Dynamic Behaviour of

Transformer Winding under

Short-Circuits

Muhammad Salman Aslam Minhas

A thesis submitted to the Faculty of Engineering, University of the Witwatersrand,

Johannesburg, in fulfilment of the requirements for the degree of Doctor of Philo-
sophy.

Johannesburg, November 2007

Declaration

I declare that this thesis is my own, unaided work, except where otherwise ac-
knowledged. It is being submitted for the degree of Doctor of Philosophy in the
University of the Witwatersrand, Johannesburg. It has not been submitted before
for any degree or examination in any other university.

Signed this day of 20

Muhammad Salman Aslam Minhas.

i
Abstract

The work presented extends and contributes to the understanding of the dynamic
behaviour of large power transformer windings under short circuits. A simple yet
accurate method of prediction of electromagnetic forces is developed and used as
input to the dynamic mechanical model. This work also explores non-linearity of the
pressboard material under dynamic loading and successfully models it to compute
characteristics like stress-strain and damping. The results of pressboard model are
used in the final model of a full transformer and the simulated predictions compare
very favourably with actual measurements. The model proves that for small radial
movements, the axial and radial behaviours are independent of each other.

ii
Acknowledgements

A very special thank you to Prof J P Reynders for all his kindness, patience, support
and guidance.

An unforgettable thank you to Prof R Iwankiewicz for his guidance and support.

Thank you to Mr Wessel Benecke and Rotek Engineering for supplying the test
transformer.

Thank you to Mr Harry Fellows for his support from the workshop in making all
the components for the experimental work.

iii
To my dad who raised me to be what I am, who guided and supported me in all my
endeavours. To my mum for her unforgettable and unselfish sacrifice to put me
through university. To my wife for her ongoing support...

To the Almighty, without whom all the above would just be a passing breeze.

iv
Contents

Declaration i

Abstract ii

Acknowledgements iii

Contents v

List of Figures xi

List of Tables xvii

List of Symbols xviii

1 Introduction 1

2 Electromagnetic forces in transformer windings 4

2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

2.2 Electromagnetic force on a conductor in a

magnetic field . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

2.3 Magnetic field in a two-winding transformer . . . . . . . . . . . . . . 5

2.4 Axial electromagnetic forces . . . . . . . . . . . . . . . . . . . . . . . 7

2.5 Radial electromagnetic forces . . . . . . . . . . . . . . . . . . . . . . 7

v
3 Prediction of axial electromagnetic forces 9

3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

3.2 Calculation of the axial force . . . . . . . . . . . . . . . . . . . . . . 9

3.3 Empirical or approximate methods . . . . . . . . . . . . . . . . . . . 10

3.3.1 Indirect measurement of axial force . . . . . . . . . . . . . . . 10

3.3.2 Residual ampere-turn method . . . . . . . . . . . . . . . . . . 15

3.4 Two-dimensional methods . . . . . . . . . . . . . . . . . . . . . . . . 19

3.4.1 Two-dimensional method of images . . . . . . . . . . . . . . . 22

3.4.2 Two-dimensional graphical image method . . . . . . . . . . . 23

3.5 Calculation using Fourier series . . . . . . . . . . . . . . . . . . . . . 23

3.6 Methods using digital computers . . . . . . . . . . . . . . . . . . . . 25

3.6.1 Smythes solution . . . . . . . . . . . . . . . . . . . . . . . . . 26

3.6.2 Rabins Solution . . . . . . . . . . . . . . . . . . . . . . . . . 27

3.6.3 Solution by analogue computer . . . . . . . . . . . . . . . . . 27

3.6.4 Image method with discrete conductors . . . . . . . . . . . . 27

3.6.5 Finite element method (FEM) . . . . . . . . . . . . . . . . . 30

3.7 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33

3.8 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33

4 Calculation of radial electromagnetic forces in concentric windings 34

4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34

4.2 Prediction of radial electromagnetic forces . . . . . . . . . . . . . . . 34

4.3 Elementary methods to calculate radial electromagnetic forces . . . . 35

vi
 4.4 Forces in the outer winding . . . . . . . . . . . . . . . . . . . . . . . 36

4.4.1 Hoop stress in disk windings . . . . . . . . . . . . . . . . . . 38

4.5 Forces on the inner winding . . . . . . . . . . . . . . . . . . . . . . . 38

4.6 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39

4.7 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40

5 Dynamic axial behaviour of transformer windings 41

5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41

5.2 Axial model of power transformer windings . . . . . . . . . . . . . . 42

5.3 Dynamic axial model . . . . . . . . . . . . . . . . . . . . . . . . . . . 43

5.4 Implementation of model . . . . . . . . . . . . . . . . . . . . . . . . 47

5.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48

6 Dynamic radial behaviour of transformer windings 49

6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49

6.2 Radial behaviour of windings . . . . . . . . . . . . . . . . . . . . . . 49

6.3 Radial model of a two-winding transformer . . . . . . . . . . . . . . 50

6.4 Kinetic energy of inner and outer disks . . . . . . . . . . . . . . . . . 54

6.5 Strain (potential) energy . . . . . . . . . . . . . . . . . . . . . . . . . 59

6.5.1 Potential energy of copper rings . . . . . . . . . . . . . . . . 59

6.5.2 Potential energy (PE) stored in pressboard ribs (radial spacers) 62

6.6 Generalized forces in radial direction . . . . . . . . . . . . . . . . . . 67

6.7 Modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69

6.8 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69

vii
7 Physical properties of pressboard under varying pre-stress 70

7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70

7.2 Physical properties of transformer board . . . . . . . . . . . . . . . . 71

7.3 Dynamic stress-strain test apparatus . . . . . . . . . . . . . . . . . . 71

7.3.1 Displacement and force sensors . . . . . . . . . . . . . . . . . 72

7.4 Experimental procedure . . . . . . . . . . . . . . . . . . . . . . . . . 74

7.5 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78

7.5.1 Stress-Strain properties at zero pre-stress . . . . . . . . . . . 78

7.5.2 Stress-Strain properties at low pre-stress (1.42 N/mm2 ) . . . 80

7.5.3 Stress-Strain properties at low to medium pre-stress (4.3 N/mm2 ) 81

7.5.4 Stress-Strain properties at medium pre-stress (7.1N/mm2 ) . . 82

7.5.5 Stress-Strain properties at medium-high pre-stress (10 N/mm2 ) 83

7.5.6 Stress-Strain properties at high pre-stress (14 N/mm2 ) . . . . 84

7.6 Modeling of experimental set up as spring-mass

system . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84

7.6.1 Mathematical model . . . . . . . . . . . . . . . . . . . . . . . 85

7.6.2 Damping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91

7.7 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92

7.8 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93

8 Combined transformer model 94

8.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94

8.2 Components of the model . . . . . . . . . . . . . . . . . . . . . . . . 95

8.2.1 Axial spacers (pressboard) . . . . . . . . . . . . . . . . . . . . 95

viii
 8.2.2 Lumped masses, springs and dashpots . . . . . . . . . . . . . 95

8.3 Combined model of a transformer winding . . . . . . . . . . . . . . . 96

8.4 Axial and radial models . . . . . . . . . . . . . . . . . . . . . . . . . 96

8.5 Coupled dynamic model . . . . . . . . . . . . . . . . . . . . . . . . . 97

8.6 Test transformer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97

8.7 Test setup and procedure . . . . . . . . . . . . . . . . . . . . . . . . 98

8.8 Simulated and measured results . . . . . . . . . . . . . . . . . . . . . 100

8.9 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105

8.10 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108

9 Conclusion and recommendations 109

9.1 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109

9.2 Recommendations for further research . . . . . . . . . . . . . . . . . 110

A Matlab source codes 115

A.1 Input code . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115

A.2 Electromagnetic force . . . . . . . . . . . . . . . . . . . . . . . . . . 121

A.2.1 Smaller routines used in electromagnetic calculation . . . . . 127

A.3 Code related to axial dynamic behaviour . . . . . . . . . . . . . . . . 128

A.4 Code related to radial dynamic behaviour . . . . . . . . . . . . . . . 130

A.4.1 Code used in radial behaviour for position of displacement . . 134

B Simulink models 136

C Test transformer design and dimensions 140

ix
C.1 Test transformer design details . . . . . . . . . . . . . . . . . . . . . 140

x
List of Figures

2.1 Force exerted on a conductor in magnetic field . . . . . . . . . . . . 5

2.2 Magnetic flux pattern of a two-winding transformer . . . . . . . . . . 6

2.3 Axial and radial forces in concentric windings when the windings are
axially non-symmetrical . . . . . . . . . . . . . . . . . . . . . . . . . 7

2.4 Radial electromagnetic forces in concentric transformer windings with

axial symmetry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

3.1 Method of calculation of radial flux by measurement of volt-per-turn

[Waters, 1966] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

3.2 Axial compression curve for untapped transformer windings [Waters,

1966; Franklin and Franklin, 1983; Waters, 1953] . . . . . . . . . . . 14

3.3 Determination of residual ampere-turns of winding tapped at top end

[Waters, 1966; Franklin and Franklin, 1983] . . . . . . . . . . . . . . 16

3.4 Arrangement of tappings and resulting residual-ampere-turns . . . . 19

3.5 Polar diagram of flux density for the middle limb of a transformer
with 10% tapped from the middle of the outer winding [Waters, 1966;
Franklin and Franklin, 1983] . . . . . . . . . . . . . . . . . . . . . . . 20

3.6 Winding representation as infinitely long two-dimensional straight

conductor [Waters, 1966; Franklin and Franklin, 1983] . . . . . . . . 21

3.7 Contiguous winding sections used in axial force calculation and images
representing the effect of the core . . . . . . . . . . . . . . . . . . . . 22

3.8 Two-dimensional representation of transformer windings with images

[Waters, 1966] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24

xi
3.9 Idealized core-type transformer with curvature of windings . . . . . . 26

3.10 Discrete image method . . . . . . . . . . . . . . . . . . . . . . . . . . 28

3.11 System of images in discrete image method for a two-winding trans-

former . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29

3.12 Flow chart of system of the images in discrete image method for a
two-winding transformer . . . . . . . . . . . . . . . . . . . . . . . . . 31

3.13 Magnetic field in the test transformer under short circuit conditions. 32

3.14 Comparison of results of discrete image method and Finite Element

Method for inner winding. . . . . . . . . . . . . . . . . . . . . . . . 32

3.15 Comparison of results of discrete image method and Finite Element

Method for outer winding. . . . . . . . . . . . . . . . . . . . . . . . 33

4.1 Two-dimension representation of concentric transformer windings for

the calculation of radial electromagnetic forces showing axial flux and
resulting radial force [Waters, 1966] . . . . . . . . . . . . . . . . . . . 35

4.2 Radial electromagnetic forces in concentric transformer windings [Wa-

ters, 1966] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36

4.3 Mean hoop stress calculating method [Waters, 1966] . . . . . . . . . 37

4.4 Comparison of results of discrete image method and finite element

method for inner winding . . . . . . . . . . . . . . . . . . . . . . . . 40

4.5 Comparison of results of discrete image method and finite element

method for outer winding . . . . . . . . . . . . . . . . . . . . . . . . 40

5.1 Physical construction of transformer winding. On the top left, a photo

shows the full winding and the right is the magnified section of the
winding showing axial spacers. The bottom section of photo shows
the duct sticks (ribs). . . . . . . . . . . . . . . . . . . . . . . . . . . . 42

5.2 Lumped mass model of a two-winding transformer . . . . . . . . . . 44

5.3 The natural frequencies of test transformer in axial direction . . . . 47

6.1 Cross-sectional view of a transformer winding . . . . . . . . . . . . . 51

xii
6.2 Radial and circumferential displacements, u and v the inner winding
ring. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52

6.3 Simplest mode of vibration breathing motion uniform radial expan-

sion and contraction . . . . . . . . . . . . . . . . . . . . . . . . . . . 52

6.4 Inner and outer winding represented as elastic rings coupled with
springs and dashpots . . . . . . . . . . . . . . . . . . . . . . . . . . . 53

6.5 Angular position of core and ribs. The ribs are represented as coupling
springs between the core and the inner winding and between the inner
and outer winding . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63

6.6 The natural frequencies of radial behaviour of test transformer . . . 69

7.1 The picture of the dynamic stress-strain tester . . . . . . . . . . . . 72

7.2 The model of the apparatus used to measure the physical properties
of pressboard . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73

7.3 Detail drawing of Stress-strain apparatus . . . . . . . . . . . . . . . 75

7.4 Displacement sensor construction . . . . . . . . . . . . . . . . . . . . 76

7.5 Force sensor construction detail . . . . . . . . . . . . . . . . . . . . . 77

7.6 Applied force, reaction force, resulting displacement, measured stress

strain characteristics and fitted stress-strain curve used in 1-mass
model calibration at pre-stress value of 0 N/mm2 . . . . . . . . . . . 79

7.7 Applied force, reaction force, resulting displacement, measured stress

strain characteristics and fitted stress-strain curve used in 1-mass
model calibration at pre-stress value of 1.42 N/mm2 . . . . . . . . . 80

7.8 Applied force, reaction force, resulting displacement, measured stress

strain characteristics and fitted stress-strain curve used in 1-mass
model calibration at pre-stress of 4.3 N/mm2 . . . . . . . . . . . . . 81

7.9 Applied force, reaction force, resulting displacement, measured stress

strain characteristics and fitted stress-strain curve used in 1-mass
model calibration at pre-stress of 7.1 N/mm2 . . . . . . . . . . . . . 82

xiii
7.10 Applied force, reaction force, resulting displacement, measured stress
strain characteristics and fitted stress-strain curve used in 1-mass
model calibration at pre-stress of 10 N/mm2 . . . . . . . . . . . . . . 83

7.11 Applied force, reaction force, resulting displacement and measured

stress strain characteristics at pre-stress value of 14 N/mm2 . . . . . 84

7.12 Experimental setup represented as spring-mass system . . . . . . . . 85

7.13 Comparison of measured and calibrated results from the model at

pre-stress of 0 N/mm2 . . . . . . . . . . . . . . . . . . . . . . . . . . 86

7.14 Comparison of measured and calibrated results from the model at

pre-stress of 1.42 N/mm2 . . . . . . . . . . . . . . . . . . . . . . . . 86

7.15 Comparison of measured and calibrated results from the model at

pre-stress of 4.3 N/mm2 . . . . . . . . . . . . . . . . . . . . . . . . . 87

7.16 Comparison of measured and calibrated results from the model at

pre-stress of 7.1 N/mm2 . . . . . . . . . . . . . . . . . . . . . . . . . 87

7.17 Comparison of measured and calibrated results from the model at

pre-stress of 10 N/mm2 . . . . . . . . . . . . . . . . . . . . . . . . . 88

7.18 Comparison of measured and calibrated results from the model at

press-stress of 14 N/mm2 . . . . . . . . . . . . . . . . . . . . . . . . 88

7.19 Schematic representation of the stress-strain properties of two non-

linear springs in a back to back arrangement . . . . . . . . . . . . . 90

7.20 Simulink model of mass-spring system . . . . . . . . . . . . . . . . . 92

8.1 Components of mechanical model of a transformer . . . . . . . . . . 95

8.2 Proposed two-winding dynamic axial model . . . . . . . . . . . . . . 98

8.3 Dynamic axial and radial models combining technique . . . . . . . . 99

8.4 The effects of friction on axial motion . . . . . . . . . . . . . . . . . 100

8.5 The effects of friction on radial motion . . . . . . . . . . . . . . . . . 100

8.6 The cross-section of the core of the test transformer . . . . . . . . . 101

xiv
8.7 The detail of the winding configuration of the test transformer . . . 102

8.8 Single phase two-winding test transformer in experimental setup . . 102

8.9 Comparison between stand-alone axial model and combined model of

the test transformer . . . . . . . . . . . . . . . . . . . . . . . . . . . 103

8.10 Single phase two-winding test transformer . . . . . . . . . . . . . . . 103

8.11 Simulated and measured results of axial displacement at a pre-stress

of 7.1 N/mm2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104

8.12 Simulated and measured results of radial displacement at a pre-stress

of 7.1 N/mm2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104

8.13 Simulated and measured results of axial displacement at a pre-stress

of 14 N/mm2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105

8.14 Simulated and measured results of radial displacement at a pre-stress

of 14 N/mm2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105

8.15 Construction of the force and displacement sensors. The force sensor
was used to measure the pre-stress and the two displacement sensors
were used for axial and radial displacement . . . . . . . . . . . . . . 107

8.16 The cross-section of the transformer disk showing the profile of the
area interfacing with the vertical spacers (pressboard) . . . . . . . . 107

B.1 Model to simulate axial behaviour of test test transformer implemen-

ted in Simulink . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136

B.2 Model of the spring used as subsystem in axial model of a winding

(Fig B.1) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137

B.3 Axial model of one disk used as a subsystem in the combined Simulink
model (Fig B.4) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137

B.4 The representation of the combined model of the test transformer

implemented in simulink . . . . . . . . . . . . . . . . . . . . . . . . . 138

B.5 Radial model of a set of inner and outer disks of the test transformer 139

C.1 Dimensions of the test transformer . . . . . . . . . . . . . . . . . . . 140

xv
C.2 Design sheet of the inner winding of the test transformer . . . . . . . 141

C.3 Design sheet of the inner winding of the test transformer . . . . . . . 142

xvi
List of Tables

3.1 Arrangement of tappings and corresponding values of [Franklin and

Franklin, 1983] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19

7.1 Applied pre-stress values. . . . . . . . . . . . . . . . . . . . . . . . . 78

7.2 Damping coefficient at different pre-stress values . . . . . . . . . . . 91

xvii
List of Symbols

F Force exerted on a conductor in magnetic field

i Current carried by the conductor in the magnetic field

B Flux density

Dm Diameter of the limb of transformer

dx thickness of the winding element

AT Ampere-turn

Br Redial flux density

Isc Symmetrical short-circuit current

It Current at which the test is conducted to measure the voltage

a Fractional difference in winding height

It Test current

Isc Symmetrical short circuit current

U Rated apparent power per limb in kVA or radial displacement of

outer winding or nominal voltage

Pc Axial compression force

ez Per unit impedance voltage

f Frequency, Hz

h Winding height

xviii
 NI Ampere turns

lef f effective length of the path of the radial flux

Permeance per unit axial length of the limb

Pa Axial force between two components of a winding

Dm Mean diameter of the transformer

k Total winding length occupied by one of the contiguous portion of

the winding

d Radial distance between two portions of one winding

the angle subtended by end from the position of the winding

ni AT of one turn of the coil

Dc Mean diameter of the turn

r1 ; r2 Distances from the winding ends 3.6

jd Current density

j imaginary operator in complex algebra

i1 ; i2 current in the inner and outer winding conductor

mean Mean hoop stress at the peak of the first half cycle of current

Wcu I 2 Rdc loss in the winding in kW at 75o at full load

pr radial force per mm of conductor

Ac cross sectional area of each conductor

nc number of conductors in each disk

Dw mean diameter of the winding

x the local displacement of mass

c damping coefficient

xix
 k spring constant

xin the local displacement of disk of inner winding

xot the local displacement of disk of outer winding

position angle determining the position of the rib

u radial displacement of inner ring (inner winding)

v circumferential displacement for inner winding

Ii moment of inertia of inner ring (inner winding)

Io moment of inertia of outer ring (outer winding)

ao constant term representing uniform radial vibration

Ti kinetic energy of the inner ring

To kinetic energy of the outer ring

T Total kinetic energy of a ring due to flexural vibration

mass density of copper

Ai cross-sectional area of the inner ring

Ao cross-sectional area of the outer ring

H Matrix related to the kinetic energy of copper ring

Ep Strain potential energy (PE)

Epi PE of inner ring

Epo PE of outer ring

i
Epins PE of inner insulation rib (between core and inner winding)

o
Epins PE of outer insulation rib (between inner and outer winding)

elasticity of copper conductor

xx
 Ii moment of inertia of the inner ring

D Matrix related to stain potential energy of the copper ring

EPec Potential energy due to extension of the ring

g(x) nonlinear stress strain characteristics of pressboard

xxi
Chapter 1

Introduction

A power transformer is the single most expensive component in generation, trans-

mission and distribution substations. The importance of the transformer is not only
due to its capital cost but the cost of undelivered energy in the event of failure,
which makes the financial losses unaffordable. The rapid growth of power systems
has given a raise to increased fault levels and the transformers should be able to
handle the fault currents to ensure the reliability of the power system.

With growth of the power systems comes not only an increase in the short circuit
levels but an increase in the system voltages. The higher voltages necessitate a
more complex winding arrangement. There has been a better understanding of
material properties in the last 15-20 years. Transformer manufacturers are exploiting
this better understanding and produce transformers of smaller size with saving of
material and transport costs. Unfortunately, this has lead to both conductor and
insulating material operating close to their limits and an increase in failure in the
early years of transformer life.

Research on the winding dynamics under short circuit conditions received a lot of
attention between 1960 and the late 1970s. Many models were developed [Patel,
1972; Tournier et al., 1964; Madin and Whitaker, 1963b; Watts, 1963] to study the
dynamic behaviour. All these models were focused on the axial movements and
radial strength was ensured by simple and experience-based rules. All the models
treated axial spacers (pressboard) as a linear springs except Patel [1972]. Patel
[1972] developed a very detailed axial model which takes into account the non-
linearity of the pressboard but neglects hysteresis and damping. He also assumes
that pressboard only offers stiffness under compression and expansion comes under
zero force. Swihart and Wright [1976] suggested a model of pressboard material.
The model was able to predict the non-linear properties of the pressboard with the

1
variation of pre-stress. However, the model was very complex and was not easily
implementable in the modeling of large power transformers.

In this thesis, a different approach was used to handle the non-linear behaviour of
pressboard. The dynamic stress-strain properties were measured and fed into the
mathematical model as a lookup table. The measured properties were simplified to
make the model manageable. Also the model presented is more realistic as the axial
and radial movements are interlinked.

The objectives of this research are:

To understand the dynamic stress-strain characteristics of transformer press-

board under varying pre-stress as the strength and stiffness of large windings
depend upon the dynamic behaviour of the pressboard. Also to measure these
characteristics under dynamic loading.

To derive a model for oil impregnated pressboard which can be used to simulate
the dynamic behaviour under varying pre-stress and transient loading of a
transformer. This model will be used to simulate a complete transformer.

To compare the simulated winding behaviour with the behaviour of an actual

winding with a view to assessing the validity of the model.

The structure of the thesis is as follows:

Chapter 2: Electromagnetic forces in transformer windings

A brief description of electromagnetic forces is given; how the electromagnetic forces
are experienced in a magnetic field; a brief outline of the forces in a two winding
transformer.

Chapter 3: Prediction of axial electromagnetic forces

The available methods of force calculation are described. The most accurate methods
are discussed in detail and compared. The method used in the thesis for force
prediction is described and compared with finite element method.

Chapter 4: calculation of radial electromagnetic forces in concentric

windings
The methods of radial force calculation are briefly described.

2
Chapter 5: Dynamic axial behaviour of transformer windings
A brief description of existing models is given. The equation of motion was de-
rived for a two concentric winding transformer. The implementation of model in
r

Simulink with non-linear pressboard is given and simulated results are presented.

Chapter 6: Dynamic radial behaviour of transformer windings

The dynamic radial behaviour of concentric windings is discussed with a brief history.
r

The equation of motion is derived and model is implemented in Simulink and the
results are described.

Chapter 7: Physical properties of pressboard under varying pre-stress

The chapter gives a brief history of the past research. The design of a dynamic stress-
strain characteristics tester is discussed. The stress-strain properties of pressboard
were measured and modeled to calibrate the model and to calculate the constants.

Chapter 8: Combined transformer model

Describes the construction of combined transformer model. Coupling of axial and
radial behaviour, test setup and test procedure, comparison of simulated and meas-
ured results.

Chapter 9: Conclusions
The findings of the thesis and further research is identified.

Additional supporting information is given in the following appendices.

TM
Appendix A: Matlab source codes
TM
The Matlab codes to predict oscillating electromagnetic forces are given. The
r

codes to calculate the inputs to the stand-alone and combined Simulink models
is also given in this appendix.

r

Appendix B: Simulink models
The combined and individual model implementation is presented in this appendix
with the detail of main and subsystems.

Appendix C: Test transformer design and dimensions

The brief description of test transformer design.

For convenience of the reader, each chapter and appendix starts with a brief intro-
duction which explains the main areas covered in the chapter or appendix.

3
Chapter 2

Electromagnetic forces in transformer

windings

2.1 Introduction
The determination of forces in the winding of transformers under short-circuit con-
ditions has been a matter of prime interest since the rapid growth in power systems
and steady increase in the size of transformers. The work of predicting theses forces
accurately has been in progress since early 1920s. The expansion of the power sys-
tems and increase in fault levels has made this issue more important than ever and to
top it is the complex geometrical arrangements of the windings due to higher trans-
mission voltages. The design and construction of the modern power transformer
depends on the accurate prediction of these forces to avoid in-service failures and to
reduce the replacement cost and cost of undelivered energy.

The methods available to calculate these forces range from being approximate to
relatively more accurate. Approximate electromagnetic force prediction techniques
are based on simplified assumptions. However, more accurate methods use more
realistic criteria for electromagnetic force calculation.

In this chapter, the nature of the electromagnetic forces in concentric windings of

transformers is discussed. The scope of this work is limited to the two-winding core
type transformers. The multi windings transformers and shell type are not discussed
here. The methods of prediction of electromagnetic forces are discussed in Chapters
3 and 4.

4
2.2 Electromagnetic force on a conductor in a
magnetic field
When a conductor carrying current is placed in a magnetic field, it experiences a
force. By Amperes Law, each rectilinear element of conductor of length dl, carrying
a current i, in a magnetic field of flux density B, perpendicular to it, is subjected to
a force d F as mentioned in Eq 2.1. This force is perpendicular to the plane formed
by the magnetic field and elemental conductor (Fig 2.1). Figure 2.1 also show the
direction of the force with respect to the magnetic field and direction of the current
(Flemings left hand rule).

d F = i B dl (2.1)

dl

dF
i

Figure 2.1: Force exerted on a conductor in magnetic field

2.3 Magnetic field in a two-winding transformer

The approximate magnetic field in a simple two-winding transformer is shown in

the Fig 2.2. Although the construction of the transformer is the simplest, the field
is by no means simple and cannot be calculated by simple methods. However, it
is evident that this type of field will produce forces tending to separate the two
windings, resulting in an outward force in the outer winding and the force on the
inner winding is inward and compressive. If the windings are of the same length and
accurately placed so that there is no axial displacement, each winding is subjected

5
to a compressive force (discussed in Section 2.4) and there is no force to move the
windings in the axial direction. Since the windings carry large currents in opposite
directions (under through faults), any displacement from the precise balance position
will lead to a large axial force tending to increase the displacement and produce more
asymmetry in the windings. So the windings are designed in such a way that they
have opposite radial forces but no unbalanced axial forces which is almost impossible
to achieve. In real transformers, the windings can have a small initial displacement
from the balanced position due to the construction limitations [Final Report: Cigre
Working Group 12-04, 1979].

Current directions

Core

Inner winding Outer winding

Figure 2.2: Magnetic flux pattern of a two-winding transformer

When a transformer is over-loaded or encounters a through fault, the primary and

secondary ampere turns are in opposition with reference to the core, but this effect
in the space between the two windings (inter winding duct) is cumulative and gives
rise to a magnetic field in the inter-winding space (duct). This cumulative magnetic
field causes leakage flux rise (flux linking to the one winding only) and mutual forces
between the windings [Norris, 1957]. These forces have two components:

Repulsive radial forces between inner and outer winding

Axial forces, which can be further divided into two types,

Axial compression; and

Unbalanced axial forces due to asymmetry

6
2.4 Axial electromagnetic forces
The radial component of leakage flux linking the windings towards the ends is mainly
responsible for the axial electromagnetic forces. In a transformer, if ampere-turns
(AT) are balanced in windings, the axial forces have a compressive nature and tend
to squeeze the winding in the middle. In axially symmetrical windings these forces
were thought of less or of no importance as the compressive strength of the winding in
the axial direction was thought to be much higher than the forces even under severe
conditions [Arturi, 1992; Say, 1958; Franklin and Franklin, 1983]. If there is an
asymmetry in the winding heights due to the tap position or for some other reasons,
the ampere-turn unbalance increases and gives rise to repulsive forces, tending to
break the winding apart from the middle. The concept of axial electromagnetic
forces is shown in Fig 2.3.

Force Outer winding

Core

Inner winding

Axial force on
inner winding
Axial component Actual force
of force on the outer
winding

Radial component
of force
Axial force on
outer winding

Figure 2.3: Axial and radial forces in concentric windings when the windings are
axially non-symmetrical

2.5 Radial electromagnetic forces

In the case of a circular conductor (the winding coil of a transformer) or a coil, the
current produces a force uniformly distributed around the periphery unless there
is magnetic asymmetry. The radial electromagnetic forces develop when the coil
current interacts with the axial component of its own magnetic flux.

In a transformer, the flux interacting with the windings occupies the space between
the two windings as shown in Fig 2.2. Consequently, the outer coil is subjected
to a pressure to extend the diameter of the coil, but the inner coil is under an
external pressure and tends to collapse to the core (Fig 2.4). The circular coils are

7
the preferable choice in a transformer as they are the strongest shape to withstand
the radial pressure mechanically [Say, 1958].

Outer winding

Core
Forces on windings

Inner winding

Inner winding
Core

Resulting force
only radial

Outer winding

Figure 2.4: Radial electromagnetic forces in concentric transformer windings with

axial symmetry

Chapters 3 and 4 deal with calculation of the axial and the radial electromagnetic
forces respectively.

8
Chapter 3

Prediction of axial electromagnetic

forces

3.1 Introduction
Forces in an axial direction can destabilize the winding causing a collapse of the
winding and fracture or displacement of the end insulation (end ring) or clamping
system. Excessive axial forces can be responsible for the bending of the conductor
between the axial spacers or by compressing the insulation to such an extent to
cause slackness and reduction of pre-stress which can lead to the displacement of
spacers and subsequent failure. The destructive nature of these forces has made the
problem of calculating the magnitude of the axial force important and has received
considerable attention from researchers since the early 1920s.

The precise solution of the radial leakage field and the axial forces in transformer
windings have been determined by various authors using a number of methods.
These methods are complex and require the use of a computer if results are to be
obtained quickly. However, elementary or simplified methods are also available.
One of the simple methods, the residual ampere-turn method, gives reliable results.
Attempts to produce closer approximations add greatly to the complexity without
a corresponding gain in accuracy.

In this chapter, different methods of predicting electromagnetic force and their suit-
ability for use in mechanical behaviour studies are briefly described.

3.2 Calculation of the axial force

The axial component or the forces in a transformer with conventional concentric
windings cannot be calculated with high accuracy by elementary methods mainly

9
because the curvature cannot be taken into account without using complex solutions
which require the use of a computer. Before computers were available, a great
deal of ingenuity was used in devising approximate methods for the calculation of
axial forces. Since the rigorous solutions were too complex to be of any practical
value, the usual approach was to make simplifying assumption, e.g. each unit length
of the circumference of the winding was a portion of an infinitely long straight
coil side. This enabled the radial component of the field to be calculated at any
point. The effect of curvature was ignored or was taken into account by the use
of empirical factors [Waters, 1966]. Measurements have shown that such methods
give fairly accurate results in many cases, but the accuracy was poor for complex
winding arrangements like single turn or high current windings and complex tapping
arrangements. In general, these methods can give the axial force upon the whole
winding or on the half or quarter of a winding with good accuracy, but are of little
value in calculating the force upon a single coil or conductor, particularly if the
winding arrangement is unusual.

However, these approximate methods are of great use to the designers since, they
indicate quickly whether or not a given arrangement of windings will result in high
axial force etc. In general, the methods used to calculate forces can be divided into
following classes [Final Report: Cigre Working Group 12-04, 1979].

Elementary methods

Simple formula methods

Sophisticated or more accurate methods

3.3 Empirical or approximate methods

These methods use empirical formulas to predict forces and are based on simplified
theory and assumptions. A correction factor, based on experience or experimental
findings, is applied to correct the results. These methods are explained below.

3.3.1 Indirect measurement of axial force

A simple method, developed by the Electrical Research Association (ERA) is ex-
plained in the book Franklin and Franklin [1983], for measurement of total axial
force on the whole or part of a concentric winding. This method does not indicate
how the force is distributed around the circumference of the winding but this is not

10
a major disadvantage as the force along the circumference is not required, instead
maximum force is needed.

Axial force calculation from radial flux measurement

If the axial flux linked with each coil of a disc winding at a given current is plotted
against the axial position, the resultant curve represents the axial compression of
the winding [Franklin and Franklin, 1983; Waters, 1966].

The flux density of the radial component of leakage field is proportional to the
derivative of axial flux with distance along the winding. The curve of axial flux
plotted against distance thus represents the integration of the radial flux density and
gives the compression curve of the winding [Franklin and Franklin, 1983; Waters,
1966].

Volt per turn method

The voltage per turn is a measure of the axial flux. The voltage of each disc coil
is measured, and divided by the number of turns in the disk this voltage per turn
is plotted against the length of the winding [Waters, 1953]. This method can only
be applied to a continuous disc winding by piercing the insulation at each crossover
and the test is most conveniently carried out with the transformer short-circuited
as for the copper-loss test.

Consider an axial force on a small axial length dx of a transformer winding as

shown in the Fig 3.1. The element of the winding under consideration is situated
at a distance x from the end a and occupies the full radial thickness of the winding.
The length of the element can be calculated as Dm around the circumference of
the winding where Dm is the diameter of the limb. If the winding has AT as the
ampere-turn per unit length, the element dx contains AT dx ampere-turns. The
axial force upon the whole element can be calculated as:

AT dx Br Dm

Where Br is the radial component of flux density.

If the ampere-turn/unit length is constant along the whole length of the portion ab,
then the total axial force on the portion is
Z b
Fab = AT Dm Br dx
a

11
 Rb
Dm a Br dx is the total radial flux passing out of the surface of the cylinder . It
can be concluded that the axial force upon any portion of a winding, having uniform
ampere-turns is given by the product of the ampere-turn per unit length and the
total radial flux.

The radial flux at a point is normal to the cylinder surface. It is the part of the
axial flux which enters the winding from one end and does not come out at the other
end. Hence it is the algebraic difference between the axial flux at the two ends of
the winding.

The axial flux at any point in the winding is proportional to the induced voltage
per turn at that point. Hence the total radial flux of any portion of the winding
and the resulting axial forces are proportional to the algebraic difference between
the induced voltage per turn at the two ends of the winding. So if the distribution
of the induced voltage per turn is known along the winding of a transformer, the
axial force on the winding can be calculated. In fact, if the induced volt-per-turn
along the winding is measured at a suitable current, the axial force on the winding
can be calculated. However, these forces are the total forces on that winding and
the variation around the circumference is not indicated.

dx Dw
Br

Core
b

Figure 3.1: Method of calculation of radial flux by measurement of volt-per-turn

[Waters, 1966]

12
Radial flux density

If the axial flux at any point x from the end of the winding is denoted by A , the
reduction or increase in flux is a short length dx is given by

da = Dm Br dx (3.1)

and
dA 1
Br =
dx Dm

If V is the rms voltage induced per turn at any point, A is the peak axial flux and
Br is the radial flux density then Equation 3.1 can be written as
dV 1
Br = (3.2)
dx 4.44f Dm

Where x and Dm are in meters and f is the frequency of the sinusoidal flux A .

If the curve of volt-per-turn is plotted against distance along the winding, the slope
of the curve represents the radial flux density at any point to the scale given by the
Equation 3.2. This method is also valid if the turns per unit length over the length
of the winding are not constant.

Volt-per-turn measurement for axial force calculation

Consider a transformer with two windings, having N I rms ampere-turns per unit
length and induced voltages per turn of V1 /N1 and V2 /N2 . The total peak radial
flux is given by
V1 V2
N1 N2
r = (3.3)
4.44f
The peak axial force upon the winding

V1 V2
N1 N2
FA = 2 N I (3.4)
4.44f
hence
rms ampere-turns rms volt-per-turn difference
Peak axial force =
4.44 f

The scale of force at 50 Hz is given by Franklin and Franklin [1983]

rms ampere-turns per mm
1 volt (rms) = kN (peak)
15750

13
 Axial compressive forces (kN)
Sum of compressions

Inner winding

Outer winding

Top Bottom

Figure 3.2: Axial compression curve for untapped transformer windings [Waters,
1966; Franklin and Franklin, 1983; Waters, 1953]

To convert the measured voltages to forces under short-circuit conditions the values
must be multiplied by (1.8Isc /It ) where Isc is the symmetrical short-circuit current
and It , the current at which the test is carried out.

To obtain the compression curve it is necessary to know the points of zero compres-
sion, and these have to be determined by inspection. This is not difficult since each
arrangement of windings produces zero points in well defined positions.

Figure 3.2 shows axial compression curves obtained on a transformer having un-
tapped windings of equal heights. There are no forces tending to separate the coils
in the axial direction. The ordinates represent the forces on coils at all points due to
the current in the windings. The shape of the curve shows that only in the end coils
are there appreciable forces, as the end coils have maximum radial flux linked to
them. The dotted curve shows the sum of the axial compressive forces for the inner
and outer windings. This method has been explained in detail by Waters [1966].
The maximum force is given by [Franklin and Franklin, 1983]
510 U
Pc = kN (3.5)
ez f h
Where:

U = nominal voltage
Pc = axial compression forces, N
ez = per unit impedance voltage
f = frequency, Hz
h = axial height of the winding in mm

14
This is the force at the peak of the first half cycle of fault current, assuming an
asymmetry factor of 1.8.

It is to be noted that the forces in a transformer winding depend only on its po-
sition and on the total ampere-turns, and not on the physical size. Thus, smaller
scaled model transformers were suitable for investigating forces. For large units
where calculations were difficult, it was more economical to produce a model and
measure the forces than to carry out elaborate calculations [Waters, 1966; Franklin
and Franklin, 1983]. However, in recent years, a significant development of FEM
software has enabled the force calculation to be accomplished easily but in cases
where the winding and tapping arrangement is complex, the FEM software takes a
long time to calculate forces on different portions of winding and other structures.

The voltage per turn method has proved very useful in detecting small accidental
axial displacements of windings from the normal position during manufacture and
transportation [Waters, 1966]. However, this may require piercing of insulation
which is not a good idea for high voltage transformers and the method is only
suitable for small lower voltage units.

3.3.2 Residual ampere-turn method

This is one of the oldest method, and follows H.O. Stephens1 method of calculating
the interleaved component or reactance . This method has a long history of use
and Say has discussed it briefly in his book [Say, 1958]. This method provides the
solution for axial forces only. The brief explanation of this method is given here but
detail explanation is given by the previous researchers [Waters, 1966; Franklin and
Franklin, 1983; Say, 1958].

Any arrangement of concentric windings in which the sum of the ampere-turns is

zero is split up into two groups each having balanced ampere-turns, one producing
an axial field and the other a radial field. The radially acting ampere-turns are
responsible for the radial flux which causes the axial force in the windings. At the
same time it causes a slight increase in the percentage reactance.

The radial ampere-turns at any point in the winding are calculated by taking the
algebraic sum of the ampere-turns of the primary and secondary windings at that
point and at the end of the windings. A curve plotted for all points is a residual or
unbalanced ampere-turn diagram from which the method gets its name. It is clear
1
Stephens H.O., Transformer Reactance and Loss with nonuniform windings. Elect. Eng., vol
53, Feb 1934, pp 346-349 cited in Waters [1966]

15
that for untapped windings of equal length and without axial displacement have no
residual ampere-turns or forces between the windings. Although there is no axial
thrust between windings, internal compressive forces and forces on the end coils still
are present.

The method of determining the distribution of radial ampere-turns is illustrated in

Fig 3.3 for a simple case of concentric winding having a fraction of the total length
tapped at the end of the outer winding. The two components I and II of Figure 3.3-b
are both balanced ampere-turn groups and, when superimposed, produce the given
ampere-turn arrangement. The diagram showing the radial ampere-turns plotted vs
distance along the winding is a triangle, as shown in Fig 3.3-c, having a maximum
value of a(N Imax ), where (N Imax ) represents the ampere-turns of either the primary
or secondary winding and a is the fractional difference in height of the windings as
shown in the Fig 3.3 (a).

a(NI)
a +

+ - + - -

I II
Windings Symmetrical Unsymmetrical Ampere turn
diagram

(a) (b) (c)

Figure 3.3: Determination of residual ampere-turns of winding tapped at top end

[Waters, 1966; Franklin and Franklin, 1983]

To determine the axial forces, it is necessary to find the radial flux produced by the
radial ampere-turns, or in other words to know the effective length of the path of
the radial flux on all points along the winding. In previous work, an assumption was
made that this length stays constant and does not vary with axial position in the
winding. This was considered to be a rough approximation to obtain simple results
[Waters, 1966; Franklin and Franklin, 1983]. However, tests show that this approx-
imation is reasonably accurate and flux does in fact follow a triangular distribution
curve of the same shape as the residual ampere-turns.

16
The calculation of the axial thrust in the case shown in Figure 3.3 can be calculated.
lef f is the effective length of the path of the radial flux. Since the mean value of
1
the radial ampere-turns is 2 a(N Imax ), the mean radial flux density at the mean
diameter of the transformer limb is

a(N Imax )
Br = 4 107 T (3.6)
2 lef f
and the axial force on either winding of N Imax ampere-turns is

2a(N Imax )2 Dm
Pa = kN (3.7)
1010 lef f

where:

Imax = maximum current (A)

Dm = mean diameter of the transformer limb (mm)
lef f = effective length of the path of the radial flux (mm)
a = fractional difference in winding heights

If

= Dm /lef f ,

is the permeance coefficient per unit axial length of limb. It gives the force
for all windings having the same properties irrespective of physical size. Also the
calculation of the ampere-turns is not difficult. In order to cover all the cases, it is
necessary to study only how varies with the properties of the core, proximity of the
tank, dimensions of the duct, dimensions of the windings and tapping arrangement.

Effective length of path for radial flux lef f

Before digital computers were available, the value of lef f was determined usually by
extending Rogowskis work on the reactance of interleaved windings [Hague, 1929].
Tests carried out on an experimental transformer by the method described by Waters
[1966] showed that expressions determined in this way cannot be applied generally
and these results were confirmed by Klichler [Waters, 1966], who pointed out that
Rogowskis method fails in many cases and suggests an empirical correction based on
experience. He concluded that the length of path for the radial flux in transformers
of normal proportions having a simple tapping arrangement of Fig 3.3 is given by

lef f = 0.222h (3.8)

17
where h is the height of the winding.

Equation 3.7, is applicable for a simple tapping arrangement and would not be used
in practice, but the same method is applicable for all tapping arrangements. The
ampere-turns must be determined, the residual ampere-turn diagram constructed or
calculated, and with the appropriate value of the axial force on the part of either
winding under each loop of the residual ampere-turn diagram can be calculated.
This has been explained by Waters [1966], Franklin and Franklin [1983] and Say
[1958].

The value of used in each case has been studied empirically [Waters, 1966, 1953],
using two transformers specially designed to suit radial flux measurements (The
transformer design is described in detail in the books [Waters, 1966] and [Franklin
and Franklin, 1983]). The values of applies exactly to these particular trans-
formers, which were designed to have widely different values of ratio and (window
height)/(core circle diameter).

The factors such as clearance between winding and core, duct width, proximity of
tank, radial thickness of the windings have an effect on , which is small. The val-
ues given should apply within narrow limits to any transformers having proportions
not too different to those of Fig 3.4. In extreme cases with large duct widths, the
accuracy decreases, as shown in Waters [1966], which gives a comparison between
this method and a more accurate computer calculation for larger changes in config-
uration.

The proximity of the tank increases the value of for the outer limbs of a three-phase
transformer, but had no significant effect on the middle limb. A limited number of
tests showed the presence of the tank did not increase the forces in the outer limbs
to values greater than those in the middle limb. The presence of the tank increases
the forces in a single-phase transformer wound on one limb, but in this case the
value of would not exceed that of the middle phase of a three-phase transformer.

The values of to be used for usual arrangements of tappings are given in Figure
3.4 and Table 3.1 for three-phase balanced loading. They apply to the middle limb,
and the total value of the force on the part of the winding.

The forces calculated with this method are not uniformly distributed around the
circumference, but concentrated in the window [Franklin and Franklin, 1983; Waters,
1953, 1966]. The plot of Figure 3.5 is for a transformer which is 10 % tapped out

18
 NI NI NI NI NI

a
a/2 a/4 a/2

a a/2 a

a/2 a/4 a/2

a b c d e

Figure 3.4: Arrangement of tappings and resulting residual-ampere-turns

Arrangement of Pa

window height window height
Tappings (kN) core circule = 4.2 core circules = 2.3
2a(N I)2
Case A Fig 3.4 1010 5.5 6.4
a(N I)2
Case B Fig 3.4 21010 5.8 6.6
a(N I)2
Case C Fig 3.4 4(1 12 a)1010
5.8 6.6
a(N I)2
Case D Fig 3.4 81010 6.0 6.8
2a(N I)2
Case E Fig 3.4 16(1 12 a)1010
6.0 6.8

Table 3.1: Arrangement of tappings and corresponding values of [Franklin and

Franklin, 1983]

at the middle of the outer winding on all three phases. The slight enhancement
of flux density is not due to the core of the adjacent limbs, but mainly due to the
ampere-turns of the windings of the side limbs.

In a three-phase transformer, the local increase of force above the mean was 25%.
Hence the greatest axial forces are in the window. For a single-phase transformer
wound on two limbs the force per unit of circumference is 50% greater than the
mean and in a three-phase transformer 25% greater than the mean value calculated
from the values of given in Table 3.1.

3.4 Two-dimensional methods

The residual ampere-turn method can provide solution for axial forces in a winding
or a part of the winding which has an asymmetrical distribution of the ampere-turns
and does not take into account the forces present in a uniform untapped winding.

19
 Limb C

o o
20 340

o o
40 320

o o
60 300

Peak flux density, gauss

o o
80 280

Limb B

o o
100 260

o o
120 240

o
140
o
220

o o
160 200

Limb A

Figure 3.5: Polar diagram of flux density for the middle limb of a transformer
with 10% tapped from the middle of the outer winding [Waters, 1966; Franklin and
Franklin, 1983]

Attempts to formulate a simple method which can help the prediction of axial com-
pression in any part of the winding have been made with some success. All the
available methods are based on two-dimensional techniques with suitable correc-
tions for curvature and the effect of the core.

To simplify the problem, Billing [1946] and Waters [1966] suggested that the wind-
ings should be represented by infinitely long straight coil sides; as shown in Figure
3.6. The forces in a length equal to the mean length of turn, are taken as represent-
ative of a transformer. The radial component of flux density near such a straight coil
side is proportional to ln ( rr21 ), and the axial component is proportional to the angle
subtended at the ends (Figure 3.6). By simple integration the axial force between
two contiguous portions of winding occupying the total axial length is calculated as
[Waters, 1966]

Pa = P (k, d) (3.9)

20
 Dm
r1

r2

Figure 3.6: Winding representation as infinitely long two-dimensional straight con-

ductor [Waters, 1966; Franklin and Franklin, 1983]

where
2(N I)2 Dm
P = kN [Waters, 1966; Franklin and Franklin, 1983]
1010 h

Where

Dm = mean diameter of the transformer, mm

h = length of the winding, mm
(N I) = ampere-turns of one winding

and (k, d) is a function of the fraction k of the total winding length occupied by one
of the contiguous portions of winding and d is the radial distance between them also
expressed as a fraction of the winding length. Therefore as determined by Waters
[1966].

k 1 + d2 1k 1 + d2
(k, d) = ln + ln (3.10)
2 k 2 + d2 2 (1 k)2 + d2

1 k 1 1 k 1 1
d tan + tan tan
d d d

21
Tables of this function are available and it is also shown how these may be used to
calculate the forces in any winding arrangement [Waters, 1966].

This method is based on the assumption that the ampere turns are concentrated
along a line at the mean diameter of the winding and d is the radial distance between
winding centers. The method produced close results in comparison with Residual
Ampere Turn method 3.3 for compressive forces [Waters, 1966].

3.4.1 Two-dimensional method of images

To apply the above method to complicated cases, it was proposed to employ the
method of images using an iron boundary as shown in Fig 3.7-b. This representation
is on one side only, and is an attempt to represent the presence of the core.

Images in axial
direction

Radial Radial images

images

Tapping
k
winding
d

h
Core window
Main
winding Windings

Images in axial
direction

(a) (b)

Figure 3.7: Contiguous winding sections used in axial force calculation and images
representing the effect of the core

Waters [1966], suggested that this method does not give accurate results if only
one side boundary is employed. Waters suggested that top and bottom boundaries
representing the yoke can be omitted without loosing accuracy.

22
3.4.2 Two-dimensional graphical image method
This method has been described by P Ignacz2 . The method is similar to one ex-
plained above but in a simplified form by an ingenious device developed by P Ignacz
(Detail of the method is given in Waters [1966]). Instead of carrying out tedious
integration of Equations 3.9 and 3.10, the force on each coil is calculated individually
using the picture given in Fig 3.6 with a simple relation developed empirically by
Ignacz.

r = C

r2
a = C ln
r1

where r and a are the radial and axial components of the force respectively and
2.04 N I
C= N IDc kg
108 h

= the angle subtended by end from the position of the winding (Fig 3.6 )
NI = AT of the winding
Dc = mean length of the turn
r1 , r2 = distances from the winding ends Fig 3.6

More detail of how this method is used is given in Waters [1966].

The method recognizes that the force on a coil, due to the winding of which it is a
part of, can not be calculated without reference to its dimensions. Curves have been
prepared (by taking dimensions into account) to enable this to be done quickly. The
effect of the core is taken into account by the method of images. The results obtained
by this method correlate well with the measured values. The method proved to be
a practical method where calculations are done by hand [Waters, 1966].

3.5 Calculation using Fourier series

Roth [Hague, 1929; Waters, 1966] was the first to attempt accurate calculations of the
forces in a transformer. To start with, he produced a solution in two dimensions only,
using double Fourier Series which became the basis of later work in two dimensions.
In 1936 he produced a solution using cylindrical co-ordinates and correctly took into
account the curvature of the windings. Computer codes of this method are available
[Waters, 1966].
2
Ignacz P, Determination of short-circuit forces in transformer windings. Institute of Electrical
Power Research, Budapest. (in Hungarian) cited in [Waters, 1966]

23
In the 2-dimensional method the windings are considered as infinitely long straight
rectangular bars having the same cross-sections as the windings and uniform current
distribution, in a closed iron duct as shown in Fig 3.8.

Image in axial
direction Iron boubdries
Xo
Windings

X1

w ind o w
Image in radial

Co re
X2 Y2 Yo
direction

Y1
Image in radial
direction
Image in axial
direction

Figure 3.8: Two-dimensional representation of transformer windings with images

[Waters, 1966]

The force on a whole winding can be calculated by taking a length of the bar equal
to the mean length of the turn of the winding. This arrangement is a close approx-
imation to the straight portions of a winding in a shell-type transformer, but it is not
accurate for a core-type transformer. Pichon and Hochart [1958] have compared the
two-dimensional method with the more accurate three-dimensional solution in cyl-
indrical co-ordinates on the computer and preferred the former for large transformers
on account of its simplicity. The forces in the window are accurately predicted by
taking the window dimensions as the iron boundary. Less accuracy is obtained for
forces on the windings away from the core window [Waters, 1966].

Considering the arrangement of Fig 3.8, the permeability of the iron may be assumed
to be infinity with only a negligible loss in accuracy. This enables the flux to enter
the walls at right angles and the field inside the duct is not changed if the iron
is replaced by an infinite series of images in all four directions as shown dotted in
Figure 3.8. The problem is reduced to calculating the flux density at any point due
to the bars and the array of images .

If A is the vector potential at any point inside the slot then the differential equations
governing the field are,

2A 2A
+ =0 (3.11)
x2 y 2

24
in the duct of the winding,
2A 2A
+ = 4jd (3.12)
x2 y 2

Where jd is the current density in the cross-sections of the windings.

Solutions of these equations may be expressed in either single or double Fourier

Series, the constants being determined by the positions of the iron boundaries,
number, position and dimensions of the windings. The mathematical derivations
are long and tedious and have been carried out by several researchers in this field,
who have continued the work of Roth and are listed in Waters [1966].

The scope of this work is not to produce the derivation, and only the final solution
is given. The solution using a single series is mathematically equivalent to the
double series method of Roth and is an attempt to simplify the final result, but now
with powerful computers the need for mathematical simplification is not essential.
However, DeKuijper [Waters, 1966] considered the single Fourier series to have some
advantages.

Pichon and Hochart [1958] have successfully adapted the original solution of Roth
for the computer codes and have indicated the programming technique. The bottom
left-hand corner of the window is taken as the origin and the forces are calculated
from the ampere-turns and the coordinates of the corners of the windings and the
iron boundary, all of which are assumed to have rectangular cross-sections as shown
in Fig 3.8. The complete solutions for the radial force and the axial force are given in
Waters [1966] and Hague [1929]. It should be noted that for the force on a part of a
winding, the part has to be considered as a separate winding with its correct value of
ampere-turns [Waters, 1966]. This method is simple and calculations can be carried
out by hand but are very time consuming hence computer use is recommended
[Waters, 1966].

3.6 Methods using digital computers

The configuration of a three-phase core-type transformer does not lend itself to
developing rigorous mathematical expressions for the leakage flux or for the forces
on the conductors at any point in the windings. In order to deal with a transformer
mathematically, it must be represented in an idealized form having axial symmetry
as shown in Fig 3.9. The core is assumed to be cylindrical and the yokes represented
by infinite planes, all having infinite permeability. An outer iron boundary may be
assumed co-axial with the limb, as shown in Fig 3.9, but the solution allows this

25
boundary to be at any distance up to infinity. The windings are co-axial with the
limb and of rectangular cross-section with uniform current density. If a winding has
parts of different current densities, it is necessary to treat these parts as separate
windings. Mathematically this presents no difficulty. Tapered windings can be dealt
with only by using a stepped representation, each step being treated as a separate
winding [Waters, 1966].

Core and winding axis

Iron boundries Iron boundries

W indow

Window
Windings

Figure 3.9: Idealized core-type transformer with curvature of windings

Experience has shown that the single limb idealization (Fig 3.9) enables highly accur-
ate estimates of flux density and electromagnetic forces. The assumptions of infinite
permeability, uniform current distribution, instead of discrete conductors and infin-
ite planes to represent the yoke, lead to errors which are negligible compared with
the errors due to differences between the design dimensions and manufactured di-
mensions of the transformer. Curvature may be taken into account correctly and
since no other assumptions were made (apart from those mentioned above), it is
possible to deal with unusual arrangements of windings. The approach to the prob-
lem is simple. More details on this method in given in Waters [1966] and Hague
[1929] which describe the method in detail.

3.6.1 Smythes solution

Roths method can be simplified considerably by assuming that the ampere-turns of
each winding are concentrated in a thin cylinder which is assumed to have the same
mean diameter as the actual winding. Based on this idea, Smythe [Waters, 1966]
suggested a solution, which simplifies the problem and makes it possible to calculate

26
the forces using an ordinary calculator.

Vein [Waters, 1966] has used Smythes solution to produce expressions for the force
in a multi-layer winding and proved that if the ampere-turns are concentrated in
thin layers the method should give accurate results.

3.6.2 Rabins Solution

Rabins solution is the simplified solution of Roths 3 which was produced by taking
into account the curvature of the windings . In 1956 Rabin [Waters, 1966] introduced
a simpler solution using a single Fourier series with coefficients which were Bessel
and Struve functions. The expression he used for the reactance calculation when
differentiated, leads to flux density and to the electromagnetic force. Rabins method
has been used in mid 1960s for software tools. The Bessel and Struve functions are
also generated by computer using appropriate expressions. Kuster [Waters, 1966]
has given a method to calculate these functions by computers, but in certain cases,
double-length arithmetic is required to obtain adequate accuracy. A more suitable
method has been provided by Chebyshew 4 which is more useful particularly when
the double-length arithmetic is not provided.

3.6.3 Solution by analogue computer

Goldenhberg [Waters, 1966] has shown how the basic equations for the flux and forces
in a transformer with axial symmetry may be solved by means of a two-dimensional
resistance network. This solution is based on a finite difference approximation of
the basic differential equations and boundary conditions. The number of resistance
elements required are very large. But if the solutions of a large number of problems
of this type was required it might be economical to set up such a network. Once the
setup is made, it would be more flexible in use than a digital computer and could
cope with different arrangements of windings, e.g. triangular shapes, which could
not be easily done with a digital computer.

3.6.4 Image method with discrete conductors

This method uses a technique similar to that mentioned in Section 3.4.1. Here each
turn of the winding is regarded as a straight conductor. A similar method has
also been used by the Cigre Working Group in the Final Report: Cigre Working
3
Roth E: (cited in Waters [1966])
4
cited in Waters [1966]

27
Group 12-04 [1979] to calculate the forces in a core type transformer. The force on a
bundle of conductors, a disk or any section of winding may be calculated by regarding
them as filaments carrying current at their geometrical centers. The cross-section of
the conductor is assumed as circular but the square and rectangular cross-sections
can be accommodated with same simplification by using the formulas which take
into account the cross-sections of different geometries [Hague, 1929]. Alternatively,
multiple circular conductors of suitable diameter can be packed into the shape of the
conductor in use. Figure 3.10 shows the conductors and the core in two-dimensional
geometry and Fig 3.11 shows the arrangement of images to take into account the
iron boundaries.

Core

Core window
Winding conduc tors

Figure 3.10: Discrete image method

This method uses a simple formula to calculate the force on each winding conduct-
or/turn. It is assumed that the turns are conductors of equal length, carrying current
and running in parallel. The force can be determined as follows,

F = Bi l sin (3.13)

Where B is the flux density at a particular turn and F is the force on that turn, and

1 i1
B = 0 (3.14)
2h

where i1 is the current in the conductor and h is the distance between the two
conductors
i1 i2
F = 0 (3.15)
h

28
 i1 i2
F = 2l.107 (3.16)
h

TM
The flow chart in Fig 3.12 shows the procedure used for developing the Matlab
code used to calculate the forces on the inner and outer winding conductors (disks
TM
or turns). The developed code is given in appendix A. Matlab was chosen
because it is a high level mathematical and engineering language with good graphical
r

capabilities. The output of this code can be easily used in Simulink for modeling
of the dynamic behaviour of the transformer which is the main objective of this
work.

Axial im ages

Iron boundries
Radial images
Radial images

Core
window
Axial images

Figure 3.11: System of images in discrete image method for a two-winding trans-
former

The forces calculated with this method do not consider the core window. Also it is
assumed that the whole length of turn Dm (where Dm = mean diameter of turn)
is situated under the yoke. This limitation can be overcome by considering that
only the conductor length under the yoke is used for the yoke side images. With

29
this method, axial and radial forces are calculated at the same time if the position
of the conductor is represented in complex form i.e. (x + jy), where x and y are
the coordinates of a two dimensional system. This is the method that was used
for calculating the time dependent forces on each disk of the winding of the test
transformer in Chapters 5 and 6 and in Appendix A. This force is the input of the
r

dynamic Simulink model used to study the behaviour of the windings under short
circuits.

3.6.5 Finite element method (FEM)

The finite element method is a method for solving problems which are usually defined
as a continuous domain either by differential equations or by equivalent global state-
ments. To make the problem manageable to numerical solution, the infinite degrees
of freedom of the system are discretized or replaced by a finite number of unknown
parameters, as a process of approximation. So the concept of Finite Element is
replacing the continuous system by a number of sub-domains or elements whose
behaviour is modeled adequately by a limited number of degrees of freedom using
processes available in the analysis of discrete systems.

r

In this work the Maxwell Finite Element Analysis (FEA) software from Ansoft
Corporation was used to calculate the force on the conductors of the test transformer.
The software does take into account the non-linearity of the iron.

The force on each disk of the test transformer was calculated with FEM (Maxwell)
and with the discrete image method at 1kA per disk. This was to compare the
accuracy of the much faster discrete image method to the more time-consuming
FEM. The magnetic field pattern (from FEM) is given in the Figure 3.13. The
comparison of the results is given in Figures 4.4 and 4.5 for the inner and outer
windings respectively. The discrete image method shows lower accuracy in calcu-
lation of radial forces and can be improved by the addition of more images in the
radial direction. The accuracy of the discrete image method can be further improved
by assuming images only for the conductors under the yokes which can not be done
in two-dimensional FEM and 3-dimensional FEM is required which is more time
consuming than 2-dimensional FEM.

30
 Input data

Position vectors
Calculate the position of the primary and
secondary: Position matrices (assuming
the core limb center as origin). Use
rectangular coordinates in complex form

Distance vector of inner winding

Calculate the distance of all the
conductors from the 1st conductor of
inner winding

Calculate force ( Fi 1, Fi 2, Fi 3 , ... ) on

all the conductors due to the
magnetic field
s generated by the the
1 t conductor

Distance vector outer winding

Calculate the distance of all the
conductors from the 1st conductor of
outer winding

Calculate force ( Fo 1, Fo 2, Fo 3 , ... )

on all the conductors due to the
magnetic field generated by the the
1st conductor of the outer winding

Calculate distance from inner to outer

winding and outer to inner winding

Calculate force on the inner

winding due to outer winding and
on outer winding due to inner
winding and add to the force on
each conductor

Calculate the forces (Fim) on each conductor due to images of

top yoke, bottom yoke and side limb

No
Add another If Fim > Fo/100, Fi/100 Stop
image

Yes

Add force to Fi and Fo

Figure 3.12: Flow chart of system of the images in discrete image method for a
two-winding transformer
.
31
Figure 3.13: Magnetic field in the test transformer under short circuit conditions.

Radial force: inner winding

10
image method
15
Force (kN)

finite element method

20

25

30
1 2 3 4 5 6
Disk no
Axial force: inner winding
10

5
Force (kN)

10
1 2 3 4 5 6
Disk no

Figure 3.14: Comparison of results of discrete image method and Finite Element
Method for inner winding.

32
 Radial force: outer winding
26

24

Force (kN)
22
image method
20 finite element method

18
1 2 3 4 5 6
Disk no
Axial force: outer winding
10

Force (kN) 0

10
1 2 3 4 5 6
Disk no

Figure 3.15: Comparison of results of discrete image method and Finite Element
Method for outer winding.

3.7 Discussion

Although the Finite Element Method is a very accurate tool to calculate the electro-
magnetic forces, the time taken to draw a transformer geometry is very significant.
Then the force on each element has to be calculated separately and the program has
to be run as many times as the number of the elements. If the transformer winding
arrangement is complex, the computation time can be extremely long.

The image method is also accurate and a fraction of time is consumed for both setup
and computation when compared to the Finite Element Method. The results of
force on each component are calculated simultaneously in the image method. Finite
Element Method is more comprehensive because it takes non-linearity of iron into
account. The accuracy of discrete image method can be improved by the addition of
more images. However, for practical calculations, 10% discrepancy is not significant.

3.8 Conclusion

The image method is accurate and less time consuming than the Finite Element
Method. It calculates the radial and axial forces simultaneously. Also the results can
r

be programmed as a force vs time waveform which are required for the Simulink
model used to study the dynamic behaviour.

33
Chapter 4

Calculation of radial electromagnetic

forces in concentric windings

4.1 Introduction
Radial forces in concentric windings of a two-winding transformer produce a hoop
stress that tends to extend the radius of the outer winding and at the same time they
produce a compressive stress in the inner winding producing buckling as shown in
Chapter 2. In this chapter the methods of calculating the electromagnetic forces in
the radial direction in a transformer with concentric windings are considered. Only
the core type transformers with cylindrical windings are discussed here. However,
these techniques can be applied to the other types of windings [Say, 1958; Franklin
and Franklin, 1983; Waters, 1966].

4.2 Prediction of radial electromagnetic forces

The radial components of the forces in a transformer with concentric windings have
never been considered significant due to the fact that radial strength of the winding
is high. Although, radial buckling has been experienced in transformers, the general
conception is that it can be avoided by a perfectly round winding cross-section
and by adequate radial supports. The bulk of the previous research deals with the
force calculation in the axial direction, since the vast majority of failures are due to
axial forces. Also axial movement is more damaging to the winding and insulation
structures than radial movements.

The methods available for the calculation of the radial forces range from the basic
empirical methods to more sophisticated methods. The empirical methods are based

34
on simplified assumptions. Corrections are added on the basis of experience or
experimental results to reduce the inaccuracies. As a result, the radial forces can be
easily and relatively accurately calculated by elementary methods especially when
the maximum radial force is required. However, the more sophisticated methods
like the two-dimensional image method and two or three dimensional Finite Element
Methods are more accurate but require the use of a computer.

4.3 Elementary methods to calculate radial electromag-

netic forces
The winding of a transformer can be represented in a two-dimensional geometry.
Figure 4.1 shows a cross section of one side of the limb of a two-winding transformer.
The outer winding lies in an axial field which causes a radial force acting outwards
tending to stretch the conductor (producing a hoop stress), and the inner winding
experiences a similar force acting inwards tending to crush or collapse it.

Distribution of
axial flux and
h radial force
Radial Force
Outer
winding
Ba
d1
Inner
winding
Radial Force

Core

Figure 4.1: Two-dimension representation of concentric transformer windings for the

calculation of radial electromagnetic forces showing axial flux and resulting radial
force [Waters, 1966]

These radial forces are slightly less at the ends of the windings due to the curving of
the magnetic flux, but the force per unit length of winding will be almost uniform
over the greater part of the winding length and it can be accurately calculated at the
middle of the winding. Since it is the maximum force which is important [Say, 1958;
Franklin and Franklin, 1983; Waters, 1966], it is convenient to ignore the curvature
of the field near the ends of the windings and assume the leakage field is uniform
along the whole length and that the radial forces will also be the same at all points
along the winding. The stresses calculated in this way will correspond to those in
the middle 90% of the winding. The reduction that occurs near the ends to about
half is of little practical importance [Waters, 1966].

35
4.4 Forces in the outer winding
Considering the simple two-dimensional picture of Fig 4.1, the axial flux density in
4107 (N I)
the leakage duct is h T if (N I) is the instantaneous value of the ampere-
turn in each winding and h is the length of the windings.

Inner winding Outer winding

Core

Radial forces
for inner and
outer windings

Figure 4.2: Radial electromagnetic forces in concentric transformer windings [Wa-

ters, 1966]

The method is based on the two-dimensional picture (Fig 4.1) of the magnetic field
used for the reactance calculation [Waters, 1966; Franklin and Franklin, 1983]. The
flux density decreases linearly from maximum value (in the duct) at the two surfaces
of the duct to zero at the other surfaces of the windings as shown in the diagram at
the right hand side of Fig 4.1. The radial force is produced by the average axial flux
density in the winding which is equal to half of the duct flux density. This radial
force acts radially outward as shown in Fig 4.2. The mean hoop stress in the outer
winding can be calculated considering the winding as a thin cylinder shown in Fig
4.3. The transverse force in two opposite halves is equivalent to the pressure on the
diameter [Waters, 1966; Franklin and Franklin, 1983], while the total force is the
equivalent to the pressure upon the circumference Dw where Dw is the diameter
of the outer winding. This force acts on both ends of the diameter AB in Fig 4.3,
i.e. on a cross sectional area of conductor equal to twice that of the whole winding.

The mean hoop stress mean in the conductor of the outer winding at the peak of
the first half cycle of short-circuit current, assuming an asymmetry factor of 1.8
[Franklin and Franklin, 1983; Waters, 1966] is,

36
 A

Pw Pw

Figure 4.3: Mean hoop stress calculating method [Waters, 1966]

0.031Wcu
mean = kN/mm2 (peak) (4.1)
he2z

where Wcu = I 2 Rdc loss in the winding in kW at rated full load at 75 C

h = axial height of the winding in mm
ez = per unit impedance voltage

The inner winding tends to become crushed against the core and it is common prac-
tice to support the winding from the core and to treat the winding as a continuous
beam with equidistant supports, ignoring the increase in strength due to curvature.
The mean radial load per mm length of the conductor of the disk coil is (after
[Franklin and Franklin, 1983]):
0.031mean Ac
W = kN/mm length (4.2)
Dw
or alternatively
510U X1
W = kN/mm length (4.3)
ez f d1 Dm N

37
where
Wcu = I 2 Rdc loss in the winding in kW at rated full load at 75 C
h = axial height of the winding in mm
Ac = cross-section of the conductor on which the force is to be determined, mm2
Dw = mean diameter of the outer winding, mm
U = rated kVA per limb
f = frequency, Hz
mean = mean hoop stress at the peak of first half cycle, kN/mm2 , from equation 4.1
d1 = equivalent duct width, mm
Dm = mean diameter of the transformer windings
(i.e. of HV and LV windings together), mm
N = number of turns in the outer winding
ez = per unit impedance voltage

Equation 4.3 gives a total force on 1mm length of the conductor occupying the full
radial thickness of the winding. In a multilayer winding, with k layers, the value for
the layer next to the duct would be (2k 1)/k times this value, the second layer
(2k 3)/k, and so on.

4.4.1 Hoop stress in disk windings

In the event of Hoop Stress in disc windings in a tightly wound disc coil, the inner
turns cannot elongate without stressing those on the outside. The transfer of stress
is considerable and instead of the stress varying from practically zero on the outside
to twice the mean value on the inside, it becomes almost uniform with a maximum
not much greater than the mean stress [Waters, 1966]. Alternatively, if the turns
of a disc coil are free to slide, then the tension must be the same at all points and
the hoop stress must be the same in all turns. Thus in an ordinary disc coil it is a
reasonable assumption to use the mean hoop stress when considering the strength
of the coil as a whole.

4.5 Forces on the inner winding

The inner winding is subjected to radial forces acting inwards as shown in Fig 4.2 and
these may be calculated by using the mean diameter. However, the inner winding
does not have a simple compressive stress equivalent to the hoop stress in the outer
winding. The modes of the failure of the windings are either by collapsing, or if it

38
is supported from the core, by bending between the supports.

If the inner winding is of the disk type then each disk is subjected to a radial force
per mm of conductor,
2 mean nc Ac
pr = kN/mm length (4.4)
Dw
where

mean = mean stress calculated in Eq 4.1

pr = radial force per mm of conductor
Ac = cross sectional area of each conductor mm2
nc = number of conductors in each disk
Dw = mean diameter of the winding (inner and outer) mm

and the total radial force

Prw = 2 mean nc Ac kN (4.5)

It is however difficult and tedious to predict the forces on each disk or turn separately
to use in a mechanical behaviour study. On the other hand, sophisticated methods
provide the solution of forces on individual turns/disks. These methods are rigorous
and deal with axial and radial force calculation simultaneously and are discussed in
Chapter 3.

4.6 Discussion

The methods discussed here are capable of calculating the cumulative radial force
on each winding but are ineffective when the force on each disk is required for the
dynamic behaviour study.

The Finite Element Method and Discrete Image Method, described in Chapter 3
are the accurate methods. The Image method with discrete conductors is the
method of choice for this work because of its ability to calculate both axial and
radial forces simultaneously and accurately. The code developed with this method is
given in Appendix A. The comparison of the results of Finite Element Method and
Image Method with Discrete Conductors for radial forces of the test transformer
are presented in Figures 4.4 and 4.5.

39
 Radial force: inner winding
10
image method
15

Force (kN)
finite element method

20

25

30
1 2 3 4 5 6
Disk no
Axial force: inner winding
10

Force (kN) 0

10
1 2 3 4 5 6
Disk no

Figure 4.4: Comparison of results of discrete image method and finite element
method for inner winding

Radial force: outer winding

26

24
Force (kN)

22
image method
20 finite element method

18
1 2 3 4 5 6
Disk no
Axial force: outer winding
10

5
Force (kN)

10
1 2 3 4 5 6
Disk no

Figure 4.5: Comparison of results of discrete image method and finite element
method for outer winding

4.7 Conclusion

The accuracy of the Image Method with Discrete Conductors depends upon the
number of radial images. The program written was able to calculate the forces on
each disk which was required for the dynamic model of Chapter 8.

The discrepancy between discrete image method and FEM is not regarded as signi-
ficant (Section 3.7, Chapter 3).

40
Chapter 5

Dynamic axial behaviour of

transformer windings

5.1 Introduction

The mathematical models to study the dynamic axial behaviour of large power trans-
formers have been available since the late 1950s and early 1960s. The models were
simplified by suitable assumptions to reduce the complexity and time of computa-
tion. Most of the models assumed pressboard to behave as a linear spring [Tournier
et al., 1964; J P Martin, 1980]. Patel [1972] assumed pressboard as a non-linear,
unidirectional spring. The model of the pressboard was further improved by Swihart
and McCormick [1980]. They developed a non-linear model including damping.

The axial model presented in this chapter, is similar to the models of previous
researchers and has used a few assumptions to simplify the problem. The pressboard
under dynamic load is treated as non-linear. The derivation of the mathematical
equations of the model was carried out by assuming the spring constant as linear
and the nonlinearity was then introduced by representing the spring characteristics
as a lookup table. The implementation of the nonlinear model is given in Chapter
8. This model is an integral component in the construction of the combined model
to study the behaviour of a transformer in reality.

41
5.2 Axial model of power transformer windings

The physical structure of the winding of a transformer consists of copper windings

insulated with craft paper (wrapped) and pressboard (clacks) used in the radial and
the axial directions respectively, as insulation and to accommodate the oil flow for
cooling of the copper conductors. The winding is made on a pressboard cylinder
(former) and the disks are supported radially by ribs as shown in Fig 5.1.

Figure 5.1: Physical construction of transformer winding. On the top left, a photo
shows the full winding and the right is the magnified section of the winding showing
axial spacers. The bottom section of photo shows the duct sticks (ribs).

The first comprehensive approach to model the axial behaviour of a transformer

winding was suggested by Tournier et al. [1962a]. The model was based on the
following assumptions.

the mass of the winding was divided into a number of lumped masses (lumped
masses were assumed rigid)
the axial insulation between the turns/disks was also lumped

The mathematical model was a set of differential equations and had to be solved
numerically. To reduce the time of computation, the number of the lumped masses
was reduced to 11 as it produced the same values for the first few natural frequencies.

42
The spring constant and damping of the pressboard insulation were determined
experimentally from the oscillograms of tests conducted on the transformer which
showed very little damping in the insulation components.

In 1963 Watts [1963] made an improvement to Touriners axial model by introdu-

cing the non-linear behaviour of pressboard. Wattss non-linear (VISCO-ELASTIC)
model was further used by Patel [1973b, 1972], in his axial model of the transformer
winding which was developed for a three phase transformer. This model included
the effects of core clamps, tie rods and the response of the tank under short cir-
cuit conditions. The model was capable of including different pre-stress levels. He
simplified the viscoelastic model to a two component model and further concluded
that under short-circuit conditions there was no internal damping in the pressboard
insulation.

The researchers [Madin and Whitaker, 1963a; Hiraishi, 1971; Ayres et al., 1975]
produced similar axial models and tested them on full scale transformers.

Swihart and Wright [1976] did an extensive study in determining the physical prop-
erties of pressboard under different pre-stress levels and suggested a very complex
model for the insulation. This model took into account the spring and damping
characteristics of the pressboard. The measurements made by Swihart et al showed
significant damping in the pressboard even after stabilization. The work also pointed
out that the behaviour of pressboard under heavy dynamic loading was considerably
different from that under lighter dynamic loads. Also there was a considerable dif-
ference between the dynamic and static characteristics of pressboard which was used
in the earlier dynamic studies. Later, Hori et al [Hori and Okuyama, 1990] produced
a two dimensional model based on the difference in applied electromagnetic force on
a winding coil on the inside and outside of the core window.

5.3 Dynamic axial model

The axial model suggested in this work is similar to the previous models with a
change to two separate windings as shown in Fig 5.2. The previous authors have
lumped the pressboard vertical spacers, but in this study, each winding disk is
assumed as a lumped mass and instead of combining the pressboard vertical spacers
together in lumped springs, each vertical spacer (the insulation between each disk) is
represented as a spring and a dashpot. The winding clamps are assumed rigid which
is realistic because the copper winding and pressboard insulation are the weakest

43
 k1 c1 K1 C1

m1 M1
x1 X1
f1 (t)+m1 g k2 c2 F1 (t)+M1 g K2 C2

m2 M2
x2 X2
f2 (t)+m2 g F2 (t)+M2 g
k3 c3 K3 C3

m3 M3

x3 X3
f3 (t)+m3 g c4 F3 (t)+M3 g
k4 K4 C4

ki ci Ki Ci

mi Mi
xi Xi
fi(t)+mig Fi(t)+Mig
k (i+1) ci+1 K(i+1) C(i+1)

kn cn Kn Cn

mn Mn
xn Xn
fn (t)+mn g Fn (t)+Mn g
k(n+1) c(n+1) C(n+1)
K(n+1)

Figure 5.2: Lumped mass model of a two-winding transformer

parts and have significantly lower stiffness than the clamp and core structure. The
top and bottom clamps are relatively heavier and stronger structures, joined with
tie rods allowing little movement of the top and bottom yoke and core clamps.

The equation of motion of the spring-mass system of Fig 5.2 can be written by apply-
ing Newtons 2nd law of motion. The direction of force and resulting displacement
are shown in Fig 5.2.

The equations of motion for lumped masses (winding disk) of the inner winding can
be written as:

44
 m1 x
1 + (c1 + c2 )x 1 c2 x 2 + (k1 + k2 )x1 k2 x2 = f1 (t) + m1 g
m2 x
2 c2 x 1 + (c2 + c3 )x 2 c3 x 3 k2 x1 + (k2 + k3 )x2 k3 x3 = f2 (t) + m2 g
.. ..
. .
mi xi ci x i1 + (ci + ci+1 )x i
ci+1 x i+1 ki xi1 + (ki + ki+1 )xi ki+1 xi+1 = fi (t) + mi g
.. ..
. .
mn x
n cn1 x n1 + (cn1 + cn )x n kn1 xn1 + (kn1 + kn )xn = fn (t) + mn g

Where:

xi = local displacement of mass

ci = is the damping coefficient
ki = spring constant

The above equations can be written in matrix form

xin + [c]x in + [k]xin = f (t) + mg

[m] (5.1)

where xin = local displacement of disk of inner winding

m1 0 0 0 0 x1

0 m2 0 0 0 x2


0 0 m3 0 0 x3
.. ..
..
[m] = . . and [xin ] = .

0 0 mi 0 xi

.. .. ..
.
. .
0 0 0 0 mn xn

45
and

(c1 + c2 ) c2 0 0 0 0

c2 (c2 + c3 ) c3 0 0 0


0 c3 (c3 + c4 ) c4 0 0
..

[c] = .

0 ci1 (ci1 + ci ) ci 0

..

.
0 0 0 cn1 cn

also [k] can be represented as

(k1 + k2 ) k2 0 0 0 0

k2 (k2 + k3 ) k3 0 0 0


0 k3 (k3 + k4 ) k4 0 0
..

[k] = .

0 ki1 (ki1 + ki ) ki 0

..

.
0 0 0 kn kn+1

Similarly, the equation of motion for the outer winding can be written as:

xot + [C]x ot + [K] xot = F(t) + mg

[M] (5.2)

Here xot is the displacement of outer winding disks

Where M, C and K are similar to the matrices m, c and k respectively.

By combining Equations 5.1 and 5.2, the equation of axial motion becomes:

+ [Cax ]X
[Max ]X + Kax X = Fax (t) + Mg (5.3)

The combined displacement X mass Max , damping Cax and stiffness Kax
matrices can be obtained by combining the matrices of inner and outer windings.

" #
xin
[X] = (5.4)
xot

" #
m 0
[Max ] = (5.5)
0 M
Similarly the stiffness and damping matrices are represented as:

46
 Axial natural frequencies
6000

5000

Frequency (Hz)
4000

3000

2000

1000

0
1 2 3 4 5 6 7 8 9 10 11 12
Number

Figure 5.3: The natural frequencies of test transformer in axial direction

" #
k 0
[Kax ] = (5.6)
0 K

" #
c 0
[Cax ] = (5.7)
0 C
The applied electromagnetic force including the static gravitational force in matrix
form is:

" #
f (t) + mg
[Fax ] = (5.8)
F (t) + M g
f(t) and F(t) are applied electromagnetic forces on the inner and outer windings, m
and M is the mass of each lumped mass of inner and outer windings respectively
and g is the gravitational acceleration.

5.4 Implementation of model

TM
Matlab code was written to construct X Kax , and Max matrices and the natural
TM
frequencies were calculated as shown in Figure 5.3. The Matlab code determines
r

the inputs to the axial dynamic model implemented in Simulink . The model
implementation is given in Appendix B. The nonlinearity of the pressboard was
included by using measured dynamic stress-strain characteristics which are discussed
in Chapter 7.

47
5.5 Conclusion
The model is intended to calculate inputs which will be used in the study of axial
and combined dynamic behaviour. The code based on the equations derived in this
chapter was used to calculate the natural frequency of the test transformer in the
axial direction.

The fundamental natural frequency of the test transformer is around 600 Hz which
shows the axial stiffness of the transformer is very high resulting in a natural fre-
quency far removed from the 100 Hz electromagnetic force frequency.

48
Chapter 6

Dynamic radial behaviour of

transformer windings

6.1 Introduction

In early years, research into the dynamics of windings was limited to the axial
behaviour of the windings. In the radial direction, the windings were assumed to be
capable of withstanding the radial stresses due to the higher stiffness. In the recent
past, it has been realized that the inner winding buckling goes unnoticed due to a lack
of means of easy visual inspection. Some electrical techniques have been available
and applied but they fail to detect the deformation, especially when the relative
deformation of the windings is small. Many indirect methods (Sweep Frequency
Response etc) have been developed to assess the deformation by measurement but
none of them are either consistent or accurate.

This chapter presents the equations governing the dynamic radial behaviour of a
two concentric windings transformer. The interaction of radial and axial behaviour
under dynamic conditions will be discussed in Chapter 8.

6.2 Radial behaviour of windings

In 1971 Hiraishi [1971] conducted experimental work to investigate the radial and
axial behaviour of transformer windings. In this work, the vibration characteristics
of the winding and natural frequencies were determined experimentally. The effects
of the natural frequency on the buckling strength were also explored. He derived
equations for the calculation of the buckling strength and plastic deformation under

49
static conditions. The radial deformation was further studied by Ayres et al. [1975].
Saravolac et al. using a static approach for radial strength based on Timoshenkos
formula for the critical load of a hinged arch [Saravolac et al., 2000]. They also
conducted experiments to understand the withstand capabilities of different types
of windings under uniform forces in the radial direction.

6.3 Radial model of a two-winding transformer

Consider a two-winding transformer with both inner and outer windings having the
same voltage and current. The windings are disk type as shown in Fig 6.1. The
radial behaviour of the windings can be investigated by considering both windings
as concentric elastic rings coupled with springs and dashpots (insulation ribs) as
shown in Fig 6.4. Although copper is not linearly elastic, the assumption that
copper rings are elastic is realistic under small displacements. The assumption is
more applicable for the copper used in modern windings which is cold worked and
can have an elasticity of 1.4 1010 P and even higher. The vibration of these rings
under a periodic force excitation can be of the following types [Timoshenko et al.,
1974].

extensional vibration with a periodic change in the radius of the ring (Figure
6.3)
flexural vibration in the plane of the ring
flexural vibration involving the displacements at right angles, out of the plane
of the ring
twist

For the inner winding, consider the flexural vibration in the plane of the ring
[Timoshenko et al., 1974] of radius r. Figure 6.2 shows the angle representing
the angular coordinates of the radial displacement, u is the radial displacement
(positive outward) and v is the tangential displacement (positive in anti clockwise
direction). Similarly for the outer winding, R is the radius, U and V are the radial
and tangential displacements respectively and will be used further in this chapter.

Due to the displacements u and v, the unit elongation of the center line of the ring
at any point is represented as [Timoshenko et al., 1974].
u v
e= + (6.1)
r r
For the general case of flexural vibrations in the plane of the ring the radial displace-
ment u can be expanded in the form of a trigonometric series [Timoshenko et al.,

50
 Figure 6.1: Cross-sectional view of a transformer winding

1974] as given in Equation 6.2,

u = ao + a1 cos + b1 sin + a2 cos 2 + b2 sin 2 + a3 cos 3 + b3 sin 3 + (6.2)

where a1 , a2 , . . . , b1 , b2 , . . . are the generalized displacements and are functions of

time. However ao represents pure uniform radial displacement represented by a
constant term. The trigonometric terms cos and sin of Equation 6.2 show the
influence of radial position. For pure flexural vibration without any radial extension,
the elongation e in Equation 6.1 is 0.

v
u= (6.3)

and the circumferential displacement of the ring can be represented as

Z
v= ud (6.4)

by integrating u and multiplying by 1, results in

1 1 1 1
v = 0a1 sin +b1 cos a2 sin 2+ b2 cos 2 a3 sin 3+ b3 cos 3. . . (6.5)
2 2 3 3
The term ao represents breathing motion of the ring with no circumferential dis-
placement and hence in Equation 6.5 is regarded as 0.

51
 ds = rd
v
u

d
r r

Figure 6.2: Radial and circumferential displacements, u and v the inner winding
ring.

ao

Figure 6.3: Simplest mode of vibration breathing motion uniform radial expansion
and contraction

Equation 6.2 can be resolved in vector form, if

ao

a1


b1


a2
q=

(6.6)
b2

a3


b3

..
.

52
 KO KO
CO
Outer CO
winding
disk Inner
Ki Ci
winding
disk
CO Ci Ki KO

KO Ki Ci CO

Ki Ci

KO
KO
CO
CO

Figure 6.4: Inner and outer winding represented as elastic rings coupled with springs
and dashpots

Where q represents the general coordinates in a vector format and the angular
coordinates in vector form are represented as u

1

cos


sin


cos 2
u = (6.7)

sin 2

cos 3


sin 3

..
.

Similarly v can also be represented in vector form of q and v . Where v is the

angular vector given as,

0

sin


cos

1
2 sin 2

v = 1 (6.8)

2 cos 2
1
sin 3
3
1
3 cos 3

..
.

53
Equation 6.2 can be written in matrix form,

u = qT [u ] = [u ]T q (6.9)

Similarly, the equation of v can also be represented in matrix format as,

v = qT [v ] = [v ]T q (6.10)

6.4 Kinetic energy of inner and outer disks

The kinetic energy of the vibrating inner and outer rings is of two types, the kinetic
energy due to simple extensional vibration (Fig 6.3) and the flexural kinetic energy
which can be represented as [Timoshenko et al., 1974],
A 2
T = u 2r (6.11)
2

Where:
T = kinetic energy of a ring due to flexural vibration
A = cross-sectional area of the ring
= mass density
r = radius of the ring
du
u = dt , velocity in the radial direction

The rings have radial and tangential movements. Hence the total velocity of the ring
includes the radial and tangential velocity u and v.
Therefore, the kinetic energy of
the inner ring due to radial and tangential motion is given in the following equation.
Z 2
i 1 2
T = Ai u + v 2 rd (6.12)
2 0

The total kinetic energy of the inner and outer winding rings due to the flexural
vibration can be represented as,

T = Ti + To
Z Z
Ai 2 2 Ao 2 2
Ti + To = 2
(u + v )rd + (U + V 2 )Rd (6.13)
2 0 2 0
Where
T i = Kinetic energy of the inner ring
T o = Kinetic Energy of the outer ring
= the mass density of winding ring
Ai , Ao = are the cross-sectional areas of inner and outer winding rings respectively
U , V =are the radial and tangential displacements of the outer winding and

54
u, v = are the radial and tangential displacements of the inner ring as shown in
Figure 6.2.

To calculate the kinetic energy T i for the inner winding ring,

u = q T u
and u 2 = q T u u T q

similarly v 2 is
v 2 = q T v v T q (6.14)

Recalling q is a function of time and is independent of .

hence
2
u + v 2 = q T u u T q + q T v v T q (6.15)

and can be represented as

= q T u Tu + v Tv q (6.16)

Integrating both sides of Equation 6.16 w.r.t ,

Z 2 Z 2 Z 2
2 2
T T T
u + v d = q u u d q + q v Tv d q (6.17)
0 0 0

To evaluate the expression given in Equation 6.17, as a first step, the vectors u
and Tu are required to be multiplied. Hence multiplying u and Tu ,

1 cos sin cos 2 sin 3

cos cos2 cos sin cos cos 2
cos sin 2

2
sin sin cos sin sin cos 2 sin sin 2
u u =
T


cos 2 cos 2 cos cos 2 sin cos2 2 cos 2 sin 2


sin 2 sin 2 cos sin 2 sin sin 2 cos 2 2
sin 2

.. .. .. .. ..
. . . . .
(6.18)
Similarly v and Tv result in the following matrix.

55

0 0 0 0 0

0 sin2 sin cos 1
12 sin cos 2
2 sin sin 2

0 cos sin cos2 21 cos sin 2 1
2 cos cos 2
v v =
T


0 1
2 sin 2 sin
1
2 sin 2 cos 1 2
4 sin 2 14 sin 2 cos 2

0 21 cos 2 sin 1
cos 2 cos 41 cos 2 sin 2 1
cos2 2
2 4
.. .. .. .. ..
. . . . .
(6.19)

Integrating Equations 6.18 and 6.19 and using the following formulae

Z 2
cos2 m d =
0
Z 2
sin2 m d =
0
Z 2
cos m sin n d = 0
0
Z 2
cos n sin m d = 0
0
Z 2
cos m cos n d = 0
0
Z 2
sin m sin n d = 0
0
(6.20)

the results of the integrals in expanded form can be represented as,

2 0 0 0 0 0

0 0 0 0 0


Z 2 0 0 0 0 0


u Tu d = 0 0 0 0 0 (6.21)
0
0 0 0 0 0


0 0 0 0 0

.. .. .. .. .. ..
. . . . . .

56
and

0 0 0 0 0 0 0

0 ( 11 )2 0 0 0 0 0


0 0 1 2
(1) 0 0 0 0

Z 2
0 0 0 ( 12 )2 0 0 0
v v d =
T

(6.22)
0 0 0 0 0 ( 12 )2 0 0

0 0 0 0 0 1 2
(3) 0


0 0 0 0 0 0 ( 13 )2

.. .. .. .. .. .. .. ..
. . . . . . . .

Adding Equations 6.21 and 6.22

2 0 0 0 0 0

0 2 0 0 0 0


Z 2 0 0 2 0 0 0

5
0
(u 2 + v 2 ) d = q 0 0 0 4 0 q T (6.23)
0
0 0 0 0 5
0
4
10
0 0 0 0 0 9


.. .. .. .. .. .. ..
. . . . . . .

By replacing the diagonal matrix of Equation 6.23 with H.

Z 2
(u 2 + v 2 ) d = q H q T (6.24)
0

Hence, the total kinetic energy for the inner ring can be represented by Equation
6.25
Ai r
T i = qi T H qi (6.25)
2

Similarly the total kinetic energy To of outer ring is given by Equation 6.26

o Ao R
T = q To H qo (6.26)
2

In Equations 6.25 and 6.26, the Ai and Ao are the areas of cross-section, and r and
R are radii of inner and outer winding rings respectively. In Equations 6.25 and
6.26, the qi and qo are the generalized coordinates (function of time) of inner and
outer rings.

57
The total kinetic energy T of the system is obtained by adding the kinetic energy
of the inner ring and the outer ring and is given in Equation 6.27.

1 T 1
T = Ti + To = qi (Ai rH) qi + qo T (Ao RH) qo (6.27)
2 | {z } 2 | {z }
Min Mot

To simplify Equation 6.27 the expressions Ai rH and Ao RH are replaced by Min

and Mot . Where Min and Mot are the mass matrices of inner and outer rings
respectively.

1 T 1
T = q i Min qi + qo T Mot qo (6.28)
2 2

Hence, the total kinetic energy for both inner and outer rings in matrix form can be
represented as,

Min O
1 T

T = Q

Q
(6.29)
2
O Mot

ao

a1


b1


a2


b2
qi ..
.
Q= =
(6.30)

A0

qo

A1

B1


A2


B2

..
.

Where Q is a vector consisting of the generalized coordinates of the inner and outer
rings.

1

Equation 6.29 has the basic form of the kinetic energy equation T = 2 mv 2 and
the diagonal matrix in the equation is the mass matrix and can be replaced with

58
the symbol Mr . Where Mr is the mass matrix of the inner and outer springs.

Min O

Mr =


(6.31)
O Mot

The equation of the kinetic energy can be written in general format as

1 T
T = Q Mr Q (6.32)
2
Mr is mass matrix and will used in the study of the radial dynamic behaviour of
the windings.

6.5 Strain (potential) energy

The potential energy (PE) is the energy stored in the system due to dynamic con-
ditions. This is the sum of potential energies stored by different components due to
their behaviour. The energy stored in the copper rings is due to their elastic be-
haviour and potential energy stored in the insulation ribs depends upon the stress-
strain characteristics of pressboard. The total potential energy can be represented
as
Ep = Epi + Epo + Epins
i o
+ Epins (6.33)

Where
Ep = total potential energy
Epi = potential energy of inner ring
Epo = potential energy of outer ring
i
Epins = potential energy of inner insulation ribs (between core and inner winding)
o
Epins = potential energy of outer insulation ribs (between inner and outer winding)

6.5.1 Potential energy of copper rings

The PE due to elastic behaviour of copper rings can be calculated as given by
Timoshenko et al., [1974]. Timoshenkos formula is given in Equation 6.34.

Z 2 2
Ii 2u
Epi = 4 + u rd (6.34)
2r 0 2
Where is the elasticity of the copper, Ii is the moment of inertia of the inner ring,
r is the radius of the inner winding and u is the displacement of the ring in the
radial direction under flexural vibration as defined in Figure 6.2.

59
From Equation 6.9, u is given as

u = qTi u (6.35)

Differentiating both sides with respect to

u
= qTi 0u (6.36)

Where 0 is the derivative of with respect to .

Differentiating both sides of Equation 6.36 with respect to again

2u
= qTi 00u (6.37)
2
Differentiating the matrix u twice w.r.t gives the following expressions.

0 0

sin cos


cos sin


2 sin 2 4 cos 2
0u =
2 cos 2
and 00u =


4 sin 2


3 sin 3 9 cos


3 cos 3 9 sin

.. ..
. .

Adding u 00 and u
1

0


0


00 3 cos 2
u + u =

(6.38)
3 sin 2


8 cos


8 sin

..
.

60
Therefore
1

0


0

2
u
T
3 cos 2
+ u = qT
00
+ = q (6.39)
2 i u u i 3 sin 2


8 cos


8 sin

..
.

Hence Equation 6.39, can be written as

2u
+ u = qTi 00u + qTi u
2
= qTi (00u + u )
| {z }
D
= qTi D (6.40)

In the above equations for simplification the vector 00u + u is replaced with D
which is related to the elastic behaviour of copper rings.

Squaring both sides of Equation 6.40

2
2u
+u =
2
= qTi DDT qi

1 0 0 3 cos 2 3 sin 2

0 0 0 0 0


0 0 0 0 0
= qTi
3 cos 2
qi
0 0 9 cos2 2 9 cos 2 sin 2


3 sin 2 0 0 9 sin 2 cos 2 2
9 sin 2

.. .. .. .. .. ..
. . . . . .

(6.41)

Integrating both sides and by using the formulae given in 6.20

61

EPec 0 0 0 0 0 0

0 0 0 0 0 0
0


0 0 0 0 0 0 0

Z 2 2 2
u
T
0 0 0 9 0 0 0
+ u d = qi qi
0 2
0 0 0 0 9 0 0

0 0 0 0 0 64 0


0 0 0 0 0 0 64

.. .. .. .. .. .. .. ..
. . . . . . . .
(6.42)

The first element of the diagonal matrix of equation 6.42 (EPec ) represents the
extension/compression mode of vibration when the central line of the ring forms
a circle of uniformly varying radius, and all cross sections move radially without
rotation (Fig 6.1). If u is the radial displacement (positive outward) of any point
on the ring of radius r. Then the unit elongation of the ring in the circumferential
direction (extensional strain) is equal to ( ur ). The potential energy (which in this
case is the energy of simple tension) is given by the Equation 6.43, where A is
the cross-sectional area of the ring [Timoshenko et al., 1974]. Hence EP ec is to be
replaced by the value given by the Equation 6.43.
Au2
EPec = 2r (6.43)
2r2
By replacing the diagonal matrix of Equation 6.42 with Kel , the potential energy
stored in the inner ring due to its elastic behaviour can be represented as

i T Ii
EP = qi Kel qi (6.44)
2r3
Similarly the elastic potential energy of the outer ring

o T Io
Ep = qo Kel qo (6.45)
2R3
This potential energy is due to the elastic behaviour of the copper rings for the inner
and outer windings and will be used to calculate the total potential energy of the
two ring system coupled with pressboard insulation ribs (radial spacers).

6.5.2 Potential energy (PE) stored in pressboard ribs (radial spacers)

The ribs or radial spacers are the components connecting the inner ring to the core
and to the outer copper ring. These ribs are normally made of pressboard material
and support the windings in the radial direction on the core.

62
The potential energy stored in insulation ribs can be calculated by assuming a spring
connecting the inner side of elastic copper ring to the core and to the outer ring.
The PE of the springs is represented by the following equation.
n
1X
Epi ins = Kj u2j (j , t) (6.46)
2
ij=1

where i = 1, 2, 3, , n depending on number of ribs

Assuming four ribs for the inner winding and two approximating functions, the radial
displacements of the inner springs are given by

Inner winding

outer winding

R
Kjo
Kji
r

f j
?j
core
Ribs between inner
Core to winding rib and outer winding

Figure 6.5: Angular position of core and ribs. The ribs are represented as coupling
springs between the core and the inner winding and between the inner and outer
winding

1 cos 1 sin 1 cos 21 sin 21 a0


1 cos 2 sin 2 cos 22 sin 22 a1

u = b
1 cos 3 sin 3 cos 23 sin 23 1
a
2
1 cos 4 sin 4 cos 24 sin 24 b2

(6.47)

The above matrix Equation can be written as:

u=aq (6.48)

Where u is the radial displacement, a is the matrix containing angular positions

of the radial spacers and 1 , 2 , are the angular positions of the ribs along the

63
circumference of the inner winding between core and winding as shown in Fig 6.5.
q is the vector of generalized coordinates of the radial movement.

Hence the PE stored in the ribs between the inner ring and the core is
1
Epi is = qT aT Ki a q (6.49)
2
Ki is the non-linear stiffness of the pressboard material for the radial spacers between
core and inner ring. For the radial spacers between the inner and outer ring, the
potential energy depends on the relative movement of the rings and is given by

n n
1X 1X
Epois = Kj (U (j , t) u(j , t))2 = K4T 4
2 2
i=1 i=1

Where u & U are the radial displacements of the inner and outer rings respectively
and is the difference between U & u and K is the stress-strain characteristics of
the pressboard radial spacers or ribs.

Assuming 6 ribs between the inner and outer rings and three approximating func-
tions for the outer winding, the radial displacement of the inner side of outer winding
radial spacer (spring) Ui can be represented in the matrix form as

1 cos 1 sin 1 cos 21 sin 21 cos 31 sin 31 a0

a1
1 cos 2 sin 2 cos 22 sin 22 cos 32 sin 32

b1
1 cos 3 sin 3 cos 23 sin 23 cos 33 sin 33
Ui =
a2
1 cos 4 sin 4 cos 24 sin 24 cos 34 sin 34
b2

1 cos 5 sin 5 cos 25 sin 25 cos 35 sin 35

a3

1 cos 6 sin 6 cos 26 sin 26 cos 36 sin 36 b3

where 1 , 2 , 3 , are the angular positions of the ribs on outer side of the inner.

The above equation can be represented as

Ui = Pi q = [Pi O]Q (6.50)

Where Q is the generalized coordinates (function of time) of radial displacement for

both inner and outer rings collectively. Pi is the matrix representing the positions
of the radial spacers (between the inner and outer winding) on the inner ring. The
circumferential positions of the radial spacers on the outer ring are same as Pi . So
the outer displacements of the outer radial spacers can be represented as

64

1 cos 1 sin 1 cos 21 sin 21 cos 31 sin 31 A0

A1
1 cos 2 sin 2 cos 22 sin 22 cos 32 sin 32

B1
1 cos 3 sin 3 cos 23 sin 23 cos 33 sin 33
Uo =
A2
1 cos 4 sin 4 cos 24 sin 24 cos 34 sin 34
B2

1 cos 5 sin 5 cos 25 sin 25 cos 35 sin 35

A3

1 cos 6 sin 6 cos 26 sin 26 cos 36 sin 36 B3

65
 Uo = Pi q = [O Po ]Q (6.51)

4 = (Uo Ui ) = [O Po ]Q [Pi O]Q = GQ (6.52)

Where G is an arbitrary variable and

G = [O Po ] [Pi O] (6.53)

in expanded form

1 cos 1 sin 1 cos 21 sin 21 cos 31 sin 31 1 cos 1 sin 1 cos 21 sin 21 cos 31 sin 31



1 cos 2 sin 2 cos 22 sin 22 cos 32 sin 32 1 cos 2 sin 2 cos 22 sin 22 cos 32 sin 32

66




1 cos 3 sin 3 cos 23 sin 23 cos 33 sin 33 1 cos 3 sin 3 cos 23 sin 23 cos 33 sin 33

G= (6.54)


1 cos 4 sin 4 cos 24 sin 24 cos 34 sin 34 1 cos 4 sin 4 cos 24 sin 24 cos 34 sin 34




1 cos 5 sin 5 cos 25 sin 25 cos 35 sin 35 1 cos 5 sin 5 cos 25 sin 25 cos 35 sin 35



1 cos 6 sin 6 cos 26 sin 26 cos 36 sin 36 1 cos 6 sin 6 cos 26 sin 26 cos 36 sin 36
Therefore the expression for 4T 4 can be written as

4T 4 = QT GT G Q (6.55)

The value of 4T 4 is a matrix of size of GT G.

The potential energy of radial spacers for the outer winding will be similar to the
PE of spacers for the inner winding and given in Equation 6.42.
1 T T
EPo ins = Q G Ko G Q (6.56)
2

The total potential energy is the sum of elastic potential energies of the copper
winding and potential energy of the insulation components and is shown as

EP = EPi el + EPo el + EPi ins + EPo ins (6.57)

| {z } | {z }
EPel EPins

Hence EP can be represented as

1 T Ii el 1 T Io el 1 1
EP = q 3
Ki q + Q 3
Ko Q + qT aTKi a q + QT GTKo G Q
2 r 2 R 2 2
(6.58)
The stiffness matrix of the radial behaviour can be extracted from the above equation
and is shown in the following expression.

Ii el Io el T T
Kr = K + K + a Ki a + G K o G (6.59)
r3 i R3 o
The matrix Kr represents the stiffness of the two coupled rings in the radial direc-
tion. The stiffness of the system is dependent on the stiffness of the copper rings
and the stress-strain characteristics of the pressboard radial spacers. This stiffness
matrix will be further used for calculation of natural frequencies and in the dynamic
r

Simulink model in Chapter 8.

6.6 Generalized forces in radial direction

The applied electromagnetic forces in the radial direction need to be resolved ac-
cording to the approximating functions used in the system of equations for the
radial behaviour. These are calculated by considering the virtual work done by the
electromagnetic forces.

If u and U are displacements and r and R are radii of the inner and outer winding
disks respectively, the virtual work can be represented as

67
 Z 2 Z 2
i
w = f (, t) u(, t)rd + f o (, t) u(, t)Rd (6.60)
0 0

by substituting the values of u U with the approximating functions, the generalized

forces for the inner winding can be represented as

Z 2
F0i (t) = fri (, t)rd
0
Z 2
F1i (t) = fri (, t) cos rd
0
Z 2
F2i (t) = fri (, t) sin rd
0
..
.

and for the outer winding

Z 2
F0o (t) = fro (, t) Rd
0
Z 2
F1o (t) = fro (, t) cos Rd
0
Z 2
F2o (t) = fro (, t) sin Rd
0
..
.

Consequently

fri

0


0

0

..
.

Fr = (6.61)
fr o


0

0


0

..
.

The equation of motion for the radial behaviour becomes,

+ Cr Q
Mr Q + Kr Q = Fr (6.62)

68
 5 Radial natural frequencies
x 10
12

10

Frequency (Hz)
6

0
0 5 10 15 20
Number of natural frequency

Figure 6.6: The natural frequencies of radial behaviour of test transformer

Where
Mr = Mass matrix for radial behaviour
Cr = Damping coefficient of insulation (pressboard ribs)
Kr = Stiffness matrix including cooper and ribs (pressboard) stiffness
Q = Displacement coordinates as function of time
fri = Radial force on the inner winding ring as a function of time
fro = Radial force on the outer winding ring as a function of time
Fr = Radial force as a function of time

6.7 Modeling
TM
A Matlab code was written which is given in Appendix A. The code can predict
the natural frequencies of the transformer in the radial direction and calculates
the input variables for the dynamic models both in the radial and axial direction
(Appendix B). The calculated natural frequencies of radial behaviour are shown in
Figure 6.6.

6.8 Conclusion
The radial model is capable of calculating natural frequencies of transformer wind-
ings which gives insight of the dynamic behaviour and highlights possible resonances.

The model is intended to provide data for input to the axial, radial and combined
r

Simulink models of the transformer.

69
Chapter 7

Physical properties of pressboard

under varying pre-stress

7.1 Introduction

Transformer board used in modern transformers has improved in density and insu-
lation properties. It is not only used as insulation but the structural integrity of the
large power transformer also depends on its mechanical properties. The board used
should possess superior compressibility and stiffness properties as it is the major
insulation and support structure in the inter-disk and end insulation.

Secondly, a pre-stress is applied on the transformer windings which compresses the

structural pressboard to remove the sponginess which can not cope with sudden
compressive forces that are applied during a short circuit or through fault condition.

The stress-strain properties of pressboard have been measured before by Patel [Pa-
tel, 1972]. However, they were measured under static load. It is believed that the
properties of the pressboard under dynamic loading are different from the static
characteristics [Swihart and McCormick, 1980]. During dynamic loading the mater-
ial is repeatedly subjected to a compressive force. The time constant of the loading
should have an effect on the stress-strain characteristics as the oil squeezed out of the
material under load and will not return before the next loading cycle. The dynamic
properties of the pressboard measured by Swihart and Wright [1976] and Swihart
and McCormick [1980] showed a non-linear behaviour and considerable damping.
However their model of the oil impregnated pressboard is complex and not user
friendly. Also the full information of the model is not available due to the fact that
the research was not published in the public domain.

70
The following procedure of measurement of stress-strain characteristics is similar to
the one published by Swihart and Wright [1976] with the difference of two pressboard
stacks instead of one. The advantage of this method is it is closer to the actual
arrangement in practical transformer and the results obtained can be easily used in
the dynamic model of a full transformer winding.

7.2 Physical properties of transformer board

To model the dynamic behaviour of a large transformer, it is mandatory to have the

dynamic properties of the structural elements. In large power transformers the main
structural element, which also insulates the windings from the core and ground, is
the pressboard. To evaluate the dynamic stress-strain characteristics, two sections
of pressboard were used in the apparatus of Fig 7.1. It is known that the properties
of the transformer board are highly nonlinear. Under static loading, it shows a high
strain in the beginning and as the load increases, strain decreases and the material
shows higher stiffness and somewhat linear stress-strain characteristics. However,
under dynamic conditions, the stiffness not only depends on the loading but also on
the time duration of the loading as it will squeeze the oil in and out of the material,
which influences the stiffness.

7.3 Dynamic stress-strain test apparatus

The apparatus of Fig 7.2 was used to test the dynamic properties of high density
transformer board. This apparatus uses a similar technique to that used by Swihart
and McCormick [1980] to measure the stress-strain characteristics. In Swiharts
[Swihart and McCormick, 1980] apparatus, only one stack of the pressboard material
was used and a perfectly elastic spring of known characteristics (very high stiffness)
was used as a second spring. The stress-strain characteristics of the pressboard were
then obtained by excluding the effects of the elastic spring in the model.

The apparatus shown in Fig 7.2 also works as a one-mass-two-spring system. It

consists of a mass M supported between two stacks of pressboard material (springs)
connected to rigid supports at the top and the bottom. The detailed drawing of the
apparatus is given in Figure 7.3. This setup is close to the reality of a large power
transformers where only the pressboard is used as support material in the winding
disks/turns wound with copper conductors.

71
The apparatus is mounted on a large damping mass and mechanically isolated from
the floor by a 30 mm thick rubber cushion. A large range of pre-stresses can be ap-
plied to the material under test by the pre-stress adjuster. The value of the pre-stress
can be varied from zero to a practical pre-stress value used in large power trans-
formers. The force impulse is applied with a soft faced hammer to avoid triggering
of the higher order frequencies. The dynamic force is measured with the force sensor
and the displacement sensor measures the resulting displacement in the vertical dir-
ection. The apparatus is capable of measuring the stress-strain characteristics of
both linear and non-linear materials.

Point where
Pre-stress adjuster impulse force
is applied
Dipslacement sensor

Force Sensor

Damping Mass

Floor isolating rubber

Figure 7.1: The picture of the dynamic stress-strain tester

7.3.1 Displacement and force sensors

The displacement and force sensors used in the stress-strain apparatus are shown in
Figures 7.4 and 7.5.

The displacement sensor is a beam type sensor made of PVC strip of 2.5mm thick-
ness. The other dimensions of the sensor are given in Figure 7.4. Two strain gauges
(top and bottom) were used to detect the deflection of the PVC strip. The strain
gauges were connected in the opposite arms of a Wheatstone Bridge. The output of

72
 Soft head hammer

Top rigid support

Press-stress adjuster

Top pressboard stack

Displacement
sensor

M Supported mass

Force sensor

Bottom pressboard stack

Bottom rigid support

Damping mass

Floor isolating rubber

Figure 7.2: The model of the apparatus used to measure the physical properties of
pressboard

the bridge was fed to an instrument amplifier which is capable of adjustable ampli-
fication of 10, 100 and 1000 times. The length of the PVC strip was kept as short as
possible to avoid unwanted vibration. The active part of the sensor was housed in
a PVC tube (shown in Figure 7.4) which was filled with silicone jelly to avoid high
frequency vibration of the active part of the sensor. The sensor was calibrated for
dynamic displacements.

The force sensor uses a simple approach of membrane deflection to measure the
force. The deflection of the 3mm thick circular steel plate produces signal in each of
four strain gauges (2 at top and two at bottom). The strain gauges are connected in
opposite arms of the Wheatstone Bridge circuit and the resulting differential signal
is fed to an instrument amplifier. The amplifier gain can be adjusted to 10, 100 or
1000 times. The force sensor was designed to be stiff so that its natural frequency
is much higher than 100Hz, which is the frequency of the applied electromagnetic
force in the transformer.

The reason for using strain gauge type instruments was due to the inherent properties
of strain gauges to cancel magnetic field effects. The element of the strain gauge, is

73
a conductor placed next to each other carrying opposite currents. This construction
of a zigzag conductor with close spacing between its segments ensures that segments
carry current in opposite directions cancelling magnetic field effects (Figure 7.4).

The use of an instrument which is not effected by the magnetic field is necessary as
under short circuit conditions, the leakage magnetic field becomes stronger and can
induce noise in the measured signal.

7.4 Experimental procedure

To measure the stress-strain characteristics of the high density transformer board

(pressboard), 3mm thick, 50 50mm pieces of pressboard were stacked to achieve
the thickness of 30mm. Figure 7.1 shows the apparatus loaded with the pressboard
material.

To evaluate the properties of the material, it was dried under vacuum and impreg-
nated with oil at 70 o C before it was fitted into the stress-strain tester. To eliminate
the effects of initial thickness loss, the material was compressed three times to a
load higher than the practical pre-stress values (10 N/mm2 ) which will be applied
during the testing. A similar prestressing process is used in industry to reduce the
initial hysteresis from the material by applying a static load more than 100% of the
pre-stress value. This was done to avoid the thickness loss (resulting from outward
flow of oil) and to remove the permanent thickness reduction which is experienced
after a drying cycle. Otherwise, the material loses its thickness during testing and
the applied pre-stress reduces and sometimes disappears.

The measurement was performed at room temperature. Only one force sensor was
used to record the applied force and the reaction force. The force applied to the
material and displacement were recorded at different pre-stresses shown in Table 7.1
and the recorded waveforms are given in Figures 7.6 to 7.11.

74
Figure 7.3: Detail drawing of Stress-strain apparatus

75
 65

Strain gauges
( 200 ohm, top and
bottom)
2 .5
12 s=
n es
ic k
Th

a. The detailed diagram of displacement sensor with dimensions in mm

b. photograph of the displacement sensor

Figure 7.4: Displacement sensor construction

76
 Bottom strain
gauge

Top strain gauge Top strain gauge

70 mm
Loading point

Bottom strain
gauge

30mm

4mm

10mm

Bottom mounting cup

Mounting
hole

Mounting
hole

Figure 7.5: Force sensor construction detail

77
 Diameter Force Stress
of contact surface
(mm) (kN) (kPa or N/mm2 )
30 0 0
30 1 1.42
30 3 4.3
30 5 7.1
30 7 10
30 10 14

Table 7.1: Applied pre-stress values.

The impact force waveshapes of Figures 7.6 to 7.11 were applied with a soft faced
hammer. The soft faced hammer was used to avoid high frequency resonances. The
resulting oscillating force and displacements were recorded with force and displace-
ment sensors in the time domain. This measurement gave results of stress-strain
characteristics of pressboard under dynamic conditions. The results of the tests at
different pre-stress levels are given in Figures 7.6 to 7.11.

7.5 Results

The measured results are shown in Figures 7.6 to 7.11. The results of the stress-
strain measurements confirm that the high density pressboard is not only non-linear
but also possess significant hysteresis. The structure of the material and the effect
of the oil impregnation has been discussed by [Patel, 1973b,a] under Visco-elastic
model of the pressboard.

7.5.1 Stress-Strain properties at zero pre-stress

Figure 7.6 shows the measured results of the pressboard material when no pre-
stress was applied. Figure 7.6-b shows that the stiffness of the material was the
highest during the first half cycle of the response and became lesser and lesser in the
subsequent cycles indicated by the longer time period of the half cycles. The negative
half cycles of the force are almost zero indicating separation of the pressboard stack
from the mass (force sensor is a part of the mass). The negative peaks of force
are very low compared to the positive peaks indicating that the expansion of the
material is almost non existent. Also the negative peaks of the force half cycles

78
are almost zero and the displacement peaks are bigger, showing that the expansion
of the material happens under no force (very small interfacial tension). Actually
this happens when the mass leaves the pressboard and separation of the mass and
pressboard occurred. Figure 7.6-c confirmed the reduction in the stiffness in the
subsequent cycles as the loop became flatter.

a. Applied force b. Force and displacement

6 6
x 10 x 10
8 8

6 6

4 4
Stress (kPa)

Stress (kPa)

2 2

0 0

-2 -2

-4 -4
-0.08 -0.06 -0.04 -0.02 0 0.02 0.04 0.06 0.08 -0.08 -0.06 -0.04 -0.02 0 0.02 0.04 0.06 0.08
Strain (mm/mm) Strain (mm/mm)
c. Stress-strain characteristics d. One element stress-strain
characteristics (red)

Figure 7.6: Applied force, reaction force, resulting displacement, measured stress
strain characteristics and fitted stress-strain curve used in 1-mass model calibration
at pre-stress value of 0 N/mm2

79
7.5.2 Stress-Strain properties at low pre-stress (1.42 N/mm2 )

The results of the measurement are shown in Fig 7.7 a, b and c. In Fig b, the
response of the material is different than the zero pre-stress case. The material in
this case, offers relatively higher stiffness but like the previous case of Fig 7.6, the
stiffness reduces in the following cycles indicated by the time period increase. Also
the material is stiffer in the positive half cycles indicated by the steeper and high
peaks of the force compared to the more flatter peaks in the negative half cycles.

Figure 7.7-c shows similar characteristics as Figure 7.6-c with a difference in the
expansion region, where the negative force is now slightly higher. The separation is
still there but it is smaller.

Time (mS)
-2 0 2 4 6 8 10 12 14 16 18
6 6

5 5

4 4

3 3

Displacement (mm)
2 2
Force (kN)

1 1

0 0

-1 -1

-2 -2

-3 -3
-2 0 2 4 6 8 10 12 14 16 18
Time (mS)

a. Applied force b. Force and displacement

6 6
x 10 x 10
8 8

6 6

4 4
Stress (kPa)
Stress (kPa)

2 2

0 0

-2 -2

-4 -4
-0.03 -0.02 -0.01 0 0.01 0.02 0.03 0.04 -0.03 -0.02 -0.01 0 0.01 0.02 0.03 0.04
Strain (mm/mm) Strain (mm/mm)

c. Stress-strain characteristics d. One element stress-strain characteristics

Figure 7.7: Applied force, reaction force, resulting displacement, measured stress
strain characteristics and fitted stress-strain curve used in 1-mass model calibration
at pre-stress value of 1.42 N/mm2

80
7.5.3 Stress-Strain properties at low to medium pre-stress (4.3 N/mm2 )

Figure 7.8 shows the results of the force vs displacement and stress-strain character-
istics at a pre-stress of 4.3 N/mm2 . Figure 7.8-b shows a change in the stress-strain
characteristics towards relatively more linear characteristics. However, there is a
slight separation below -1.2kN. Also the material offers a higher stiffness than pre-
vious cases. The material still shows similar characteristics to the previous cases, a
lower stiffness in the negative half cycles compared to the positive half cycles.

6
a. Applied force 6
b. Force and displacement
x 10 x 10
8 8

6 6

4 4
Stress (kPa)

Stress (kPa)

2 2

0 0

-2 -2

-4 -4
-0.03 -0.02 -0.01 0 0.01 0.02 0.03 0.04 -0.03 -0.02 -0.01 0 0.01 0.02 0.03 0.04
Strain (mm/mm) Strain

c. Stress-strain characteristics d. One element stress-strain characteristics

Figure 7.8: Applied force, reaction force, resulting displacement, measured stress
strain characteristics and fitted stress-strain curve used in 1-mass model calibration
at pre-stress of 4.3 N/mm2

81
7.5.4 Stress-Strain properties at medium pre-stress (7.1N/mm2 )

Figure 7.9 shows that under the medium values of pre-stress, the material possesses
a higher stiffness than the previous case. The separation is reduced and the material
is very close to linear in terms of stress-strain properties. Also the higher frequency
of the vibration is indicative of higher stiffness and the displacement under the same
force is smaller.

a. Applied force c. Force and displacement

6
6
x 10 x 10
8 8
measured
average
6 6 fitted

4 4
Stress (kPa)

Stress (kPa)

2 2

0 0

-2 -2

-4 -4
-0.03 -0.02 -0.01 0 0.01 0.02 0.03 0.04 -0.03 -0.02 -0.01 0 0.01 0.02 0.03 0.04
Strain (mm/mm) Strain (mm/mm)

b. Stress-strain characteristics d. One element stress-strain characteristics

(black)

Figure 7.9: Applied force, reaction force, resulting displacement, measured stress
strain characteristics and fitted stress-strain curve used in 1-mass model calibration
at pre-stress of 7.1 N/mm2

82
7.5.5 Stress-Strain properties at medium-high pre-stress (10 N/mm2 )

Figure 7.10 shows the material under this pre-stress of 10N/mm2 behaves almost
like a linear spring with hysteresis. The peak values of displacement in positive and
negative half cycles are still not similar but the difference is not as large as in the
previous cases. The stiffness still decreases with time.

6
a. Applied force 6
c. Force and displacement
x 10 x 10
8 8
measured
average
6 6
fitted

4 4
Stress (kPa)

2 2
Stress (kPa)

0 0

-2 -2

-4 -4

-6 -6
-0.03 -0.02 -0.01 0 0.01 0.02 0.03 0.04 -0.03 -0.02 -0.01 0 0.01 0.02 0.03 0.04
Strain (mm/mm) Strain (mm/mm)
b. Stress-strain characteristics d. One element stress-strain characteristics
(black)

Figure 7.10: Applied force, reaction force, resulting displacement, measured stress
strain characteristics and fitted stress-strain curve used in 1-mass model calibration
at pre-stress of 10 N/mm2

83
7.5.6 Stress-Strain properties at high pre-stress (14 N/mm2 )

Figure 7.11 represents the behaviour of the material at the maximum pre-stress
applied to the samples. The resulting stress-strain characteristics are close to the
desirable values. There is no separation and the average stress-strain curve of the
material is almost linear (if hysteresis is omitted). The material offers the highest
stiffness both in the positive and negative half cycles.

6
a. Applied force 6
c. Force and displacement
x 10 x 10
8 8
measured
average
6 6 fitted

4 4
Stress (kPa)

Stress (kPa)

2 2

0 0

-2 -2

-4 -4
-0.03 -0.02 -0.01 0 0.01 0.02 0.03 0.04 -0.03 -0.02 -0.01 0 0.01 0.02 0.03 0.04
Strain (mm/mm) Strain (mm/mm)

b. Stress-strain characteristics d. One element stress-strain characteristics

(black)

Figure 7.11: Applied force, reaction force, resulting displacement and measured
stress strain characteristics at pre-stress value of 14 N/mm2

7.6 Modeling of experimental set up as spring-mass

system

The dynamic stress-strain measurement can be modeled by the spring mass model of
Figure 7.12-b with two linear springs supporting a mass and being attached to a rigid

84
structure. In reality, the pressboard material represented as springs is non-linear.
It has non-linear stress-strain characteristics and a complex damping behaviour.
The objective of this modeling is to separate the stress-strain properties and the
damping coefficient from the measured data of the previous section which will be
used in further models of full transformer.

F
F

M M
x x

a. Simplified apparatus representation b. Equivalent spring-mass system

Figure 7.12: Experimental setup represented as spring-mass system

7.6.1 Mathematical model

The results of the measurements presented in Section 7.5 are the stress-strain prop-
erties of the whole system consisted of two stacks of pressboard (non-linear springs
with damping) and a mass between them. The simplified equivalent diagram of the
system is given in Figure 7.12-b. The equation of motion for the spring mass system
can be written as:

M x
+ C x + g(x) x = F (t) (7.1)

Where:
M = mass,
C = damping coefficient
g(x) = stress-strain characteristics of pressboard
F (t)= force as a function of time
In Equation 7.1, g(x) represents the stress-strain characteristics of the non linear
pressboard and was determined by a lookup table.

The stress-strain characteristics of the pressboard at different pre-stresses measured

in Section 7.5 are available in the form of loops due to the hysteresis of the material
which makes the model very complex.

85
 Applied Force
6

kN
2

2 0 2 4 6 8 10
Time S x 10
3

Displacement: Calibrated model

4

2
mm

4
2 0 2 4 6 8 10
Time S x 10
3

Measured displacement
4

2
mm

4
2 0 2 4 6 8 10
Time S x 10
3

Figure 7.13: Comparison of measured and calibrated results from the model at
pre-stress of 0 N/mm2

Applied Force
6

4
kN

2 0 2 4 6 8 10
Time S 3
x 10
Displacement: Calibrated model
2

1
mm

2
2 0 2 4 6 8 10
Time S 3
x 10
Measured displacement
2

1
mm

2
2 0 2 4 6 8 10
Time> S 3
x 10

Figure 7.14: Comparison of measured and calibrated results from the model at
pre-stress of 1.42 N/mm2

86
 Force
6

kN
2

2 0 2 4 6 8 10
Time S 3
x 10
Modeled
2

1
mm

2
2 0 2 4 6 8 10
Time S 3
x 10
Measured
2

1
mm

2
2 0 2 4 6 8 10
Time S 3
x 10

Figure 7.15: Comparison of measured and calibrated results from the model at
pre-stress of 4.3 N/mm2

Applied force
6

4
kN

2 0 2 4 6 8 10
Time S x 10
3

Displacement: Calibrated model

2

1
mm

2
2 0 2 4 6 8 10
Time S x 10
3

Measured displacement
2

1
mm

2
2 0 2 4 6 8 10
Time S x 10
3

Figure 7.16: Comparison of measured and calibrated results from the model at
pre-stress of 7.1 N/mm2

87
 Applied force
6

kN
2

2 0 2 4 6 8 10
Time S 3
x 10
Displacement: Calibrated model
2

1
mm

2
2 0 2 4 6 8 10
Time S 3
x 10
Measured displacement
2

1
mm

2
2 0 2 4 6 8 10
Time S 3
x 10

Figure 7.17: Comparison of measured and calibrated results from the model at
pre-stress of 10 N/mm2

Applied force
6

4
kN

2 0 2 4 6 8 10
Time S x 10
3

Displacement: Calibrated model

2

1
mm

2
2 0 2 4 6 8 10
Time S x 10
3

Measured displacement
2

1
mm

2
2 0 2 4 6 8 10
Time S x 10
3

Figure 7.18: Comparison of measured and calibrated results from the model at
press-stress of 14 N/mm2

88
It is been established in the literature and here in the results of the measurements
that the properties of the pressboard not only vary under different stress but have
a significant hysteresis [Swihart and McCormick, 1980]. In this work, to simplify
the problem, the hysteresis was neglected and an average curve (as shown in the
sub-figures d of Figures 7.6 to 7.11) passing through the hysteresis loop was used
for further calculations. This was not a big limitation as the difference in response
only becomes significant later on in time. In reality, the first few peaks of the
displacement are off interest (under transient conditions). After the first few peaks,
the current transient and resulting forces become smaller and are not as destructive.
Also, it is believed that the peak of the first half cycle is the highest peak and is the
main cause of winding failures in power transformers [IEC 60076, 1992].

The nonlinear springs of Figure 7.12 are directional springs and have different char-
acteristics under compression and expansion. In the model, the two springs are
connected to the mass and can be considered as back to back. When the impulse
load is applied to the mass, the bottom spring gets compressed while the top spring
expands. Similarly when the bottom spring relaxes the top spring goes into axial
contraction. A schematic representation of the collective response of both of the
springs is of a push-pull type. The simplified stress-strain characteristics of the sys-
tem of springs are given in Figure 7.19. The stress-strain characteristics measured
in Section 7.5 is the combined effect of the two springs. Therefore the properties of
a single spring are required for simulation. Due to the nonlinear behaviour of the
material there is no easy way to extract the stress-strain properties of a single spring
from the measured data. Also the properties of the material change with different
pre-stress, which makes the problem more complicated and no single technique was
sufficient to extract the properties of a single spring. Hence the model approach
was adopted. The model was calibrated to the measured response to determine
the stress-strain characteristics and damping coefficient for a single spring. It is an
adaptive model.

During the model adaptation process, curves were fitted and the properties of one
spring were extracted. Figures d in Figures 7.6 to 7.11 show the stress-strain char-
acteristics required to produce a curve similar to the measured one.

As a result of simplifications made in the model, there is a large discrepancy in fre-

quency between predicted and measured frequency of oscillation particularly, when
pre-stress is absent or very low. However, the objective of this work is to predict the
maximum displacement which is the first peak and subsequent peaks are of second-
ary importance. For all values of pre-stress, the co-relation between measured and
predicted maximum displacement is good.

89
In the cases of pre-stress of 0, 1.42 and 4.3 N/mm2 , only one side of the pressboard
stacks shows stiffness. The expanding spring has negligible or no contribution to the
oscillatory motion. Hence the stress-strain curve (of a single pressboard component
spring) for these pre-stresses is the average curve without the hysteresis. As the
pre-stress increases, the expansion of the so-called springs becomes a reaction force.
In the case of pre-stress of 7.1 N/mm2 , the force contribution of the expanding
pressboard part becomes about 15% of the measured curve, the lower curve of Figure
7.9. An expansion effect of 30 % of the measured stress-strain characteristics came
from the expanding pressboard component at a pre-stress of 10 kN and 14 kN/mm2 .
At these pre-stress values, the average curve of the measured stress-strain values is
more linear.

In the measurement of the stress-strain characteristics in Section 7.5, the damping

and the stiffness of the displacement sensor was neglected. It was noticed that when
the displacement was large there was some extra stiffness in the system which could
possibly be due to the displacement sensor. The flexible arm length of the displace-
ment sensor was only 65mm which under larger displacements may create stiffness
issues. Hence care was taken to ensure the sensor arm remained in a horizontal
position before the load is applied. This means that the stress-strain characteristics
may not have been measured with a high accuracy. However, the same sensors were
used for the transformer experiment and the errors are likely to be consistent in
both sets of measurements.

upper
spring
stress

strain

lower
spring

Figure 7.19: Schematic representation of the stress-strain properties of two nonlinear

springs in a back to back arrangement

90
 Pre-stress Damping coefficient
N/mm2 N.s/m
1.42 170
4.3 170
7.1 125
10 90
14 90

Table 7.2: Damping coefficient at different pre-stress values

7.6.2 Damping

The results of the experimental work (Figures b of Figures 7.6 to 7.11) are shown as
force and displacement curves. The results for displacement vs time indicate that
the material has a significant damping. The damping is maximum in the first cycle
and decreases gradually thereafter as illustrated in Table 7.2. This behaviour of the
material can be attributed to the oil squeeze-out of the pressboard under the initial
force impact. The inward flow of the oil has to happen under normal atmospheric
pressure. This is either very little or almost non-existent in the short space of time
before the second cycle of the transient starts, which again removes the major portion
of the remaining oil. The evaluation of this non-linear damping, coupled with the
hysteretic damping (as the material has hysteresis) is difficult and no simple method
can be used for the prediction of damping. The damping coefficient was needed to
fit the behaviour under different pre-stress values. Therefore, the model calibration
approach for the prediction of the damping coefficient was used. The damping
coefficient is larger at low pre-stress compared to the higher pre-stress value. It
must be noted that the viscus damping of the oil, in which the windings move, is
neglected in this work.

r

The mathematical model of the spring-mass system was implemented in Simulink
as shown in Figure 7.20. The results obtained after the calibration of the model
are presented in Figures 7.13 to 7.18. The stress-strain curves to be used in the
further simulations are given in part-d of Figures 7.6 to 7.11. The stress-strain
characteristics (fitted waveform part-d Figures 7.6 to 7.11) and damping coefficient
values obtained from the calibrated model will be used in the modeling of an actual
transformer.

91
Impulse force
(Chapter 7) Mass= 2.3 kg x1000
Insulation stack
[t1,f1] thickness = 30 mm
1 1
From -K- -K-
s s
Workspace
M_i Intg 1 Intg 2 mm

-K- x_i

Damp 0 ins_thick

-1 -K- Constant2 -K-

Damping
-3 ins_a3 Switch2 -1
(Table 7.2)
K4 -2
[t,d]

From
Workspace1
-K- K3

ins_a2 Switch1 0 Pressboard Measured

stress-strain value of
Constant1 lookup table
Pressboard stack displacement
area = 707 mm2

Figure 7.20: Simulink model of mass-spring system

7.7 Discussion
The experimental measurements shown in Section 7.5 indicate that high density
pressboard possesses non-linear properties with hysteresis and complex damping.
However, the linear approximation of damping is appropriate for initial cycles.

It is noted that the material shows an initial higher stiffness under dynamic loading
which can be attributed to the squeezing out of oil as mentioned by Patel [1972].
Subsequently, the material shows lower stiffness and consequently a lower frequency
of vibration with time.

The properties appear in their most complex form in the first case where there is
no pre-stress. The time period of the first half cycle is much smaller than the later
ones. This is due to a considerable change in stiffness which is attributed to the
outward oil flow from the pressboard material. Once the oil is out, the material has
a very low stiffness and low frequency of vibration compared to the oil saturated
situation in the first half cycle.

As the pre-stress increases the non-linearity of the material decreases (Figures 7.6
to 7.11). This is evident from Fig c of Figure 7.6 where there is no negative force
and the highest negative force is observed in Fig 7.11.

92
The reaction effect of the expanding pressboard section increases with the pre-stress.
This is evident from the measured stress-strain characteristics because the press-
board non-linearity reduces with the increase of pre-stress. This is the reason for
the declining slope of the average stress-strain curve representing each cycle of vi-
bration as seen in the Figure-c of Figures 7.6-7.11.

Due to the inaccuracies and limitations of the sensors, the curve fitting was not very
accurate in cases of larger displacements. This could be prevented in future work
by choosing non-contact type sensors for displacement.

To simplify the model, damping has been considered as constant for each value of
pre-stress rather than using a function dependant on stress, time and degree of oil
impregnation. The simplification has still permitted a reasonable prediction of the
first and highest displacement peak.

7.8 Conclusions

1. The pressboard material is highly non-linear under low pre-stress. As the pre-
stress increases the material becomes more and more linear. This is due to the
changed degree of oil impregnation of the pressboard.

2. The oil contained in oil impregnated pressboard is removed by impulse or

loading and the material becomes softer. However, the oil can move back into
the pressboard and restore the stiffness if sufficient time is allowed.

3. It was not easy to have a simple equation for predicting stress-strain charac-
teristics of one spring. Hence the technique of model calibration is suggested.

4. The oil impregnated pressboard possesses a complex damping. The damping

coefficient was predicted by the model calibration.

5. The damping properties of the oil impregnated pressboard decrease with in-
crease of pre-stress.

6. The reaction effect of the expanding pressboard (non-linear spring) increases

with the increase in the pre-stress.

7. Non-contact type sensors should be used for more accurate measurements.

8. This work shows that a sufficient amount of pre-stress is necessary in a trans-

former if a near linear mechanical behaviour of pressboard is desired.

93
Chapter 8

Combined transformer model

8.1 Introduction
In the past, researchers have investigated the axial and radial models separately.
In a real transformer, these behaviours are coupled and believed to have an effect
on each other. The electromagnetic forces in a transformer have radial and axial
components due to the positions of the turns in the winding. During a short circuit,
the winding movements result in a change in relative position of the turns. This
results in change in the forces. In this work, the movement of the windings are
assumed very small. Hence the forces will remain same throughout the dynamic
phenomenon.

The dynamic radial forces can vary due to the axial movement and vice versa. The
movement of the windings in axial direction winding may be limited due to the
friction offered by the structure of the windings and core.

The implementation of a combined transformer model is discussed in this chapter.

The radial and axial models developed in previous Chapters 5 and 6 are combined to
assess the behaviour of the transformer winding when subjected to pulsatory forces
produced by the short circuit currents. The method of combining of the models is
r

also presented in this chapter. The combined model was implemented in Simulink .
The detail of the implementation of the model is given in appendix B. This work is
limited to helical and disk type windings where the pressboard axial spacers are used
as inter-disk insulation. The helical and disk type windings are the most common
windings in large power transformers.

94
8.2 Components of the model
The active part of a power transformer consists of a number of components like
core, yoke, clamping structure, axial and radial insulation, copper conductors and
the winding end insulation. The core and yokes consist of steel laminations stacked
together. The core and yoke are made of magnetic material to cage the main flux.
The clamping structure of the windings consists of core clamps and tie bars. They
are made of either magnetic or non-magnetic steel of high tensile strength. During
short circuits, deformation in the clamping structure is rare. The weakest part in
the winding is the axial and radial insulation (pressboard), which provides most
of the flexibility and damping under the influence of the compressive forces during
the fault conditions especially for helical and disk type windings. The mechanical
representation of the insulation and winding conductors is discussed in the following
sections.

8.2.1 Axial spacers (pressboard)

The axial spacer is used in the winding to maintain the winding structure and for
inter-disk insulation. In this work, the axial, radial and end insulation is represented
as a non-liner directional spring connected to a dashpot in parallel as shown in Fig
8.1 (a). The oil impregnated pressboard insulation has non-linear characteristics
under dynamic load conditions which has been discussed in detail in Chapter 7.

K C mass

(a) (b)

Figure 8.1: Components of mechanical model of a transformer

8.2.2 Lumped masses, springs and dashpots

The mass of the whole winding can be lumped into several smaller portions to reduce
the degrees of freedom of the system. The lumped parameters should be close to the
actual structure so that the dynamic behaviour predicted by the lumped mass model
is sufficiently close to the actual behaviour. The idea is to reduce the computational
time and to make the model more manageable. The similar techniques have been
adopted in literature [Tournier et al., 1964, 1962a,b; Patel, 1972, 1973b]. The authors

95
in their models lumped few disks of the disk type winding as one mass to make the
problem simpler for the computational process. However, in this study each disk
is represented as a separate mass which is more realistic and comparable with the
actual large power transformer where each disk of the winding is connected to the
next disk with the axial spacer (pressboard). Similar technique is followed for the
lumped non-linear springs and one set of axial spacers (pressboard) between two
disks is treated as one spring. The same rule applies to the damping effect of the
vertical spacer.

8.3 Combined model of a transformer winding

To establish a combined model of a winding, the following components/steps are
required:

electromagnetic forces calculation to predict

axial forces
radial electromagnetic forces

dynamic models

dynamic axial model

dynamic radial model

combined model

The prediction of the electromagnetic forces, the factors influencing them and meth-
TM
ods of calculation are discussed in Chapters 3 and 4. A Matlab code was written
to calculate these forces is given in Appendix A. The code generates a time vary-
ing waveform which was further used as the input to determine the response of the
r

dynamic model. The dynamic model was implemented and simulated in Simulink .

8.4 Axial and radial models

The arrangement of components and the mathematical representation of radial and
axial models is given in Chapters 5 and 6 respectively. The models are built with
the components discussed in the Section 8.2.2.

96
8.5 Coupled dynamic model
The suggested coupled dynamic model of a transformer is discussed here along with
coupling technique. Figure 8.2 shows the components of the coupled two-winding
transformer model including the friction with core which may retard the axial mo-
tion.

In a perfectly designed and manufactured transformer, the windings are always under
compressive forces even during short-circuits. During the dynamic phenomenon
(under short circuits) the axial forces will become different from the applied pre-
stress and may not remain the same for the whole winding especially when the
windings are of different heights or in the event of geometric mismatch of their
centers. This could result in magnetic unbalance resulting in unbalanced forces in
the upward or downward directions. The forces which are working against the pre-
stress will produce a very high or no pressure in the magnetically unbalanced regions
of the winding. The areas where the pressure is higher than the pre-stress the radial
movement will be impeded due to the increased friction on the surface of the vertical
spacer surfaces as shown in Figure 8.4. However, the forces on the unsupported
length of the conductors will remain same as the electromagnetic force. On the
other hand, the movement of the inner winding towards the core limb (inward),
will produce enhanced friction and resistance to the axial movement (Fig 8.5). This
phenomenon is only applicable to the inner winding provided there is no permanent
deformation during the short circuit. The flow chart of Figure 8.3 shows the coupling
r

technique used to couple the axial and radial models in Simulink . The friction
between the core and the inner winding can be regarded as Coulomb damping. Patel
[1972] has suggested a method to determine the coefficient of friction experimentally.
In this work the friction coefficient determined by Patel was used.

8.6 Test transformer

For model response validation, a single phase, two-winding test transformer was built
and is shown in Figure C.1. The diagram of the core section and winding detail is
given in Figure 8.6 and 8.7. The detail of the design of the test transformer is given
in Appendix C. The transformer has a 1:1 ratio. Both inner and outer windings
were of the disk type having 6 disks each with 5 turns in each disk. The inter-disk
insulation (clacks or vertical spacer thickness) is 10 mm. A larger thickness of 10 mm
was used to reduce the stiffness in the axial direction and was not due to the voltage
between the disks which are not more than 20V under normal operating conditions.
The other advantage of identical windings was to ensure zero initial displacement

97
 c1 k1 C1 K1
f1(t)+m1 g F1 (t)+M1 g

m1 M1
ccore x1 X1
c2 C2
k2 K2
f2 (t)+m2 g
F2 (t)+M2 g
m2 M2
ccore x2 X2
c3 k3 C3 K3

f3 (t)+m3 g F3 (t)+M3 g
m3 M3
ccore
x3 X3
c4 C4 K4
k1

ci ki Ci Ki

fi(t)+mig Fi(t)+Mig
mi Mi
ccore xi Xi
c(i+1) C(i+1)
k(i+1) K(i+1)

c(n-1) C(n-1)
K (n-1) K(n-1)
fn (t)+mn g Fn(t)+Mng

mn Mn
ccore
xn Xn
cn Cn Kn
kn

Figure 8.2: Proposed two-winding dynamic axial model

which could have produced unexpected forces. The disadvantage of the disk type
windings is its stiffness (in radial direction) which is very high and results in a very
small radial displacement.

The inner winding of the transformer was permanently short circuited and the cur-
rent was injected in the outer winding.

8.7 Test setup and procedure

The circuit diagram of the test setup is given in Figure 8.10. A three phase alternator
was used to supply the single phase current. The alternator was rated 450kVA and
was driven by a forced air cooled motor. The terminal of the voltage of the alternator

98
 Figure 8.3: Dynamic axial and radial models combining technique

could be controlled manually and was used to control the current. A maximum short
circuit current of 4kA could be drawn from the alternator at a voltage of 130V.

An electronically controlled dc powered switch was used to energize the short-

circuited transformer. The current switching mechanism was able to operate around
100 s of zero crossing of the voltage waveform. It was necessary to close the switch
at the zero crossing for a steady increase of first half cycle of the current giving the
maximum displacement in the transformer windings. The current was injected into
the transformer using a break-before-make mechanism where the supply of the mo-
tor driving the alternator was Switched Off before applying the short circuit on the
terminals of the alternator. The maximum peak current injected was under 4kA.
Current higher than 4kA was not possible due to the susceptibility of malfunction-
ing of the electronic switch as the higher terminal voltage was required for higher
current but the zero crossing switch was not capable of handling the electromagnetic
noise generated at higher voltages.

The test transformer was equipped with two displacement sensors (Chapter 7). The
first was used to measure the axial movement at the lower most disk of the outer
winding. The second displacement sensor was used to pick up the radial displacement
of the same disk.

99
 Direction of
motion

Outer copper disk

Inner copper disk
Insulation rib

Insulation rib
Core

Radial force

Friction at
core-inner disk
interface

Figure 8.4: The effects of friction on axial motion

Vertical Force

Copper disk

Clack (vertical insulation

Radial motion component)
(horizental)

Copper disk

Vertical Force

Figure 8.5: The effects of friction on radial motion

8.8 Simulated and measured results

Figures 8.11 and 8.13 show the simulated and measured results of axial displacement
of the bottom disk of the test transformer at a medium pre-stress of 7.1 N/mm2 and
a high pre-stress of 14 N/mm2 . The results qualitatively co-relate with the simulated
results. However, quantitatively the measured displacement is 20%-30% larger than
the simulated values.

Figures 8.12 and 8.14 show the measured and simulated displacements in radial
direction. The measured displacements are significantly larger than the simulated
radial displacements. The larger displacement is due to the difficulty in separating
the in-phase larger axial displacement from the very small radial displacement.

Another important observation was, that the interaction between the radial and axial

100
 3 mm

8 mm 45
mm

8 mm 60
mm

8 mm 80 mm

8 mm 100 mm

110 mm

120 mm
50 mm 120 mm

120 mm

Figure 8.6: The cross-section of the core of the test transformer

forces was negligible. During the period of forced vibration, there is no significant
difference between the axial displacements of the combined and stand-alone models
as illustrated in Figure 8.9. However, during free vibration, the combined model
shows a higher damping and a rapid decay of vibration which is also clearly obvious
from Figure 8.9. This is due to the axial friction which is modeled in the combined
model but neglected in the stand-alone axial model.

101
 2 mm inner
cylinder
220 mm dia

8 mm rib
Duct 8 mm
2 mm inner cylinder outer
winding (inner dia 274mm)
Outer cylinder 1mm
18 mm outer winding
8 mm rib

End insulation

Winding disks 5
turns each

Disk where sensors

were attached

Figure 8.7: The detail of the winding configuration of the test transformer

Figure 8.8: Single phase two-winding test transformer in experimental setup

102
 Combineed Model: Axial displacement of top disk
0.01

0
Displacement (mm)

0.01

0.02

0.03

0.04

0.05
0 0.02 0.04 0.06 0.08 0.1 0.12
Time (S)

Axial displacement of top disk in standalone axial moxel

0.01

0
Displacement (mm)

0.01

0.02

0.03

0.04

0.05
0 0.02 0.04 0.06 0.08 0.1 0.12
Time (S)

Figure 8.9: Comparison between stand-alone axial model and combined model of
the test transformer

Zero crossing switch

Outer Inner
M A winding
winding
3-ph, 450kVA, 50 Hz
Alternator set

Test Transformer

Figure 8.10: Single phase two-winding test transformer

103
 Axial displacement: Modeled
0.05
mm

0.05
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1
S
Axial displacement: Measured
0.05
mm

0.05
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1
S

Figure 8.11: Simulated and measured results of axial displacement at a pre-stress of

7.1 N/mm2

3
x 10 Radial displacement: Modeled
1

0.5
mm

0.5

1
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1
S
Radial displacement: Measured

0.02
mm

0.02

0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1
S

Figure 8.12: Simulated and measured results of radial displacement at a pre-stress

of 7.1 N/mm2

104
 Axial displacement: Modeled

0.04
0.02

mm
0
0.02
0.04
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1
S
Axial displacement: Measured

0.04
0.02
mm

0
0.02
0.04
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1
S

Figure 8.13: Simulated and measured results of axial displacement at a pre-stress of

14 N/mm2
3
x 10 Radial displacement: Modeled
1
mm

1
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1
S
Radial displacement: Measured
0.02
mm

0.02

0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1
S

Figure 8.14: Simulated and measured results of radial displacement at a pre-stress

of 14 N/mm2

8.9 Discussion
There is no significant difference between the axial response of the combined and
stand-alone models (Fig 8.9). This can be attributed to the stiffness of the trans-
former resulting in a very small displacement in the radial direction. The combined
model of low stiffness designs and in larger transformers (where larger radial dis-
placement are expected) can be different from stand-alone models due the axial
friction forces between the core and the inner winding.

The modeling of hysteresis is approximate and the recovery time of the pressboard
has not been included in the model, nor has the visco-elastic behaviour of the oil im-
pregnated pressboard been comprehensively modeled. It is to be expected therefore,

105
that there will be a discrepancy between measured and predicted displacements. Un-
der the circumstances, a discrepancy of 20% to 30% can be taken as pointing to a
reasonably representative model.

A loss in pre-stress was observed during the successive application of current. This
shows that in-service, a transformer may lose its applied pre-stress, especially after
the event of a short-circuit and become more susceptible to damage.

The measured axial displacement in Figures 8.11 and 8.13 is larger (20%-30%) than
the simulated displacement. This could be due to the inaccuracy of measurement,
hysteresis (which was neglected), disk profile, recovery time of pressboard and visco-
elastic behaviour of impregnated pressboard.

Inaccuracy: The inaccuracy can be reduced by using more accurate instruments

designed to record very small displacements. The interference from the leakage
magnetic field in transformers may make these measurements more difficult.

Hysteresis: It has been established that the pressboard is a non-linear material and
has significant hysteresis (Chapter 7). After the initial impact of force (during
first half cycle), the pressboard between the disks is squeezed and reduced in
thickness. The pressboard material does not return to its normal thickness
before the second half cycle of the force but shows a longer time constant for
recovery. This is evident from the recorded waveforms in Figures 8.11 and
8.13, where it is clear that the thickness is reduced and it is only restored to
its original value after several cycles of oscillation.

Profile: The interface profile of the mass in the stress-strain measuring rig was
smooth. Contrary to the test transformers disk profile (the mass element)
which was built by radially wound, paper covered coils of 5 conductors (2mm
thick) as shown in the Figure 8.16. It is common practice that the edges of
the conductors are made round to reduce the chances of high electric stresses.
Hence the vertical spacer (pressboard), will exhibit a lower stiffness and a
larger displacement, due to the profile of the disk

Time of recovery: The time of recovery of the pressboard material is also longer
than the period of the applied forces. This in turn showed a reduction in
the stiffness in the later cycles of the force and the recovery of the material
dimensions was due to the decaying force. If the force peaks were of similar
magnitude of the 1st peak, the recovery of the pressboard material would have
been even slower.

Visco-elastic behaviour: An impact force squeezes oil out of the impregnated

106
 pressboard. The process of oil returning back into the pressboard material is
slow because it happens at a pressure close to the atmospheric pressure [Patel,
1972]. The oil squeeze-out reduces the stiffness of the material which recovers
with inward flow of oil.

Figure 8.15: Construction of the force and displacement sensors. The force sensor
was used to measure the pre-stress and the two displacement sensors were used for
axial and radial displacement

Paper covering
Copper conductor

Disk-pressboard
interfaces

Figure 8.16: The cross-section of the transformer disk showing the profile of the area
interfacing with the vertical spacers (pressboard)

The measured radial displacements are significantly larger than the simulated radial
displacements. This was due to the complexity of coupled behaviour of the trans-
former and the inaccuracy of the instrumentation used. The radial displacement
was expected to be a few m, which is extremely difficult to measure with the type
of instrumentation used.

The displacement sensor employed to measure the radial displacement was a strain
gauge type and is shown in the Figure 7.4. The force sensor was also a strain gauge

107
type. The sensors were used for their simplicity and inherited property of not being
effected by the magnetic field. The sensors had the capability of picking up some
effects of the axial movement which was in phase with the radial response. Qualit-
atively, the measured and modeled radial displacement show reasonable correlation.

The test transformer was of very high stiffness and had a high natural frequency.
Also the injected current was not large enough to produce significant displacements,
which was one of the reasons for the simulated displacement not going negative.

8.10 Conclusions
1. The difference between simulated and measured response is within 20%-30%
for axial behaviour. In view of the complexity of the model and challenges in
measurement, this discrepancy is to be expected.

2. Qualitatively, the measured radial response was similar to the simulated dis-
placement. The difference between the measured and predicted radial re-
sponses is due to the complexity of measurement where it was difficult to
separate the axial and radial movements. The larger measured radial dis-
placement could also be due to the lose turns especially the outer turn of the
disk where the displacement sensor was installed.

3. Bearing in mind the complexity of the mechanical system, predicted dynamic

behaviour determined from the model, has been shown to be realistic and
representative of the transformer.

4. There was no significant difference in response of the combined and stand-alone

models of test the transformer. This can be attributed to the high stiffness of
the test transformer where the predicted radial displacement was negligible.
In low stiffness designs or in larger transformers, where radial displacements
may be larger, the responses of the models may be significantly different.

5. Loss in pre-stress between the successive applications of current showed that,

after a short circuit, a transformer may loose its applied pre-stress.

6. A significant hysteresis was shown by the pressboard material and the stiffness
recovery time was long. This may lead to a failure if short circuits are frequent.

7. If a large number of approximating functions are used, the model may require
greater computer resources to execute the model.

108
Chapter 9

Conclusion and recommendations

9.1 Conclusion

1. In terms of the objectives of this research,

(a) It is been shown that the dynamic behaviour of a transformer winding is

dependant on the dynamic characteristics of the oil impregnated press-
board and paper which form the electrical insulation system and the
mechanical support system for the conductors in the coils of the winding.
(b) These properties of the pressboard have been measured and after reason-
able simplification have been incorporated in a comprehensive model in
which the basic element is a single disk in a winding with axial and radial
pressboard spacers.
(c) The model has been tested experimentally and the discrepancy between
the measured and predicted displacements range between 20% and 30%.
Bearing in mind the complexity of the model and difficulties in the meas-
urements, this discrepancy is to be expected. It can be assumed that
with further refinements to the model and improvements in the measur-
ing techniques, this discrepancy can be significantly reduced.

2. In addition the research has highlighted the following issues:

(a) The axial movements of the windings produce much larger displacements
than those in the radial direction. Although the radial forces are higher
than the axial forces, the stiffness in the radial direction is also signific-
antly higher than the axial stiffness, resulting in very small radial dis-
placements.
(b) Physically, the axial and radial behaviours are coupled but the high ra-
dial stiffness allows very small radial displacements and little or no extra

109
 friction to the axial movements. Hence, when the radial displacements
are small, the radial and axial dynamic behaviours are independent of
each other and can be studied separately. This statement is also valid for
transformers where windings are designed as self supporting in the radial
direction and likely to have very small radial displacements.
(c) The dynamic stress-strain characteristics show significant hysteresis, res-
ulting in a slower recovery to the initial size and stiffness compared to
the pulsating electromagnetic forces. Even with a very stable material
(where the hysteresis is reduced to a much smaller value and the pre-
stress is maintained), under a long duration fault, after first few cycles,
the stiffness offered by the pressboard will reduce, resulting in larger dis-
placements which could lead to a mechanical failure of the windings.
(d) Traditionally, the tripping time of the protection of power transformers
is based on the heating effects of the conductors used in the windings.
These have a longer time constant than the stiffness recovery time of
the pressboard. Older transformers are even more susceptible to this
type of failure due to the loss of pre-stress due to ageing [Krause, 2003].
Faster and intelligent tripping is required to avoid failures due to reduced
stiffness and pre-stress.

9.2 Recommendations for further research

This work has explored the nonlinear mechanical behaviour of transformer press-
borad and implemented it in a model of a complete transformer. Patels model
[Patel, 1972] has been significantly extended. There is now a solid base upon which
further research can be conducted:

Dynamic properties of the pressboard

Techniques need to be developed to determine the properties of the pressboard

to a higher degree of accuracy under all representative values of pre-stress

Special non-contact type displacement sensors will be needed to prevent inter-

ference with the behaviour of the pressboard

The electromagnetic forces in a 50 Hz system have a frequency of 100 Hz.

This necessitates the measurement of dynamic stress-strain characteristics at
or close the frequency of 100 Hz to determine realistic parameters of modeling.

110
The separation of axial and radial movement

This could possibly be achieved by installing sensors at the axial geometrical center
of the winding where little or no forced axial movement is expected. This applies
when the windings are stiff in the axial direction, have no initial displacement and
are of the same height. Special non-contact type sensors may provide the solution

The conditions under which the axial and radial behaviours can be studied inde-
pendently need further investigation.

The extension of this work to include layer type winding needs to be studied.

111
References

C M Arturi. Force calculation in transformer windings under unbalanced MMFs by

a non-linear finite element code. In IEEE Transactions on Magnetics, volume 28,
pages 13631366, March 1992.

R E Ayres, G O Usry, M R Patel, and R B Steel. Dynamic measurement during

short circuit testing of transformers, part 2- test results ans analysis. In IEEE
Trasactions on Power Apparatus and Systems, volume PAS-94, pages 198206,
1975.

E Billing. Mechanical stress in transformer windings. In IEE Transactions, pages

227248, 1946.

Final Report: Cigre Working Group 12-04. Calculation of short-circuit forces in

transformers. In Electra, number 67, pages 2975. Cigre, 1979.

A C Franklin and D P Franklin. The J & P Transformer Book. Butterworth & Co.
(Publishers) Ltd, England, 11th edition, 1983.

B Hague. Electromagnetic Problems in Electrical Engineering. Oxford University

Press London: Humpherey Milford, 1929.

K Hiraishi. Mechanical strength of transformer winding under short-circuit condi-

tions. In IEEE Transactions on Power Apparatus and Systems, pages 23812390,
October 1971.

Y Hori and K Okuyama. Axial vibration analysis of transformer winding under

short circuit conditions. pages 433451. IEEE Transactions on Power Apparatus
and Systems, March/April 1990.

IEC 60076. Power Transformers - Ability of withstand Short-circuit. IEC, 2nd

edition, 1992. Part-5.

Y V Tournier J P Martin, J Wendling. Specifics of the short-circuit behavior of

concentrically-wound transformers and autotransfromers. In IEEE Transactions
on Power Apparatus and Systems, volume PAS-99, pages 20692077, Nov/Dec
1980.

112
C Krause. The variation of the clamping pressure of power transformers due to
operational temperature cycles and loss of pressure with time. In Cigre Study
Committee A2 Transformers Colloquium, Merinda. CIGRE, 2003.

A B Madin and J D Whitaker. The dynamic behaviour of transformer winding

under axial short-circuit. In IEE Proceedings, volume 110, pages 535550, March
1963a.

A B Madin and J D Whitaker. The dynamic behaviour of transformer winding

under axial short-circuit forces. volume 110, pages 535550, March 1963b.

E T Norris. Mechanical strength of power transformers in service. In IEE Proceed-

ings, pages 289300. IEE, Feb 1957.

M R Patel. Dynamic response of transformer windings and clamps under axial short-
circuit forces. PhD thesis, Rensselaer Polytechnic Institute, Troy, New York, 1972.

M R Patel. Dynamic response of power transformers under axial short circuit forces
part-1 winding and clamps as individual components. In IEEE Transactions on
Power Apparatus and Systems, volume PAS-92, pages 15581566, 1973a.

M R Patel. Dynamic response of power transformers under axial short circuit forces
part-II winding and clamps as a combined system. In IEEE Trasactions on Power
Apparatus and Systems, volume PAS-92, pages 15671575, 1973b.

A Pichon and B Hochart. Problems concerning the behaviour of transformers on

short-circuit. Technical Report 142, Cigre, 1958.

M R Saravolac, P A Vertigen, C A Sumner, and W H Siew. Design verification

criteria for evaluating the short circuit withstand capability if transformer inner
windings. In Cigre 2000, paper 12-208, 2000.

M G Say. The Performance and design of Alternating Current Machines. Sir Issak
Pitman & Sons, Ltd, London, 3rd edition, 1958.

D O Swihart and L S McCormick. Short circuit vibration of shell form power

transformer. In IEEE Transactions on Power Apparatus and Systems, volume
PAS-99, pages 800810, March/April 1980.

D O Swihart and D V Wright. Dynamic stiffness and damping of transformer

pressboard during axial short-circuit vibration. In IEEE Trasactions on Power
Apparatus and Systems, volume PAS-95, pages 721728, March/April 1976.

S Timoshenko, DH Young, and W Weaver(JR). Vibration Problems in Engineering.

John Wiley & Sons, Inc., 4th edition, 1974.

113
Y Tournier, G Ebersohl, A Ciniero, S Yakov, A B Madin, and J D Whitaker.
A study of the dynamic behaviour of transformer windings under short-circuit

conditions. In Cigre Conference Internationale des Grands Reseaux Electriques `
a
Haute Tension, number 143, 1962a.

Y Tournier, G Ebersohl, A Ciniero, S Yakov, A B Madin, and J D Whitaker.

Supplement to paper 143. In Cigre Conference Internationale des Grands Reseaux

Electriques `
a Haute Tension, number 143a, 1962b.

Y Tournier, M Richard, A Ciniero, S Yakov, A B Madin, and J D Whitaker. A

study of the dynamic behaviour of transformer windings under short-circuit con-

ditions. In Cigre Conference Internationale des Grands Reseaux Electriques `
a
Haute Tension, number 134, 1964.

M Waters. The measurement and calculation of electromagnetic forces in concentric

transformer windings. In IEEE Transactions, pages 3546, 1953.

M Waters. The Short-Circuit Strength of Power Transformers. Macdonald & Co

(Publishers) Ltd, Golf House, 2 Portman Street, London, W.1, 1966.

G B Watts. A mathematical treatment of dynamic behaviour of a power-transformer

winding under axial short-circuit forces. In IEE Proceedings, volume 110, pages
551560, 1963.

114
Appendix A

Matlab source codes

This code is where the model is defined and all input parameters are entered for
r

electromagnetic forces, and dynamic calculation for input to the Simulink models.

A.1 Input code

% I n p u t s f o r t h e programmes t o c a l c u l a t e t h e e l e c t r o m a g n e t i c f o r c e s ,
% m e c h a n i c a l c h a r a c t e r i s t i c s o f t h e t r a n s f o r m e r i n a x i a l and r a d i a l
% d i r e c t i o n and i n p u t s t o t h e s i m u l i n k model .
% This f i l e i s f o r m a t t e d t o be p r i n t e d i n t h e t h e s i s ( i n l a t e x )
clear
% Winding d e t a i l s ( d i m e n s i o n s o f w i n d i n g )
% T o t a l number o f d i s k s i n i n n e r w i n d i n g
d i s k s i n =5;
% T o t a l number o f d i s k s i n o u t e r w i n d i n g ( U n i f o r m d i s k s assumed )
d i s k s o t =5;

% Number o f t u r n s i n e a c h d i s k o f i n n e r w i n d i n g
t u r n s i n =5;
% Number o f t u r n s i n e a c h d i s k o f o u t e r w i n d i n g
t u r n s o t =5;

% Number o f r a d i a l r i b s o f i n n e r w i n d i n g = number o f c l o c k s
r i b s i n= 4;
% A x i a l t h i c k n e s s of each r i b i n i n n e r winding .
r i b s t i n = 121 e 3; %m
% R a d i a l width of each r i b i n i n n e r winding
r i b s w i n = 121 e 3; %m

% Number o f r a d i a l r i b s o f o u t e r w i n d i n g
r i b s o t = 6;

115
% A x i a l t h i c k n e s s of each r i b of o u t e r winding
r i b s t o t = 121 e 3; %m
% R a d i a l width of each r i b of the o u t e r winding
r i b s w o t = 121 e 3; %m

% Axial i n s u l a t i o n dimensions of inner winding

% Thickness of the i n n e r clack
c l a c k t i n = 5 1 e 3; %m
% Width o f t h e i n n e r c l a c k
c l a c k w i n = 40 1 e 3; % m
% Length of the i n n e r c l a c k
c l a c k l i n = 60 1 e 3; % m

% Axial i n s u l a t i o n dimensions of outer winding

% Thickness of the outer clack
c l a c k t o t = 5 1 e 3; %m
% Width o f t h e o u t e r c l a c k
c l a c k w o t = 40 1 e 3; % m
% Length of the o u t e r c l a c k
c l a c k l o t = 60 1 e 3; % m

% Paper i n s u l a t i o n t h i c k n e s s wrapped on t h e c o n d u c t o r
% T h i c k n e s s o f t h e p a p e r i n s u l a t i o n wrapped on t h e i n n e r w i n d i n g
t i n s i n = 0.3 1 e 3; %m
% T h i c k n e s s o f t h e p a p e r i n s u l a t i o n wrapped on t h e o u t e r w i n d i n g
t i n s o t = 0.3 1 e 3; %m

% Winding c o n d u c t o r d i m e n s i o n s
% Width o f i n n e r w i n d i n g c o n d u c t o r
w i n = 12 1 e 3; %m
% Thickness of outer winding conductor
t in = 2 1 e 3; %m
% Total t h i c k n e s s of the i n n e r conductor i n c l u d i n g i n s u l a t i o n t h i c k n e s s
cond in = w in t i n ;
% Inn er conductor area i n c l u d i n g paper i n s u l a t i o n
i n c o n d a r p= c o n d i n 2 t i n s i n ;

% Width o f o u t e r w i n d i n g c o n d u c t o r
w ot = 12 1 e 3; %m
% Thickness of outer winding conductor
t ot = 2 1 e 3; %m
% Total t h i c k n e s s of inner winding conductor i n c l u d i n g i n s u l a t i o n
c o n d o t = w ot t o t ;
% O ut e r w i n d i n g c o n d u c t o r a r e a i n c l u d i n g p a p e r i n s u l a t i o n

116
o t c o n d a r p= c o n d i n 2 t i n s i n ;

% Total t h i c k n e s s of i n n e r conductor i f used i n a bundle

% Number o f c o n d u c t o r s i n t h e b u n d l e i n i n n e r w i n d i n g
cond bundle p = 1;
% Number o f c o n d u c t o r s i n t h e b u n d l e o f o u t e r w i n d i n g
cond bundle s = 1;

% Core and i n s u l a t i o n c y l i n d e r d i m e n s i o n s
% E f f e c t i v e diameter of core
c o r e d i a = 120 1 e 3; %m
% Thickness of inner winding c y l i n d e r
c y l t h i c k i n = 40 1 e 3; %m
% Thickness of the outer winding c y l i n d e r
c y l t h i c k o t = 40 1 e 3; %m
% I n s u l a t i o n between c o r e and i n n e r w i n d i n g
i n s c o r e w i n d i n = 501 e3 %m

%I n s u l a t i o n d i m e n s i o n s on w i n d i n g end i n s u l a t i o n
% I n n e r w i n d i n g i n s u l a t i o n on t o p s i d e o f w i n d i n g
i n s w i n d t p i n = 60 1 e 3; %m
% I n n e r w i n d i n g i n s u l a t i o n on t h e bottom s i d e o f window
i n s w i n d b t i n = 60 1 e 3; %m
% O ut e r w i n d i n g i n s u l a t i o n on t h e t o p o f t h e w i n d i n g
i n s w i n d t p o t = 60 1 e 3; %m
% O ut e r w i n d i n g i n s u l a t i o n on t h e bottom o f t h e w i n d i n g
i n s w i n d b t o t = 60 1 e 3; %m

%I n n e r r a d i u s o f i n n e r w i n d i n g
r w i n d i n i n = c o r e d i a /2 + i n s c o r e w i n d i n + c y l t h i c k i n + r i b s t i n ;
% O ut e r r a d i u s o f i n n e r w i n d i n g
r w i n d i n o t = r w i n d i n i n + t u r n s i n ( t i n + 2 t i n s i n ) ;
% Inner r a d i u s of outer winding
r wind ot in = r wind in ot + cyl thick ot + ribs t ot ;
% O ut e r r a d i u s o f o u t e r w i n d i n g
r w i n d o t o t = r w i n d o t i n + t u r n s o t ( t o t + 2 t i n s o t ) ;
% Radial t h i c k n e s s of inner winding
wind t in = r wind in ot r wind in in ;
% Radial t h i c k n e s s of inner winding
wind t ot = r wind ot ot r wind ot in ;
% Mean r a d i u s o f i n n e r w i n d i n g
m e a n r w i n d i n = 1/2( r w i n d i n i n + r w i n d i n o t ) ;
% Mean r a d i u s o f o u t e r w i n d i n g
mean r wind ot = 1/2( r w i n d o t i n + r w i n d o t o t ) ;

117
% Mean l e n g t h o f i n n e r w i n d i n g t u r n
m e a n l e n g t h o f t u r n i n = m e a n r w i n d i n 2 p i ;
% Mean l e n g t h o f o u t e r w i n d i n g t u r n
m e a n l e n g t h o f t u r n o t = m e a n r w i n d o t 2 p i ;
% Mean l e n g t h o f w h o l e w i n d i n g t u r n
m e a n l e n g t h w i n d t u r n = 1 / 2 ( m e a n l e n g t h o f t u r n i n +m e a n l e n g t h o f t u r n o t ) ;

% R a d i a l Model INPUT DATA%

% Number o f a p p r o x i m a t i n g f u n c t i o n s
a p p r o x f u n =4;
% Mass d e n s i t y o f c o p p e r =8920 Kg/ s q m
m d e n s i t y =8920;
% Young s Modulus o f c o p p e r ( E l a s t i c i t y o f c o p p e r )
E=1.685 e10 ;
% S t i f f n e s s o f i n s u l a t i o n ( P a t e l s 100000 p s i c o n v e r t e d t o N/ s q m)
K=1e5 4 . 4 5 / ( 2 5 . 4 / 1 0 0 0 ) 2 ;

% I n t e r t u r n i n s u l a t i o n in inner winding
i n s i n t e r t u r n i n = c l a c k t i n + 2 t i n s i n ;
% I n t e r t u r n i n s u l a t i o n in outer winding
i n s i n t e r t u r n o t = c l a c k t o t + 2 t i n s o t ;

a r e a i n= t u r n s i n c o n d i n % xs e c t i o n a r e a o f i n n e r w i n d i n g d i s k
a r e a o t= t u r n s o t c o n d o t ; % xs e c t i o n a r e a o f o u t e r w i n d i n g d i s k

% I n p u t f o r t h e E l e c t r o m a g n e t i c f o r c e s Code ( e m f o r c e .m)
% f o r p r e d i c t i o n o f e l e c t r o m a g n e t i c f o r c e s u s i n g method o f I m a g e s

i i n = 1000; % current in inner winding

i o t = 1000; % current of outer winding
i o t = i o t ; % Changing s i g n o f t h e c u r r e n t
%( w i n d i n g p o l a r i t y )
Taw=20; % Taw=L/R= X / ( 2 . p i . R . f )
f =50; % F r e q u e n c y o f t h e s u p p l y ( 5 0 Hz )

%d i s t a n c e o f p r i m a r y w i n d i n g from t o p y o k e t o t h e c e n t e r o f 1 s t c o n d u c t o r
d i n t p = i n s w i n d t p i n + 0.5 w in + t i n s i n ;
% d i s t a n c e o f p r i m a r y w i n d i n g edge from bottom y o k e
d i n b t = i n s w i n d b t i n + 0.5 w in + t i n s i n ;
% d i s t a n c e o f s e c o n d a r y w i n d i n g edge from t o p y o k e
d o t t p = i n s w i n d t p o t + 0 . 5 w ot + t i n s o t ;
% d i s t a n c e o f s e c o n d a r y w i n d i n g edge from bottom y o k e
d o t b t = i n s w i n d t p o t + 0 . 5 w ot + t i n s o t ;
n i m a g e=5 ; % no o f i m a g e s

118
omega=2 p i f ; % omega

%CONSTANTS
C=210 7; % Constant of f o r c e

% FORCE ON PRIMARY DUE TO PRIMARY

% Under s i n u s o i d a l c u r r e n t s
F inr =[]; % F o r c e a s t i m e f u n c t i o n on i n n e r w i n d i n g
F otr =[]; % F o r c e a s f u n c t i o n o f t i m e on o u t e r w i n d i n g
time = [ ] ; % Time v e c t o r

f o r t =0: 0 . 0 0 1 : 0 . 1 ; % Duration of f a u l t
% Current as time f u n c t i o n
I i n = i i n d i s k s i n ( s i n ( omega tp i / 2 ) + exp( t / (Taw/omega ) ) ) ;
% Current as time f u n c t i o n
I o t =i o t d i s k s o t ( s i n ( omega tp i / 2 ) + exp( t / (Taw/omega ) ) ) ;
t i m e =[ t i m e t ] ; % Time m a t r i x
em force02 ; % em f o r c e c a l c u l a t i o n code
% Adding f o r c e on e a c h c o n d u c t o r due t o a l l i n n e r c o n d u c t o r s
F inr = [ F inr , F in ] ;
% Adding f o r c e on e a c h c o n d u c t o r due t o a l l o u t e r c o n d u c t o r s
F otr = [ F otr , F ot ] ;
end
F inr = F inr ;
F otr = F otr ;
T = time ;

ax mech02 % M e c h a n i c a l b e h a v i o u r c a l c u l a t i o n code i n a x i a l d i r e c t i o n
rd mech03 % M e c h a n i c a l b e h a v i o u r c a l c u l a t i o n code i n r a d i a l d i r e c t i o n
ang var % A n g l e d e f i n i n g code a l o n g t h e c i r c u m f e r e n c e o f t h e d i s k

s i g n =1

%Components f o r SIMULINK model

%Damping= damping c o e f f i c i e n t volume o f t h e d i s k
d i n =1750.03 a r e a i n m e a n l e n g t h o f t u r n i n 1000%0 ;
d o t =3;
d r = 3;

% unit conversion
stress strain ;
s t r e s s =175 s t r e s s ;
s t r a i n=s t r a i n ;
q =1;

119
mu = 0 . 5 ;

%PLOTTING
s u b p l o t ( 2 , 3 , 1 ) , p l o t ( time , r e a l ( F i n r ) )
grid
x l a b e l ( Time ( S ) )
y l a b e l ( F o r c e (N) )
t i t l e ( R a d i a l f o r c e on i n n e r w i n d i n g )

s u b p l o t ( 2 , 3 , 2 ) , p l o t ( time , imag ( F i n r ) )
grid
x l a b e l ( Time ( S ) )
y l a b e l ( F o r c e (N) )
t i t l e ( A x i a l f o r c e on i n n e r w i n d i n g )

s u b p l o t ( 2 , 3 , 4 ) , p l o t ( time , r e a l ( F o t r ) )
grid
x l a b e l ( Time ( S ) )
y l a b e l ( F o r c e (N) )
t i t l e ( R a d i a l f o r c e on o u t e r w i n d i n g )

s u b p l o t ( 2 , 3 , 5 ) , p l o t ( time , imag ( F o t r ) )
grid
x l a b e l ( Time ( S ) )
y l a b e l ( F o r c e (N) )
t i t l e ( A x i a l f o r c e on o u t e r w i n d i n g )

s u b p l o t ( 2 , 3 , 3 ) , bar ( a x n a t f r e )
grid
x l a b e l ( Number )
y l a b e l ( F r e q u e n c y ( Hz ) )
title ( Axial natural frequencies )

s u b p l o t ( 2 , 3 , 6 ) , bar ( r d n a t f r e )
grid
x l a b e l ( Number o f n a t u r a l f r e q u e n c y )
y l a b e l ( F r e q u e n c y ( Hz ) )
t i t l e ( Radial natural frequencies )

120
A.2 Electromagnetic force

% e m f o r c e 0 2 .m
% T h i s code c a l c u l a t e s t h e v a l u e o f t h e e l e c t r o m a g n e t i c f o r c e a p p l i e d on
% t h e e a c h d i s k / c o n d u c t o r o f w i n d i n g i n a two w i n d i n g t r a n s f o r m e r . The
% i n p u t t o t h i s code comes from t h e inp mod f i l e

% % % %FORCE ON PRIMARY DUE TO PRIMARY

% o v e r a l l a x i a l t h i c k n e s s of i n n e r winding d i s k i n c l u d i n g i n s u l a t i o n ( paper
% & clack )
w d i n = w i n + 2 t i n s i n + c l a c k t i n ;
% o v e r a l l a x i a l t h i c k n e s s of outer winding d i s k i n c l u d i n g i n s u l a t i o n ( paper
% & clack )
wd ot = w ot + 2 t i n s o t + c l a c k t o t ;
% t o t a l a x i a l l e n g t h o f t o t a l no o f d i s k s ( i n n e r )
l n i n = w d i n ( d i s k s i n 1);
% t o t a l a x i a l l e n g t h o f t o t a l no o f d i s k s ( o u t e r )
l n o t = wd ot ( d i s k s o t 1);
%
% d i s t a n c e t o t h e c e n t e r o f bottom d i s k ( i n n e r w i n d i n g )
d bt in = l n i n /2;
% d i s t a n c e to the c e n t e r of the top d i s k ( i n n e r winding )
d tp in = l n i n /2;
%
% d i s t a n c e t o t h e c e n t e r o f bottom d i s k ( o u t e r w i n d i n g )
d b t o t = l n o t /2;
% d i s t a n c e t o t h e c e n t e r o f bottom d i s k ( o u t e r w i n d i n g )
d tp ot = l n o t /2;
%C a l c u l a t i n g h e i g h t o f window
l w i n d o w = d i n t p + ( l n i n + w i n + 2 t i n s i n ) +d i n b t ;

% Position of inner winding

d i n ( 1 : d i s k s i n ) = d t p i n : w d i n : d b t i n ;
% Position matrix of inner winding
pos wind in = mean r wind in + d in j ;

% Position of outer winding

d o t ( 1 : d i s k s o t ) = d t p o t : wd ot : d b t o t ;
pos wind ot = mean r wind ot + d ot j ;

c i n=C m e a n l e n g t h w i n d t u r n I i n 2 ;
c o t=C m e a n l e n g t h w i n d t u r n I o t 2 ;
c i n o t=C m e a n l e n g t h w i n d t u r n I i n I o t ;

121
F in = zeros ( d i s k s i n , 1 ) ; % F o r c e on t h e i n n e r w i n d i n g
F ot = zeros ( d i s k s o t , 1 ) ; % F o r c e on t h e o u t e r w i n d i n g

%FORCE ON INNER WINDING DUE TO INNER WINDING%%%%%

mat 1=p o s w i n d i n ;
mat 2=p o s w i n d i n ;
d mtom ; % C a l c u l a t i n g d i s t a n c e from m a t r i x 1 t o m a t r i x 2
Cp=c i n ;
f mtom ; % C a l c u l a t i n g t h e f o r c e on e a c h i n n e r w i n d i n g t u r n ( f mtom )
F in = Force ; % T o t a l f o r c e on i n n e r w i n d i n g

% FORCE ON OUTER WINDING DUE TO OUTER WINDING %%%%

%D i s t a n c e { d b t o t } i n mm t o t h e c e n t e r o f t h e f a r most s e c o n d a r y
%c o n d u c t o r t o w a r d s t h e bottom

mat 1=p o s w i n d o t ;
mat 2=p o s w i n d o t ;
d mtom ;
Cp=c o t ;
f mtom ;
F ot = Force ;

% FORCE BETWEEN INNER AND OUTER %%%%

% On i n n e r w i n d i n g due t o o u t e r w i n d i n g%%%
mat 1=p o s w i n d i n ;
mat 2=p o s w i n d o t ;
d mtom ;
Cp=c i n o t ;
f mtom ;
F i n=F i n + F o r c e ; % Adding f o r c e t o t h e t o t a l f o r c e

%%% On o u t e r w i n d i n g due t o i n n e r w i n d i n g %%%

mat 1=p o s w i n d o t ;
mat 2=p o s w i n d i n ;
Cp=c i n o t ;
d mtom ;
f mtom ;
F o t= F o t + F o r c e ; % Adding f o r c e t o t h e t o t a l f o r c e

% FORCE ON THE INNER WINDING DUE TO ITS IMAGE POSITIONS %%%%

%F o r c e due t o t h e i m a g e s o f t o p p l a n e on i n n e r w i n d i n g
mat 1=p o s w i n d i n ;

122
 f o r a =1: n i m a g e ;
mat 2=p o s w i n d i n +( 1)( a +1)(2 d i n t p + l n i n+ ( a 1)2 l w i n d o w ) j ;
mat n2=1 r e a l ( mat 2)+ j imag ( mat 2 ) ;

%C a l c u l a t i o n s o f f o r c e s on
%i n n e r w i n d i n g
%d m m ;
d mtom ;
Cp=c i n ;
f mtom ;
%F i n = F i n + c o n j ( F o r c e ) ;
F in = F in + Force ;
%Fup1 = F o r c e ;
%f o r c e c a l c u l a t i o n i n ve x d i r e c t i o n
%d mtonx ;
%d m nx ;
d mtonx ;
f mtom ;
%F i n = F i n + c o n j ( F o r c e ) ;
F in = F in + Force ;
% %Fup2=F o r c e ;
end

% %F o r c e due t o i m a g e s o f bottom p l a n e

f o r a =1: n i m a g e ;
mat 1=p o s w i n d i n ;
mat 2=p o s w i n d i n (1)( a +1)(2 d i n b t +l n i n +(a 1)2 l w i n d o w ) j ;
mat n2=1 r e a l ( mat 2)+ j imag ( mat 2 ) ;
d mtom ;
%d m m ;
Cp=c i n ;
f mtom ;
%F i n = F i n + c o n j ( F o r c e ) ;
F in = F in + Force ;
%Fdn1= F o r c e ;
%f o r c e c a l c u l a t i o n i n ve x d i r e c t i o n
d mtonx ;
%d m nx ;
f mtom ;
%F i n= F i n + c o n j ( F o r c e ) ;
F in = F in + Force ;
%Fdn2 = F o r c e ;
end

123
%%%%%Y AXIS IMAGES OF OUTER WINDING

mat 1 = p o s w i n d o t ;

f o r a= 1 : n i m a g e ;
mat 2 = p o s w i n d o t +( 1)( a +1)(2 d o t t p + l n o t+ ( a 1)2 l w i n d o w ) j ;
mat n2=1 r e a l ( mat 2)+ j imag ( mat 2 ) ;

%C a l c u l a t i o n s o f f o r c e s due t o i m a g e s on
% secondary

d mtom ;
Cp=c o t ;
f mtom ;
F ot = F ot + Force ;

%f o r c e c a l c u l a t i o n i n ve x d i r e c t i o n
d mtonx ;
%d m nx ;
f mtom ;
F ot = F ot + Force ;
end

%F o r c e due t o i m a g e s o f bottom p l a n e
f o r a =1: n i m a g e ;
mat 1=p o s w i n d o t ;
mat 2=p o s w i n d o t ( 1)( a +1)(2 d o t b t + l n o t + ( a 1)2 l w i n d o w ) j ;
mat n2= 1 r e a l ( mat 2)+ j imag ( mat 2 ) ;
%d m m ;
d mtom ;
f mtom ;
F ot = F ot + Force ;

%f o r c e c a l c u l a t i o n i n ve x d i r e c t i o n
d mtonx ;
f mtom ;
F ot = F ot + Force ;
end

%F o r c e s on p r i m a r y due t o s e c o n d a r y i m a g e s
%Top i m a g e s
mat 1=p o s w i n d i n ;

124
 % inner winding
f o r a =1: n i m a g e ;
mat 2=p o s w i n d o t +( 1)( a +1)(2 d o t t p+ l n o t + ( a 1)2 l w i n d o w ) j ;
mat n2=1 r e a l ( mat 2)+ j imag ( mat 2 ) ;

% outer winding
d mtom ;
Cp=c i n o t ;
f mtom ;
F in = F in + Force ;
%f o r c e c a l c u l a t i o n i n ve x d i r e c t i o n
d mtonx ;
f mtom ;
F in = F in + Force ;
end
% Bottom i m a g e s
f o r a =1: n i m a g e ;
mat 1=p o s w i n d i n ;
mat 2=p o s w i n d o t (1)( a +1)(2 d o t b t + l n o t +(a 1)2 l w i n d o w ) j ;
mat n2= 1 r e a l ( mat 2)+ j imag ( mat 2 ) ;
d mtom ;
f mtom ;
F in = F in + Force ;

%%f o r c e c a l c u l a t i o n i n ve x d i r e c t i o n
d mtonx ;
f mtom ;
F in = F in + Force ;
end

%F o r c e s on s e c o n d a r y due t o p r i m a r y i m a g e s

mat 1=p o s w i n d o t ;

f o r a =1: n i m a g e ;
mat 2=p o s w i n d i n + ( 1)( a +1)(2 d i n t p + l n i n+ ( a 1)2 l w i n d o w ) j ;
mat n2=1 r e a l ( mat 2)+ j imag ( mat 2 ) ;

%C a l c u l a t i o n s o f
% primary
%d m m ;
d mtom ;
Cp=c i n o t ;
f mtom ;

125
 F ot = F ot + Force ;
%f o r c e c a l c u l a t i o n i n ve x d i r e c t i o n
d mtonx ;
%d m nx ;
f mtom ;
F o t = F o t +F o r c e ;

end

% %F o r c e on i n n e r w i n d i n g due t o i m a g e s o f bottom p l a n e

f o r a =1: n i m a g e ;
mat 1=p o s w i n d o t ;
mat 2=p o s w i n d i n ( 1)( a +1)(2 d i n b t+ l n i n +(a 1)2 l w i n d o w ) j ;
mat n2=1 r e a l ( mat 2)+ j imag ( mat 2 ) ;
%d m m ;
d mtom ;
Cp=c i n o t ;
f mtom ;
F ot = F ot + Force ;

%f o r c e c a l c u l a t i o n i n ve x d i r e c t i o n
%d m nx ;
d mtonx ;
f mtom ;
F ot = F ot + Force ;
end

% Xa x i s i m a g e s
% F o r c e s on i n n e r w i n d i n g due t o i t s own i m a g e s
mat 1=p o s w i n d i n ;
mat 2=1 r e a l ( p o s w i n d i n )+ j imag ( p o s w i n d i n ) ;

%C a l c u l a t i o n s o f f o r c e s on
% on i n n e r w i n d i n g due t o i n n e r w i n d i n g
%d m m ;
d mtom ;
Cp=c i n ;
f mtom ;
F in = F in + Force ;

% F o r c e on i n n e r w i n d i n g due t o o u t e r w i n d i n g
mat 2=1 r e a l ( p o s w i n d o t )+ j imag ( p o s w i n d o t ) ;
%d m m ;

126
 d mtom ;
Cp=c i n o t ;
f mtom ;
F in = F in + Force ;

% F o r c e on o u t e r w i n d i n g due t o i n n e r w i n d i n g
mat 1=p o s w i n d o t ;
mat 2=1 r e a l ( p o s w i n d i n )+ j imag ( p o s w i n d i n ) ;
%d m m ;
d mtom ;
Cp=c i n o t ;
f mtom ;
F ot = F ot + Force ;

% F o r c e on s e c o n d a r y due t o s e c o n d a r y

mat 2=1 r e a l ( p o s w i n d o t )+ j imag ( p o s w i n d o t ) ;

%d m m ;
d mtom ;
Cp=c o t ;
f mtom ;
F ot = F ot + Force ;

A.2.1 Smaller routines used in electromagnetic calculation

%d mtom , r o u t i n e t o c a l c u l a t e d i s t a n c e

clear dist ;
d i s t=z e r o s ( l e n g t h ( mat 1 ) , l e n g t h ( mat 2 ) ) ;
clear x ; clear y ;
f o r x =1: l e n g t h ( mat 1 ) ;
f o r y =1: l e n g t h ( mat 2 ) ;
d i s t ( x , y)=mat 1 ( x)mat 2 ( y ) ;
end
end

127
%d mtonx : c a l c u l a t i o n o f d i s t a n c e f o r x s i d e
d i s t=z e r o s ( l e n g t h ( mat 1 ) , l e n g t h ( mat n2 ) ) ;
clear x ; clear y ;
f o r x =1: l e n g t h ( mat 1 ) ;
f o r y =1: l e n g t h ( mat n2 ) ;
d i s t ( x , y)=mat 1 ( x ) mat n2 ( y ) ;
end
end

%F o r c e s between two w i n d i n g s

f o r c e=z e r o s ( s i z e ( d i s t ) ) ;

f o r x =1: l e n g t h ( mat 1 ) ;
f o r y =1: l e n g t h ( mat 2 ) ;

i f d i s t ( x , y )==0;
f o r c e ( x , y )=0;
e l s e i f d i s t ( x , y )=0;
f o r c e ( x , y)=Cp/ d i s t ( x , y ) ;
end
end
end
F o r c e=sum( f o r c e , 2 ) ;

A.3 Code related to axial dynamic behaviour

%%ax mech02A x i a l dynamic b e h a v i o u r r e l a t e d code

%% Code f o r c a l c u l a t i n g t h e n a t u r a l f r e q u e n c i e s o f p r i m a r y and s e c o n d a r y
%% w i n d i n g s . Only two w i n d i n g s a r e a n a l y z e d . The code a l s o g e n e r a t e s t h e
%% o u t p u t f o r s i m u l i n k model .

% D i s k mass ( t h e mass o f i n s u l a t i o n i s s u p p o s e d t o be 20% o f c o p p e r mass )

mass disk in = 1.2( mean lengthofturn in ) t u r n s i n cond in m density ;
mass disk ot = 1.2( mean lengthofturn ot ) turns ot cond ot m density ;

% A x i a l i n s u l a t i o n t h i c k n e s s ( between t h e d i s k s )
a x i n s t i n = c l a c k t i n +2 t i n s i n ;
a x i n s t o t = c l a c k t o t +2 t i n s o t ;

% Area o f t h e c l a c k s u p p o r t i n g i n n e r w i n d i n g , o n l y t h o s e r i b s a r e
% e n t e r e d h e r e which a r e used f o r c l a c k s
% The a r e a o f c l a c k ( a x i a l i n s u l a t i o n ) i n c o n t a c t w i t h t h e w i n d i n g

128
ax ins area in = wind t in clack w in ribs in ;
ax ins area ot = wind t ot clack w ot ribs ot ;

% Mass M a t r i x
d i a g m a s s i n = [ m a s s d i s k i n ones (1 , d i s k s i n ) ] ;
diag mass ot = [ m a s s d i s k o t ones (1 , d i s k s o t ) ] ;
d i a g m a s s a = [ d i a g m a s s i n , d i a g m a s s o t ] ; % d i a g o n a l o f a x i a l mass m a t r i x
mass a = diag ( d i a g m a s s a ) ; % mass m a t r i x ( a x i a l )

% S t i f f n e s s Matrix
% T h i c k n e s s o f i n s u l a t i o n between d i s k s
i n s t h i c k i n =[ i n s w i n d t p i n , c l a c k t i n o n e s ( 1 , ( d i s k s i n 1 ) ) . . .
, ins wind bt in ];
i n s t h i c k o t =[ i n s w i n d t p o t , c l a c k t o t o n e s ( 1 , ( d i s k s o t 1 ) ) . . .
, ins wind bt ot ];

% Diagonal of s t i f f n e s s matrix f o r inner winding

% E q u i v a l e n t s t i f f n e s s = K Area / L e n g t h o f i n s u l a t i o n
f o r a =1: d i s k s i n +1
i n s s t i f f i n (1 , a ) = K a x i n s a r e a i n / i n s t h i c k i n (1 , a ) ;
end

f o r a =1: d i s k s i n ;
d i a s t i f f i n ( 1 , a ) = i n s s t i f f i n ( 1 , a)+ i n s s t i f f i n ( 1 , a +1);
end

s u b d i a s t i f f i n = i n s s t i f f i n ( 2 : d i s k s i n ) ;%s u b d i a m e t e r o f i n n e r w i n d i n g
s t i f f i n a x = ( diag ( d i a s t i f f i n ) . . .
d i a g ( s u b d i a s t i f f i n ,1) d i a g ( s u b d i a s t i f f i n , 1 ) )

% S t i f f n e s s matrix of outer winding

f o r a =1: d i s k s o t +1
i n s s t i f f o t ( 1 , a)= K a x i n s a r e a o t / i n s t h i c k o t ( 1 , a ) ;
end

f o r a =1: d i s k s o t ;
d i a s t i f f o t ( 1 , a)= i n s s t i f f o t ( 1 , a)+ i n s s t i f f o t ( 1 , a +1);
end

dia stiff = [ dia stiff in , dia stiff ot ];

sub dia stiff ot = i n s s t i f f o t (2: disks ot ); % sub diameter outer

s t i f f o t a x = ( d i a g ( d i a s t i f f o t ) d i a g ( s u b d i a s t i f f o t , 1)...
d i a g ( s u b d i a s t i f f o t , 1 ) )

129
% Main d i a g o n a l o f s t i f f n e s s m a t r i x
dia stiff = [ dia stiff in , dia stiff ot ];
% Subd i a g o n a l o f s t i f f n e s s m a t r i x
s u b d i a s t i f f =[ s u b d i a s t i f f i n , 0 , sub dia stiff ot ];
s t i f f a x = ( d i a g ( d i a s t i f f ) d i a g ( s u b d i a s t i f f , 1)d i a g ( s u b d i a s t i f f , 1 ) ) ;

ax omega=e i g ( s t i f f a x / m a s s a ) ;
a x n a t f r e=s q r t ( ax omega ) / ( 2 p i ) ;

A.4 Code related to radial dynamic behaviour

% R a d i a l model u s i n g e x p a n s i o n o f r i n g s
% The f i l e i s same a s rd mech03 b u t t h e mass and s t i f f n e s s m a t r i x a r e
% corrected

%%%%R a d i a l Model I n p u t Data %%%%

a r e a i n= p i m e a n r w i n d i n 2 ; %r a d i a l a r e a o f i n n e r w i n d i n g d i s k
a r e a o t= p i m e a n r w i n d o t 2 ; %r a d i a l a r e a o f o u t e r w i n d i n g d i s k

x s e c i n = t u r n s i n i n c o n d a r p ; %c r o s s s e c t i o n a l a r e a o f i n n e r w i n d i n g
x s e c o t = t u r n s o t o t c o n d a r p ; %c r o s s s e c t i o n a l a r e a o f o u t e r w i n d i n g

% t o t a l area of each i n n e r r i b i n c o n t a c t with winding

r i b a r e a i n = r i b s w i n w in ;
% t o t a l area of each o u t e r r i b i n c o n t a c t with winding
r i b a r e a o t = r i b s w o t w ot ;

%%%Moment o f I n e r t i a o f i n n e r and o u t e r d i s k s
%%I n n e r d i s k
I i = t u r n s i n w i n t i n 3 / 1 2 ; % per conductor of inner winding
I o = t u r n s o t w ot t o t 3 / 1 2 ; % of outer winding

%%%Mass M a t r i x i n n e r w i n d i n g
%d i a g o n a l m a t r i x
d i a g 1 =[2 p i ] ;
f o r m=1: a p p r o x f u n ;
d i a g 1 =[ d i a g 1 ( ( 1 /m)2+1) p i ]
d i a g 1 =[ d i a g 1 ( ( 1 /m)2+1) p i ]
end
% mass o f i n n e r r i n g
diag m in = m density x s e c i n mean r wind in diag1 ;
% mass o f o u t e r r i n g

130
diag m ot = m density x s e c o t mean r wind in diag1 ;
d i a m a s s =[ d i a g m i n , d i a g m o t ] ;
m a s s r=d i a g ( d i a m a s s ) ; % mass m a t r i x ( i n c l u d i n g i n n e r and o u t e r w i n d i n g )

f o r x =1: l e n g t h ( d i a m a s s ) ;
m a s s r a d ( x , x ) = 1/ d i a m a s s ( x ) ;
end

%%P o t e n t i a l e n e r g y
%E l a s t i c S t i f f n e s s M a t r i x
%p o t e n t i a l e n e r g y due t o e x t e n s i o n o f i n n e r r i n g
u1 = a r e a i n E p i / m e a n r w i n d i n ;
%p o t e n t i a l e n e r g y due t o e x t e n s i o n o f o u t e r r i n g
U1 = a r e a o t E p i / m e a n r w i n d o t ;

f o r x =1: a p p r o x f u n
u2 ( x ) = ( x 2 1)2 p i ;
end

diag2 =[];
f o r x =1: a p p r o x f u n
d i a g 2= [ d i a g 2 , u2 ( 1 , x ) , u2 ( 1 , x ) ] ; % d i a g o n a l o f PE m a t r i x
end
%d i a i n = [ u1 , d i a g 1 ] ;% d i a g o n a l o f e l a s t i c s t i f f n e s s matrix for inner ring
%%%%INNER RING
%%% P o t e n t i a l E n e r g y due t o c o p p e r component o f w i n d i n g
s t i f f e l i n = E I i / m e a n r w i n d i n 3 d i a g 2 ;
% P . E due t o e l a s t i c b e h a v i o u r o f i n n e r w i n d i n g
s t i f f e l i n = [ u1 , s t i f f e l i n ] ;
s t i f f e l o t = E I o / m e a n r w i n d o t 3 d i a g 2 ;
% P . E due t o e l a s t i c b e h a v i o u r o f o u r w i n d i n g
s t i f f e l o t = [ U1 , s t i f f e l o t ] ;
s t i f f e l a s t= [ s t i f f e l i n , s t i f f e l o t ] ;

s t i f f e l a s t =d i a g ( s t i f f e l a s t ) ; % S t i f f due t o e l a s t i c p o t e n t i a l e n e r g y

%%% s t i f f n e s s due t o i n s u l a t i o n ( p r e s s b o a r d / p a p e r )

s t i f f 1 =[];
f o r t h e t a = 0 : ( 2 p i / r i b s i n ) : ( 2 pi (2 p i / r i b s i n ) )
f o r x =1: a p p r o x f u n
F1 ( 1 , x ) =t h e t a x ;
end

131
 s t i f f 0 =[];
f o r y =1: a p p r o x f u n
s t i f f 0 = [ s t i f f 0 , cos ( F1 ( 1 , y ) ) , s i n ( F1 ( 1 , y ) ) ] ;
end
s t i f f 1 =[ s t i f f 1 ; s t i f f 0 ] ;
end

s t i f f i n n e r=s t i f f 1 ;
u0 = o n e s ( l e n g t h ( s t i f f i n n e r ( : , 1 ) ) + 1 , 1 ) ;
u1 = z e r o s ( 1 , l e n g t h ( s t i f f i n n e r ( 1 , : ) ) ) ;
u1 = [ u1 ; s t i f f i n n e r ] ;
s t i f f i n n e r = [ u0 , u1 ] ; %s t i f f n e s s o f i n s u l a t i o n r i b s o f i n n e r r i n g

u= s t i f f i n n e r ;
% p o s i t i o n m a t r i x w i t h z e r o e n t r i e s ( i n c l u d e s i n n e r and o u t e r w i n d i n g )
p o s i n r = [ s t i f f i n n e r , z e r o s ( ( r i b s i n +1) , ( a p p r o x f u n 2 + 1 ) ) ] ;
% pressboard r i b s p o s i t i o n inner winding
pos in = pos inr pos inr ;
% s t i f f n e s s of the i n n e r winding f o r n a t u r a l f r e q u e n c i e s
s t i f f i n = K r i b a r e a i n r i b s i n 1/ r i b s t i n p o s i n r p o s i n r ;

%%%%OUTER RING

s t i f f 1 =[];
f o r t h e t a = 0 : ( 2 p i / r i b s o t ) : ( 2 pi (2 p i / r i b s o t ) ) ;
f o r x =1: a p p r o x f u n
F1 ( 1 , x ) =t h e t a x ;
end

s t i f f 0 =[];
f o r y =1: a p p r o x f u n
s t i f f 0 = [ s t i f f 0 , cos ( F1 ( 1 , y ) ) , s i n ( F1 ( 1 , y ) ) ] ;
end
s t i f f 1 =[ s t i f f 1 ; s t i f f 0 ] ;
end
%c o l 1 o f t h e r i b s p o s i t i o n m a t r i x o f o u t e r r i n g
u0 = o n e s ( l e n g t h ( s t i f f 1 ( : , 1 ) ) + 1 , 1 ) ;
%( p a r t i a l ) row 1 o f t h e r i b s p o s i t i o n m a t r i x o f o u t e r r i n g
u1 = z e r o s ( 1 , l e n g t h ( s t i f f 1 ( 1 , : ) ) ) ;
s t i f f o t r = [ u1 ; s t i f f 1 ] ;
s t i f f o t r = [ u0 , s t i f f o t r ] ;
% producing a matrix \ d e l t a containing the d i f f e r e n c e of displacement of
% outer r i b s
s t i f f o t r = [ s t i f f 1 , s t i f f 1 ] ;

132
s i z e 0=z e r o s ( r i b s o t , l e n g t h ( m a s s r ) a p p r o x f u n 4 ) ;
s t i f f o t =[ s i z e 0 , s t i f f o t r ] ;

%P o s i t i o n o f r i b s o f o u t e r w i n d i n g
pos otr = ( s t i f f o t ) ( s t i f f o t ); % o f s p r i n g p o s i t i o n m a t r i x 2
pos ot = pos otr ;
% assumed l i n e a r s t i f f n e s s o f w i n d i n g t o c a l c u l a t e n a t u r a l f r e q u e n c i e s
s t i f f o t = K r i b a r e a o t r i b s o t 1/ r i b s t o t p o s o t r ;

s t i f f= s t i f f i n +s t i f f o t +s t i f f e l a s t ;
s t i f f i n s = s t i f f i n +s t i f f o t ;

%R a d i a l p o s i t i o n o f i n s u l a t i o n f o r s i m u l i n k

p o s r = p o s i n +p o s o t ;

% V e c t o r s u & v and U & V

angl = 0;
f o r x =1: a p p r o x f u n
an gl e ( 1 , x ) = a n g l x ;
end
angle u =[1];
angle v =[1];
f o r y =1: a p p r o x f u n
a n g l e u = [ a n g l e u , cos ( an gle ( 1 , y ) ) , s i n ( angle ( 1 , y ) ) ] ;
% a n g l e v = [ a n g l e v , 2/ y s i n ( a n g l e ( 1 , y ) ) , 2/ y c o s ( a n g l e ( 1 , y ) ) ] ;
end

%%%%OUTPUTS

size ( mass r ) ;
size ( s t i f f );
rd omega = s q r t ( e i g ( s t i f f / m a s s r ) ) ;
r d n a t f r e=s o r t ( rd omega / ( 2 p i ) )
bar ( r d n a t f r e )

%%%

f i n = ( real ( F inr ));

f ot = ( real ( F otr ));

a0=z e r o s ( a p p r o x f u n 2 , l e n g t h ( r e a l ( f i n ) ) ) ;
a0=a0 ;

133
%F o r c e s on d i s k 1
F r 1 i n = [ f i n ( : , 1 ) , a0 ] ;
F r 1 o t = [ f o t ( : , 1 ) , a0 ] ;
%F r 1 i n = [ f i n ( 1 , : ) ; a0 ] ;
%F r 1 o t = [ f i n ( 1 , : ) ; a0 ] ;
F r = [ Fr1 in , Fr1 ot ] ;
%F o r c e s on d i s k 2

A.4.1 Code used in radial behaviour for position of displacement

% To p r o d u c e m a t r i x f o r p o s i t i o n
%I n n e r r i n g p o s i t i o n s where d i s p l a c e m e n t i s measured
%( i n t h e m i d d l e o i f a x i a l s p a c e r s )

ang1 = [ ] ;
%posn = 8
posn=a p p r o x f u n 2 ;
ang le = 2 p i / posn ;
f o r t h e t a = 0 : an gl e : ( 2 pi an gl e )
f o r x =1: a p p r o x f u n
F1 ( 1 , x ) = [ t h e t a x ] ;
end

ang0 = [ 1 ] ;
f o r y =1: a p p r o x f u n
ang0= [ ang0 , cos ( F1 ( 1 , y ) ) , s i n ( F1 ( 1 , y ) ) ] ;
end
ang1 =[ ang1 ; ang0 ] ;
end
r 1 = z e r o s ( 1 , l e n g t h ( ang1 ) ) ;
r 1 ( 1 , 1 ) =1;
ang1 =[ r 1 ; ang1 ] ;

%Out e r r i n g p o s i t i o n s where d o s p l a c e m e n t i s measured

ang2 = [ ] ;
posn2 = 8
ang le = 2 p i / posn2 ;
f o r t h e t a = 0 : an gl e : ( 2 pi an gle )
f o r x =1: a p p r o x f u n
F1 ( 1 , x ) = [ t h e t a x ] ;
end

ang0 = [ 1 ] ;
f o r y =1: a p p r o x f u n

134
 ang0= [ ang0 , cos ( F1 ( 1 , y ) ) , s i n ( F1 ( 1 , y ) ) ] ;
end
ang2 =[ ang2 ; ang0 ] ;
end
ang2 =[ r 1 ; ang2 ] ;

135
Appendix B

Simulink models

The models and their components (subsystems) are presented in this chapter.

Force [Time imag(F_otr(:,1))]

1 1
From -K-
s s xi11
Workspace6
disk_mass Intg 13 Intg 17
Out In
0
-K-
spring1 Constant1
Damp6

Out In
Force [Time imag(F_otr(:,2))]
spring13
1 1
From -K-
s s
Workspace7
disk_mass6 Intg 18 Intg 19

-K-

Damp7 Out In

spring17

Out In

spring16
Force [Time imag(F_otr(:,3))]
1 1
From -K-
s s
Workspace8
disk_mass7 Intg 22 Intg 23

-K-

Damp9 Out In

spring19

Out In

spring18

Force [Time imag(F_otr(:,4))]

1 1
From -K-
s s
Workspace9
disk_mass9 Intg 24 Intg 14

-K-

Damp10
Out In

spring21
Out In

spring20
Force [Time imag(F_otr(:,5))]
1 1
From -K-
s s
Workspace10
disk_mass10 Intg 16 Intg 15

-K-

Damp11 Out In

spring23

Out In

spring22
Force [Time imag(F_otr(:,6))]
1 1
From -K-
s s
Workspace11
disk_mass8 Intg 20 Intg 21
-K-

Damp8 Out In 0

spring2 Constant

Out In

spring15

Figure B.1: Model to simulate axial behaviour of test test transformer implemented
in Simulink

136
 ins_thick4

-K- -K-

ins_a5 K5
1 1
Out ins_thick2 In
Sign
Switch1
-1 -K- -1 -K-

-6 ins_a6 K6 -5

Figure B.2: Model of the spring used as subsystem in axial model of a winding (Fig
B.1)

-K-

1 rev Sign
Fa(t) 1 1
-K- 1
s s
disk_mass1 Intg 3 x(n)
Intg 4
2
ri multyp -K-
Damp1 Out In
3
spring3 x(n-1)

Out In
R 2 4
spring2 x(n+1)

Figure B.3: Axial model of one disk used as a subsystem in the combined Simulink
model (Fig B.4)

137
Figure B.4: The representation of the combined model of the test transformer im-
plemented in simulink

138
Figure B.5: Radial model of a set of inner and outer disks of the test transformer

139
Appendix C

Test transformer design and

dimensions

C.1 Test transformer design details

The details of the core and winding design are given in this appendix. The Figure C.1. The
winding winding design sheet is given in the Figure C.2 and the Figure C.3 is the winding sheet
of outer winding.

640

CORE 120

120

240 WINDINGS

120

Core diameter = 120 mm

Note: All dimensions in mm

Figure C.1: Dimensions of the test transformer

140
Figure C.2: Design sheet of the inner winding of the test transformer

141
Figure C.3: Design sheet of the inner winding of the test transformer

142