Dynamic Loads On Francis Turbine

Dynamic loads on Francis turbines
Ingebjørg Valkvæ
Master of Science in Mechanical Engineering

Submission date: June 2016
Supervisor: Pål Tore Selbo Storli, EPT
Co-supervisor: Rakel Ellingsen, EPT
Ole Gunnar Dahlhaug, EPT
Norwegian University of Science and Technology

Department of Energy and Process Engineering
Acknowledgements
This master thesis was written at the Waterpower Laboratory at the Norwegian
University of Science and Technology (NTNU) during the spring semester 2016. The
thesis is a continuation of a project assignment carried out by the author during
the corresponding fall semester. Execution of the laboratory experiment and the
computer simulations were performed partly together with PhD Candidate Rakel
Ellingsen and Postdoc. Fellow Chirag Trivedi.
I take this opportunity to gratefully acknowledge the assistance and contributions of
several persons. First of all, I would like to thank my supervisor Pål-Tore Storli for
his valuable guidance and reassurance during troubled times this year. His scientic
approach have helped me to complete the given tasks. I would also like to thank
my co-supervisors Rakel Ellingsen and Ole Gunnar Dahlhaug for sharing their ideas
on the subject. Further thanks to master student Einar Agnalt and Chirag Trivedi.
Einar have produced the computer aided design (CAD) models used in my work, and
spent a lot of time on the laboratory preparations. Chirag conducted the necessary
computational uid dynamics (CFD) simulation and provided me with his results.
I would not have been able to nalize my work without your help. I am also very
grateful to all the other Ph.D. Candidates at the Waterpower Laboratory for their
good advices, including a special thanks to Ph.D. Candidate Carl W. Bergan for the
much needed help with computer programs and diculties in the laboratory.
Finally, I want to express my gratitude to my family and boyfriend for their endless
support throughout the time spent at NTNU. In addition, thanks to all my fellow
students in the Mechanical Engineering class and at the Waterpower Laboratory. It
has been an enjoyable journey.
Ingebjørg Valkvæ
Trondheim, 24.06.2016
i
ii
Abstract
New installations of unregulated power to the electrical grid have led to more uc-
tuations in the grid frequency. Hydropower plants with large reservoirs can provide
services to balance the frequency by increasing or reducing production in accordance
to what is needed. However, such balancing services are putting stress on the tur-
bines by exposing them to dynamic loads. In the recent years, several high head
Francis turbines have been taken out of operation due to cracks in the runner. It
is thought to be because the turbines are not designed with sucient measures to
handle all the types of dynamic loads appearing during variations in operation. Anal-
ysis of stresses caused by uid-structure interaction (FSI) is therefore of increasing
interest in the hydropower industry.
In this thesis, pressure and strain measurements have been performed on the Francis
test rig in the Waterpower Laboratory at NTNU. The results were further used as
validation of a FSI analysis conducted on the same turbine model in the software
program ANSYS. The rotational speed of the runner was decreased by 60 revolutions
per minute (RPM) over a time period of 2 seconds to illustrate how an estimated
change in grid frequency impact stresses in the turbine.
The results from the FSI simulation showed high stresses both on the leading and
trailing edge of the runner blades, where the maximum stress was found on the
trailing edge tip towards the shroud. A decrease in runner speed resulted in a
decrease in material stresses. At the end time, the maximum equivalent stress has
a magnitude of 5 MPa, which is way below the yielding limit for the material.
Validation of the CFD simulation showed sucient correlation between the measured
and simulated pressures. However, it was dicult to achieve a comparison of strain
results with the strain gage, and the FSI results should therefore be further validated
before they can be used as background for more research.
iii
iv
Sammendrag
Nye installasjoner av ikke-regulerbar elektrisitet i strømnettet har ført til større

svingninger i nettfrekvensen. Vannkraftverk med store reservoarer kan tilby tjenester
for å balansere frekvensen ved å øke eller redusere produksjonen i henhold til behovet.
Slike balansetjenester påfører spenninger i turbinene ved å utsette dem for dynamiske
belastninger. I de siste årene har ere høytrykk Francisturbiner måtte tas ut av drift
på grunn av sprekker i løpehjulet. Det antas å være fordi turbinene ikke er utformet
for å håndtere alle typer av dynamiske belastninger som opptrer i løpet av variasjoner
i driften. Analyse av spenninger forårsaket av uid-struktur interaksjon er derfor av
økende interesse.
I denne oppgaven har trykk- og deformasjonsmålinger blitt gjennomført på Francis
testriggen i Vannkraftlaboratoriet ved NTNU. Resultatene ble videre brukt som
validering av en FSI analyse utført på samme turbinmodell i programmet ANSYS.
Rotasjonshastigheten til løpehjulet ble redusert med 60 RPM i løpet av en periode på
2 sekunder for å illustrere hvordan en estimert forandring i nettfrekvensen påvirker
spenninger i turbinen.
Resultatene fra FSI simuleringen viste høye spenninger både i forkant og bakkant av
turbinbladene, hvor den maksimale belastning ble funnet på spissen av bakkanten
mot turbinringen. En reduksjon i rotasjonshastigheten resulterte i en reduksjon i
materialspenningene. Ved endetiden har den maksimal spenningen en størrelse på
5 MPa, noe som er langt under det som er ytgrensen til materialet. Validering av
CFD simuleringen viste tilstrekkelig korrelasjon mellom målt og simulert trykk. Det
var imidlertid vanskelig å oppnå en sammenligning av deformasjonsresultatene ved
hjelp av strekklappen, og FSI resultatene bør derfor ytterligere valideres før de kan
brukes som bakgrunn for videre forskning.
v
vi
Contents
List of Tables xi
List of Figures xiii
Nomenclature xv
1 Introduction 1
1.1 Objective . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2 Previous and ongoing work . . . . . . . . . . . . . . . . . . . . . . . 2
2 Theoretical background 5
2.1 Mechanical properties and failure of metals . . . . . . . . . . . . . . 5
2.1.1 Deformation . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.1.2 Cracks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.1.3 Fatigue . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.2 Energy generation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.2.1 Synchronous generators . . . . . . . . . . . . . . . . . . . . . 8
2.2.2 The Nordic system . . . . . . . . . . . . . . . . . . . . . . . . 8
2.2.3 Implementation of renewable energy sources . . . . . . . . . . 10
2.2.4 Frequency control . . . . . . . . . . . . . . . . . . . . . . . . 11
2.3 The Francis turbine . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.3.1 Main components . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.3.2 Operation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.3.3 Turbine performance . . . . . . . . . . . . . . . . . . . . . . . 17
2.3.4 Francis runner . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.3.5 Material and design . . . . . . . . . . . . . . . . . . . . . . . 18
2.4 Loads on a Francis turbine . . . . . . . . . . . . . . . . . . . . . . . 19
2.4.1 Torque oscillations . . . . . . . . . . . . . . . . . . . . . . . . 19
2.4.2 Pressure pulsations . . . . . . . . . . . . . . . . . . . . . . . . 21
2.4.3 Reduction of turbine lifetime . . . . . . . . . . . . . . . . . . 23
3 Investigation tools 25
3.1 Model testing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
3.1.1 Sensor fundamentals . . . . . . . . . . . . . . . . . . . . . . . 26
vii
3.1.2 Discrete sampling and sampling rate . . . . . . . . . . . . . . 26
3.1.3 Filtering and smoothing of data . . . . . . . . . . . . . . . . . 27
3.2 Numerical analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
3.2.1 Computational uid dynamics . . . . . . . . . . . . . . . . . 28
3.2.2 Computational structural mechanics . . . . . . . . . . . . . . 28
3.2.3 Fluid-structure interaction . . . . . . . . . . . . . . . . . . . . 29
3.2.4 Numerical errors and uncertainties . . . . . . . . . . . . . . . 30
4 Facilities and methods 33

4.1 The Waterpower Laboratory . . . . . . . . . . . . . . . . . . . . . . 33
4.1.1 Francis test rig . . . . . . . . . . . . . . . . . . . . . . . . . . 34
4.1.2 Generator parameters . . . . . . . . . . . . . . . . . . . . . . 36
4.2 Laboratory experiment . . . . . . . . . . . . . . . . . . . . . . . . . . 36
4.2.1 Instrumentation . . . . . . . . . . . . . . . . . . . . . . . . . 36
4.2.2 Calibration . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
4.2.3 Variation in rotational speed . . . . . . . . . . . . . . . . . . 40
4.2.4 Data acquisition . . . . . . . . . . . . . . . . . . . . . . . . . 42
4.3 Computer simulation . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
4.3.1 CFD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
4.3.2 FEA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
4.3.3 FSI . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
5 Uncertainty analysis 51
5.1 Uncertainties in the calibrations . . . . . . . . . . . . . . . . . . . . . 51
5.1.1 Static pressure transducers . . . . . . . . . . . . . . . . . . . 51
5.1.2 Torque . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
5.1.3 Rotational speed sensor . . . . . . . . . . . . . . . . . . . . . 53
5.2 Uncertainties in the tests . . . . . . . . . . . . . . . . . . . . . . . . 53
5.2.1 Static pressure transducers . . . . . . . . . . . . . . . . . . . 54
5.2.2 Torque transducers . . . . . . . . . . . . . . . . . . . . . . . . 54
5.2.3 Rotational speed sensor . . . . . . . . . . . . . . . . . . . . . 55
6 Results 57
6.1 Turbine torque . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
6.2 CFD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
6.2.1 Validation of CFD . . . . . . . . . . . . . . . . . . . . . . . . 58
6.3 FSI . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
6.3.1 Mesh independence test . . . . . . . . . . . . . . . . . . . . . 65
6.3.2 Validation of FSI . . . . . . . . . . . . . . . . . . . . . . . . . 67
6.3.3 FSI results . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
7 Discussion 73
7.1 Known error sources . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
7.1.1 Rotational speed functions . . . . . . . . . . . . . . . . . . . 73
7.1.2 Operation points . . . . . . . . . . . . . . . . . . . . . . . . . 74
7.1.3 Time synchronization of measurements . . . . . . . . . . . . . 75
viii
7.1.4 Sensor positions . . . . . . . . . . . . . . . . . . . . . . . . . 75
7.2 Computer simulations . . . . . . . . . . . . . . . . . . . . . . . . . . 75
7.2.1 Time step . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
7.2.2 y+ values from CFD simulation . . . . . . . . . . . . . . . . . 75
7.2.3 Computational model geometry . . . . . . . . . . . . . . . . . 76
7.2.4 Mesh independence test . . . . . . . . . . . . . . . . . . . . . 76
7.3 FSI . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
7.3.1 Pressure mapping . . . . . . . . . . . . . . . . . . . . . . . . 77
7.3.2 Deformation of runner . . . . . . . . . . . . . . . . . . . . . . 77
8 Conclusion 79
9 Further work 81
APPENDICES I
A Risk assessment I
B Sensor locations and BEP from Francis-99 XIV
C Error analysis XVI
C.1 Determination of uncertainties in model tests . . . . . . . . . . . . . XVII
D Calibration reports XIX

E Procedure for operation with variable RPM XLVIII
F LabView program for controlling RPM LII
G Additional results LIV
G.1 Mesh independence test . . . . . . . . . . . . . . . . . . . . . . . . . LV
G.2 von-Mises yield criterion . . . . . . . . . . . . . . . . . . . . . . . . . LV
G.3 Validation of CFD . . . . . . . . . . . . . . . . . . . . . . . . . . . . LVI
ix
x
List of Tables
4.1 Mechanical properties of the Tokke runner [12] . . . . . . . . . . . . 35

4.2 Technical generator parameters [37] . . . . . . . . . . . . . . . . . . . 36
4.3 Measuring equipment . . . . . . . . . . . . . . . . . . . . . . . . . . 38
4.4 Measuring equipment [37] . . . . . . . . . . . . . . . . . . . . . . . . 39
4.5 Boundary conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
4.6 Mesh independence test . . . . . . . . . . . . . . . . . . . . . . . . . 47
5.1 Total calibration uncertainties of pressure transducers . . . . . . . . 52

5.2 95% condence interval uncertainties . . . . . . . . . . . . . . . . . . 54
6.1 Probe results from dierent surfaces/edges on a runner blade . . . . 71
7.1 Operating points . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74

7.2 y+ values for the CFD model runner . . . . . . . . . . . . . . . . . . 76
C.1 Component errors in the calibration of an instrument . . . . . . . . . XVII

C.2 Component errors in the test . . . . . . . . . . . . . . . . . . . . . . XVIII
xi
xii
List of Figures
2.1 Engineering stress-strain-curve . . . . . . . . . . . . . . . . . . . . . 6

2.2 Coordinating production and consumption [43] . . . . . . . . . . . . 10
2.3 Number of minutes outside 49.9 - 50.1 Hz per week [43] . . . . . . . 11
2.4 How dierent reserves balance the system . . . . . . . . . . . . . . . 14
2.5 Illustration of a Francis turbine [33] . . . . . . . . . . . . . . . . . . 15
2.6 Velocity diagrams at the inlet and outlet of the runner [19] . . . . . 17
2.7 Net torque on the turbine [25] . . . . . . . . . . . . . . . . . . . . . . 20
2.8 Flow eld at inlet of runner [34] . . . . . . . . . . . . . . . . . . . . . 22
2.9 Statistics of cracks in Francis runners from 1960 to 2000 [22] . . . . 24
3.1 Measurement sensors . . . . . . . . . . . . . . . . . . . . . . . . . . . 26

3.2 Attempt at smoothing data using the Savitzky-Golay method [8] . . 27
4.1 A section of the open loop system at NTNU . . . . . . . . . . . . . . 34

4.2 Dimensions of the Tokke model runner . . . . . . . . . . . . . . . . . 35
4.3 Onboard pressure sensors mounted into the runner hub . . . . . . . 37
4.4 Pressure sensors in the vaneless space and in the draft tube cone . . 38
4.5 Pressure sensors in pipe line . . . . . . . . . . . . . . . . . . . . . . . 38
4.6 Sensors for calculating running point . . . . . . . . . . . . . . . . . . 39
4.7 Strain gage attached to one of the runner blades . . . . . . . . . . . 40
4.8 Variation in RPM . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
4.9 Schematic representation of RPM disturbance system . . . . . . . . 41
4.10 Wet CAD model of the Tokke turbine . . . . . . . . . . . . . . . . . 43
4.11 Locations of pressure sensors in the CFD model . . . . . . . . . . . . 45
4.12 Structural CAD model of the Tokke runner . . . . . . . . . . . . . . 46
4.13 Conditions and loads on structural model . . . . . . . . . . . . . . . 48
5.1 Condence interval of repeated measurement . . . . . . . . . . . . . 53
6.1 Correlation between rotational speed, turbine torque and power output 57
6.2 Comparison of pressure results from sensor PT02 . . . . . . . . . . . 59
6.3 Sensors mounted in the runner hub . . . . . . . . . . . . . . . . . . . 60
xiii
6.9 Mesh independence test - Leading edge . . . . . . . . . . . . . . . . . 66
6.10 Mesh on runner blade . . . . . . . . . . . . . . . . . . . . . . . . . . 67
6.11 Strain gage results from laboratory measurement . . . . . . . . . . . 67
6.12 Strain path . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
6.13 Imported pressure in ANSYS Mechanical . . . . . . . . . . . . . . . 69
6.14 Maximum stress on runner over time [Pa] . . . . . . . . . . . . . . . 69
6.15 Maximum stress in runner at last time step [Pa] . . . . . . . . . . . 70
6.16 Maximum stress in blades at last time step [Pa] . . . . . . . . . . . . 70
6.17 Stress distribution on pressure side of runner blade [Pa] . . . . . . . 71
6.18 Stress distribution on suction side of runner blade [Pa] . . . . . . . . 71
7.1 Rotational speed functions . . . . . . . . . . . . . . . . . . . . . . . . 74
B.1 Coordinates of pressure sensors . . . . . . . . . . . . . . . . . . . . . XIV

B.2 Coordinates of strain gage sensor . . . . . . . . . . . . . . . . . . . . XV
B.3 Best eciency point from the second Francis-99 workshop [2] . . . . XV
G.1 Element metrics showing the quality of the mesh elements . . . . . . LIV
G.2 Mesh independence test - Max stress . . . . . . . . . . . . . . . . . . LV
G.3 Comparison of pressure results from sensor PT12 . . . . . . . . . . . LVI
G.4 Comparison of pressure results from sensor PT14 . . . . . . . . . . . LVII
xiv
Nomenclature
Abbreviations
3D Three-Dimensional
BEP Best Eciency Point
CAD Computer Aided Design
CFD Computational Fluid Dynamics
FEA Finite Element Analysis
FEM Finite Element Method
FSI Fluid-Structure Interaction
HCF High Cycle Fatigue
LCF Low Cycle Fatigue
NTNU Norwegian University of Science and Technology
RPM Revolutions Per Minute
RSI Rotor-Stator Interaction
TSO Transmission System Operator
Greek symbols
δ Magnetic torque angle ◦
Strain -
η Eciency %
γ Guide vane angle ◦
µ Kinematic viscosity Ns/m2

ω Angular velocity 1/s
ρ Density kg/m3
xv
σ Stress Pa
Other symbols
bp Permanent speed droop %
ur Friction velocity m/s
A Area m2
D Diameter m
E Modulus of elasticity Pa
F Force N
f Frequency Hz
g Gravity m /s
2
H Net head m
J Moment of inertia kg m2
l Length m
N Rotor speed RPM
P Power N m/s
p Pressure kPa
po Number of stator poles -
Q Discharge m /s
3
T Torque Nm
y Distance from node to wall m
y+ Non-dim. wall distance -
Subscripts
0 Original
d Dampening
el Electrical
gen Generated
h Hydraulic
i Instantaneous
load Demand
m Magnetic
xvi
nom Nominal
ran Random
sys System
tot Total
xvii
xviii
Chapter 1
Introduction
Today's energy system is characterized by large time variations in consumption and

production of electricity. This is partly due to the increasing installation of power
from renewable energy sources, like wind turbines and solar cells, to the electrical
grid. As a consequence, the grid frequency is uctuating and there is more need
for balancing services like the ones that can be provided by hydropower plants.
However, such balancing services are exposing the rotating machinery to dynamic
loads and increases the turbine stresses. These dynamic loads come in addition to
other loads caused by pressure pulsations in the waterway. Calculation of material
stresses in a turbine runner due to dynamic loads is uncertain, yet these stresses are
crucial when dimensioning the runner for operation and for prediction of runner life.
In recent years, several turbines have received major damages just after a short
period in operation [45]. This applies in particular to high head Francis turbines,
the most commonly used turbine in Norway. The reasons behind the damages are
not fully known, but dynamic loads are one of the likely causes. From the early days
of hydropower production, turbines were designed to run continuously, operating at
a rated head and discharge, while pumped-storage plants or gas turbines provided
the regulatory power. The introduction of a new energy legislation in the beginning
of the 1990s aected the energy market and hydropower operation. The current
market requires a great exibility in operation, and operation outside of the design
point happens quite often.
An understanding of uid-structure interaction in turbines have become more essen-
tial since the dierent turbine loads are mainly induced by the internal uid ow.
Fluid phenomena like rotor-stator interaction (RSI) and draft tube vortex rope are
under investigation, but the link between uid dynamics and structural mechanics
is still not well established. Safe operation of a plant requires that the machine com-
ponents are optimised with respect to both static and dynamic loading in the design
phase. Accordingly, the numerical methods for prediction and analysis of dynamic
loads must be improved.
1
1.1. OBJECTIVE
1.1 Objective
This master thesis is a continuation of the author's project thesis aiming for a FSI
analysis of a Francis turbine. Lack of time to perform PIV measurements, together
with problems with execution of the CFD simulation, led to alterations in the original
tasks of the master thesis. This did not change the purpose of the work and therefore
did not require any large rearrangements. A new objective and new tasks were made
in accordance with supervisor Pål-Tore Storli and are stated below.
The objective of the master thesis is to perform a FSI analysis of a Francis runner
model subjected to variable speed of rotation, as well as pressure and strain mea-
surements on the physical turbine model. The model is installed at the Waterpower
laboratory at NTNU and is the subject of much research to come.
The following tasks are to be considered:
1. Literature study on dynamic loads in a Francis turbine, in addition to pressure
and strain measurement techniques.
2. Perform transient pressure and strain measurements in the Francis test rig at
chosen operating conditions.
3. Perform a FSI simulation using test conditions from the measurements as
inputs.
4. Validate the simulation with the model measurements.
1.2 Previous and ongoing work
Unsteady ow in Francis turbines have been investigated for over 50 years. Dierent
aspects have been studied, trying to determine what are causing the various load
phenomena, how turbine material reacts and how further damage can be prevented
through better design and operation restrictions. The researchers have mainly fo-
cused on pressure uctuations in the ow path during steady state operation, such as
those caused by rotor-stator interaction and vortex rope in the draft tube. However,
since the operation of hydropower plants have changed after the introduction of the
new energy law in 1991, dynamic loads and fatigue in Francis turbines have become
a major eld of study.
In 2010, Frunzaverde et al. [27] presented a failure analysis of a broken Francis
turbine runner blade. The failure had appeared just some months after a welding
repair work on a fatigue crack initiated near the trailing egde at the junction with
the hub. They conducted a metallographic investigation on a piece of the damaged
blade and concluded that the cracking of the blade was caused by improper welding
conditions and high material stresses. In addition to this investigation, they per-
formed numerical computations with static nite element analysis (FEA) in order to
2
1.2. PREVIOUS AND ONGOING WORK
evaluate the deformation and stress distribution on the blade. In 2012, Seidel et al.
[40] evaluated RSI-induced stresses in Francis runners. They describe the equipment
and procedure of strain gage measurements in Francis prototype runners, and how
this is dependent on head and specic speed. The measurement data was compared
with simulations of Francis runner dynamics to predict dynamic stresses caused by
RSI. Furthermore, in 2014, Seidel et al. [41] published a paper called "Dynamic
load in Francis runners and their impact on fatigue life". Summarized ndings of
recent investigations through experiments and CFD analysis enabling Francis run-
ners which combine high eciency and a robust mechanical design, were presented.
Luna-Ramírez et al. [35] performed a failure analysis of moving blades in a Francis
runner in 2015. The analysis consisted of determination of pressure on the blade
surface using CFD, calculation of stress distribution in the runner at dierent operat-
ing conditions with the nite element method (FEM), and a simple fatigue analysis.
Their results showed a large concentration of stresses in the T-joint between the
blade and the hub during steady state operation at dierent loads.
The Waterpower Laboratory at NTNU have a long history on hydropower research,
and over 400 master thesis have been written concerning dierent hydropower issues
and developments. Kobro [34] wrote his PhD thesis on "Measurement of pressure
pulsations in Francis turbines" in 2010. The conducted measurements were per-
formed by means of onboard measuring equipment both in model runners and full-
scale prototype runners. Resulting data sets were used to investigate the dynamic
pressure and strain in the runners. The analysis results from both model and proto-
type runner showed that the wake leaving the guide vanes is the most severe source
of dynamic pressure in the runner. Even though the draft tube vortex rope pulsa-
tion propagates upstream the runner, it did not appear as a signicant frequency in
the strain measurements. In 2012, Bergmann-Paulsen [18] carried out a static FSI
analysis on selected turbine components to study sediment erosion and develop an
improved turbine design. The results showed a stress distribution which coincided
with the energy transfer along the runner blade. In addition to that for dierent
designs, the stress was relatively low compared to the criteria for hydraulic turbines.
The following year, Hovland [32] investigated pressure pulsations and stress in a high
head Francis model turbine. She performed model tests on the Tokke turbine with
simultaneously measurements of pressure at dierent parts of the waterway. Then,
spectral analysis was performed on all measurements to map out frequencies and
corresponding amplitudes at various load conditions. RSI induced pulsations dom-
inated the vaneless space and runner channels. Pressure pulsations was dampened
across the runner, and sensors at the trailing edge displayed large values compared
to the sensors positioned at the outlet. In 2014, Haga [29] conducted pressure mea-
surements on the Tokke runner during transient start and stop procedures. A Fast
Fourier transform analysis was used to identify pressure oscillations. The results
showed that the largest frequency experienced was the blade passing frequency dur-
ing start/stop and operation at best eciency point(BEP) and part load.
PhD Candidate Rakel Ellingsen [25] and Associate Professor Pål-Tore Storli [45],
are currently investigating how today's variable grid frequency impact components
3
1.2. PREVIOUS AND ONGOING WORK
in hydropower plants with regard to dynamic loads. They have conducted eld mea-
surements and simulations in MATLAB on torque oscillations to calculate stresses
in a Francis turbine. Simulation results gave a torque variation in the range of 3-5%
of the set point value, meaning a 5% increase in the dynamic loads imparted on
the turbine. The work is a part of a longer term goal, namely to identifying stress
oscillations in a Francis runner operating at oscillating speed of rotation due to grid
frequency variations. Furthermore, Postdoc. Fellow Chirag Trivedi is performing a
uid-structure analysis of a high head Francis turbine as a part of the Francis-99
project [2] [49]. Detailed mechanical analysis on the runner will be carried out un-
der dierent operating conditions. Load variation, start-stop, total load rejection,
and spin-no-load caused by penetration of intermittent power into the power grid
network summarizes the transients being studied. Trivedi is a former PhD Candi-
date from Luleå University of Technology in Sweden, where he wrote his thesis on
"Investigations of transient pressure loading on a high head Francis turbine".
4
Chapter 2
Theoretical background
2.1 Mechanical properties and failure of metals
Many materials are subjected to forces or loads when in service. In such situations it
is necessary to know the characteristics of the material being used and to design the
related component so that any excessive deformation and fracture are prevented. A
material's mechanical behaviour is reected by the relationship between an applied
force/load and deformation. There are three ways a load may be applied: tension,
compression, and shear [23]. Its magnitude may be constant over time or uctuate
continuously. To compare specimens of dierent sizes, load is calculated per unit
area, also called normalization to the area. Force divided by area is dened as stress
[30]. In tension and compression tests, the relevant area is that perpendicular to
the force, while in shear or torsion tests, the area is perpendicular to the axis of
rotation. As a result of a tensile or compressive stress, a change in dimensions or
deformation elongation occur. To enable comparison of specimens with dierent
length, the elongation is also normalized, but to the length L. This is dened as
strain. Equations of stress, σ , and strain, , are given below:
F
σ= (2.1)
A0
li − l0 ∆l
= = (2.2)
l0 l0
F is here the instantaneous force applied perpendicular to the cross section, A0 is

the original cross-sectional area prior to an applied load, l0 is the original length
of the material, and li is the instantaneous length. Strain can be positive (tensile)
or negative (compression), and is sometimes expressed in units such as mm/mm
although it is dimensionless [23].
5
2.1. MECHANICAL PROPERTIES AND FAILURE OF METALS
2.1.1 Deformation
Two types of material deformation exists; elastic and plastic. Elastic deformation is
characterized by Hooke's law (Equation (2.3)), where stress and strain are propor-
tional to each other and result in a linear relationship in a plot [23].
σ =E· (2.3)
E is the Modulus of Elasticity or Young's modulus, and a large E reects a sti

material. With this type of deformation, the material will resume its original form
when the applied force is removed. The graph in Figure 2.1 represents the rela-
tionship between stress and strain. Past the point of yielding (Y), stress and strain
are no longer proportional. Permanent, non-reversible plastic deformation strikes.
The stress increases to continue the plastic deformation to a maximum, and then
decrease to a fracture. Tensile strength (TS) corresponds to the maximum stress;
if the stress is held at this point, fracture will occur. The results are dependent on
variables such as the material's composition, microscopic imperfections, the way the
component is manufactured, the rate of loading, and the temperature during the
loading [30].
Figure 2.1: Engineering stress-strain-curve
2.1.2 Cracks
Ductility is a measure of the degree of plastic deformation before fracture [23]. Duc-
tile materials often have high energy absorption and withstand substantial plastic
6
2.1. MECHANICAL PROPERTIES AND FAILURE OF METALS
deformation before they fracture. A material that have low energy adsorption and
tolerates very little or no plastic deformation before a fracture is called brittle. A
ductile fracture is preferred if a fracture should occur, because such fractures are eas-
ier to detect due to the time domain of the plastic deformation. Prevention methods
can then be performed when these types of cracks are detected. It is said to be a
"stable" crack. In contrast, a brittle fracture occurs suddenly due to rapid crack
propagation and are therefore called "unstable" cracks.
The measured fracture strengths for the majority of brittle materials are signicantly
lower than those predicted by theoretical calculations based on atomic bonding ener-
gies. This is because of the existence of microscopic aws or cracks that exist under
normal conditions at the surface or within the body of a material. A stress may
be amplied or concentrated at a crack tip, and these aws are therefore sometimes
called stress raisers. However, stress amplication is not restricted to microscopic
defects; it may occur at macroscopic internal discontinuities, at sharp corners, and
at notches in large structures.
2.1.3 Fatigue
Fatigue is fracture which occurs during dynamic and variable loading. Under these
circumstances, the fracture may appear at a stress level which is lower than the
yield and tensile strength. The fracture often happens after a long period of cyclic
tension. Fatigue is estimated to be the cause of about 90% of all metal failures [23].
It occurs suddenly, without any warning, and it is cut as a brittle fracture, even for
ductile materials. There is seldom any plastic deformation before such a failure.
In order to specify a safe strength for metallic material under repeated loading,
it is necessary to determine a limit below which no failure can be detected after
applying a load for a specic number of cycles. By using a testing machine, a series
of specimens can each be subjected to a predetermined stress and cycled to failure.
The results are plotted as a graph representing the stress S on the vertical axis
and the number of cycles to failure N on the horizontal axis. The graph is called
a S-N diagram. The fatigue behaviour can be classied into two domains. One is
associated with relatively high loads that produce not only elastic strain but also
some plastic strain during each cycle. As a result, fatigue lives are relatively short;
this domain is called low-cycle fatigue (LCF) and occurs at less than approximately
104 to 105 cycles [23]. For lower stress levels creating totally elastic deformations,
longer life result. This is termed high-cycle fatigue (HCF) where large numbers of
cycles are required to produce fatigue failure. It is related to fatigue lives greater
than about 104 to 105 cycles.
Cracks related to fatigue failure almost always initiate on the surface of a component
at some point of stress concentration. Crack nucleation sites include sharp llets,
surface scratches, threads, keyways, dents, and the like. In addition, cyclic loading
can produce microscopic surface discontinuities as a consequence of dislocation slip
of atoms that may also act as stress raisers, and thereby as crack initiation sites.
7
2.2. ENERGY GENERATION
The design of a component can have a huge inuence on its fatigue characteristics.
The sharper the geometrical discontinuity on a surface, the more severe the stress
concentration. Probability of fatigue failure might be reduced by avoiding (when
possible) structural irregularities or by making modications in design where sud-
den contour changes leading to sharp corners are eliminated - for example, designing
rounded llets with large radii of curvature at the point where there is a change in
diameter. Also, it has been observed that improving the surface nish by polishing,
imposing residual compressive stresses within a thin outer surface layer, and con-
ducting case hardening, all are techniques that will enhance fatigue life signicantly
[23].
2.2 Energy generation
For more than 150 years, hydropower has been a signicant contributor to energy
generation. It is by far the most ecient method of large scale electric power pro-
duction. Dependent on which type of turbine being employed in a power plant, the
conversion eciency can be as high as 95% for large installations [10]. Common
to hydropower aggregates is that they can provide rapid eect grants and should
thereby be able to meet uctuations in energy consumption in the electrical system.
This applies in particular to high pressure stations with large magazines [19].
2.2.1 Synchronous generators

The primary function of a hydropower generator is to convert the rotation of the shaft
into electric power [52]. The basic process of generating electricity is to rotate a series
of coils inside a magnetic eld or vice versa. This leads to movement of electrons
inside conductors, which produces electrical current. A synchronous generator have
synchronized waveform of the generated voltage and rotation of the generator. Each
peak of the sinusoidal waveform corresponds to a physical position of the rotor. The
frequency is governed by Equation (2.4), where f is the frequency, N is the rotor
speed, and po is the number of poles formed by the stator windings.
N · po
f= (2.4)
120
2.2.2 The Nordic system

The Norwegian energy system is unique in that approximately all electricity is gen-
erated through hydropower. In 2010, this share was 95%, which means that Norway
is the sixth largest hydropower producer in the world [16]. Norway have more than
800 magazines corresponding to 85 TWh. And with a high share of power installed,
30.1 GW, the Norwegian energy system have a relative large exibility when it comes
8
to production based on demand [9]. In 1991, the energy system and energy market
in Norway became liberalized. This implied that all got the right to choose their
electricity supplier, and that the electricity market became competitive. Character-
istics such as eciency enhancement and greater cost consciousness in the operation
of the power grid followed.
The Norwegian grid is connected with Sweden, Finland and east Denmark, and
forms a joint synchronous system, the Nordic power system [36]. However, the elec-
tricity production diers considerably among the Nordic countries. As mentioned,
in Norway nearly all electricity is generated from hydropower. Sweden and Finland
use a combination of conventional thermal power, hydropower and nuclear power,
with hydropower stations located mainly in northern areas and thermal power in
the south [28]. Denmark relies primarily on conventional thermal power, but wind
power is contributing with an increasing part of the demand for energy. The Nordic
system is tuned to a frequency of 50.0 Hz, which means that the machines connected
to the grid must be set so that they deliver/use a power of 50 Hz [16]. The system
frequency is a continuously changing variable which is determined and controlled
by real time balance between system demand and total generation. It indicates the
"health condition" of the power grid. If the frequency is lower than 50 Hz, this is a
sign of insucient production. If the frequency is higher than 50 Hz, it implies that
the production should be adjusted down. Should the deviation from 50 Hz become
too large, components connected to the grid may automatically disconnect or be
damaged.
Electricity needs to be produced in the same moment as it is used, and the con-
sumption varies every minute as electrical apparatus are being adopted. Generally,
the consumption of electricity is especially high in the morning and in the afternoon,
and lower during night-time. The consumption is also signicantly higher during the
winter than the summer, since Norwegians use electricity for heating [16]. Statnett
is Norway's transmission system operator (TSO). This means that the company fa-
cilitates the power market by making it physically possible to transport power from
sellers to buyers. They are also responsible for keeping the Norwegian system in
balance. To ensure that production of power at any given time is equal to consump-
tion, Statnett demands that all electricity producers inform about their expected
production one day in advance.
9
Figure 2.2: Coordinating production and consumption [43]
2.2.3 Implementation of renewable energy sources

The change towards a more environmental friendly future have led to replacement of
many non-renewable energy sources like oil and gas with renewable energy sources.
In July 2005, the European Economic Area (EEA) committee decided to include
the Renewable Directive of the European Union (EU) into the EEA agreement [31].
The directive states that the share of renewable energy in EU should increase up to
20% from 2005 to 2020. Norway and Sweden have close collaboration on the power
system development and are committed by the deal. Furthermore, in 2012, the
electricity certicate market was established, where green certicates are intended
to stimulate expansion of new renewable energy production equivalent to 24.6 TWh
yearly within 2020 [31].
When it is desired to phase in electricity from renewable sources to the grid, a number
of new challenges turn up. Renewable energy production is often characterized by
variations over time in supply, tends to be unpredictable, and is hard to regulate.
This implies in particular to solar and wind energy, where electricity produced from
solar cells and wind turbines is characterized through production which seldom can
be planned. Wind energy as a contributor to the electrical system will vary from
hour to hour, while sun energy may vary within a couple of hours, due to the varying
access dependent on weather conditions. Consequently, the information requirement
from Statnett pose diculties for renewable energy producers which only can predict
their production a few hours ahead of time.
In addition does the renewable electricity generation behave quite dierent from
traditional, centralized generation facilities, in that most of these sources do not
contribute to systems reserves and to the total system inertia [47]. Inertia in the
10
rotating masses of synchronous aggregates decides the instant frequency response

with respect to dierences in the total power balance. When a frequency incident
occurs, the synchronous machines will absorb or inject kinetic energy from or into
the grid to correct the frequency deviation. The lower the system inertia is, the
easier the grid frequency reacts on changes in generation and load patterns. The
quality of the Nordic grid frequency has gradually become worse during the last
years, meaning that an increase in frequency variations outside the normal range of
49.9 - 50.1 Hz has been observed [43]. This trend can be seen in Figure 2.3 and is
most likely related to the utilization of new non-regulated energy sources, among
other reasons.
Figure 2.3: Number of minutes outside 49.9 - 50.1 Hz per week [43]
2.2.4 Frequency control

After the day-ahead trade is settled and the production plans are decided by the
TSOs, imbalances can occur due to changes in the predicted consumption and pro-
duction, faults in the systems or outages of transmission lines and plants. Balancing
within the hour is therefore necessary to be able to maintain a sucient system
quality. This is done through intraday trade or through the use of reserves. Intra-
day trade (Elbas) was implemented in Norway in 2009, and makes it possible for
balance responsible entities to adjust their plans closer to real time [53]. Instanta-
neous changes are handled by reserves for frequency control. In the Nordic system,
one can distinguish between primary, secondary and tertiary control.
Frequency control in a power system can actually be divided into two phases. In
the rst phase, inertial response, the frequency controllers of the power plants are
not activated. Instead, the generators release or absorb their kinetic energy to
adjust the change in frequency. In a second phase, the frequency is rst stabilized
and then restored to the nominal frequency by the reserves. If a large frequency
11
disturbance occur, additional measures, like automatic load shedding, can be taken
into consideration [47].
Inertial response
The frequency variation directly after a generation-load imbalance (∆P) is deter-
mined by Equation (2.5), where Pgen is the generated power, Pload is the power
demand, wel is the electrical angular frequency, and Jsys is the total inertia of the
system [47]. The right-hand side represents the derivative of the kinetic energy
stored in all the generators of the power system.
d( 12 Jsys · wel
2
)
Pgen − Pload = (2.5)
dt
The system inertia is dependent on two things: the number of operating generators
and the inertia of each of these generators. Synchronous machines found in the
standard power plants can contribute to this inertia based on the direct coupling
between their rotational speed and the electrical frequency. On the other hand,
wind turbines and photovoltaic units are equipped with converters that decouples
the motion of the generator from the grid frequency, and therefore do not deliver any
inertial response [38]. Replacing conventional generation by wind and solar power
will thus result in a lower system inertia, which can lead to a high initial rate of
change of frequency.
Primary frequency control

Primary frequency control plays a role between the rst seconds (<30 s) after a fre-
quency disturbance [24]. It is a result of automatic control of the synchronous ma-
chines, and immediately opposes frequency deviations without intervention through
supervisory control or operator action. The frequency is regulated using a device
called governor which regulates the power input of the turbine according to the ro-
tational speed, in order to ensure stable operation of the generator. The governor
action is dependent on two parameters, the dead band and the speed droop. The
dead band of the governor decides the minimum amount of change in frequency
needed before the governor action is activated [51]. While the speed droop is the
ratio of the relative change in frequency to the relative change in power output. The
mathematical expression for permanent speed droop, bp is presented in Equation
(2.6), where ∆f is the change in frequency, fnom is the nominal frequency (50 Hz
on the Nordic grid), ∆P is the change in power at the power plant, and Pnom is the
nominal power at the power plant. The droop is always positive to ensure stable
operation, and the maximum value for the speed droop in the Nordic region is cur-
rently set to 6% [45]. The formula indicates that a drop in frequency gives increased
12
output.
∆f
/fnom
bp = ∆P/
(2.6)
Pnom
Primary frequency control results in a new stationary position of the frequency that
diers from the nominal value. Therefore, additional power must be added so that
the frequency reaches its set-point value of 50.0 Hz. This is done by the secondary
frequency control.
Secondary frequency control

Secondary frequency control governs the allocation of loading among the accessible
power plants. It follows after the primary frequency control, and reacts in the time
scale of minutes (15 s - 15 min) [24]. Typically, it adjusts utilization of load after
generation. The secondary reserves have been manually controlled in the Nordic
market for a long period. But since the quality of the electrical grid frequency has
gone down during the last decade, Statnett implemented a new market for automatic
secondary governing in 2013 [45].
Tertiary generation control

Tertiary generation control dominates minutes to hours (>15 min) after a frequency
disturbance, and it is usually a planned event in anticipation of expected load changes
[24]. It ensure combination of optimal energy production that minimizes costs while
power demand in the market becomes saturated. The tertiary reserve is retrieved
from the regulated electricity market, a common balance market for the Nordic
electricity market. System administrators buy the power needed to balance the
power system, and the providers gives a proce to regulate the power production up
or down, and then have 15 minutes before the trade must be implemented.
13
2.3. THE FRANCIS TURBINE
Figure 2.4: How dierent reserves balance the system
2.3 The Francis turbine
With approximately 60% of the global hydropower capacity in the world, Francis
turbines are the most widely used type of hydro turbines [3]. The Francis turbine
operates at heads between 15 and 700 meters and medium volume ow. It can
quickly switch from pump to generation mode during changes in power demand,
and the turbine is widely used to stabilize power grid operation [19].
2.3.1 Main components

Many of the components in a turbine construction are tailor made, and not all
of them are found in every turbine. Manufacturers may also construct some details
dierent. Moreover, the turbine constructions depend on the turbine size [33]. Below
is a list of the main components found in a Francis turbine.
14
Figure 2.5: Illustration of a Francis turbine [33]
• Spiral casing: The spiral casing works as the water conduit between the
penstock and the regulating mechanism. It serves to distribute the water
equally around the circumference through the stay vanes and in towards the
guide vanes. Since the cross-sectional area of the spiral casing is decreasing,
the passing water is constantly accelerating and obtains a larger rotation.
• Stay ring and stay vanes: The stay ring consists of an upper and lower ring
connected by the stay vanes. The purpose of the xed stay vanes is to absorb
the axial forces on the inside of the spiral casing, and they are shaped such
that they almost don't aect the water ow.
• Guide vanes: The purpose of the adjustable guide vanes is to achieve a best
possible uniform ow pattern with increasing rotation in towards the runner.
In addition are the guide vanes used to regulate the water volume.
• Covers: The covers are fastened to the stay ring of the spiral casing. They
15
are designed with a high stiness to assure minimum deformations caused by

the water pressure. This is very important for achieving a small clearance
gap between the guide vanes and the facing plates of the covers. The gap
between the runner and the covers is also made as small as possible. Turbine
covers support the guide vane trunnion bearings. In addition, the upper covers
supports the regulating ring bearing, the labyrinth ring, the turbine bearing
and the shaft seal box, while the lower cover supports the lower labyrinth ring
and the draft tube cone.
• Labyrinth seals: Undesirable water leakage through the gaps between the
runner and the covers cause reduced eciency, but by placing labyrinth obsta-
cles in the water way the amount of lost water can be reduced by creating a
drop in pressure. The labyrinth consist of two parts, a static seal attached to
the covers and a rotating part attached to the runner. For high head turbines,
the leakage ow can by used as cooling water after it has been ltered through
the labyrinth seal, which will increase the overall eciency.
• Runner: The runner consists of a hub, a shroud and several blades connecting
them together. The whole power output takes place in the turbine runner, as
the runner transfers torque and power from the water to the turbine shaft
through its rotation.
• Draft tube: The draft tube work as the water conduit from the runner to the
outlet gate. It gradually reduces the velocity due to an increase of its cross-
section, and thereby converting kinetic energy at the runner outlet to pressure
energy at the draft tube outlet.
2.3.2 Operation
The Francis turbine is a type of reaction turbine, meaning that the turbine is com-
pletely submerged in water and experience a drop in pressure from the inlet to the
outlet of the runner [19]. About half of the the specic energy at the inlet is kinetic
energy and the other half is pressure energy. Through the runner, the energy is
converted to mechanical energy partly due to the pressure drop and partly from the
impulse forces caused by changes in the direction of the relative velocity vectors.
The ow pattern at an operating point is created by the inow towards the runner
blades and the outow from the runner into the draft tube. Flow characteristics
are idealistic only in a limited part of the operating range, which includes high
eciency and low uctuations of pressure and output power. O-design operating
conditions with an inow angle that deviates from the ideal angle comes with swirl,
ow separation and backow in the draft tube [41]. The velocity triangles of a
Francis runner is shown in Figure 2.6, where w ~ is directed in relative direction along
the blades and ~u is directed in tangential direction. The absolute velocity is the sum
of the two, ~c = ~u + w
~ [19].
16
Figure 2.6: Velocity diagrams at the inlet and outlet of the runner [19]
2.3.3 Turbine performance

During both project planning and operation of plants, turbine performance charac-
teristics of output and eciency are needed to estimate the number and size of units
required, quality of the production, and economic feasibility. The equations below
are used for computation of the important quantities hydraulic eciency η (Equa-
tion (2.7)) and head H (Equation (2.9)). τ is the turbine torque, ω is the angular
velocity of the runner, and ρ is the density of water. ∆p represents the pressure
dierence from inlet of the spiral casing to the outlet of the draft tube, while A is
the cross sectional area at those specic locations.
τ ·ω
ηM = (2.7)
pM · Q
ρ · Q2 1 1
pM = ∆p + · ( 2 − 2) (2.8)
2 A1 A2
pM
H= (2.9)
ρ·g
17
2.3.4 Francis runner

The shape of a runner blade cross-section is like a thin airfoil. So when water ows
over it, a low pressure is induced on one side (suction side) and a high pressure on the
other side (pressure side). The Francis runner has xed blades, usually between seven
and thirteen dependent on the operating head [48]. High head turbines are required
to have a higher number of runner blades based on strength considerations. By
increasing the number of blades, the pressure loading on the blades will be reduced,
cavitation is avoided, and separation is prevented at the runner inlet during low
loads. However, it also increases friction losses due to a larger amount of contact
surfaces. Overall, the runner blades experience more stresses than other turbine
parts because of the small distance between them and the large pressure variation
from pressure side to suction side. They are therefore shaped in such a way that the
main part of the hydraulic energy is utilized at the beginning of the blade, meaning
that it's thicker at the inlet than at the outlet of the runner.
2.3.5 Material and design

The Francis components can be divided into pressurized static parts and movable
stressed parts. The pressurized static parts includes the spiral casing, top and bot-
tom cover, stay ring and draft tube, while the movable stressed parts includes the
guide vanes, shaft, runner and labyrinth seals. For high head turbines, the stress
carrying guide vanes and runner blades are often made of "ne grain high tensile
strength carbon steel" [20]. The choice of material is based upon maximum stress
and number of pressurized cycles experienced during operation. In contrast are com-
ponents that require much welding, such as the covers and the draft tube, usually
made of "ne grain low tensile stress with low carbon content steel". The develop-
ment of new material technology over the years have improved cavitation resistance
and increased the ductility of Francis turbines. Consequently, a higher stress level
is tolerated in the stationary pressurized components. But for materials used in
rotating parts with high frequency loads and on wet surfaces exposed to high ow
rate, it is important that the residual stress level is held low to achieve a long turbine
lifetime. When manufacturing a turbine, the weld quality must be perfected in or-
der to avoid defects and brittleness in the heat aected zones of the material. Thus,
in addition to improvements in casting and welding controls, limitations are often
made in the chemical compositions as a safety procedure [21]. The advent of CAD
and manufacturing occurred at the end of the 1970s and 1980s, and made many
of the advances seen today possible. The use of CAD enhances the ability of an
organization to work in a competitive environment to design site specic turbines.
Modern CFD ow analysis, FEA techniques for engineering, and computer numeri-
cally controlled (CNC) in the industry, have signicantly improved turbine eciency
and output accuracy. CFD analysis is today conducted by all turbine manufacturers
as a part of the design process, and turbine geometry is now very close to design
geometry thanks to CNC.
18
2.4. LOADS ON A FRANCIS TURBINE
2.4 Loads on a Francis turbine
The operation regime of a hydraulic turbine can be divided into steady and transient
state. The steady state is dened as stable operation at a xed operational point,
either BEP, part load or overload [35]. The turbine operates at a constant head,
speed, load, and guide vane opening, while forces such as the static weight of the
runner including the water weight, residual stresses and dynamic forces acts on
the turbine. Residual stresses can be generated during construction and assembly
of the turbine, but also during operation of the turbine due to bad welding and
uneven heating. The eect of residual stress is amplied on complex shapes or in
certain turbine components. Dynamic forces occurs with rotation of the unit as
a result of the combination of unbalance and misalignment with other transient
perturbations, like intermittent water ow or unwanted objects in the water. Static
forces are constant in magnitude, direction and frequency, while dynamic forces are
random, non-periodic forces with dierent directions, amplitude and frequencies.
The transient state is characterized by change in head, load or guide vane opening
related to starting, synchronization, changing load, stopping, load rejections, and
runaway speed. During such operation, vibrations do not follow a specic pattern
but develop depending on the amount of water going through the turbine.
2.4.1 Torque oscillations

In a hydropower plant, the turbine runner and the generator rotor are connected by
the shaft and are rotating in the same direction. During operation, the hydraulic
torque Th from the water acting on the runner will try to accelerate the turbine,
while the magnetic torque Tm acting on the generator will try to decelerate the
turbine [25]. The summation of the dierent torques result in a net torque pulling
on the turbine runner. As expressed in Equation (2.10), Newton's second law claims
that a net torque is proportional to a change in angular velocity. In the equation,
the right side consists of the mass moment of inertia J and the change in angular
velocity of the runner ω . The left side is the summation of torques T, where Td
is the dampening torque from a mechanism in the generator that can cause both
acceleration and deceleration. If the summation is equal to zero, the angular velocity
will be constant. But when the accelerating torques are larger than the decelerating
torques, the angular velocity will increase, or vice versa.
dω X
J = T = Th − Tm − Td (2.10)
dt
19
Figure 2.7: Net torque on the turbine [25]
The grid frequency determines how fast the generator stator magnetic eld rotates.
When power is generated, the magnetic eld of the rotor pulls on the magnetic eld
of the stator. In phase, these elds are separated by the magnetic torque angle δ .
If the grid frequency increases, the stator magnetic eld catches up with the rotor
eld, decreasing δ and thereby decreasing the torque acting on the rotor from the
stator. Equation (2.11) expresses how the torque acting on the rotor is connected
to the magnitude of δ . Here, Tnom is the nominal torque and δnom is the nominal
torque angle. As variations in grid frequency leads to variations in δ , generators
are often equipped with a dampening mechanism to dampen these variations in
magnetic angle. This counteraction is described in Equation (2.12), where md is a
constant linking the magnitude of torque to the magnitude of the time derivative of
the magnetic angle.
sin(δ)
Tm = Tnom (2.11)
sin(δnom )
dδ
Td = md (2.12)
dt
The hydraulic torque Th is a function of the hydraulic power Ph and angular velocity
ω , as expressed in Equation (2.13) [25]. H represents the net head over the runner,
Q is the volume ow, ρ is water density, g is the gravitational constant, and ηh is
the hydraulic eciency.
Ph ρgQHηh
Th = = (2.13)
ω ω
A variation in grid frequency will lead to changes in both ω and Th . This is due to
the fact that ω changes with frequency since generators are synchronous machines,
and Th changes based on the permanent speed droop. A decrease in frequency means
a decrease in ω and this will reduce Th according to Equation (2.13). The permanent
20
speed droop in the governor will work against the decrease in frequency by opening
the guide vanes such that Q and Ph increases, thus reducing Th even more. Hence,
the speed droop mechanism amplies the changes in Th . If Th increases while Tm is
kept constant, the rotational speed of the turbine will increase. This means that the
generator rotor will rotate faster than the magnetic eld in the stator given by the
grid frequency. δ and the induced torque in the generator will then increase until
the hydraulic torque and the induced torque are equal once again. An increase in
grid frequency will lead to an increase in rotational speed of the magnetic eld in
the generator stator. δ will then be reduced, which reduces the magnetic torque and
the decelerating torque. The turbine's speed of rotation will further increase until
the torques are equal and the grid frequency is proportional to the speed of rotation.
Even if the unit is operating at steady state, the inuence of grid frequency variations
leads to unsteady torque oscillations. The frequency varies much and relatively
fast. These oscillations become dynamic loads on the rotating masses and increases
material stresses in the turbine parts. They are characterized as slow oscillations
and come in addition to all the other dynamic loads explained further below [45].
2.4.2 Pressure pulsations

When a Francis turbine is running at BEP, it will normally be inuenced by oscil-
lations caused by RSI [32]. However, when the operation moves outside of BEP,
additional phenomena such as vortex breakdown and separation around curved lines
appear. O-design pressure pulsations can have large amplitudes that may cause
serious damage to the structural runner. The magnitude of these oscillations is de-
pendent on design, operational patterns and dynamic response of the system. They
can occur periodically or stochastically. Experimental results have shown that low
load oscillations often have a stochastic nature, while high load oscillations occur
periodically [40]. Pressure pulsations can be divided into two categories; pulsations
based on rotational oscillations from the turbine rotation (linked to the guide vanes,
runner vanes, runner and draft tube) or pulsations based on elastic oscillations [29].
The elastic oscillations are a result of the ow changing through the turbine and
causing a dynamic pressure dierence, either an acceleration or retardation of the
ow. This pressure wave will travel through the waterways of the plants and is
dependent on adjustment of valves or vanes.
Rotor-stator interaction
Rotor-stator interaction can be characterized by two dierent phenomena: wake/rotor
interaction and potential rotor/stator interaction [34].
• Wake/rotor interaction can be assumed steady in the stator reference frame,
while they are unsteady in the rotor reference frame because the rotor is passing
through wakes which are leaving the stator.
21
• Both stator and rotor experience non-uniform pressure components which re-
sults in unsteadiness from potential rotor/stator interaction.
The combination of these two phenomena creates a non-uniform velocity eld en-
tering the runner. As the radial clearance between the runner inlet and guide vane
outlet (the vanless space) decreases, the level of pressure pulsations will increase.
Figure 2.8: Flow eld at inlet of runner [34]
Draft tube surge

Draft tube pressure pulsations in Francis turbines is a familiar phenomena. The
ow pattern in the draft tube is determined by the direction of the ow exiting the
rotating runner. At BEP, the direction is almost parallel to the shaft axis. But at
points diering from BEP, a tangential velocity component is present, which is the
source of several phenomena [34]:
• At loads below BEP, the direction of the tangential ow is in the same direction
as the runner rotation.
• At some loads, a cavitating rotating vortex rope with helical shape can be
observed.
• At loads above BEP, the direction of the tangential ow is in opposite direction
of the runner rotation.
• At even higher loads, this results in a cavitated vortex core.
The part load vortex rope lead to forces uctuating radially in the draft tube. This
may cause vibrations in the system, and fatigue after a while. At some hydropower
plants, the rotating vortex rope is so strong that it results in a pressure uctuation
which can propagate upstream the runner, creating mass ow uctuations in the wa-
terway. These uctuations will increase elastic pressure waves. And if they coincide
with resonance frequencies in the system, they can cause major uctuations. The
overload vortex core may cause mass ow uctuations that can propagate through
the entire waterway, causing power from the unit to uctuate.
22
Vortex shedding
Von Kármán vortex shedding comes from unsteady ow separation o blu bodies
in a uid stream and creates alternating low-pressure vortices on the downstream
side of the body [34]. As a consequence, the body is subjected to an oscillating force
perpendicular to the ow direction. Such vortices are one of the components of the
guide vane wake, and is also known to result in severe noise from Francis runner
blade trailing edges.
2.4.3 Reduction of turbine lifetime

Turbine components experience individual wear and tear during operation and the
sustainability of these components may vary on type of loads they are experiencing.
Main causes for turbine damage are fatigue, material defects, cavitation, and erosion,
and these are well known issues for high head turbines due to high pressure and the
pressure variation under dynamic load. Even though each unit is designed with a
sucient margin of safety to withstand normal stresses, high vibrations and dynamic
stresses result in gradual development of cracks. In Francis runners, fatigue cracks
tends to occur either relative early in life or after decades of operation. The failure
mechanism is said to be a combination of LCF and HCF [27]. LCF originates from
start-stop cycles, which can propagate fatigue cracks from manufacturing defects or
initiate fatigue cracks at areas of high stress concentration. With rotational speed up
to several hundreds RPM, the amount of vibration or HCF cycles caused by wakes
behind the stay- and guide vanes increases to many millions per day. Consequently,
when a crack has reached the fatigue crack growth critical size with respect to HCF
loading, it may cause catastrophic failure very fast compared to the design life of the
turbine runner. During steady state, several analyses have shown that the largest
stresses appear where the blade is xed to the hub or shroud near the trailing edge
[27] [35] [18]. At this T-joint geometry, the level of stress and strain is further
intensied due to notch eects. They are thereby the most critical locations for
fatigue. Figure 2.9 presents statistics of cracks in Francis runners during the last 50
years collected from 368 Francis aggregates. It shows a clear improvement from the
1960s to the 1990s due to upgrades in design techniques. However, the amount of
runners which have experienced cracking have increased over the last 20 years. The
reasons for this is most likely that the runners are not designed to withstand the
dynamic loads exerted on them during frequent variations in operation.
23
Figure 2.9: Statistics of cracks in Francis runners from 1960 to 2000 [22]
24
Chapter 3
Investigation tools
Experimental measurements of transients combined with a numerical analysis can

provide reasonable research outputs together. Detailed experiments on transients
are often limited due to the risk of damaging the machine. Therefore, a necessary
set of experimental results can be used for validation of a numerical model, and
further analysis can be carried out numerically. The uncertainty of the numerical
results will increase or decrease with increasing or decreasing deviation from the
nearest known physical model.
3.1 Model testing
On-site testing can be complicated and costly, and most prototype machines are too
large to be tested in a laboratory. Therefore, it is very valuable to have the ability to
perform tests on downscaled models of prototypes. It gives a greater exibility to do
repeated measurements at dierent operation points. For model testing of hydraulic
turbines, an international standard IEC 60193 have been developed which is used as
guidelines for measurements, calibration, and uncertainty analysis [42]. Prior to a
model test, calibration of measurement equipment is necessary to make sure that the
readings from the instrument are consistent with other measurements, to determine
the accuracy of the instrument reading, and to establish if the instrument can be
trusted. The resulting uncertainty in a measurement is a numerical estimate of the
dispersion of values that could reasonably be attributed to the quantity [6]. It pro-
vides the quality of a measurement and oer the opportunity to assess and minimize
the risks and possible consequences of bad decisions [6]. Further information about
calibration and uncertainty analysis can be found in Appendix C.
25
3.1. MODEL TESTING
3.1.1 Sensor fundamentals

Pressure measurement
A pressure measurement can be described as either static or dynamic, and also by
the type of measurement being performed where one can distinguish between three
types: absolute, gauge, and dierential [13]. Pressure sensors, or transducers, have a
sensing element of constant area and respond to force applied to this area by the uid
pressure. The force deects a bellows, diaphragm, or Bourdon tube. In turn, these
deections, tensions, or strains are transformed to electrical outputs proportional to
the pressure through a transduction method. Common outputs are mV, V, mA, and
frequency.
Strain measurement
The most common way to measure strain is with a strain gage, a device whose elec-
trical resistance varies in proportion to the amount of strain in the device [14]. The
resistance change occurs during all conditions of static and dynamic strain. Usually,
the magnitude of measured strain is very small and is therefore often expressed as
micro strain ( · 10−6 ). This is also the reason for why strain gages almost always
are used in a bridge conguration with a voltage excitation source.
(a) Pressure transducer (b) Strain gage
Figure 3.1: Measurement sensors
3.1.2 Discrete sampling and sampling rate

Digital data-acquisition systems record signals at discrete times, resulting in a deduc-
tion of actual measured behaviour between samples. The rate of the measurements
is called the sampling rate. By choosing a poor sampling rate, the results can turn
out misleading when analysing time-varying data. Information from the original
signal can be lost, and unwanted components can appear during the reconstruction
of the signal. It is therefore important to choose a tting sampling rate based on
which result sizes that one wants to analyse and which rate the sensors being used
are able to perform at.
26
3.2. NUMERICAL ANALYSIS
3.1.3 Filtering and smoothing of data

A moving average lter smooths the given data by replacing each point with the
average of the neighbouring data points within a specic range [8]. The response of
the smoothing is calculated by the dierence equation,
1
ys (i) = [y(i + N ) + y(i + N − 1) + ... + y(i − N )] (3.1)
2N + 1
where ys (i) is the smoothed value for the ith data point, N is the number of neigh-
bouring data points on each side of ys (i), and 2N+1 is the span.
Savitzky-Golay ltering is a generalization of the moving average method. The lter
coecient are derived by performing an unweighted linear least-square t using a
polynomial of a certain degree. A high degree polynomial makes it possible to achieve
a high level of smoothing without attenuation of data features. The method is very
eective for frequency data since it preserves the high-frequency components of the
data, but it can be less successful than a moving average lter at rejecting noise.
Figure 3.2: Attempt at smoothing data using the Savitzky-Golay method [8]
3.2 Numerical analysis
As the computational power of computers increases, the application of high-end

numerical methods continues to grow. CFD has emerged as a reliable tool due to its
great accuracy and exibility in simulating various and complex three-dimensional
(3D) ow phenomena. It has become a standard method to evaluate the feasibility of
hydro turbines, and minimize much of the time and money spent for experimental
investigations. In combination with FEA, CFD can be used to investigate uid-
structure interaction, which have shown to be a crucial analysis for risk reduction
27
during the design process of turbines. FSI problems play prominent roles in many
scientic and engineering elds, yet a comprehensive study of such problems remains
a challenge due to their strong non-linearity and multidisciplinary nature.
3.2.1 Computational uid dynamics

Computational uid dynamics is the science of predicting uid ow, heat and mass
transfer, chemical reactions, and related phenomena by the means of computer-based
simulation. The most common CFD technique used in commercially available pro-
grams is the nite volume method [17]. It has an advantage in memory use and speed
for major problems, turbulent ows, higher speed ows, and source term dominated
ows. The nite volume procedure starts with dividing the regions of interest into
small control volumes. Integration of the governing dierential equations of uid
ow (the Navier-Stokes equations, the mass and energy conservation equations, and
the turbulence equations) over the set of control volumes, in additions to application
of the divergence theorem, follows. To evaluate the derivative terms, values at the
control volume faces are needed. The result is a set of linear algebraic equations,
one for each control volume, which are further solved iteratively or simultaneously.
Turbulence modelling is a key issue in the majority of CFD simulations. Most engi-
neering applications are turbulent and therefore require a suitable turbulence model.
ANSYS CFX
ANSYS CFX is a high-performance CFD software tool based on the nite volume
method that gives out accurate and reliable solutions across a wide range of CFD
and multi-physics applications. It is known for its outstanding robustness, accuracy
and speed when it comes to rotating machinery such as pumps, fans, compressors,
and turbines [4]. The software have been applied to solve a variety of uid problems
for over 20 years.
3.2.2 Computational structural mechanics

Structural mechanics is computation of deections, deformations and internal forces
or stresses within structures, either for design purposes or performance evaluation of
existing structures. The nite element method is the dominant discretization tech-
nique in structural mechanics, where partial dierential equation (PDE) problems
are translated into a set of linear algebraic equations for steady state problems and
a set of ordinary dierential equations for transient problems [1]. The basic method
involves decomposition of the domain into non-overlapping elements of simple ge-
ometry. The response of each element is expressed in terms of a nite number of
degrees of freedom characterized as the value of an unknown function or functions,
at a set of nodes. Systematically recombination of all elements at the nodes, thus
forming a global system of equations, follows for the nal calculation of the model
28
solution. FEM is a good choice for solving PDEs over complicated domains when
the domain changes, when the desired precision varies over the whole domain, or
when the solutions is missing smoothness.
ANSYS Mechanical
ANSYS Mechanical is a FEA tool that enables the user to analyse complex models
and solve advanced mechanical problems. It is used for structural analysis, including
linear, non-linear and dynamic studies [5]. Both steady and transient problems can
be solved in Mechanical. A static analysis is used for static loading conditions,
and non-linear behaviour such as large deections, large strain, contact, plasticity,
hyper-elasticity, and creep. A dynamic analysis includes mass and damping eects.
In ANSYS, one can distinguish between three dierent dynamic analyses: modal,
harmonic and transient dynamic.
3.2.3 Fluid-structure interaction

Fluid-structure interaction is a multi-physics coupling between the laws that express
uid dynamics and structural mechanics. This phenomena is characterized by stable
or oscillatory interactions between a deformable or moving structure and an inter-
nal or surrounding uid ow. When a uid ow encounters a structure, stresses
and strains appear in the solid body, which are forces that can lead to permanent
deformations. Dependent on the pressure and velocity of the ow and the material
properties of the structure, these deformations can be very large or quite small [7].
If the deformations of the structure are small and variations in time are slow, the
behaviour of the uid will not be largely aected by the deformations, and the focus
can be on the resultant stresses in the solid parts (one-way FSI). However, if the
variations in time are fast, pressure waves can appear in the uid even for small
structural deformations. For large structural deformations, the velocity and pres-
sure elds of the uid will change as a result, and the problem needs to be treated
as a bidirectionally coupled multi-physics analysis (two-way FSI).
ANSYS software provides several ways to perform a FSI simulation. One way is
to combine CFX and Mechanical through ANSYS Workbench. When dealing with
an unidirectional coupling (one-way FSI), the dominating eld can be simulated
without considering the other physical eld. As for transient uid ow through a
Francis turbine, it can be simulated without considering the structural motion and
without signicant loss of accuracy. The transient pressure distribution from the
CFD analysis can then be used as surface load for the structure dynamic analysis.
ACT Transient FSI

ACT Transient is a numerical method that can be used to conduct a one-way FSI
simulation in ANSYS. The method is an advantage when the work is shared between
29
two persons, since the uid and structural analyses can be performed on separate
computers. It does not create one giant project with enormous amounts of data, but
instead the CFD calculation is closed and is a pure one-way coupling. Three data les
are created for result transfer to ANSYS Mechanical containing the mesh, the step
end times and the nodal load values of each time step [26]. This enables the user to
interpolate transient CFD results from the CFD mesh to the FEM mesh in each time
step of a transient structural analysis. To get a consistent workow the extension
automatically creates the same time step settings as used in the underlying CFD
simulation. The CFD results from one calculation can be used for many Mechanical
runs. However, the downside with using this numerical method is that the viscous
forces/shear forces are not accounted for in the analysis.
3.2.4 Numerical errors and uncertainties

When identifying and validating modelling and simulation results, evaluation of er-
rors and uncertainties is very important. Error is dened as a recognisable deciency
in a numerical model that is not caused by lack of knowledge, while uncertainty is
dened as a potential deciency in a numerical model that is caused by lack of knowl-
edge [50]. The amount of errors and uncertainties decides the quality of the results
obtained. One can distinguish between dierent errors and uncertainties based on
how they occur:
Causes of errors:
• Numerical errors and round of errors, discretization errors and iterative con-
vergence errors
• Coding errors and mistakes or "bugs" in the software
• User errors and human errors through incorrect use of the software
Causes of uncertainties:
• Input uncertainties and inaccuracies due to limited information or approximate
representation of geometry, boundary conditions, material/uid properties,
etc.
• Model uncertainties and discrepancies between real ow and the numerical
model due to insucient representation of physical or chemical processes, or
due to simplied assumptions in the modelling process
Mesh grid
The solution of numerical models is dependent on the mesh grid. It is therefore
always recommended to adapt the grid until the solution is independent of the
mesh. However, this is sometimes dicult due to limited computational resources.
30
When grid independence is obtained, the most coarse and independent mesh should
be chosen for further analysis.
When dealing with a CFD model, ne grid resolution is required close to the walls
to obtain an accurate solution for the boundary layer. The non-dimensional wall
distance value y+ dened in Equation (3.2) is used as a requirement, where ur is
the friction velocity at the nearest wall, y is the distance from the node to the nearest
wall, and ν is the local kinematic viscosity [18]. Dierent turbulence models have
dierent y+ requirements, and for example, the k-epsilon model uses 30<y+<300.
ur · y
y+ = (3.2)
ν
As a general rule of thumb, it is desirable with ne grids in areas of interest. And
elements with very slender proportions should be avoided.
Time step
A sucient time step is important in transient CFD simulations. The time step
should be small enough to capture the transient phenomena occurring and small
enough for the simulation to converge. Devices with many blades should have small
times step size to resolve the interaction between the blades and surrounding, non-
rotating geometry. However, a small time step increases the complexity of the sim-
ulation and needs more time and computer resources to achieve a full solution.
31
32
Chapter 4
Facilities and methods
Pressure and strain measurements on a Francis turbine model and a FSI simulation
of the same turbine have been carried out to investigate how variable rotational
speed aect the load pattern and corresponding stresses in the turbine runner. The
FSI analysis was performed in the software program ANSYS by combining ANSYS
CFX and ANSYS Mechanical. Results from the laboratory measurements were
used as inputs in the CFD and FEM analyses, and further used for validation of the
solutions.
4.1 The Waterpower Laboratory
The Waterpower Laboratory at NTNU is built around a main piping system and
include three test rigs: test rig for Francis, Kaplan and reversible pump turbines,
test rig for Pelton turbines, and test rig for micro-, mini-, and small turbines [46].
There are installed two pumps in the basement that operates separately, in parallel
or series to achieve a broad area of volume ow and head. Each pump is driven
by a 315 kW motor and a frequency converter [2]. Available pumping power is 700
kW and the maximum ow rate is 1.1 m3 /s. It is possible to obtain dierent run
conditions by utilizing dierent piping systems, and the dierent operating modes
include both open and closed loop. The closed loop system can be pressurized to
maximum 100 m, while the open loop system has a maximum head of 16 m. A
section of the open loop system with the Francis test rig can be seen in Figure 4.1.
A lower reservoir is situated under the oor in the laboratory and has a capacity of
450 m3 . An upper reservoir is situated on the top oor of the building and consists
of two large tanks with an u-shaped channel between them. This is used when
running an open loop. The laboratory also have two main tanks, one high pressure
and one low pressure tank, situated upstream and downstream of the Francis rig,
respectively. The high pressure tank is used as a pressure reservoir when running
33
4.1. THE WATERPOWER LABORATORY
closed and open loop, dampening pump eects and delaying the inuences on the
turbine from changes in pump speed. The tank may also function as an air cushion
when running the rig from the upper reservoir. The low pressure tank acts as the
tail water in the system.
Figure 4.1: A section of the open loop system at NTNU
4.1.1 Francis test rig

A model of Tokke hydropower plant is installed as a part of the Francis test rig in
the laboratory. The arrangement from the spiral case to the draft tube is set to
be a true model of the hydropower plant, and the model is scaled down to 1:5.1
of the prototype [2]. Cross sectional areas at the turbine inlet and outlet sections
are 0.0872 m2 and 0.236 m2 respectively, while the reference diameter of the model
runner is 0.349 m [49]. Additional dimensions of the model runner designed by the
Waterpower Laboratory can be found in Figure 4.2. The Francis turbine is installed
with a vertical shaft and a DC generator of 352 kW. It is equipped with 14 xed
stay vanes, 28 adjustable guide vanes and an elbow-type draft tube. The runner
consists of both half and full-length blades, and sum up to 30 blades in total. They
are twisted 180 degrees around from inlet to outlet, and the blade thickness at the
trailing edge is 3 mm [11].
34
4.1. THE WATERPOWER LABORATORY
Figure 4.2: Dimensions of the Tokke model runner
The Francis runner is constructed out of dierent bronze materials, where the blades
are made of cast tin-bronze JM 3, while hub and shroud are made of bronze JM 7 [39].
The JM 3 alloy consists of 88% copper and 12% tin, and the JM 7 alloy consist of 80%
copper, 10% aluminum, 5% iron, and 5% nickel [12]. Properties of the two materials
are found in Table 4.1. Both are well suited for construction of turbine runners. By
adding tin in the alloy, one gets a harder and more durable bronze. Tin-bronze is
resistant to corrosion in water, and is often used in marine industry. Aluminium
bronze alloy is characterized by good impact strength, excellent corrosion resistance
to acids and salt water, and is malleable. The main concern when manufacturing
the runner blades was to make sure they would withstand dierent temperatures
and pressures, and thereby not deform during operation. In addition to nancial
reasons, tin-bronze was chosen as material based on this requirement.
Material JM 3 JM 7
Density [g/m3 ] 8.8 7.6
Yield strength [MPa] > 150 > 260
Tensile strength [MPa] > 270 > 590
Young's modulus [MPa] 105 1.1 × 105
Poisson's ratio 0.34 0.32
Thermal conductivity [W/m◦ C] 50 65
Resistivity [nΩm] (20◦ C) 200 190
Table 4.1: Mechanical properties of the Tokke runner [12]
35
4.2. LABORATORY EXPERIMENT
4.1.2 Generator parameters

The Francis test rig includes a IGG6258 generator manufactured by Siemens [37].
It has the following technical parameters:
Parameter Value
Rated speed [rpm] 1560
Rated armature voltage [V] 420
Rated output [kW] 352
Rated torque [Nm] 2150
Maximum eld weakening speed [rpm] 2000
Rated current [A] 890
Eciency [%] 92
Maximum speed [rpm] 2600
Table 4.2: Technical generator parameters [37]
The output wire from the generator is connected to a rectier cabinet/DC converter
of type Simoreg 6RA70 DC Master, also manufactured by Simens [44]. Next, the
converter is connected to the power grid, which supplies and demands power to and
from the generator to maintain the chosen operating point. The rotational speed of
the generator is xed by the converter.
4.2 Laboratory experiment
The Francis test rig was operated with an open loop water circuit to obtain a con-
dition similar to the prototype without signicant variation of the eective head
during transients. Risk assessment of the experiment is found in Appendix A. A
script constructed in the software program LabVIEW from National Instruments
(NI) made it possible to control the runner speed while the guide vane angle was
held constant. Pressure at several points in the turbine and strain in one of the
runner blades were measured. Wiring of the total laboratory set-up was completed
by master student Einar Agnalt.
4.2.1 Instrumentation
Pressure sensors
Recently, ve new pressure sensors were ush-mounted into the hub of the Tokke
model runner for better durability during operation of the rig. They are placed right
in the middle of two hydraulic channels. The placement, with sensor notations, can
be seen in Figure 2. Pressure sensor PT11 is mounted close to the inlet of the
36
runner and sensor PT14 close the outlet. Likewise, sensor PT12 is placed upstream
the splitter blade and sensor PT13 downstream. The fth sensor, PT10, is mounted
in the neighbouring hydraulic channel. Pulsations from RSI, the change in channel
cross section, and phase dierence in pressure pulses will be accounted for when the
pressure sensors are placed like this.
Figure 4.3: Onboard pressure sensors mounted into the runner hub
Two pressure sensors, PT20-21, were ush-mounted in the vaneless space between
the guide vanes and the runner blades. Another four pressure sensors, PT30-31,
were mounted into the draft tube cone. The latter pressure sensors are piezoelectric,
suitable for measuring dynamic curves or pulsations. Figure 3 illustrates were all
these sensors can be found on the turbine, and the placement is set to register
pressure pulsations from RSI and draft tube surge.
37
Figure 4.4: Pressure sensors in the vaneless space and in the draft tube cone
In addition, three pressure sensors, PT01, PT02, and PT03, were ush-mounted to
the penstock, the inlet of the spiral casing, and the draft tube outlet, respectively.
Figure 4.5: Pressure sensors in pipe line
Notation Transducer Type Range

PT10-14 Kulite XTM-190S(M) Piezoresistive sensor 0-3.5 bara
PT20-21 Kulite XTL-190(M) Piezoresistive sensor 0-10 bara
PT30-33 Kistler 701A Piezoelectric sensor 0-250 bar
PT01-03 Kulite HKM-375(M) Piezoresistive sensor 0-1.7 bara
Table 4.3: Measuring equipment
Operation point sensors

Sensors recording operating conditions are placed around on the Francis test rig. A
magnetic ow meter, FT40, is mounted at the turbine inlet pipeline to measure the
discharge to the turbine. Pressure sensor PT40 is used to measure the pressure at the
38
casing inlet, while pressure sensor PT41 is used to measure dierential pressure (∆
p) across the turbine. Other sensors are included to register the generator torque
(WT40), friction bearing on shaft (WT41), RPM on shaft (ST40), and angle of
guide vane (ZT40). The ST40 measuring system contains a photocell (in the form
of an optical fork) and a circular disc with one cut xed to the generator shaft, and
measures 1024 pulse per rotation.
Figure 4.6: Sensors for calculating running point
Notation Instrument Type Range

FT40 Krohne Aquaux IFS 4000 Flow meter 0.15-500 l/s
Di. pressure
PT40 Fuji Electric FHCW36W1-AKCAY 0-1000 kPa
transducer
Di. pressure
PT41 Fuji Electric FHCW36W1 0-1000 kPa
transducer
WT40 Hottinger Z6FC3 Load cell 0-2000 Nm
WT41 Hottinger Z6FC3 Load cell 0-16 Nm
ST40 OPB 960T51 Optical fork 1024 pulse
ZT40 Stegman AG612 Angle transducer 1 pulse
Table 4.4: Measuring equipment [37]
Strain gage
A semiconductor strain gage was used to measure strain in one of the full-length
runner blades. This is placed on the pressure side of the runner blade, up towards
the joint between blade and hub, as pointed out in the Figure 4.7. This position is
based on earlier FEA results showing where the highest stresses usually occurs. It
have been glued on with a quick-drying glue, and further covered with the same glue
to assure a reliably working strain gage. The attachment method is temporarily,
and should be improved before similar measurements shall be conducted in the
laboratory.
39
Figure 4.7: Strain gage attached to one of the runner blades
4.2.2 Calibration
Calibration of the instruments used in the experiment were performed before the
measurements were conducted. This included calibration of the magnetic ow me-
ter, the weighing tank, generator- and friction-torque measurement sensors, and
pressure sensors. In the Waterpower Laboratory, pressure head is calibrated with a
manometer, ow by weighing and torque by weights on the level arm. The set up for
the rotational speed measurement system does not need calibration since the only
value that is measured is time. Achieved calibration reports are found in Appendix
D. Uncertainties from the calibrations were further used to perform an uncertainty
analysis, which is presented in Chapter 5.
The dynamic pressure transducers mounted in the draft tube cone were not cali-
brated due to lack of dynamic calibration equipment. Instead, factory calibration
data was used. Furthermore, calibration constants for the strain gage had to be the-
oretically calculated. Output voltages with corresponding strain were plotted and
tted with a calibration curve created by the use of a third degree function. The es-
timated uncertainty of this calibration curve is 0.25%. Because the strain gage only
has been theoretically calibrated, the absolute values provided by the measurements
should be handled with caution.
4.2.3 Variation in rotational speed

To control the rotational speed of the turbine runner, a script was designed in Lab-
VIEW. This script can be further studied in Appendix F. It uses a single sinusoidal
oscillation given by Equation (4.1), that is suppose to represent a quick decrease in
40
turbine speed caused by a variation in grid frequency.
1 1
y(t) = A · [ sin(πf t + φ) + ] + (y1 + A) (4.1)
2 2
A is the amplitude of the runner speed oscillation, f is the ordinary frequency of the
oscillation, and φ species where in its cycle the oscillation is at t = 0.
Figure 4.8: Variation in RPM
The LabVIEW values were transferred to a current output card NI 9265, connected
to a NI cDAQ, which created a 4-20 mA signal. This signal was sent to the "Francis
Generator signal"-box (located near the Francis runner in the laboratory) by con-
necting cables to the "disturbance RPM" input channels. The box further sends the
signal down to a rectier cabinet/DC converter in the basement. Then, DC power
is transmitted up to the turbine generator, deciding how the generator should act.
Figure 4.9 describes the set-up of the system in the laboratory.
Figure 4.9: Schematic representation of RPM disturbance system
The sine curve started at 273 RPM and increased up to the synchronous speed of
333 RPM. To stabilize the system characteristics, operation at this point was run
for approximately three minutes before the runner speed was decreased down to 273
again. Only the performance during the decrease was of interest for this master
thesis. The decrease of 60 RPM had a time period of 2 seconds, which was limited
41
4.3. COMPUTER SIMULATION
by how quick the laboratory machines can be run and how long time it would take to
simulate the same change in ANSYS. The generator in the Waterpower Laboratory
have a maximum torque limit of 2150 Nm at stationary operation, and can tolerate
a change of +10% to -20% in rotational speed during a short time period.
4.2.4 Data acquisition

Measurement data were acquired through LabVIEW and a NI cDAQ 9178 with eight
modules, including:
• Five NI 9239 modules connected to PT10-14, PT30-33, FT40, PT40-41, WT40-
41, and ZT40.
• Two bridge NI 9237 modules connected to PT20-21 and PT01-03.
• One NI 9401 module connected to ST40.
Due to the mounting of new pressure sensors in the runner, a slip-ring have been
installed on the turbine shaft to ensure synchronous sampling of both rotating and
stationary domain. The onboard sensors, which are wired through the slip ring, are
connected to an amplier to increase the voltage range going into the module. The
draft tube pressure sensors are connected to a converter that transforms the signal
from resistance per bar to voltage per bar.
A new logging program was made in LabVIEW with help from PhD Candidate Carl
Bergan in order to sample all signals simultaneously. Every signal, except from the
rotational speed, was sampled at 5 kHz. In total, 10 repetitions of the same tran-
sient measurement were carried out to be able to estimate the random uncertainty.
The values written to le were un-calibrated voltage values which afterwards were
manipulated to give a qualitative representation of the behaviour of the measured
quantities.
4.3 Computer simulation
All the geometries used in the numerical analyses were drawn in the CAD program
PTC Creo. The geometries represents a true copy of the Tokke turbine model in
the Waterpower Laboratory. The FSI analysis is a combination of a transient uid
analysis of the ow passage done in ANSYS CFX and a transient structural analysis
of the turbine runner done in ANSYS Mechanical. It is a one-way interaction,
meaning that deformations in the structure caused by the uid were not taken into
account in the ow analysis. To run the simulations, one of NTNU's supercomputers
was used, providing a quicker process.
42
4.3.1 CFD
A transient and turbulent CFD simulation was perfomed in ANSYS CFX by Postdoc.
Fellow Chirag Trivedi. Since it was conducted by another person than the author of
this thesis, only the basics of the analysis will be explained in the upcoming sections.
Further explanation can be found from the Francis-99 workshop 2 and corresponding
papers [2].
Geometry
The CFD simulation of the Tokke turbine model included the wetted surfaces of the
spiral casing, stay vanes, guide vanes, runner vanes, and draft tube. The CAD model
can be seen in Figure 4.10. When including the whole ow passage of the Francis
turbine, it is easier to obtain accurate predictions of the pressure distributions and
pressure pulsations on the runner blades. The turbine model was divided into four
domains: stationary domain 1 consisting of the spiral casing, stationary domain 2
consisting of the guide vanes, rotating domain consisting of the runner, and station-
ary domain 3 consisting of the draft tube. These domains were connected by using
appropriate interfaces.
Figure 4.10: Wet CAD model of the Tokke turbine
Labyrinths seals were not included due to the fact that modelling of uid ow in the
labyrinth seals require very ne meshing and substantial computational power. The
ow leakage losses in the turbine are thereby not investigated through CFD.
Mesh
The mesh was prepared using ICEM CFD with a tolerance ≥ 10−4 . On all the
domains, a hexahedral mesh grid was created by using 3D structured multi-blocks.
The total number of elements is 8128515 for the complete turbine, with 1 338 645, 2
579 790, 3 477 600, and 732 480, for the spiral casing, runner, guide vanes, and draft
43
tube, respectively. Fine mesh was created close to the boundaries and in complex
passages of the turbine.
Fluid properties and boundary conditions

An incompressible Newtonian uid was dened as the medium, with density, viscos-
ity, and temperature consistent with water properties. Buoyancy Model was set to
"Non Buoyant" and Reference Pressure to 1 atm. The standard k − model was
used, which is one of the most common turbulence models. It includes two extra
transport equations to represent the turbulent properties of the ow: the turbulent
kinetic energy, k, and the rate of dissipation of kinetic energy, epsilon. The model
applies the y+ method where a logarithmic overlap law is used to describe the bound-
ary layer nearby the wall. The Turbulent Wall Functions were set to "Scalable" for
all the domains. In addition, all the walls were dened with no-slip and smooth wall
boundary conditions.
Pressure at the inlet of the spiral casing and the runner speed were dened as bound-
ary conditions by using average measurement values from one of the experimental
measurements. Instead of using a sine curve to describe the runner speed, a straight
line was created between the two speeds of 333 RPM and 273 RPM. This was mainly
to simplify the simulation, however, it creates a small dierence between the results
from the laboratory and the results from the simulation.
Time Casing inlet

T [s] pressure Runner
[kPa] speed [rpm]
Start time 0 135 333
Transient start time 1 135 333
Transient end time 3 133 273
End time 4 133 273
Table 4.5: Boundary conditions
It needs to be mentioned that the pressure boundaries in the table above were taken
from a single measurement performed early in the semester and reect pressure
results from a RPM variation over 3 seconds. In contrast, the transient time domain
in the CFD simulation is 2 seconds. This was caused by incorrect information ow
between the dierent persons involved in this thesis. The CFD simulation had
already started running with these boundaries when the error was detected, leading
to conduction of new laboratory measurements with a time domain of 2 seconds.
Time step size

In this case, the transient component in the simulation is the turbine runner, thus
the time step is dened through the runner speed. The time step size is given
44
by Equation (4.2), where the value 0.1592 represents the conversion factor between
radians and revolutions. Each time step is equivalent to 0.5◦ of the runner revolution,
and results were recovered every 100th time step during the CFD simulation.
1
runnerspeed × 0.1592
timestep = (4.2)
360
Validation of CFD model

A validation of the CFD model was performed to check if the model is adequate for
practical purposes. This was done by comparing the results with the available exper-
imental data to check if the reality was represented accurately enough. Plots from
dierent pressure sensors were compared to pressure plots from specic coordinate
points in the CFD model. These coordinates, with related location uncertainties,
can be found in Appendix B and are illustrated with red crosses in the gure below.
Figure 4.11: Locations of pressure sensors in the CFD model
4.3.2 FEA
The structural analysis of the turbine runner was performed in ANSYS Mechanical
through the Transient Structural model system. The runner stand out as the most
vulnerable component since it is exposed to many sources of dynamic excitation. To
avoid failure, it is therefore of great interest to predict its dynamic behaviour.
45
Geometry
The 3D model consists of hub and shroud with labyrinth seals, cone, full length
runner blades, and splitter blades. It is shown in Figure 4.12. By using the whole
runner geometry, the inuence of all the components and the neighbouring blades
will be accounted for in the transient analysis. Holes for bolts were removed to
simplify the model, assuming that this would not signicantly inuence the results.
In ANSYS DesignModeler, all the parts were connected by creating one multibody
part, which ensures connection between all parts without using separate connection
regions.
(a) Runner (b) Blades
Figure 4.12: Structural CAD model of the Tokke runner
A feature called Virtual Topology in Mechanical lets the user group faces and/or
edges together, allowing to form virtual cells in order to reduce or improve surfaces.
Instead of permanently changing the geometry, virtual cells were generated automat-
ically with a low search intensity and then slightly modied. This provided nicer
surfaces to work with when creating the mesh.
Material properties
A new material library was created under Engineering Data in ANSYS Workbench.
JM 3 and JM 7 were characterized with the material properties from Table 4.1.
Material selection for each structural part was done in ANSYS Mechanical. By
using the correct material instead of using the default material "Structural steel", it
creates an improved foundation to compare experimental results with the computer
simulation.
Mesh
Since the author decided to create one multibody part, the runner parts share the
same mesh nodes in the connecting regions. Tetrahedral shaped mesh elements were
46
created on all of the runner bodies as this provided the best possible solution for this
specic model. On narrow faces, local sizing controls were implemented to decrease
the size of the elements. It is important with a ne mesh grid on areas where high
stresses are expected, like the llets and trailing edge of the runner blades. Complex
areas like the labyrinth seals also require small mesh elements to achieve a satisfying
mesh quality. These areas were therefore targeted. The smallest element size used
was 2 mm. Pinch Control were also inserted to repair a bad surface edge on the
leading edge of the runner blades.
To be able to conduct a mesh independence test, several dierent mesh grids were cre-
ated in separate modules. The Relevance and Span Angle centres were changed from
Coarse to Medium, and the Advanced Size Function called Curvature was turned o
and on to provide additional control over the global mesh sizing. Maximum element
size was set to 7 mm. A list of the mesh grids included in the independence analysis
are presented in Table 4.6. Stress probes, which outputs the specic stress of the
point where it is placed, were attached to one blade to compare the quality of the
mesh grids. Finer mesh grids than the one mentioned were created in the process,
but had to be rejected due to time and computation limitations.
Mesh # nodes Functions applied

1 1 786 521 Coarse Relevance Center and Coarse Span Angle Center
Medium Relevance Center and Medium Span Angle
2 2 024 838
Center
Medium Relevance Center and Medium Span Angle
3 2 582 903 Center, in addition to a smaller element size for the blade,
hub and shroud surfaces
Curvature, Coarse Relevance Center, and Coarse Span
4 3 526 317
Angle Center
Table 4.6: Mesh independence test
Boundary conditions
To ensure that the structural 3D model acts the same way as the physical model,
constraints and appropriate boundary conditions were applied to the runner. The
boundary conditions implemented are listed below:
• Fixed support on the surfaces connecting the hub and the shaft
• Acceleration due to gravity (g)
• Variable rotational speed implemented as modied data from the laboratory
log
The boundary condition Fixed support locks the displacement of the mesh nodes
on the related surfaces and replicates the impact a shaft would have had on the
47
structure. Standard earth gravity dened by ANSYS was added in negative z-

direction. Further, the same tabular data as used in the CFD analysis was used
as transient rotational velocity. It was important that the velocity functions were
equal for the two models to make sure that the pressure loads would map onto the
correct surfaces at each time step. The rotational velocity (rotational inertia) and
gravity (runner weight) together increases the load picture by imparting inertia force
on the runner. Axial pressure forces on the outer hub, shroud and labyrinth seals
were neglected from the analysis due to the complexity of the transient pressure
distribution and insucient information.
4.3.3 FSI
A one-way transient simulation was conducted. In contrast, a two-way FSI analysis
is time consuming and require a huge amount of computer resources. With the
current sti materials in the Tokke runner, it is also possible that the deformations
caused by the uid pressure would not have had a large impact on the uid ow
behaviour.
The ACT Transient FSI method was applied since the work was shared between
two persons, the author and Chirag Trivedi. First of all, the ACT extension had
to be downloaded, installed and implemented into ANSYS. Further, macros were
run via the CFD Post Macro calculator to create the CFD data les in the right
format. Then a FSI surface macro was used to export surface pressures from the hub,
shroud, blades and cone. In ANSYS Mechanical - Transient Structural, pressure was
imported as a load through the ACT extension and mapped onto the corresponding
surfaces of the structural runner model. Figure 4.13 presents an overview of the
boundary conditions, including the surface pressure, imparted on the runner.
Figure 4.13: Conditions and loads on structural model
48
Equivalent stress and strain were inserted as solver outputs to analyse stresses and
strain in the runner caused by uid-structure interaction. In addition, dierent
probes were added to extract results at specic coordinates and surfaces.
Validation of FSI
A validation of the FSI analysis was performed by comparing the results from the
strain gage measurements with strain results from the computational model. Strain
gage measurements are a powerful validation method because their results represent
area integral measures of the eect of pressure elds onto the structure. Coordinate
systems were added at the strain gage coordinates, which can be found in Appendix
B. Then, a path was created between the two points and snapped to the nearest
mesh nodes to follow the blade during rotation of the runner.
49
50
Chapter 5
Uncertainty analysis
The dierent sources of errors that contribute to uncertainty during model tests
are listed in Table C.1 and Table C.2 in Appendix C. This chapter provides the
most important uncertainties related to the results presented further on. The total
uncertainties are calculated to give a picture of how trustworthy the laboratory
results are. The experiment was repeated 10 times to be able to retrieve a standard
deviation.
The author have chosen to present the uncertainties related to the pressure, torque
and rotational speed measurements. Uncertainty analysis of the strain measurement
is not included since the strain gage only was calibrated theoretically, and thereby
lack information to perform a complete analysis. But as mentioned in subsection
4.2.2, the estimated uncertainty of the strain gage calibration curve is 0.25%.
5.1 Uncertainties in the calibrations
5.1.1 Static pressure transducers

The total systematic uncertainties of the pressure transducer calibrations were cal-
culated with 95% condence through the use of the calibration LabVIEW program
in the Waterpower Laboratory. This uncertainty includes the randomness in the cal-
ibrated instrument during the calibration, systematic uncertainty in the instrument
(dead weight manometer), and due to regression analysis to t the calibration points
to a linear calibration equation. The uncertainties are listed in Table 5.1 below.
51
5.1. UNCERTAINTIES IN THE CALIBRATIONS
Sensor max(f )[%] Sensor max(f )[%]

preg preg
PT10 ±0.4128 PT21 ±0.0527
PT12 ±0.3989 PT20 ±0.0396
PT13 ±0.3901 PT02 ±0.1942
PT14 ±0.3972 PT03 ±0.0488
Table 5.1: Total calibration uncertainties of pressure transducers
5.1.2 Torque
The uncertainties of the instruments used in the torque calibrations were not in-
cluded in the calibration LabVIEW program, and thus the total calibration un-
certainties have to be calculated theoretically. To calibrate the torques, weights
calibrated by the Norwegian Metrology Service were applied.
• The systematic uncertainty in the weights used for the calibrations, fτW , is
taken into account by using the calibrated kilogram values instead of the stated
kilogram values of the weights.
• Generator torque: The length of the torque arm was measured to be 0.9979
m. The related uncertainties of the shaft diameter, the rod, the calliper, and
the band width, are assumed to be ±0.5 µm, ±0.5 µm, ±5 µm, and ±5 µm, re-
spectively [44]. By combining the uncertainties one obtains fτGarm =±0.0005%
as the systematic uncertainty in the length of the arm.
Friction torque : The length of the torque arm was measured to be 0.1469
m. The related uncertainties of the shaft diameter, the calliper, and the block
width, are assumed to be ±5 µm, ±5 µm, and ±5 µm, respectively [44]. By
combining the uncertainties, one obtains fτFarm =±0.0042% as the systematic
uncertainty in the length of the arm.
• The systematic and random uncertainties in the instruments, fτc and fτd , are
included in fτreg . The maximum uncertainties in the regression lines were
max(fτGreg )=0.9298% for a generator torque of 62.23 Nm and max(fτFreg )=
6.1554% for a friction torque of 0.9454 Nm.
Combining the listed uncertainties above gives the total random uncertainties of:
q
max(fτGcal ) = ± +(fτGarm )2 + max(fτGreg )2 = ±0.9298% (5.1)
q
max(fτFcal ) = ± (fτFarm )2 + max(fτFreg )2 = ±6.1554% (5.2)
These are the maximum uncertainties, and do not fully represent the actual uncer-
tainties of the operating torque ranges.
52
5.2. UNCERTAINTIES IN THE TESTS
5.1.3 Rotational speed sensor

The rotational speed measurement system is not calibrated since it is assumed to
be accurate. However, there is a systematic uncertainty of ±0.025% linked to the
rotational speed in the Waterpower Laboratory [42].
5.2 Uncertainties in the tests
Figure 5.1 illustrates how the 10 measurements were merged/synchronized together

to nd the average curve and related 95% condence interval. This is done for all of
the measurement results.
(a) Synchronized measurement results
(b) Average measurement with 95% condence interval
Figure 5.1: Condence interval of repeated measurement
53
5.2.1 Static pressure transducers

• fpg is equal to fpcal , and is found in Table 5.1.
• fph is caused by the drift of the output signal over time. This error is ignored
here.
• fpk may be ignored since the calibrations of the static pressure transducers were
conducted using the same data acquisition system planned for the experiment.
• fpl is found by using the 95% condence interval for the logged data from the
tests. The maximum random uncertainties during the test are presented in
Table 5.2. A major part of this uncertainty may come from unsynchronized
measurements in time.
Sensor max(f )[%], 0-1 s max(f )[%], 1-3 s max(f )[%], 3-4 s
pl pl pl
PT10 ±0.007 ±0.012 ±0.024
PT12 ±0.015 ±0.016 ±0.017
PT13 ±0.030 ±0.024 ±0.019
PT14 ±0.090 ±0.067 ±0.057
PT21 ±0.003 ±0.004 ±0.003
PT20 ±0.002 ±0.003 ±0.003
PT02 ±0.001 ±0.001 ±0.001
PT03 ±0.045 ±0.037 ±0.032
Table 5.2: 95% condence interval uncertainties
5.2.2 Torque transducers

• max(fτGcal )=±0.9298% and max(fτFcal )=±6.1554%
• The calibration from the manufacturer is assumed to be correct, and there-
fore is the uncertainty fτh , caused by drifting of the output signal over time,
ignored.
• None physical sizes exists, as the torque is given directly by the voltage signal
from the weight cell. fτj can therefore be ignored.
• fτk may also be ignored since the conditions during the test were the same as
during the calibration.
• max(fτGl ) and max(fτFl ) are found in Table 5.3.
Gl Gl Gl
WT40 ±0.004 ±0.005 ±0.003
WT41 ±0.022 ±0.022 ±0.023
54
5.2.3 Rotational speed sensor

The maximum random uncertainty for the rotational speed measurements can be
found in Table 5.4.
nl nl nl
ST40 ±0.003 ±0.002 ±0.001
55
56
Chapter 6
Results
6.1 Turbine torque
In the laboratory, the torque on the generator and friction torque on the axial
bearing were measured. The torque that the runner is able to deliver is the sum
of these two since they both act on the same shaft. In Figure 6.1, the normalised
rotational speed, turbine torque and power of the unit during operation is seen. As
the speed of rotation decreases with 60 RPM, the torque on the turbine increases
with approximately 157 Nm, stating an increase of 25%.
P = ω(τG + τF ) (6.1)
Figure 6.1: Correlation between rotational speed, turbine torque and power output
57
6.2. CFD
6.2 CFD
The CFD simulation results received from Chirag starts at 0.7007 seconds, and not
at 0 seconds, due to unsatisfying results in-between that time range. To compare
pressure and strain results from the experiment and the simulations, the sensors co-
ordinates and structural model runner had to be rotated approximately 340 degrees
before use to account for the missing time steps in the CFD simulation. In addition,
the FSI results have only included 2/3 of the CFD pressure results due to reasons
explained further on in the upcoming sections.
6.2.1 Validation of CFD

The experimental pressure plots represents the average of 10 equal measurements.
As shown by the gures below, there is a dierence of around 20 kPa between the
results from the pressure sensors in the laboratory and the results from the CFD
simulation, where the CFD simulation gives the highest pressure. The inlet pressure
in the CFD simulation was set by the boundary condition of 135 kPa. This was given
by the average value of pressure sensor PT40 from a single measurement conducted
in the beginning of this work, which in turn was replaced by new measurements later
on. But when nding this boundary pressure value it was not taken into account
that one have to subtract the pressure height, ρgZ = 20.3 kPa, from the pressure
retrieved directly from the measurement for sensor PT40, due to the way the sensor
is connected to the pipe. The new measurements gave an inlet pressure of 137.4 kPa,
which is equal to 117.1 kPa when subtracting the pressure height. One can therefore
say that most of the dierence between the results lays in the pressure size of 137.4
- 135 kPa = 2.4 kPa plus the pressure height of 20.3 kPa. However, it is clear that
the majority of the pressure results follow the same trend. Pressure is here given as
static relative pressure in all the plots, expect for the dynamic representation of the
experimental pressure in the draft tube cone.
Spiral casing inlet, PT02

Figure 6.2 shows that the CFD plot miss out on some distinct changes in pressure
compared to the measurement plot, especially at around 2.5 seconds and after 3
seconds. This may be based on the time step size, boundary conditions, and/or
solution models dened in the simulation. The pressure dierence on the vertical
axis at the starting point is approximately 17.7 kPa. Even so, the total variation in
pressure during the decrease in rotational speed is approximately 2.5 kPa for both
result plots. This imply a sucient correlation between the measurements and the
simulation.
58
6.2. CFD
(a) LAB
(b) CFD
Figure 6.2: Comparison of pressure results from sensor PT02
Runner, PT10-14
Figure 6.3 shows the placement of each pressure transducer mounted into the runner
hub. Sensor PT11 stopped working right before the experiment was performed, but
that was not crucial for the validation of the CFD model.
59
6.2. CFD
Figure 6.3: Sensors mounted in the runner hub
At the inlet of the runner between a full length blade and a splitter blade, one can
nd sensor PT10. While decreasing the runner speed, the pressure at the runner
inlet also decreases, as seen in Figure 6.4. The change in the experimental result is
equal to a drop of approximately 19.5 kPa in contrast to the pressure variation of
approximately 9 kPa for the CFD result. This is a dierence of considerable size.
The vertical dierence between the pressure plots is approximately 21 kPa.
It can be seen that the pressure is not fully stabilized before 3.5 seconds, and not after
3 seconds, which is the time when the rotational speed variation stops. The shape
change in the end of the CFD plot is probably bases on the coarse plot distribution,
and do not represent an actual increase in pressure.
60
6.2. CFD
(a) LAB
(b) CFD
Downstream the splitter blade towards the outlet of the runner, the pressure is
increasing as the rotational speed is decreasing. Here, the increase for both plots
is approximately 1.1 kPa. The vertical dierence between the pressure plots is
approximately 16.5 kPa. The changes in the curve shape at the beginning and the
end of the CFD plot are also most likely based on the coarse result distribution
caused by few time steps. Pressure plots from sensor PT12 and PT14 can be found
in Appendix G if further study of pressure in the runner is of interest.
61
6.2. CFD
(a) LAB
(b) CFD
Vanless space, PT20-21

In the vaneless space, there is a pressure drop of approximately 10.5 kPa when the
runner speed decreases. The plots have a satisfying trend correlation. Ant the
vertical dierence between the pressure plots is approximately 20 kPa, which is a
good t with the PT40 input error.
62
6.2. CFD
(a) LAB
(b) CFD
Draft tube cone, PT30-33

The pressure sensors used in the draft tube cone are dynamic sensors, which means
that they only react to pressure changes, and do not measure static pressure. It
is therefore not correct to compare Figure 6.7a and Figure 6.7b by looking at the
dierence in Pa. But the plots follow the same trend, which can imply that the
simulated ow in the draft tube is acceptable and captures the most important
disturbances.
63
6.2. CFD
(a) LAB
(b) CFD
Draft tube outlet, PT03

Figure 6.8 represents the pressure distribution near the outlet of the draft tube, and
shows a phase shift between the two plots. This is most likely because the placement
of the read-o point on the CFD model is incorrect. The pressure dierence between
the two plots is approximately 9.7 kPa when considering the highest peaks in the
plots.
64
6.3. FSI
(a) LAB
(b) CFD
6.3 FSI
6.3.1 Mesh independence test

Figure 6.9 presents the mesh independence test conducted by the author, where
Figure 6.9b illustrates results at time 0.9 seconds. The test was performed over a
time range of approximately 0.4 seconds; the rst few seconds of the total analysis.
Results are achieved from a stress probe placed at a point of high stress on the leading
edge of a runner blade. By studying the plots one can see that mesh number 1 clearly
stands out from the other mesh grids, and it is concluded to be an insucient mesh
with low credibility. Mesh number 3 outputs the highest stress overall, however, it
peaks to a stress value that might be misleading. This is further illustrated in Figure
65
6.3. FSI
G.2 in Appendix G.
(a) Stress vs. time
(b) Stress vs. mesh size
Figure 6.9: Mesh independence test - Leading edge
Mesh independence was unfortunately not reached, and the author had to base the
selection of mesh mainly on the quality of the mesh grids. Mesh number 4 was
thereby chosen for further analysis, with 3 526 317 elements and 5 225 788 nodes.
Figure 6.10 shows how a part of the turbine runner is meshed. The mesh quality
was checked by inspecting the Element Quality diagram, where 0 represents a bad
element and 1 is a perfect element. This diagram can be found in Appendix G. The
majority of the elements for mesh 4 were good, with a average element quality of
0.81645 and a standard deviation of 0.11483. The worst shaped elements were found
on the labyrinths seals and on some of the leading edges of the splitter blades towards
the shroud. Since the labyrinth ow was neglected from the analysis, the mesh at
the labyrinth seals is not critical for the FSI results. The worst mesh element has a
quality of 0.015, while the best has a quality of 1.
66
6.3. FSI
Figure 6.10: Mesh on runner blade
6.3.2 Validation of FSI

Due to concerns about the reability of the strain gage, the strain gage baseline was
measured before and after the the test rig was run, while the runner was at rest and
lled with water. The rst measurement result showed an oset of approximately
-85×10−6 . This continued to increase during the time the test rig was in operation,
and in the end, the strain gage measurement showed a baseline of -100×10−6 . How-
ever, this last oset was logged after an additional 10 runs that not are included
in this thesis. As a consequence, it is dicult to decide how much the strain gage
output has changed between each measurement. The author chose therefore to only
present the result from the rst measurement, and subtract the oset of -85×10−6 .
Figure 6.11: Strain gage results from laboratory measurement
67
6.3. FSI
The strain result shown in Figure 6.11 indicate compression of the material since
the values are negative, and have a strain range of 2.8×10−6 over the time period
0.7 to 2.4 seconds. The author was not able to predict what are causing the increase
in the curve after 2.4 seconds, but the compression is decreasing.
The best way to compare measured and simulated stresses for a Francis runner is
to compare the strain components in direction of the strain gage [40]. A path that
followed the direction of the actual strain gage was attached to the surface mesh, but
this only solved strain along the length of the path, and not versus time. A picture
of the strain path is shown in Figure 6.12. The author managed to get solutions
from the rst and last time step, which have average strain values of 6.4×10−6
and 3.68×10−6 , respectively. The positive values imply a decreasing tension in the
runner blade, which is a contradiction to what the strain gage results are showing.
Lack of time caused this validation to remain incomplete. Further work have to be
conducted to get a proper validation of the stress and strain in the runner.
Figure 6.12: Strain path
6.3.3 FSI results

The pressure mapped onto the structural runner only represents uid pressure from
0.7 seconds to 3 seconds. The last part from 3 to 4 seconds was dicult to implement
in a continuous transient analysis with the specic FSI method applied. This was
because the CFX results were handed over to the author in three separate result
les, where the last time range had been set as the rst solver steps of the solution
since this part was nished solving rst. When trying to merge the solutions, the
last time range was interpreted as the rst time range which is obviously wrong, so
this part of the CFD solution had to be rejected. Another problem appearing when
trying to merge the result les, was that only the full time steps could be used, not
the partial time steps. As a consequence, the FSI simulation has been conducted
68
6.3. FSI
with as few as 11 time steps over a time period of 2.3 seconds.

Figure 6.13 illustrates how the pressure from the CFD simulation is mapped onto the
structural runner. The pressure distribution shows the range for the last time step.
As expected, the pressure decrease through the runner due to energy conversion.
The maximum pressure input at the last time step has a magnitude of 71.5 kPa.
Figure 6.13: Imported pressure in ANSYS Mechanical
Equivalent stress
The stress results were analysed as von-Mises equivalent stress. Von-Mises stress is
widely used by designers to check if their design can withstand a given load condition.
The design will fail if the maximum value of von-Mises stress induced in the material
is higher than the strength of the material. Further information on the von-Mises
stress criterion is found in Appendix G.
Figure 6.14: Maximum stress on runner over time [Pa]
69
6.3. FSI
Results from the FSI simulation shows that the material stresses in the runner
decreases as the rotational speed decreases. The maximum stress found in the runner
decreases from 8.4 MPa to 5 MPa, equal to a stress decrease of 40.4% during the
time period of 2.3 seconds. The runner blades experience the highest stress, while
the minimum stress is found in the cone.
Figure 6.15: Maximum stress in runner at last time step [Pa]
Figure 6.16: Maximum stress in blades at last time step [Pa]
The materials JM 7 and JM 3 have yielding limits of 260 Mpa and 150 MPa. As
explained in section 2.1, this is dened as the stress at which a material begins to
deform plastically. With maximum stress of 8.4 MPa in the turbine runner, it is
very unlikely for the runner to deform as a result of the pressure forces acting on
the runner during variation in rotational speed.
70
6.3. FSI
Figure 6.17: Stress distribution on pressure side of runner blade [Pa]
Figure 6.18: Stress distribution on suction side of runner blade [Pa]
On the runner blades, the highest stresses appear at the leading and trailing edges.
The maximum stress is found at the trailing edges at the distinct tip towards the
shroud. This is not a surprise, as the blade is at its thinnest here. For the leading
edges, the stress is especially high in the joint between blade and shroud on the pres-
sure side of the blade. Table 6.1 presents results extracted from specic placements
around on a runner blade at the start and the end time of the simulation. Edge 1 is
the lower edge towards the shroud, while edge 2 is the upper edge towards the hub.
Leading egde 1 Leading edge 2 Trailing egde 1 Trailing egde 2

4.97 MPa 2.84 MPa 7.07 MPa 2.28 MPa
3.62 MPa 1.49 MPa 4.09 MPa 1.76 MPa
Table 6.1: Probe results from dierent surfaces/edges on a runner blade
71
6.3. FSI
72
Chapter 7
Discussion
7.1 Known error sources
7.1.1 Rotational speed functions

As mentioned, the rotational speed functions used in the laboratory and in the
simulations are dierent. The one used for the laboratory measurements have a
smooth transition in the change between steady and transient operation. It uses
a sine wave to illustrate the variation in RPM. This was mainly to conduct safe
operation of the test rig, thus sparing the generator for drastic changes, in addition
to the fact the the grid frequency act in a way similar to a sine curve. To simplify the
CFD analysis, the runner speed variation was constructed from four points, instead
of using a direct input from the measurements. By looking at Figure 7.1, it is
clear that the pressure results will be aected by this dierence during the transient
change. Before 2.25 seconds, the runner speed is highest for the sine curve, while
after 2.25 seconds, the straight curve gives out the highest speed. By drawing a link
between pressure and rotational speed, one can say that the pressure will change
faster for the CFD simulation, which is clearly seen in Figure 6.2, Figure 6.4, and
Figure 6.6.
73
7.1. KNOWN ERROR SOURCES
Figure 7.1: Rotational speed functions
7.1.2 Operation points

Table 7.1 shows the dierence in experimental and computational operating values
at the starting point. In the laboratory, the guide vane angle and the runner speed
were set to 9.84◦ and 333 RPM, respectively. But uctuations and uncertainties in
the laboratory LabVIEW program caused the average values to get small osets.
Parameter CFD Laboratory

Guide vane angle [◦ ] 9.84 9.87
Net head [m] 13.97 12.63
Discharge [m3 /s] 0.2 0.197
Runner angular speed [rpm] 333 332.8
Hydraulic eciency [%] 82.09 85.9
Atmospheric pressure [kPa] 101.3 100.7
Casing inlet pressure-abs [kPa] 236.33 217.8
Draft tube outlet pressure-abs [kPa] 116.57 96.0
Table 7.1: Operating points
Net head and eciency of the Francis turbine is calculated by using Equation (2.9)
and Equation (2.7). The experimental value of ∆p in the equations is provided by
pressure sensor PT41 and represents the dierence in pressure from inlet to outlet of
the model turbine. The dierence in the boundary pressures between the CFD and
the laboratory measurements was discussed in the the previous chapter and is the
basis for the majority of the deviations in pressure and strain results. It is important
also to remember that the numerical estimation of the parameters is inuenced
by many other factors such as turbulence modelling, combined performance of the
domains, interface modelling, and performance of the mesh grid under dierent ow
conditions.
74
7.2. COMPUTER SIMULATIONS
7.1.3 Time synchronization of measurements

The measurements were logged without using a triggering time system, meaning
that the author decided when each measurement series should start and stop. As
a consequence, the measurement series starts at dierent times, and it has been
dicult to decide exactly when the change in RPM starts for each measurement.
When synchronizing the measurements to nd the average measurement result, the
10 equal measurements do not lie exactly on top of each other, thus increasing the
uncertainty of the laboratory result.
7.1.4 Sensor positions

The sensor coordinates found in Appendix B comes with uncertainties provided in
mm. The surveying of the sensor locations in the laboratory have not been accurately
executed. This cause dierences in the results from the experimental measurements
and the computational simulations. Since the rst deciseconds of the CFD simulation
was removed, the coordinates related to the runner had to be rotated approximately
340 degrees around the vertical axis, which may have increased the uncertainty of
the coordinate placements in the computational models.
7.2 Computer simulations
7.2.1 Time step

The pressure measurements conducted in the laboratory have a sampling rate of 5
kHz which gives an accurate representation of the results, independent of revolutions
of the runner. In contrast, the CFD simulation have a time step given by Equation
4.2, leading to jagged and less accurate pressure results. However, a simulation with
many time steps over a long time range takes much time and requires enormous
computer capacity to calculate.
7.2.2 y+ values from CFD simulation

The use of scalable wall functions in ANSYS CFX leads to production of consistent
results for grids of varying y+. They activate the local usage of the log-law in
regions where the y+ is suciently small, in connection with the standard wall
function approach in coarser y+ regions. Table 7.2 presents an overview of dierent
y+ values in the turbine runner at dierent times. They represents the average y+
values from each node on the specied surface. All the y+ values are well below the
maximum y+ requirement of 300, thus should be able to give an accurate prediction
of boundary layer ow.
75
7.3. FSI
Region 0.75 1 2 3 4
Hub 66.8 66.8 84.5 110.8 111.6
Full-length blades 47.5 47.5 52.0 58.5 59.1
Splitter blades 25.0 25.0 27.7 29.3 29.6
Shroud 155.1 155.2 148.7 134.0 133.7
Table 7.2: y+ values for the CFD model runner
7.2.3 Computational model geometry

Areas which do not aect the result values should have been smoothed out to simplify
the meshing process and avoid increasing mesh densities for the structural model.
This applies in particular to the labyrinth seals, due to the fact that the labyrinth
ow was neglected in the CFD simulation and the axial thrust forces were neglected
from the FEM simulation. Furthermore, the author could have given more thought
to the idea of looking at only 1/15 of the turbine runner to reduce the complexity
of the analysis.
7.2.4 Mesh independence test

A mesh grid independent of the results was not achieved in this work. As a conse-
quence, the magnitudes of result values are uncertain. However, the author created
mesh grids based on knowledge from previous similar studies, with ner mesh cells in
areas where it was expected to see high stress and strain. Lack of computer capacity
and time to perform a simulation with a large number of mesh elements, made it
more dicult to gain a perfect mesh.
7.3 FSI
There are several ways to perform a FSI analysis in ANSYS. If the CFX and Me-
chanical analyses are performed in the same ANSYS Workbench, the modules could
be directly connected and results would be able to be transferred back and fourth.
This was not the case for this project. The author received three CFX result les
which together supplied the total solution. Then, there were two possible ways to
conduct the FSI analysis; either save separate pressure les for each component and
each time step and import into Mechanical as external data, or use the ACT Tran-
sient FSI extension and save three les to import into Mechanical. For transient
analysis, the ACT Transient method is much easier and less time consuming than
using external data. But for this case, the solver steps of the CFX result les were
saved in the wrong order, in addition to separated in three les, which made it more
dicult to create the needed load les.
76
7.3. FSI
7.3.1 Pressure mapping

Since the mesh grids (uid and solid) on the runner were created in separate workspaces,
it is not ensured that the mesh elements corresponds at the uid-solid interface for
accurate transmission of the pressure loads. The CFD runner consist of 2 579 790
mesh elements while the FEM model consist of 3 526 317 mesh elements. Unfortu-
nately, the author was unable to retrieve mapping statistics due to problems with
the simulation.
7.3.2 Deformation of runner

Compared to a prototype turbine runner, the small model runner dimensions gives
out very low stresses. The model runner in the Waterpower laboratory also have sti
materials to ensure a long lifetime, which increases the resistance towards deforma-
tion. However, it is worth noting that the curved trailing edge of the runner blades
in the model runner makes it more sensitive to dynamic pressure loading compared
to a runner with straight cut trailing edges.
77
7.3. FSI
78
Chapter 8
Conclusion
In this thesis, a transient FSI analysis of a Francis runner has been carried out.
The runner was subjected to a decrease in rotational speed of 60 RPM over a time
period of 2 seconds while the guide vane opening was held constant. The aim was to
investigate how variations in grid frequency impact stresses in the turbine. Results
were validated against measured values from an experiment conducted on the same
turbine model in the Waterpower Laboratory. The simulation results showed that
the overall equivalent stress in the runner decrease as the speed decreases. The
highest stress was found at the blade trailing edge towards the shroud, where the
blade is at its thinnest. At the beginning of the simulation, the magnitude of this
stress value was 8.4 MPa. It decreased down to 5 Mpa, stating a stress change of
40.4%.
Validation of the CFD simulation showed a satisfying correlation between measured
and simulated pressure through the turbine. This conclusion is drawn when looking
beyond the fact that the input boundary conditions to the CFD simulation had
an error of known size. However, the author did not manage to validate stress
and strain in the structural model due to diculties with the FSI simulation and
uncertainties related to the physical strain gage. Therefore, further work is needed
to fully investigate the complete result history.
79
80
Chapter 9
Further work
• Proper validation of the stress results in the runner is needed. Using a strain
gage is good way to conduct this validation even though the author did not
manage to do so within the time limit given by this thesis.
• In order to obtain a more detailed picture of the dynamic behaviour of the
turbine, realistic oscillations of the runner should be examined, allowing to gain
insight into more complex uid-structure interaction phenomena that might
occur.
• Additional operating points outside BEP could be of interest, where several
ow characteristic are more active.
• Conducting a frequency analysis on the laboratory results allow detection of
pulsations from RSI and other disturbances in the ow. This makes it easier
to conclude on what the results are based upon.
• A mesh grid that is independent of the results should be achieved. This will
give a more accurate computational solution.
• Conducting a two-way FSI simulation would provide more trustworthy results.
The optimal way to perform such an analysis is to conduct both FEM and
CFD in the same workspace to make sure the geometries and mesh grids are
corresponding.
• Leakage ow through the labyrinth seals and additional axial forces on hub,
shroud and labyrinth seals should be included in the FSI analysis.
81
Bibliography
[1] Filtering and smoothing data. http://www.me.berkeley.edu/~lwlin/me128/

FEMNotes.pdf. Accessed: 17.02.2016.
[2] Francis-99, second workshop (2016), test case. https://www.ntnu.edu/nvks/
test-case. Accessed: 18.04.2016.
[3] Francis hydro turbine. https://renewables.gepower.com/hydro-power/
large-hydropower-solutions/hydro-turbines/francis-turbine.html.
Accessed: 03.02.2016.
[4] Ansys cfx. http://www.ansys.com/Products/Fluids/ANSYS+CFX, . Accessed:
25.01.2016.
[5] Ansys mechanical. http://www.ansys.com/Products/Structures/ANSYS+
Mechanical, . Accessed: 25.01.2016.
[6] Purpose of a calibration? https://msl.irl.cri.nz/services/
temperature-and-humidity/humidity-calibration-service/
calibration-certificate/purpose-of-a- . Accessed: 22.05.2016.
[7] Dening uid-strucutre interactions. https://www.comsol.com/
multiphysics/fluid-structure-interaction. Accessed: 05.05.2016.
[8] Introduction to nite element modeling. http://se.mathworks.com/help/
curvefit/smoothing-data.html. Accessed: 25.04.2016.
[9] Kraftmarkedet. http://www.fornybar.no/kraftmarkedet. Accessed: 27-01-
2016.
[10] Hydroelectric power. http://www.mpoweruk.com/hydro_power.htm. Accessed:
01.06.2016.
[11] Francis-99. https://www.ntnu.edu/web/norwegian-hydropower-center/
francis-99. Accessed: 15.02.2016.
[12] Eo stÅl as - tekniska data. http://www.eostal.se/Templates/
SearchResult1.aspx?CMS_SearchString=teknisk+data+bronser. Accessed:
11.05.2016.
82
BIBLIOGRAPHY
[13] Pressure measurement overview. http://www.ni.com/white-paper/13034/

en/. Accessed: 04.05.2016.
[14] Measuring strain with strain gages. http://www.ni.com/white-paper/3642/
en/. Accessed: 04.05.2016.
[15] What is von mises stress? http://www.learnengineering.org/2012/12/
what-is-von-mises-stress.html. Accessed: 05.06.2016.
[16] Stine Andersen, Martha Marie Øberg, Siri Veila, and Helene Sundheim. Ener-
giskolen, Lærehefte. Statnett, 8 2014.
[17] André Bakker. Lecture 5 - solution methods. applied computational uid dy-
namics, 2002. http://www.bakker.org, Fluent Inc.
[18] Jonas Bergmann-Paulsen. Fsi-analysis of a francis turbine. 2012.
[19] Hermod Brekke. Pumper og turbiner. NTNU, Vannkraftlaboratoriet, 2003.

[20] Hermod Brekke. Konstruksjon av pumper og turbiner. Waterpower laboratory,
NTNU, 2008.
[21] Hermod Brekke. Design, performance and maintenance of francis turbines.
Global Journal of Research In Engineering, 13(5), 2013.
[22] Bjarne Børresen. Sprekker i løpehjul. analyser, forebygging og erfaringer.
[23] William D Callister, David G Rethwisch, et al. Materials science and engineer-
ing: an introduction, volume 7. Wiley New York, 2007.
[24] A Deepak, CJ Roger, and J Gerald. The impact of hydroelectric power and other
forms of generation on grid frequency stability for the wecc region. HydroVision
International, Sacramento, CA, July, pages 1922, 2011.
[25] Rakel Ellingsen and Pål-Tore Storli. Simulations of the dynamic load in a fran-
cis runner based on measurements of grid frequency variations. International
Journal of Fluid Machinery and Systems, 8(2):102112, 2015.
[26] Christopher Frey. One-way uid-structure-interaction enhanced by act, 2014.
[27] D Frunzaverde, S Muntean, G Marginean, V Campian, L Mar³avina, R Terzi,

and V erban. Failure analysis of a francis turbine runner. In IOP Confer-
ence Series: Earth and Environmental Science, volume 12, page 012115. IOP
Publishing, 2010.
[28] Ove S Grande, G Doorman, and BH Bakken. Exchange of balancing resources

between the nordic synchronous system and the netherlands/germany/poland.
SINTEF Energy Research, Project, (12X535):02, 2008.
[29] Sigurd Tangerud Haga. Dynamic load on high head francis turbines during
start/stop. 2014.
83
BIBLIOGRAPHY
[30] Russell C Hibbeler. Statics and mechanics of materials. Pearson Higher Ed,
2013.
[31] Frida Holand. Fremtidens frekvensstabilitet i kraftsystemet: roterende reserver,

primærreserver og ny uregulerbar kraft i synkronområdet norden i 2020. 2014.
[32] Julie Marie Hovland. Pressure pulsations and stress in a high head francis model
turbine. 2013.
[33] Arne Kjølle. Mechanical equipment. Waterpower laboratory, NTNU, 2001.

[34] Einar Kobro. Measurement of pressure pulsations in francis turbines. 2010.
[35] Alberto Luna-Ramírez, Alfonso Campos-Amezcua, Oscar Dorantes-Gómez,

Zdzislaw Mazur-Czerwiec, and Rodolfo Muñoz-Quezada. Failure analysis of
runner blades in a francis hydraulic turbinecase study. Engineering Failure
Analysis, 59:314325, 2016.
[36] Camilla Odelbrink. Design of an alternative frequency controller for hydropower
stations: Reducing the 60-second oating in the grid frequency of the nordic
power system. 2014.
[37] Jørgen Ramdal and Ole Gunnar Dahlhaug. Procedures for the Francis Turbine
Test Rig. Waterpower Laboratory, NTNU, 2006.
[38] Trent Ratzla. Eects Future Renewable Installations Will Have on System
Synchronous and Synthetic Inertia. PhD thesis, TU Delft, Delft University of
Technology, 2012.
[39] Kristoer Rundhaug. Development of new mechanical design of model turbines.

2014.
[40] U Seidel, B Hübner, J Löad, and P Faigle. Evaluation of rsi-induced stresses in

francis runners. In IOP Conference Series: Earth and Environmental Science,
volume 15, page 052010. IOP Publishing, 2012.
[41] U Seidel, C Mende, B Hübner, W Weber, and A Otto. Dynamic loads in francis
runners and their impact on fatigue life. In IOP Conference Series: Earth and
Environmental Science, volume 22, page 032054. IOP Publishing, 2014.
[42] IEC Standard. 60193. IEC: International Electrotechnical Commission. Hy-
draulic turbines, storage pumps, and pump-turbines. Model acceptance tests.
Publication data, pages 1101, 1999.
[43] Systemdrifts- og Markedsutviklingsplan 2014-2020. Tiltaksplan for sikker og ef-
fektiv drift av kraftverksystemet. Statnett, 2014.
[44] Pål-Tore Storli. Modelltest av francis turbin i vannkraftlaboratoriet ved ntnu.
2006.
84
BIBLIOGRAPHY
[45] PT Storli and TK Nielsen. Dynamic load on a francis turbine runner from
simulations based on measurements. In IOP Conference Series: Earth and
Environmental Science, volume 22, page 032056. IOP Publishing, 2014.
[46] Andrea Stranna. Testing of rpt in pumping mode of operation. 2013.
[47] Pieter Tielens and Dirk Van Hertem. Grid inertia and frequency control in
power systems with high penetration of renewables. status: published, 2012.
[48] Chirag Trivedi, Bhupendra Gandhi, and Cervantes J Michel. Eect of transients
on francis turbine runner life: a review. Journal of Hydraulic Research, 51(2):
121132, 2013.
[49] Chirag Trivedi, Michel J Cervantes, and Ole G Dahlhaug. Experimental and
numerical studies of a high-head francis turbine: A review of the francis-99 test
case. Energies, 9(2):74, 2016.
[50] Henk Kaarle Versteeg and Weeratunge Malalasekera. An introduction to com-

putational uid dynamics: the nite volume method. Pearson Education, 2007.
[51] Pierre JC Vogler-Finck and Wolf-Gerrit Früh. Evolution of primary frequency
control requirements in great britain with increasing wind generation. Interna-
tional Journal of Electrical Power & Energy Systems, 73:377388, 2015.
[52] Hermann-Josef Wagner and Jyotirmay Mathur. Introduction to hydro energy
systems: basics, technology and operation. Springer Science & Business Media,
2011.
[53] Ivar Wangensteen. Power system economics: the nordic electricity market. 2012.
85
BIBLIOGRAPHY
86
Appendix A
Risk assessment
I
INNHOLDSFORTEGNELSE
1 INNLEDNING .................................................................................................................... 1
2 ORGANISERING................................................................................................................ 1
3 RISIKOSTYRING AV PROSJEKTET ...................................................................................... 1
4 TEGNINGER, FOTO, BESKRIVELSER AV FORSØKSOPPSETT .............................................. 1
5 EVAKUERING FRA FORSØKSOPPSETNINGEN................................................................... 2
6 VARSLING......................................................................................................................... 2
6.1 Før forsøkskjøring ............................................................................................................ 2
6.2 Ved uønskede hendelser ................................................................................................. 2
7 VURDERING AV TEKNISK SIKKERHET ............................................................................... 3
7.1 Fareidentifikasjon, HAZOP............................................................................................... 3
7.2 Brannfarlig, reaksjonsfarlig og trykksatt stoff og gass .................................................... 3
7.3 Trykkpåkjent utstyr ......................................................................................................... 3
7.4 Påvirkning av ytre miljø (utslipp til luft/vann, støy, temperatur, rystelser, lukt) ........... 4
7.5 Stråling............................................................................................................................. 4
7.6 Bruk og behandling av kjemikalier .................................................................................. 4
7.7 El sikkerhet (behov for å avvike fra gjeldende forskrifter og normer)............................ 4
8 VURDERING AV OPERASJONELL SIKKERHET.................................................................... 4
8.1 Prosedyre HAZOP ............................................................................................................ 4
8.2 Drifts og nødstopps prosedyre ........................................................................................ 4
8.3 Opplæring av operatører................................................................................................. 5
8.4 Tekniske modifikasjoner.................................................................................................. 5
8.5 Personlig verneutstyr ...................................................................................................... 5
8.6 Generelt ........................................................................................................................... 5
8.7 Sikkerhetsutrustning ....................................................................................................... 5
8.8 Spesielle tiltak.................................................................................................................. 5
9 TALLFESTING AV RESTRISIKO – RISIKOMATRISE ............................................................. 5
10 KONKLUSJON ................................................................................................................... 1
11 LOVER FORSKRIFTER OG PÅLEGG SOM GJELDER ............................................................ 6
12 DOKUMENTASJON........................................................................................................... 6
13 VEILEDNING TIL RAPPORTMAL........................................................................................ 7
1 INNLEDNING
Stasjonære og transiente trykkmålinger skal gjennomføres flere steder i Francis riggen.
Forsøket skal forgå i April 2016.
2 KONKLUSJON
Riggen er bygget til god laboratorium praksis (GLP).
Apparaturkortet får en gyldighet på 4 måneder

Forsøk pågår kort får en gyldighet på 4 måneder
3 ORGANISERING
Rolle NTNU
Prosjektleder Pål-Tore Storli/
Ole Gunnar Dahlhaug
Apparaturansvarlig Bård Brandåstrø
Romansvarlig Halvor Haukvik
HMS koordinator Morten Grønli
HMS ansvarlig (linjeleder): Olav Bolland
4 RISIKOSTYRING AV PROSJEKTET
Hovedaktiviteter risikostyring Nødvendige tiltak, dokumentasjon DTG
Prosjekt initiering Prosjekt initiering mal x
Skjema for Veiledningsmøte med
Veiledningsmøte x
pre-risikovurdering
Fareidentifikasjon – HAZID
Innledende risikovurdering x
Skjema grovanalyse
Prosess-HAZOP
Vurdering av teknisk sikkerhet x
Tekniske dokumentasjoner
Prosedyre-HAZOP
Vurdering av operasjonell sikkerhet x
Opplæringsplan for operatører
Uavhengig kontroll
Sluttvurdering, kvalitetssikring Utstedelse av apparaturkort
Utstedelse av forsøk pågår kort
5 TEGNINGER, FOTO, BESKRIVELSER AV FORSØKSOPPSETT

Vedlegg:
Prosess og Instrumenterings Diagram, (PID) skal inneholde:
 Alle komponenter i forsøksoppsetningen
 Komponentliste med spesifikasjoner
 Tegninger og bilder som beskriver forsøksoppsetningen.
Hvor oppholder operatør seg, hvor er gassflasker, avstegningsventiler for vann/luft.
Annen dokumentasjon som beskriver oppsett og virkemåte.
1
6 EVAKUERING FRA FORSØKSOPPSETNINGEN
Evakuering skjer på signal fra alarmklokker eller lokale gassalarmstasjon med egen lokal
varsling med lyd og lys utenfor aktuelle rom, se 6.2
Evakuering fra rigg området foregår igjennom merkede nødutganger til møteplass, (hjørnet
gamle kjemi/kjelhuset eller parkeringsplass 1a-b.)
Aksjon på rigg ved evakuering: Slå av luft og vanntilførsel
7 VARSLING
7.1 Før forsøkskjøring
Varsling per e-post, til Liste iept-experiments@ivt.ntnu.no
I e-posten skal det stå::
 Navn på forsøksleder:
 Navn på forsøksrigg:
 Tid for start: (dato og klokkelslett)
 Tid for stop: (dato og klokkelslett)
All forsøkskjøringen skal planlegges og legges inn i aktivitetskalender for lab. Forsøksleder
må få bekreftelse på at forsøkene er klarert med øvrig labdrift før forsøk kan iverksettes.
7.2 Ved uønskede hendelser

BRANN
Ved brann en ikke selv er i stand til å slukke med rimelige lokalt tilgjengelige slukkemidler,
skal nærmeste brannalarm utløses og arealet evakueres raskest mulig. En skal så være
tilgjengelig for brannvesen/bygningsvaktmester for å påvise brannsted.
Om mulig varsles så:
NTNU SINTEF
Morten Grønli, Mob: 918 97 515
Olav Bolland: Mob: 918 97 209
NTNU – SINTEF Beredskapstelefon 800 80 388
GASSALARM
Ved gassalarm skal gassflasker stenges umiddelbart og området ventileres. Klarer man ikke
innen rimelig tid å få ned nivået på gasskonsentrasjonen så utløses brannalarm og laben
evakueres. Dedikert personell og eller brannvesen sjekker så lekkasjested for å fastslå om
det er mulig å tette lekkasje og lufte ut området på en forsvarlig måte.
Varslingsrekkefølge som i overstående punkt.
PERSONSKADE
 Førstehjelpsutstyr i Brann/førstehjelpsstasjoner,
 Rop på hjelp,
 Start livreddende førstehjelp
 Ring 113 hvis det er eller det er tvil om det er alvorlig skade.
2
ANDRE UØNSKEDE HENDELSER (AVVIK)
NTNU:
Rapportering av uønskede hendelser, Innsida, avviksmeldinger
https://innsida.ntnu.no/wiki/-/wiki/Norsk/Melde+avvik
SINTEF:
Synergi
8 VURDERING AV TEKNISK SIKKERHET

8.1 Fareidentifikasjon, HAZOP
Se kapittel 13 ”Veiledning til rapport mal.
Forsøksoppsetningen deles inn i følgende noder:
Node 1 Rørsystem med pumpe
Node 2 Roterende utstyr (turbin og gir)
Node 3 Hydraulikk
Vedlegg, skjema: Hazop_mal
Vurdering
Node1:
 Overtrykksventil som slår ut dersom trykket i systemet blir for høyt.
 Rørelementer er eksternt levert og godkjent for aktuelt trykk.
Node2:
 Roterende utstyr står utilgjengelig for folk. Dvs det er innkapslet eller man må klatre
for å nå opp til det.
Node3:
 Trykk i slanger og rør (olje/vann) Hydraulikkslanger er ikke egenprodusert
 Trykksatt utstyr er sertifisert og kjøpt inn av eksterne leverandører
8.2 Brannfarlig, reaksjonsfarlig og trykksatt stoff og gass

Inneholder forsøkene brannfarlig, reaksjonsfarlig og trykksatt stoff
JA Trykksatt hydraulikkolje, trykksatt vann
Vurdering: Arbeidsmedium er vann. Alle rør er levert av eksternt firma med prøvesertifikat
Hydraulikk til hydrostatisk lager. Hyllevare komponenter, de er dermed ikke egenprodusert
8.3 Trykkpåkjent utstyr

Inneholder forsøksoppsetningen trykkpåkjent utstyr:
JA Utstyret trykk-testes i henhold til norm og dokumenteres
Vurdering: Prøvesertifikat for trykktesting finnes i labperm.
3
8.4 Påvirkning av ytre miljø (utslipp til luft/vann, støy, temperatur, rystelser, lukt)
Se kapittel 13 ”Veiledning til rapport mal..
NEI
8.5 Stråling
NEI
Vedlegg:
Vurdering:
8.6 Bruk og behandling av kjemikalier

JA Hydraulikkolje
Vedlegg: Sikkerhetsdatablad
Vurdering: Hydraulikkolje, mineralsk olje. Datablad er vedlagt.
8.7 El sikkerhet (behov for å avvike fra gjeldende forskrifter og normer)

NEI
9 VURDERING AV OPERASJONELL SIKKERHET

Sikrer at etablerte prosedyrer dekker alle identifiserte risikoforhold som må håndteres
gjennom operasjonelle barrierer og at operatører og teknisk utførende har tilstrekkelig
kompetanse.
9.1 Prosedyre HAZOP

Metoden er en undersøkelse av operasjonsprosedyrer, og identifiserer årsaker og farekilder for
operasjonelle problemer.
Vedlegg: HAZOP_MAL_Prosedyre
Vurdering:
9.2 Drifts og nødstopps prosedyre

Driftsprosedyren er en sjekkliste som skal fylles ut for hvert forsøk.
Nødstopp prosedyren skal sette forsøksoppsetningen i en harmløs tilstand ved uforutsette
hendelser.
Vedlegg: Procedure for running experiments
Nødstopp prosedyre:
4
9.3 Opplæring av operatører
Dokument som viser Opplæringsplan for operatører utarbeides for alle forøksoppsetninger.
 Kjøring av pumpesystem
Vedlegg: Opplæringsplan for operatører
9.4 Tekniske modifikasjoner
9.5 Personlig verneutstyr

 Det er påbudt med vernebriller i sonen anlegget er plassert i.
9.6 Generelt
9.7 Sikkerhetsutrustning
9.8 Spesielle tiltak
10 TALLFESTING AV RESTRISIKO – RISIKOMATRISE

Risikomatrisen vil gi en visualisering og en samlet oversikt over aktivitetens risikoforhold slik
at ledelse og brukere får et mest mulig komplett bilde av risikoforhold.
IDnr Aktivitet-hendelse Frekv-Sans Kons RV

1 Lekkasje i Hydraulikk 1 A A1
2 Fremmedlegemer i vannet 1 A A1
3 Rørbrudd 1 A A1
4 Roterende Aksling 1 B B1
Vurdering restrisiko: Det er liten restrisiko ved forsøkene, foruten at trykk-satt vann og olje
fordrer bruk av vernebriller. Fremmedlegemer i vannet gir liten risiko for personskade, men kan
føre til store skader på maskineri.
5
11 LOVER FORSKRIFTER OG PÅLEGG SOM GJELDER
Se http://www.arbeidstilsynet.no/regelverk/index.html
 Lov om tilsyn med elektriske anlegg og elektrisk utstyr (1929)
 Arbeidsmiljøloven
 Forskrift om systematisk helse-, miljø- og sikkerhetsarbeid (HMS Internkontrollforskrift)
 Forskrift om sikkerhet ved arbeid og drift av elektriske anlegg (FSE 2006)
 Forskrift om elektriske forsyningsanlegg (FEF 2006)
 Forskrift om utstyr og sikkerhetssystem til bruk i eksplosjonsfarlig område NEK 420
 Forskrift om håndtering av brannfarlig, reaksjonsfarlig og trykksatt stoff samt utstyr og
anlegg som benyttes ved håndteringen
 Forskrift om Håndtering av eksplosjonsfarlig stoff
 Forskrift om bruk av arbeidsutstyr.
 Forskrift om Arbeidsplasser og arbeidslokaler
 Forskrift om Bruk av personlig verneutstyr på arbeidsplassen
 Forskrift om Helse og sikkerhet i eksplosjonsfarlige atmosfærer
 Forskrift om Høytrykksspyling
 Forskrift om Maskiner
 Forskrift om Sikkerhetsskilting og signalgivning på arbeidsplassen
 Forskrift om Stillaser, stiger og arbeid på tak m.m.
 Forskrift om Sveising, termisk skjæring, termisk sprøyting, kullbuemeisling, lodding og
sliping (varmt arbeid)
 Forskrift om Tekniske innretninger
 Forskrift om Tungt og ensformig arbeid
 Forskrift om Vern mot eksponering for kjemikalier på arbeidsplassen
(Kjemikalieforskriften)
 Forskrift om Vern mot kunstig optisk stråling på arbeidsplassen
 Forskrift om Vern mot mekaniske vibrasjoner
 Forskrift om Vern mot støy på arbeidsplassen
Veiledninger fra arbeidstilsynet

se: http://www.arbeidstilsynet.no/regelverk/veiledninger.html
12 DOKUMENTASJON
 Tegninger, foto, beskrivelser av forsøksoppsetningen
 Hazop_mal
 Sertifikat for trykkpåkjent utstyr
 Håndtering avfall i NTNU
 Sikker bruk av LASERE, retningslinje
 HAZOP_MAL_Prosedyre
 Forsøksprosedyre
 Opplæringsplan for operatører
 Skjema for sikker jobb analyse, (SJA)
 Apparaturkortet
 Forsøk pågår kort
6
13 VEILEDNING TIL RAPPORTMAL
Kapittel 7 Vurdering av teknisk sikkerhet
Sikre at design av apparatur er optimalisert i forhold til teknisk sikkerhet.
Identifisere risikoforhold knyttet til valgt design, og eventuelt å initiere re-design for å sikre
at størst mulig andel av risiko elimineres gjennom teknisk sikkerhet.
Punktene skal beskrive hva forsøksoppsetningen faktisk er i stand til å tåle og aksept for
utslipp.
7.1 Fareidentifikasjon, HAZOP

Forsøksoppsetningen deles inn i noder: (eks Motorenhet, pumpeenhet, kjøleenhet.)
Ved hjelp av ledeord identifiseres årsak, konsekvens og sikkerhetstiltak. Konkluderes det
med at tiltak er nødvendig anbefales disse på bakgrunn av dette. Tiltakene lukkes når de er
utført og Hazop sluttføres.
(eks ”No flow”, årsak: rør er deformert, konsekvens: pumpe går varm,
sikkerhetsforanstaltning: måling av flow med kobling opp mot nødstopp eller hvis
konsekvensen ikke er kritisk benyttes manuell overvåkning og punktet legges inn i den
operasjonelle prosedyren.)
7.2 Brannfarlig, reaksjonsfarlig og trykksatt stoff.

I henhold til Forskrift om håndtering av brannfarlig, reaksjonsfarlig og trykksatt stoff samt
utstyr og anlegg som benyttes ved håndteringen
Brannfarlig stoff: Fast, flytende eller gassformig stoff, stoffblanding, samt stoff som
forekommer i kombinasjoner av slike tilstander, som i kraft av sitt flammepunkt, kontakt
med andre stoffer, trykk, temperatur eller andre kjemiske egenskaper representerer en fare
for brann.
Reaksjonsfarlig stoff: Fast, flytende, eller gassformig stoff, stoffblanding, samt stoff som
forekommer i kombinasjoner av slike tilstander, som ved kontakt med vann, ved sitt trykk,
temperatur eller andre kjemiske forhold, representerer en fare for farlig reaksjon, eksplosjon
eller utslipp av farlig gass, damp, støv eller tåke.
Trykksatt stoff: Annet fast, flytende eller gassformig stoff eller stoffblanding enn brann- eller
reaksjonsfarlig stoff, som er under trykk, og som derved kan representere en fare ved
ukontrollert utslipp.
Nærmere kriterier for klassifisering av brannfarlig, reaksjonsfarlig og trykksatt stoff er

fastsatt i vedlegg 1 i veiledningen til forskriften ”Brannfarlig, reaksjonsfarlig og trykksatt
stoff”
http://www.dsb.no/Global/Publikasjoner/2009/Veiledning/Generell%20veiledning.pdf
http://www.dsb.no/Global/Publikasjoner/2010/Tema/Temaveiledning_bruk_av_farlig_stoff_Del_1.p
df
Rigg og areal skal gjennomgås med hensyn på vurdering av Ex sone
 Sone 0: Alltid eksplosiv atmosfære, for eksempel inne i tanker med gass,
brennbar væske.
 Sone 1: Primær sone, tidvis eksplosiv atmosfære for eksempel et fylle tappe
punkt
7
 Sone 2: Sekundert utslippssted, kan få eksplosiv atmosfære ved uhell, for
eksempel ved flenser, ventiler og koblingspunkt
7.4 Påvirkning av ytre miljø

Med forurensning forstås: tilførsel av fast stoff, væske eller gass til luft, vann eller i grunnen
støy og rystelser påvirkning av temperaturen som er eller kan være til skade eller ulempe for
miljøet.
Regelverk: http://www.lovdata.no/all/hl-19810313-006.html#6
NTNU retningslinjer for avfall se: http://www.ntnu.no/hms/retningslinjer/HMSR18B.pdf
7.5 Stråling
Stråling defineres som
Ioniserende stråling: Elektromagnetisk stråling (i strålevernsammenheng med bølgelengde
<100 nm) eller hurtige atomære partikler (f.eks alfa- og beta-partikler) som har evne til å
ionisere atomer eller molekyler
Ikke-ioniserende stråling: Elektromagnetisk stråling (bølgelengde >100 nm), og ultralyd1,
som har liten eller ingen evne til å ionisere.
Strålekilder: Alle ioniserende og sterke ikke-ioniserende strålekilder.
Ioniserende strålekilder: Kilder som avgir ioniserende stråling, f.eks alle typer radioaktive
kilder, røntgenapparater, elektronmikroskop
Sterke ikke-ioniserende strålekilder: Kilder som avgir sterk ikke-ioniserende stråling som
kan skade helse og/eller ytre miljø, f.eks laser klasse 3B og 4, MR2-systemer, UVC3-kilder,
kraftige IR-kilder4
1 Ultralyd er akustisk stråling (”lyd”) over det hørbare frekvensområdet (>20 kHz). I strålevernforskriften er
ultralyd omtalt sammen med elektromagnetisk ikke-ioniserende stråling.
2 MR (eg. NMR) - kjernemagnetisk resonans, metode som nyttes til å «avbilde» indre strukturer i ulike
materialer.
3 UVC er elektromagnetisk stråling i bølgelengdeområdet 100-280 nm.
4 IR er elektromagnetisk stråling i bølgelengdeområdet 700 nm – 1 mm.
For hver laser skal det finnes en informasjonsperm(HMSRV3404B) som skal inneholde:
 Generell informasjon
 Navn på instrumentansvarlig og stedfortreder, og lokal strålevernskoordinator
 Sentrale data om apparaturen
 Instrumentspesifikk dokumentasjon
 Referanser til (evt kopier av) datablader, strålevernbestemmelser, o.l.
 Vurderinger av risikomomenter
 Instruks for brukere
 Instruks for praktisk bruk; oppstart, drift, avstenging, sikkerhetsforholdsregler,
loggføring, avlåsing, evt. bruk av strålingsmåler, osv.
 Nødprosedyrer
Se ellers retningslinjen til NTNU for laser: http://www.ntnu.no/hms/retningslinjer/HMSR34B.pdf
7.6 Bruk og behandling av kjemikalier.

Her forstås kjemikalier som grunnstoff som kan utgjøre en fare for arbeidstakers sikkerhet
og helse.
Se ellers: http://www.lovdata.no/cgi-wift/ldles?doc=/sf/sf/sf-20010430-0443.html
Sikkerhetsdatablar skal være i forøkenes HMS perm og kjemikaliene registrert i
Stoffkartoteket.
8
Kapittel 8 Vurdering av operasjonell sikkerhet
Sikrer at etablerte prosedyrer dekker alle identifiserte risikoforhold som må håndteres
gjennom operasjonelle barrierer og at operatører og teknisk utførende har tilstrekkelig
kompetanse.
8.1 Prosedyre Hazop

Prosedyre-HAZOP gjennomføres som en systematisk gjennomgang av den aktuelle
prosedyren ved hjelp av fastlagt HAZOP-metodikk og definerte ledeord. Prosedyren brytes
ned i enkeltstående arbeidsoperasjoner (noder) og analyseres ved hjelp av ledeordene for å
avdekke mulige avvik, uklarheter eller kilder til mangelfull gjennomføring og feil.
8.2 Drifts og nødstopp prosedyrer

Utarbeides for alle forsøksoppsetninger.
Driftsprosedyren skal stegvis beskrive gjennomføringen av et forsøk, inndelt i oppstart, under
drift og avslutning. Prosedyren skal beskrive forutsetninger og tilstand for start,
driftsparametere med hvor store avvik som tillates før forsøket avbrytes og hvilken tilstand
riggen skal forlates.
Nødstopp-prosedyre beskriver hvordan en nødstopp skal skje, (utført av uinnvidde),
hva som skjer, (strøm/gass tilførsel) og
hvilke hendelser som skal aktivere nødstopp, (brannalarm, lekkasje).
Kapittel 9 Risikomatrise Tallfesting av restrisiko

For å synliggjøre samlet risiko, jevnfør skjema for risikovurdering, plottes hver enkelt
aktivitets verdi for sannsynlighet og konsekvens inn i risikomatrisen. Bruk aktivitetens IDnr.
Eksempel: Hvis aktivitet med IDnr. 1 har fått en risikoverdi D3 (sannsynlighet 3 x konsekvens
D) settes aktivitetens IDnr i risikomatrisens felt for 3D. Slik settes alle aktivitetenes
risikoverdier (IDnr) inn i risikomatrisen.
I risikomatrisen er ulike grader av risiko merket med rød, gul eller grønn. Når en aktivitets
risiko havner på rød (= uakseptabel risiko), skal risikoreduserende tiltak gjennomføres. Ny
vurdering gjennomføres etter at tiltak er iverksatt for å se om risikoverdien er kommet ned
på akseptabelt nivå.
Svært E1 E2 E3 E4 E5
alvorlig
Alvorlig D1 D2 D3 D4 D5
KONSEKVENS
Moderat C1 C2 C3 C4 C5
Liten B1 B2 B3 B4 B5
Svært A1 A2 A3 A4 A5
liten
Svært liten Liten Middels Stor Svært Stor
SANSYNLIGHET
9
Prinsipp over akseptkriterium. Forklaring av fargene som er brukt i risikomatrisen.
Farge Beskrivelse
Rød Uakseptabel risiko. Tiltak skal gjennomføres for å redusere risikoen.
Gul Vurderingsområde. Tiltak skal vurderes.
Grønn Akseptabel risiko. Tiltak kan vurderes ut fra andre hensyn.
10
Appendix B
Sensor locations and BEP from

Francis-99
To compare pressure and strain results from the laboratory measurements and the
simulations, accurate locations of the sensors are needed. The laboratory coordinates
of each sensor are listed in the tables below:
Figure B.1: Coordinates of pressure sensors
XIV
Figure B.2: Coordinates of strain gage sensor
The following data is collected from the Francis-99 workshop 2 [2]. Figure B.3
shows acquired ow parameters during BEP operating point, with corresponding
uncertainties. The uncertainty in hydraulic eciency is the total uncertainty. This
was used as a base when deciding on which operating point to run the experiment
at.
Figure B.3: Best eciency point from the second Francis-99 workshop [2]
XV
Appendix C
Error analysis
According to IEC60193 [42], the error in a measurement is dened as the dierence

between the measured value and the true, physical value of the measurement. All
measurements of a physical size have an uncertainty that come from systematic
and random errors. This uncertainty is dened as the area where the true value is
expected to lie within. IEC60193 have set this probability to be the 95% condence
interval. There are three types of errors to be considered:
• Spurious errors
• Random errors
• Systematic errors
Spurious errors can be human errors or instrument malfunction, which invalidate
a measurement. If the error is too small to make the results obviously invalid, the
point should be repeated or some rejection criteria may be applied.
Random errors are caused by many, small, independent inuences that prevent a
set of measured values for a system with unchanged input value from delivering the
same reading for the measured size. As a consequence, the measurements deviate
from their mean value in accordance with the probability laws, such that the distri-
bution normally approaches Gaussian distribution as the number of measurements
increases. Random errors are dependent on the care taken during the measurements,
the number of measurements and the operating conditions. The area of uncertainty
for the random errors is estimated on the basis of statistical methods. When the
number of measurements is low, the statistical results that are based on the as-
sumption of a normal distribution, must be corrected by the means of the Student's
t value. This is a factor which compensates for the increasing uncertainty in the
standard deviation for a given condence interval, when the sample size is reduced.
Systematic errors are errors which cannot be reduced by increasing the amount
of measurements if the instruments used and the conditions for the measurements
XVI
stay unchanged. They are dependent on remaining errors of the instruments or
the measurement system at the beginning of the tests. Estimation of uncertainty
associated with systematic errors can not be experimentally evaluated without a
change of equipment or by changing the conditions for the measurements. The
alternative is to make subjective judgements based on experience and consideration
of the equipment being used. To estimate the uncertainty, primarily the sources
which can have aected the measured values needs to be identied. The main sources
of uncertainty for an instrument are hysteresis, linearity, accuracy, zero oset, and
drift. Next step is to assign limits for the uncertainty that can be accounted for
in each single inuence. Even so, the possible values of systematic components
essentially have a Gaussian distribution, such that the calculation of the systematic
uncertainty from the individual systematic uncertainties is done by the root-sum-
square method.
Total uncertainty in a measurement is obtained by combining systematic and
random uncertainties. Together these denes the range where the true value of the
measurement lays with a 95% probability. Assuming that both the systematic and
the random errors have the same probability distribution, they can be combined
through the root-sum-square method.
C.1 Determination of uncertainties in model tests
The dierent sources of errors that contribute to the uncertainty during calibration
of an instrument are listed in Table C.1, and are dened by IEC 60193. X indicates
the property measured by the instrument.
Error Description
±fXa Systematic error of the primary calibration method
±fXb Random error of the primary calibration method
±fXc Systematic error (repeatability) of the secondary instrument
±fXd Random error of the secondary instrument
±fXe Physical phenomena and external inuences
±fXf Error in physical properties
Table C.1: Component errors in the calibration of an instrument
The uncertainties of the test itself may be separated into the errors given by Table
C.2.
XVII
Error Description
±fXcal Systematic error in the calibration
±fXh Additional systematic error in the instrument
±fXj Error in physical properties
±fXks Systematic errors due to physical phenomena and external inuences
±fXkr Random errors due to physical phenomena and external inuences
±fXl Random error in the repeatability of the secondary instrument
Table C.2: Component errors in the test
XVIII
Appendix D
Calibration reports
XIX
file:///M:/5.klasse/Masteroppgave/Calibration/cDAQ4Mod4-ai1_1.html
CALIBRATION REPORT
CALIBRATION PROPERTIES
Calibrated by: Katarina Kloster and Ingebjørg Valkvæ
Type/Producer: Kulite HKM-375M
SN: 8240-4-888
Range: 0-1.7 bar a
Unit: kPa
CALIBRATION SOURCE PROPERTIES

Type/Producer: Pressurements deadweight tester P3023-6-P
SN: 66611
Uncertainty [%]: 0,008
POLY FIT EQUATION:

Y= + 230.02390217E-3X^0 + 22.55034497E+3X^1
CALIBRATION SUMARY:
Max Uncertainty : 0.194185 [%]
Max Uncertainty : 0.221837 [kPa]
RSQ : 0.999991
Calibration points : 15
Figure 1 : Calibration chart (The uncertainty band is multiplied by 10 )
_______________________________________
Katarina Kloster and Ingebjørg Valkvæ
1 av 2 20.04.2016 17:40
file:///M:/5.klasse/Masteroppgave/Calibration/cDAQ4Mod4-ai1_1.html
CALIBRATION VALUES
Best Poly Fit Deviation Uncertainty Uncertainty

Value [kPa] Voltage [V]
[kPa] [kPa] [%] [kPa]
100.000000 0.004422 99.947240 0.052760 0.194185 0.194185
110.015107 0.004868 110.002871 0.012236 0.164734 0.181233
120.030214 0.005313 120.051003 -0.020789 0.140486 0.168626
130.045321 0.005753 129.962240 0.083081 0.121674 0.158232
140.060428 0.006196 139.948236 0.112193 0.107029 0.149906
160.090643 0.007084 159.966808 0.123835 0.087311 0.139776
180.120857 0.007987 180.332795 -0.211938 0.069897 0.125899
200.151071 0.008875 200.374684 -0.223613 0.060147 0.120385
220.181285 0.009764 220.401110 -0.219825 0.051930 0.114340
240.211499 0.010638 240.128541 0.082958 0.052992 0.127294
260.241714 0.011522 260.050969 0.190744 0.071828 0.186926
270.256821 0.011970 270.157619 0.099202 0.059701 0.161345
280.271928 0.012439 280.722866 -0.450939 0.079151 0.221837
290.287035 0.012864 290.320206 -0.033171 0.070310 0.204100
300.302142 0.013289 299.898876 0.403266 0.064545 0.193829
COMMENTS:
The uncertainty is calculated with 95% confidence. The uncertainty includes the randomness in the calibrated instrument during the calibration, systematic
uncertainty in the instrument or property which the instrument under calibration is compared with (dead weight manometer, calibrated weights etc.), and due to
regression analysis to fit the calibration points to a linear calibration equation.The calculated uncertainty can be used as the total systematic uncertianty of the
calibrated instrument with the given calibration equation.
2 av 2 20.04.2016 17:40
file:///M:/5.klasse/Masteroppgave/Calibration/cDAQ4Mod4-ai2_8240...
CALIBRATION REPORT
Calibrated by: Katarina Kloster and Ingebjørg Valkvæ
Type/Producer: Kulite HKM-375M
SN: 8240-4-887
Range: 0-1.7 bar a
Unit: kPa

Type/Producer: Pressurements deadweight tester P3023-6-P
SN: 66611
POLY FIT EQUATION:

Y= -4.25163513E+0X^0 + 22.85839685E+3X^1
CALIBRATION SUMARY:
RSQ : 1.000000
_______________________________________
Katarina Kloster and Ingebjørg Valkvæ
1 av 2 20.04.2016 17:40
file:///M:/5.klasse/Masteroppgave/Calibration/cDAQ4Mod4-ai2_8240...
CALIBRATION VALUES
Best Poly Fit Deviation Uncertainty Uncertainty

Value [kPa] Voltage [V]
[kPa] [kPa] [%] [kPa]
49.924465 0.002370 49.923452 0.001012 0.038917 0.019429
54.932018 0.002589 54.927796 0.004222 0.033517 0.018412
59.939572 0.002810 59.978672 -0.039100 0.048805 0.029253
64.947125 0.003027 64.940910 0.006215 0.025936 0.016845
69.954679 0.003245 69.933418 0.021261 0.021804 0.015253
79.969786 0.003683 79.944248 0.025538 0.022136 0.017702
89.984893 0.004122 89.979020 0.005873 0.038883 0.034989
100.000000 0.004560 99.989738 0.010262 0.011933 0.011933
120.030214 0.005437 120.039231 -0.009017 0.009671 0.011608
140.060428 0.006314 140.082665 -0.022237 0.011533 0.016153
160.090643 0.007191 160.120424 -0.029781 0.016164 0.025877
170.105750 0.007628 170.123544 -0.017795 0.027658 0.047047
180.120857 0.008066 180.119551 0.001306 0.024571 0.044258
190.135964 0.008503 190.124050 0.011914 0.012618 0.023991
200.151071 0.008941 200.120744 0.030327 0.012288 0.024594
COMMENTS:
2 av 2 20.04.2016 17:40
9.5.2016 Onboard14ai0_83171201.html
CALIBRATION REPORT
Calibrated by: Katarina Kloster
Type/Producer: Kulite XTM190SM
SN: 83171201
Range: 03,5bar a
Unit: kPa
Type/Producer: Pressurements deadweight tester P30236P
SN: 66611
POLY FIT EQUATION:
Y= 6,79035476E+0X^0 + 37,37745937E+0X^1
CALIBRATION SUMARY:
Max Uncertainty : 0,412824 [%]
Max Uncertainty : 0,255793 [kPa]
RSQ : 0,999978
_______________________________________
Katarina Kloster
file:///C:/Users/katarik/Dropbox/Masteroppgave/Kalibrering/Onboard%20trykksensorer/Onboard14ai0_83171201.html 1/2
9.5.2016 Onboard14ai0_83171201.html
CALIBRATION VALUES
Best Poly Uncertainty
Value [kPa] Voltage [V] Deviation [kPa] Uncertainty [%]
Fit [kPa] [kPa]
52,024465 1,573476 52,022183 0,002281 0,412824 0,214769
57,032018 1,707366 57,026649 0,005369 0,359201 0,204860
62,039572 1,841416 62,037098 0,002473 0,314763 0,195277
67,047125 1,975587 67,052056 0,004930 0,277559 0,186095
72,054679 2,109498 72,057322 0,002643 0,246191 0,177392
82,069786 2,377531 82,075730 0,005944 0,196966 0,161649
92,084893 2,645303 92,084357 0,000536 0,161537 0,148751
102,100000 2,913199 102,097618 0,002382 0,136593 0,139461
122,130214 3,448776 122,116143 0,014071 0,110093 0,134457
142,160428 3,983703 142,110359 0,050069 0,104357 0,148355
162,190643 4,517960 162,079517 0,111125 0,109034 0,176843
172,205750 4,784111 172,027565 0,178184 0,113004 0,194600
182,220857 5,079663 183,074541 0,853684 0,118517 0,215962
192,235964 5,318855 192,014942 0,221022 0,121911 0,234357
202,251071 5,585231 201,971383 0,279688 0,126473 0,255793
COMMENTS:
The uncertainty is calculated with 95% confidence. The uncertainty includes the randomness in the calibrated instrument during the calibration,
systematic uncertainty in the instrument or property which the instrument under calibration is compared with (dead weight manometer, calibrated
weights etc.), and due to regression analysis to fit the calibration points to a linear calibration equation.The calculated uncertainty can be used as
the total systematic uncertianty of the calibrated instrument with the given calibration equation.
9.5.2016 Onboard14ai1_83171202.html
CALIBRATION REPORT
SN: 83171202
Range: 03,5 bar a
Unit: kPa
Type/Producer: Pressurements deadweight tester P32231
SN: 66256
POLY FIT EQUATION:
Y= + 5,63531073E+0X^0 + 36,92939840E+0X^1
CALIBRATION SUMARY:
RSQ : 0,999979
_______________________________________
Katarina Kloster
9.5.2016 Onboard14ai1_83171202.html
CALIBRATION VALUES
Fit [kPa] [kPa]
52,024465 1,256370 52,032286 0,007822 0,397652 0,206876
57,032018 1,391836 57,034973 0,002955 0,346012 0,197338
62,039572 1,527381 62,040570 0,000998 0,303218 0,188115
67,047125 1,662928 67,046244 0,000881 0,267408 0,179289
72,054679 1,798475 72,051911 0,002768 0,237190 0,170906
82,069786 2,069613 82,064873 0,004913 0,189777 0,155750
92,084893 2,340708 92,076251 0,008641 0,155644 0,143324
102,100000 2,611900 102,091213 0,008787 0,131613 0,134377
122,130214 3,154053 122,112587 0,017627 0,106088 0,129565
142,160428 3,695751 142,117161 0,043267 0,100571 0,142972
162,190643 4,236547 162,088431 0,102212 0,105101 0,170464
172,205750 4,506808 172,069010 0,136739 0,108910 0,187550
182,220857 4,803963 183,042758 0,821901 0,114162 0,208028
192,235964 5,046814 192,011100 0,224864 0,117468 0,225815
202,251071 5,316436 201,968094 0,282976 0,121860 0,246463
COMMENTS:
9.5.2016 Onboard14ai2_83171203.html
CALIBRATION REPORT
SN: 83171203
Range: 03,5 bar a
Unit: kPa
SN: 66256
POLY FIT EQUATION:
Y= 15,59293487E+0X^0 + 37,22867417E+0X^1
CALIBRATION SUMARY:
RSQ : 0,999979
_______________________________________
Katarina Kloster
9.5.2016 Onboard14ai2_83171203.html
CALIBRATION VALUES
Fit [kPa] [kPa]
52,024465 1,814818 51,970319 0,054146 0,398919 0,207535
57,032018 1,950244 57,012067 0,019951 0,346983 0,197892
62,039572 2,085131 62,033734 0,005838 0,304028 0,188618
67,047125 2,219728 67,044592 0,002533 0,268106 0,179757
72,054679 2,354344 72,056172 0,001493 0,237794 0,171342
82,069786 2,623907 82,091649 0,021863 0,190226 0,156118
92,084893 2,893109 92,113675 0,028782 0,156004 0,143656
102,100000 3,161908 102,120721 0,020721 0,131935 0,134706
122,130214 3,699622 122,139075 0,008861 0,106367 0,129906
142,160428 4,236823 142,138375 0,022054 0,100845 0,143362
162,190643 4,773052 162,101471 0,089172 0,105357 0,170880
172,205750 5,040319 172,051455 0,154295 0,109149 0,187961
182,220857 5,335225 183,030432 0,809576 0,114433 0,208520
192,235964 5,575989 191,993727 0,242237 0,117751 0,226360
202,251071 5,843424 201,950000 0,301071 0,122156 0,247063
COMMENTS:
9.5.2016 Onboard14ai3_83171204.html
CALIBRATION REPORT
SN: 83171204
Range: 03,5bar a
Unit: kPa
SN: 66256
POLY FIT EQUATION:
Y= 8,69549532E+0X^0 + 37,28439544E+0X^1
CALIBRATION SUMARY:
RSQ : 0,999980
_______________________________________
Katarina Kloster
9.5.2016 Onboard14ai3_83171204.html
CALIBRATION VALUES
Fit [kPa] [kPa]
52,024465 1,628072 52,006172 0,018293 0,390050 0,202922
57,032018 1,762727 57,026725 0,005293 0,339339 0,193532
62,039572 1,897090 62,036351 0,003221 0,297364 0,184483
67,047125 2,031375 67,043106 0,004019 0,262243 0,175826
72,054679 2,165738 72,052747 0,001932 0,232601 0,167600
82,069786 2,434635 82,078403 0,008617 0,186090 0,152724
92,084893 2,703210 92,092051 0,007158 0,152624 0,140544
102,100000 2,971733 102,103759 0,003759 0,129071 0,131782
122,130214 3,508752 122,126199 0,004015 0,104054 0,127082
142,160428 4,044996 142,119741 0,040688 0,098644 0,140233
162,190643 4,580491 162,085326 0,105317 0,103065 0,167162
172,205750 4,848974 172,095584 0,110166 0,106921 0,184124
182,220857 5,142000 183,020852 0,799995 0,111955 0,204006
192,235964 5,382836 192,000301 0,235662 0,115197 0,221450
202,251071 5,649968 201,960148 0,290923 0,119508 0,241706
COMMENTS:
9.5.2016 Onboard56ai0_83171205.html
CALIBRATION REPORT
SN: 83171205
Range: 03,5 bar a
Unit: kPa
SN: 66256
POLY FIT EQUATION:
Y= + 5,87661454E+0X^0 + 37,17518896E+0X^1
CALIBRATION SUMARY:
RSQ : 0,999979
_______________________________________
Katarina Kloster
9.5.2016 Onboard56ai0_83171205.html
CALIBRATION VALUES
Fit [kPa] [kPa]
52,024465 1,241435 52,027208 0,002744 0,397164 0,206623
57,032018 1,376092 57,033111 0,001093 0,345577 0,197090
62,039572 1,510832 62,042062 0,002491 0,302827 0,187873
67,047125 1,645553 67,050344 0,003219 0,267057 0,179054
72,054679 1,780273 72,058608 0,003929 0,236873 0,170678
82,069786 2,049690 82,074230 0,004445 0,189520 0,155539
92,084893 2,318990 92,085523 0,000630 0,155435 0,143132
102,100000 2,588248 102,095209 0,004791 0,131444 0,134204
122,130214 3,126414 122,101637 0,028577 0,105952 0,129399
142,160428 3,664240 142,095420 0,065008 0,100428 0,142769
162,190643 4,201565 162,070604 0,120038 0,104951 0,170221
172,205750 4,470140 172,054923 0,150827 0,108760 0,187292
182,220857 4,765781 183,045421 0,824565 0,114021 0,207770
192,235964 5,007269 192,022798 0,213165 0,117331 0,225553
202,251071 5,275393 201,990363 0,260708 0,121729 0,246198
COMMENTS:
9.5.2016 Vaneless1 ai0_V453734.html
CALIBRATION REPORT
Type/Producer: Kulite
SN: V453734
Range: 010 bar a
Unit: kPa
SN: 66611
POLY FIT EQUATION:
Y= + 98,86246121E+0X^0 70,35007630E+3X^1
CALIBRATION SUMARY:
RSQ : 1,000000
_______________________________________
Katarina Kloster
file:///C:/Users/katarik/Dropbox/Masteroppgave/Kalibrering/Vaneless/Vaneless1%20ai0_V453734.html 1/2
9.5.2016 Vaneless1 ai0_V453734.html
CALIBRATION VALUES
Fit [kPa] [kPa]
51,954465 0,000667 51,945614 0,008851 0,039565 0,020556
56,962018 0,000596 56,961839 0,000179 0,034545 0,019677
61,969572 0,000525 61,963083 0,006489 0,030565 0,018941
66,977125 0,000453 66,976076 0,001049 0,027085 0,018141
71,984679 0,000382 71,985625 0,000946 0,024602 0,017709
76,992232 0,000311 76,994447 0,002215 0,023104 0,017788
81,999786 0,000240 81,999547 0,000239 0,020736 0,017004
92,014893 0,000097 92,016276 0,001383 0,017514 0,016115
102,030000 0,000045 102,031898 0,001898 0,015502 0,015816
122,060214 0,000330 122,071054 0,010840 0,013133 0,016030
142,090428 0,000615 142,099123 0,008695 0,013976 0,019858
162,120643 0,000899 162,128092 0,007450 0,015579 0,025257
172,135750 0,001042 172,139068 0,003318 0,013677 0,023543
182,150857 0,001184 182,149320 0,001536 0,013885 0,025292
192,165964 0,001326 192,159500 0,006464 0,014314 0,027506
202,181071 0,001468 202,169133 0,011937 0,014618 0,029554
COMMENTS:
file:///C:/Users/katarik/Dropbox/Masteroppgave/Kalibrering/Vaneless/Vaneless1%20ai0_V453734.html 2/2
9.5.2016 Vaneless2ai1_V453733.html
CALIBRATION REPORT
Calibrated by: Einar Agnalt and Katarina Kloster
Type/Producer: Kulite
SN: V453733
Range: 010 bar a
Unit: kPa
SN: 66611
POLY FIT EQUATION:
Y= + 53,37801232E+0X^0 77,26353937E+3X^1
CALIBRATION SUMARY:
RSQ : 1,000000
_______________________________________
Einar Agnalt
file:///C:/Users/katarik/Dropbox/Masteroppgave/Kalibrering/Vaneless/Vaneless2ai1_V453733.html 1/2
9.5.2016 Vaneless2ai1_V453733.html
CALIBRATION VALUES
Fit [kPa] [kPa]
51,954465 0,000019 51,919861 0,034604 0,052726 0,027393
61,969572 0,000111 61,942501 0,027070 0,031512 0,019528
56,962018 0,000047 56,971794 0,009776 0,037707 0,021479
66,977125 0,000176 66,967565 0,009560 0,027897 0,018685
71,984679 0,000241 71,996309 0,011631 0,024981 0,017983
76,992232 0,000306 77,016601 0,024369 0,022538 0,017353
81,999786 0,000371 82,016920 0,017134 0,020525 0,016831
92,014893 0,000500 92,024780 0,009887 0,017451 0,016058
102,030000 0,000630 102,024453 0,005547 0,015408 0,015721
122,060214 0,000889 122,077012 0,016797 0,013541 0,016528
142,090428 0,001148 142,099463 0,009035 0,014582 0,020720
162,120643 0,001407 162,121678 0,001035 0,013619 0,022079
172,135750 0,001537 172,134553 0,001196 0,013961 0,024033
182,150857 0,001667 182,147625 0,003231 0,014328 0,026098
192,165964 0,001796 192,154310 0,011654 0,014900 0,028633
202,181071 0,001926 202,174270 0,006801 0,015060 0,030448
COMMENTS:
file:///C:/Users/katarik/Dropbox/Masteroppgave/Kalibrering/Vaneless/Vaneless2ai1_V453733.html 2/2
file:///M:/5.klasse/Masteroppgave/Calibration/flow.html
CALIBRATION REPORT
Calibrated by: Rakel Ellingsen
Type/Producer: Krohne Optiflux
SN: xxxxxx
Range: 0 - 450 l/s
Unit: l/s

Type/Producer: Weighing method
SN: xxxxx
Uncertainty [%]: 0
POLY FIT EQUATION:

Y= -147.80000000E-3X^0 + 73.95900000E-3X^1
CALIBRATION SUMARY:
Max Uncertainty : 0.000201 [l/s]
RSQ : 1.000000
_______________________________________
Rakel Ellingsen
1 av 2 04.04.2016 15:03
file:///M:/5.klasse/Masteroppgave/Calibration/flow.html
CALIBRATION VALUES
Best Poly Fit Uncertainty Uncertainty

Value [l/s] Voltage [V] Deviation [l/s]
[l/s] [%] [l/s]
0.245718 5.320763 0.245718 0.000000 0.033218 0.000082
0.276933 5.742822 0.276933 0.000000 0.040414 0.000112
0.294971 5.986709 0.294971 0.000000 0.050470 0.000149
0.332043 6.487957 0.332043 0.000000 0.048292 0.000160
0.345482 6.669667 0.345482 0.000000 0.055743 0.000193
0.374027 7.055631 0.374027 0.000000 0.050539 0.000189
0.388063 7.245409 0.388063 0.000000 0.048331 0.000188
0.392152 7.300698 0.392152 0.000000 0.043560 0.000171
0.405216 7.477328 0.405216 0.000000 0.046113 0.000187
0.431016 7.826172 0.431016 0.000000 0.041826 0.000180
0.420916 7.689607 0.420916 0.000000 0.041396 0.000174
0.410390 7.547291 0.410390 0.000000 0.046184 0.000190
0.308215 6.165786 0.308215 0.000000 0.049692 0.000153
0.270213 5.651954 0.270213 0.000000 0.040581 0.000110
0.292572 5.954272 0.292572 0.000000 0.047872 0.000140
0.327318 6.424069 0.327318 0.000000 0.047770 0.000156
0.340066 6.596432 0.340066 0.000000 0.051606 0.000175
0.354110 6.786332 0.354110 0.000000 0.056734 0.000201
COMMENTS:
2 av 2 04.04.2016 15:03
20.4.2016 kalibrering_jan_16_inl.html
CALIBRATION REPORT
Calibrated by: Katarina Kloster og Einar Agnalt
Type/Producer: FHCW36W1AKCAY
SN: 5332548
Range: 05 bar
Unit: kPa
SN: 66256
POLY FIT EQUATION:
Y= 124.73172075E+0X^0 + 62.78902544E+0X^1
CALIBRATION SUMARY:
Max Uncertainty : Inf [%]
RSQ : 0.999998
_______________________________________
Katarina Kloster og Einar Agnalt
file:///C:/Users/katarik/Dropbox/Masteroppgave/Kalibrering/kalibrering_jan_16_inl.html 1/3
CALIBRATION VALUES
Fit [kPa] [kPa]
0.000000 1.989080 0.160656 0.160656 Inf NaN
100.151071 3.578962 99.987845 0.163226 0.134289 0.134492
110.166178 3.743256 110.303652 0.137474 0.114797 0.126467
120.181285 3.900404 120.170842 0.010443 0.105924 0.127301
130.196392 4.060813 130.242772 0.046379 0.100161 0.130406
140.211499 4.218979 140.173829 0.037670 0.086589 0.121408
150.226606 4.380136 150.292744 0.066137 0.084123 0.126376
160.241714 4.534776 160.002415 0.239298 0.072724 0.116534
170.256821 4.696255 170.141530 0.115290 0.063848 0.108705
180.271928 4.854628 180.085658 0.186269 0.057295 0.103287
190.287035 5.014961 190.152824 0.134211 0.061818 0.117633
200.302142 5.175049 200.204544 0.097598 0.053114 0.106388
220.332356 5.492106 220.112279 0.220077 0.047365 0.104360
240.362570 5.813185 240.272473 0.090098 0.048022 0.115427
260.392784 6.128190 260.051351 0.341433 0.035409 0.092203
280.422999 6.450688 280.300676 0.122323 0.041027 0.115048
300.453213 6.768101 300.230743 0.222470 0.035348 0.106203
320.483427 7.086405 320.216727 0.266700 0.032946 0.105586
340.513641 7.406337 340.304984 0.208658 0.030893 0.105194
360.543855 7.725477 360.343458 0.200397 0.028619 0.103185
380.574070 8.045315 380.425738 0.148331 0.028039 0.106709
400.604284 8.365655 400.539600 0.064684 0.027864 0.111624
410.619391 8.525062 410.548615 0.070776 0.028611 0.117484
420.634498 8.683610 420.503659 0.130839 0.025597 0.107671
430.649605 8.845366 430.660171 0.010566 0.025353 0.109181
440.664712 9.004142 440.629605 0.035108 0.026875 0.118430
450.679819 9.163264 450.620724 0.059095 0.025478 0.114826
460.694926 9.323779 460.699302 0.004376 0.025810 0.118907
470.710033 9.483799 470.746745 0.036711 0.026220 0.123420
480.725141 9.645973 480.929496 0.204355 0.026831 0.128984
490.740248 9.807166 491.050672 0.310424 0.026084 0.128006
500.755355 9.966263 501.040222 0.284867 0.026580 0.133101
450.679819 9.166879 450.847667 0.167847 0.026375 0.118865
400.604284 8.368133 400.695183 0.090899 0.030958 0.124020
350.528748 7.569534 350.551950 0.023202 0.034976 0.122600
300.453213 6.773381 300.562240 0.109027 0.042605 0.128008
250.377677 5.976573 250.531496 0.153819 0.044278 0.110861
200.302142 5.181682 200.621028 0.318886 0.061709 0.123605
180.271928 4.860407 180.448500 0.176572 0.076228 0.137417
160.241714 4.545116 160.651694 0.409981 0.094615 0.151612
140.211499 4.221640 140.340960 0.129461 0.104742 0.146861
120.181285 3.902883 120.326522 0.145237 0.124271 0.149350
100.151071 3.584595 100.341535 0.190464 0.133964 0.134166
0.000000 1.986324 0.012348 0.012348 Inf NaN
COMMENTS:
20.4.2016 kalibrering_jan_16_diff.html
CALIBRATION REPORT
Calibrated by: Katarina Kloster og Einar Agnalt
Type/Producer: FHCW36W1AKCAY
SN: 5332544
Range: 05 bar
Unit: kPa
SN: 66256
POLY FIT EQUATION:
Y= 124.21648772E+0X^0 + 62.75555460E+0X^1
CALIBRATION SUMARY:
Max Uncertainty : Inf [%]
RSQ : 0.999999
_______________________________________
Katarina Kloster og Einar Agnalt
file:///C:/Users/katarik/Dropbox/Masteroppgave/Kalibrering/kalibrering_jan_16_diff.html 1/3
CALIBRATION VALUES
Fit [kPa] [kPa]
0.000000 1.983013 0.228610 0.228610 Inf NaN
100.151071 3.577036 100.262421 0.111350 0.128657 0.128851
110.166178 3.731596 109.961878 0.204300 0.119484 0.131631
120.181285 3.892490 120.058866 0.122419 0.107005 0.128601
130.196392 4.053484 130.162127 0.034265 0.108468 0.141222
140.211499 4.217842 140.476531 0.265032 0.099120 0.138977
150.226606 4.372979 150.212207 0.014400 0.078841 0.118440
160.241714 4.532001 160.191745 0.049968 0.075042 0.120249
170.256821 4.697466 170.575583 0.318762 0.077902 0.132633
180.271928 4.853362 180.358939 0.087011 0.067465 0.121620
190.287035 5.013529 190.410310 0.123275 0.056855 0.108187
200.302142 5.170485 200.260147 0.041995 0.063287 0.126765
220.332356 5.490434 220.338717 0.006360 0.058905 0.129786
240.362570 5.807117 240.212379 0.150191 0.043605 0.104810
260.392784 6.128818 260.400884 0.008099 0.039227 0.102144
300.453213 6.763980 300.260855 0.192358 0.034784 0.104509
340.513641 7.402037 340.302422 0.211219 0.030146 0.102653
360.543855 7.724922 360.565301 0.021445 0.028525 0.102845
380.574070 8.040798 380.388275 0.185795 0.026882 0.102307
400.604284 8.360019 400.421157 0.183127 0.024920 0.099831
410.619391 8.519253 410.413980 0.205411 0.026045 0.106944
420.634498 8.681173 420.575319 0.059179 0.023399 0.098424
430.649605 8.839896 430.536116 0.113489 0.023991 0.103315
440.664712 8.999816 440.571938 0.092774 0.022058 0.097201
450.679819 9.160680 450.667085 0.012734 0.020104 0.090605
460.694926 9.320513 460.697503 0.002577 0.020977 0.096638
470.710033 9.480630 470.745727 0.035694 0.020517 0.096577
480.725141 9.639165 480.694632 0.030509 0.021043 0.101159
490.740248 9.800494 490.818957 0.078709 0.020369 0.099958
500.755355 9.960186 500.840529 0.085174 0.019844 0.099369
490.740248 9.799922 490.783029 0.042781 0.020729 0.101727
470.710033 9.481527 470.801998 0.091964 0.021540 0.101393
450.679819 9.163949 450.872191 0.192371 0.020908 0.094229
430.649605 8.844076 430.798404 0.148799 0.022939 0.098786
400.604284 8.363564 400.643591 0.039307 0.023370 0.093623
350.528748 7.567459 350.683625 0.154877 0.025636 0.089860
300.453213 6.769154 300.585540 0.132327 0.030927 0.092920
250.377677 5.971826 250.548757 0.171080 0.043283 0.108370
200.302142 5.170563 200.265034 0.037108 0.072995 0.146210
180.271928 4.848850 180.075789 0.196139 0.078231 0.141028
160.241714 4.529384 160.027494 0.214220 0.118082 0.189217
140.211499 4.215810 140.349015 0.137515 0.107308 0.150458
120.181285 3.897584 120.378550 0.197265 0.133665 0.160640
100.151071 3.575702 100.178686 0.027615 0.130027 0.130223

0.000000 1.973691 0.356400 0.356400 Inf NaN
COMMENTS:
Appendix E
Procedure for operation with

variable RPM
XLVIII
1 Oscillating RPM or torque
This is a procedure explaining how to produce oscillations in RPM or gen-
erator torque in the Francis test rig at the Waterpower Laboratory. The
general idea is to send an oscillating RPM or torque current signal to the
rectifier cabinet controlling the generator power. Both options are explained
in this document.
Set-up of the system

1. Connect a 9-36 voltage DC power source and banana plug
cables to the NI 9265 module
The module has a current output ranging from 0 to 20 mA and needs
a power source to operate.
2. Connect the module to a cDAQ
3. Find the chosen input channels on the ”Francis Generator

signals” box
The box is located down by the Francis runner, and the signals from
the box goes to the rectifier cabinet in the basement which transfers
DC power to the generator. If the RPM is to be controlled, the ca-
bles from the NI 9265 module shall be connected to the ”disturbance
RPM”-input. And if the torque is to be controlled, the cables must
be connected to the ”disturbance TRQ”-input. The different options
are seen in Figure 1. Be aware that the input signal starts at 4mA.
4. Connect other cables if necessary

It is also possible to measure the voltage and current sent to the rec-
tifier cabinet by connecting a suitable module to the ”Generator volt-
age/current” input.
5. Connect the cDAQ to a computer with an usb cable
6. Open your LabVIEW project and connect it to the right

cDAQ channel
In the control room

1. Start up the test rig and go to planned operational point
See other procedures if needed.
2. Activate the ”speed/torque korreksjon”

It is found under ”Francis-generator” and ”Kommandoer” on one of
the computers. See Figure 3. When this is activated, the RPM/torque
is controlled by the current signal sent from the NI module to the
rectifier cabinet.
1
How to run the test rig
1. Start up the LabVIEW program which controls the oscilla-
tions
The output from the NI module is given by the coding you have pro-
vided in LabVIEW. For example, a sine curve can be used.
2. An change of 1 mA is equivalent with 6.25 RPM

Given that the NI 9265 module only goes up to 20 mA and you have
to start at 4 mA, you can only increase/decrease the signal with total
of 16 mA, which is equal to 100 RPM.
3. Always make sure the changes are okay Try to start with a small
increase or decrease. Follow the RPM/torque movement on the screens
in the control room.
4. Change to the original RPM and stop the oscillations

To avoid a jump in RPM, let the LabVIEW program run the RPM
back to its starting point before you shut it down.
5. Deactivate the ”speed/torque correction”
Figure 1: ”Francis Generator signals” box
2
Figure 2: Set-up of the system
Figure 3: Screen shot from the computer in the control room
3
Appendix F
LabView program for

controlling RPM
The LabVIEW program shown on the next page has been produced by PhD Candi-
date Carl Bergan. It uses the application MatchScript to be able to write a script,
and further uses the script outputs to create a waveform signal. The signal is trans-
ferred to a CompactDAQ and corresponding module when the user choose to deploy
it. This specic script produces a signal starting at 4 mA.
LII
Appendix G
Additional results
Figure G.1: Element metrics showing the quality of the mesh elements
LIV
G.1 Mesh independence test
(a) Stress vs. time
(b) Stress vs. mesh size
Figure G.2: Mesh independence test - Max stress
G.2 von-Mises yield criterion
The von-Mises yield criterion states that yielding will occur whenever the distortion
energy in a unit volume is equal to the distortion energy in the same volume when
uniaxially stressed to the yield strength [15]. A scalar invariant is derived as von-
Mises equivalent stress:
r
1
σe = [(σ1 − σ2 )2 + (σ2 − σ3 )2 + (σ3 − σ1 )2 ] (G.1)
2
LV
When the von-Mises equivalent stress exceeds the uniaxial material yield strength,
general yielding will occur.
G.3 Validation of CFD
(a) LAB
(b) CFD
Figure G.3: Comparison of pressure results from sensor PT12
LVI
(a) LAB
(b) CFD
Figure G.4: Comparison of pressure results from sensor PT14
LVII

Dynamic Loads On Francis Turbine

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Dynamic Loads On Francis Turbine

Uploaded by

Copyright:

Available Formats

Dynamic loads on Francis turbines

Master of Science in Mechanical Engineering

Norwegian University of Science and Technology

Nye installasjoner av ikke-regulerbar elektrisitet i strømnettet har ført til større

4 Facilities and methods 33

D Calibration reports XIX

4.1 Mechanical properties of the Tokke runner [12] . . . . . . . . . . . . 35

5.1 Total calibration uncertainties of pressure transducers . . . . . . . . 52

6.1 Probe results from dierent surfaces/edges on a runner blade . . . . 71

7.1 Operating points . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74

C.1 Component errors in the calibration of an instrument . . . . . . . . . XVII

2.1 Engineering stress-strain-curve . . . . . . . . . . . . . . . . . . . . . 6

3.1 Measurement sensors . . . . . . . . . . . . . . . . . . . . . . . . . . . 26

4.1 A section of the open loop system at NTNU . . . . . . . . . . . . . . 34

5.1 Condence interval of repeated measurement . . . . . . . . . . . . . 53

7.1 Rotational speed functions . . . . . . . . . . . . . . . . . . . . . . . . 74

B.1 Coordinates of pressure sensors . . . . . . . . . . . . . . . . . . . . . XIV

µ Kinematic viscosity Ns/m2

Today's energy system is characterized by large time variations in consumption and

1.2 Previous and ongoing work

2.1 Mechanical properties and failure of metals

F is here the instantaneous force applied perpendicular to the cross section, A0 is

E is the Modulus of Elasticity or Young's modulus, and a large E reects a sti

Figure 2.1: Engineering stress-strain-curve

2.2 Energy generation

2.2.1 Synchronous generators

2.2.2 The Nordic system

Figure 2.2: Coordinating production and consumption [43]

2.2.3 Implementation of renewable energy sources

rotating masses of synchronous aggregates decides the instant frequency response

2.2.4 Frequency control

Primary frequency control

Secondary frequency control

Tertiary generation control

Figure 2.4: How dierent reserves balance the system

2.3 The Francis turbine

2.3.1 Main components

Figure 2.5: Illustration of a Francis turbine [33]

are designed with a high stiness to assure minimum deformations caused by

2.3.3 Turbine performance

2.3.4 Francis runner

2.3.5 Material and design

2.4 Loads on a Francis turbine

2.4.1 Torque oscillations

Figure 2.7: Net torque on the turbine [25]

2.4.2 Pressure pulsations

Figure 2.8: Flow eld at inlet of runner [34]

Draft tube surge

2.4.3 Reduction of turbine lifetime

Experimental measurements of transients combined with a numerical analysis can

3.1 Model testing

3.1.1 Sensor fundamentals

(a) Pressure transducer (b) Strain gage

Figure 3.1: Measurement sensors

3.1.2 Discrete sampling and sampling rate

3.1.3 Filtering and smoothing of data

3.2 Numerical analysis

As the computational power of computers increases, the application of high-end

3.2.1 Computational uid dynamics

3.2.2 Computational structural mechanics

3.2.3 Fluid-structure interaction

ACT Transient FSI

6.1 Probe results from dierent surfaces/edges on a runner blade . . . . 71

5.1 Condence interval of repeated measurement . . . . . . . . . . . . . 53

E is the Modulus of Elasticity or Young's modulus, and a large E reects a sti

Figure 2.4: How dierent reserves balance the system

are designed with a high stiness to assure minimum deformations caused by

Figure 2.8: Flow eld at inlet of runner [34]

3.2.1 Computational uid dynamics

structure. Standard earth gravity dened by ANSYS was added in negative z-

(b) Average measurement with 95% condence interval

Figure 5.1: Condence interval of repeated measurement

Table 5.2: 95% condence interval uncertainties

Table 5.3: 95% condence interval uncertainties

Table 5.4: 95% condence interval uncertainties