FINITE ELEMENTS FOR THE ANALYSIS OF

ROTOR-DYNAMIC SYSTEMS THAT INCLUDE

GYROSCOPIC EFFECTS

A thesis submitted for the degreeof Doctor of Philosophy

By

Simon Jones

Department of Mechanical Engineering, Brunel University

April 2005
 l

ABSTRACT

This thesis presents new finite element formulations for the analysis of rotor-
dynamic systemsthat include the effects of gyroscopic influence. Euler-Bernoulli
finite elements have been created for both shaft and propeller descriptions. In

addition to the gyroscopic effects, centrifugal stiffening has been considered for
the propeller elements. The principle of virtual work has been used to determine
the equationsof motion and formulate element matrices.

The proposed element matrices have been incorporated in the VIBRATIO suite
of vibration analysis software in order to test the formulations. The software uses

an innovative hybrid modelling technique that enables the user to analyse various
dynamic problems including rotating beam elements with rigid body

attachments.

A model of a ship's drive shaft has been created in VIBRATIO for comparison

against a verified ANSYS model. Results for forced vibration shaft analysis
show excellent correlation between VIBRATIO's Euler shaft formulation and
ANSYS's Timoshenko formulation.

Incremental analyses of propeller systems using the novel gyroscopic

formulation show gyroscopic effects of flexible blade attachments, and also the

changing mode shapes and frequencies due to centrifugal stiffening. Results

show gyroscopic and centrifugal stiffening effects must not be ignored for an
accuratepropeller analysis.
 J
n--na

TABLE OF CONTENTS

2
ABSTRACT ..........................................................................................................

TABLE OF CONTENTS 3
.....................................................................................

NOMENCLATURE 7
.............................................................................................
SYMBOLS 7
.................................................................................................................................
SUBSCRIPTS 8
............................................................................................................................

9
LIST OF FIGURES .............................................................................................
FIGURES 9
...................................................................................................................................
TABLES 9
.....................................................................................................................................
GRAPHS 9
....................................................................................................................................

ACKNOWLEDGEMENT 12
.................................................................................

CHAPTER 1 13
.......................................................................................................

13
INTRODUCTION .......................................................................................................................

CHAPTER 2 17
.......................................................................................................

LITERATURE REVIEW 17
...........................................................................................................
2.1 INTRODUCTION 17
..............................................................................................................
2.2 FLEXIBLE SHAFT MODELLING 17
...................................................................................
2.3 FLEXIBLE PROPELLER BLADE & DISC MODELLING 25
.............................................
2.4 THE AIMS OF CURRENT RESEARCH 35
..........................................................................
 4
.. u.. t..III3

CHAPTER 3 36
.......................................................................................................

FINITE ELEMENT HYBRID MODELLING OF FLEXIBLE SHAFTING SYSTEMS 36

....
3.1 INTRODUCTION 36
..............................................................................................................
3.2 THE FINITE ELEMENT METHOD 36
.................................................................................
3.3 RIGID BODY THEORY 38
...................................................................................................
3.3.1 Definitions and Assumptions 38
......................................................................................
3.3.2 Equations of Motion 38
...................................................................................................
3.3.3 Stiffness and Damping Systems 39
.................................................................................
3.3.4 Generalisation of the Equations of Linear Momentum 40
...............................................
3.3.5Generalisation
of theEquationsof Angular Momentum............................................
41
3.4 ASSEMBLY OF EQUATIONS 42
.........................................................................................
3.4.1 Force - Acceleration Equations 42
.................................................................................
.
3.4.2 Moment Equations 45
.....................................................................................................
.
3.5 THE FEA FORMULATION FOR BEAMS 46
.....................................................................
.
3.5.1 Stiffness Matrices 46
......................................................................................................
.
3.5.2 Mass Matrices 47
............................................................................................................
.
3.5.3 Co-ordinate Transformation for the Built-In Condition 49
............................................. .
3.5.4 Co-ordinate Transformation for a Pin-Jointed Connection 51
.........................................
3.6 PROGRAMMING CONSIDERATIONS 52
..........................................................................
3.7 SUMMARY 53
.......................................................................................................................

CHAPTER 4 55
.......................................................................................................

SHAFT GYROSCOPIC BEAM ELEMENT MATRIX DERIVATION 55

...............................
4.1 INTRODUCTION 55
..............................................................................................................
4.2 THEORY 56
............................................................................................................................
4.2.1 Beam Elements (Three Dimensional) 56
.........................................................................
4.2.2 Equations of Motion 57
...................................................................................................
4.2.3 Gyroscopic Beam Element Matrix Derivation 58
............................................................
4.2.4 Beam Element Matrix 63
.................................................................................................
4.3 ANSYS GYROSCOPIC DAMPING MATRIX (ANSYS Theory) 64
...................................
4.4 SUMMARY 65
.......................................................................................................................

CHAPTER 5 66
.......................................................................................................

GYROSCOPIC PROPELLER ELEMENT MATRIX DERIVATION 66

.................................
5.1 INTRODUCTION
66
..............................................................................................................
 J

5.2 THEORY 66
............................................................................................................................
5.2.1 Equations of Motion 66
...................................................................................................
5.2.2 Gyroscopic Propeller Element Derivation 67
..................................................................
5.2.3 Gyroscopic Propeller Element Matrix 72
........................................................................
5.3 SUMMARY 73
.......................................................................................................................

CHAPTER 6 75
.......................................................................................................

CENTRIFUGAL STIFFENING OF PROPELLER ELEMENT MATRIX DERIVATION75

6.1 INTRODUCTION 75
..............................................................................................................
6.2 THEORY 76
............................................................................................................................
6.2.1 Equations of Motion 76
...................................................................................................
6.2.2 Centrifugal Stiffening Propeller Element Derivation 76
.................................................
6.2.3 Centrifugal Stiffening Propeller Element Matrix 80
........................................................
6.3 SUMMARY 81
...................................................................................................................

CHAPTER 7 82
.......................................................................................................

RESULTS 1: GYROSCOPIC SHAFT ELEMENT 82

.................................................................
7.1 INTRODUCTION 82
..............................................................................................................
7.2 LLOYD'S SHIP DRIVE SHAFT 83
.......................................................................................
7.3 BEAM ELEMENT RESULTS 85
...........................................................................................
7.4 SUMMARY 89
.......................................................................................................................

CHAPTER 8 91
.......................................................................................................

RESULTS 2: GYROSCOPIC PROPELLER ELEMENT 91

......................................................
8.1 INTRODUCTION 91
..............................................................................................................
8.2 EIGENVALUE RESULTS 91
................................................................................................
8.3 FORCED FREQUENCY RESULTS 96
.................................................................................
8.4 SUMMARY 104
.....................................................................................................................
Table of Contents 6

CHAPTER 9 106
.....................................................................................................

RESULTS 3: CENTRIFUGAL STIFFENING ELEMENT 106

..................................................
9.1 INTRODUCTION 106
............................................................................................................
9.2 EIGENVALUE RESULTS 107
..............................................................................................
9.3 FORCED FREQUENCY RESULTS 111
...............................................................................
9.4 COMBINED GYROSCOPIC & CENTRIFUGAL STIFFENING FORCED
FREQUENCY RESULTS 116
......................................................................................................
9.5 SUMMARY 119
.....................................................................................................................

CHAPTER 10 122
...................................................................................................

DISCUSSIONS & CONCLUSIONS 122

........................................................................................
10.1 DISCUSSIONS 122
..............................................................................................................
10.2 CONCLUSIONS 126
............................................................................................................
10.3 FURTHER WORK 128
.........................................................................................................

REFERENCES 130
.................................................................................................

APPENDIX A 139
...................................................................................................

ELEMENT MATRICES 139

..........................................................................................................
A. 1 Mass Element Bending Matrix 141
........................................................................................
A. 2 Mass Element Axial Deformation Matrix 141
.......................................................................
A. 3 Mass Element Torsional Deformation Matrix 142
.................................................................
A. 4 Stiffness Element Bending Matrix 142
...................................................................................
A. 5 Stiffness Element Axial Deformation Matrix 143
..................................................................
A. 6 Stiffness Element Torsional Deformation Matrix 143
............................................................
A. 7 Gyroscopic Element Bending Matrix for Shaft 144
...............................................................
A. 8 Gyroscopic Element Bending Matrix for Propeller 144
.........................................................
A. 9 Centrifugal Stiffening Element Matrix for Propeller 145
.......................................................
Nomenclature 7

NOMENCLATURE

SYMBOLS
A cross sectional area m2
As shear area (As =A/ 2) m2
C global damping matrix N-s/m
Ce elemental damping matrix N-s/m
E Young's modulus N/m2
F global external force vector N
Fe elemental external force vector N
J
J global inertia matrix kg m2
G shearmodulus N/m2
G global gyroscopic matrix
Ge elemental gyroscopic matrix
Ge elemental propeller gyroscopic matrix
v
I secondmoment of area m4
K global stiffness matrix N/m
Ke elemental stiffness matrix N/m
Ke elemental propeller stiffness matrix N/m
v

Ke': elemental propeller centrifugal stiffening matrix N/m

I element length m
M global mass matrix kg
Me elemental mass matrix kg
Mep elemental propeller massmatrix kg

N=0 Note: N is used to introduce the gyroscopic cross-coupling terms.

r radius M.
s radial distance to elemental massunder consideration m
u shaft deflections of geometric centre m
uM shaft deflections of the masscentre m
lie shaft deflections of a beam element M.
V(x) matrix of elementshapefunctions
x, y, z local coordinates for element m
Nomenclature 8

xg, yg,: ýg global coordinate axes with origin at root of beam (Fig 5.1) m
X, Y, Z global coordinates m

a rotation about the x axis rad

13 rotation about the y axis rad
y rotation about the z axis rad

9P polar mass moment of inertia per unit length kg"m

p mass per unit length kg/m

p mass density kg/m3

S2 shaft rotational velocity rad/s

Ö= a
and O=
at ax

SUBSCRIPTS
G gyroscopic
e element
p propeller
magn magnetic
List of Figures 9

LIST OF FIGURES

FIGURES
Figure 3.1: SchematicRepresentationof a Multi-Body System
Figure 3.2: Bodies i andj Connectedby Spring k,
Figure 4.1: ShapeFunctions
Figure 4.2: Shaft Definitions
Figure 4.3: Shape Function Matrix Geometry
Figure 5.1: Local Axis System

Figure 6.1: Geometry

Figure 6.2: Element Location

Figure 7.1: Lloyd's Drive Shaft Model

Figure 8.1: Rectangular Cross-Section of a Beam Element
Figure 10.1: Gyroscopic Disc Example

TABLES
Table 8.1: Eigenfrequencies for Standard Euler Beams

Table 8.2: Eigenfrequenciesfor SquareGyroscopic Euler Propeller

Table 8.3: Gyroscopic Eigenfrequenciesfor Rectangular Euler Propeller
Table 8.4: Gyroscopic Eigenfrequenciesfor Rectangular Euler Propeller of
Varying Width
Table 8.5: First Three Static Propeller Eigenfrequencies
Table 9.1: Eigenfrequency Comparison with Results from Wright et al., [1982]
Table 9.2: Euler Propeller EigenfrequenciesIncluding Centrifugal Stiffening

GRAPHS
Graph 7.1: Axial X Displacement Of Shaft End
List of Figures 10

Graph 7.2: Vertical Y Displacement Of Shaft End

Graph 7.3: Horizontal Z Displacement Of Shaft End

Graph 7.4: Rotational Alpha Displacement Of Shaft End

Graph 7.5: Rotational Beta Displacement Of Shaft End

Graph 7.6: Rotational Gamma Displacement Of Shaft End

Graph 8.1: Y Translation Of Propeller Tip (S. R. 50)

Graph 8.2: Z Translation Of Propeller Tip (S. R. 50)

Graph 8.3: ß Rotation Of Propeller Tip (S.R. 50)

Graph 8.4: y Rotation Of Propeller Tip (S.R. 50)

Graph 8.5: Y Translation Of Steel Propeller Tip
Graph 8.6: Z Translation Of Steel Propeller Tip
Graph 8.7: ß Rotation Of Steel Propeller Tip
Graph 8.8: y Rotation Of Steel Propeller Tip
Graph 8.9: Y Translation Of Aluminium Propeller Tip
Graph 8.10: Z Translation Of Aluminium Propeller Tip

Graph 8.11: ß Rotation Of Aluminium Propeller Tip

Graph 8.12: y Rotation Of Aluminium Propeller Tip

Graph 9.1: PercentageChangein Eigenfrequencies(S.R. 100)

Graph 9.2: PercentageChangein Eigenfrequencies(S.R. 50)
Graph 9.3: PercentageChange in Eigenfrequencies(S.R. 25)
Graph 9.4: PercentageChangein Eigenfrequencies(S.R. 10)
Graph 9.5: Y Translation Of Propeller Tip (S.R. 100)
Graph 9.6: y Rotation Of Propeller Tip (S.R. 100)
Graph 9.7: Y Translation Of Propeller Tip (S.R. 50)
Graph 9.8: y Rotation Of Propeller Tip (S.R. 50)
Graph 9.9: Y Translation Of Propeller Tip (S.R. 25)
Graph 9.10: y Rotation Of Propeller Tip (S.R. 25)
Graph 9.11: Y Translation Of Propeller Tip (S.R. 10)
Graph 9.12: y Rotation Of Propeller Tip (S.R. 10)
Graph 9.13: Y Translation Of Aluminium Propeller Tip
Graph 9.14: 7 Rotation Of Aluminium Propeller Tip
Graph 9.15: Y Translation Of Aluminium Propeller Tip
List of Figures 11

Graph 9.16: Z Translation Of Aluminium Propeller Tip

Graph 9.17: ß Rotation Of Aluminium Propeller Tip

Graph 9.18: y Rotation Of Aluminium Propeller Tip
Acknowledgement 12

ACKNOWLEDGEMENT

The author would like to expresshis indebted thanks to the following people and

organisationswithout which this researchwould not have been possible:

" Prof. I. Esat, for his encouragement, supervision and advice throughout
the duration of this research.

" Brunel University for sponsoring this research and providing the

necessaryresources.
" My family and friends who have patiently awaited the completion of this

work, providing encouragementthroughout.

Chapter 1: Introduction 13

CHAPTER I

INTRODUCTION

Flexible mechanical systems are gaining an increasing significance as more

large-scale machinery is being built. The majority of modern machines
incorporate some form of flexible rotor-bearing system in order to control and
distribute mechanical power and many of these include propeller attachments.

Examples include wind turbines, marine propulsion systems and turbo-

machinery. Current design trends for rotating equipment aim to heighten

efficiency by reducing weight and increasing operating speeds. These goals are
being made more attainable by a greater understanding of rotor-dynamic
behaviour and improved methods for predicting system responses. The ground

theory of mechanisms and machines dates back to the early twentieth century,
but the dynamic analysis of flexible mechanisms is often too complex for an

analytical solution. Thus it is with the increased power of modern computers that
new methods of behaviour prediction are being developed for flexible

mechanisms.

Vibration analysis is essential in the design and analysis of rotating machinery.

The majority of vibrations are caused by rotation related sources of some
description, (normally imbalance) consequently the forces are synchronous to the

rotational speed.Thus, forced vibration analysis is fundamental in the design and

analysis of rotating machinery. The two most commonly used methods of forced
frequency analysis are the finite element, or transfer matrix method. However

since certain effects (including, gyroscopic, centrifugal stiffening, and fluid

bearings) are dependant on the rotational speed the methods require

computational assembly and inversion of large matrices at each frequency step.

Chapter 1: Introduction 14

This is computationally expensive and inefficient however with the advancesin

modern computing speedit is rapidly becoming less of a problem.

The transfer matrix approach allows for a continuous representation of the shaft

system and produces results in good agreement to experimental work. Its main
advantage is the small amount of computer memory and power required to
analyse systems.However, the equations of motion are not explicitly written and
some experimental work is usually required in obtaining the transfer matrices.
Therefore, as computers have become exponentially more powerful, finite

element methods are now largely replacing those basedon transfer matrices. This
is especially true during initial system design stageswhen transfer matrices may
be difficult to verify.

The finite element method provides a methodical approach for the discretization

of a continuum. It can provide a solution for many types of complicated systems

including fluid flows, heat exchange, static mechanical stresses, or dynamic

mechanical systems, including those examined in this study. In each case the
system is divided into smaller contacting regions known as elements which are
described mathematically. For rotor-dynamic systems finite element models aim

to describe them mathematically using a system of differential equations. As with

all finite element modelling this can only ever be an approximation, although
potentially a very good one. Subsequently engineers seek to simplify systems,
making assumptions that reduce the mathematical complexity of models, ideally
without compromising the results. Two such common simplifications are the
omission of gyroscopic effects, and the disregard for the exact geometry of

propeller blades, which are instead treated as attached rigid or flexible discs with
the same value of inertia.

There have been analyses performed on systems that incorporate flexible blade

arrays (propellers) however these are usually small blades attached to the
periphery of a disc. This is becausetraditionally the majority of research in the
field of rotor-dynamic shaft analysis is based around gas turbines. This is due to

the abundant use of gas turbines in modern industry and the inherent resonance
Chanter 1: Introduction 15

problems associated with their design. Accordingly, there are considerable

financial gains to be made from improvements in gas turbine design, especially
those that increase the efficiency of large power stations. This research has

resulted in most modem gas turbines operating beyond their first and second
order rigid body critical speeds. Further advancements are pushing operating
frequencies towards even higher speedsapproaching the bending modes of the

rotating group. This has been facilitated by improved balancing techniques,

bearing designs and computer simulations. Moreover, designers face demandsto

achieve these improvements whilst reducing both the initial and operating costs.
Hence, when researchis performed on propeller attachments,it is mostly done on

systems that best represent a gas turbine rotor. Other analyses tend to focus on
the behaviour of helicopter blades. These are very lightweight and so in
consideration of this researchersnormally do not include gyroscopic effects of
the blade, assuming them to be negligible, but are more concerned with the
effects of the blade tip masses.The need for a model that takes into account the
interaction between the shaft and blades is paramount to describe systemssuch as

wind turbines or ship propellers.

With the abundant availability of fast computers it is now cost effective to apply
finite element techniques to this broader range of problems, where traditionally it

was perhaps more cost effective to solve them in an empirical, heuristic fashion.
Producing a universal code to encompass all possible dynamical systems is

arguably the ultimate goal for a dynamic finite element analysis package.
However, the factors governing the behaviour of various dynamical systems are

so copious and interrelated that the task might appear almost impossible.
Additionally a computer program capable of considering each and every aspect

of all systems no matter how small the consequencewould be computationally

expensive. One solution is to use modules within the code so that various types
of problem can be analysed in isolation. However, all these separate analysis
modules still have to be developed and written.

The aim of the present investigation is to develop a reliable computational

technique for the analysis of the dynamic responseof flexible shaft and
propeller
Chapter 1: Introduction 16

systems, paying particular attention to the often-ignored subject of gyroscopic

response. The modelling of shaft and shaft-propeller systems has been achieved

using the finite element method (FEM). Firstly, a gyroscopic finite element

capable of describing shaft systems has been derived. The derivation principles

used have been further developed and extended in order to create a finite element
capable of modelling and predicting the response of rotating propeller blades,
including the effects of gyroscopic moments. This new approach for describing

the behaviour of propeller elements using an Euler-Bernoulli finite element that

incorporates gyroscopic behaviour in this manner has not previously been

reported. The effect of centrifugal stiffening on propeller blades has also been

considered in the analysis. This is due to the possible size and speed of some
blades which could be analysed in a more universal code. Finally, the developed

elements have been incorporated into a hybrid finite element model in order that
they can be verified.
Chapter 2: Literature Review 17

CHAPTER 2

LITERATURE REVIEW

2.1 INTRODUCTION
This chapter presents a survey of key research conducted in the field of the
dynamics of flexible shafts and propellers with respect to finite element

modelling. For dynamic shaft or propeller systems finite element code usually
falls into one of two categories using either Euler-Bernoulli beam theory, which
ignores the effect of shear stress, or Timoshenko theory, which includes the

shear effect. Although both methods are very similar, and in some papers they
are treated as almost interchangeablewith shear effects being added to classical
Euler beam theory, in general the Timoshenko method is considered to be

superior for thick shafts. Section 2.2 presentsthe researchwork conducted on the
modelling of flexible shafts, in particular that which incorporates the effects of
gyroscopic behaviour. Section 2.3 reviews the researchperformed on the various
methods of modelling propeller blades as flexible beam elements, once again
focusing on work that includes any gyroscopic behaviour. This chapter

concludes in section 2.4 by describing the aims and scope of the present research
against the background of previous research.

2.2 FLEXIBLE SHAFT MODELLING

Much of the early research into the finite element computer analysis of rotor-
dynamic systems was developed with the intent of analysing turbo-machinery in

power plants. The abundant use of turbo-machinery creates considerable

financial gains if benefits in both weight and performance are found,
especially
when this enables operation at supercritical speeds. For this section on shaft
Chanter 2: Literature Review 18

modelling it is the gyroscopic effect that is of main concern due its tendency to
be omitted from many finite element models. This is of particular interest for

system types where the rotational inertia or speed is great enough that its

inclusion is significant. Also of interest are various possible methods of

combining finite elements into a computer code that is flexible enough to analyse
different problem types.

The theory of beams and vibrations was developed in the late 1800's and early
1900's. However, the large number of equations that neededto be solved limited
its application to complex programs. Therefore, research into finite element

modelling of dynamic systemsdidn't really begin to flourish until the 1960's and
1970's, which coincided with the availability of modem computers on which to

calculate results. Rankine, W. A., [1879] published the first recorded attempt to
discuss the nature of a rotor-dynamic shaft. Neglecting the Coriolis acceleration
for a uniform frictionless shaft, he concluded that the shaft should be stable
below its first critical speed, in neutral stability, or "indifferent equilibrium" at
the critical speed and unstable above it. Jeffcott, H. H., [1919] investigated the
effect of unbalance on rotating shafts and, unlike Rankine, allowed for the
possibility of successful operation above the first critical speed. By the early
1920's some rotors were operating above their first critical speeds,but it was yet
to be understood why some rotors could operate at these speeds while other
designs would fail. Over the coming years much research was performed in the

study of whirling, oil whip, unbalance, shafts and bearings, whilst trying to
explain observedphenomenon.

Green, R.B., [1948] was one of the first to investigate the gyroscopic effects on
the critical speedsof flexible rotor-disc systemsusing analytical techniques. Due
to the difficulties associatedwith calculating analytical solutions, many graphical

plots to aid the design of early rotor-disc systems were created. These were for
cantilever, simply supported single and dual disc systems and infinite-disc
systems. The paper shows the gyroscopic effect for all degrees of freedom and
possible ratios of shaft rotating frequency to whirl or precessionfrequency.
Chanter 2: Literature Review 19

With the advancements in modern computing Kapur, K. K., [1966], Morton,

P.G., [1968], and Ruhl, R. L., [1970] published the earliest concise methods of

partial differential equations for finite element analysis. Ruhl, R. L., and Booker,

J.F., [1972] took this work further, producing an Euler-Bernoulli finite element

model for a turbo-rotor system with the provision for a rigid disc attachment.
The finite element method developed by Ruhl was used to study rotor stability

and unbalance response, but only the effects of elastic bending energy and

translational kinetic energy were considered. The work by Ruhl was later

improved upon by Nelson, H. D., and McVaugh, J.M., [1976], including the

effects of rotary inertia, gyroscopic moments and axial load, for disc-shaft

Later, Zorzi, E. S., and Nelson, H. D., [1977] included the effects of
systems.
internal damping to the beam elements. Zorzi, E. S., and Nelson, H. D., [1980]

culminate their work by modelling the effect of constant axial torque on the

model, producing a corresponding axial stiffness matrix. Although constant

torque may be suitable for a gas turbine shaft, it must be noted that this is not

sufficient to model an internal combustion engine shaft, for example, for which a
time dependant torque will be required.

Davis, R., Henshell, R.D., and Warburton, G.B., [1972], wrote one of the first

early works on Timoshenko finite beam elements for rotor-dynamic analysis.

Previous to this Timoshenko beams had mostly been used to analyse vibrations
in non-rotating structures. The work was performed becauseof the tendency for
Euler-Bernoulli beam theory to overestimate the frequency of vibration. It was

noted that this overestimation increasedwith the ratio of beam depth to vibration

wavelength. The paper derives matrices for a Timoshenko beam element, sets
out boundary conditions and performs convergence tests. Finally, there is some
discussion over the shear coefficient K, which is often a point of contention in
the analysis of Timoshenko beams. For a more detailed derivation of the
Timoshenko shear coefficient in relation to various material cross sections the

author refers to Cowper, G.R., [1966] and Timoshenko, S.P., [1922].

Thomas, D. L., Wilson, J.M., and Wilson R.R., [1973] also published early work

on tapered Timoshenko finite beam elements. They produced an element with

Chanter 2: Literature Review 20

three degrees of freedom at each node, comparing it to earlier elements

containing only two. This allowed for translational motion, shear deformation
and the rotation of the cross-section.The work concluded by suggestingthat it is
in fact the simplest Timoshenko beam elements (i. e. those concerned only with
bending) that are most useful for a general finite element code. However, for the

analysis of structures in which shear and rotary inertias are important, the
presentedelement will render better results.

Thomas, J., and Abbas, B. A. H., [1975] presented a Timoshenko finite element
that enabled the user to apply geometric, boundary, and force conditions
correctly, thus giving an accurate representation of a Timoshenko beam. The
author claims to produce a model that gives far superior results than those given
by Kapur, K., [1966] and Davis, R., Henshell, R.D., and Warburton, G.B.,
[1972].

Nelson, H. D., [1980] continues the development of finite elements for rotor-
dynamic analysis using Timoshenko beam theory. The paper generalizes

previous works, establishing Timoshenko matrices using the shape functions

proposed in Nelson, H. D., [1977]. The model includes translational and
rotational inertia, gyroscopic moments, bending and shear deformation, as well

as axial loads, but ignores internal damping. It is shown that accuracy improves
with the number of finite elements used. The methods used can easily be
incorporated into existing computer programs using Euler beam theory with

negligible increases in computation time and storage requirements.

Gasch, R., [1976] presented a concise work on the modelling of vibrations in

large turbo rotors using the finite element method. The aim of the work
was to
produce a much more accurate analysis of large turbo rotors than that provided
by transfer matrix methods. Gasch uses the principle of virtual
work to
determine the Euler beam element matrices. His formulation includes
provision
for gyroscopic forces, internal damping, fluid film forces and
unbalance response
terms amongst others in a thorough analysis. However, it does not consider the

effects of propeller blades, instead treating the rotors as rigid discs, which for
Chapter2: LiteratureReview 21

this application is acceptable. Gasch also reduces computational effort by using a

banded system of matrices that allows the large number of zeros in the global

system matrix to be ignored.

Rouch, K. E., and Kao, J.S. [19791 developed stiffness and mass matrices for a

tapered beam finite element, additionally including gyroscopic effects by means

of a skew-symmetric damping matrix. Nodal end variables were used to include

the shear deflection, and rotary inertia effects were included in the energy

function, effectively creating a Timoshenko beam. Finally, a reduction of the

shear nodal variables was performed prior to global assembly in order to reduce

computation time. The accuracy of the approach was verified through analysis of

simple cantilever beams and simply supported rotating shafts.

The most common formulation for finding the critical speeds of a rotor is based

on the assumption that both rotational inertia and shear deformations are

negligible. However, the consequence of neglecting these effects is that the

critical speeds coincide with the natural frequencies of the non-rotating shaft.
Genta, G., [1985] goes further to state that because the effects of rotational
inertia and shear deformations have an opposite effect on critical speeds, models

which only include rotational inertia (which has a lesser effect) can be less

accurate than those which ignore both. Genta also states that shear and rotational
inertia effects are stronger at higher critical speeds, in particular during

supercritical running. The tests performed were done for a simple uniform shaft

such that an eigenvalue analysis was compared against a closed form solution.
Genta also noted that the inclusion of shear deformation underaxial loads was of

little practical importance as it introduced very small corrections.

Kim, Y. D., and Lee, C.W., [1986] furthered the previous work on Euler finite
beam elements, including the gyroscopic effects. They produced a matrix

reduction technique, which they claim substantially reduces computation time.

Although very significant at the time and still worth incorporating, it is perhaps
becoming less important due to the exponential increase in computing speed

seenover past years.

Chapter 2: Literature Review 22

Sauer, G., and Wolf, M., [1989] worked on the gyroscopic effect of rotating
disc-shaft systems.Included in their work is a method for calculating the natural

eigenfrequenciesof such a shaft. Using a base excited pump rotor they showed
the effect of separation, whereby the natural resonant frequency of a shaft is
separated into two resonant frequencies approximately equidistant above and
below the original natural frequency. It was also observed that at certain

excitation frequencies the response in the direction normal to the plane of

excitation is higher than that parallel to it, and for a broad frequency range the
responseis dominated by displacementsnormal to the excitation. This is a well-
known gyroscopic phenomenon caused by coupling of the moments. It is noted
that with the gyroscopic effects omitted for the particular `simple' problem
studied there is no responsenormal to the plane of excitation. Sauer and Wolf
continue to show how to analytically determine the frequencies at which the
excitation parallel to the plane of excitation is equal to the responsenormal to the
plane.

Chen, L. W., and Ku, D. M., [1991] developed a Timoshenko finite beam element

with three nodes for the analysis of the natural whirl speedsof rotating shafts.
Axial deformations are not considered, hence each node has four degrees of
freedom; two translational and two rotational. The purpose of the investigation

was to develop a COTimoshenko beam element in place of the more usual Cl

class elements. The finite element model presented provided and accurate
representation of rotating shaft systems. Interestingly, Chen and Ku noted that
the shear effect has a considerable effect on shafts where the end conditions are
hinged-clamped or clamped-clamped. The effect is much less significant for

clamped-free or hinged-hinged conditions.

Gmür, T. C., and Rodrigues J.D., [1991] proposed linearly tapered finite elements
for the modelling of rotor bearing systems. The elements include the effects
of
translational and rotational inertia, gyroscopic moments, internal damping, shear
deformations and mass eccentricity. Changing the shape functions
used created
linear, quadratic and cubic elements. Results showed accuracy improving
with
Chapter 2: Literature Review 23

the order of element used. The eight degree of freedom cubic elements are
published and are suitable for use analysing cylindrical or conical rotors.

Hong, S-W., and Park, J-H., [1999] proposed a new method to obtain exact

solutions for multi-stepped rotor bearing systems. An exact dynamic beam

element matrix in Laplace domain for a Timoshenko shaft was derived. The
most important advantage of the method was to reduce the system matrix size.
This was achieved because the method allows a uniform shaft segment,

regardlessof length, to be modelled by a single element without loss of accuracy

or incursion of error.

Mohiuddin, M. A., and Khulief, Y. A., [1999] presented a finite element

formulation for a rotor-bearing system. The model accounts for gyroscopic

effects and the inertial coupling between bending and torsional deformations.
This appears to be the first work where inertial coupling has been included in

this manner. The work also presents a truncation method to reduce the order of
the model. The reduced order matrices are shown to preserve the selected lower

modes of the system, while significantly reducing computation time.

Luczko, J., [2002] developed a geometrically non-linear model for rotating

shafts, with internal resonance and self excited vibration. He states that in order
to analyse the physical phenomenon of shafts rotating close to critical speeds, it
is necessary to use non-linear models. This is due to the large vibration
amplitudes that occur at critical speeds and accordingly, geometrical non-
linearities need to be considered. Another factor in the non-linearity noted is the

influence of axial forces on the transverse vibrations. In the analysis it is an

Euler type model of a slender shaft that is considered. The solution is based on

the Galerkin method. The model allows investigation of coupled torsional,

longitudinal and transverse vibrations. Results demonstrate torque has an
important effect on stability regions and in general reduces a systems
critical
speed. Some dynamical characteristics showing the effects of internal resonance

and self excited vibrations were obtained. Results also showed that at critical
Chapter 2: Literature Review 24

speeds there could be vast differences with results obtained using a standard
Euler model.

Zou, C-P., Hua, H-X., and Chen, D-S., [2002] proposed a modal synthesis

method for the analysis of rotor-bearing systems with multi-branched shafting.

The method was specifically developed to tackle internal combustion engines,

where the number of connected shafts makes them difficult to model by

conventional means. Although modelling could be achieved using existing
techniques,very large matrices are neededfor the models and as such, computers

of great capacity are required. The final is

model a shafting systemwhereby rigid
couplings and flexible connections are combined. As such, the method is similar
to the hybrid modelling technique used here. Additionally, the higher order
modes were neglected to further reduce computation requirements.The proposed
method gives results with good agreement to those found by the transfer matrix

method and full FEM for the lower order frequencies.

Xiong, G. L., Yi, J.M., Zeng, C., Guo, H. K., and Li, L. X., [2003] studied the

gyroscopic effects of the spindle on the characteristics of a milling system. The

method used was finite elements based on Timoshenko beams. This is not the
first time Timoshenko elements have included gyroscopic moments. In fact, the

method uses matrices originally developed by Nelson, H. D., [1980], but it is

considered to be the first analysis of a milling machine in this manner and full
matrices are provided.

During the evolution of rotor-dynamic knowledge many insightful books have

been written on the subject, with authors attempting to encapsulate all of the

most relevant work that is commonly used. Although most of these works omit
finer points such as the gyroscopic effect or influences of fluid film bearings,

many proffer the reader a good insight into computer modelling and finite
elements for rotor-dynamic applications. Some of the more concise works
include Zienkiewicz, O.C., [1971], Thomson, W.T., [1993] originally
published
in 1972, Lee, C-W., [1993], and Zienkiewicz, O.C., and Taylor, R.L., [2000].
Chapter 2: Literature Review 25

These publications prove a valuable source of reference during rotor-dynamic

research.

2.3 FLEXIBLE PROPELLER BLADE & DISC MODELLING

Flexible blade modelling is typically performed using one of two common

methods; either the blades are individually ignored and a rigid or flexible disc
attachmentof the same inertia is used (disc models), or the blades are considered
to be short and attached to a rigid or flexible disc (bladed disc models). This is
becausethese methods best describe gas turbines, which are probably the most

common problem analysed by the rotor-dynamic finite element method.

Consequently,many of the modelling techniques are not applicable for a flexible
blade attached to a rotating shaft with no disc. Nevertheless, there is some

research in this area; normally regarding wind turbines, ship propellers, or

helicopter rotors (which usually include a tip mass in the model). However, at

present these works tend to ignore the gyroscopic effects due to blade bending in
order to simplify the problems.

The first attempt to analyse the vibrations of shaft-disc assemblies was probably

made by Lamb, F. R. S., and Southwell, R. V., [1921], who investigated the

vibrations of a flexible disc while studying the failure of turbine discs. The

analytical formulation proposed was aimed at finding the first critical speed and
it included both centrifugal stiffening and the effects of bending. The work was a
joint production but the important formulation was credited to Southwell. The
study of vibrations of rotating cantilever beams also originated from work by
Southwell, but he worked with Gough (Southwell, and Gough., [1921]) on this.

Based on the Rayleigh energy theorem they suggested a simple equation to

estimate the natural frequencies of rotating cantilever beams. This is now known
as the Southwell equation.

Handelman, G., Boyce, W., and Cohen, H., [1958], analytically investigated the

effect of a tip mass on the transverse vibrations of a uniform cantilever beam.

The work was specificallydoneto understandthe behaviourof helicopterblades,
Chapter 2: Literature Review 26

considering both the increase in centrifugal force, which stiffens the blade, and
the increase in kinetic energy. They were probably the first to observe that the
centrifugal stiffening effect largely governs the first mode of vibration, whilst for
the higher modes it is the elastic behaviour of the material that is the dominant
factor. They were unable to determine at which mode the interchange between
the two mechanismsoccurs.

Carnegie, W., [1959] derived an early expression based on energy methods for

the work done due to centrifugal effects when a rotating cantilever blade

executes small vibrations. The formulation included the total potential and

kinetic energy for vibration purposes. Being formulated when almost all rotor-

dynamic analysis was centred on the analysis of gas turbines, the blades were

treated as being attached to a rotating disc. By determining the potential energy

due to centrifugal force and previously derived expressions for stationary beams,

Carnegie was able to use Rayleigh's method to formulate an approximate

expression for the first critical frequency. Carnegie, W., [1964] then derived an

expression that allowed for the inclusion of rotary inertia and shear deflections.
Carnegie, W., Stirling, C., and Fleming, J., [1965-66] investigated the centrifugal

stiffening effect further by comparing theoretical results with experimental ones

from a high speed test rig using the finite difference method. Results were

deemed `satisfactory', noting that in practice the blade root was less than rigid
leading to lower frequency values than the theoretical predictions. There was an

obvious agreement for a relative rise in frequency with speed above the
`standstill' value for any given shaft. Furthermore, it was observed that the

assumption of (frequency)2 - (speed)2 being linear seemed to be a satisfactory

approximation for theoretical and experimental results over modes one to four

for the various blades examined. Finally, it was noted that theoretical
inaccuracies could be reduced for the higher order modes to that of the first, but

only at the expense of a considerable increase in work. Rao, J. S., and Carnegie,
W., [1969] then continued to study the vibrations of rotating cantilever beams,

this time using a non-linear approach for a bladed disc model. Magari, P.J., and
Shultz, L. A., [1987] also used the potential energy theorem to produce
a rotating
Chapter 2: Literature Review 27

twisted blade finite element for the analysis of helicopter blades, but matrices
were not given.

Dokainish, M. A., and Rawtani, S., [1971] investigated the vibration response of

rotating cantilever plates using finite elements. The plates were mounted on the

periphery of a rotating disc and although centrifugal forces were considered,

gyroscopic moments were ignored. Because of the rectangular geometry of the

plate, flat triangular elements of constant thickness were chosen to describe it.

The increase in bending stiffness due to the centrifugal force was calculated from

the in plane stresses induced in the plate. The natural frequencies of the blades

were shown to be significantly higher than those of non-rotating blades.

Peters,D.A., [1973] produced extensive analytical work on the free vibrations of

rotating uniform cantilever beams. He did this in order to simplify results from
previous models so that approximate solutions for the main modes of vibration

were easily obtainable. Although finite element eigenvalues were obtainable at

the time, the large number of modes produced by an eigenvalue analysis of a
system containing a number of elements makes results very difficult to interpret.
Although the analysis is very detailed, it only considers the natural frequencies

of various blade types and does not consider gyroscopic influences on vibrations.

Kumar, R., [1974] investigated the vibrations of space booms under a centrifugal
force field, using the Myklestad method. Analysis showed that the difference in

natural frequency caused by centrifugal stiffening was highest for the first mode

and becomes insignificant for higher modes. This is in agreement with results

produced by Handelman, G., Boyce, W., and Cohen, H., [1958]. These

centrifugal variations are larger for out of plane vibrations than in plane

vibrations. Otherwise, this paper produces results that despite being applicable to
the sizing of a space boom, are largely irrelevant to the research here.

Stafford, R.O., and Giurgiutiu, V., [1975] developed a number of semi-analytic

methods for rotating Timoshenko beams acting as propellers. The analysis

included centrifugal force, shear and rotary inertia, but ignored
axial motion and
Chanter 2: Literature Review 28

eliminated the Coriolis forces. The majority of analysis was based around the
transfer matrix method. The usual Euler beam theory assumptionsare used, but
first order corrections were used in order to account for shear and rotary inertia.
They noted that at a frequency of 50 rad/s the total correction due to shear and

rotary inertia only changedthe vibration frequency by approximately 3%. Of this

correction, the largest change (68% of total correction) was due to shear effects,
it
so was concluded for it
many applications would be sufficient to only include
the shearcorrections.

Jones,L. H., [1975] presentedthe method of integral equations for determination

of the eigenvalues of a rotating beam with tip mass. The work was performed
becauseof the difficulties of finding the closed form solution to the eigenvalue

problem. The primary aim was to compute improvable lower bounds for the
eigenvalues of the reduced problem, as finding the exact solution was still
proving to be problematic. The work concluded finding improved lower bounds
for the second eigenvalue. The lower bounds agreed with approximate results

obtained previously by methods including Rayleigh-Ritz. Furthermore, the

Volterra equation method used offered substantial computational advantages

over the Fredholm equation method. At the time of publication these

computational advantageswere highly significant.

Hoa, S.V., [1978] uses the energy method to develop a finite element model for a
flexible curved blade with a weighted edge. This model is primarily aimed at the

analysis of auto cooling fans. However, Hoa, S.V., [1979] then extends this

method using work by Wang, J.T. S., Mahrenholtz, 0., and Böhm, J., [1976] to

create a more universal finite element code utilising Euler beam theory for the

analysis of flexible blades with tip masses. Although gyroscopic bending effects

are ignored, centrifugal stiffening is considered and a corresponding matrix

developed. This accounts for the effect of the propeller element's own
mass
putting itself and any other connected elements towards the root of the blade

under tension as the system rotates. Although the centrifugal stiffening

formulation is well presented, discrepancies have been found in the final
matrix.
Chapter 2: Literature Review 29

Putter, S., and Manor, H., [1978] investigated the natural frequencies of radial

rotating beams mounted on a rotating disc. The centrifugal force for a uniform
tapered beam is considered in the work, the derivation being very similar to that

proposed by Hoa, S.V., [1979]. Unlike Hoa who uses axial stress to produce a

stiffening matrix in addition to the standard stiffness matrix for bending, Putter

and Manor derive an energy matrix from the centrifugal effect. They then use
this to calculate stress at the root of the beam (where it is greatest) in order to

produce an expression for an approximate maximum permissible speed of a

radial rotating beam, dependant on the material's strength. Detailed aspects of

working designs will influence maximum permissible speeds in ways not

accountable by a formulation such as this.

Hodges, D. H., [1979] developed a finite element method for non-uniform

rotating propeller beams, which he coined the Ritz Finite Element Method, due
to its similarities to the Ritz analytical method. Although effective at converging

to the correct solution for the first natural frequency, there was no attempt to
include either centrifugal stiffening or gyroscopic effects. It was suggested the

method be used as a method of analysing non-linear and non-conservative

systems.

Wright, A. D., et al [1982] also considered centrifugally stiffened beams, taking

into account previous works by Hoa, S.V., [1979], Peters, D. A., [1973] and

Hodges, D. H., [1979]. Wright et al used the Frobenius method to calculate exact
frequencies and mode shapes for centrifugally stiffened beams where both
flexural rigidity and mass distribution varied linearly. The results were tabulated

and compared with other solutions including a conventional finite element code.
Where zero root offset and zero tip mass are used for ease of comparison, it can
be seen that results correspond almost exactly to both Hoa's and Hodges'

methods. Also published is a wide variety of results for other test geometries for
the verification of future codes. Furthermore, the authors observe that centrifugal

stiffening has a decreasing influence with increasing mode numbers. The

mathematics of centrifugal stiffening and corresponding effects has been further

investigated in Chapter 6.
Chapter 2: Literature Review 30

Stephen, N. G., and Wang, P.J., [1986] continued along the lines of earlier
boundary value solutions to tackle problems where the material cross sections
did not possestwo axes of symmetry. They did this in order to find the deformed

shape of loaded turbine blades before performing vibration analysis. Three-

dimensional problems were reduced to two-dimensional boundary value

problems in order to predict the curvature of the blade centreline, taking into
account the centrifugal forces on these rotating cantilevers. Yet since finite
element modelling was beginning to become commonplace, the value at the
stage is difficult to quantify. As the author himself states,"Apart from classical
interest, the solution should provide a test for approximate methods of analysis".

Bauer, H. F., and Eidel, W., [1988] investigated the vibrational behaviour of a

uniform beam orientated perpendicular to the axis of rotation. The fundamental

frequency for beams was determined by using approximated mode shape
functions and solved using the Ritz-Galerkin method. The analysis was
performed for all possible combinations of free, clamped, hinged, and guided
boundary end conditions, in order to investigate the effects of spin speed, hub

radius and aspect ratio on the various geometries. Centrifugal and Coriolis
effects were considered. The most important conclusion was that the spin speed
has a pronounced influence on the natural frequency of a rotating beam.
Depending on boundary conditions the natural frequency may increase or
decrease with an increase in spin speed.

Yigit, Scott and Galip-Ulsoy [1988] investigated the flexural motion of a radial

rotating beam attached to a rigid body. Deriving fully coupled non-linear

equations using Hamilton's principle and Euler beams. Note that the term
`coupled' is used to denote a system in which elastic and rigid body motions are
interdependent.As with previous investigations, centrifugal forces were included

and deemed a necessity, but gyroscopic effects were not. A torque profile was
used to drive the body so that the rigid body motion was an unknown prior to the

solution. Using both linearised analysis and numerical solution of differential

Chapter 2: Literature Review 31

equations, it was found that uncoupled equations could lead to substantially

incorrect results, particularly with regard to frequencies.

Sakata, M., et al [1989] investigated the vibration of a bladed flexible rotor due

to gyroscopic moment. The work was principally aimed at describing a jet

engine's response to a change in angular velocity produced by the aircraft
changing direction. Experimental and finite element work was performed using
flexible blades, disc and shaft. The model has been simplified such that only
eight degrees of freedom are used and full finite element derivations and results

are not given. However, results show reasonable agreement between

experimental and computational results.

Naguleswaran, S., [1994] studied the vibrations of uniform Euler-Bernoulli

beams under centrifugally induced tension. Results have been tabulated in order

to serve as data for the development of problems with more complicated flexural

rigidity and/or mass distributions. In and out of plane vibrations are considered
to be uncoupled, due to a principle axis of the blade cross-section being assumed

parallel to the axis of rotation, and torsional motion is not considered. The mode
shape equation was solved using Frobenius' method, with the general solution
being expressed by four linearly independent shape function polynomials.
Natural frequencies are shown to increase with blade offset (from the axis of

rotation) and/or rotational speed.

Genta, G., and Tonoli, A., [1996] developed a disc finite element for the analysis

of rotor-dynamic behaviour. The disc is treated as being thin (two-dimensional)

and flexible, taking account of centrifugal forces, gyroscopic effects, and
thermo-elastic effects. The element is developed in a complex co-ordinate
formulation. However, there should be no problems obtaining the
relevant
matrices in real co-ordinate form. It is shown that although linear shape
functions are suitable for the analysis of a disc, in the case of blade
arrays, cubic
functions must be used. With a low number of degrees of freedom the
element
model is said to be very useful for complex rotor-dynamic computations. Genta,
Chapter 2: Literature Review 32

G., and Tonoli, A., [1997] then expanded the model to accommodate bladed-
discs.

Jacquet-Richardet, G., Ferraris, G., and Rieutord, P., [1996] were also concerned

with the reduction of computation time. They investigated the vibration modes of
flexible bladed disc-shaft systems. To reduce computation time global non-

rotating mode shapes are used in a modal analysis to evaluate the dynamic

response. Gyroscopic and centrifugal stiffening effects are accounted for and all

possible couplings are allowed. Very good agreement was shown between this

and a full mathematical model, with the intended advantages in computation

time being achieved.

Al-Ansary, M. D., [1998] investigated the effects of rotary inertia on the flexural

vibrations of rotating cantilever beams, taking into account longitudinal

elasticity. Galerkin's method was used to create a simple formulation

demonstrating that for a typical helicopter blade the extensional force may
increase by up to 10% if the rotary inertia contribution is retained in the analysis.
This relationship may be used in the initial design stages of rotating beams.

Yoo, H. H., and Shin, S.H., [1998] investigated the vibrations of rotating
cantilever beams. They developed linear equations of motion based on a new
dynamic modelling method which employs a hybrid set of deformation
variables. The Rayleigh-Ritz assumed mode method is used to approximate the
hybrid variables. Stretching and bending terms were coupled gyroscopically.

Natural frequencies were shown to increase as angular speed and hub radius
increased. The chord-wise bending proved to be coupled with the stretching

motion of the beams, with different natural frequencies being exhibited,

especially for high angular velocities. This coupling effect was proved to become

negligible as the beam's slenderness ratio increased. Since the gyroscopic effects

are closely related to centrifugal stiffening this is to be expected. Beam element

matrices were not published and results were not compared against other data
due to a lack of published material.
Chanter 2: Literature Review 33

Bir, G., and Stol, K., [1999] investigated the operating modes of a teetered-rotor

wind turbine. They examined a two-bladed wind turbine becausethe gyroscopic

asymmetry can make the dynamics quite distinct from a turbine with three or
more blades. Traditional modelling of wind turbines analyses the modes of
vibration with the blades not rotating. This is becauseof the difficulty associated
with modelling coupled stationary and rotating components. However, this
method ignores the centrifugal and gyroscopic effects. The system was modelled
as having four degreesof freedom and it was shown that centrifugal stiffening
dictates rotor flap modes, whilst gyroscopic effects control yaw and teeter

modes. Blade flexibility and aerodynamic effects were ignored.

Aleyaasin, M., Ebrahimi, M., and Whalley, R., [2000] presented a paper on

vibrations of rotating shafts by frequency domain hybrid modelling. The

problem analysed was that of a cantilevered rotating marine propeller. The
distributed-lumped model for the system was derived from the transfer matrices.
Gyroscopic effects were considered and results showed this reduced order model

accurately representedthe system. The modelling technique was essentially a

simplified version of that used in this study becausethe system is derived from
transfer matrices instead of being modelled.

Huang, B. W., and Huang, J.H., [2001] investigated the effects of centrifugal and
Coriolis forces on a mistuned bladed disc. The blades were approximated as

cantilever beams, with ten modal co-ordinates considered for each blade. Results

were a little inconclusive but showed that the Coriolis forces might enhancethe
localization vibration. This phenomenon is seen in mistuned systems where the

vibration amplitude of individual defected blades may be seriously affected by

the local irregularities.

Yoo, H. H., Park, J.H., and Park, J., [2001] derived the equations of motion for
a
pre-twisted cantilever blade, using a method that employs hybrid deformation
variables. The method uses Rayleigh Ritz theory to approximate the deformation

within the system. Shear, rotary inertia and gyroscopic effects were considered

negligible. Results are expressed in dimensionless form, showing the gap

Chapter 2: Literature Review 34

between natural frequency loci increasing with pre-twist angle and also that the
hub radius significantly affects natural frequencies.

Chen, C.L., and Chen, L. W., [2002] investigated the random vibrations of a

rotating cantilever blade with external and internal damping using the finite

element method. Euter thin beam elements were used to describe the system,
with base excitation (earthquake simulation) and white noise being the main
methods of excitation. There was however, no attempt to describe the gyroscopic
moments induced by such a system.

Yang, J.B., Jiang, L. J., and Chen, D. CH., [2004] investigated the dynamic

modelling and control of a rotating Euler-Bernoulli beam. A finite element

model is produced containing the effects of centrifugal stiffening as derived by

Hoa, S.V., [1979]. Gyroscopic effects are however ignored as the study focuses

on the active vibration control of rotating beams. It is indicated that the scheme
in
can produce excellent results the suppressionof vibrations.

Chatelet, E., D'Ambrosio, F., and Jacquet-Richardet, G., [2005] investigated a

global modelling approach to the analysis of flexible bladed systems. The study
focuses on the interaction between shafts and propellers and as such uses a fine

finite element mesh which is of course very computationally expensive.

Gyroscopic and centrifugal stiffening effects are accounted for in the analysis.
The model's complexity drives the work towards reduction techniques to
improve computation time. The results show that the behaviour of complex

systems may be poorly modelled using traditional modelling techniques based on

one-dimensional beam approaches. However these one dimensional approaches

have not used the gyroscopic formulations developed here.
Chapter 2: Literature Review 35

2.4 THE AIMS OF CURRENT RESEARCH

The main objectives of this researchare:

9 To develop and/or validate a finite beam element that can be used to

analyse forced shaft vibrations of shaft systems with complex geometry,
including gyroscopic effects.

9 To develop a finite beam element that can be used to analyse forced

propeller vibrations. The element must include the gyroscopic effect, as

for a propeller it can be quite large. The technique should be effective in

allowing the modelling of propellers to be easily achieved in place of the

current trend to model them as discs where a gyroscopic effect is

expected.

" To investigate and validate the effect of centrifugal stiffening on propeller

blades. Specifically, the perceived effect of a propeller blades' stiffening

as it experiences an axial tension due to the forces imposed on it under

rotation. Including the effect in the model.

" To perform simulations of dynamic behaviour of mechanical shaft and

propeller systems in order to demonstratethe capability of the proposed

finite elements.
Chapter 3: Finite Element Hybrid Modelling 36

CHAPTER 3

FINITE ELEMENT HYBRID MODELLING OF FLEXIBLE

SHAFTING SYSTEMS

3.1 INTRODUCTION
This chapter presents the Euler Newton formulation of the vibrational behaviour

of a multi-body system interconnected by discrete stiffness elements. The

formulation is extended to incorporate flexible shafting systems. It is assumed

that flexible shafts are connected to rigid bodies and the connection can be either
built-in or pin jointed. The described methods are used in the VIBRATIO suite of
vibration analysis software which has been used to incorporate elements
developed in Chapters 4 to 6. A typical multi-rigid-body system supported or
interconnected by discrete spring elements and beams is shown in Figure 3.1

ýý ". ý. ý,
ý,

Figure 3.1: Schematic Representation of a Multi-Body System

3.2 THE FINITE ELEMENT METHOD

The finite element method provides a systematic approach for the discretization

of a continuum. It can provide a solution for many different types of complicated

systems including fluid flows, heat exchange, static mechanical stresses, or
dynamic mechanical systems, and the rotor-dynamic systems examined here.
Chapter 3: Finite Element Hybrid Modelling 37

These problem types are very complex and would be difficult to analyse by other
techniques (classical methods).

In each case the system is divided into small contacting regions known as

elements.Described mathematically, adjacent elementsare effectively allowed to

touch each other without overlapping. Usually the element shapesare chosen to
be as simple as possible, but when grouped together they can still be used to
describe a complex geometry. In the case of dynamic rotor-bearing systems,the

system is approximated by finite degree of freedom beam elements whose

motions are described by ordinary differential equations. This method of
describing a complex system as a group of small attached elements is more

straightforward to analyse than if it were treated as a single unit. This is because

each element is mathematically more easily expressedthan the whole system.
Numerically, combining a number of elements to model the whole system and

solving them by computer is a routine task that can be performed very easily
comparedto a classical analysis.

For each element the governing equations (usually in differential form) are

transformed into algebraic equations called the element equations. For simplicity
these are often expressed in matrix form and thus tend to be referred to as the

element matrices. When fully assembled these element matrices represent an

approximation of the physical system, with the equations expressing a balance of
some physical property. For the rotor-dynamic analysis performed here, the
expressed property is energy. The ease of use of the finite element method

revolves around the fact that the element equations are identical for each element

of the same type. Consequently, element equations need only be derived for each
typical element type, not for every element. It is normally possible to describe

even a complex system with only a limited number of element types, perhaps

only one or two. Thus the entire problem of solving a set of algebraic equations
is reduced to finding an adequate set of equations for the simple elements
used;
combining them into the appropriate system equations and solving them as a

system. The approach used produces large system matrices that would be very
Chapter 3: Finite Element Hybrid Modelling 38

difficult to solve by hand, however the assembly and solution of these matrices is

a routine task that is ideally suited to the modem computer.

3.3 RIGID BODY THEORY

3.3.1 Definitions and Assumptions

" The model is geometrically linear: i. e. the geometry of the system does

not change during vibration, thus the vibration amplitudes are assumedto
be small.

" Gyroscopic effects are assumedto be small and negligible for the rigid
bodies, but significant for shafting systems.

" For FEA of shafting systems all the assumptions relating to linear
elasticity are assumed.

9 The time dependenteffects of hyper-elastic materials are excluded.

" It is generally assumedthat the mounting (or spring) has zero length.

" Throughout the analysis it is assumedthat the stiffness of springs in their

principal axes of deflection are uncoupled.
These assumptionsare acceptablefor most of the engineering vibration problems
where the vibration amplitude is small, as is the case with most systems.

3.3.2 Equations of Motion

In order to set up the equations of motion for a dynamic system the following are

required:
" Generalisation of the equations of external forces and internal reactions.
(Internal reactions due to damping and stiffness elements)

" Generalisation of the equations of linear momentum.

(Force - acceleration equations)

9 Generalisation of the equations of angular momentum.

(Turning moment equations)
Chapter 3: Finite Element Hybrid Modelling 39

3.3.3 Stiffness and Damping Systems

The presented stiffness matrix formulations apply equally to the damping matrix.
This is achieved by replacing the stiffness parameters with the corresponding
damping parameters.

Assuming the stiffnesses are described in a Cartesian local axis system that

coincides with the principal axes system, then the force f acting on the spring
may be expressedas:

f=kx (3.1)

Where k is the stiffness matrix and x is the displacementvector.

Generally it is most convenient to describe the behaviour of a system in the

global axis system OXYZ. However this is not a prerequisite as it is equally

possible to set equations of each body in its own frame, in this study all stiffness

matrices are expressed in a common global axes frame. Therefore the individual

matrices have been transformed accordingly. Since the principal axes of the local
and global axes are orthogonal, an orthogonal transformation exists between the
two frames. Thus a displacement vector x in local axes coordinates can be

expressed as vector X in the global axes system using T, a transformation matrix

which may be described by three Euler rotations.

Giving:

X=Tx (3.2)

Therefore, the reaction force F in the global system, by pre multiplying Equation
3.1 by T, can be expressedas:

F= Tkx (3.3)
Chapter 3: Finite Element Hybrid Modelling 40

Now since:

x= TTX (3.4)

F= TkTTX (3.5)

Introducing a new matrix notation K, to representsthe global stiffness matrix:

K=TkTT (3.6)

It can be seenthat:

F= KX (3.7)

3.3.4 Generalisation of the Equations of Linear Momentum

If the massmatrix in the Euler Newton formulation is obtained relative to the
axis passing through its centre of mass, then the subsection of the mass matrix
corresponding to linear momentum is a diagonal matrix containing the mass

elements.

Then it can be statedthat:

hý =mi (3.8)

Whereh, is linearmomentum,m is a diagonalmassmatrix, and x is the velocity

vector of the body.

Using the transformation expressed previously, the global matrix H, can be

expressedas:
H, = TmTTX (3.9)
Chapter 3: Finite Element Hybrid Modelling 41

Since:
M= TmTT (3.10)

The force acting on a body i, which can be described as the rate of change of
momentum is:

Force, = H; aý'=m; i (3.11)

=

3.3.5 Generalisation of the Equations of Angular Momentum

Similarly, the equations of moment of momentum may be expressedas:
ha=jä (3.12)

Here hQis the angular momentum, and j may or may not be a diagonal matrix.

Once again this can be transformed into the global axes system giving:

Hp = TjTTÄ (3.13)

Introducing a new matrix notation J:

J=TjTT (3.14)

It can be seenthat:

Ha = JA (3.15)

The vector differentiation of HQwith respect to time gives the moment vector in
the global axes system:
OH°
Momenta = 11a= +AxHQ (3.16)
at
The term Ax Ha contains the product of angular velocity (usually referred to
as
the gyroscopic effect). For most vibration problems this is small and may be
Chapter 3: Finite Element Hybrid Modelling 42

ignored. However, in this work the effect has been included for both shaft and

propeller finite elements.

3.4 ASSEMBLY OF EQUATIONS

3.4.1 Force - Acceleration Equations

In order to assemblethe equations of motion, the internal forces acting on the

individual bodies due to their motion relative to each other are required. Figure
3.2 shows two bodies (i andj) in motion.

Figure 3.2: Bodies i and j Connected by Spring k,.

The motion of the axis system belonging to body i is given by the vector x; and

the angular rotation of the axes by vector a;. Similarly, the motion of body j is
described by xj and aj.

For small motions, displacements of the end points of springs on each body
described in the axes frame of each body, are given by:

di= xi + ai x rpi (3.17)

dj =xj + aix rpi (3.18)
Chapter 3: Finite Element Hybrid Modelling 43

Where rj,, and r,; are the coordinates of the spring attachment relative to bodies i

andj respectively.

d; can be expressedin matrix form including the cross product term as:
dx; 0
x; Z,,;
a,
-yv;
d,, = y; + -zr, 0 x,,; ß; (3.19)
d_, 7i yy1 -xpi 0 r;

Or in short using new notation as:

d, = x; + Ry,a; (3.20)

And dj can be written in matrix form including its cross product term as:

d4 Xj 0 : a. i
ri -y,,
d}j = y; + -zYj 0 x,3 ßj (3.21)
d zj yp3 -XP 0 yi

Which in short hand becomes:

dj= xj + Ry1a (3.22)
j

The relative displacement of the bodies is given by:

d=dj -d; (3.23)

Reaction forces due to the relative displacements on each body are respectively

given by:
F, =k,. d Fj = -k, d (3.24)

And moments by:

Moment; = r; xFj and Momenta= rj x Fj (3.25)
Chapter 3: Finite Element Hybrid Modelling 44

The crossproducts can be expressedin matrix form so that:

Momenta, 0 -: ' yp, F,

Moment., = 0 -xp; F;; (3.26)
v;
Moment, -yp; x'O 0 F_;

Noting that the matrix in Equation 3.26 is the transposeof the matrix in Equation
3.19 it can be statedthat:

Moment, = RT,F; (3.27)

Again for Moments the cross products can be expressed in matrix form so that:

Momenta 0 -z pý yý,ý F4
Moments 0 -XPj F, (3.28)
f_ pj j
Momentrj -yrj xyj 0 F-j

Noting also that the matrix in Equation 3.28 is the transpose of the matrix in

Equation 3.21 it can be stated that:

Moment =Rj Fj (3.29)

Now the equations of motion can be compiled as:

m, z, + (k, di - k,.dj) = Fj (3.30)

Here F; is the vector of external forces acting on body i.

And:

mj xj - (kr di - krdj) = Fj (3.31)

Then substituting dj and d; from Equations 3.20 and 3.22 then F; can be

expressedas:
mi x, + kr (x; + Rp; aj) - kr (xj + Rrj aj) = Fr (3.32)
Chapter 3: Finite Element Hybrid Modelling 45

Similarly:

mj xj +k, (xj + Rpjaj)- k, (x; + Rp; a;)=Fj (3.33)

Expanding gives:

m11, +krx; + krR, a1- k, xj - k, Rpj aj =F; (3.34)

x
mj -14x; - kr Rpi ai + k, xj + kr Rpj aj = Fj (3.35)

Equations 3.34 and 3.35 are the Force - Acceleration equations.

3.4.2 Moment Equations

Writing the moment equation for body i, where it is the external moment acting

on body i:

Jjä; + RT; (k,. di-k, dj)=Moment; (3.36)

Similarly :

(k, d; - kr dj) = Moment;

Jjäj - RPM (3.37)

Substituting d; & dj as before and expanding gives:

J; ä; + Rp, k, xi+ RTP,

krRp; a; - Rv; k, xj - Rp; k, Rpj aj =Moment, (3.38)

Jjäj - Rp. krx; - Rp. k, R1 a; + Ry k, xj + RP k,. Rpýaý =Moments (3.39)

Equations 3.38 and 3.39 are the moment equations.

Chapter 3: Finite Element Hybrid Modelling 46

Expressing Equations 3.34 and 3.38 in matrix form gives:

0 z; kr krR,,; kr krRý, Force;

1m; +
x;
_ RT k
ý xj
_
0 J. ä.iNr RT k rk r0R aj VJ r RTP.I kRa.
r PJ J Moment.
(3.40)

Similarly Equations 3.35 and 3.39 are expressedas:

0 xj kr krR kr krR Forced

1mj _ Rl, kr
pi x;
+
Rf kr R
pj xj_
Momenti
0 Jj äj r,krR p; a; aj
jkrR pj
(3.41)

The overall equations of motion are now complete. It is worth noting that the

stiffness and damping matrices are identical in structure. Thus in order to obtain
the damping matrix, one simply replaces stiffness coefficients with the

corresponding damping coefficients.

3.5 THE FEA FORMULATION FOR BEAMS

The FEA formulations for standard Euler or Timoshenko beams are both widely

available in the literature. These formulations will therefore only be stated for
Euler beams. The full matrices used are given in Appendix A.

3.5.1 Stiffness Matrices

Assuming suffixes 1 and 2 represent each endpoint of a beam, it and 0 represent

unit displacements,then the bending and lateral motion for the coupled y and
gamma axes are given by:
Fy., 12 61 -12 61
Mri EI 61 4J2 -61 212 Br1 (3.42)
-
1
Fy2
M72
13 -12
61
-61
2/Z
12 -61 u, 2
412 orz
-61

The same matrix can be rotated and used to represent bending in the second
lateral axis (in this casethe z and beta).
Chapter 3: Finite Element Hybrid Modelling 47

The axial deformation is given by:

IF', l_EALI1 1
-T: rx, (3.43)
F2 1JIilx2Jr
1-1

Torsional deformation is given by:

tM.
Mý'J=GJ[ 1' (3.44)
2 11Ito., i

3.5.2 Mass Matrices

The mass matrix corresponding to the above stiffness matrix, where u is the

massper unit length, is given as:

Fri 156 221 54 ü,.
-131 1
My1 u1 22/ 412 131 Or, (3.45)
-3/Z
Fy2 420 54 131 156 -22/ ü52
M, 4/2 1$
2 -131 -3/2 -22/

For axial motion:

{F=} [1 }
(3.46)
6 2]{üj,

And for the axial rotation:

{M. [1
(3.47)
6 2]1ö
2

Where j' is the rotational inertial per unit length. This can be expressedas pr 2

for cylindrical beam elements.

The equations of motion for beams (or shafts) are described in their local axes
frames. In order to assemblethe shaft equations with the rigid body formulations

given earlier, a set of co-ordinate transformations are necessary.In the Cartesian

co-ordinate system there are three possible displacements and three rotations at
each end of a beam element. Thus twelve co-ordinates are used to represent the
end deflections of each beam element.
Chapter 3: Finite Element Hybrid Modelling- 48

T
(zýi eai e), 1x2 uy2 11. 0 oý
Ztyl 11--l 81 1 a (3.48)
x2

Writing this in short hand gives:

(zh ý 02)T
Zvi (3.49)

To start the assembly process and find the overall equations of motion, it is

convenient to convert the beam deformations into the global axis co-ordinate
system for all the matrices. The transformation matrix for this operation is given
by:
TS

Ts (3.50)
T=
TS
TS

Each sub matrix Ts contains the direction cosines between the beams co-

ordinate system and the global co-ordinate system.

After the transformation the deflection vector is expressedas:

T
(1111 8l
1112 B2) (3.51)

The elemental mass and stiffness matrices are transformed into the global axis
frame by the following operations:

K=T TKj (3.52)

M=T TM (3.53)
CT

Assuming the assembly of all the elemental stiffness and massmatrices gives the

stiffness and mass matrices (Ks,, ft , M,., ) for the overall shaft, then to attach a
beam to a rigid body, the deformation vector at the point of attachment (Vector
3.51) has to be transformed to the motion of the mass centre of the rigid body.
The following sectionsgive the formulations for the co-ordinate transformations.
Chapter 3: Finite Element Hybrid Modelling 49

3.5.3 Co-ordinate Transformation for the Built-In Condition

Two equations are needed to relate the motion of the beam to the rigid body

attachment. The first relates the movement at the point of attachment expressed
in terms of rigid body motion, to the motion of the beam at the point of

attachment, expressed in terms of the FEA formulation. The second equation

relates the angular motion of the rigid body and the angular motion of the beam
at the attached node.

Assuming u and 0 to be the beam node displacements and rotations

respectively, x and a to be the rigid body displacements and rotations

respectively, and r3 the vector from rigid body's centre of origin to the point of

shaft connection, then connecting beam i to body j the first equation is:

u; = xi + of x rs (3.54)

Where the vector rs is:

r, =fa b c} (3.55)

The second equation relates the angular motions of the rigid body and the shaft

end. Assuming both the rigid body and beam element are expressed in the same

co-ordinate systemthen:

0i =Ui (3.56)

Thus the transformation may be written in matrix form as:

ums, 1000c -b x,
U, 010 -c 0a yj
try, -a 0 :, (3.57)
BQ, _001b
0 0 0 1 0 0 j a,
Bp, 0 0 0 0 1 0 ßj
0ý 0 0 0 0 0 1 yj

Which may be expressedin shorthand as:

{}=(s3{Xi}
(3.58)
+
Chapter 3: Finite Element Hybrid Modelling 50

If the local reactions at the point of attachment are expressedabout the centre of

mass,then the following relationship may be written:

Fý 100000 fj
Fy, 010000 f

F, (3.59)
TQ, _001000f,
0 -c b100 tai
Tn, c0 010 tq,
-a
Tn a0001 try
-b

Or once again in short as:

{}=[s]r{} (3.60)

Where the vector on the right hand side of the equation is the reaction forces at
the point of attachment and the vector on the left-hand side of the equation
representsthese reactions in relation to the rigid body axis system.

Now extending the principle it can be seen that if both ends of the shaft were

connectedto rigid bodies, then the relevant transformation would be:

u,, xi

B,, [S,, 1 aj (3.61)

_
U21 L SkJ
xk
ez1 at

Here the first end of beam i is connected to body j and the second end of beam i
is connected to body k.

Thus if the beam connecting the two rigid bodies is made of a number of finite

elements,then the transformation matrix can be written as:

UI1 xi
©
Is, 1 ai (3.62)

.. I..
Sk xr
ýý ak

This is the final transformation matrix, obtained in order to provide a co-ordinate

system consistentwith the rigid body axis system.

Chapter 3: Finite Element Hybrid Modelling 51

If the transformation matrix is called R, then the stiffness matrix in this co-

ordinate system could be expressedas:

K! =RT KShafý
R (3.63)
SJ,
Global.
an

Similarly, the massmatrix would be expressed:

TMsi,
Mr =R \3.64
clnalsnqn afR

These matrices can now be assembledinto the system global mass and stiffness

matrices. It is important to note that the first and last six variables relate to the
existing rigid body motions. The other variables are `new' variables. Therefore,
the assemblyoperation has to be carried out with great care.

3.5.4 Co-ordinate Transformation for a Pin-Jointed Connection

Here the connection is treated as being pin jointed for lateral rotations, but axial

shaft rotation is still coupled with that of the rigid body. This can be used in the

modelling of a system containing a mechanical connection akin to a universal

joint, or with minor modifications could be converted to a truly pin jointed
connection. Again connecting beam i to body j the motion at the point of the

attachment is:

u; =xj+ajxr, (3.65)

However, now the second equation does not exist for lateral rotations, since there

is no relationship between the rigid body and rotations of shaft end nodes:
0, ý aý (3.66)

This means that the transformation matrix relevant to this case is the top three

rows and six columns of the transformation matrix S, used for the built in
condition. If the bending of the ends of the shaft is not coupled with the rigid
body, a rigid body motion will result for the beam and the resulting stiffness

matrix will be ill conditioned. This is due to the rotational rigid body motion
under torsion. In order to eliminate the rigid body motion in torsion, it may be
Chapter 3: Finite Element Hybrid Modelling 52

assumedthat the torsion of the shaft is equal to the component of the rigid body
rotational motion along the beam. This can be expressed in terms of the dot
product between the direction cosines of the beam axis, and the angular rotation
a, of the rigid body.

Thus:
0ai =u"a (3.67)

Where u= {u1 u2 it3} and a={ a1 ßj yjI. Here u contains the direction cosines

of the beam along its length.

The transformation may thus be written in matrix form as:

xi
U. 1 0 0 0 c -b 0 Ollyj
uy, 0 1 0 -c 0 a 0 0 :j
tr 0 0 1 b 0 0 0 a, (3.68)
-a
Oa, _ 0 0
0 0 0 u, U2 Zl3 Yl
8ß, 0 0 0 0 0 0 1 0 iJ
By, 0 0 0 0 0 0 0 1 0m
Br1

It can be seen that at the point of connection there are now eight degrees of
freedom. This is due to the fact that six belong to the rigid body and six to the

shaft node, giving a total of twelve degrees of freedom. However, three

displacements of the shaft node and rigid body are identical (and coupled)
as is
the axial rotation, so four degrees of freedom cancel. The rest of the formulation

to transform the mass and stiffness matrices into the global axis system is the

same as that of the built-in condition (Equations 3.60 - 3.64).

3.6 PROGRAMMING CONSIDERATIONS

The equations developed in this chapter have been formulated
and structured
such that they can be used for developing general vibration analysis software. As
Chapter 3: Finite Element Hybrid Modelling 53

the equations given for the rigid body motion refer to bodies i andj, they can be
placed in the global axis system accordingly.
The solution of equations 3.40 and 3.41 requires four (6x6) sub matrices to be

placed in the correct positions in the global system matrix as follows:

kr krRp;
placed starting from position (6(i-1)+1,6(i-1)+1)
RTpr k rrr rk R.pr

k,. kRpj
placed starting from position (6(i-1)+1,6(j-1)+1)
RT k
PJ "
RTPJkR.
+ PJ

kr krRý,;
_ [Rnkr placed starting from position (6(j-1)+1,6(i-1)+1)
ir R.pý

kr krR'"
[Rkr Placed starting from position 6Ü-1)+1' 6(j-1)+1)
RT k R.
Pi Pi

It is necessary that the entries into the global stiffness matrix are additive (i. e.

each new entry is added to what has previously been recorded in order to account
for the contributions of different springs/shafts). Thus it is easiest to construct

the global stiffness matrix using a counting loop. The data structure for each

spring needs to refer the spring to the attached mass number. In this way, a
spring stiffness sub matrix can be located in the global matrix according to the
mass numbers the spring ends are attached to. These principles apply equally to
the damping and inertia matrices, which also need to be expressed in the global

axis frame using a matrix transformation before they are located in the global

matrices.

3.7 SUMMARY
This chapter presents a method of analysing general multi-rigid body systems
inter connected by linear springs and shafts (beams). The mathematical

modelling is for small vibrations where non-linear geometry effects and

Chapter 3: Finite Element Hybrid Modelling 54

gyroscopic couplings (between rigid body masses) are assumed to be small.

Also, the deflection characteristics of mountings are assumed to be linear. There
is a shortage of published material on general mathematical modelling of
flexibly supported multi-rigid-body systems for vibration analysis. Additionally,
little published material has been found on hybrid modelling where rigid bodies

of finite size (not point masses) are modelled together with FEA representations

of flexible beam elements. This chapter gives a precise formulation suitable for

computational implementation; the formulation forms the core of the vibration

analysis suite VIBRATIO which has been used to incorporate elements derived
in Chapters 4,5 and 6. Although rigid bodies are not mandatory for the

verification of beam elements, the inclusion in the software creates a more

universal modelling technique, making it easier to analyse different system

types.
Chapter 4: Gyroscopic Beam Element 55

CHAPTER 4

SHAFT GYROSCOPIC BEAM ELEMENT MATRIX

DERIVATION

4.1 INTRODUCTION
Euler beam elements are often used in finite element analysis but rarely include

the relatively minor gyroscopic coupling effects. Although there are a number of

papers that publish a gyroscopic beam element matrix for an Euler beam

(examples include Nelson, H. D., and McVaugh, J.M., [1976], Kim, Y. D., and

Lee, C. W., [1986], and also Zorzi, E. S., and Nelson, H. D., [1977]) there are none

showing the full derivation. Here, the full derivation is given in order to provide

the reader with a comprehensive understanding of a gyroscopic beam element,

before further progressing the work to develop a gyroscopic propeller element in

the following chapter. In this case the matrix has been derived using the energy

methods explained by Gasch, R., [1976].

In practice it can be extremely difficult to determine the exact solution of the

differential equations that describe a complex system. Hence the finite element
is
method an approximate solution. In the theory of finite element beams this
approximation derives mainly from the fact that the element positions are
determined by predefined shape functions that only approximately satisfy the
differential equations and relevant boundary conditions. The higher the order of

shape function that is used, the better the approximation that will be achieved.
Other approximations often arise due to the omission of various effects because

of the complexity of creating an all encompassing element that accounts for

Chapter 4: Gyroscopic Beam Element 56

everything. One good example of a simplification omission is the gyroscopic

effect, which is investigated and included here.

4.2 THEORY
4.2.1 Beam Elements (Three Dimensional)

The general three dimensional beam element can be thought of as being

described by four lower order displacements or elements combined. Assuming

the element to lie along the X axis, then the four components can be described as
follows:

" Bar element, which describes axial deformation.

" Torsional element, which describestorsional deformation.

" X-Y beam element, which describesbending in the X-Y plane.

" X-Z beam element, which describesbending in the X-Z plane.

In this analysis the beam elements are assumed to be straight bars with uniform

cross-section and two nodes (one at each end). There are twelve degrees of
freedom per element, six at each node (three displacements and three rotations).

The shape functions are required to define the longitudinal position of the beam

elements for given rotations and deflections of nodes, while the node coordinates

are functions of time. Many papers ignore axial and torsional deformations in
their derivations, producing element matrices that show only the eight degrees of
freedom, which are those associated with bending. These bending matrices
are
still identical to the bending parts of twelve degree of freedom systems, but care

must be taken with axis systems when making comparison.

The X-Y and X-Z bending beam elements have been constructed using Euler-
Bernoulli beam theory. Euler-Bernoulli theory assumes the normal stresses
(Qy and o. ) are zero along the longitudinal surfaces of the beam. This

assumption is generally considered reasonable for slender beams, which have

one predominant dimension. Furthermore, the theory assumes that under
Chapter 4: Gyroscopic Beam Element 57

deformation the transverse cross-section remains perpendicular to the deformed

central line, meaning that there is neither shear deformation nor shear stress.

4.2.2 Equations of Motion

Gasch, R., [1976] states:"The equations of motion can be found by the principle

of virtual work, which statesthat the work done by external forces acting on our
system and the work done by the internal forces must vanish for any virtual
displacement."

Therefore:

aw=o (4.1)

The virtual work öW is summed up as follows:

$auTd1EI(uh_cýNu)dx+ Jo11T
aw= jöu"TEIu"dx
+ iüdx
,o
JauTdeüdx faurk5Nudx JUTC,
+ - - gUdX

+ T
Öu? BQu' ÖuiT1 O Nu']/ }
inu
-}-[ÖuT -

ölT (CS2u + BS)ü) +t {_c22 $oUT

+j] nt
COS
JUECdX

fau', [_c2ouTnis }
uc.,dxsinf)t+ cossit_n2auT,,
-c22 zesinslt], (4.2)

From which the gyroscopic virtual work is defined as:

aWG
=- J[aUTcOPNÜ1, (4.3)

Since the gyroscopic bending moments are dependant on the angular velocity of
the beam elements, the resultant element matrix will be added to the damping

matrix in the global system. For this reason, the gyroscopic coupling matrix is

often referred to as the gyroscopic damping matrix.

Chapter 4: Gyroscopic Beam Element 58

4.2.3 Gyroscopic Beam Element Matrix Derivation

For the purposes of this analysis the following element shape functions have

been used:

0 1

hl
I uo =1

h2
00=1

h3
=1

9, =1

Figure 4.1: Shape Functions

Where:

=1 (4.4)

1-3ý'+2ý' (4.5)

hz -1(ý-2ý2+ (4.6)

k= 3ý2 - 2ý3 (4.7)

h4=l (ý 2- ý3) (4.8)

Chanter 4: Gyroscopic Beam Element 59

Figure 4.2: Shaft Definitions

By introducing a body frame such as Figure 4.2, the description of motion is

reduced to the analysis of the motion of a coordinate system, the position of

which is given at any time by the displacement vector w and its corresponding

rotation matrix. Thus the element derivation can now be completed with the

element lying in its own local axis system. Furthermore, the axis of rotation

can be assumed to lie along the beam's axis of symmetry for simplicity. This

is now much simpler to describe mathematically and the element can be

placed into the global axis system by using a transformation matrix as part of

the software assembly process.

The equations of motion for the gyroscopic effect can now be derived for a

using
singleelement the virtual work principle definedearlier:

awe=_J[euT0Nu}1dx (4.9)
Chapter 4: Gyroscopic Beam Element 60

Since:
aau'(x,t) = V'T (X). aus (t) (4.10)

Then the gyroscopic virtual work can be described as:

$[V'nOpNia'], dx (4.11)
awG=-aue

Also:

ü(x, t)= VT(X)"üe(t) (4.12)

Thus the gyroscopic virtual work for an element can be expressedas:

i
f [V'flOpNV'T ]c" (4.13)
awG,_ -aUe0 tie

Since the rotational speedof the shaft and polar moment of inertia per unit length

are constantswithin the integration, this can be expressedas:

f
nOp [VNV'T] dx-üe (4.14)
awG,=-aue 0

Expressing GQ as follows:

Ge = 0Op J[VPNV?
TI dx (4.15)
0

Then:
MG, = -aur, "Ge "ite (4.16)

Where G. is now defined as the gyroscopic beam element matrix.

For a flexible cylindrical beam element whose axis of symmetry coincides with
that of rotation, a twelve degree of freedom system is used. If x is the axis of

symmetry then by the geometric displacement principles shown in Figure 4.3, VT

becomes:
Chapter 4: Gyroscopic Beam Element 61

0 hZ 000h, 0 h, 0
r00h,
V0h, (4.17)
000 -h2 0 h3 000 -hý

The displacementand velocity components are defined as follows:

ue =(xo Yo zo %A Yo xi Yl :, ci Ql YO (4.18)

ue xo Yo o &'o A ro x Yi ± 6i A Y, ()4.19
e= .

v lgure 4.3: shape Function Matrix Geometry

Chapter 4: Gyroscopic Beam Element _62

Thus multiplying out the matrices in Equation 4.15 gives:

0 0 0 0 0 0 0 0 0 0 0 0
0 0 h,'h,' 0 h,'h2 0 0 0 h,'h; 0 h,'h; 0
0 'k 0 0 0 h,'14 0 -h, 'h3 0 0 0 h,'h;
-h,
0 0 0 0 0 0 0 0 0 0 0 0
0 ' 0 0 0 1414 0 -hZh; 0 0 0 14h,'
-hZh,
0 0 0 -1414 0 0 0 -1414 0 -14h. 0
V 'NV 'T = -k/4
0 0 0 0 0 0 0 0 0 0 0 0
0 0 h;h1' 0 1414 0 0 0 h;h; 0 loh{ 0
0 0 0 0 1414 0 -1414 0 0 0 /412;
-1412;
0 0 0 0 0 0 0 0 0 0 0 0
0 '
h, 0 0 0 h4'12 0 -N13 0 0 0 h'h'
44
-h,
0 0 -h412' 0 -1412 0 0 0 -1413 0 -141; 0

(4.20)

For a cylindrical beam element which has its axis of symmetry coinciding with

the axis of rotation, the polar mass moment of inertia per unit length is defined as
follows:
ýtr
OP=2 (4.21)

Therefore:

G'2r2 [V'NV'T]
Ge = öx (4.22)
0

Given that:
ah ah ö
_ aý ax and ý=x (4.23) & (4.24)
äx 1

Then it can be seenthat:

ah ah 1 (4.25)
ax __
aý 1

Also:

O' = (4.26)
ax
 63
Chapter 4: Gyroscopic Beam Element

Thus the integrals from within the matrix are as follows: (4.27) to (4.36)

1hh
/I ax 1414ax
5 101
0 0
I
jz
hl Jh2h47x
=-
J 101 0
301
0

J/414ax=-51 fh3h3öx51
0

f kh4 Jhh; ax
=101
0
101 0
1 212
hZh2ax 1h4h4ax
J= 151 0
151
0

4.2.4 Beam Element Matrix

Inserting terms 4.27 to 4.36 into Equation 4.22 gives the gyroscopic damping

matrix for a beam element:

000000000000
00 36 0 -31 000 -36 0 -31 0
0 000 -31 0 36 000 -31
-36
000000000000
0 31 000 412 0 -31 000 -12
0"f 00 31 0 -412 000 -31 0 12 0 (4.37)
G`
601 000000000000
00 -36 0 31 000 36 0 31 0
0 36 000 31 0 -36 000 31
000000000000
0 31 000 -12 0 -31 000 412
00 31 0 12 000 -31 0 -412 0

This is the complete gyroscopic beam element matrix for a cylindrical element

which rotates about its axis of symmetry. The element has been defined as lying

along the x-axis such that gyroscopic coupling occurs in the y and z axes. It can
Chapter 4: Gyroscopic Beam Element 64

be seen that it is z and beta deflections that create moments about y and gamma,

and vice-versa.

4.3 ANSYS GYROSCOPIC DAMPING MATRIX (ANSYS Theory)

Below is the ANSYS gyroscopic matrix as defined in the ANSYS Theory

Reference Element Library. ANSYS refers to the matrix as a gyroscopic

damping matrix. Although the gyroscopic effect is not strictly that of damping, it
is dependant on the velocity component in the same way as damping and is thus

often referred to as such. The element type is PIPE 16 - elastic straight pipe.

Whilst it shares its stiffness and mass matrices with the standard BEAM 4

ANSYS element, the gyroscopic matrix is unique. Unfortunately, a derivation or

reference to other material has not been given, so the matrix is included purely
for comparison purposes. Note that ANSYS uses the same axis system for beam

elements as that used here.

0 0 0 0 0 0 0 0 0 0 0 0
0 0 g 0 h 0 0 0 -g 0 h 0
0 -g 0 0 0 h 0 g 0 0 0 h
0 0 0 0 0 0 0 0 0 0 0 0
0 -h 0 0 0 i 0 h 0 0 0 j
0 0 -h 0 -i 0 0 0 h 0 -j 0
G= 2S2p Al
kuvsrs 0 0 0 0 0 0 0 0 0 0 0 0
0 0 -g 0 -h 0 0 0 g 0 0
-h
0 g 0 0 0 -h 0 -g 0 0 0 -h
0 0 0 0 0 0 0 0 0 0 0 0
0 -h 0 0 0 j 0 h 0 0 0 i
0 0 -h 0 -j 0 0 0 h 0 0
-i
Where:

6/582 -(1/10-1/2S)R2 (2/15+1/65+1/3; 2)R2

1 (1+ S) (1+ S)2
Chapter4: Gyroscopic Beam Element 65

)R2 12E1
-(1/30+1/65-1/6S2 R= I/A S-GA
(l+s)'- l2

4.4 SUMMARY
The mathematical formulation of an Euler-Bernoulli finite beam element that can

be used to model a shaft and includes gyroscopic effects has been realised. The

standard stiffness and mass matrix derivations of an Euler-Bernoulli beam

element are widely published in literature and have not been given here.
However, the full matrices are given in Appendix A for reference and the author

refers to Thomson, W. T., [1993] or Gasch, R., [1976] if the reader requires a
complete derivation. The gyroscopic matrix has also been previously published
in literature. However, an extensive search has not found the source of
derivation. Due to the importance of the gyroscopic effect in this work and the
fact that in Chapter 4 the principles used in this chapter are developed and

extended to create an element capable of modelling the gyroscopic effects of a

propeller, the full derivation has been given. The equivalent ANSYS gyroscopic

matrix has also been given as ANSYS has been used to verify the shaft

modelling code VIBRATIO into which this gyroscopic matrix has been
incorporated.
Chapter 5: Gyroscopic Propeller Element 66

CHAPTER 5

GYROSCOPIC PROPELLER ELEMENT MATRIX

DERIVATION

5.1 INTRODUCTION
This chapter presents the formulation of a gyroscopic propeller element using
Euler-Bernoulli theory. The element derivation is based on the same principles as

that of the standard beam element derivation shown previously in Chapter 4.

However, now the axis of rotation no longer lies along the axis of symmetry
belonging to the element, but is instead perpendicular to it. This change does not

affect the standard Euler mass and stiffness matrices, but for gyroscopic

behaviour the effect of changing this rotational axis is much more significant,

making the polar moment of inertia per unit length a variable within the
integration. The polar moment of inertia now increases quadratically with the

radial distance away from the axis of rotation. Thus the radial distance of the

element under consideration from the axis of rotation must be considered and

accounted for in the matrix derivation.

5.2 THEORY
5.2.1 Equations of Motion
As before in Chapter 4 the virtual work equation presented by Gasch, R., [1976]
is used to derive the equations of motion. The principle of virtual work was first
formulated by Bernoulli, J.J. and may be stated as follows: If a system in

equilibrium tinder the action of a set of forces is given a virtual displacement, the
virtual work done by the forces will be : ero. This is more easily understood as
the work done by the external forces acting on a system and the work done by the
Chapter 5: Gyroscopic Propeller Element 67

internal forces must vanish for a virtual displacement. The statement assumesan
imaginary virtual displacement given instantaneously, thus there are no

significant changesin geometry and the system is assumedto behave linearly.

Therefore:

aw=o (5.1)

As before (in Chapter 4) the virtual work aW for an Euler beam is summed up as
follows:
ý
t{oUnTEIu'dx+

8W =f au"' d; EI (ü" - flNu") dx + fuT/, üdx

1o

jc7UT
deüdx- Jeurk5Nudx JaUTCnIQgnUdX
+ -
+[auT inu + öu'T oaü' öu'TSZo Nn'], }
-

au' (CS2u + Bnü) + {-522 SauTpcdxcost

faUT, nt [_f22CjUTM46c c2t cit], }

-522 ucsdxsin + cos -S22öfTme, sin (5.2)

From which the gyroscopic virtual work is defined as:

aWG J[auT)ONü}, (5.3)

5.2.2 Gyroscopic Propeller Element Derivation

Gyroscopic bending moments are proportional to the angular velocity of an

element. The moments are also coupled such that for an element spinning about
the x-axis, a positive rotation in orientation about beta would produce a negative

moment about gamma. Note that energy is not being removed from the system
by this action, but the moments are simply transposed from one axis to another.
However, becausethis action is proportional to angular velocity the gyroscopic

matrix is added to the system damping matrix, and for this reason is often
Chapter 5: Gyroscopic Propeller Element 68

referred to as the gyroscopic damping matrix even though it is not damping in the

conventional sense. In Equation 5.3 this velocity dependence appears as the

ii term, which is defined as follows:

Z )'o Mio ro ý5 3'i ý«A Y) (5.4)

-X . i

For the purposes of this analysis the cubic element shape functions Equations

4.4-4.8 as shown in Figure 4.1 have been used.

The equations of motion for the gyroscopic effect can now be derived for a single

element using the virtual work principle defined earlier:

aWG
=_f [auTcooNu'], (5.5)

Since:
au'(x, t) = V'r (x) " au, (t) (5.6)

And:

u(X, t) = VT (X). ü, (t) (5.7)

Thus the gyroscopic virtual work for an element can be expressedas:

I
awý,-au' $[v'coNv1T]
ax.
u (5.8)
o 0

Since the rotational speedof the shaft is a constant within the integration, this

can be expressedas:
1
öWG, = -öue S2f [V'OpNV'r ] äx ü, (5.9)
"
0

Expressing Ge as:

I
Ge, = fl J[V'6pNV'T] 8x (5.10)
0
Chapter 5: Gyroscopic Propeller Element 69

Then:
aWGý Ge
vv
=-aue . "Ue
(5.11)
Where GQPisnow defined as the gyroscopic beam element matrix for a propeller.

In order to describe a propeller-shaft system mathematically, it is first necessary

to choose an appropriate axis system. For this study it has been chosen that the

shaft should lie along the global X-axis, whilst the propeller blades radiate
perpendicularly from the shaft. However, within their local axis systems the

propeller and shaft elements can be made to fit any axis system the author

chooses, so long as the appropriate matrix transformation is performed during

global assembly stages. This can even be taken so far as to both propeller and

shaft elements sharing the same local axis system, only to be placed

perpendicular to one another on assembly. Thus the propeller and shaft elements
are described in the same local axis system, where the element's axis of
symmetry is about the x-axis. It is possible to create sub-assemblies for each
shaft or propeller blade from however many elements are required before

combining them into the global system. This can reduce assembly complications
during programming, so that the main consideration is the radial orientation of

propeller elements about the X-axis in order to allow for multi-blade systems.
However, this is not a prerequisite and it is equally possible to assembleelements
into the global matrix in any order the reader chooses.

For a twelve degree of freedom system where x is the axis of rotation but the

element's axis of symmetry lies vertically parallel to the y-axis, x and gamma are
coupled, and z and alpha are coupled. This geometry is shown in Figure 5.1 and

consequentlythe matrix of element shape functions VT becomes:

00h, 0 h2 000h, 0 h4 0V0h000

(5.12)
-h2 0 h, 000 -hs
Chapter 5: Gyroscopic Propeller Element 70

Figure 5.1: Local Axis System

Note: a, b and s are global distances from the axis of rotation

In order to account for the polar moment of inertia per unit length, each propeller

blade element is treated as being parallel to the x-axis when attributed to its own
local axis system. Therefore, the radial distance to the element under

consideration can always be found, whilst coordinates and an angle can be used
to determine the orientation of the various propeller blades to the global axis
system. The local coordinate system is shown in Figure 5.1 above.

Now since the polar mass moment of inertia per unit length is defined as:
op 1
= us (5.13)
Where is the massper unit length. And s is the radial distance to the elemental
,u
massunderconsideration.

We can expressthe gyroscopic beam element matrix as:

I
Gep = f [Vts2NV'T
] dx (5.14)
F'
0
Chapter 5: Gyroscopic Propeller Element 71

Given that ý=x1 for the element under consideration, then from the Figure 5.1

it can be seenthat:

s =a+ ýl (5.15)

s =a+ ý(b - a) (5.16)

s= a(1-ý)+bý (5.17)

s2 = a2 + ý(2ab 2a 2) + ý2 (a 2 +b2 -2ab) (5.18)

-

Here:

of _a (5.19)

Given that:
ah ah aý
_ (5.20)
ax aýax

And:

(5.21)

Then it can be seenthat:

ah_"1ah 1 (5.22)

Thus the integrals for the various matrix terms are as follows:
'1, 6(2a2+3ab+2b2)
aX Jyý, 0, ax -4ab+2b2)
s21, = 35 s =1(-5a2
ý' 70
oo
1
fhs2/4 ax -1(-2a2 +4ab+5b2) JI4s2hx 12(9a2+3ab+2b2)
=
70 105

Jh,'s2148x -6(2a2 + 3ab+ 2b2) Jhs2h3 ax -1(2a2 -4ab - 5b2)

= 35 2 =
0 70
0
 Chapter 5: Gyroscopic Propeller Element 72

+ab+3b2) Ns2hä öx 1(5a2+ 4ab - 2b22)

f hzs2h4ax -12(3a2
= 210 70
0 0
Is2 6(2a2 +3ab+2b2) Jhäs2hä8x +3ab+9b2)
ax =12(2a2 105
_
o35
o 0

(5.23) to (5.32)

Now multiplying out equation 5.11 it can be seenthat:

0 0 0 0 0 0 0 0 0 0 0 0
0 0 '
's2h, 0 's2h2
h, 0 0 0 h,'s2h; 0 h,'s2h; 0
-h,
'
0 -h,'s2h, 0 0 0 h,'s2h2 0 -h, 's2h; 0 0 0 h;s2h,
0 0 0 0 0 0 0 0 0 0 0 0
0 -h2s2h, 0 0 0 hZs2h2 0 -h s2h; 0 0 0 h2s2h;
0 0 ' 0 0 0 0 -h2s=h; 0 -h2s=h; 0
-h2s2h, -h2s2h2
V's2NV'T - 0 0 0 0 0 0 0 0 0 0 0 0
0 0 k's2h1' 0 h's2hz 0 0 0 hys2h3 0 h4s2h4 0
0 -h;
'
s2h, 0 0 0 h3s2h2 -h, s2h; 0 0 0 0 h;s=h;
0 0 0 0 0 0 0 0 0 0 0 0
'
0 -h; s2h, 0 0 0 h;s2h2 -h; S2h 0 0 0 0 h,s=h;
0 0 '
s2h, 0 -h, s2h2 0 0 0 -h4's2h; 0 -h4s=h; 0
-h;
(5.33)

5.2.3 Gyroscopic Propeller Element Matrix

Inserting the terms 5.23 to 5.32 into the relevant positions in matrix 5.33, it can
be seen that the gyroscopic beam element matrix for a propeller blade

perpendicular to the axis of rotation with distances a and b defined in Figure 5.1
is as follows:
Chapter 5: Gyroscopic Propeller Element 73

0 0 0 0 o o 0 0 0 0 0
36(2a2+
0 -3/(-2a2 +
0 0 0 -3G(2az+ 0 31(-5a=- 0
D 0
3ab+2b2) 4ab+5b2) 3ab+2b') 4ab+2b=)
36(2a'+ 31(-5a' -
0 -36(2a2 + 0 0 0 -31(-2a2 + 0 0 0 0
3ab+2b') 4ab+5b2) 3ab+2b2) 4ab+2b2)
0 0 0 0 0 0 0 0 0 0 0 0
31(-2a2+ 212(9a2+ 31(2a2- -12(3a +
0 0 0 0 0 0 0 0
4ab+5b2) 3ab+2b2) 4ab-5b2) ab+3b=)
3/(-2a2 + 212(9a' + 3/(2a /_( +
0 0 0 0 0 0 0 0
4ab+5b) 3ab+2b') 4ab-5b=) ab+3b=)
G`' In
210 0 0 0 0 0 0 0 0 0 0 0 0

-36(2a2 + -3l(2a2 - 36(2a2+ 31(5a2+

0 0 0 0 0 0 0
3ab+2b) 4ab-5b2) 3ab+2b2) 4ab-2b2)
36(2a2+ -36(2a2 + 31(5a2+
0 0 0 0 -31(2a' - 0 0 0 0
3ab+2b2) 4ab-5b2) 3ab+2b2) 4ab-2b2)

0 0 0 0 0 0 0 0 0 0 0 0
-12(3a2 + -3/(5a2 + 2/I(2a2 +
0 -3/(-5a2 - 0 0 0 0 0 0 0
4ab + 2b2 ) ab+3b') 4ab-2b2) 3ab+9b=)
/''(3a2+ -31(5a1+ -21=(2a=+
0 0 -31(-5a2- 0 0 0 0 0 0
4ab+2b) ab+3b2) 4ab-2b2) 3ab+9b7)

(5.34)

5.3 SUMMARY
A novel technique that creates an Euler-Bernoulli finite element that can be used

to model a rotating propeller blade including the gyroscopic effects has been

realised. The method developed to include gyroscopic effects in propeller models

is similar to the derivation of the standard gyroscopic matrix given in Chapter 4.

The key difference is the inclusion of the polar moment of inertia per unit length

as a variable within the integration. This approach and the resulting gyroscopic

matrix have not been previously published.

The main stiffness and mass matrices do not change from that of a standard
Euler-Bernoulli shaft element and have been given in Appendix A. With the

mass, stiffness and gyroscopic coupling effect accounted for the modelling of a

propeller blade is almost complete. However, during operation there is additional

centrifugal stiffening imposed on a propeller rotating about an axis perpendicular
to its axis of symmetry. These stiffening effects are investigated separately in
Chapter 6 since they are in addition to the standard matrices given here and do

not replace them. The mathematical formulation used to incorporate the

Chapter 5: Gyroscopic Propeller Element 74

gyroscopic effect created by propeller deflections could be applied to the

derivation of a Timoshenko finite element if required. However, as the

gyroscopic effect is itself quite small the difference between a Timoshenko and

Euler-Bernoulli model of the effect will be negligible.

Chapter 6: Centrifugal Stiffening Propeller Element 75

CHAPTER 6

CENTRIFUGAL STIFFENING OF PROPELLER ELEMENT

MATRIX DERIVATION

6.1 INTRODUCTION

Propeller elements are invariably subjected to axial forces under rotation. These

axial forces are created by the angular acceleration the blade undergoes during

rotation. So that the axial force in a propeller element will be greatest at the
beam's root as this point has to exert enough centrifugal force to support the

angular acceleration of the entire blade, the force will fall to zero at the beam's

tip. Even for a uniform beam element, the change in force from root to tip will

not be linear because the force is proportional to both the mass and radius

squared. Hoa, S.V., [1979] produced a paper titled: "Vibration of a Rotating

Beam with Tip Mass", which contains a method for deriving the stiffening effect

on a uniform propeller element due to centrifugal forces. The paper also includes

provision for the additional stiffening effect due to a blade tip mass.

Here the effect of centrifugal stiffening has been considered to warrant

investigation for possible inclusion in the formulation of an accurate propeller

element. It is assumed that for a heavy element or one rotating quickly the
centrifugal stiffening effect has the potential to make significant changes to the

ensuing bending mode shapes and frequencies. Furthermore, the full derivation
has been included because discrepancies were found with results published by
Hoa, S.V., [1979].
Chapter 6: Centrifugal Stiffening Propeller Element 76

6.2 THEORY
6.2.1 Equations of Motion

Under rotation radial beam elements (propeller elements) are stiffened due to the

stressescreatedby centrifugal forces. For a beam which lies parallel to the y-axis
and rotates about the x-axis, the centrifugal forces create the stresses

o and z, in the neutral surface. The following strain energy equation proved by

Kapoor, K. K., and Hartz, B. J., [1966] shows that the strain energy stored in the

element is higher than the bending strain energy by the amount:

aw"=2 1{6x(au/ax)'+2z,
(au lay)}dv
lax)(Ott (6.1)
v

In the caseof slenderbeams, on/ c3y=0 thus the equation simplifies to:

awcs f {ax(2u1ax)2}
dyd (6.2)
=i

6.2.2 Centrifugal Stiffening Propeller Element Derivation

The stress ßx can be calculated as 6x = F, /A where Fxis defined as the axial

r
force acting on any section at a distance x from the inner edge of the element,

and A is the area of element this force is applied to.

I, I u1C W. A. VtRl111CIIy
Chapter 6: Centrifugal Stiffening Propeller Element 77

E1 >IE / ýIHIE H >IHIE H

r

ni"1
0

n"l x

Figure 6.2: Element Location

Given the geometric definitions shown in Figures 6.1 and 6.2, where nt is the

total number of elements and n is number of elements before the element under

consideration, it can be seen that the force acting on any section at the distance

y from the element's end is:

I nil
f
Fx=A pn'(r+nl+x)dx+A f 2(r+xg)dx& (6.3)
pc
x (n+l)I

Equation 6.3 can be solved to give the axial stress, Qx as:

FYA (x2 + 2x(l n+ r) -1(l (ni2 - n2) + 2r(m n)))

tTx =_ 1o12 - (6.4)
x2

As before in Chapters 4 and 5, the displacement u is approximated by a third

order polynomial (As with Equations 4.4-4.8):

u=c, +c2x+c3X2+c4X3 (6.5)

Expressing this displacement in matrix form, where V(x) is the matrix of

polynomial shapefunctions and u, are the corresponding coordinate

displacementsexpressedin matrix form, gives:

u(x, 1)= VT(x). ue(t) (6.6)

Chapter6: CentrifugalStiffeningPropellerElement 78

Using the standard bending shape functions h, to h4described in Chapter 5

gives:

h, =1-3e2+2s3 & h2 =-1(s-2c2+c3) (6.7) & (6.8)

h3= (3E2 2s3) & h4 = -1(-s2 + c3) (6.9) & (6.10)

-

Hencethe matrix of shapefunctions for this geometry becomes:

0 h2 000h, 0 h, 0
VT -00h, (6.11)
01000 -h2 0 h, 000 -h4

Where the displacementsare defined as:

'UNH-To YozrogAYoxqY,; gAY. )

(6.12)

Now given that:

aWCs=2 J{Qx(au / ax)2}dxdyd:

(6.13)

And:

u'(x, t) = V, T(X). U.(t) (6.14)

Then it can be seen:

M, ! ue(t)"K, - ue(t) (6.15)

H

Where:

Kew = faVVTdxdyd:
(6.16)
0

Or:

Keo =A jcTxV'V'Tdx
(6.17)
0
Chapter 6: Centrifugal Stiffening Propeller Element 79

Thus the elemental centrifugal stiffness matrix Ken becomes:

0 0 0 0 0 0 0 0 0 0 0 0
0 v, h;h1' 0 0 0 0 a, l; h; 0 0 0 -x, 1411;
0 0 Qkk 0 6, kk 0 0 0 o h1'h 0 Q, l; liq 0
0 0 0 0 0 0 0 0 0 0 0 0
0 0 CA/4 0 a, 1414 0 0 0 Q, J4h; 0 a. 1, h, 0
0 -C k/4 0 0 0 6, h;hZ 0 -Ql1Lh 0 0 0 chh;
K,. A
0 0 0 0 0 0 0 0 0 0 0 0
0
0 a, 1414 0 0 0 -a14h3 0 a; 13h; 0 0 0 -Q; h; h,
0 0 a, 1414 0 6rh2h3 0 0 0 Q, h3h; 0 cr 11h; 0
0 0 0 0 0 0 0 0 0 0 0 0
0 0 6rh, 'h; 0 a, 14h; 0 0 0 Q, 1411" 0 c, h, h; 0
'
0 -a 14h,
,
0 0 0 Q=loh; 0 -a, 14h,' 0 0 0 c h;hj

(6.18)
Where Qx has been previously defined in Equation 6.4.

Now using the previous differential definition shown in Chapter 5 and given as

follows:
ah ah"1
= al X6.1ýý

The integrals are defined as:

fa. ,, pn23[l(7m2 -7n2 - 7n - 2) + 7r(2ni - 2n - 1)]

(6.20)
,J 35
0

fa,, kkdx p)221[1(7m2- 7n2 -14n-5)+14r(ni-n-1)]

(6.21)
140
0

pQ23[l(7m2- 7n2-7n-2)+7r(2m-2n-1)]
(6.22)
"dx
6X = 35

jah'h: dx
--p21[l(7m2-7n2+2)+14r(m-n)] (6.23)
,,
o140

frhI4cix
I- pcI212[1(14m2
-14n2 -7n-2)+7r(4ni-4n-1)] (6.24)
210
0

pc221[1(7m2
- 7n2 -14n - 5) + 14r(m -n
aXhi14dx_ -1)] (6.25)
140
0f
Chapter 6: Centrifugal Stiffening Propeller Element 80

Johhdx PS1212[1(7rn2
-7n2 -7n-3)+7r(2m-2n-1)] (6.26)
0
420

pf123[I(7m2 -7n2 -7n-2)+7r(2m-2n-1)] (6.27)

6x _
35
of
2l[1(7m2
JQ h'h'dx -7n2 +2)+14r(ni-n)] (6.28)
x34 = 140
0

212[1(14m2-14n2-21n-9)+7r(4m-4n-3)]
ja h'h'dx= (6.29)
x44 210
0

6.2.3 Centrifugal Stiffening Propeller Element Matrix

The centrifugal stiffening matrix for a rotating propeller can more simply be
describedin matrix form as:

0 0 0 0 0 0 0 0 0 0 0
0 A 0 0 0 -D 0 -A 0 0 0 -E
0 0 A 0 D 0 0 0 -A 0 E 0
0 0 0 0 0 0 0 0 0 0 0 0
0 0 D 0 B 0 0 0 -D 0 F 0
0 -D 0 0 0 B 0 D 0 0 0 F
K=ea p2l
0 0 0 0 0 0 0 0 0 0 0 0
0 -A 0 0 0 D 0 A 0 0 0 E
0 0 -A 0 -D 0 0 0 A 0 0
-E
0 0 0 0 0 0 0 0 0 0 0 0
0 0 E 0 F 0 0 0 -E 0 C 0
0 -E 0 0 0 F 0 E 0 0 0 C

(6.30)
Where:
3[1(7m2-7n2 -7n - 2)+7r(2m 2n
A_ - -1)]
35

12[1(14x2-14n2 -7n-2)+7r(4m-4n-1)]
B=
210

C -12[l(14r2 -14n2 -2In-9)+7r(4m-4n-3)]

210
Chapter 6: Centrifugal Stiffening Propeller Element 81

D -7n2 -14n-5)+14r(m-n-1)]
--1[1(7nt2 140

-7n2 + 2) +14r(m - n)]

E --1[l(7m2
140

I2[1(7n,2-7n2 -7n-3)+7r(2m-2n-1)]
F_-
420

6.3 SUMMARY
The mathematical formulation of an Euler-Bernoulli finite element, that can be

used to model the centrifugal stiffening effects of a rotating propeller beam, has
been achieved. The resulting matrix does not differ greatly from that published
by Hoa, S.V., [1979]. However, most of the terms within the matrix have at least

one or two signs that have been published as negative instead of positive and vice

versa. These differences are caused by a minor publishing/integration error

within the original work whereby d in Table 1 should have been expressed as d=

(R'+ n). It should also be noted that this stiffness matrix is in addition to the
-
standard Euler-Bernoulli bending, axial and torsional stiffness matrices and does

not replace them. This is obviously so since the matrix is dependant on omega

squared, and thus without propeller rotation all terms will equal zero. The
standard Euler beam stiffness matrices are given in Appendix A, but have not
been derived since they are widely published.

Handelman, G., Boyce, W., and Cohen, H., [1958] and also Kumar, R., [1974]
have investigated the centrifugal stiffening effect. Concluding that the increase in

natural frequencies caused by the increase in stiffness is most significant for the
first mode of vibration. An investigation into this claim with published results
has been included in Chapter 9.
Chapter 7: Gyroscopic Shaft Element Results 82

CHAPTER 7

RESULTS 1: GYROSCOPIC SHAFT ELEMENT

7.1 INTRODUCTION
The Euler beam derivation in Chapter 4 describes the gyroscopic effect of an

element bending away from its axis of rotation. Assuming a shaft to lie along the

x-axis and be excited in the vertical y-plane (either through y or Gamma) without

rotation one would expect to see displacements in this vertical plane only. Even

if rotation is included in the model this will still be the case, unless the

gyroscopic effects have also been accounted for. Only when the gyroscopic

effects are included will the displacements in the vertical plane become coupled

with the horizontal plane and induce vibrations in this direction also. This
coupling effect also changes the natural frequencies system, thus where the
determination of the exact natural frequency is critical it is vital that gyroscopic

effects are included.

In order to verify the computer shaft modelling procedure including gyroscopic

results, analyses have been performed on a large ship drive shaft simulation. This
system was chosen becauseof a working collaboration with Lloyd's Register of
Shipping during which VIBRATIO was verified against results produced by
ANSYS 5.1 and Lloyd's experimental data. Since the ANSYS results have been

validated a direct comparison between these and VIBRATIO results is possible.

The ship drive shaft is a highly complex model which makes differences in

results immediately apparent. For shaft element validation an analysis has been
performed omitting the possible disc or propeller attachments. This method is
chosen in order to ensure that the relatively small gyroscopic effects of the shaft
Chanter 7: Gvrosconic Shaft Element Results 83

elements are not disguised by the large gyroscopic forces associated with such an

attachment. This has also been done because of difficulties modelling such a

system in ANSYS 5.1. Since ANSYS does not use the hybrid rigid body

modelling technique employed here (Chapter 3), a rotational point inertia cannot
be simply attached to the free end. Instead a very wide beam element of the same

inertia has to be used which in turn creates an ill-conditioned stiffness matrix.

For the same reason the axial displacement spring has not been offset from the

shaft but is instead treated as being at the node.

7.2 LLOYD'S SHIP DRIVE SHAFT

A model of a ship's drive shaft has been created in both ANSYS and Vibratio;

the shaft is 48.01 metres long and manufactured from steel of density 7800
kg/m3, Young's Modulus 2.07 x 1011 N/m2, and a Poisson's ratio of 0.3. The
initial 32.5 metres of shaft connected to the engine is 0.47 metres in diameter, the

centre 14.8 metre section is 0.55 metres in diameter with a 0.15 metre diameter
hollow centre, and the final 0.71 metres of shaft that would normally be

connected to the propeller has a diameter of 1 metre. There are nine support

positions along its length, but the propeller end is unsupported and free. The

shaft itself rotates at a frequency of 40 Hz. At the opposite end of the shaft to the
free end there is a simulated engine. The engine is made from ten rotational

inertias, each connected with torsion springs, where the final (tenth) inertia is

attached to the shaft. Six of the inertias (numbered four through to nine) are

harmonically excited with real forces proportional to omega, where omega is the

engine frequency. The harmonic excitation is torsional acting along the axis of
the shaft. The propeller attachment point is also harmonically excited at a
frequency proportional to omega squared in the vertical plane. Figure 7.1 below

shows this in more detail.

Chapter 7: Gyroscopic Shaft Element Results 84

NOT TO SCALE j'2 --- CC--100 kN. s/m

lm Diameter 0.71m Cy=100 kN. s/m
C, =100 kN. s/m
2
0.75m
Ky=750 MN/m
2" -----------
-ý- KZ=480 MN/m
0.75m
Cy=100 kN. s/m
0.55m Out Diameter 12° CZ 100 kN. s/m
0.15m In Diameter 11.8m

0.75m Ky=750 MN/m

KZ 600 MN/m
2"0.75 Cy 100 kN. s/m
m
Cý 100 kN. s/m
8

8.56 m

1 14
5

4.96 m

5.21 m

0.47m Diameter
s
Shaft Dimensions 5.71m Lateral y and
E= 207 GPa z deflections
'ý-ý-- fixed at
s
p= 7800 kg/m3 4.96m nodes.
Poisson's Ratio = 0.3
----- ---- K, 7-490.5MN/m
3
3.10 m

Number of
--
elementsper ta- 61.98MN.mhad
section. 1. -6.437 0 kg. m2
-_----_-_-_-.
Ka- 86.55 MN. m/rad
I, - 0.3020 kg.m2 ----------ý-
a,
K. - 48.46 MN. m/rad
I. - 0.1349 kg.m2
----------- -
K, - 48.46 MN. m/rad

Harmonically 1ý- 0.2962kg.m2 -------------

excited engine I. - 48.46
MN.mirad
inertias. 1ý- 0.2963kg.m2 ---- -_N _-- All degreesof
Ka- 48,46 MN. mhad freedom fixed
1..- 0.1530kg.m2 ---------_ at nodes except
K. - 48.46 MN. mhad for axial
I; 0.2990 kg-m2
rotation alpha.
K, - 78.80MN.mhad
1. - 0.0539kg.m2 ---- M _-_--
K, -15.59 MN. m/rad
Imo-0.3360kg.m' ------_ý____
K. - 15.59MN mhad
Iý- 0.4295 kgm= -------------

figure 7.1: Lloyd's Drive Shaft Model

Chanter 7: Gyroscopic Shaft Element Results 85

7.3 BEAM ELEMENT RESULTS

The following results compare an ANSYS model against a geometrically
identical one developed in VIBRATIO using the gyroscopic beam element

formulation derived in Chapter 4. ANSYS has of course been extensively

verified by other means, so is a recognised method to verify elements. Unlike the

Euler beam elements developed for this study ANSYS uses Timoshenko beam

theory and its own gyroscopic matrix that is presented in Chapter 4.

Unfortunately, no reference as to how this gyroscopic matrix is derived is given

in the ANSYS theory manual and despite an extensive literature review its source

has not been found. However, as it has been extensively tested and verified
including an experimental model for this particular problem it is assumed that the

ANSYS results are correct.

For this type of model VIBRATIO is capable of producing many different results

including displacement, velocity, acceleration, stress and eigenvalue.

Eigenvalues have not been compared for this model because the forced vibration

results are superior, checking both the natural response frequencies and mode

shapes simultaneously. The displacement values are published since these are the

simplest to interpret. Stress results have not been compared since they rely on
further mathematic calculations that may differ from those used by ANSYS.

VIBRATIO produces displacement results for each degree of freedom at each

shaft node. The Lloyd's shaft has nineteen nodes giving one hundred and
fourteen different sets of results to compare with the ANSYS model. Results

correlate with those produced by ANSYS so it is pointless publishing all the

available graphs. Thus, only the six degrees of freedom belonging to the free
(unsupported) propeller end are shown in Figures 7.1 to 7.6. However it can be
taken as read that results at other shaft positions have equally good or better

correlation with ANSYS than those of the free end.

It can be seen that correlation between results is almost perfect with those

produced by ANSYS. However, for lateral deflections there is a slight frequency

shift in responseof approximately 0.5Hz towards the upper end of the frequency
Chanter 7: Gyroscopic Shaft Element Results 86

analysis. This shift increases with frequency and is a known phenomenon that

exists between Euler and Timoshenko modelling techniques. Davis, R., et al.,
[1972] previously noted the tendency for Euler theory to overestimate vibration

frequency increases with the ratio of beam depth to the wavelength of vibration.

Displacements for both models can be seen to be of the same magnitude.

Furthermore the reader should note the small blips in the graphs around 36 Hz

(vertical) and 44 Hz (horizontal). This small detail is of interest because entirely

different modelling techniques reproduce the same minor disturbances on a

complex system.

Figures 7.2 and 7.6 are very similar as are those of 7.3 and 7.5. This is due to the
direct relationship between the vertical deflection and corresponding Gamma

rotation, and the same relationship that exists for the horizontal deflection with

Beta. However despite only the gyroscopic effect coupling the vertical and
horizontal responses, the two graphs are not of the same shape as may be

expected.

%31aLu 1.1; t-t ial it LISplace1nent ui Jnait L' na

Chapter 7: Gyroscopic Shaft Element Results 87

--- -r-- " --- " -- ------ -- -ýr--------_.......... ».......,

Chapter 7: Gyroscopic Shaft Element Results 88

"I apu / .?. awaaawiiai r%ipIIa 1'I Ia% IIIvut "I ouait ruu

v. ulsu . -. aw auawuu Lb{u ai.. r. u'WlIllcul VI 1511411 GIIU

Chapter 7: Gyroscopic Shaft Element Results 89

Graph 7.6: Rotational kiamma Displacement vi matt End

7.4 SUMMARY
This chapter presents results demonstrating and verifying the Euler beam element
derived and shown in Chapter 4. The verification of a gyroscopic shaft element is

straightforward, with much research having been previously performed in this

area. The method chosen was comparison against ANSYS results using an
ANSYS model that had previously been verified by Lloyd's Register of Shipping

against experimental data.

Using the Lloyd's model of a ship's drive shaft it can be seen that results using

the gyroscopic shaft elements show very good correlation between ANSYS and
VIBRATIO formulations. Given that the two formulations use completely
different theory (ANSYS uses Timoshenko and VIBRATIO uses Euler) and

solvers, these results are exceptional. This particular drive shaft has been chosen

as a reliable benchmark, because the problem's complexity immediately

highlights differences, and also previous corroboration with Lloyd's Register of

Shipping has verified results. Since the ANSYS simulation has been verified by

Lloyd's Register of Shipping against a full model it is reliable to conclude that

the VIBRATIO gyroscopic shaft formulations and computer coding are correct.
Chapter 7: Gyroscopic Shaft Element Results 90

As such further tests have not been performed on this gyroscopic shaft element
formulation.

It has previously been published that Timoshenko theory is superior to Euler

theory for thick beams, for example; Lee, C. W., [1993]. However these results

show that for shafts up to 0.5 metres in diameter rotating at up to 40 Hz Euler

theory achieves almost identical results. Being such a long shaft one may initially

consider it to still be slender however with so many support positions along its
length this is not really the case. Although further work could be performed on a

Timoshenko model to highlight the differences, the aim of deriving and verifying

an Euler shaft model with gyroscopic capabilities has been realised, and results

are considered accurate.

Chapter 8: Gvrosconc Propeller Element Results 91

CHAPTER 8

RESULTS 2: GYROSCOPIC PROPELLER ELEMENT

8.1 INTRODUCTION
This chapter presents results of the verification of the gyroscopic element
formulations. Perhaps the most obvious method of verification for a propeller

element is comparison to experimental results. Unfortunately this can also be the

most problematic technique. The main difficulty of using an experiment for

verification is experimental error which can never be fully eliminated. Even

assuming this error could be reduced to less than five percent, its effect may still
be greater than the gyroscopic effects. Thus, at best experimental methods could
be used to verify trends. Thus experimental results have not been used and a

computational technique is employed. Eigenvalue and forced frequency results

have been compared against theoretical values for simple cantilevers and then a

parametric study has been performed.

8.2 EIGENVALUE RESULTS

Results have been obtained for simple cantilever beams since the natural
frequencies are widely available. The beams have then been treated as rotating

about an axis perpendicular to their axis of symmetry as in propellers for

comparison. Using Euler beam theory Thomson, W. T., [1993] shows the first

three modes of vibration can be calculated as follows:

El
(O ±(flnl)2 (ö. 1
Chapter 8: Gyroscopic Propeller Element Results 92

Where for a cantilever beam the following coefficients are given:

Fundamental (Al)" = 3.52

SecondMode (/3,1)2= 22.0

Third Mode (ß, 1)2= 61.7

Figure 8.1 below depicts a typical rectangular cross-section of a beam element so

that ly and 1. can be calculated. For a beam of differing width and thickness the

natural frequencies will not be the same in each direction as J. and 1: will be
distinct.

thickness ---==- "-"="-"-- -"-z

II

width

Figure 8.1: Rectangular Cross-Section of a Beam Element

lvidth3 x thickness
Iy = (8.2)
12

width x thickness3
I- (8.3)
` _ 12

Since the massper unit length increases linearly with the width and thickness in
Equation 8.1 the natural frequency about ly is independent of the beam thickness,

and the natural frequency about 1_is independent of the beam width. This is
only
true for a beam of rectangular cross-section. Thus the following values given in
Table 8.1 are for various beam thicknesses only. This of course is not true when

gyroscopic coupling is considered since the natural frequencies about 1y and 1;

become coupled. For simplicity each beam is divided into three
elements of
Chapter 8: Gyroscopic Propeller Element Results 93

equal length. The total length is maintained at one metre in all cases, and the
slendernessratio is given since this is the ultimate determinate of the natural
frequency. The propeller is fixed at the root as in the case of a cantilever. The

material is steel where Young's Modulus = 207 GPa, Poisson's Ratio = 0.3, and
Density = 7800 kg/m3. Natural frequencies are given in radians per second. One

can see that with error values of just 1.24 % for the third natural frequency of

vibration there is little point in increasing the number of finite elements further
for this type of test.

Thick Slender Exact Exact Exact PC PC PC Err. Err. Err.

-ness -ness (01 (02 (03 Cul W2 (03 % % %
(m) Ratio
w1 w2 C03
0.01 100 52.3 327 918 52.3 329 929 0.10 0.49 1.24
0.02 50 105 654 1835 105 658 1858 0.10 0.49 1.24
0.04 25 209 1309 3670 209 1315 3716 0.10 0.49 1.24
0.10 10 523 3271 9175 523 3287 9289 0.10 0.49 1.24
[able if. t: Ligentrequencies for Standard Euler Beams

Not surprisingly the inclusion of gyroscopic effects (assuming there is a propeller

rotation speed) complicates the eigenfrequencies somewhat. Due to coupling in

the damping matrix the width and thickness are now not independent of each

other for eigenfrequencies in each axis of symmetry. Thus the width and
thickness must be considered together. In order not to complicate the problem

unnecessarilyinitial results are for systems where the thickness and width of the
beamsare the same.

Table 8.2 (below) shows the effect of adding a propeller rotation speed to the

eigenfrequenciesof two simple propeller beams. Where propeller rotation speed

is included the first six natural frequencies have been provided. The
reason for
this is the tendency for the gyroscopic effect to `split' the natural frequencies.
This is a normal gyroscopic phenomenon that results clearly show in Table 8.2.
For example the first natural frequency of the 0.01 x 0-01m propeller is 52.3

rad/s. Once the propeller speed is set to 250 rpm this eigenfrequeney
splits into
Chapter 8: Gyroscopic Propeller Element Results 94

two eigenfrequencieseither side of the original, 45.4 rad/s and 59.0 rad/s. This
trend continues throughout results and can be seen to be true for the second and
third eigenfrequenciesalso. Another point the reader should note is that as the
beam becomes thicker (and therefore stiffer) the splitting of eigenfrequencies
becomes less pronounced despite the associated increase in inertia. This can be

stated since although the numerical difference is similar the eigenfrequencies are
higher so the percentagechangeis smaller.

Thickness Width Speed PC PC PC PC PC PC

(m) (m) (rpm) (01 (02 w3 (04 (05 (06
0.01 0.01 0 52.3 329 929
0.01 0.01 250 45.4 59.0 253 400 655 1116
0.01 0.01 500 39.0 65.2 194 462 484 1134
0.02 0.02 0 105 658 1858
0.02 0.02 250 97.7 111 581 731 1584 2070
0.02 0.02 500 90.9 118 507 800 1311 2231
0.04 0.04 0 209 1315 3716
0.04 0.04 250 202 216 1239 1390 3454 3944
0.04 0.04 500 195 223 1162 1462 3168 4139
0.10 0.10 0 523 3287 9289
0.10 0.10 250 516 530 3212 3363 9037 9528
0.10 0.10 500 509 537 3135 3438 8772 9752

Fable tß.2: Eigentrequencies for Square Gyroscopic Euler Propeller

The interaction and `splitting' of natural eigenfrequencies due to gyroscopic

coupling is however much harder predict intuitively for systems where the width
and thickness of the propeller element are not the same.This is due to interaction
with two sets of eigenfrequencies which differ in ly and 1,. Table 8.3 shows
results for such a system.

The first three eigenfrequenciesin Iy (for a 0.1m thick beam) will in fact be 523,
3287 and 9289 rad/s respectively as shown in Table 8.1, however these do
not
necessarily appear as the higher order eigenfrequencies in Table 8.3 since the
Chapter 8: Gyroscopic Propeller Element Results 95

fourth natural mode in 11 is often achieved first. The `splitting' of

eigenfrequencies is no longer seenas was the case in the square propeller blades;
instead the eigenfrequencies tend to shift to a higher or lower value. Results

show this shift to be intuitively unpredictable, so it is vital as analysis such as this

is performed. As before as the beam thickness and stiffness increase the change
in eigenfrequenciestends to become less pronounced.

Thickness Width Speed PC PC PC PC PC PC

(m) (m) (rpm) o w2 0)3 (04 w5 C06

0.01 0.10 0 52.3 329 523 929 2092 3288

0.01 0.10 500 52.2 321 528 919 1976 3134
0.01 0.10 1000 51.9 300 541 889 1735 2934
0.01 0.10 1500 51.5 274 554 849 1535 2803
0.02 0.10 0 105 523 658 1858 3288 4184
0.02 0.10 500 104 512 667 1821 3241 4089
0.02 0.10 1000 104 487 688 1717 3098 3933
0.02 0.10 1500 103 456 73 1577 2916 3856
0.04 0.10 0 209 523 1315 3288 3716 8368
0.04 0.10 500 209 522 1309 3210 3767 7767
0.04 0.10 1000 207 520 1293 3038 3867 6882
0.04 0.10 1500 205 517 1266 2837 3967 6154
Table 8.3: Gyroscopic Eigenfrequencies for Rectangular Euler Propeller

Furthermore since the width and thickness of the beam are no longer independent

of one another it is not easy to tabulate results according to the slendernessratio

as is the case for centrifugal stiffening (Chapter 9, Table 9.2). This is clearly
demonstrated below in Table 8.4 where a steel cantilever blade of the
same
thickness but differing width shows different eigenvalues for the first three

modes of vibration about J. For a stationary propeller these three modes of

vibration are associatedonly with the thickness of the blade. It can be seen that
Chapter 8: Gyroscopic Propeller Element Results 96

the thinner the width the greater the effect becomes as the blade becomes more
flexible and natural frequencies in both axis become closer together.

Thickness Width Speed PC PC PC

(m) (m) (rpm) wt 0)2 w3

0.01 0.075 0 52.3 329 929

0.01 0.075 500 52.1 308 903
0.01 0.075 1000 51.6 273 833
0.01 0.075 1500 50.8 239 748
0.01 0.100 0 52.3 329 929
0.01 0.100 500 52.2 321 919
0.01 0.100 1000 51.9 300 889
0.01 0.100 1500 51.5 274 849
0.01 0.125 0 52.3 329 929
0.01 0.125 500 52.2 324 927
0.01 0.125 1000 52.0 311 923
0.01 0.125 1500 51.8 293 918
Fable ii. 4: Gyroscopic Ligenfrequencies for Rectangular Euler Propeller of
Varying Width

8.3 FORCED FREQUENCY RESULTS

The eigenvalue results given in Section 8.2 are adequate for the understanding of

gyroscopic coupling, however it is considered beneficial to demonstrate effects

graphically as well. For this purpose forced frequency results for various models
have been produced. Initial results (Graphs 8.1-8.4) have been formed to

simultaneously display the gyroscopic coupling and `splitting' phenomenon. As

such a square section propeller blade has been used with a slenderness ratio
(S.R.) of 50. Specifically it is one metre long and two centimetres thick (for both

width and thickness). A nominal force and moment has been applied to Yand yat
the tip node. No damping has been included and as such amplitude (especially
peak amplitudes at resonant frequencies) should be ignored. This propeller model
Chapter 8: Gyroscopic Propeller Element Results 97

has been run at various speeds including the analysis of a stationary propeller.

Results for the first mode of vibration which is shown to split into two natural
frequencies as the propeller rotates are displayed below (Graphs 8.1-8.4). X and

a results are not shown since they are zero. In addition to the splitting which can
be clearly seen, gyroscopic coupling is displayed. This shows as the lack of

response in the Z 8
and axis when the propeller is not rotating.

ý-.. -Jr.. ... _. --. »...,. »...... -.... f..... a.. all ý. 3. jm. JU)

- -I- -'""" ""

Chapter 8: Gyroscopic Propeller Element Results 98

vx apu v. . r "%. a.. v.. v... vý.... w .. f. w. av vp

v. "p.. va. J .......... v... -t, l... a,. , all w. .. _U

The `normal' gyroscopic phenomena have been demonstrated, with both the

splitting of natural frequencies and coupling of perpendicular axis shown.

However a square section propeller blade is an unlikely proposition, thus the
following results are for two rectangular propellers. A simple one metre propeller

has been chosen, ten centimetres wide and one centimetre thick. Although these
Chapter 8: Gyroscopic Propeller Element Results 99

dimensions are fictional a steel and aluminium comparison has be drawn since
these are two of the most common propeller materials. Again a nominal force

and moment has been applied to Y and y at the tip and no damping has is
included. Results (Graphs 8.5-8.12) show resonant peaks and demonstrate the

need for gyroscopic modelling by the complication of results. X and a results are
not depicted since they are zero.

The material properties are as follows:

Steel:
Young's Modulus = 207 GPa, Poisson's Ratio = 0.3, and Density = 7800 kg/m3
Aluminium (7075-T6):
Young's Modulus = 71.7 GPa, Poisson's Ratio = 0.33, and Density = 2810 kg/m3

Ironically since both Young's Modulus and Density are reduced for Aluminium

the eigenvalues for both materials are very similar. This is shown in Table 8.5
below. As such the frequency response graphs look very similar although with
the same force applied the aluminium propeller deflects more as it has a lower
stiffness.

Thickness Width Length Frequency Frequency Frequency

(m) (m) (m) ail rad/s 0)2 rad/s ca Z rad/s
Steel 0.01 0.10 1.0 52.3 329 523
Aluminium 0.01 0.10 1.0 51.3 322 513
l able s. o: tirst l hree static Propeller Eigenfrequencies
Chapter 8: Gyroscopic Propeller Element Results 100

Graph 8.5: Y translation ut Steel rropeller tip

urapn am; c, i ransiauun vi meei rropeiier lip

Chapter 8: Gyroscopic Propeller Element Results 101

"rapt, a. / -p mutation vi meei rropeuer i ip

%Jl 0.0. r AVLilllull Vl JICCI rrupeuer lip

Chapter 8: Gyroscopic Propeller Element Results 102

0.20

0.18

0 16
.
0 14

0.12

0 10
.
= 0.08 -- - -- ---- -

0.06

0.04 --
0.02

0.00
0 100 200 300 400 500 600

Response Frequency (rads )

Propeller Rotation Speed (r. p. m. )
-0 -500 - 1000 -- 1500

Graph 8.9: Y Translation Of Aluminium Propeller Tip

Graph 8.10: Z Translation Of Aluminium Propeller Tip

Chapter 8: Gyroscopic Propeller Element Results 103

urapn a. 1 i: p tcotauon ui Aiumimum rropeuer i ip

"rapu o. ih: y muuauuu vi tiiuiuuuuiu rrupeuer i ip

Gyroscopic coupling is immediately evident as a response around 525-550 rad/s,

this is only seen in results where rotation is present since this eigenvalue belongs

to the Iz axis not the Ij, axis in which the system is excited. The response becomes

more significant in amplitude with rotation frequency as the gyroscopic forces

increase. Coupling is also evident in the Z and 8 results which are only present

when there is rotation. Furthermore the gyroscopic effect changes the natural
Chapter 8: Gyroscopic Propeller Element Results 104

frequencies of the propeller with rotation speed, although interestingly in this

instancethe effect on the first mode of vibration is minimal.

8.4 SUMMARY
This chapter presents results in the form of both eigenvalues and forced
frequency response graphs for the gyroscopic propeller element derived using

Euler beam theory in Chapter 5. The validation of such an element is not

straightforward since a verified model or other theoretical results to compare

against do not exist. Time and money have prevented the inclusion of an
experimental study in this work. However since the work is analytical a

verification examining the gyroscopic coupling effect is itself is considered

sufficient given that the gyroscopic effects are not too difficult to understand.
Thus simple systems have been examined in order to check the derived matrix

with all results being as expected.

Changes in system behaviour agree with the expected gyroscopic response for a

simple beam in all circumstances. Firstly gyroscopic coupling is demonstrated,

whereby lateral vibrations in one plane are translated creating coupled vibrations
in a perpendicular plane. Secondly the `splitting' of natural frequencies has been
demonstrated. This shows as a single natural beam frequency becoming two

either side of the original, the separation increasing with rotation speed. Where

more complex (non-symmetric) propeller blades are analysed this effect often
becomes a frequency shift instead. And finally all amplitudes are as expected,

with vibration amplitudes in perpendicular axes being approximately equal. This

is important since as various whirl conditions are achieved at different speeds the

lateral displacements will tend to be equal. Thus if these displacements vastly

differed the gyroscopic coupling effect would likely prove to be incorrect.

Finally results show that for propellers lateral changes in system response due to

gyroscopic forces must not be ignored due to the possibility of significant

changes in system behaviour. The effect is greater the closer the width and
thickness of the blade are dimensionally since the coupled natural frequencies in
Chapter 8: Gyroscopic Propeller Element Results 105

both axes are then closer together. The change is also more significant for more
flexible blades.
Chapter 9: Centrifugal Stiffening Propeller Element Results 106

CHAPTER 9

RESULTS 3: CENTRIFUGAL STIFFENING ELEMENT

9.1 INTRODUCTION
This chapter presents results showing the effects of centrifugal stiffening of

propeller systems using the formulation developed from Hoa, S.V., [1979] in

Chapter 6. For the purpose of this investigation all gyroscopic effects have been

eliminated such that the centrifugal effect is studied in isolation. This is in order
to simplify the problem and save confusion between the gyroscopic and

centrifugal stiffening effects.

It can be seen that if tests are initiated with a stationary propeller and then the

rotation frequency is increased, one would expect to the resonant vibration

frequencies of the propeller increase due to the additional stiffening effect.
Handelman, G., Boyce, W., and Cohen, H., [1958] and Kumar, R., [1974]

showed that the increase in natural frequencies caused by the increase in

stiffness, is most significant for the first mode of vibration. However this may

only be true for certain types of system and as such is an expected system

response to be investigated.

This type of stressstiffening is most significant for very slender beams that have

a very low initial lateral stiffness, thus the effect will also be investigated for
beams of differing slendernessratios in order to determine at what point it may
become insignificant.
 Chapter 9: Centrifugal Stiffening Propeller Element Results 107

9.2 EIGENVALUE RESULTS

All results have been produced for systems where the propeller is 1 metre long

and split into three finite elements of equal length (since three elements are

shown to provide good accuracy in Chapter 8). The thickness of the blade has

then been changed in order to provide results for systems of different slenderness

ratios. Since gyroscopic effects have been removed there is no coupling between

the two bending planes (Figure 8.1), thus for a rectangular section the natural
frequencies for the thickness and width are independent of each other. As such

the width and associated natural frequencies are not given in the results. Once

again steel is used where Young's Modulus = 207 GPa, Poisson's Ratio = 0.3,

Density 7800 kg/m. 3

and =

Thickness Slenderness Speed o Wright C02 Wright w3 Wright

(m) Ratio (rpm) (Rad/s) et al. Rad/s et al. Rad/s et al.
0.01 100 0 52.3 52.3 329 328 929 918
0.01 100 500 77.3 77.4 355 354 954 944
0.01 100 1000 124 124 423 422 1031 1018
0.01 100 1500 175 175 518 517 1145 1131
fable 9. l: >rigenirequency comparison with Results from Wright et al.,
[19821

Table 9.1 shows eigenfrequencies for a slender beam with results compared to
those of Wright et al., [1982]. Results from Wright et al have been interpolated
from Table 3 in the published paper. Results show excellent agreement despite

the use of just three finite elements for this model. As such further comparison
against other results has been deemed unnecessary as this study is focused on
propeller behaviour. A complete comparison of different methods of analysing
the effects of centrifugal stiffening is given by Wright et al., [1982].
 Propeller Element Results 108
Chapter 9: Centrifugal Stiffening

Thickness Slenderness Speed wl (02 (03

(m) Ratio (rpm) Rad/s Rad/s (Rad/s)
0.01 100 0 52.3 329 929

0.01 100 500 77.3 355 954

0.01 100 1000 124 423 1031

0.01 100 1500 175 518 1145

0.02 50 0 105 658 1858

0.02 50 500 119 671 1871

0.02 50 1000 155 709 1911

0.02 50 1500 199 770 1975

0.04 25 0 209 1315 3716

0.04 25 500 217 1322 3723

0.04 25 1000 238 1342 3743

0.04 25 1500 270 1374 3776

0.10 10 0 523 3287 9289

0.10 10 500 526 3290 9292

0.10 10 1000 535 3298 9300

0.10 10 1500 550 3312 9314

Table 9.2: Euler Propeller Eigenfrequencies Incluuing

Centrifugal Stiffening

The first thing to notice about centrifugal stiffening is that it can be very
for relatively slow rotation speeds. Also since the stiffening is
significant even
dependant on the rotation speed squared, as the rotation speed increases the

effect becomesconsiderably more important although not exponentially so.

Centrifugal stiffening also proves to be far more significant for slender

propellers. This is due to the low bending stiffness associated with a slender

propeller, thus the additional effect of centrifugal stiffening is more significant.

This is most obvious comparing eigenfrequcncies for the thick and thinnest (10

and 100 slenderness ratio) propellers in Table 9.2. Due to the modelling
techniqueused it can be seenthat eigenfrequcnciesfor the thick propeller arc
Chapter 9: Centrifugal Stiffening Propeller Element Results 109

exactly ten times that of the thin propeller when there is no rotation. However
this trend does not continue once rotation is invoked and therefore centrifugal

stiffening effects are included. In fact for the thick shaft eigenfrequencies hardly

change, while mode one for the slender shaft shows the eigenfrequency more

than triple by 1500 rpm.

For further clarification the percentage change in eigenfrequencies is displayed

graphically below (Graphs 9.1-9.4) for all four slenderness ratios up to a

propeller rotation speed of 10000 rpm. This clearly shows how the effect can be

is much greater for a slender beam where a 1967 % increase in eigenfrequency is

seen for the first mode of vibration for the one centimetre thick propeller at

10000 rpm, while for the ten centimetre thick propeller under the same

conditions a 137 % increase is seen.

UU)
Chapter 9: Centrifugal Stiffening Propeller Element Results 110

Graph 9.1: rercentage unange in Ligentrequencies (N. K. 5U)

vrapu 7. j: rcrceutage %-uauge in r igenireyuencles (',N.k. L)

Chapter 9: Centrifugal Stiffening Propeller Element Results Ill

Graph 9.4: rercentage . nange in r, igenirequencies k:5.it. iu)

9.3 FORCED FREQUENCY RESULTS

The eigenvalue results given in Section 9.2 are sufficient for the verification and

understanding of centrifugal stiffening, however it is deemed beneficial to

demonstrate the effects graphically as well. For this purpose forced frequency

results for various models of different thickness have been produced. The models

used echo eigenvalue results using a propeller blade with slenderness ratios
(S. R. ) of 100,50,25 & 10. Specifically it is one metre long and one, two or four

centimetres thick, with each blade split into three finite elements of equal length.

The blade root is fixed (like a cantilever) and a nominal force and moment has

been applied to Y and y at the tip node. Since gyroscopic effects have been

eliminated there will be no results for X and a so the propeller blade width does

not effect results. X and a results are also zero so only Y and y results are given
(Graphs 9.5-9.12). No damping has been included and as such amplitude (that is

peak amplitudes at resonant frequencies) should be ignored. The models have

been run at various speeds including the analysis of a stationary propeller.

Results show as the rotation speed increases so does the resonant frequency in all

cases. As expected these match the eigenfrequencies. Although it is true to say

Chapter 9: Centrifugal Stiffening Propeller Element Results 112

that the frequency shift of the higher modes are less significant than those of
lower modes in terms of percentage frequency shift, the actual shift in frequency

in radians per second is in fact be seen to be very similar. Results also clearly

display the effect of increasing the beam thickness. Once a slenderness ratio of

25 is reached the centrifugal stiffening effect makes very little difference to the

predicted response frequencies at these rotation speeds (Graphs 9.9-9.10). This is

so much so in fact that results for a propeller with a slenderness ratio of 10

(Graphs 9.11-9.12) appear to overwrite one another due to the scale required to

fit the first three modes of vibration on the graph.

Graph 9.5: Y 'translation Ut Propeller Tip (S. R. 100)

Chapter 9: Centrifugal Stiffening Propeller Element Results 113

urapn v. o: y noianon vi rropener i Ip (no.K. iuu)

vrapu 7. i; Ii 14ii L uuu vi IF FUNViI F1 111 IJ. K. 'U)

Chapter 9: Centrifugal Stiffening Propeller Element Results 114

txrapu Y.a: y nuIauun vi rrupeuer i Ip k3. ic. : )u)

1A aiiJiaiiuii
v ar 7.7.1 v, IU VII i IN kC. [c. L')
Chapter 9: Centrifugal Stiffening Propeller Element Results 115

k, ruplI 7.1V: r i%utatluu v ia vpclac1 i lip k,: P. ik. A. 71

vrapn '. i i: Yi ransianon vi rropeiier i ip (ýo.R. iu)

 116
Chapter 9: Centrifugal Stiffening Propeller Element Results

vrapn Y. 1L: 7 notation vi rrupener i ip k3. rc. ivy

9.4 COMBINED GYROSCOPIC & CENTRIFUGAL STIFFENING

FORCED FREQUENCY RESULTS

Graphs 9.13 & 9.14 below show the effects of centrifugal stiffening on the same

aluminium propeller described in Chapter 8 (Graphs 8.9-8.12) where the

gyroscopic effects are shown. It can clearly be seen that at the same low rotation

speeds used for the gyroscopic tests the centrifugal stiffening effect is somewhat
different to the gyroscopic effect however the order of magnitude by which

natural frequencies change is approximately equal.

A combined model containing both gyroscopic and centrifugal stiffening effects

is then shown in Graphs 9.15-9.18 below. The graphs are very similar in shape to

Graphs 8.9-8.12 which show only gyroscopic effects, however centrifugal

stiffening is apparent in the increased natural frequencies. For clarity a direct

comparison can be made with Graphs 8.9,9.13, & 9.15, which show gyroscopic,

centrifugal stiffening and combined effects for the same system respectively. It is

most obvious for the first mode of vibration since this hardly changes under

gyroscopic influence alone. Also since in some cases both effects can conspire to
increase a natural frequency there are significant increases seen.
Chapter 9: Centrifugal Stiffening Propeller Element Results 117

hrapn 9. I :Yi ransianon vi Aiumtnium rropeuer i ip

vraNu 7. I t: rI uIaawu v r+. uumill uIu IF roprurr lip

Chapter 9: Centrifugal Stiffening Propeller Element Results 118

urapn g. in: ii ranstauon vi Aiuminium rropeuer i Ip

k-rapu 7.1o; 1 rausiauvu ", Hiummium rropeiier lip

 119
Chapter 9: Centrifugal Stiffening Propeller Element Results

V1ilpll 7.1 /. F.! 1%Vl4991Vaa va AaY aaaaaaa Yaa" a avN'... -. .. I.

VI I JII 7.10.1 AxusatIVII V A4U413gEsIUIl 1 UIJ Raua 1 III

9.5 SUMMARY

This chapter presents results for propellers including the effects of centrifugal

stiffening derived in Chapter 6. Eigenvalues have been verified against results

published by Wright et al., [1982]. As expected the increase in bending mode

stiffness due to the axial stress caused by centrifugal stiffening increases the

natural response frequencies, this continues to increase with rotation speed. This
Chanter 9: Centrifugal Stiffening Propeller Element Results 120

is most significant for slender beams where the normal bending stiffness is

comparatively low. As the propeller beam becomes progressively thicker and the
slendernessratio is reduced the effect becomes far less significant and for a very
thick propeller at low rotation speeds it could be ignored. However if rotation
speedsare high it will still be necessaryto include the effect, and as changes are
often significant it is better practice to always include this type of effect.

The effect of centrifugal stiffening will also change the bending mode shape
since the blade becomes stiffer at the root than the tip. However the number of
modes does not change; it is the resonant frequency of response which increases
from a standard prediction. Thus for modelling techniques whereby the computer

model is `corrected' to match experimental results it may be deemed possible to

ignore the effects and compensate in later model `correction'. However if

analysis at different speeds is required it is very important to include the effects

of centrifugal stiffening in the computer model.

Handelman, G., Boyce, W., and Cohen, H., [1958] and Kumar, R., [1974] stated
that the increasein natural frequencies causedby the increase in stiffness, is most
significant for the first mode of vibration. However this is not strictly true as the
actual frequency shift is approximately equal for all three of the first modes. It is
true to state that the frequency shift in higher modes is less significant than for
the lower modes in terms of percentage change; but this is of course just a
different way of representing the data. Results have not been previously

published showing that the shift in frequency is approximately equal for the first
three modes in this manner.

Finally it is shown that at the same rotation speed the gyroscopic

effect is
approximately equal to that of centrifugal stiffening for propellers in terms of
numerical frequency shift for symmetric propellers. Due to this it is important to
include both effects in a model; especially as for some
modes of vibration both
effects will increasea particular natural frequency making the potential total shift
very large indeed. It should be noted that because gyroscopic coupling
sometimes lowers a natural frequency this will not always be the
case. As the
Chapter 9: Centrifugal Stiffening Propeller Element Results 121

increase in computation time for such additions is now minimal due to

improvements in modern computer speeds it is considered wise to include both

effects during modelling.

Chapter 10: Discussions & Conclusions 122

CHAPTER 10

DISCUSSIONS & CONCLUSIONS

10.1 DISCUSSIONS
Over the past forty or so years since computers have become available for

scientific use much effort has been devoted to the study of flexible beam

elements for rotor-dynamic analysis. The analysis is of great importance mainly

due to the need of improving efficiencies of power plants since vast sums of

money are involved. However there are many other problems that this type of

analysis is applicable to, some of which include manufacturing equipment, drive

shafts, rotor-arms for helicopters, propellers, and turbine blades.

To reduce design costs and improve understanding of such systems complete

dynamic modelling of such systems has to be achieved. In recent years since

computers has become exponentially more powerful the Finite Element Method
has been regarded as the best choice for the approximation of such continuums.
The method is very flexible permitting the analysis of structures with complex

geometry using a limited number of simple elements. Finite Element modelling

is now long established and there are many possible methods of achieving the

end goal, the method chosen for this is

study the rigid body and flexible element
hybrid modelling technique described in Chapter 3. This type of modelling

proves much more flexible than a system such as ANSYS uses since it allows the
inclusion of rotational point inertias or spring attachments at a distance from the

shaft node. Achieving this in ANSYS creates an ill-conditioned stiffness matrix

becausein each case a finite element of very high stiffness will have to be
used.
Chapter 10: Discussions & Conclusions 123

In the past the programming of a mathematical finite element model demanded a

balance between the requirements of high accuracy and computational efficiency.

In order to achieve good accuracy a detailed model incorporating all possible

effects is required. However this type of detail would create computational

problems limiting the practical value of the model. As such much previous

research has focused on reducing computational time whilst retaining accuracy.

However now since computers are now easily powerful enough to deal with

complex modelling techniques the principal issue often boils down to the

inclusion of all possible system idiosyncrasies in order to ensure an accurate

model. The main objective of this research was to improve the understanding of
propeller behaviour since over the course of time there has been considerably
less research in the area of rotor-dynamic propeller elements than of shaft

elements. To this end the main two considerations are the gyroscopic and
centrifugal stiffening effects, because the gyroscopic effect is very different to
that of a rotating shaft and the centrifugal stiffening effect does not exist at all in

shaft elements.

The theory presentedin this thesis permits propeller blades to be modelled using
Euler-Bernoulli beam elements. This is a much faster technique than models

which employ a fine mesh of small elements. To this end a novel mathematical
approach has been used to describe the gyroscopic bending moments and forces
of such an element. The literature survey in Chapter 2 shows that an attempt to
describe the gyroscopic effects in this manner has not been previously attempted.
The method is fully described in Chapter 5. The technique is based on the
derivation of a standardEuler gyroscopic matrix for an element rotating about its

own axis of symmetry as described in Chapter 4. However the crucial difference

when a propeller type element is being considered is that the axis of rotation is
now perpendicular to the elements axis of symmetry. Thus the polar moment of
inertia per unit length now becomes a variable within the integration. This

moment of inertia increasesquadratically with the distance away from the axis of
rotation. Thus the radial distance of the elemental mass under consideration from
the axis of rotation must be considered in the matrix derivation. The final
result
Chapter 10: Discussions & Conclusions 124

shows the effect to be directly dependant on the rotation speed and the radial
distancesquaredas can be expected.

Gyroscopic behaviour is often considered a difficult subject to understand

because forces and moments do not necessarily occur in the direction one might

initially expect. Additionally where rotation speeds are small or the polar

moment of inertia is small it is often possible to produce an accurate model

without the inclusion of gyroscopic effects and thus they can get ignored.
Gyroscopic motion occurs whenever the axis about which a body is spinning is
itself rotating about another axis. A common example of this is a disc rotating on

a shaft as shown in Figure 10.1 below. Where the moment is applied about the : -

axis the reaction is not about the same axis as it would be if the rotor were

stationary, but is instead about the y-axis, perpendicular to both the applied

moment and axis of rotation.

Y
Reaction Moment
X
(Precession Axis)
. -ý
S2

Applied Moment

Figure 10.1: Gyroscopic Disc Example

In addition to the coupling effect described above whereby

moments (or
vibrations) in one axis translate to create moments in another, the gyroscopic
forces have a second major effect. That is the splitting of
natural frequencies.
Gyroscopic forces cause a single natural frequency to split into two frequencies,
for a symmetrical system these new natural frequencies
tend to be either side of
Chanter 10: Discussions& Conclusions 125

the original with the spacing increasing with the rotation speed. This is of great
importance in the analysis of rotor-dynamic systems as the change in natural
frequencies will also change acceptable running conditions. These effects are
fully shown for simple propellers in Chapter 8.

The literature survey in Chapter 2 showed there has been more research centred

on the centrifugal stiffening effect of propeller elements than the gyroscopic

effects. It is the considered opinion of this author that this is due to the

centrifugal stiffening not only changing the natural frequencies but also
increasing stresses especially at the blade's root. As such the centrifugal

stiffening effect presented in this thesis is based on a paper by Hoa, S.V., [1979].
A full derivation is given in Chapter 6 both to improve understanding and
because discrepancies were found in the original publication. The final stiffening

effect is dependant on the square of the propeller rotation speed, however it must
be noted that this stiffening effect is in addition to the static stiffness of the beam

and does not replace it.

Unlike the complications involved with the gyroscopic effect, centrifugal

stiffening is relatively straightforward to understand. In essencethe rotation of a

propeller blade createsan axial tension along the length of the blade. This will be
greatest at the root and fall to zero at the propeller tip. Since the force will be
proportional to the rotation speed squared multiplied by the distance from the
axis of rotation. The axial force createsan axial stresswhich in turn increasesthe
bending stiffness of the propeller blade. This in turn increases the natural
frequencies of the system. The effect is proportionally much greater the more

slender the initial propeller blade since in this type of system the static bending
stiffness will be very small.

The additional bending stiffness does not have the exact same effect on all the

natural frequencies, although all the bending frequencies do increase.

Handelman, G., Boyce, W., and Cohen, H., [1958] and Kumar, R., [1974]
stated
that the increasein natural frequencies caused by the increase in stiffness, is most
significant for the first mode of vibration. However this is not strictly true
as the
Chapter 10: Discussions & Conclusions 126

frequency shift is approximately equal for all three of the first modes, which has

not been previously reported. However it would be correct to state that the
frequency shifts of the higher modes are less significant than those of lower

modes in terms of percentagefrequency shift.

The inclusion of gyroscopic and centrifugal stiffening terms described in this

work makes the modelling of propeller blades using Euler finite beam elements a

viable proposition. This is a convenient technique that allows a blade to be

modelled very simply as a series of connected elements rather than using a fine

mesh of elements. Since rectangular section elements can easily be derived a

propeller can be modelled to include a changing cross-section size along its

length and any twist along its axis of symmetry. Euler beam theory is not limited

to the round or rectangular matrices used in this study (for shafts or propellers

respectively). Should the analysis of a different cross section shape be required it

is possible to determine the second moment of area and shear modulus without

the need for major changes. Thus the modelling procedure becomes even more
flexible. The associated changes in bending mode shapes and frequencies can

then be accounted for.

10.2 CONCLUSIONS
The aim of the present researchto develop an Euler finite beam element capable

of modelling propellers with particular attention given to the gyroscopic and

centrifugal stiffening effects has been achieved. The results of this work have
particular applications to wind turbine blades, helicopter rotors, or other
propellers. Furthermore some more research into the standard gyroscopic shaft
formulation has been performed.

The findings of the researchcan be summarised as follows:

9 For shaft analysis it has previously been published that Timoshenko

theory is superior to Euler theory for thick beams, for example; Lee,
C.W., [1993]. However results in Chapter 7 show that for
shafts up to 0.5
Chapter 10: Discussions & Conclusions 127

metres in diameter rotating at up to 40 Hz Euler theory achieves almost

identical results and is more than adequate for computational analysis.
Even though at 48 metres long one may assume the shaft to be slender

this is not really case since it has so many support positions along its
length.

" For the Euler gyroscopic propeller element coupling is demonstrated,

whereby lateral vibrations in one plane are translated creating coupled

vibrations in a perpendicular plane. Also the `splitting' of natural

frequencies has been established in Chapter 8. This shows as a single

natural beam frequency becoming two natural frequencies either side of

the original, the separation increasing with rotation speed. Where more

complex (non-symmetric) propeller blades are analysed this effect often

becomes a frequency shift instead.

Results show that for propellers lateral changes in system responsedue to

gyroscopic forces must not be ignored due to the possibility of significant

changes in system behaviour. The effect is greater the closer the width
and thickness of the blade are dimensionally, since the coupled natural
frequencies in both axes are then closer together. The change is
also more
significant for more flexible (slender) blades.

" The increase in bending stiffness due to centrifugal stiffening increases

the natural frequencies of vibration. These frequencies continue to
increase with rotation speed although not exponentially so
since the
propeller has a static bending stiffness also.

" Frequency change is most significant for slender beams where the
static
bending stiffness is comparatively low. As the propeller beam becomes

progressively thicker and the slenderness ratio is reduced the effect

becomes far less insignificant and if rotation
speeds are also very low
centrifugal stiffening could be ignored.
Chapter 10: Discussions & Conclusions 128

" The gyroscopic effect may increase or decrease a natural frequency, thus
in some instancesthe combined increase from gyroscopic and centrifugal

stiffening will create a dramatic frequency change from a static

prediction.

9 Finally it should be noted that for propeller analysis it is equally

important to include gyroscopic effects and centrifugal stiffening effects

since both are approximately equal in magnitude. The methods used here

are considered superior to the simplistic approach at attaching a rotational

point disc inertia.

10.3 FURTHER WORK

The primary aim of developing a finite element capable of simulating a propeller
including gyroscopic motion due to bending and centrifugal stiffening due to

centrifugal forces has been realised. However there are further refinements that

could be achieved.

" Firstly the techniques described could be used to produce a Timoshenko

propeller element, doing this and using the same solver would be useful
to demonstrateany limitations between Timoshenko and Euler models.

" Secondly although the main gyroscopic effect due to bending has been
formulated there are two more possible displacements that will induce

gyroscopic moments. There is rotation of the element about its own axis
of symmetry to consider. This will produce a gyroscopic moment similar
in direction and derivation to that produced here. Rough
calculations
show that for most blades the change in inertia will be less than 5% of
that due to bending for the same degree of rotation of the blade's
root,
thus the effect is considerably smaller than that due to bending. There is
also the moment produced by an axial extension of a propeller blade. For
Chapter 10: Discussions & Conclusions 129

many blade types this will be insignificant since the axial extension will
be close to zero. However a universal code should be capable of

analysing all blade types including highly flexible materials which will
extend, or indeed blades with exceedingly high rotation speeds.

" Finally an experimental model could be set up in order to fully validate

the gyroscopic coupling effects on a propeller blade, however it is this

authors belief that since the work done is analytical experimental methods
are unnecessary with the modelling type of verification already
performed being sufficient.
References 130

REFERENCES

Abbas, B. A. H., [1979] "Simple Finite Elements for Dynamic Analysis of Thick
Pre-Twisted Blades" Aeronautical Journal, Vol. 83, pp 450-453

Al-Ansary, M. D., [1998] "Flexural Vibrations of Rotating Beams Considering

Rotary Inertia" Journal of Computers and Structures, Vol. 69, pp 321-328

Aleyaasin, M., Ebrahimi, M., and Whalley, R., [2000] "Flexural Vibration of
Rotating Shafts by Frequency Domain Hybrid Modelling" Journal of Computers

and Structures, Vol. 79, pp 319-331

ANSYS Theory reference 001099 Ninth Edition 14.16.5 SAS IP, Inc. 0

Bauer, H. F., and Eidel, W., [1988] "Vibration of a Rotating Uniform Beam, Part
II: Orientation Perpendicular to the Axis of Rotation" Journal of Sound and
Vibration, Vol. 122(2), pp 357-375

Bir, G., and Stol, K., [1999] "Operating Modes of a Teetered-Rotor Wind
Turbine" International Modal Analysis Conference NREL/CP-500-25983

Carnegie, W., [1959] "Vibrations of Rotating Cantilever Blading: Theoretical

Approaches to the Frequency Problem Based on Energy Methods" Journal of
Mechanical Engineering Science, Vol. 1(3), pp 235-240

Carnegie, W., [1964] "Vibrations of Pre-Twisted Cantilever Blading Allowing

for Rotary Inertia and Shear Deflection" Journal of Mechanical Engineering
Science,Vol. 6(2), pp 105-109
References 131

Carnegie, W., Stirling, C., and Fleming, J., [1965-66] "Vibration Characteristics

of Turbine Blading Under Rotation; Results of an Initial Investigation and

Details of a High-Speed Test Installation" Proceedings of the Institute of
Mechanical Engineers, Vol. 180(3), pp 124-132

Chatelet, E., D'Ambrosio, F., and Jacquet-Richardet, G., [2005] "Toward global

modelling approaches for dynamic analyses of rotating assemblies of

turbomachines" Journal of Sound and Vibration, Vol. 282(1-2), pp 163-178

Chen, C.L., and Chen, L. W., [2002] "Random Vibration of a Rotating Blade with
External and Internal Damping by the Finite Element Method" Journal of Sound

and Vibration, Vol. 252(4), pp 697-715

Chen, L. W., and Ku, D. M., [1991] "Finite Element Analysis of Natural Whirl
Speeds of Rotating Shafts" Journal of Computers and Structures, Vol. 40, pp

741-747

Cowper, G. R., [1966] "The Shear Coefficient in Timoshenko's Beam Theory"

Journal of Applied Mechanics, pp 335-340

Davis, R., Henshell, R.D., and Warburton, G.B., [1972] "A Timoshenko Beam
Element" Journal of Sound and Vibration, Vol. 22(4), pp 475-487

Dokainish, M. A., and Rawtani, S., [1971] "Vibration Analysis of Rotating

Cantilever Plates" International Journal for Numerical Methods in Engineering

Vol. 3, pp 233-248

Esat, 1.1., [1993] "Mathematical Modelling of Small Amplitude Oscillation

of
Multi-Body Systems" Journal of machine vibration Vol. 2, pp 30-35

Esat, I. I., et at. [1997] "Hybrid Modelling; Mixed Rigid

and Flexible Systems"
ASME Journal, 97-AA-91
References 132

Esat, I. I., [2002] Vibratio Computer Code Formulations

Esat, I. I., Banisoleiman, K., and Jones, S., [2002] "Modelling and Analysis of
Stresses in a Flexible Shafting System Under Harmonic Excitations" Integrated

Design and Process Technology Conference June 2002 Society for Design and

Process Science

Gasch, R., [1976] "Vibration of Large Turbo-Rotors in Fluid-Film Bearings on

an Elastic Foundation" Journal of Sound and Vibration, Vol. 47(1), pp 53-73

Genta, G., [1985] "Consistent Matrices in Rotor Dynamics" Meccanica, Vol. 20,

pp 235-248

Genta, G., and Tonoli, A., [1996] "A Harmonic Finite Element for the Analysis

of Flexural, Torsional and Axial Rotordynamic Behaviour of Discs" Journal of

Sound and Vibration, Vol. 196(1), pp 19-43

Genta, G., and Tonoli, A., [1997] "A Harmonic Finite Element for the Analysis

of Flexural, Torsional and Axial Rotordynamic Behaviour of Blade Arrays"

Journal of Sound and Vibration, Vol. 207(5), pp 693-720

Gmür, T. C., and Rodrigues J.D., [1991] "Shaft Finite Elements for Rotor
Dynamics Analysis" Transactions of the ASME, Journal of Vibration and
Acoustics, Vol. 113, pp 482-493

Green, R.B., [1948] "Gyroscopic Effects on the Critical Speeds of Flexible

Rotors" Journal ofApplied Mechanics, Vol. 70, pp 369-376

Handelman, G., Boyce, W., and Cohen, H., [1958] "Vibrations

of a Uniform
Rotating Beam with Tip Mass" Proceedings of the Third U.S. National Congress

of Applied Mechanics, pp 175-180

References 133

Hoa, S.V., [1978] "Vibration Frequency of a Curved Beam with Tip Mass"
Journal of Soundand Vibration, Vol. 61(3), pp 427-436

Hoa, S.V., [1979] "Vibration of a Rotating Beam with Tip Mass" Journal of
Sound and Vibration, Vol. 67(3), pp 369-381

Hodges, D. H., [1979] "Vibration and Responseof Non-Uniform Rotating Beams

with Discontinuities" Journal of the American Helicopter Society, Vol. 24(5), pp

43-50

Hong, S-W., and Park, J-H., [1999] "Dynamic Analysis of Multi-Stepped

Distributed ParameterRotor-Bearing Systems" Journal of Sound and Vibration,
Vol. 227(4), pp 769-785

Huang, B. W., and Huang, J.H., [2001] "Mode Localization in a Rotating

Mistuned Turbo Disk with Coriolis Effect" International Journal of Mechanical
Sciences, Vol. 43, pp 1643-1660

Jacquet-Richardet, G., Ferraris, G., and Rieutord, P., [1996] "Frequencies and
Modes of Rotating Flexible Bladed Disc-Shaft Assemblies: A Global Cyclic
Symmetry Approach" Journal of Sound and Vibration, Vol. 191(5), pp 901-915

Jeffcott, H. H., [1919] "The Lateral Vibration of Loaded Shafts in the

Neighbourhood of a Whirling Speed - The Effect of Want of Balance"
Philosophical Magazine Series 6, Vol. 37, pp 304

Jones, L. H., [1975] "The Transverse Vibration of a Rotating Beam

with Tip
Mass: The Method of Integral Equations" Quarterly of Applied Mathematics,
Vol. XXXIII(3), pp193-203

Jones, S., and Esat, I., [2003] "Gyroscopic Beam Element Matrix Derivation
and
Results" Societyfor Design and Process Science; Integrated Design
and Process
Technology Conference, December 2003.
References 134

Kapoor, K. K., and Hartz, B.J., [1966] "Stability of Plates Using the Finite
Element Method" Journal of the Engineering Mechanics Division, Proceedings

of the American Society of Civil Engineers EM2, pp 177-195

Kapur, K. K., [1966] "Vibrations of a Timoshenko Beam, Using a Finite Element

Approach" Journal of the Acoustical Society of America Vol. 40(5) pp 1058-
1063

Kim, Y. D., and Lee, C. W., [1986] "Finite Element Analysis of Rotor Bearing
Systems Using a Modal Transformation Matrix" Journal of Sound and Vibration,

Vol. 111(3), pp 441-456

Kumar, R., [1974] "Vibrations of Space Booms Under Centrifugal Force Field"
Canadian Aeronautics and Space Institute Transactions Vol. 7(1), pp 1-5

Lamb, F. R. S., and Southwell, R. V., [1921] "The Vibrations of a Spinning Disk"

pp 272-280

Lee, C. W., [1993]. "Vibration Analysis of Rotors" Klinver Academic Publishers

Dordrecht. ISBN 0-7923-2300-9

Luczko, J., [2002] "A Geometrically Non-Linear Model of Rotating Shafts with
Internal Resonanceand Self-Excited Vibration" Journal of Sound and Vibration,
Vol. 255(3), pp 433-456

Magari, P.J., and Shultz, L. A., [1987] "Development of a Rotating Blade Finite
Element with an Application to the Analysis of Helicopter Rotorsystems"
American Institute of Aeronautics and Astronautics, Inc., pp 17-27

Mohiuddin, M. A., and Khulief, Y. A., [1999] "Coupled Bending Torsional

Vibration of Rotors Using Finite Element" Journal of Sound
and Vibration, Vol.
223(2), pp 297-316
References 135

Morton, P.G., [1968] "Influence of Coupled Asymmetric Bearings on the Motion

of a Massive Flexible Rotor" Applied Mechanics Group. Proc. Institute of
Mechanical Engineers Vol. 182(1), pp 255

Naguleswaran, S., [1994] "Lateral Vibration of a Centrifugally Tensioned

Uniform Euler-Bernoulli Beam" Journal of Sound and Vibration, Vol. 176(5), pp
613-624

Nelson, H. D., [1977] `°A Finite Rotating Shaft Element Using Timoshenko Beam
Theory" Engineering Research Centre Report ERC-R-77023, Arizona State
University, pp 61

Nelson, H.D., [1980] "A Finite Rotating Shaft Element Using Timoshenko Beam
Theory" Journal of Mechanical Design, Vol. 102, pp 793-803

Nelson, H. D., and McVaugh, J.M., [1976] "The Dynamics of Rotor-Bearing

Systems Using Finite Elements" Journal of Engineering for Industry, Vol. 98, pp
593-600

Peters, D. A., [1973] "An Approximate Solution for the Free Vibrations of
Rotating Uniform Cantilever Beams" NASA Technical Memorandum, N78-
33289

Putter, S., and Manor, H., [1978] "Natural Frequencies of Radial Rotating
Beams" Journal of Sound and Vibration Vol. 56(2), pp 175-185

Rankine, W.A., [1879] "On the Centrifugal Force of Rotating Shafts" Engineer,
London Vol. 27, pp 249

Rao, J.S., and Carnegie, W., [1969] "Non-Linear Vibrations of Rotating

Cantilever Beams" The Aeronautical Journal of the Royal Aeronautical Society,
Vol. 74, pp 161-165
References 136

Rouch, K. E., and Kao, J.S. [1979] "A Tapered Beam Finite Element for Rotor
Dynamics Analysis" Journal of Sound and Vibration, Vol. 66(1), pp 119-140

Ruhl, R.L., [1970] "Dynamics of Distributed Parameter Turbo-Rotor Systems:

Transfer Matrix and Finite Element Techniques" PhD Thesis, Cornell University,
Ithaca, N. Y.

Ruhl, R.L., and Booker, J.F., [1972] "A Finite Element Model for Distributed
Parameter Turbo-Rotor Systems" Transactions of the ASME, Journal of
Engineering for Industry, Vol. 94. pp 126-132

Sakata, M., Kimura, K., Park, S.K.,. and Ohnabe, H., [1989] "Vibration of
Bladed Flexible Rotor Due to Gyroscopic Moment" Journal of Sound and
Vibration, Vol. 131(3), pp 417-430

Sauer, G., and Wolf, M., [1989] "Finite Element Analysis of Gyroscopic
Effects" Finite Elements in Analysis and Design, Vol. 5, pp 131-140

Southwell, and Gough., [1921] "The Free Transverse Vibration of Airscrew

Blades" British A. R. C. Report and Memoranda, No. 766

Stafford, R.O., and Giurgiutiu, V., [1975] "Semi-Analytic Methods for Rotating
Timoshenko Beams" International Journal of Mechanical Science, Vol. 17,
pp
719-727

Stephen, N. G., and Wang, P.J., [1986] "Stretching and Bending of Rotating
Beam" Journal of Applied Mechanics, Vol. 53, pp 869-872

Thomas, D.L., Wilson, J.M., and Wilson R.R., [1973] "Timoshenko Beam Finite
Elements" Journal of Sound and Vibration, Vol. 31(3), pp 315-330

Thomas, J., and Abbas, B. A. H., [1975] "Finite Element Model for Dynamic
Analysis of Timoshenko Beam" Journal
of Sound And Vibration, Vol. 41(3),
pp291-299
References 137

Thomson, W. T., [19931 "Theory of Vibration with Applications" Fourth Edition,

Published by Chapman & Hall, ISBN 0 412 54620 5

Timoshenko, S.P., [19221 "On the Transverse Vibrations of Bars of Uniform

Cross-Section" Philosophical Magazine, Vol. 43, pp 125-131

Wang, J.T. S., Mahrenholtz, 0., and Böhm, J., [1976] "Extended Galerkin's
Method for Rotating Beam Vibrations Using Legendre Polynomials" Solid

Mechanics Archives, Vol. 1, pp341-365

Wright, A. D., Smith, C.E., Thresher, RW., and Wang, J.L. C., [1982] "Vibration
Modes of Centrifugally Stiffened Beams" Journal of Applied Mechanics, Vol.

49, pp 197-202

Xiong, G.L., Yi, J.M., Zeng, C., Guo, H. K., and Li, L. X., [2003] "Study of the
Gyroscopic Effect of the Spindle on the Stability Characteristics of the Milling
System" Journal ofMaterials Processing Technology, Vol. 138, pp 379-384

Yang, J.B., Jiang, L. J., and Chen, D. CH., [2004] "Dynamic Modelling and
Control of a Rotating Euler-Bernoulli Beam" Journal of Sound and Vibration,
Vol. 274, pp 863-875

Yigit, A., Scott, R.A., and Galip-Ulsoy, A., [1988] "Flexural Motion of a
Radially Rotating Beam Attached to a Rigid Body" Journal of Sound and
Vibration, Vol. 121(2),pp 201-210

Yoo, H.H., and Shin, S.H., [1998) "Vibration Analysis of Rotating Cantilever
Beams"Journal of Soundand Vibration, Vol. 212(5), pp 807-828

Yoo, H.H., Park, J.H., and Park, J., [2001] "Vibration Analysis of Rotating Pre-
Twisted Blades"Journal of Computersand Structures,Vol. 79, pp 1811-1819
References 138

Zienkiewicz, O.C., [1971] "The Finite Element Method in Engineering Science"

2ndEdition, McGraw-Hill Publishing Company Ltd.

Zienkiewicz, O.C., and Morgan, K, [1983] "Finite Elements and

Approximation" John Wiley & Sons Inc. 0-471-89089-8

Zienkiewicz, O.C., and Taylor, RL., [1988] "The Finite Element Method: Basic
Formulation and Linear Problems" 4`" Edition, McGraw-Hill Publishing
Company Ltd. Vol. 1

Zienkiewicz, O.C., and Taylor, R.L., [2000] "The Finite Element Method: The
Basics" 5thEdition, Butterworth-Heinemann Publishing Company Ltd. Vol. 1

Zorzi, E.S., and Nelson, H. D., [1977] "Finite Element Simulation of Rotor-
Bearing Systems with Internal Damping" Journal of Engineering for Power,
Vol. 99(1), pp 71-76

Zorzi, E. S., and Nelson, H. D., [1980] "The Dynamics of Rotor-Bearing Systems

with Axial Torque; A Finite Element Approach" Transactions of the ASME,

Journal ofMechanical Design, Vol. 102, pp 158-161

Zou, C-P., Hua, H-X, and Chen, D-S., [2002] "Modal Synthesis Method of
Lateral Vibration Analysis for Rotor-Bearing System" Journal of Computers and
Structures, Vol. 80, pp 2537-2549
Appendix: Element Matrices 139

APPENDIX

ELEMENT MATRICES

Figure A. 1: Element Axis System

The displacementvector is as follows for all matrices:

(. fJ A
tlk = X'0 YO ZU a0 rU vl Zl al 71)
'Xl .
Appendix: Element Matrices 140

Figure A. 2: Gyroscopic Propeller Definitions

Note: a, b and s are global distancesfrom the axis of rotation

m"1
i
Hx
nl

Figure A. 3: Centrifugal Stiffening Propeller Definitions

Appendix: Element Matrices 141

A. 1 Mass Element Bending Matrix

0 0 0 0 0 0 0 0 0 0 0 0
13111 1112,, 9111 1312N
0 0 0 0 0 0 0 0
35 210 70 420
131,[ 1112/1 91/1 1312,,
0 0 0 _ 210 0 0 0 0 0
35 70 420
0 0 0 0 0 0 0 0 0 0 0 0
1112,, 1.1 13/2,, 13U
0 0 _ 210 0 0 0 0 - 0 - , 0
105 420 140
11121[ <[ 131211 1'1[
0 0 0 0 0 0 0 0
210 105 420 _ 140
M - 0
ký_ 0 0 0 0 0 0 0 0 0 0 0
9J/[ 1312p 1311[ 1112,,
0 0 0 0 0 0 0 0
70 420 35 210
9111 1312,, 131N 11121t
0 0 0 _ 0 0 0 0 0
70 420 35 210
0 0 0 0 0 0 0 0 0 0 0 0
1312p I3j[ 1112,
u Pp
0 0 0 - 140 0 0 0 0 0
420 210 105
1312p l/2_ I' ft
0 _ 420 0 0 0 _11- 0 _l 210 0 0 0
140 105

A. 2 Mass Element Axial Deformation Matrix

0 0 0 0 0 0 0 0 0 0
3 6
0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0
IMk:
6 0 0 0 0 0 3 0 0 0 0 0

0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0
Appendix: Element Matrices 142

A. 3 M ass Element Torsional Deformation M atrix

0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0

0 0 0 I.IrI 0 0 0 0 0 F`rýl 0 0
6 12
0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0
M_
ký 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0

0 0 0 L1r2L 0 0 0 0 0 1ur21 0 0
12 6
0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0

A. 4 Stiffness Element Bending Matrix

o o 0 0 0 0 0 0 0 0 0 0
12E1 6EI 12E1 6E1
0 0 0 0 0 0 0 0
13 P _ P
lz
12EI 1 12E1 LEI
0 0 0 -6E 0 0 0 0 0
13 _ 12
0 0 0 0 0 0 0 0 0 0 0 0
6E1 4EI 6E1 2EI
0 0 _ 1Z 0 0 0 0 0 0
1 12 1
6E1 4EI 6EI 2EI
0 0 0 0 0 0 0 0
12 1 12 1
[Kkl =
0 0 0 0 0 0 0 0 0 0 0 0
12E1 6E1 12E1 6E1
0 - 0 0 0 - 12 0 0 0 0
1, 13 - z
12EI 6EI 12E1 6E1
0 0 0 0 0 0 0 0
1' P
0 0 0 0 0 0 0 0 0 0 0 0
6ý 1 2E1 6E1 4EI
0 0 _ 0 0 0 0 0 0
12 1
62I 2j EI
0 0 0 0 o o 4E1
_6 P o o
1 1
Appendix: Element Matrices 143

A. 5 Stiffness Element Axial Deformation Matrix

EA 0 0 0 0 0
0 0 0 0 0

0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0
L Kk7
1_-
EA EA
0 0 0 0 0 0 0 0 0 0
-
0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0

A. 6 Stiffness Element Torsional Deformation Matrix

0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0
1J GJ
0 0 0 0 0 0 0 0 0 0

0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0
ýKýý_
0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0
GJ GJ
0 0 0 0 0 0 0 0 0 0

0 0 0 0. 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0
Appendix: Element Matrices 144

A. 7 Gyroscopic Element Bending Matrix for Shaft

0 0 0 0 0 0 0 0 0 0 0 0
0 0 36 0 -31 0 0 0 -36 0 -31 0
0 0 0 0 -31 0 36 0 0 0 -31
-36
0 0 0 0 0 0 0 0 0 0 0 0
0 31 0 0 0 412 0 -31 0 0 0 _12
cpr2 0 0 31 0 -412 0 0 0 -31 0 1Z 0
[Gk]-
601 0 0 0 0 0 0 0 0 0 0 0 0
0 0 -36 0 31 0 0 0 36 0 31 0
0 36 0 0 0 31 0 -36 0 0 0 31
0 0 0 0 0 0 0 0 0 0 0 0
0 31 0 0 0 _12 0 -31 0 0 0 412
0 0 31 0 12 0 0 0 -31 0 -412 0

A. 8 Gyroscopic Element Bending Matrix for Propeller

0 0 0 0 0 0 0 0
o 0 0 0
36(2x' +
0 0 0 -3/(-2a2 + 0 0 0 -36(2a2 + 0 31(-5a' - 0
3ab+2b) 4ab+5b2) 3ab+2b7) 4ah+2b7)

0 -36(2a2 + 0 0 0 -31(-2a' + 0 36(2a2+ 0 0 0

31(-5a2-
3ab+2b) 4ab+5b') 3ab+2b2) 4ab+2b')
0 0 0 0 0 0 0 0 0 0 0 0
31(-2a2+ 212(9a'+ 31(2a2 '(3a'+
0 0 0 0 0 0 0 0 -!
4ab+5b) 3ab+2b) 4ab-5b') ab+3b')
31(_2a2+ 212(9a2+ 31(2a2 ! 2(3x2 +
0 0 0 0 0 0 0 0
4ab+5b=) 3ab+2b) 4ab-5b') ab+3b°)
G,, =
0 0 0 0 0 0 0 0 0 0 0 0
-36(2x2 + -31(2a2 - 36(2"' + 31(5a'+
0 0 0 0 0 0 0 0
3ab+2b2) 4ab-5b) 3ab+2b') 4ab-2b2)
36(2a2+ -31(2a' - -36(2a' + 3/(5at +
0 0 0 0 0 0 0 0
3ab+2b2) 4ab-5b') 3ab+2b') 4ah-2b2)
0 0 0 0 0 0 0 0 0 0 0 0
'(3a'+ (5a'+ 2! '(2a'+
0 -3/(-5a2 - 0 0 0 -! 0 -3! 0 0 0
4ab+2b) ab+3b2) 4ab-2b') 3ab+9b1)
-31(-5a2- 12(3a'+ -31(5x' + +
0 0 0 0 0 0 0 -2l2(2a' 0
4ab+2b2) ab+3b=) 4ab-2b2) Sah+9b2)
Appendix: Element Matrices 145

A. 9 Centrifugal Stiffening Element Matrix for Propeller

0 0 0 0 0 0 0 0 0 0 0 0
0 A 0 0 0 -D 0 -A 0 0 0 -E
0 0 A 0 D 0 0 0 -A 0 E 0
0 0 0 0 0 0 0 0 0 0 0 0
0 0 D 0 B 0 0 0 -D 0 F 0
0 -D 0 0 0 B 0 D 0 0 0 F
Kew _ 1,nz
0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 D 0 A 0 0 0 E
-A
0 0 -A 0 -D 0 0 0 A 0 -E 0
0 0 0 0 0 0 0 0 0 0 0 0
0 0 E 0 F 0 0 0 -E 0 C 0
0 -E 0 0 0 F 0 E 0 0 0 C

Where:
7n 2) + 7r(2ni 2n 1)]
A_3[1(7m' -7n2 - - - -
35

B -12[1(14x2 -14n2 -7n-2)+7r(4m-4n-1)]

210

12[1(14m2
-14n2 -21n-9)+7r(4m-4n-3)]
C_
210

1[1(7m2-7n2 -14n-5)+14r(m-n-1)]
D=-
140

-7n2 + 2) +14r(m - n)]

E --1[1(7nz2
140

12[1(7m2-7n2 - 7n - 3) + 7r(2ni - 2n 1)]

ý, -
-- 420