Establishment of an open 3D

steam turbine flutter test case

DI QI

Department of Energy Technology

EGI-2016-084 MSC EKV1166
School of Industrial Technology and Management
Royal Institute of Technology
Stockholm, Sweden

1
 Master of Science Thesis EGI-2016-084 MSC
EKV1166

Establishment of an open 3D steam turbine

flutter test case

DI QI
Approved Examiner Supervisor

2016-10-31 Paul Petrie-Repar Paul Petrie-Repar

C ommissioner C ontact person

Abstract
An open three-dimensional (3D) flutter test case for steam turbines is presented. Unlike previous research
on turbine flutter, the geometry is an open resource and is from a test case originally presented by
Durham University. The geometry of the test case includes the stator, rotor and diffuser, which is
representative of the aerodynamic characteristics of modern steam turbine blading. The average inlet flow
conditions are total pressure 27 kPa and total temperature 340 K which are typical for the last stage. The
average static pressure at the exit of the diffuser is 8800 Pa. It also provides the typical flow conditions for
the last stage steam turbine.

The aim of current study is to define a 3D test case for open realistic steam turbine blades flutter analysis.
Commercial numerical tool ANSYS CFX was used to solve Reynolds-averaged Navier-Stokes (RANS)
equations for viscous flow and Laminar equations for inviscid flow for steady and unsteady state. The
defined mode shape for the test case was the first flap bending mode fixed at the hub. Multi-row steady
state simulations with mixing plane were performed for different length of rotor exit. Reflecting waves
was found to influence both steady and unsteady simulation at the mixing plane and rotor outlet. Only the
rotor was considered for the flutter analysis. The plots of normalised aerodynamic damping and local
work coefficient for different Inter Blade Phase Angle (IBPA) were calculated. It showed that -90 degrees
IBPA was least stable. Unsteady aerodynamic work was done mainly on the tip region of rotor blade.
Initial results of tip clearance flow were only studied in the steady state.

This thesis work is partly included in paper written for European Turbomachinery Conference (ETC). It
is verified by comparing the results obtained from CFX and LUFT (Linearized Unsteady Flow solver for
Turbomachinery) solvers.

Key words: Flutter, steam turbine, last stage, CFD

2
Sammanfattning
Ett testfall för analys av svängningar i ångturbinsblad utifrån en öppen tredimensionell (3D) modell
presenteras i denna rapport. Till skillnad från tidigare undersökningar av turbinbladssvängningar är
geometrin i detta fall allmänt tillgänglig via ett testfall ursprungligen presenterat av Durham University.
Denna geometri inkluderar stator, rotor och diffusor, vilket är representativt för de aerodynamiska
egenskaperna hos moderna ångturbiners bladuppsättning. Inloppsförhållandena var ett totaltryck på i
genomsnitt 27 kPa och en total temperatur på i genomsnitt 340 K, vilket är typiskt för det sista
turbinsteget. Det genomsnittliga statiska trycket vid diffusorns utlopp var 8800 Pa, också det typiskt för
förhållandena vid det sista steget i en ångturbin.

Studiens syfte var att definiera ett 3D-testfall för realistisk analys av öppen turbinbladssvängning.
Kommersiell mjukvara (ANSYS CFX) användes för att lösa RANS-ekvationer (Reynolds-averaged Navier
Stokes) för visköst flöde och för ekvationer gällande laminärt flöde vid icke-visköst flöde, för stationära
och icke-stationära fall. Testfallets definierade böjmod var första böjmod med turbinbladets ena ände
fixerad. Stationära simuleringar av multipla turbinrader med mixing-plane-metoden genomfördes för olika
längd på rotorutloppet. Vågreflektion upptäcktes påverka både stationär och icke-stationär simulering vid
mixing plane-läget och rotorutloppet.

För svängningsanalysen av turbinbladen undersöktes endast rotorn. Här togs diagram för normaliserad
aerodynamisk dämpning och lokal arbetskoefficient fram för olika IBPA (inter-blade phase angle). Detta
visade att en IBPA på -90 grader var minst stabil. Icke-stationärt aerodynamiskt arbete utfördes framför
allt på rotorbladens ändregioner. S.k. tip-clearance-flöde studerades endast för stationärt flöde.

Detta examensarbete ingår delvis i en artikel skriven för European Turbomachinery Conference (ETC).
Det verifieras genom jämförelse av de resultat som erhållits från CFX och LUFT (Linearized Unsteady
Flow solver for Turbomachinery).

Nyckelord: vibration, ångturbin, sista turbinsteg, CFD

3
 Acknowledgement

Time flies for two years in KTH. It is the time that I will never forget. I will never forget all my teachers
and friends help me to finish my master thesis work.

To my supervisor Paul Petrie-Repar. Thank you for offering me the interesting topic and I have learned a
lot from your knowledge and ways of thinking. I really appreciate your thoughtful help and detailed
explanation.

To PhD students Tobias Gezork and Mauricio Gutierrez Salas. Thank you guys for helping me start from
a beginner for this field of knowledge. Thanks for your patience and humor, that encourage me to
overcome a lot of my difficult time in thesis work.

To China Scholarship Council (CSC) and KTH. Thanks for offering the scholarship to study in KTH so
that I can have the chance to experience a new life and learn a lot for my field.

To all my friends in master thesis room. Thanks for your kind help and encouragement. I was having a
very good time working with you.

To my parents, relatives and friends. Thanks for your warm care and encouragement. You are always my
backbone to go for my future.

4
Table of Contents
Abstract ...................................................................................................................................................................................... 2
Acknowledgement................................................................................................................................................................... 4
Figures ........................................................................................................................................................................................ 7
Tables........................................................................................................................................................................................ 10
Nomenclature......................................................................................................................................................................... 11
1 Introduction .................................................................................................................................................................. 14
2 Background ................................................................................................................................................................... 15
2.1 Flutter analysis .................................................................................................................................................. 15
2.1.1 Fundamentals of flutter............................................................................................................................. 15
2.1.2 Flutter analysis methods ........................................................................................................................... 16
2.1.3 Determination of flutter stability............................................................................................................ 18
2.1.4 Fourier transformation method.............................................................................................................. 19
2.2 Mixing plane ...................................................................................................................................................... 20
2.3 Open geometry................................................................................................................................................. 20
3 Methodology................................................................................................................................................................. 23
3.1 Introduction of numerical tools................................................................................................................... 23
3.2 Workflow ........................................................................................................................................................... 23
3.3 Flow-path setup................................................................................................................................................ 24
3.4 Mesh generation ............................................................................................................................................... 25
3.5 Modal analysis ................................................................................................................................................... 26
3.6 Solver setup ....................................................................................................................................................... 28
3.6.1 Steady ............................................................................................................................................................. 28
3.6.2 Unsteady........................................................................................................................................................ 30
4 Steady state results and discussion .......................................................................................................................... 31
4.1 Influence of mixing plane .............................................................................................................................. 31
4.1.1 Comparison of interface velocity constraint ....................................................................................... 31
4.1.2 Comparison of stationary and rotating diffusers ............................................................................... 33
4.1.3 Comparison of different length of rotor exit...................................................................................... 35
4.2 Steady state results for no tip clearance ..................................................................................................... 36
4.2.1 Comparison with reference...................................................................................................................... 36
4.2.2 Steady state results and discussion ......................................................................................................... 37
4.3 Steady state results for 0.5% tip clearance ................................................................................................ 39
4.3.1 Comparison with reference...................................................................................................................... 39
4.3.2 Tip clearance steady state results and discussion ............................................................................... 40
4.4 Grid convergence............................................................................................................................................. 41
4.4.1 Grid convergence for no tip.................................................................................................................... 41

5
 4.4.2 Grid convergence for tip .......................................................................................................................... 43
5 Unsteady state results and discussion..................................................................................................................... 45
5.1 Initial steady results preparation .................................................................................................................. 45
5.2 Results and discussion for no tip clearance .............................................................................................. 46
6 Conclusions ................................................................................................................................................................... 49
7 Future work................................................................................................................................................................... 50
8 Bibliography .................................................................................................................................................................. 51
Appendix I: Cases description and summary ................................................................................................................ 53
Steady state cases.............................................................................................................................................................. 53
Unsteady state cases ........................................................................................................................................................ 53
Appendix II: ETC paper ..................................................................................................................................................... 54

6
Figures
Figure 1 Steam Turbine Last stage ................................................................................................................................... 14
Figure 2 Collar’s triangle of forces.................................................................................................................................... 15
Figure 3 Campbell diagram................................................................................................................................................. 15
Figure 4 Nodal diameter and travelling wave mode shape ........................................................................................ 16
Figure 5 First order mode shapes ..................................................................................................................................... 17
Figure 6 Comparison of phase-shifted periodic boundaries and periodic boundary conditions for 90
degree IBPA.................................................................................................................................................................. 19
Figure 7 Stage geometry and Boundary .......................................................................................................................... 21
Figure 8 Blade profiles at the tip ....................................................................................................................................... 21
Figure 9 Blade profiles at mid-span ................................................................................................................................ 21
Figure 10 Blade profiles at hub.......................................................................................................................................... 21
Figure 11 Workflow of steady and unsteady analysis .................................................................................................. 23
Figure 12 Schematic of stage and boundaries ............................................................................................................... 24
Figure 13 Comparison of different end wall setup ...................................................................................................... 24
Figure 14 Rotor and diffuser mesh at mid-span & detailed leading and trailing edge mesh............................ 25
Figure 15 Rotor and diffuser mesh at shroud ............................................................................................................... 25
Figure 16 Rotor and diffuser mesh at hub .................................................................................................................... 25
Figure 17 Stator mesh at mid-span & detailed leading and trailing edge mesh.................................................... 26
Figure 18 Stator mesh at shroud ....................................................................................................................................... 26
Figure 19 Stator mesh at hub ............................................................................................................................................. 26
Figure 20 Comparison of with and without spin softening....................................................................................... 26
Figure 21 Displacement of mode shape ......................................................................................................................... 27
Figure 22 Nondisplaced and displaced FEM mesh ..................................................................................................... 27
Figure 23 Grid convergence for modal analysis ........................................................................................................... 28
Figure 24 Solver model and boundary conditions ....................................................................................................... 28
Figure 25 Inlet total pressure and velocity [4] ............................................................................................................... 29
Figure 26 Inlet total temperature and viscosity [4]....................................................................................................... 29
Figure 27 Total mesh displacement at rotor shroud.................................................................................................... 30
Figure 28 Comparison for stator 90% span blade loading for stator ..................................................................... 31
Figure 29 Comparison for rotor 90% span blade loading for rotor........................................................................ 31
Figure 30 Absolute Mach number from hub to shroud for different mixing plane ........................................... 32
Figure 31 Static Pressure from hub to shroud for different mixing plane ............................................................ 32
Figure 32 Absolute total pressure from hub to shroud for different mixing plane ............................................ 32
Figure 33 Absolute Total Temperature from hub to shroud for different mixing plane .................................. 32
Figure 34 Absolute flow alpha........................................................................................................................................... 32

7
Figure 35 Absolute Mach number at rotor-diffuser mixing plane (RANS) .......................................................... 33
Figure 36 Total Pressure at rotor-diffuser mixing plane (RANS)............................................................................ 33
Figure 37 Total temperature at rotor-diffuser mixing plane (RANS) ..................................................................... 34
Figure 38 Absolute Mach number at rotor-diffuser mixing plane (Inviscid)........................................................ 34
Figure 39 Absolute total pressure at rotor-diffuser mixing plane (inviscid) ......................................................... 34
Figure 40 Absolute total temperature at rotor-diffuser mixing plane (inviscid) .................................................. 34
Figure 41 Blade loading at 50% span............................................................................................................................... 34
Figure 42 Blade loading at 90% span............................................................................................................................... 34
Figure 43 First harmonic amplitude of pressure at the mixing plane ..................................................................... 35
Figure 44 Density Gradient for rotor at 50% span for short case........................................................................... 35
Figure 45 Density Gradient for rotor at 50% span for long case ............................................................................ 35
Figure 46 Rotor blade loading at 50% span................................................................................................................... 36
Figure 47 Rotor blade loading at 90% span................................................................................................................... 36
Figure 48 Density gradient for one chord rotor exit at 50% span........................................................................... 36
Figure 49 Density gradient for two chord rotor exit at 50% span........................................................................... 36
Figure 50 Comparison of total pressure at comparison plane [4]............................................................................ 37
Figure 51 Comparison of Swirl alpha at comparison plane [4] ................................................................................ 37
Figure 52 Mach number and pressure at typical positions......................................................................................... 38
Figure 53 Mach number contour for rotor 5% span................................................................................................... 38
Figure 54 Mach number contour at rotor 50% span................................................................................................... 39
Figure 55 Mach number contour at rotor 90% span................................................................................................... 39
Figure 56 Mach number contour at rotor 95% span................................................................................................... 39
Figure 57 Comparison of 0.5% tip and no tip at rotor outlet................................................................................... 40
Figure 58 Blade loading comparison for 0.5% tip and no tip at 50% span .......................................................... 40
Figure 59 Blade loading comparison for 0.5% tip and no tip at 90% span .......................................................... 40
Figure 60 Blade loading comparison for 0.5% tip and no tip at 95% span .......................................................... 41
Figure 61 Blade loading comparison for 0.5% tip and no tip at 99% span .......................................................... 41
Figure 62 visualization of tip clearance flow.................................................................................................................. 41
Figure 63 Blade loading for stator at 50% span............................................................................................................ 42
Figure 64 Blade loading for stator at 90% span............................................................................................................ 42
Figure 65 Blade loading for stator at 95% span............................................................................................................ 42
Figure 66 Blade loading for rotor at 50% span............................................................................................................. 42
Figure 67 Blade loading for stator at 90% span............................................................................................................ 43
Figure 68 Blade loading for stator at 95% span............................................................................................................ 43
Figure 69 Different tip cells at 50% span for 0.5% tip............................................................................................... 43
Figure 70 Different tip cells at 90% span for 0.5% tip............................................................................................... 43

8
Figure 71 Different tip cells at 95% span for 0.5% tip............................................................................................... 44
Figure 72 Tip meshes for 0.5% tip................................................................................................................................... 44
Figure 73 Blade loading comparison at 30% span for initial steady results .......................................................... 45
Figure 74 Blade loading comparison at 50% span for initial steady result ............................................................ 45
Figure 75 Blade loading comparison at 90% span for initial steady results .......................................................... 45
Figure 76 Normalized aerodynamic damping ............................................................................................................... 46
Figure 77 Schlieren plot of short rotor 50% span at maximum mesh displacement ......................................... 47
Figure 78 Schlieren plot of long rotor 50% span at maximum mesh displacement ........................................... 47
Figure 79 Normalised local work coefficient for -90 degree at 50% span pressure side................................... 47
Figure 80 Normalised local work coefficient for -90 degree at 50% span suction side..................................... 47
Figure 81 Normalised local work coefficient for -90 degree at 90% span pressure side................................... 47
Figure 82 Normalised local work coefficient for -90 degree at 90% span suction side..................................... 47
Figure 83 Local wall work density versus span ............................................................................................................. 48
Figure 84 Schematic of stage and boundaries ............................................................................................................... 55
Figure 85 Mode shape.......................................................................................................................................................... 55
Figure 86 Relative Mach number at 50% span（LUFT）........................................................................................ 58
Figure 87 Relative Mach number at 90% span (LUFT) ............................................................................................. 58
Figure 88 Blade loading at rotor 50% span.................................................................................................................... 58
Figure 89 Blade loading at rotor 90% span.................................................................................................................... 58
Figure 90 Relative Mach number at 50% （CFX inviscid）.................................................................................... 59
Figure 91 Relative Mach number at rotor exit .............................................................................................................. 59
Figure 92 Amplitude of first harmonic pressure at the mixing plane ..................................................................... 60
Figure 93 Schlieren plots for CFX short rotor at 50% span..................................................................................... 60
Figure 94 Schlieren plots for CFX long RANS case at 50% span........................................................................... 60
Figure 95 Schlieren plots for LUFT short RANS case at 50% span....................................................................... 60
Figure 96 Blade loading comparison for single short rotor at 50% span.............................................................. 61
Figure 97 Normalized aerodynamic damping comparison........................................................................................ 61
Figure 98 Local wall work density versus span for IBPA -90 degree ..................................................................... 61
Figure 99 Normalised local work coefficient for IBPA -90 degree at 90% span............................................... 61

9
Tables
Table 1 The stage features .................................................................................................................................................. 21
Table 2 Stator inlet profile .................................................................................................................................................. 22
Table 3 Setting of Fluent from Durham......................................................................................................................... 22
Table 4 Key points Coordinates of the end wall .......................................................................................................... 24
Table 5 Blade material Properties ..................................................................................................................................... 27
Table 6 Fluid properties ...................................................................................................................................................... 29
Table 7 Summary of solver setting ................................................................................................................................... 29
Table 8 summary of steady state cases ............................................................................................................................ 37
Table 9 Steady state cases summary for tip clearance flow........................................................................................ 40
Table 10 Mesh information for grid convergence without tip clearance ............................................................... 42
Table 11 Mesh information for different tip cells for 0.5% tip................................................................................ 43
Table 12 Detail of meshes .................................................................................................................................................. 57
Table 13 Summary of steady state cases and solutions ............................................................................................... 57

10
Nomenclature
Latin Symbols
b blade height

c blade chord length

f modal frequency

F force

G Structural damping matrix

h amplitude

K stiffness matrix

m mass

M mass matrix

[M] diagonal mass matrix

n nodal diameter

n vector normal to surface element

N number of blades

p pressure

𝑝𝑝̅ mean pressure

𝑝𝑝 time dependent perturbation pressure

𝑝𝑝𝑝 complex pressure amplitude

r radius, radial coordinate

Re Reynold number

S stress stiffening matrix

t time

V flow velocity

𝜔𝜔∗ reduced frequency

W work

x,y,z cartesian coordinates

X displacement vector

𝑋𝑋̇ first derivative of displacement vector, velocity

𝑋𝑋̈ second derivative of displacement vector, acceleration

11
Greek Symbols
α hmax/c
π number “pi” 3.1415927…

ρ density

σ Inter-blade phase angle

ωs rotational frequency

Ξ Normalised aerodynamic damping

Δ difference

Subscripts
aero aerodynamic

damping related to damping

external external excitation source

i imaginary unit

max maximum

ref reference

wd wall work density

Abbreviations
3D three dimensional

ASME American Society of Mechanical Engineers

ATM automatic topology and meshing

BSDA back surface deflection angle

CFD computational fluid dynamics

CTP constant total pressure

deg degrees

DOF degree of freedom

DSTC Durham Stage Test Case

EO engine order

FEM finite element method

FEA finite element analysis

12
FT Fourier transformation

HCF high cycle fatigue

HP high pressure

IBPA interblade phase angle

IP intermediate pressure

K Kelvin

Kg kilogram

KTH Kungliga Tekniska Högskolan (Royal Institute of Technology)

LP low pressure

LSBs last stage blades

LUFT linearized unsteady flow solver for turbomachinery

ND nodal diameter

rad radian

RANS Reynolds-averaged Navier–Stokes

RMS root mean square

SAV stage average velocity

TWM travelling wave mode

13
1 Introduction
Flutter is a self-excited vibration due to the interaction between the aerodynamic and structural forces.
Flutter analysis is becoming more important and challenging for steam turbines. After years of efforts to
improve the efficiency of high and intermediate pressure (HP and IP) sections, low pressure(LP) section
of steam turbines have attracted manufactures interest since it can produce power accounting for up to 50%
of total energy extracted from steam. One approach is to lengthen the last stage blades(LSBs) to increase
the power output. However, since the longest blade in the last stage has low structural frequency and
supersonic tip speeds (Petrie-Repar 2015), this approach can make LSBs more susceptible to flutter. And
aggressive design has also been attracted to improve flow performance level. Some studies have been
made to conduct steam turbine flutter analysis. Masserey et,al (Masserey 2012) made an analysis of causes
of ND37 cracks in the root attachments. It is caused by flutter excitation under the circumstance of -55
degrees Inter Blade Phase Angle (IBPA). Rice et,al (Rice 2009) made a flutter analysis of low pressure
steam turbine blade with tip shroud for first family vibration. It is found that tip shroud destabilized the
blade for nodal diameter -15.

Figure 1 Steam Turbine Last stage

The last stage blades design of steam turbine is one of the most important elements because it determines
turbine’s overall performance, dimensions and number of casings. Due to its importance, the low pressure
last stage blading geometry is always kept as a secret and protected by manufactures. Unlike previous
researches on turbine flutter, the geometry used in the report is an open resource and is from a test case
originally presented by Durham University (Burton 2014). The geometry of the test case includes the
stator, rotor and diffuser, which is representative of the aerodynamic characteristics of modern steam
turbine blading. It is the only open geometry for author’s knowledge for steam turbine.
Tip clearance flow is a flow caused by pressure difference between pressure side and suction side of blades.
This flow exiting the tip clearance interacts with the incoming flow to form a tip vortex and creating
performance and stability losses. The tip region is important for LSBs because most unsteady
aerodynamic work done here leads to flutter (Petrie-Repar 2015). As a result, Tip clearance flow is
influential for LSB flutter analysis. Burton et,al (Z. I. Burton 2014) in Durham University also did tip
clearance study on the geometry I used in this thesis report. However, she focused on studying tip
clearance effect on flow in exhaust hood for steady state.
The aim of current study is to define a three dimensional (3D) test case for open realistic steam turbine
blades flutter analysis. Steady and unsteady simulations were performed to study flow field effects and
aerodynamic damping. Due to time limitation, tip clearance flow effects are also investigated on the steady
state. Different from previous study Burton did (Burton 2014), different tip clearance gaps were included
to show the effects on tip clearance flow.

14
2 Background
2.1 Flutter analysis
2.1.1 Fundamentals of flutter
Aeroelastisity is the phenomenon of the interaction between a structure and flow. The system is always
balanced by inertial forces, elastic forces and aerodynamic forces. When the balance is disrupted, unstable
vibration occurs. It has been explained in the collar’s triangle of forces.

Figure 2 Collar’s triangle of forces

There are mainly two types of dynamic aeroelasiticity in turbomachinery, flutter and forced response.
Flutter is a self-excited and self-maintained phenomenon that unsteady fluid flow keeps feeding the energy
to structures and leads to the unstable vibration. This phenomenon can be accompanied by an exponential
increase of oscillation amplitude which leads to material failure. The difference between flutter and forced
response is about excitation source. The excitation source of forced response is from external which is
always from the disturbance of rotor-stator interaction.
The difference between flutter and forced response can be explained well in the Campbell diagram. It
represents rotor speed versus vibration frequency. When rotor speed meets the blade mode line, the point
is where flutter natural frequency occurs. Since forced response excitation source is from external,
oscillation happens when rotor speed, Engine Order (EO) and blade mode line meet.

Figure 3 Campbell diagram

15
These aeroelastic phenomena can be expressed and explained also by the equation:
𝐌𝐌𝑿𝑿̈ + 𝑮𝑮𝑿𝑿̇ + 𝑲𝑲𝑲𝑲 = 𝑭𝑭𝒂𝒂𝒂𝒂𝒂𝒂𝒂𝒂(𝒕𝒕) (1)

where M, G and K denote mass, structural damping, stiffness matrixes respectively. X is the modal
coordinates vector and Faero is aerodynamic forces. The left side of the equation represents structural part
and right side represents aerodynamic part.
Aerodynamic forces contain two terms: 𝑭𝑭𝒅𝒅𝒅𝒅𝒅𝒅𝒅𝒅𝒅𝒅𝒅𝒅𝒅𝒅 is from the interaction between blade itself and flow.
𝑭𝑭𝒆𝒆𝒆𝒆𝒆𝒆𝒆𝒆𝒆𝒆𝒆𝒆𝒆𝒆𝒆𝒆 is from the external excitation source. From the explanation of the flutter above, the aeroelastic
equation can be modified for flutter:
𝐌𝐌𝑿𝑿̈ + 𝑮𝑮𝑿𝑿̇ + 𝑲𝑲𝑲𝑲 = 𝑭𝑭𝒅𝒅𝒅𝒅𝒅𝒅𝒅𝒅𝒅𝒅𝒅𝒅𝒅𝒅 (𝒕𝒕) (2)

The equation above can solve the aerodynamic damping and determine stability of the system.
Flutter analysis is not just about one single blade but a full blade row. This is because the flow around a
blade is affected by blade itself and the neighbor blades. It is called aerodynamic coupling. The coupling
effect is mainly caused by adjacent blades. A traditional approach to determine the eigenmode for a blade
row is called Travelling Wave Mode (TWM). It assumes that each blade is identical (tuned blades), each
blade has only one degree of freedom (DOF), unsteady flow perturbations are linear with blade vibration.
For one travelling wave mode, all blades have same oscillation amplitude and frequency. Constant phase
shift is between two adjacent blades and it is called Inter Blade Phase Angle (IBPA):
2𝜋𝜋𝜋𝜋 (3)
σ=
𝑁𝑁
where n is nodal diameter and N is number of blades.
Nodal diameter is the diameter formed by points which are stationary in the vibration. It is an integer and
defined as:
−N/2 ≤ n ≤ N/2 (4)

When n is positive, it is forward travelling wave. When n is negative, it is backward travelling wave. Figure
4 shows the relationship between nodal diameter and travelling wave mode shape.

Nodal diameter=1 Nodal diameter=3

Figure 4 Nodal diameter and travelling wave mode shape

2.1.2 Flutter analysis methods

The equation above can be solved by coupled treatment, which means the effects of aerodynamic forces
on vibration modes are considered. But it consumes high computational cost.

16
There is a common way to simplify the solution of equation. When structure forces dominate the blade
motions, it can decouple the structural and aerodynamic effects. It means aerodynamic forces do not alter
the deformation of blades and structural analysis of mode shapes and frequency without considering
aerodynamic forces can be used to solve aerodynamic forces. Since the last stage of steam turbine has a
high mass ratio, this method can be used in the master thesis work.
In order to calculate aerodynamic damping forces, modal analysis of blade and unsteady flow analysis
should be taken into consideration.
Based on the decoupled treatment, modal analysis can be used to solve the equation. Modal analysis is to
determine the natural mode shapes and frequency for free vibration. Damping is ignored and no force is
considered, which can be expressed in the formula:
𝐌𝐌𝑿𝑿̈ + 𝑲𝑲𝑲𝑲 = 𝟎𝟎 (5)

Finite element analysis (FEA) is always considered as the first method to solve the equation above. FEA is
a numerical technical that discretizes the distribution of a variable by dividing the complex geometry into
small and simple elements. Based on it, complex physical problems can be broken into a series of matrix
equations.
�𝑒𝑒 𝑖𝑖𝑖𝑖𝑖𝑖. The equation can be modified as:
Assume the solution is 𝑿𝑿 = 𝒖𝒖
�−𝐌𝐌𝑤𝑤 𝟐𝟐 + 𝑲𝑲 �{𝒖𝒖
�} = 𝟎𝟎 (6)

where 𝒖𝒖
� is the eigenvector and 𝜔𝜔 is the eigenvalue.
For finite element analysis, each natural frequency has a mode shape. There are three first-order mode
shapes, flapwise bending, edgewise bending and torsion. They can be seen in the Figure 5.

Figure 5 First order mode shapes

For present work, the first order flapwise mode shape will be used in the flutter analysis, which has the
lowest natural frequency of all mode shapes.
When turbine blade is rotating, the blade has larger bending stiffness because rotation leads to stiffening
effects for centrifugal forces. This phenomenon is called pre-stressed effect. The equation can be
modified:
�−𝐌𝐌𝑤𝑤 𝟐𝟐 + 𝑲𝑲 + [𝑺𝑺]�{𝒖𝒖
�} = 𝟎𝟎 (7)

where S is stress stiffening matrix.

There is another effect related with rotating blade and it is called spin softening. It is the phenomena that
a part vibrating in the plane of rotation tends to soften the assembly, like twisted blade. The equation can
be changed:
�−𝐌𝐌𝑤𝑤 𝟐𝟐 + 𝑲𝑲 − 𝑤𝑤𝑠𝑠𝟐𝟐 [𝑴𝑴]�{𝒖𝒖
�} = 𝟎𝟎 (8)

where [M] is diagonal mass matrix and 𝑤𝑤𝑠𝑠 is rotational speed.

17
Different numerical methods can be applied in the unsteady flow analysis, time-marching method and
linearized method. Time-marching method is the most popular one for solving unsteady flow. It is the
method to discretize the unsteady equation through a computational grid until it obtains steady-state
solution or periodic unsteady solution. This method can solve 3D Euler and Navier-stokes equations and
reliably predict complex flows in the turbomachinery. However, large size of computational demands
limits its large application, especially for 3D problems. To reduce the computational size and maximize
accuracy, many methods has been developed. Among these, the most widely used one is phase-shifted
boundary conditions. It simplifies the whole annulus to one single passage for circumferential periodic
disturbances.
Compared with non-linear time-accurate method, linearized method is developed to provide a high
efficiency. This method is also called frequency-domain method, which is introduced with a pseudo-time
derivative and then can be solved within a single or several frequencies. It assumes that the unsteady
perturbations are relatively small when compared to the mean flow. However, time-linearized methods
have the disadvantage of neglecting non-linear effects. It is less accurate for cases with shocks or flow
separation.
In this master thesis work, time-marching and phase-shifted boundary conditions will be both used to
obtain the reliable flow condition and less computational time.

2.1.3 Determination of flutter stability

There are some parameters used to describe and judge the stability of flutter system.

Reduced frequency a non-dimensional parameter that describes the degree of unsteadiness of the problem.
It is defined as:

𝑤𝑤𝑤𝑤 2𝜋𝜋Δ𝑡𝑡𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓 (9)

𝑤𝑤 ∗ = =
𝑉𝑉𝑟𝑟𝑟𝑟𝑟𝑟 Δ𝑡𝑡𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏
w = 2πf ( 10 )

Where f is modal frequency, c is chord length, 𝑉𝑉𝑟𝑟𝑟𝑟𝑟𝑟 is relative velocity at outlet for turbine. The critical
value of reduced frequency is 0.2 to 0.5. Before numerical methods were used in the flutter analysis,
reduced frequency is an essential parameter to decide. The higher value means higher unsteadiness of the
flow. In the thesis, the modal frequency will be modified to adapt to the real machine reduced frequency
𝑤𝑤 ∗ to 0.3.

To determine if fluid flow is feeding the energy to blade, unsteady work that flow does to blades should
be calculated. For simplification, simple harmonic motion is assumed for blade motion.

𝐱𝐱 = 𝐡𝐡cos(wt) ( 11 )

𝒙𝒙̇ = −𝑤𝑤𝒉𝒉sin(wt) ( 12 )

𝒏𝒏
� = 𝒏𝒏
� + 𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛(𝑤𝑤𝑤𝑤) ( 13 )

Where h is amplitude and n is a normal value.

The local unsteady pressure is composed of a mean value and a pressure perturbation.
𝑝𝑝 = 𝑝𝑝̅ + 𝑅𝑅𝑅𝑅{𝑝𝑝� exp (𝑖𝑖 𝑤𝑤 𝑡𝑡)} = 𝑝𝑝̅ + 𝑝𝑝𝑟𝑟 cos(𝑤𝑤𝑤𝑤) − 𝑝𝑝𝑖𝑖 sin(𝑤𝑤𝑤𝑤) ( 14 )

As a result, unsteady work can be calculated:

18
 𝑇𝑇 ( 15 )
𝑊𝑊𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎 = � � −𝑝𝑝𝑛𝑛� ∙ 𝑥𝑥̇ 𝑑𝑑𝑑𝑑 𝑑𝑑𝑑𝑑 = − � 𝜋𝜋𝒉𝒉 ∙ 𝑛𝑛�𝑝𝑝𝑖𝑖 𝑑𝑑𝑑𝑑
𝐴𝐴 0 𝐴𝐴

Non-dimensional aerodynamic damping is defined as:

−𝑊𝑊𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎 ( 16 )
Ξ=
𝜋𝜋𝜋𝜋𝛼𝛼 2 𝑐𝑐 2 𝑝𝑝𝑟𝑟𝑟𝑟𝑟𝑟
Where 𝛼𝛼 = ℎ 𝑚𝑚𝑚𝑚𝑚𝑚 /𝑐𝑐 and ℎ 𝑚𝑚𝑚𝑚𝑚𝑚 is the maximum blade displacement, b is the blade height. 𝑝𝑝𝑟𝑟𝑟𝑟𝑟𝑟 is the total
pressure at the inlet minus static pressure at the inlet. It shows the overall stability of the blade. Non-
dimensional aerodynamic damping is used in the thesis work to predict if the rotor under the risk of
flutter. When aerodynamic damping is positive, it shows the system is not stable because fluid flow is
providing the energy. Otherwise, it is stable.
In order to detect the local stability of the blade, local work coefficient can be used:

−𝒉𝒉 ∙ 𝑛𝑛�𝑝𝑝𝑖𝑖 ( 17 )
𝑤𝑤 =
𝛼𝛼 2 𝑐𝑐𝑝𝑝𝑟𝑟𝑟𝑟𝑟𝑟
In CFX, there is also a parameter which shares similar definition to detect the local stability, called Wall
Work Density. However, different from definition of unsteady work, wall work density is defined as
“work done by a moving wall onto the fluid per unit area” (ANSYS, Variables and Predefined
Expressions Available in CEL Expressions n.d.).

𝑤𝑤𝑤𝑤𝑤𝑤 = 𝑝𝑝𝒏𝒏∆𝑥𝑥 ( 18 )

Wall work density can be also used to determine the unsteady work.

( 19 )
𝑊𝑊 = � 𝑤𝑤𝑤𝑤𝑤𝑤 𝑑𝑑𝑑𝑑

This solution will be used in the thesis work to calculate the unsteady work and aerodynamic damping.

2.1.4 Fourier transformation method

Although many methods have been utilized to reduce the computational time, for example nodal diameter
and periodic boundary condition, more than two passages should be included in the simulation. Fourier
Transformation(FT) method can be applied to reduce the number of passages and two are enough for any
nodal diameters in CFX. The Figure 6 shows the difference of using phase-shifted periodic boundaries
and periodic boundary conditions for 90 degree inter blade phase angle.

Figure 6 Comparison of phase-shifted periodic boundaries and periodic boundary conditions for 90 degree IBPA

19
Fourier Transformation(FT) method in ANSYS CFX uses phase-shifted periodic boundaries, which
stores unsteady flow history. The unsteady flow characteristics are compressed and stored by temporal
Fourier series decomposition.
For flutter analysis, the disturbance from the periodic motion of wall boundary is only considered for
interest and the disturbance from stator-rotor interaction is not included. The feature of interaction can be
analyzed in the forced response.
In this master thesis, FT is mainly used as the method to solve unsteady simulations for its advantages of
saving time and memory. Time transformation is used to verify the accuracy of FT method.

2.2 Mixing plane

Mixing plane is one of the interface treatment that enables multiple frames to be calculated at the same
time. Take the case in this thesis report as an example. Stator, rotor and diffuser will be treated in different
ways as they are in different frames. As a result, the unsteady state of multi rows can be calculated as
steady state for each frame. This method includes circumferentially averaging fluxes at the interface.
Mixing plane method has been widely used in turbine treatment of multi-row simulations (Guénette 2012)
(Arnone 1994) (Stein 2016) (P. M. Petrie-Repar 2014).
However, the drawback of this method is obvious as well. The unsteady characteristics will be erased by
averaging, like wakes and clearance effects from upstream. Loss will appear when flow has great gradient.
Mass, momentum or energy at the interface will not be conservative (Arnone 1994). For CFX, the mixing
plane can cause one-time mixing loss. “The loss is equivalent to assuming that the physical mixing
supplied by the relative motion between components is sufficiently large to case any upstream velocity
profile to mix out prior to entering the downstream machine component (ANSYS 2013).” There is a
phenomenon that is detected in previous study occurring in this thesis report as well. Arnone et,al
(Arnone 1994) found that pit-averaging can lead to reflection in shock system. He applied one approach
that separates stator and rotor and creates a special mixing plane to eliminate the reflection effect. As
realistic geometry should be considered in this report, the method cannot be used. The other method is to
make mixing plane far apart from blade. This approach will be applied for interface at the rotor and
diffuser.
In Burton’s thesis report, she talked about different treatment of interfaces, including mixing plane
(Burton 2014). She also compared different methods, but for isolated last stage, mixing plane method is
the only one utilized.

2.3 Open geometry

The low pressure last stage blading geometry is always kept as a secret and protected by manufactures.
The geometry in this thesis is from an open source in Durham University (Burton 2014). To author’s
knowledge, it is the only geometry for steam turbine last stage found in the literature review that is open
to the public.
The stage design is based on a genetic blade generated by Alstom Power in Rugby and Durham University,
which is 20 years old, 50Hz. Burton (Burton 2014) modified it to provide a typical inflow conditions for
the exhaust hood. It is composed of two parts, stator and rotor. The stage can purely present the
aerodynamical characteristics for the modern blading and it cannot be used for manufactures.
The stage basic geometry features are shown in Table 1. The hub, mid-span and shroud of the stage used
in the thesis report are shown in the following figures. The modified stator has less back surface deflection
angle (BSDA) and it reduced the strength of the strength of shock structure and suppress the separation at
the trailing edge. The rotor is highly twisted and the difference between rotor hub and tip stagger angle is

20
about 60 degrees. The number of rotor blades has increased from 60 to 65, which decreases the loss from
the gap of adjacent blades. All these methods of modification can provide a higher efficiency and a
uniform rotor static pressure outlet.

Table 1 The stage features

Stator Rotor
Blade Count 60 65
RPM - 3000
Blade Length[m] 0.79 0.92
Hub Diameter[m] 0.77 -

Figure 8 Blade profiles at the tip

Figure 7 Stage geometry and Boundary

Figure 9 Blade profiles at mid-span Figure 10 Blade profiles at hub

21
The stator inlet conditions include the representative profiles of total pressure, total temperature and three
velocity components. The file is generated by Alstom Power from previous multi-stage calculation. Inlet
parameters vary with the span, which is shown in the Table 2. The rotor outlet static pressure is modified
to 8800 Pa in order to eliminate the loss.
Table 2 Stator inlet profile

Zoe used the ANSYS Fluent to run the simulation. Standard k-ϵ is used as the Turbulence Model to
approximate the turbulence characteristics of flow. The setting of simulation is shown in the Table 3. To
facilitate the convergence, the author used the trick of increase the rotating speed step by step. The same
turbulence model and trick will be used in the running simulation as well.
Table 3 Setting of Fluent from Durham

The main objective of Zoe’s work is to study the flow in steam turbine exhaust hood. The master thesis
report only uses the geometry and boundary conditions which are freely provided by Durham University.
And only steady state results can be applied in the master thesis as a reference to verify the results this
report obtained.

22
3 Methodology
3.1 Introduction of numerical tools
The simulation for present work applied ANSYS 17.0 as numerical tools. ANSYS 17.0 has a package of
turbomachinery analysis. DesignModeler is the first step to prepare the geometry for mesh generation.
TurboGrid is a software designed to provide high-quality hexahedral meshes for rotating machines. It has
fully automatic topology and meshing option even for complex blade shapes, which can maintain high
level of mesh quality and efficiency at the same time. For tip clearance at the shroud, TurboGrid can
specify the distance of the gap and the grid can be set to match or mismatch the center of blade thickness.
ANSYS APDL is to perform the Finite Element Analysis (FEA) for blades. In present work, only modal
analysis is conducted by APDL. For running the simulation, ANSYS CFX is the main set of software.
CFX-pre is preprocessor to set up boundary conditions, simulation model etc. CFX-solver is to solve and
monitor the results. CFD-post is the post-processor to visualize and synthesize results. Matlab2015b is
also used for statistics post-processing.
Compared to previous versions, ANSYS 17.0 has new features for blade row modelling. Fourier
transformation can solve multi-stage and multi-disturbance transient blade row problems. And the speed
is 10x faster than before, which saves computational time and resources. These features give the
opportunity to explore more phenomenon in my case.

3.2 Workflow
The thesis work can be divided into three parts, preparation, steady state and unsteady state. Geometry
files are imported to set the flow path and then generate meshes. Aforementioned boundary conditions
are used in the steady state analysis for both tip and no tip. Meanwhile, modal analysis is conducted to
provide the mesh displacement. Subsequently, results from steady state are used as initial results for
unsteady analysis. For limited time, unsteady state is only run for tip. The Figure 11 shows the workflow
of present thesis work.

Figure 11 Workflow of steady and unsteady analysis

23
3.3 Flow-path setup
Geometry file is already given as ‘.iges’ format from Durham University. The file includes stator and rotor
blade geometry. In order to import the geometry into TurboGrid to generate mesh, shroud and hub
should be defined in DesignModeler. Different from definition of domains, the stage is split into three
domains, stator, rotor and diffuser.
However, since hub and shroud lines from the original file disappear in DesignModeler, the position
boundaries are drawn according to limited setting parameters from Durham University by author of this
master thesis report. Inlet is located at 2.5 times chord length of stator at hub upstream the stator leading
edge. Outlet is at 2.5 times chord length of rotor at hub downstream the rotor trailing edge. The
stator/rotor interface is defined in the middle of two blades. Rotor/diffuser interface is defined at 0.2
rotor chord downstream the rotor trailing edge. These two interfaces are set as mixing plane. The plane
upstream on chord length of outlet boundary is comparison plane and it is used to compare parameters to
verify results. The position of hub and shroud is set according to hub diameter and stator and rotor blade
length, which is shown in Table 1. The key points coordinates of the end wall are shown in Table 4. The
end wall is drawn by linking the key points with straight lines.

Figure 12 Schematic of stage and boundaries Figure 13 Comparison of different end wall setup

The difference between these two end wall setups can be seen in Figure 13. The main difference appears
at the inlet length of stator, shroud position of the rotor tip and hub and shroud lean angles of the diffuser.
In my end wall definition, straight lines are used instead of some bump lines in Durham settings. And
outlet boundary is set to be parallel to rotor and diffuser mixing plane instead of parallel to inlet boundary.
All simulation results in this master thesis report are based on the end walls defined by author.

Table 4 Key points Coordinates of the end wall

(z,r) Hub (mm) Shroud (mm)

1 (-68,712.70) (-68,1331.6)
2 (600.50,712.70) (68,1331.6)
3 (728.92,738.38) (374.96,1536.37)
4 (1188.05,911.79) (1098.09,2030.52)

24
3.4 Mesh generation
As described before, the stage has been divided into three parts, stator, rotor and diffuser. For meshing,
rotor and diffuser are combined as one domain to generate mesh. In TurboGrid, there is a topology
method called “automated topology and mesh (ATM) optimized”. The topology used in ATM is an H-O
grid, which makes the O-grid for boundaries layers and hexahedral meshes for other parts. This topology
method can offer the high-quality mesh. Meanwhile, mesh size can be controlled. The first element width
is set as Y plus 40 for wall functions to resolve boundary layer flow. The expansion rate is used to control
the number of cells for boundary layers. In CFX, it is said that “As a general rule, any important shear
layer in the flow (boundary layer, mixing layer, free jets, wakes, and so on) should be resolved with at least
10 nodes normal to the layer. This is a very challenging requirement that often requires the use of grids
that are aligned with the shear layers. (ANSYS 2013)” As a result, at least ten nodes have been made in the
boundary layers and detailed mesh for boundary layers can be seen in Figure 14 and Figure 17. Meshes in
the figures shows the coarse one. The total nodes for stator are 512239 and for rotor and diffuser are
566482.

Figure 14 Rotor and diffuser mesh at mid-span & detailed leading and trailing edge mesh

Figure 15 Rotor and diffuser mesh at shroud Figure 16 Rotor and diffuser mesh at hub

25
 Figure 17 Stator mesh at mid-span & detailed leading and trailing edge mesh

Figure 18 Stator mesh at shroud Figure 19 Stator mesh at hub

3.5 Modal analysis

In ANSYS APDL, it assumes that rotor blade is fixed at a rigid hub and the mode shape is first flap
bending mode. Pre-stressed and spin softening are considered in the modal analysis. Material of blade is
typical and set as titanium alloy, whose properties are shown in Table 5. Spin softening in this case did
not change the frequency a lot, which can be seen in Figure 20.

Figure 20 Comparison of with and without spin softening

26
 Table 5 Blade material Properties

Titanium alloy
Density 4620 kg m-3
Young’s modulus 9.6E10 Pa
Poisson’s ratio 0.36

Figure 21 Displacement of mode shape

The calculated modal frequency is 89.84Hz for rotation speed at 3000 rpm and maximum total
displacement is 1.32 meter. Figure 22 shows the nondisplaced and displaced FE mesh. Different element
meshing sizes are applied in the grid convergence for modal analysis. In Figure 23, it can be seen that
cases with size smaller than 0.006m share similar results of frequency. Considering both results precision
and computational time, mesh element size of 0.006m is taken for modal analysis and unsteady analysis in
the next step.

Figure 22 Nondisplaced and displaced FEM mesh

The modal frequency should be modified to match reduced frequency of 𝜔𝜔∗ =0.3. It is because the
modified blade geometry did not consider the structural mechanics and the calculated modal frequency is
too low for real situation. The frequency can be calculated as:
𝜔𝜔∗ 𝑉𝑉
f=
2𝜋𝜋𝜋𝜋
where f is modal frequency. c is chord length at mid-span. V is relative velocity at turbine outlet. The
modified modal frequency is 132.08 Hz.

27
 Figure 23 Grid convergence for modal analysis

3.6 Solver setup

3.6.1 Steady
In CFX-pre, stator, rotor and diffuser are combined together to make the solver setting. The following
Figure 24 shows the model geometry and boundary conditions.

Figure 24 Solver model and boundary conditions

Flow-path setup has mentioned the position of all boundaries. In this part, boundary conditions will be
specified. Inlet conditions are offered by a file of total pressure, velocity components and total
temperature varying with inlet radius. It can be seen in Figure 25 and Figure 26. Outlet is set static
pressure 8800 Pa and the value is the average over the whole outlet. Interfaces between stator and rotor
and between rotor and diffuser are defined as mixing-plane (stage). The stage model obtains average
fluxes at circumferential direction at the interface, which enables multi-rows prediction for steady state
simulation. At the stage interface, the average static pressure is set to be average band pressure for both
sides of upstream and downstream of the interface (ANSYS 2013). There are two ways to treat velocity at
the downstream, static average velocity and constant total pressure. They will be compared and discussed
in Chapter 4. Passage boundaries are set as rotational periodic and rotate at axis z. The setting allows one

28
passage for solving the case and it saves computational time and resources. Other boundaries like hub,
shroud and blades are set as no slip wall.

Figure 25 Inlet total pressure and velocity (Burton 2014) Figure 26 Inlet total temperature and viscosity (Burton 2014)

Stator and diffuser are stationary and rotor rotates at axis z with rotation speed of 3000 rpm. Fluid flow in
the last stage is steam and it is approximated as ideal gas. It is a common simplification. Multi-phase of
wet steam can be research as a secondary effect. For present work, wet steam is only seen as ideal gas. The
following Table 6 shows the steam properties that are used in the solver. The most common k-ϵ
turbulence model is used to calculate the turbulent characteristics of flow. For this model, 10% turbulence
intensity is specified. Scalable wall function is used to calculate the boundary layer flows. After about 300
iterations, the solution converges.
Table 6 Fluid properties

Cp 4153 J/kgK
Gamma 1.12
Thermal conductivity 0.061 W/mK
Dynamic Viscosity 1.032E-5 Pas

Table 7 Summary of solver setting

Working fluid Water data, ideal gas

Turbulence Model Standard K-e
Turbulence Intensity 10%
Wall function Scalable
Turbulence numerics High resolution
Iteration Approx. 300

To facilitate converge, setting of first order discretization was initially used. Back pressure gradually
decreased from 16000 Pa to 8800 Pa and rotational speed increased from 1000 rpm to 3000 rpm until all
residuals are below 1E-4. High resolution for turbulence model is used after it converged. The number of
iteration steps is about 300 for steady state simulation.

29
3.6.2 Unsteady

3.6.2.1 Solver setup

For unsteady simulation, rotor is isolated from full stage. Steady state should be prepared for unsteady
simulation as initial results. Steady state for single rotor is calculated with suitable inlet and outlet
boundary conditions. Inlet and outlet boundary conditions were supposed to be extracted from mixing
plane. Since noise from mixing plane is found to affect flow condition, inlet profile is obtained from about
2% downstream of mixing plane, which will be further discussed in Chapter 5.1. Steady state simulation
converged to 10^-5 to ensure the flow field is same with full stage steady state in order to eliminate
fluctuation effect from steady.
Fourier transformation is chosen as the main method for calculating unsteadiness of different IBPA cases,
which is available in CFX Transient Blade Row models. And time transformation is also used to verify
results of Fourier transformation, especially for least stable IBPA.
As mesh deformation is periodic, time period is the reciprocal of modified modal frequency of rotor blade.
Time step for each time period is set as 65 for odd passage numbers of rotor.

3.6.2.2 Mesh motion and deformation

The mode shape aforementioned is applied to CFX unsteady simulation. When blade displacement
appears, mesh is forced to move or deform. In order to obtain results as accurate as possible, deformation
should be reduced. Ansys CFX can control mesh motion and deformation. The option of Regions of
Motion Specified is set relative to initial mesh. Mesh stiffness is recommended to be variable in CFX
solver modelling to protect mesh distribution. One custom expression is applied for mesh stiffness value,
which is inversely proportional to volume of finite volumes. The value can help to maintain mesh quality
and reduce elements inverting.
The scaling factor of mesh motion is chosen to be 0.001 to make the maximum displacement of blade is
about 1% of mid-span chord length, which could catch unsteady fluctuation and cause no big mesh
deformation at the same time. Mesh motion for rotor shroud and hub is set as unspecified, which means
there is no constraints on nodes motion.
The maximum blade displacement is about 1.316mm at the tip. It is shown in Figure 27.

Figure 27 Total mesh displacement at rotor shroud

30
4 Steady state results and discussion
In steady state analysis, cases of different lengths of rotor exit are conducted to study mixing plane’s
effects. The steady state for no tip clearance gap is also calculated by inviscid flow model and the results
are included in Appendix II. And two tip gap cases are included to investigate effects of tip clearance flow
effects.

4.1 Influence of mixing plane

Multi-row steady-state simulation of stator, rotor and diffuser is performed. Mixing plane should be used
between stator and rotor and between rotor and diffuser so that steady-state solutions can be obtained for
different domains in each reference frame. In order to validate mixing plane, different treatments of
velocity and reflection of interfaces are investigated.

4.1.1 Comparison of interface velocity constraint

In CFX, there are two ways to treat velocity on the downstream side for mixing plane, stage average
velocity(SAV) and constant total pressure(CTP). The difference between these two treatments is average
band velocity and average band total pressure calculated for interface respectively. Comparison of these
two treatments has been made for 90% span blade loading for stator and rotor and different parameters
from hub to shroud for mixing planes.
The flow angle is defined as:
𝛼𝛼 = 𝑡𝑡𝑡𝑡𝑡𝑡 −1 (𝑢𝑢𝜃𝜃⁄𝑢𝑢𝑥𝑥 )
𝛽𝛽 = 𝑠𝑠𝑠𝑠𝑠𝑠−1 (𝑢𝑢𝑟𝑟 /𝑈𝑈)

Where 𝑈𝑈 = �(𝑢𝑢2𝑥𝑥 + 𝑢𝑢2𝑦𝑦 + 𝑢𝑢2𝑧𝑧 ).

From comparison of Figure 28 to Figure 34, no significant variation is seen in plots. For the case in the
report, these two ways do not alter results a lot. As comparison should be obtained in Appendix II with
LUFT code, stage average velocity is chosen for this thesis report, which means that velocity at
downstream of mixing plane is calculated as average band velocity.

Figure 28 Comparison for stator 90% span blade loading for stator Figure 29 Comparison for rotor 90% span blade loading for rotor

31
Figure 30 Absolute Mach number from hub to shroud for different Figure 31 Static Pressure from hub to shroud for different mixing
mixing plane plane

Figure 32 Absolute total pressure from hub to shroud for different Figure 33 Absolute Total Temperature from hub to shroud for
mixing plane different mixing plane

Figure 34 Absolute flow alpha

32
4.1.2 Comparison of stationary and rotating diffusers
To validate the mixing plane method, three cases are performed to examine the interface between rotor
and diffuser. The first case is the base case, which has short rotor exit. In the second case, rotor and
diffuser are set as one domain and diffuser rotates with rotor. As no slip wall is used for viscous fluid,
counter-rotating wall is specified for diffuser hub and shroud in order to obtain equivalent flow solution.
It means that diffuser hub and shroud are assumed to be stationary at stationary frame. As a comparison,
the third case also set rotor and diffuser as one rotating domain, but it does not split hub and shroud of
rotor and diffuser. Figure 35 to Figure 37 show different mass-averaged absolute parameters from hub to
shroud at rotor-diffuser mixing plane. For long rotor cases, data from same plane with short rotor exit is
extracted and compared.
For two sides of interface of short case, total pressure and Mach number at diffuser inlet is lower than that
at rotor exit. It is because pitch-wise averaging has caused the entropy loss at the mixing plane. Total
temperature for both sides are consistent and this means that energy is conservative for the mixing plane.
However, it can be seen from Figure 38 to Figure 40 that for inviscid flow, these parameters agree well for
two sides of mixing plane. Mixing plane did not cause entropy loss for inviscid case. It is because laminar
solver used for inviscid case erased out turbulence model and diffusion effects. This makes the wake and
shock at the trailing edge of rotor weaker than that RANS captured. It can be figured out clearly from the
figure of first harmonic amplitude of pressure, Figure 43. It is defined as the first harmonic amplitude of
the spatial pressure distribution for each radius. It is used to capture the flow parameter gradient and big
value means mixing plane erased out more unsteady features that can cause more loss. In Figure 43, it is
clear to see that inviscid flow model has much less amplitude of RANS case. As a result, that is the reason
difference of mixing plane loss for two flow models.
Fluxes for long case without splitting at mixing plane are more similar to those for short case rotor exit.
However, results for long case with splitting differ greatly from those for other cases. It is because the
change of flow velocity from rotor to diffuser at hub and shroud created a blockage of flow field. It can
be seen also from the blade loading figures, Figure 41 and Figure 42, which causes effects mainly on the
trailing edge of rotor.

Figure 35 Absolute Mach number at rotor-diffuser mixing plane Figure 36 Total Pressure at rotor-diffuser mixing plane (RANS)
(RANS)

33
Figure 37 Total temperature at rotor-diffuser mixing plane (RANS) Figure 38 Absolute Mach number at rotor-diffuser mixing plane
(Inviscid)

Figure 39 Absolute total pressure at rotor-diffuser mixing plane Figure 40 Absolute total temperature at rotor-diffuser mixing plane
(inviscid) (inviscid)

Figure 41 Blade loading at 50% span (RANS) Figure 42 Blade loading at 90% span (RANS)

34
 Figure 43 First harmonic amplitude of pressure at the mixing plane

Schlieren plot is made for both cases at 50% rotor span, which is the plot of density gradient. At the trailing
edge of rotor, wave reflection can be figured out, especially compared to Schlieren plot for second case.
This can also be examined in blade loading figure. The wave reflection provides pressure fluctuation for
rotor trailing edge. The reflection is because of boundary location. Rotor exit plane is located only about
0.2 chord downstream the rotor trailing edge. As a result, different length of rotor exit will be investigated
to eliminate effect from wave reflection in Chapter 4.1.3.

Figure 44 Density Gradient for rotor at 50% span for short case Figure 45 Density Gradient for rotor at 50% span for long case

4.1.3 Comparison of different length of rotor exit

To reduce flow reflection, longer rotor exit is applied. Cases of one chord and two chord long rotor exit
are compared for rotor blade loading at different spans, Figure 46 and Figure 47. Since comparison made
above shows long case without splitting is more similar to short case, this long case is chosen to compare
with different length cases. At 50% and 90% span, results of four lengths rotor exit show that longer exit
decreases flow reflection, especially for transonic tip region. But these cases results do not exactly agree
with case of rotor-diffuser as long case as effects from mixing plane.

35
 Figure 46 Rotor blade loading at 50% span Figure 47 Rotor blade loading at 90% span

Figure 48 and Figure 49 are Schlieren plots of one and two chord rotor exit at 50% span. No obvious wave
reflection can be seen in plots.

Figure 48 Density gradient for one chord rotor exit at 50% span Figure 49 Density gradient for two chord rotor exit at 50% span

4.2 Steady state results for no tip clearance

4.2.1 Comparison with reference
In Durham Stage Test Case(DSTC), the stage is split into two domains, stator and rotor (Burton 2014).
The setting is similar with one domain case in this thesis. Comparison will be made between DSTC and
one domain case. Total pressure and swirl angle alpha at comparison plane are extracted from CFX results
and compared with reference, which are shown in Figure 50 and Figure 51.
In order to provide a representative boundary profile for exhaust hood inlet, DSTC validated its results by
comparing with other published resources, (Beevers 2010)and (Gardzilewicz 2003). Although the end wall
definition is different from original setup, from Figure 50 and Figure 51, it can be seen that results agree
with the reference well and flow field conditions for different definitions of end wall do not vary too
much.

36
 Figure 50 Comparison of total pressure at comparison plane (Burton 2014)

Figure 51 Comparison of Swirl alpha at comparison plane (Burton 2014)

4.2.2 Steady state results and discussion

Long no-splitting case is reasonable to be used to analyze steady state results because of no wave
reflection and direct comparison with DSTC. It has a typical static pressure and Mach number distribution
in steam turbine last-stage. Typical plots of Mach number and pressure is shown in Figure 52. Summary of
steady cases is shown in Table 8.
Table 8 summary of steady state cases

Rotor Domain Short Long No-splitting

Mass(kg/s) 85.14 85.15
Power (MW) 11.97 12.07
First cell Y+ 40 40
Total to static efficiency (%) 88.13 85.43

37
 Figure 52 Mach number and pressure at typical positions

The averaged Mach number at stator inlet is 0.35. The Mach number is accelerated in the stator to about
1.1. The Mach number at stator exit has a great gradient, which at hub is higher than that at shroud. It
causes increase of Mach number in the rotating frame at the tip and hub for rotor inlet.

Figure 53 Mach number contour for rotor 5% span

At the hub of rotor blade, low pressure produces a low root reaction and that causes flow separation at
suction side. From axial plane, greatest separation can be seen at rotor 5% span. Figure 53 shows this
phenomenon. At the tip of rotor blade, higher pressure drop induces high relative exit Mach number,
which triggers shocks within blade passages. At the diffuser outlet, Mach number drops to 0.38.
Figure 55 and Figure 56 shows the typical flow at the tip region in the last stage of steam turbine (Petrie-
Repar 2015). There are two oblique shocks for rotors starting at the trailing edge. One is the passage
shock and flow after it starts to expand. The other one spreads to downstream and points to the outlet.
This shock deaccelerates the flow at the suction side.
Mass flow calculated is conservative. Mass flow at inlet is 85.1544 kg/s and it is 85.4464 kg/s at outlet.
The difference mass flow at inlet and outlet is small and the value is similar to 86.6 kg/s that DSTC
provides (Z. I. Burton 2014).

38
 Figure 54 Mach number contour at rotor 50% span

Figure 55 Mach number contour at rotor 90% span

Figure 56 Mach number contour at rotor 95% span

4.3 Steady state results for 0.5% tip clearance

4.3.1 Comparison with reference
In order to compare the results with reference directly, the tip ratio is defined as the radial direction
distance to blade height (Z. I. Burton 2014). It means the radial direction distances for 0.5% is 4.6mm.
The normal distance for tip gap is about 2.7 mm, which is typical for steam turbine.
Figure 57 shows the comparison between literature and my simulation from CFX. The results agree well
with each other and tip clearance has a great effect on the span over 90%. The total pressure raises at the
tip, evident to all tip clearance flow.

39
 Figure 57 Comparison of 0.5% tip and no tip at rotor outlet

4.3.2 Tip clearance steady state results and discussion

Tip clearance steady state is performed for 0.5% and 1% tip gap both short and long case. Based on
previous study on splitting and non-splitting case, long case here is non-splitting case. Tip clearance
effects on blade loadings can be seen for 50% span to 99% span, from Figure 58 to Figure 61. At the 50%
span, blade loading does not change a lot. It is because tip clearance effects are limited at the tip region. At
90% and 95% span, suction side pressure changes, especially for rotor trailing edge. But as rotor blade is
long scale at the span, the main effect of tip clearance flow is mainly over 98% span. At 99% span, the
blade loading for 0.5% tip gap changes more than other spans, but the difference is relatively small. 1%
blade loading changes more than 0.5% tip gap case. With increasing tip gap, pressure at pressure side
decreases more. It means that tip clearance flow decreases the pressure drop between pressure side and
suction side at the tip part.
Table 9 Steady state cases summary for tip clearance flow

Rotor Domain 0.5% tip short 0.5% tip long 1% tip short 1% tip long
Mass(kg/s) 85.06 85.03 85.21 85.21
Power (MW) 11.61 11.7 11.62 11.57
First cell Y+ 40 40 40 40
Total to static efficiency (%) 86.58% 85.76% 86.32% 85.46%

Figure 58 Blade loading comparison for 0.5% tip and no tip at 50% Figure 59 Blade loading comparison for 0.5% tip and no tip at 90%
span span

40
Figure 60 Blade loading comparison for 0.5% tip and no tip at 95% Figure 61 Blade loading comparison for 0.5% tip and no tip at 99%
span span

Figure 62 shows the Mach number contour of tip region at the mid chord. It can be seen in the figure that
Mach number increases in the tip region and that is due to tip clearance flow from pressure side to suction
side. When tip gap increases, flow velocity in the tip region increases as well. For all three tips, a low Much
number region can be seen in the downstream of tip flow. With larger velocity at the tip region, the region
appears at further downstream place. There is a high Mach number region near the blade. When tip gap
increases, the Mach number of this region becomes higher. This can explain the reason why there is
pressure propagation of rotor suction side for different tip gaps.

No tip 0.5% tip 1% tip

Figure 62 visualization of tip clearance flow

4.4 Grid convergence

4.4.1 Grid convergence for no tip
Three levels of mesh resolution have been investigated for different spans. Blade loadings for stator and
rotor are shown from Figure 63 to Figure 68. Five levels are grids from 1 million nodes to 3 million nodes,
which is to explore the relation between nodes of mesh and grid convergence performance. The detailed
information is displayed in Table 10. The upper limit of nodes for grid convergence is made by CFX
software.
From blade loading curves, it can be recognized that there is no big variation for these three levels. The
difference appears mainly at the trailing edge of the suction side. Cases 1M is acceptable. For the reason of
fastening computation, 1M is used for unsteady state as the base grid.

41
 Table 10 Mesh information for grid convergence without tip clearance

1M 2M 3M
Stator Nodes 406436 905061 1295259
Nodes on hub 4782 8961 11669
Cells in radial 85 101 111
Cells in boundary layer 11 11 15
Rotor & Nodes 566482 1208720 1857052
diffuser
Nodes on hub 6587 10407 14176
Cells in radial 86 116 131
Cells in boundary layer 11 11 12
Total Nodes 972918 2113781 3152311

Figure 63 Blade loading for stator at 50% span Figure 64 Blade loading for stator at 90% span

Figure 65 Blade loading for stator at 95% span Figure 66 Blade loading for rotor at 50% span

42
 Figure 67 Blade loading for stator at 90% span Figure 68 Blade loading for stator at 95% span

4.4.2 Grid convergence for tip

Stator mesh and domain topology are unchanged as the described 1M case in chapter4.4.1. Three different
mesh cells are generated and investigated for 0.5% tip clearance. Mesh information for tip clearance is
shown in and Table 11. The blade loadings are identical for different cells in the tip region. It shows that
the grid convergence works well for cases with tip clearance. The tip mesh is shown in Figure 72.

Table 11 Mesh information for different tip cells for 0.5% tip

8800Pa 0.5% tip Coarse Medium Fine

Rotor Nodes 438163 462628 487093

Nodes on hub 3946 3946 3946

Cells at tip clearance 20 25 30
Min. Face angle (deg) 27.3621 27.3621 27.3621

Figure 69 Different tip cells at 50% span for 0.5% tip Figure 70 Different tip cells at 90% span for 0.5% tip

43
Figure 71 Different tip cells at 95% span for 0.5% tip Figure 72 Tip meshes for 0.5% tip

44
5 Unsteady state results and discussion
ANSYS CFX is used to calculate the single rotor for different travelling wave modes. Since reflection has
been detected in steady state, both short and long rotor cases will be investigated to show rotor exit
effects on unsteady results. Inviscid flow model is also calculated to compare with RANS model. In
Appendix II, LUFT code is used to verify the solver of ANSYS CFX for unsteady state.

5.1 Initial steady results preparation

As Chapter 3.6.2.1 mentioned, steady state results of single rotor should be prepared as the initial results
for unsteady simulation. Boundary profiles were extracted from multi-row steady state results for single
rotor domain to set as unsteady inlet and outlet conditions. When they were extracted directly from
mixing plane, it shows big difference at the leading edge of blade loading, as shown in Figure 73, Figure 74
and Figure 75. It is because the noise of mixing plane provides the error mainly on the rotor inlet.

Figure 73 Blade loading comparison at 30% span for initial steady Figure 74 Blade loading comparison at 50% span for initial steady
results result

One method is applied to eliminate the mixing plane effect that boundary inlet profile is extracted from
2% downstream the mixing plane instead, which is still far from rotor blade. The mass circumferential
averaged quantities are applied from hub to shroud. It provides a good agreement for blade loading, which
eliminates the effects from different flow field of multi and single row steady state. In this thesis report,
short RANS case is taken as the example to show comparison of blade loading at 30%, 50% and 90%
span.

Figure 75 Blade loading comparison at 90% span for initial steady results

45
Depending on different geometry and solvers, the position of modified boundary has 1% difference. But
all of them are far away from the rotor blade and it is the place that rotor passage flow does not change
the circumferential distribution of inlet parameters.

5.2 Results and discussion for no tip clearance

RANS short, RANS long, inviscid one chord and long cases are compared for normalized aerodynamic
damping, as shown in Figure 76. Unfortunately, CFX solution did not converge for +/- 90 degrees
Fourier Transformation for RANS and inviscid long cases. It is possibly because long rotor outlet made
solution hard to converge. However, the solver works for time transformation. Comparison of cases of 0
and 180 degrees for both solvers is made and it is found that the difference is less than 2%. Data are
available in Appendix I. As a result, it is reasonable for time transformation results of +/- 90 degrees to be
accepted for comparison. In Figure 76, long cases for both RANS and inviscid methods share similar
shape of curves. But shorter rotor cases have different results, which means the aerodynamic damping
results are very sensitive to different length of rotor outlet. It is because reflection from rotor outlet from
both steady state and unsteady state disturbs unsteady flow, which changes aerodynamic work on blades.
And reflection effects are different for different rotor lengths. The biggest difference appears at +/- 90
degrees IBPA, which means these two IBPAs are very sensitive to wave reflection effects. The wave
reflection can be directly seen in the Schlieren plot of both short and long RANS cases at 50% span,
Figure 77 and Figure 78.

Figure 76 Normalized aerodynamic damping

Negative value for normalised aerodynamic damping means rotor blade is aerodynamically unstable. All
case curves have similar trend of curves. The least stable position appears at -90 degrees IBPA. The result
is consistent with literatures (Masserey 2012) (Rice 2009). However, due to wave reflection effects, it is
hard to determine if rotor blade is unstable at the IBPA of -90 degrees. Long cases of RANS and inviscid
are most trustable for no deflection detected. The minimum values of normalised aerodynamic damping
for these two cases are -0.0036 and -0.1256 respectively. This means that the blade could experience
flutter vibration at this point. But mechanical damping should be also considered as another criterion to
determine the flutter risk. Meanwhile, more IBPAs could be calculated and the normalised aerodynamic
damping of other IBPAs could be lower than that of -90 degrees. It means that the blade could be under
higher risk of flutter at other IBPAs.

46
Figure 77 Schlieren plot of short rotor 50% span at maximum mesh Figure 78 Schlieren plot of long rotor 50% span at maximum mesh
displacement displacement

Local wall work density for different flow solvers is shown for least stable IBPA at 50% and 90% span for
both pressure side and suction side, from Figure 79 to Figure 82. Positive local wall work density means
energy are provided by aerodynamic forces to blade and it makes blade unstable. The suction side curve
shapes are different. This is because reflection mainly influence the suction side.

Figure 79 Normalised local work coefficient for -90 degree at 50% Figure 80 Normalised local work coefficient for -90 degree at 50%
span pressure side span suction side

Figure 81 Normalised local work coefficient for -90 degree at 90% Figure 82 Normalised local work coefficient for -90 degree at 90%
span pressure side span suction side

47
For long cases of RANS and inviscid, the suction side is quite similar at the trailing edge of the rotor at 90%
span for local word coefficient. It can also explain the phenomenon that in local wall work density versus
span plot, these two cases share similar curves, Figure 83.
In Figure 83, it can be seen that most work done on the blade is at the tip region of blade. When wall
work density is positive, it means blade is providing work on the fluid. Otherwise, blade is absorbing
energy from fluid. The unstable work done on the blade is mainly from the tip over 80% span. For the
span between 60% and 80%, the rotor blade is under the aerodynamic damping. From the figure, it can
also be seen that reflection mainly influences the tip region as well.

Figure 83 Local wall work density versus span

48
6 Conclusions
A 3D steam turbine flutter test case has been established. The test case includes many features that are
important for steam turbine flutter calculations, such as a 3D long twisted blade, transonic flow at the
rotor exit, high stagger angle near the tip, mixing planes close to the blade and a blade with low natural
frequency.
The initial results are quite different for cases of different lengths of rotor exit. It is suspected that the
reason is steady and unsteady flow reflections from the mixing planes and the rotor outlet. The least stable
IBPA appears at -90 degrees, which could cause flutter vibration. Unsteady aerodynamic work is done
mainly on the tip region of rotor blade.
Initial steady results for tip clearance flow shows that typical gap of 0.5% did not show big impact on the
flow field.

49
7 Future work
The test case is established for flutter analysis based on a 3D open last stage blade of steam turbine.
However, due to time limit, more work could be done. More IBPAs can be done for unsteady cases to
make the aerodynamic damping curve more compete. It can investigate the least stable value for rotor
blade. In addition, ANSYS v17.0 provides non-reflective option in outlet, which could be used to
eliminate the effect of wave reflection. Tip clearance flow could be included in the unsteady simulation to
investigate the effects on aerodynamic damping. Different tip gaps can be investigated for unsteady as well.
Many interesting problems can be investigated based on this test case. For the established test case, the
utilization of mixing plane has averaged out all propagations, like wake and shock. But according to
Huang et,al (Huang 2006), the stator flow influences much on the behavior of aeroelastic behavior of LP
rotor in steam turbine and isolated blade row is not recommended for flutter prediction. As a result, multi-
row unsteady calculation could be included in future work.
One of the limit of the test case is that the reflection from stator to rotor cannot be eliminated. More
researched can be conducted to investigate how to apply non-reflecting boundary conditions on the
mixing plane.
For current test case, flutter analysis is made at the designed operating point. But flutter risk is dependent
on steam turbine loading, which happens at the case of off-design point. More work could be done to
study the flutter margin for different operating points.
Panovsky and Kielb (Panovsky 1998) pointed out that the mode shapes are the most important effects on
blade flutter analysis. The test case is using the first bending mode shape and sets hub fixed at the hub.
More mode shapes can be done on the blade to investigate the flutter risk.
The tip shroud is a new design for last stage blades to eliminate the tip clearance flow and provide better
flow performance. Rice et,al (Rice 2009) figured out that the tip shroud could destabilize the blade and be
more prone to flutter. Tip shroud could be added in the geometry and study the effect on flutter.

50
8 Bibliography

ANSYS. 2013.
"Ansys CFX-Solver Modeling Guide."

ANSYS.
"Variables and Predefined Expressions Available in CEL Expressions."
Accessed 08 13, 2016.
https://www.sharcnet.ca/Software/Ansys/16.2.3/en-us/help/cfx_ref/i1304139.html.

Arnone, A. and Benvenuti, E. 1994.

"Three-Dimensional Navier-Stokes Analysis Of A Two-Stage Gas Turbine."
ASME International Gas Turbine and Aeroengine Congress and Exposition. The Hague. ISBN: 978-0-
7918-7883-5

Beevers, A., Congiu, F., Pengue, F. and Mokulys, T. 2010.

"An Analysis of the Merits of CFD for the Performance Prediction of a Low Pressure Steam Turbine Radial Diffuser."
ASME Turbo Expo: Power for Land, Sea, and Air. Glasgow. ISBN: 978-0-7918-4402-1

Burton, Z. 2014.
"Analysis of Low Pressure Steam Turbine Diffuser and Exhaust Hood System."
PhD thesis. Durham Unniversity.

Burton, Z., Ingram, G.L. and Hogg, S. 2014.

"The Influence of Condenser Pressure Variation and Tip Leakage on LP Steam Turbine Exhaust Hood Flows."
Proceedings of the Institution of Mechanical Engineers,228(4): 370-379.

Gardzilewicz, A., Swirydczuk, J., Badur, J. , Karcz, M., Werner, R. and Szyrejko, C. 2003.
"Methodology of CFD computations applied for analysing ow through steam turbine exhaust hoods."
Transactions of the Institute of Fluid-Flow Machinery 113: 157-168.

Guénette, V., Houde, S., Ciocan, G.D., Dumas, G., Huang, J. and Deschênes, C. 2012.
"Numerical prediction of a bulb turbine performance hill chart through RANS simulations."
26th IAHRSymposiumonHydraulicMachineryandSystems.

51
Huang, X. Q., He, L. and Bell, D. L. 2006.
"Influence of upstream stator on the rotor flutter stability in a low pressure steam turbine stage."
Proceedings of the Institution of Mechanical Engineers, Part A: Journal of Power and Energy. 25-35.

Masserey, P.A., McBean, I. and Lorini, H. 2012.

"Analysis and improvement of vibrational behavior on the ND37 A last stage blade."
VGB Power Tech Journal 8.

Panovsky, J. and Kielb, R. E. 1998.

"A Design Method to Prevent Low Pressure Turbine Blade Flutter."
ASME Gas Turbine Conference and Exhibition. Stockholm. ISBN: 978-0-7918-7866-8

Petrie-Repar, P., Fuhrer, C., Grübel, M. and Vogt, D.M. 2015.

"Two-dimensional steam turbine flutter test case."
14th International Symposium on Unsteady Aerodynamics. Stockholm.

Petrie-Repar, P., Makhnov, V., Shabrov, N., Smirnov, E., Galaev, S. and Eliseev, K. 2014.
"Advanced Flutter Analysis Of A Long Shrouded Steam Turbine Blade."
Proceedings of ASME Turbo Expo. Düsseldorf. ISBN: 978-0-7918-4577-6

Rice, T., Bell, D. and Singh, G. 2009.

"Identification of the Stability Margin Between Safe Operation and the Onset of Blade Flutter."
Journal of Turbomachinery 131.

Stein, P., Pfoster, C., Sell, M., Galpin, P. and Hansen, T. 2016.
"Computational Fluid Dynamics Modeling of Low Pressure Steam Turbine Radial Diffuser Flow by Using a Novel
Multiple Mixing Plane Based Coupling— Simulation and Validation."
Journal of Engineering for Gas Turbines and Power 138.

52
Appendix I: Cases description and summary

Steady state cases

Solver Rotor exit Tip clearance Mass flow(kg/s) Power (MW) Total to static efficiency (%)
No 85.14 11.97 88.13
RANS Short 0.5% 85.06 11.61 86.58
1% 85.21 11.62 86.32
RANS 1 chord No 85.15 11.96 88.17
RANS 2 chord No 85.15 11.93 88.55
RANS Long No 85.14 11.68 88.18
No 85.15 12.07 85.43
RANS Long-No splitting 0.5% 85.03 11.7 85.76
1% 85.21 11.57 85.46
Inviscid 1 chord No 85.56 12.26 90.6
Inviscid Long No 85.56 12.26 90.56

Unsteady state cases

Normalised aerodynamic damping
Solver Rotor exit
180 (deg) 90 (deg) 0 (deg) -90 (deg)
RANS Short 0.558 0.784 0.1804 0.0778
RANS Long 0.8387 0.7275 0.4907 -0.0036
Inviscid 1 chord 0.6644 0.527 0.2424 -0.4064
Inviscid Long 0.7125/0.7017 0.6854 0.4/0.4084 -0.1252
Note: Bold number is calculated by time transformation. Otherwise, it is calculated by Fourier transformation.

53
Appendix II: ETC paper
Establishment of an open 3D steam turbine flutter test case
D. Qi - P. Petrie-Repar -T. Gezork
[email protected] , [email protected], [email protected]
(Department of Energy Technology, Royal Institute of Technology, Stockholm, Sweden)
Abstract
This paper introduces an open three-dimensional (3D) flutter test case for steam turbines. The
test case is fully described and initial results are presented. The steam turbine last stage blading
geometry is taken from a test case originally presented by Durham University. The stage is
representative of the aerodynamic characteristics of modern steam turbine blading. To the
authors’ knowledge, this is the first time that a steam turbine flutter test case is presented based
on an open 3D realistic blade geometry. ANSYS CFX and LUFT (Linearized Unsteady Flow
solver for Turbomachinery) were both applied to calculate inviscid and RANS steady and
unsteady flow solutions. 3D non-reflecting boundary condition is applied for LUFT. Plots of
aerodynamic damping versus inter-blade phase angle and plots of the local work coefficient on
the blade for critical cases are presented. There are significant differences in the results
calculated by CFX and LUFT. It is speculated that the main reason for the differences are flow
reflection effects from the mixing plane and the rotor outlet for both steady and unsteady
simulations.
KEYWORDS
Flutter, Steam turbine, Last stage, CFD, aeroelasticity, aeromechanics
Nomenclature
LUFT Linearized Unsteady Flow solver for Turbomachinery
RANS Reynolds-averaged Navier–Stokes equations
w∗ Reduced Frequency
3D Three Dimensional
OEMs Original equipment manufacturers
IBPA Inter Blade Phase Angle
INTRODUCTION
Flutter is a self-excited vibration due to the interaction between the aerodynamic and structural forces.
Usually, the last stage blades are the only part to be concerned with flutter risk in steam turbine (Rice et.al,
2009). These last stage blades are typically over one-meter long and are susceptible to flutter because of
their low structural frequency and supersonic tip speeds. In recent years, blade flutter risk has caught more
attention because of the manufactures interest to increase turbine output and improve performance level
(Rice et.al, 2009). Three-dimensional unsteady inviscid (Masserey et.al, 2012) (Rice et.al, 2009) and
URANS (Stüer et.al, 2008) (Petrie-Repar, 2014) flow simulations for flutter analysis have been performed
previously.
However, the blade profiles and the flow properties presented by OEMs are always protected and kept as
a secret. These results cannot be repeated and compared. Therefore, a flutter test case based on an open
geometry for a steam turbine last stage is defined in this paper. The test case can be used by other
researchers to evaluate and improve their methods.

54
The geometry for this test case is an open resource from Durham University, based on a generic blade,
generated by Alstom Power in Rugby and modified by Durham University (Burton, 2014). The geometry
was used by Durham University to provide representative inflows for exhaust hood and computational
modelling of exhaust hood flows is mainly concerned in their work. It is the first time that a steam turbine
flutter test case is established for an open 3D realistic blade geometry.
TEST CASE DEFINITION
The test case is based on Durham University blade geometry, which is available online (Burton, 2014).
Durham University used the geometry to define representative inflows for modelling exhaust hood flows.
The turbine stage is composed of three parts in this paper, the stator, rotor and diffuser. This is different
from the Durham configuration where the rotor and diffuser were considered to be one part. The
schematic of the stage and boundaries is shown in Figure 84. The stage is representative of the
aerodynamical characteristics of modern steam turbine. The rotor is 0.92m long and highly twisted. The
stagger angle at the tip is about 60 degrees. The rotor speed is 3000 rpm.
The mixing plane is defined for interface between the stator and rotor and that between the rotor and
diffuser to connect domains at different frames. There are three definitions of rotor domain according to
the position of rotor exit. The short rotor domain has the exit at about 0.2 chord downstream of the rotor
trailing edge. The one chord rotor domain exit is extended to one chord. The rotor and diffuser are set as
one domain for long rotor domain. The definition of domains can be seen in Figure 84.

Figure 84: Schematic of stage and boundaries Figure 85: Mode shape

The stator inlet conditions include the representative profiles of total pressure, total temperature and three
velocity components. The inlet boundary conditions are given by Alstom Power from previous multi-stage
calculations and are typical for the last stage of a steam turbine and are available at the Durham website
(Burton, 2014). The average inlet flow conditions are total pressure 27 kPa and total temperature 340 K.
The rotor outlet static pressure is 8800 Pa. The fluid flow in the last stage is wet steam and simply
approximated as ideal gas. The ratio of specific heat and dynamic viscosity is assumed as 1.12 and
1.032·10−5 Pas.
To obtain the blade mode, the rotor blade is assumed to be fixed at a rigid hub and the first flap bending
mode is considered as the mode shape (Figure 85). The fixed hub boundary condition can be used as an
approximation for blade dominated mode shapes when negligible modal coupling through the disk is
present. Pre-stress and spin softening are considered in the modal analysis. The material of blade is set as
titanium alloy and this is typical for a steam turbine blade. The calculated modal frequency is 89.84Hz for

55
the rotation speed of 3000 rpm. Subsequently, the modal frequency was modified to match reduced
frequency of w ∗=0.3. Flutter onset is more likely at low reduce frequency, thus real blades have higher
frequency. The adjusted w ∗ is reasonable since the modified blade geometry from Durham did not
consider the structural mechanics and the calculated modal frequency is too low for real situation. The
frequency can be calculated as:
𝜔𝜔∗ 𝑉𝑉 (1)
f=
2𝜋𝜋𝜋𝜋
where f is modal frequency. c is chord length at mid-span. V is relative velocity at turbine outlet. The
modified modal frequency is 132.08 Hz.
The energy method is employed to perform the flutter analysis. That is the blades are assumed to be tuned
and the aeroelastic eigen modes are travelling wave mode defined by the interblade phase angle. The Inter
Blade Phase Angle (IBPA) is defined as:
2𝜋𝜋𝜋𝜋 (2)
σ=
𝑁𝑁
where n is nodal diameter and N is number of blades. The nodal diameter is the diameter formed by
points which are stationary on a disk in the vibration. It is an integer and defined as:
−N/2 ≤ n ≤ N/2 (3)
Positive n corresponds to forward travelling waves. Negative n corresponds to backward travelling waves.
And positive waves are counted in the velocity direction of the blade row.
To determine if the blade at certain travelling mode is stable, the unsteady work that the flow does to the
blades for the prescribed travelling wave mode is calculated. The non-dimensional aerodynamic damping
is calculated to determine if the rotor is at risk of flutter. It is defined as:
−𝑊𝑊𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎 (4)
Ξ=
𝜋𝜋𝜋𝜋𝛼𝛼 2 𝑐𝑐 2 𝑝𝑝𝑑𝑑𝑑𝑑𝑑𝑑
where 𝛼𝛼 = ℎ 𝑚𝑚𝑚𝑚𝑚𝑚 /𝑐𝑐 and hmax is the maximum blade displacement, c is the chord at the mid-span, b is the
blade height, 𝑝𝑝𝑟𝑟𝑟𝑟𝑟𝑟 is the total pressure at the inlet minus static pressure at inlet. The normalized
aerodynamic damping shows the overall stability of the blade. When aerodynamic damping is negative, the
system is not stable because fluid flow is providing energy which is transmitted to the blade. Otherwise, it
is meaning that vibratory energy of the blade is dissipated by the flow. In order to show the local stability
of the blade, the local work coefficient will be used:
−𝒉𝒉 ∙ 𝑛𝑛�𝑝𝑝𝑖𝑖 (5)
𝑤𝑤 =
𝛼𝛼 2 𝑐𝑐𝑝𝑝𝑑𝑑𝑑𝑑𝑑𝑑
where h is local displacement vector, n� is the local normal vector and 𝑝𝑝𝑖𝑖 is the imaginary component of
the local unsteady pressure.
METHODS
The aim of current study is to define an open 3D test case for realistic steam turbine blade flutter analysis.
The test case is analyzed with two solvers: CFX and LUFT and two flow models: inviscid and RANS. The
LUFT code is capable of performing steady-state flow simulations and linearized unsteady flow
simulations. The code has been validated for flutter analysis (Petrie-Repar et.al, 2007) (Petrie-Repar et.al,
2006) (Petrie-Repar, 2012). The flow model used by the code are the 3D URANS flow equations with the
Spalart and Allmaras turbulence model. The turbulence model is fully linearized for the unsteady flow
simulations. 3D non-reflecting boundary conditions are applied at the inlet and outlet of each domain for
steady and unsteady simulations. In addition, the commercial flow solver ANSYS CFX 17.0 is used as

56
numerical tool. The most common k-ϵ turbulence model is used and Scalable wall functions are used to
calculate the boundary layer flows. 10% turbulence intensity is specified for both solvers.
ANSYS TurboGrid is used as meshing tool. The rotor and diffuser are combined as one domain to
generate the CFD grid. The topology consists of an O-grid for boundaries layers and hexahedral meshes
for other parts. The expansion rate is used to control the number of cells for boundary layers to meet the
requirement of at least 10 cells in the boundary layer for CFX (ANSYS, 2013). The detail of meshes for
the cases are shown in Table 12. As LUFT does not use wall function to resolve boundary layers, smaller
wall cell height is applied.

Table 12: Detail of meshes

Flow model Number of Wall cell height Minimum angle
nodes (mm) (degree)
Stator Inviscid 312564 2 29.8283
Rotor Inviscid 313820 2 38.8137
Diffuser Inviscid 46221 2 60.772
CFX Stator RANS 452704 0.1 29.8088
CFX Rotor RANS 504948 0.1 37.049
CFX Diffuser RANS 91698 0.1 55.514
LUFT Stator RANS 437424 0.02 29.8558
LUFT Rotor RANS 493164 0.02 37.0186
LUFT Diffuser RANS 68324 0.02 55.4224

RESULTS
Steady Computation
Steady-state simulations were performed using the solvers ANSYS CFX and LUFT and using inviscid and
RANS k-ϵ/SA models. Summary of cases and solution is shown in Table 13. RANS simulation was
calculated by short and long rotor case. In order to investigate reflections at the mixing plane, three rotor
meshes with the exit plane at different locations were used. The case setting is same for both CFX and
LUFT code.

Table 13: Summary of steady state cases and solutions

Numeric tool CFX CFX CFX CFX LUFT LUFT LUFT
Flow Equations RANS RANS Inviscid Inviscid RANS Inviscid Inviscid
k-ϵ k-ϵ Laminar Laminar SA Laminar Laminar
Rotor Domain Short Long 1 chord Long Short 1 chord Long
Mass(kg/s) 85.14 85.15 85.56 85.56 85.5 86.05 86.11
Power (MW) 11.97 12.07 12.26 12.26 11.63 11.78 11.79
First cell Y+ 40 40 - - 1.0 - -
Total to static 88.13 85.43 90.599 90.56 85.44 85.72 85.36
efficiency (%)

57
To make numerical results calculated by different codes and different meshes comparable, a standard
definition of span position is provided in the test case. Shroud is defined as 100% span and hub is defined
as 0% span. As hub line is parallel to axis and shroud line is inclined, the span between hub and shroud is
linearly separated according to the slope of shroud line. In other words, x% span is the conical plane that
has a x% slope of shroud line.

Figure 86: Relative Mach number at 50% span Figure 87: Relative Mach number at 90%
（ LUFT） span (LUFT)
CFX and LUFT have similar Mach number contour plot for RANS at 50% and 90% span. There are two
oblique shocks for rotors starting at the trailing edge as shown in Figure 86 and Figure 87. One is the
passage shock to the neighbor blade pressure side and flow after it starts to expand. The other one spreads
to downstream and points to the outlet. This shock deaccelerates the flow at the suction side. The relative
Mach number at the rotor exit is mostly transonic Figure 91. This is the typical flow pattern for a last
stage steam turbine.
The blade loading at 50% and 90% span is shown, Figure 88 and Figure 89. For blade loading at 50%
span, there is an obvious difference between CFX inviscid flow and other cases at the rotor leading edge
suction side. This phenomenon is because for CFX inviscid solution, flow is separated at this position,
Figure 90.

Figure 88: Blade loading at rotor 50% span Figure 89: Blade loading at rotor 90% span

58
Figure 90: Relative Mach number at 50% （ CFX Figure 91: Relative Mach number at rotor exit
inviscid）
Mixing Plane
Mixing planes are used to connect the flow domains between adjacent rows when there is a difference in
the rotational speed of the rows, for example the stator and the rotor. The pitchwise averaged flow
properties are calculated as a function of radius either side of the mixing plane and these pitchwise average
flow properties are used (after adjusting the flow by the relative rotational speed) as boundary conditions
for the adjacent row. In this process, variations in the pitchwise direction are not transferred to adjacent
rows as this would result in an unsteady flow perturbation at the boundary of the adjacent row and it
would not be possible to calculate a steady state solution. Even though these pitchwise variations in the
flow are not transferred to adjacent rows it is important that these variations are treated properly at the
mixing plane to ensure that there are no unphysical flow reflections. The LUFT code applies a steady non-
reflecting boundary condition at the mixing plane to ensure that these pitchwise variations are allowed to
exit the domain without reflection. In order to examine the flow reflections at the mixing plane, the first
harmonic of the pressure in the pitchwise direction from the inviscid flow simulations at the location one
chord downstream from the rotor trailing edge was examined. In Figure 92 the amplitude of the pressure
harmonic is shown from LUFT and CFX solutions calculated with the one chord rotor and the long rotor
at the location of the exit of the one chord rotor. The solutions for the one chord rotor were taken
directly from the rotor exit while the solution from the two chord rotor is extracted at the location of the
exit plane of the one chord rotor. This component of the flow is not transferred to the downstream
diffuser domain. It can be seen that there is a good agreement between the LUFT solutions at this
location. This suggests that the non-reflecting boundary condition is working well at the mixing plane.
However, the CFX solutions differ but there is a reasonable agreement between the CFX solution
extracted from the long rotor mesh and the LUFT solutions. This suggests that the harmonic component
predicted by the LUFT solutions is correct. There is probably a steady flow reflection at the mixing plane
within the CFX one chord simulation. This steady flow reflection can travel upstream and influence the
flow at the blade. This probably explains the difference between the blading loadings on the suction side
near the trailing edge calculated by CFX in Figure 89. This is a common problem for transonic flows
(Arnone and Benvenuti, 1994).

59
Figure 92: Amplitude of first harmonic pressure Figure 93: Schlieren plots for CFX short rotor
at the mixing plane at 50% span

The steady flow reflections can be seen clearly from Schlieren plots of CFX short case at rotor exit
(Figure 93). When rotor and diffuser are set as one domain, the flow field was not constrained and the
reflection effects did not appear (Figure 94). It appears that the LUFT solution is not affected by flow
reflections from the mixing plane (Figure 95). Flow reflections at the stator exit within the CFX solution
can also be seen from the Schlieren plots.

Figure 94: Schlieren plots for CFX long RANS Figure 95: Schlieren plots for LUFT short
case at 50% span RANS case at 50% span

Unsteady Computation
CFX and linearized flow solver LUFT were used to calculate the rotor unsteady flow for different
travelling wave modes. These simulations only included the rotor domain to save computational time and
resources. Inlet and outlet boundary profiles from multi-row steady state results were extracted from the
rotor domain to set the flow boundary conditions for the unsteady flow simulations. Steady state
simulations for the single rotor with extracted boundary conditions were performed to set up the unsteady
simulations. However, when boundary profiles were extracted directly from mixing plane, there was a
difference in the blade loading from the single rotor steady results at the leading edge, as shown in Figure
96. It was speculated that the flow at the mixing plane is affected by flow reflections. Modified boundary
inlet flow conditions were extracted from 2% downstream the mixing plane instead and mass
circumferential averaged quantities are applied from hub to shroud, which is still far from rotor blade.

60
This provided a good agreement for blade loading as seen in Figure 96, which eliminates the effects from
different flow field of multi and single row steady state.

Figure 96: Blade loading comparison for single Figure 97:Normalized aerodynamic damping
short rotor at 50% span comparison
The aerodynamic damping calculated for the RANS short, inviscid flow one chord and long cases with
CFX and LUFT are shown in Figure 97. The CFX Fourier transformation method is used to reduce
computational time. Unfortunately, it was not possible to calculate CFX solutions for +/- 90 degrees with
the Fourier transformation method, however, it was possible to obtain CFX solutions in the time domain.
The unsteady solutions for 0 and 180 degrees for both methods agreed well. The CFX solutions were only
calculated at four ibpas while the LUFT solutions were calculated are possible 65 ibpas. All LUFT
solutions were stable and only two CFX solutions were unstable. The aerodynamic damping was also
calculated at the blade natural frequency (89.84 Hz) with LUFT for inviscid flow on the one chord
domain. Although the damping values are lower the blade row is still stable.
The biggest discrepancy between the LUFT and CFX solutions in Figure 97 occurs at -90 degrees.
Figure 98 shows a plot of damping versus span for the various solutions. It can be seen that most of the
aerodynamic work is done near the tip and there is a clear difference between the CFX and LUFT
solution. To examine this difference in more detail, the local work coefficient on the blade at 90% span at
-90 degrees is examined in Figure 99. Positive local work indicated that energy is added to the blade and
this is unstable. There is a good agreement between the LUFT and CFX solutions on the pressure side
however there are large differences in the solutions on the suction side. The LUFT solutions are
predicting stable work on the suction side while the CFX solutions are predicting unstable work on the
suction side. These differences are probably due to unsteady flow reflections from the rotor exit where the
flow is transonic. The LUFT solver does apply an exact 3D non-reflecting boundary condition at the inlet
and outlet. It can be seen that the LUFT solutions are independent of the location of the rotor exit,
however, the CFX solution are dependent on the exit location.

Figure 98: Local wall work density versus span Figure 99：Normalised local work coefficient
for IBPA -90 degree for IBPA -90 degree at 90% span

61
DISCUSSION
At this stage, solutions for this new test case have only been calculated using two different methods and
there are some major differences between the solutions. Comparisons with solutions from other methods
would help to develop a consensus what is the correct solutions. The authors are planning to set up a
website at KTH where a full description of the test case can be found and also the meshes used in this
paper will be available for downloading. It is hoped that by making this data publically available that other
researchers will be encouraged to calculate solutions for the test case which they will also share with the
scientific community.
The test case could also be extended to include other features which are important for steam turbine
blades such as the influence of shrouded blades and the effects of tip clearance. The test case could also
be used to explore how various parameters affect the flutter stability for a 3D representative blade such as
diffuser exit pressure, mode shape and reduced frequency. This test case could also be used to study the
influence of multi-row effects on flutter as the stator row is included.
The nature blade frequency was calculated to be 89.84 Hz. This was increased to 132.08 Hz for the initial
unsteady flow simulations to achieve a reduced frequency of 0.3 base on full chord as it was thought that
this is more representative of actual steam turbines. However, the LUFT code predicted that all travelling
wave modes were stable at this frequency. Unsteady flow simulations were also performed at 89.84 Hz
and all these cases were also stable. Some modern steam turbines can be significantly longer and have
higher stagger angles at the tip than the current test case. It would also be interesting to have another open
test case for a longer blade, however, the authors are not aware of any open geometry for a longer blade.
CONCLUSION
A 3D steam turbine flutter test case has been established. The test case includes many features that are
important for steam turbine flutter calculations, such as a 3D long twisted blade, transonic flow at the
rotor exit, high stagger angle near the tip, mixing planes close to the blade and a blade with low natural
frequency.
Initial results from two CFD methods were presented and there were major differences in the calculated
solutions. It is suspected that the reason for the differences are steady and unsteady flow reflections from
the mixing planes and the rotor outlet. There is a need for more solutions to be calculated for this test
case in order to determine a consensus on the correct solution.
It appears that the LUFT solution is better for this test case as its solutions are not affected by unphysical
flow reflections from the mixing planes or the rotor outlet for the unsteady flow simulations.
ACKNOWLEDGEMENTS
The financial support from China Scholarship Council (CSC) for first author is acknowledged.
REFERENCE
ANSYS. (2013). ANSYS CFX-Solver Modeling Guide. Published in the U.S.A
Arnone, A., Benvenuti, E., (1994). Three-dimensional Navier-Stokes analysis of a two-stage gas turbine. Proceedings
of the ASME 1994 International Gas Turbine and Aeroengine Congress and Exposition. Hague,
Netherlands.
Burton, Z., (2014). Analysis of Low Pressure Steam Turbine Diffuser and Exhaust Hood Systems. PhD thesis,
Durham University.
Masserey, P. A., McBean, I. and Lorini, H. (2012). Analysis and improvement of vibrational behavior on the ND37
A last stage blade. VGB Power Tech Journal (8).

62
Petrie-Repar, P. J. (2012). Establishment of a steam turbine flutter test case. Proceedings of 13th International
Symposium on Unsteady Aerodynamics, Aeroacosutics and Aeroelasticity of Turbomachinery, Tokyo,
Japan.
Petrie-Repar, P. J., McGhee, A. M. and Jacobs, P. A. (2007). Three-dimensional viscous flutter analysis of standard
configuration 10. Proceedings of ASME Turbo Expo 2007: Power for Land, Sea, and Air, Montreal, Canada.
Petrie-Repar, P. J., McGhee, A. M., Jacobs, P. A., and Gollan, R. (2006). Analytical maps of aerodynamic
damping as a function of operating condition for a compressor profile. Proceedings of ASME Turbo Expo 2006:
Power for Land, Sea, and Air, Barcelona, Spain.
Petrie-Repar, P.J., Makhnov, V., Shabrov, N. Smirnov, E. Galaev, S. and Eliseev, K. (2014). Advanced
Flutter Analysis of A Long Shrouded Steam Turbine Blade. Proceedings of ASME Turbo Expo 2014: Turbine
Technical Conference and Exposition, Düsseldorf, Germany.
Rice, T., Bell, D. and Singh, G., (2009). Identification of the Stability Margin Between Safe Operation and the Onset
of Blade Flutter. ASME. J. Turbomach. 131(1):011009.
Stüer, H., Schmitt, S., and Ashcroft, G. (2008). Aerodynamic mistuning of structurally coupled blades. Proceedings
of ASME Turbo Expo 2008: Power for Land, Sea and Air, Berlin, Germany.

63