
GV6Stress(1,m)=Me*a3+Off3;
FrekA{3,m}=PressFreq;
end
l=12;
end

if l == 4
l=13;
end

l=l-8;

if l == 5
l=1;
end

end
i=0;
b=0;
for i=1:length(GV4Volts);
if GV4Volts{1,i} >0
b=b+1;
CGV4{1,b}=GV4Volts{1,i};
FrekAC{1,b}=FrekA{1,i};
CGV4Stress(1,b)=GV4Stress(1,i);
end
end
i=0;
b=0;
for i=1:length(LabVolts);
if GV6Volts{1,i} >0
b=b+1;
CGV6{1,b}=GV6Volts{1,i};
FrekAC{3,b}=FrekA{3,i};
CGV5Stress(1,b)=GV5Stress(1,i);
end
end
i=0;
b=0;
for i=1:length(LabVolts);
if LabVolts{1,i} >0
b=b+1;
CLab{1,b}=LabVolts{1,i};
FrekAC{2,b}=FrekA{2,i};
CGV6Stress(1,b)=GV6Stress(1,i);
end
end

79
G4=cell2mat(CGV4);
G6=cell2mat(CGV6);
GLab=cell2mat(CLab);
for i=1:118
G4A(1,i)=G4(1,i);
G6A(1,i)=G6(1,i);
GLA(1,i)=GLab(1,i);
G4ST(1,i)=CGV4Stress(1,i);
G5ST(1,i)=CGV5Stress(1,i);
G6ST(1,i)=CGV6Stress(1,i);
FrequencyRaw{1,i}=FrekAC{1,i};
FrequencyRaw{2,i}=FrekAC{2,i};
FrequencyRaw{3,i}=FrekAC{3,i};
end
for i=119:167
G4A(1,i)=G4(1,i+1);
G6A(1,i)=G6(1,i+1);
GLA(1,i)=GLab(1,i+1);
G4ST(1,i)=CGV4Stress(1,i+1);
G5ST(1,i)=CGV5Stress(1,i+1);
G6ST(1,i)=CGV6Stress(1,i+1);
FrequencyRaw{1,i}=FrekAC{1,i+1};
FrequencyRaw{2,i}=FrekAC{2,i+1};
FrequencyRaw{3,i}=FrekAC{3,i+1};
end
for i=168:232
G4A(1,i)=G4(1,i+2);
G6A(1,i)=G6(1,i+2);
GLA(1,i)=GLab(1,i+2);
G4ST(1,i)=CGV4Stress(1,i+2);
G5ST(1,i)=CGV5Stress(1,i+2);
G6ST(1,i)=CGV6Stress(1,i+2);
FrequencyRaw{1,i}=FrekAC{1,i+2};
FrequencyRaw{2,i}=FrekAC{2,i+2};
FrequencyRaw{3,i}=FrekAC{3,i+2};
end
for i=233:254
G4A(1,i)=G4(1,i+3);
G6A(1,i)=G6(1,i+3);
GLA(1,i)=GLab(1,i+3);
G4ST(1,i)=CGV4Stress(1,i+3);
G5ST(1,i)=CGV5Stress(1,i+3);
G6ST(1,i)=CGV6Stress(1,i+3);
FrequencyRaw{1,i}=FrekAC{1,i+3};
FrequencyRaw{2,i}=FrekAC{2,i+3};
FrequencyRaw{3,i}=FrekAC{3,i+3};
end

numb=1000;
x=linspace(0.120,0.28,numb);
Speed=linspace(227,530,numb);
%1-17 18-34 35-50 51-69 70-90 91-109 110-134 135-160 161-189 190-217
%218-239 240-254
%Create new interpolated Q_ed based on Ned
q(:,12)=interp1(Ned(1:17),Qed(1:17),x,'PCHIP','extrap');
q(:,11)=interp1(Ned(18:34),Qed(18:34),x,'PCHIP','extrap');
q(:,10)=interp1(Ned(35:50),Qed(35:50),x,'PCHIP','extrap');
q(:,9)=interp1(Ned(51:69),Qed(51:69),x,'PCHIP','extrap');
q(:,8)=interp1(Ned(70:90),Qed(70:90),x,'PCHIP','extrap');
q(:,7)=interp1(Ned(91:109),Qed(91:109),x,'PCHIP','extrap');

80
q(:,6)=interp1(Ned(110:134),Qed(110:134),x,'PCHIP','extrap');
q(:,5)=interp1(Ned(135:160),Qed(135:160),x,'PCHIP','extrap');
q(:,4)=interp1(Ned(161:189),Qed(161:189),x,'PCHIP','extrap');
q(:,3)=interp1(Ned(190:217),Qed(190:217),x,'PCHIP','extrap');
q(:,2)=interp1(Ned(218:239),Qed(218:239),x,'PCHIP','extrap');
q(:,1)=interp1(Ned(240:254),Qed(240:254),x,'PCHIP','extrap');

%Create new interpolated Q_ed based on RPM

qS(:,12)=interp1(RPM(1:17),Qed(1:17),Speed,'PCHIP','extrap');
qS(:,11)=interp1(RPM(18:34),Qed(18:34),Speed,'PCHIP','extrap');
qS(:,10)=interp1(RPM(35:50),Qed(35:50),Speed,'PCHIP','extrap');
qS(:,9)=interp1(RPM(51:69),Qed(51:69),Speed,'PCHIP','extrap');
qS(:,8)=interp1(RPM(70:90),Qed(70:90),Speed,'PCHIP','extrap');
qS(:,7)=interp1(RPM(91:109),Qed(91:109),Speed,'PCHIP','extrap');
qS(:,6)=interp1(RPM(110:134),Qed(110:134),Speed,'PCHIP','extrap');
qS(:,5)=interp1(RPM(135:160),Qed(135:160),Speed,'PCHIP','extrap');
qS(:,4)=interp1(RPM(161:189),Qed(161:189),Speed,'PCHIP','extrap');
qS(:,3)=interp1(RPM(190:217),Qed(190:217),Speed,'PCHIP','extrap');
qS(:,2)=interp1(RPM(218:239),Qed(218:239),Speed,'PCHIP','extrap');
qS(:,1)=interp1(RPM(240:254),Qed(240:254),Speed,'PCHIP','extrap');

%Create new interpolated Eff based on NED

ef(:,12)=interp1(Ned(1:17),Eff(1:17),x,'PCHIP','extrap');
ef(:,11)=interp1(Ned(18:34),Eff(18:34),x,'PCHIP','extrap');
ef(:,10)=interp1(Ned(35:50),Eff(35:50),x,'PCHIP','extrap');
ef(:,9)=interp1(Ned(51:69),Eff(51:69),x,'PCHIP','extrap');
ef(:,8)=interp1(Ned(70:90),Eff(70:90),x,'PCHIP','extrap');
ef(:,7)=interp1(Ned(91:109),Eff(91:109),x,'PCHIP','extrap');
ef(:,6)=interp1(Ned(110:134),Eff(110:134),x,'PCHIP','extrap');
ef(:,5)=interp1(Ned(135:160),Eff(135:160),x,'PCHIP','extrap');
ef(:,4)=interp1(Ned(161:189),Eff(161:189),x,'PCHIP','extrap');
ef(:,3)=interp1(Ned(190:217),Eff(190:217),x,'PCHIP','extrap');
ef(:,2)=interp1(Ned(218:239),Eff(218:239),x,'PCHIP','extrap');
ef(:,1)=interp1(Ned(240:254),Eff(240:254),x,'PCHIP','extrap');

%Create new interpolated RPM based on NED

r(:,12)=interp1(Ned(1:17),RPM(1:17),x,'PCHIP','extrap');
r(:,11)=interp1(Ned(18:34),RPM(18:34),x,'PCHIP','extrap');
r(:,10)=interp1(Ned(35:50),RPM(35:50),x,'PCHIP','extrap');
r(:,9)=interp1(Ned(51:69),RPM(51:69),x,'PCHIP','extrap');
r(:,8)=interp1(Ned(70:90),RPM(70:90),x,'PCHIP','extrap');
r(:,7)=interp1(Ned(91:109),RPM(91:109),x,'PCHIP','extrap');
r(:,6)=interp1(Ned(110:134),RPM(110:134),x,'PCHIP','extrap');
r(:,5)=interp1(Ned(135:160),RPM(135:160),x,'PCHIP','extrap');
r(:,4)=interp1(Ned(161:189),RPM(161:189),x,'PCHIP','extrap');
r(:,3)=interp1(Ned(190:217),RPM(190:217),x,'PCHIP','extrap');
r(:,2)=interp1(Ned(218:239),RPM(218:239),x,'PCHIP','extrap');
r(:,1)=interp1(Ned(240:254),RPM(240:254),x,'PCHIP','extrap');

%Create new interpolated Pressure amplitude values based on Ned

GV4(:,12)=interp1(Ned(1:17),G4A(1:17),x,'PCHIP');
GV4(:,11)=interp1(Ned(18:34),G4A(18:34),x,'PCHIP');
GV4(:,10)=interp1(Ned(35:50),G4A(35:50),x,'PCHIP');
GV4(:,9)=interp1(Ned(51:69),G4A(51:69),x,'PCHIP');
GV4(:,8)=interp1(Ned(70:90),G4A(70:90),x,'PCHIP');
GV4(:,7)=interp1(Ned(91:109),G4A(91:109),x,'PCHIP');
GV4(:,6)=interp1(Ned(110:134),G4A(110:134),x,'PCHIP');
GV4(:,5)=interp1(Ned(135:160),G4A(135:160),x,'PCHIP');
GV4(:,4)=interp1(Ned(161:189),G4A(161:189),x,'PCHIP');

81
GV4(:,3)=interp1(Ned(190:217),G4A(190:217),x,'PCHIP');
GV4(:,2)=interp1(Ned(218:239),G4A(218:239),x,'PCHIP');
GV4(:,1)=interp1(Ned(240:254),G4A(240:254),x,'PCHIP');

%Create new interpolated mean Pressure values based on Ned

GV4STR(:,12)=interp1(Ned(1:17),G4ST(1:17),x,'PCHIP');
GV4STR(:,11)=interp1(Ned(18:34),G4ST(18:34),x,'PCHIP');
GV4STR(:,10)=interp1(Ned(35:50),G4ST(35:50),x,'PCHIP');
GV4STR(:,9)=interp1(Ned(51:69),G4ST(51:69),x,'PCHIP');
GV4STR(:,8)=interp1(Ned(70:90),G4ST(70:90),x,'PCHIP');
GV4STR(:,7)=interp1(Ned(91:109),G4ST(91:109),x,'PCHIP');
GV4STR(:,6)=interp1(Ned(110:134),G4ST(110:134),x,'PCHIP');
GV4STR(:,5)=interp1(Ned(135:160),G4ST(135:160),x,'PCHIP');
GV4STR(:,4)=interp1(Ned(161:189),G4ST(161:189),x,'PCHIP');
GV4STR(:,3)=interp1(Ned(190:217),G4ST(190:217),x,'PCHIP');
GV4STR(:,2)=interp1(Ned(218:239),G4ST(218:239),x,'PCHIP');
GV4STR(:,1)=interp1(Ned(240:254),G4ST(240:254),x,'PCHIP');

%Create new interpolated Pressure amplitude values based on Ned

GV6(:,12)=interp1(Ned(1:17),G6A(1:17),x,'PCHIP','extrap');
GV6(:,11)=interp1(Ned(18:34),G6A(18:34),x,'PCHIP','extrap');
GV6(:,10)=interp1(Ned(35:50),G6A(35:50),x,'PCHIP','extrap');
GV6(:,9)=interp1(Ned(51:69),G6A(51:69),x,'PCHIP','extrap');
GV6(:,8)=interp1(Ned(70:90),G6A(70:90),x,'PCHIP','extrap');
GV6(:,7)=interp1(Ned(91:109),G6A(91:109),x,'PCHIP','extrap');
GV6(:,6)=interp1(Ned(110:134),G6A(110:134),x,'PCHIP','extrap');
GV6(:,5)=interp1(Ned(135:160),G6A(135:160),x,'PCHIP','extrap');
GV6(:,4)=interp1(Ned(161:189),G6A(161:189),x,'PCHIP','extrap');
GV6(:,3)=interp1(Ned(190:217),G6A(190:217),x,'PCHIP','extrap');
GV6(:,2)=interp1(Ned(218:239),G6A(218:239),x,'PCHIP','extrap');
GV6(:,1)=interp1(Ned(240:254),G6A(240:254),x,'PCHIP','extrap');

%Create new interpolated mean Pressure values based on Ned

GV6STR(:,12)=interp1(Ned(1:17),G6ST(1:17),x,'PCHIP');
GV6STR(:,11)=interp1(Ned(18:34),G6ST(18:34),x,'PCHIP');
GV6STR(:,10)=interp1(Ned(35:50),G6ST(35:50),x,'PCHIP');
GV6STR(:,9)=interp1(Ned(51:69),G6ST(51:69),x,'PCHIP');
GV6STR(:,8)=interp1(Ned(70:90),G6ST(70:90),x,'PCHIP');
GV6STR(:,7)=interp1(Ned(91:109),G6ST(91:109),x,'PCHIP');
GV6STR(:,6)=interp1(Ned(110:134),G6ST(110:134),x,'PCHIP');
GV6STR(:,5)=interp1(Ned(135:160),G6ST(135:160),x,'PCHIP');
GV6STR(:,4)=interp1(Ned(161:189),G6ST(161:189),x,'PCHIP');
GV6STR(:,3)=interp1(Ned(190:217),G6ST(190:217),x,'PCHIP');
GV6STR(:,2)=interp1(Ned(218:239),G6ST(218:239),x,'PCHIP');
GV6STR(:,1)=interp1(Ned(240:254),G6ST(240:254),x,'PCHIP');

%Create new interpolated Pressure amplitude values based on Ned

GV5(:,12)=interp1(Ned(1:17),GLA(1:17),x,'PCHIP','extrap');
GV5(:,11)=interp1(Ned(18:34),GLA(18:34),x,'PCHIP','extrap');
GV5(:,10)=interp1(Ned(35:50),GLA(35:50),x,'PCHIP','extrap');
GV5(:,9)=interp1(Ned(51:69),GLA(51:69),x,'PCHIP','extrap');
GV5(:,8)=interp1(Ned(70:90),GLA(70:90),x,'PCHIP','extrap');
GV5(:,7)=interp1(Ned(91:109),GLA(91:109),x,'PCHIP','extrap');
GV5(:,6)=interp1(Ned(110:134),GLA(110:134),x,'PCHIP','extrap');
GV5(:,5)=interp1(Ned(135:160),GLA(135:160),x,'PCHIP','extrap');
GV5(:,4)=interp1(Ned(161:189),GLA(161:189),x,'PCHIP','extrap');
GV5(:,3)=interp1(Ned(190:217),GLA(190:217),x,'PCHIP','extrap');
GV5(:,2)=interp1(Ned(218:239),GLA(218:239),x,'PCHIP','extrap');
GV5(:,1)=interp1(Ned(240:254),GLA(240:254),x,'PCHIP','extrap');

82
%Create new interpolated mean Pressure values based on Ned
GV5STR(:,12)=interp1(Ned(1:17),G5ST(1:17),x,'PCHIP');
GV5STR(:,11)=interp1(Ned(18:34),G5ST(18:34),x,'PCHIP');
GV5STR(:,10)=interp1(Ned(35:50),G5ST(35:50),x,'PCHIP');
GV5STR(:,9)=interp1(Ned(51:69),G5ST(51:69),x,'PCHIP');
GV5STR(:,8)=interp1(Ned(70:90),G5ST(70:90),x,'PCHIP');
GV5STR(:,7)=interp1(Ned(91:109),G5ST(91:109),x,'PCHIP');
GV5STR(:,6)=interp1(Ned(110:134),G5ST(110:134),x,'PCHIP');
GV5STR(:,5)=interp1(Ned(135:160),G5ST(135:160),x,'PCHIP');
GV5STR(:,4)=interp1(Ned(161:189),G5ST(161:189),x,'PCHIP');
GV5STR(:,3)=interp1(Ned(190:217),G5ST(190:217),x,'PCHIP');
GV5STR(:,2)=interp1(Ned(218:239),G5ST(218:239),x,'PCHIP');
GV5STR(:,1)=interp1(Ned(240:254),G5ST(240:254),x,'PCHIP');

%Create new interpolated Pressure amplitude values based on RPM

GV4S(:,12)=interp1(RPM(1:17),G4A(1:17),Speed,'PCHIP');
GV4S(:,11)=interp1(RPM(18:34),G4A(18:34),Speed,'PCHIP');
GV4S(:,10)=interp1(RPM(35:50),G4A(35:50),Speed,'PCHIP');
GV4S(:,9)=interp1(RPM(51:69),G4A(51:69),Speed,'PCHIP');
GV4S(:,8)=interp1(RPM(70:90),G4A(70:90),Speed,'PCHIP');
GV4S(:,7)=interp1(RPM(91:109),G4A(91:109),Speed,'PCHIP');
GV4S(:,6)=interp1(RPM(110:134),G4A(110:134),Speed,'PCHIP');
GV4S(:,5)=interp1(RPM(135:160),G4A(135:160),Speed,'PCHIP');
GV4S(:,4)=interp1(RPM(161:189),G4A(161:189),Speed,'PCHIP');
GV4S(:,3)=interp1(RPM(190:217),G4A(190:217),Speed,'PCHIP');
GV4S(:,2)=interp1(RPM(218:239),G4A(218:239),Speed,'PCHIP');
GV4S(:,1)=interp1(RPM(240:254),G4A(240:254),Speed,'PCHIP');

%Create new interpolated Pressure amplitude values based on RPM

GV5S(:,12)=interp1(RPM(1:17),GLA(1:17),Speed,'PCHIP');
GV5S(:,11)=interp1(RPM(18:34),GLA(18:34),Speed,'PCHIP');
GV5S(:,10)=interp1(RPM(35:50),GLA(35:50),Speed,'PCHIP');
GV5S(:,9)=interp1(RPM(51:69),GLA(51:69),Speed,'PCHIP');
GV5S(:,8)=interp1(RPM(70:90),GLA(70:90),Speed,'PCHIP');
GV5S(:,7)=interp1(RPM(91:109),GLA(91:109),Speed,'PCHIP');
GV5S(:,6)=interp1(RPM(110:134),GLA(110:134),Speed,'PCHIP');
GV5S(:,5)=interp1(RPM(135:160),GLA(135:160),Speed,'PCHIP');
GV5S(:,4)=interp1(RPM(161:189),GLA(161:189),Speed,'PCHIP');
GV5S(:,3)=interp1(RPM(190:217),GLA(190:217),Speed,'PCHIP');
GV5S(:,2)=interp1(RPM(218:239),GLA(218:239),Speed,'PCHIP');
GV5S(:,1)=interp1(RPM(240:254),GLA(240:254),Speed,'PCHIP');

%Create new interpolated Pressure amplitude values based on RPM

GV6S(:,12)=interp1(RPM(1:17),G6A(1:17),Speed,'PCHIP');
GV6S(:,11)=interp1(RPM(18:34),G6A(18:34),Speed,'PCHIP');
GV6S(:,10)=interp1(RPM(35:50),G6A(35:50),Speed,'PCHIP');
GV6S(:,9)=interp1(RPM(51:69),G6A(51:69),Speed,'PCHIP');
GV6S(:,8)=interp1(RPM(70:90),G6A(70:90),Speed,'PCHIP');
GV6S(:,7)=interp1(RPM(91:109),G6A(91:109),Speed,'PCHIP');
GV6S(:,6)=interp1(RPM(110:134),G6A(110:134),Speed,'PCHIP');
GV6S(:,5)=interp1(RPM(135:160),G6A(135:160),Speed,'PCHIP');
GV6S(:,4)=interp1(RPM(161:189),G6A(161:189),Speed,'PCHIP');
GV6S(:,3)=interp1(RPM(190:217),G6A(190:217),Speed,'PCHIP');
GV6S(:,2)=interp1(RPM(218:239),G6A(218:239),Speed,'PCHIP');
GV6S(:,1)=interp1(RPM(240:254),G6A(240:254),Speed,'PCHIP');

%Include the calibration constant to convert voltage change to pressure

change in kpa
RGV4=GV4*a1;
RGV5=GV5*a2;
RGV6=GV6*a3;

83
RGV4S=GV4S*a1;
RGV5S=GV5S*a2;
RGV6S=GV6S*a3;

%For BEP=GV10 0,1785NED

GALL=RGV4+RGV5+RGV6;
GV4=GV4/G4A(1,42);
GV5=GV5/GLA(1,42);
GV6=GV6/G6A(1,42);
GALLS=GV4S+GV5S+GV6S;
GV4S=GV4S/G4A(1,42);
GV5S=GV5S/GLA(1,42);
GV6S=GV6S/G6A(1,42);

BEPGV4=G4A(1,42)*a1;
BEPGV5=GLA(1,42)*a2;
BEPGV6=G6A(1,42)*a3;

%Interpolate values for plotting based on NED and QED

N=zeros(numb,numb);
Q=N;
E=N;
for t=1:numb
N(:,t)=x;
Q(t,:)=linspace(0.207,0.02,numb);
end
for t=1:numb
EFF(t,:)=interp1(q(t,:),ef(t,:),Q(t,:),'PCHIP',0);
RRPM(t,:)=interp1(q(t,:),r(t,:),Q(t,:),'PCHIP',0);
GALLP(t,:)=interp1(q(t,:),GALL(t,:),Q(t,:),'PCHIP',0);
GV4P(t,:)=interp1(q(t,:),GV4(t,:),Q(t,:),'PCHIP',0);
GV5P(t,:)=interp1(q(t,:),GV5(t,:),Q(t,:),'PCHIP',0);
GV6P(t,:)=interp1(q(t,:),GV6(t,:),Q(t,:),'PCHIP',0);
Stress4(t,:)=interp1(q(t,:),GV4STR(t,:),Q(t,:),'PCHIP',0);
Stress5(t,:)=interp1(q(t,:),GV5STR(t,:),Q(t,:),'PCHIP',0);
Stress6(t,:)=interp1(q(t,:),GV6STR(t,:),Q(t,:),'PCHIP',0);
RGV4P(t,:)=interp1(q(t,:),RGV4(t,:),Q(t,:),'PCHIP',0);
RGV5P(t,:)=interp1(q(t,:),RGV5(t,:),Q(t,:),'PCHIP',0);
RGV6P(t,:)=interp1(q(t,:),RGV6(t,:),Q(t,:),'PCHIP',0);
end

%Interpolate values for plotting based on RPM and QED

NS=zeros(numb,numb);
QS=NS;
ES=NS;

for t=1:numb
NS(:,t)=Speed;
QS(t,:)=linspace(0.207,0.02,numb);
end
for t=1:numb
GV4PS(t,:)=interp1(qS(t,:),GV4S(t,:),QS(t,:),'PCHIP',0);
GV5PS(t,:)=interp1(qS(t,:),GV5S(t,:),QS(t,:),'PCHIP',0);
GV6PS(t,:)=interp1(qS(t,:),GV6S(t,:),QS(t,:),'PCHIP',0);
RGV4PS(t,:)=interp1(qS(t,:),RGV4S(t,:),QS(t,:),'PCHIP',0);
RGV5PS(t,:)=interp1(qS(t,:),RGV5S(t,:),QS(t,:),'PCHIP',0);
RGV6PS(t,:)=interp1(qS(t,:),RGV6S(t,:),QS(t,:),'PCHIP',0);

84
end

for i=1:1000
for k=1:1000
if EFF(i,k)< 10
GV4P(i,k)=0;
GV5P(i,k)=0;
GV6P(i,k)=0;
RGV4P(i,k)=0;
RGV5P(i,k)=0;
RGV6P(i,k)=0;
end
end
end

%Beyond this point there is only code to create different plots

%Decide value of lines in the hill chart

%Efflines=[10 25 40 60 70 75 80 85 90 91 92 93 93.3 93.6 93.8 93.85 93.88
93.903 93.95 94 95];
%Speedlines=[210 220 230 240 250 260 270 280 290 300 310 320 330 340 350
360 370 380 390 400 410 420 430 450 470 490 510 530 550 570 600];
%Pressurelines=[0.1 0.4 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5 1.7 1.9 2.2 2.4
2.6 2.8 3 3.3 4 5 5.6 6 8 10 20];
%SmallPL=[0.5 0.6 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5 1.6 2 3 4];
%PressAllLines=[3 3.3 3.6 3.9 4 4.2 4.4 4.6 4.8 5 5.2 5.5 6 7 10 15];
%StressLines=[120 125 130 135 140 145 150 155 160 165 170 175 180 185 190
195 200 205 210];
%Plotting hill chart
%contourf(N,Q,EFF,Efflines)
%Plotting wierd speed chart
% contourf(N,Q,RRPM,Speedlines)

%Plotting pressure diagrams

%contourf(N,Q,GV4P,Pressurelines)
%contourf(N,Q,GV5P,Pressurelines)
%contourf(N,Q,GV6P,Pressurelines)
%contourf(N,Q,GV4P,SmallPL)
%contourf(N,Q,GV5P,SmallPL)
%contourf(N,Q,GV6P,SmallPL)
%1:17 18:34 35:50 51:69 70:90 91:109 110:134 135:160 161:189 190:217
%218:239 239:254
%Plotting guide vane angles
hold on;
plot(Ned(1:17),Qed(1:17),'-k');
plot(Ned(18:34),Qed(18:34),'-k');
plot(Ned(35:50),Qed(35:50),'-k');
plot(Ned(51:69),Qed(51:69),'-k');
plot(Ned(70:90),Qed(70:90),'-k');
plot(Ned(91:109),Qed(91:109),'-k');
plot(Ned(110:134),Qed(110:134),'-k');
plot(Ned(135:160),Qed(135:160),'-k');
plot(Ned(161:189),Qed(161:189),'-k');
plot(Ned(190:217),Qed(190:217),'-k');
plot(Ned(218:239),Qed(218:239),'-k');
plot(Ned(240:254),Qed(240:254),'-k');

%Plotting guide vane angles(RPM)

%hold on;
%plot(RPM(1:17),Qed(1:17),'-k');
%plot(RPM(18:34),Qed(18:34),'-k');

85
%plot(RPM(35:50),Qed(35:50),'-k');
%plot(RPM(51:69),Qed(51:69),'-k');
%plot(RPM(70:90),Qed(70:90),'-k');
%plot(RPM(91:109),Qed(91:109),'-k');
%plot(RPM(110:134),Qed(110:134),'-k');
%plot(RPM(135:160),Qed(135:160),'-k');
%plot(RPM(161:189),Qed(161:189),'-k');
%plot(RPM(190:217),Qed(190:217),'-k');
%plot(RPM(218:239),Qed(218:239),'-k');
%plot(RPM(240:254),Qed(240:254),'-k');

xlim([0.12 0.28]);
ylim([0.02 0.207])
% Create ylabel
ylabel('Q_E_D');

% Create xlabel
xlabel('N_E_D');

% Create title
%title({'Pressure Pulsation Diagram for GV6'});

RunNed(1)=Ned(17);
RunNed(2)=Ned(34);
RunNed(3)=Ned(50);
RunNed(4)=Ned(69);
RunNed(5)=Ned(90);
RunNed(6)=Ned(109);
RunNed(7)=Ned(134);
RunNed(8)=Ned(160);
RunNed(9)=Ned(189);
RunNed(10)=Ned(217);
RunNed(11)=Ned(239);
RunNed(12)=Ned(254);

RunQed(1)=Qed(17);
RunQed(2)=Qed(34);
RunQed(3)=Qed(50);
RunQed(4)=Qed(69);
RunQed(5)=Qed(90);
RunQed(6)=Qed(109);
RunQed(7)=Qed(134);
RunQed(8)=Qed(160);
RunQed(9)=Qed(189);
RunQed(10)=Qed(217);
RunQed(11)=Qed(239);
RunQed(12)=Qed(254);
%Plotting runaway line
plot(RunNed(1:12),RunQed(1:12),'-k');

%Plotting pressure pulsations based on RPM and guide vane opening

%for i=1:12
%hold on
%plot3(i.*ones(1,1000),Speed/3.389,GV4S(:,i));
%end

86
 Stress levels
%run exceltilhill.m first

Kof=323.8095238;
%1000 to convert from KPa to MPa
Mises4M=Stress4*Kof/1000;
Mises4A=RGV4P*Kof/1000;
Mises5M=Stress5*Kof/1000;
Mises5A=RGV5P*Kof/1000;
Mises6M=Stress6*Kof/1000;
Mises6A=RGV6P*Kof/1000;
ATM=98.5*Kof/1000;

MisesAVG=(Mises4A+Mises5A+Mises6A)/3;

MisesAVGMean=(Mises4M+Mises5M+Mises6M)/3;

MisesAmp=[0.1 0.2 0.3 0.4 0.45 0.5 0.55 0.6 0.65 0.7 0.8 0.9 1 1.1 1.2 1.4
1.7 2 3];
MisesMean=[35 40 42 44 46 48 50 52 54 56 58 60 62 64 67 70 75 80];
%contourf(N,Q,MisesAVG,MisesAmp);
%contourf(N,Q,MisesAVGMean,MisesMean);
UTS=910;
KT=1.36; %strain intensity factor

for i=1:1000
for k=1:1000
StressEff4(i,k)=Mises4A(i,k)*((UTS)/((UTS)-Mises4M(i,k)));
StressEff5(i,k)=Mises5A(i,k)*((UTS)/((UTS)-Mises5M(i,k)));
StressEff6(i,k)=Mises6A(i,k)*((UTS)/((UTS)-Mises6M(i,k)));
StressEffAVG(i,k)=MisesAVG(i,k)*((UTS)/((UTS)-MisesAVGMean(i,k)));
end
end

for i=1:1000
for k=1:1000
if Mises4M(i,k)==0;
ATME=0;
else
ATME=ATM;
end
StressEff4LCF(i,k)=(Mises4M(i,k)+Mises4A(i,k)-ATME)*((UTS)/((UTS)-ATM));
if Mises5M(i,k)==0;
ATME=0;
else
ATME=ATM;
end
StressEff5LCF(i,k)=(Mises5M(i,k)+Mises5A(i,k)-ATME)*((UTS)/((UTS)-ATM));
if Mises6M(i,k)==0;
ATME=0;
else
ATME=ATM;
end
StressEff6LCF(i,k)=(Mises6M(i,k)+Mises6M(i,k)-ATME)*((UTS)/((UTS)-ATM));
if MisesAVGMean(i,k)==0;
ATME=0;
else

87
 ATME=ATM;
end
StressEffAVGLCF(i,k)=(MisesAVGMean(i,k)+MisesAVG(i,k)-ATME)*((UTS)/((UTS)-
ATM));
end
end

SLevels=[0.2 0.4 0.6 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5 1.6 1.8 2];
%contourf(N,Q,StressEff4,SLevels)

k1HCF=13.62;
m1HCF=5;
for i=1:1000;
for k=1:1000;
CF4HCF(i,k)=10^(k1HCF-m1HCF*log10(2*KT*StressEff4(i,k)));
CF5HCF(i,k)=10^(k1HCF-m1HCF*log10(2*KT*StressEff5(i,k)));
CF6HCF(i,k)=10^(k1HCF-m1HCF*log10(2*KT*StressEff6(i,k)));
CFAVGHCF(i,k)=10^(k1HCF-m1HCF*log10(2*KT*StressEffAVG(i,k)));
if CF4HCF(i,k)>10^50
CF4HCF(i,k)=0;
end
if CF5HCF(i,k)>10^50
CF5HCF(i,k)=0;
end
if CF6HCF(i,k)>10^50
CF6HCF(i,k)=0;
end
if CFAVGHCF(i,k)>10^50
CFAVGHCF(i,k)=0;
end
end
end

k1LCF=10.97;
m1LCF=3;
for i=1:1000;
for k=1:1000;
CF4LCF(i,k)=10^(k1LCF-m1LCF*log10(KT*StressEff4LCF(i,k)));
CF5LCF(i,k)=10^(k1LCF-m1LCF*log10(KT*StressEff5LCF(i,k)));
CF6LCF(i,k)=10^(k1LCF-m1LCF*log10(KT*StressEff6LCF(i,k)));
CFAVGLCF(i,k)=10^(k1LCF-m1LCF*log10(KT*StressEffAVGLCF(i,k)));

if CFAVGLCF(i,k)>10^50
CFAVGLCF(i,k)=0;
end

LCFDPC(i,k)=1/CFAVGLCF(i,k);
end
end
CyclesToFailure=[10 10^2 10^3 5*10^3 10^4 5*10^4 10^5 2*10^5 3*10^5 4*10^5
5*10^5 6*10^12 10^6 5*10^6 10^7 10^8]*10^7;
ScaledCyclesToFailure=[0.1 1 5 10 50 100 500 1000 2000 3000 4000 5000 6000
8000 10^4 10^5 2*10^5 5*10^5];
LCFCyclesToFailure=[1 5 10 20 30 40 50 60 70 80 90 10^2 120 140 160 180
200 250 300 350 10^3]*10^5;

%contourf(N,Q,CF4HCF,CyclesToFailure)
%contourf(N,Q,CF5HCF,CyclesToFailure)
%contourf(N,Q,CF6HCF,CyclesToFailure)
%contourf(N,Q,CF4LCF,LCFCyclesToFailure)

88
%contourf(N,Q,CF5LCF,LCFCyclesToFailure)
%contourf(N,Q,CF6LCF,LCFCyclesToFailure)
%contourf(N,Q,CFAVGHCF/(10^9),ScaledCyclesToFailure)
%contourf(N,Q,CFAVGHCF/(10^3),ScaledCyclesToFailure)

CF4LCFScaled=CF4LCF/(10^6);
CF5LCFScaled=CF5LCF/(10^6);
CF6LCFScaled=CF6LCF/(10^6);
CFAVGLCFScaled=CFAVGLCF/(10^6);

ScaledLCFCyclesToFailure=[0.2 0.4 0.6 0.8 1 1.2 1.4 2 3 4 5 6 7 8 9 10 11

12 13 14 20 30 40 50 60 70 80 90 10^2 120 140 160 180 200 ];
%contourf(N,Q,CF4LCFScaled,ScaledLCFCyclesToFailure)
%contourf(N,Q,CF5LCFScaled,ScaledLCFCyclesToFailure)
%contourf(N,Q,CF6LCFSclaed,ScaledLCFCyclesToFailure)
%contourf(N,Q,CFAVGLCFScaled,ScaledLCFCyclesToFailure)

RelLCFDamage=[0.001 0.01 0.05 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2
1.4 1.7 2 3 4 5 6 7 8 9 10 20 30 40 50];
%contourf(N,Q,LCFDPC,RelLCFDamage*10^-7)
%contourf(N,Q,LCFDPC/(3.487*10^-7),RelLCFDamage)
for i=1:1000;
for k=1:1000;
DPH(i,k)=1680*RRPM(i,k)/CFAVGHCF(i,k);

end
end

for i=1:1000;
for k=1:1000;
if isinf(DPH(i,k))==1
DPH(i,k)=0;
end
if isinf(LCFDPC(i,k))==1
LCFDPC(i,k)=0;
end

end
end

DPHScaled=DPH*10^6;
DamageNumber=[0.01 0.06 0.08 0.1 0.2 0.3 0.4 0.5 1 5 10 50 200 400 600 800
2000];
%contourf(N,Q,DPHScaled/0.142,DamageNumber)
RelDamage=[0.01 0.05 0.1 0.2 0.3 0.4 0.6 0.8 1 1.2 1.6 2 5 10 50 100 200
300 1000 2000 10000];
set(axes,'BoxStyle','full','FontSize',16,'Layer','top','XGrid','on','YGrid'
,'on');
hold on;
contourf(N,Q,DPHScaled/0.142,RelDamage)

xlabel('N_E_D - Dimensionless Speed Factor')

ylabel('Q_E_D - Dimensionless Discharge Factor')

for i=1:1000;
for k=1:1000;
HoursToOneStartStop(i,k)=CFAVGHCF(i,k)/(RRPM(i,k)*1680*CFAVGLCF(i,k));

89
if isinf(HoursToOneStartStop(i,k))==1
HoursToOneStartStop(i,k)=0;
end
end
end

Hours=[0.001 0.01 0.05 0.1 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 2.2 2.5 2.8
3 3.5 4 5 6 7 8 9 10];
%contourf(N,Q,HoursToOneStartStop,Hours)

for i=1:1000
for k=1:1000

DamagePerSession(i,k)=DPH(i,k)*2.456+(1/CFAVGLCF(i,k));

end
end

DPSScaled=DamagePerSession*10^6;
Dscaled=[0.001 0.01 0.05 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.4 1.7 3 5
8 10 50 100 500 1000 2000 3000];
DscaledLow=[0.001 0.01 0.05 0.10 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 0.6
0.7 0.8 0.9 1 1.4 1.7 3 5];
%contourf(N,Q,DPSScaled,Dscaled)
hold on;
plot(RunNed(1:12),RunQed(1:12),'-k');
plot(Ned(1:17),Qed(1:17),'-k');
plot(Ned(18:34),Qed(18:34),'-k');
plot(Ned(35:50),Qed(35:50),'-k');
plot(Ned(51:69),Qed(51:69),'-k');
plot(Ned(70:90),Qed(70:90),'-k');
plot(Ned(91:109),Qed(91:109),'-k');
plot(Ned(110:134),Qed(110:134),'-k');
plot(Ned(135:160),Qed(135:160),'-k');
plot(Ned(161:189),Qed(161:189),'-k');
plot(Ned(190:217),Qed(190:217),'-k');
plot(Ned(218:239),Qed(218:239),'-k');
plot(Ned(240:254),Qed(240:254),'-k');

Frequencies
i=0;
k=0;
for i=1:length(Trykk)
if length(Trykk{1,i})>190000
k=k+1;
Pressure{1,k}=Trykk{1,i};
end
end

i=0;
k=0;
for k=1:3
for i=1:length(Pressure)/3;
P{k,i}=Pressure{1,(3*i)-3+k};

90
 end
end

i=0;
for i=1:145
PC{1,i}=P{1,i};
PC{2,i}=P{2,i};
PC{3,i}=P{3,i};
PC{4,i}=HILL(5,i);
end
for i=146:168
PC{1,i}=P{1,i+1};
PC{2,i}=P{2,i+1};
PC{3,i}=P{3,i+1};
PC{4,i}=HILL(5,i);
end
for i=169:254
PC{1,i}=P{1,i+2};
PC{2,i}=P{2,i+2};
PC{3,i}=P{3,i+2};
PC{4,i}=HILL(5,i);

end

BEPVolt(1,1)=G4A(1,42);
BEPVolt(2,1)=GLA(1,42);
BEPVolt(3,1)=G6A(1,42);

A(1,1)=a1;
A(2,1)=a2;
A(3,1)=a3;

for i=1:3
FA=P{i,82};
[pxx,f]=JKwelch(FA,5000,length(FA)/40,[]);
hold on;
pxx=pxx*A(i,1);
f=60*f/PC{4,82};
plot(f,pxx)
xlim([0 70])
end

Welch
This script was made by Johannes Kverno

function [ pxx, f ] = JKwelch( Signal, Fs, window, noverlap )

% Welch [pxx, f] = JKwelch(Signal, Fs, window, noverlap)
% An attempt to create a function to do the spectral analysis of a signal

S1=sum(window);
S2=sum(window.^2);
ENBW=Fs*(S2/(S1^2));

91
[pxx,f]=pwelch(detrend(Signal),window,floor(noverlap*length(window)),[],Fs)
; %Signal, Window, noverlap, frequency, Sample rate

pxx=pxx.*ENBW;
pxx=sqrt(pxx)*sqrt(2);
end

Runaway
%Exceltilhill og FrekvensAnalyse må kjøres først
clearvars Trun TrunA
Runaway=TDMS_readTDMSFile('C:\Users\eirilo\Documents\MATLAB\testmalingerRUN
AWAY.tdms');
b=0;
for i=3:length(Runaway.data)
if length(Runaway.data{1,i})>140000
b=b+1;
TR{1,b}=Runaway.data{1,i};
end

end
RunPM=[295 308 323 340 363 368 379 380 386 390 398 408 425 432.5 441
457.5];
GVopening=[0.440 0.484 0.572 0.660 0.791 0.879 0.967 1.0111 1.055 1.143
1.231 1.407 1.758 1.9779 2.242 3.0326];
for i=1:7
Trun{1,i}=TR{1,(3*i)-2};
Trun{2,i}=TR{1,(3*i)-1};
Trun{3,i}=TR{1,(3*i)};
Trun{4,i}=RunPM(1,i);
Trun{5,i}=GVopening(1,i);
end
%1

Trun{1,8}=PC{1,255};
Trun{2,8}=PC{2,255};
Trun{3,8}=PC{3,255};
Trun{4,8}=RunPM(1,8);
Trun{5,8}=GVopening(1,8);
for i=9:13
Trun{1,i}=TR{1,(3*(i-1))-2};
Trun{2,i}=TR{1,(3*(i-1))-1};
Trun{3,i}=TR{1,(3*(i-1))};
Trun{4,i}=RunPM(1,i);
Trun{5,i}=GVopening(1,i);
end

%2
Trun{1,14}=PC{1,240};
Trun{2,14}=PC{2,240};
Trun{3,14}=PC{3,240};
Trun{4,14}=RunPM(1,14);
Trun{5,14}=GVopening(1,14);
%2.2
Trun{1,15}=TR{1,37};
Trun{2,15}=TR{1,38};

92
Trun{3,15}=TR{1,39};
Trun{4,15}=RunPM(1,15);
Trun{5,15}=GVopening(1,15);

%3
Trun{1,16}=PC{1,217};
Trun{2,16}=PC{2,217};
Trun{3,16}=PC{3,217};
Trun{4,16}=RunPM(1,16);
Trun{5,16}=GVopening(1,16);

A(1,1)=a1;
A(2,1)=a2;
A(3,1)=a3;

for m=1:3
for k=1:length(Trun)

T=Trun{1,k};

[N,e] = histcounts(T,1000);
Me = mean(T);
NumberOfPoints = sum(N);
temp=0;
conf=0.97;
SNLTRYKK(m,k)=Me;
for i=1:length(N)
temp=temp +N(i); value=temp/NumberOfPoints;
if value >=(1-conf)/2
lower=e(i);
break
end
end

for j=1:length(N)
temp=temp +N(j); value=temp/NumberOfPoints;
if value >=conf+(1-conf)/2
upper=e(j);
break
end
end

VoltAmp=(upper-lower)/2;
SNLAmp(m,k)=VoltAmp;

TrunA(m,k)=VoltAmp*A(m,1);
end
end
for i=1:length(Trun);
TrunA(4,i)=Trun{4,i};
TrunA(5,i)=Trun{5,i};

93
end

TrunAR=TrunA;
TrunAR(1,:)=TrunA(1,:)/(G4A(1,42)*A(1,1));
TrunAR(2,:)=TrunA(1,:)/(GLA(1,42)*A(2,1));
TrunAR(3,:)=TrunA(1,:)/(G6A(1,42)*A(3,1));

FA=Trun{1,4};
[pxx,f]=JKwelch(FA,5000,length(FA)/40,50);
f=f*60/Trun{4,4};
hold on;
plot(f,pxx);

for i=1:3
FA=Trun{i,4};
[pxx(:,i),f(:,i)]=JKwelch(FA,5000,length(FA)/40,50);
hold on;
pxx(:,i)=pxx(:,i)*A(i);
f(:,i)=f(:,i)*60/Trun{4,4};
plot3(i.*ones(1,length(pxx)),f(:,i),pxx(:,i));
xlim([0 70])
end

FA=Trun{1,4};
[pxx,f]=JKwelch(FA,5000,length(FA)/20,50);
f=f*60/Trun{4,4};
hold on;
plot(f,pxx);
MaxP=max(pxx)
for i=1:length(pxx)
if pxx(i,1)==MaxP

MainFreq=f(i)
end
end

for i=1:3
FA=Trun{i,1};
[pxx(:,i),f(:,i)]=JKwelch(FA,5000,length(FA)/30,50);
hold on;
pxx(:,i)=pxx(:,i)*A(i);
f(:,i)=f(:,i)*60/Trun{4,1};
plot3(i.*ones(1,length(pxx)),f(:,i),pxx(:,i));
xlim([0 70])
zlim([1 3])
end

94
95
Appendix C – Calibration reports
Calibration report for GV4 – Meas XP5

96
97
Calibration report for GV5 – Kulite XTE

98
99
Calibration report for GV6 – Kulite XTE

100
101
Calibration report for Friction Torque – TW T2

102
103
Calibration report for Generator Torque – WT T2

104
105
Eirik Lødemel Fatigue loads in a Francis turbine runner