Damage Tolerance Design for Wing Components

101
Damage Tolerance Design Damage Tolerance Design for Wing Components Procedure Standardization Bernardo Vilhena Gavinho Lourenço Dissertação para obtenção do Grau de Mestre em Engenharia Aeroespacial Júri Presidente: Professor Fernando José Parracho Lau Orientador: Professor Filipe Szolnoky Ramos Pinto Cunha Co-orientador: Professor Luís Filipe Galrão dos Reis Vogais: Professor Pedro da Graça Tavares Álvares Serrão Professor Ricardo António Lamberto Duarte Cláudio Outubro de 2010

Transcript of Damage Tolerance Design for Wing Components

Damage Tolerance Design

Damage Tolerance Design for Wing Components – Procedure

Standardization

Bernardo Vilhena Gavinho Lourenço

Dissertação para obtenção do Grau de Mestre em

Engenharia Aeroespacial

Júri

Presidente: Professor Fernando José Parracho Lau

Orientador: Professor Filipe Szolnoky Ramos Pinto Cunha

Co-orientador: Professor Luís Filipe Galrão dos Reis

Vogais: Professor Pedro da Graça Tavares Álvares Serrão

Professor Ricardo António Lamberto Duarte Cláudio

Outubro de 2010

i

Acknowledgements

The completion of this thesis was only possible with the support and guidance of many people

to whom I wish to extend my acknowledgment.

First of all, I wish to thank professors Filipe Cunha and Luís Reis for their support, guidance

and availability throughout this thesis. I also wish to thank engineers Carlos Rodrigues and Rui Pereira

for their guidance and availability during my stay in OGMA. I also wish to acknowledge Dr. Wanhill,

from the Nationaal Lucht-en Ruimtevaartlaboratorium (NLR) in the Netherlands, who kindly suggested

some important documents on Initial Damage Characterization.

A special thanks to my family and friends for their outstanding support and patience, without

them it would have been impossible.

ii

iii

Abstract

Fatigue analysis of mechanical components, in many cases, leads to underestimated lives and

thus greater costs, so a different philosophy was developed, Damage Tolerance Design. Using this

theory, the designer no longer assumes a perfect component but rather the existence of an initial

damage that is allowed to propagate. However, that damage is detected and repaired within the safety

limits placed.

The purpose of this thesis will be the definition and standardization of procedures to be

followed by an aircraft contractor, OGMA, in order to implement a Damage Tolerant Design for the

components. Once the procedure is approved by the Aeronautical Authorities it will have significant

impact on costs reduction, without compromising the safety of the repairs.

The thesis is mostly focused towards wing structures, particularly riveted joints. To do so,

stress concentration factors, residual strength determination, crack growth analysis and inspection

intervals must be correctly defined, determined and analyzed. Furthermore, the initial damage on the

structure must also be correctly assumed, in terms of shape, size, direction and quantity.

Computational methods for determining some of these parameters are mandatory.

For the concretization of the created procedure, a wing panel was analyzed using Damage

Tolerance principles.

Keywords: Damage Tolerance, Crack Propagation, Riveted Joint, Inspection Chart.

iv

v

Resumo

Uma análise de fadiga aos componentes mecânicos conduz, em muitas situações, a vidas

subestimadas e consequentemente a maiores custos, sendo por isso adoptada uma filosofia

diferente, a Tolerância ao Dano. Usando esta teoria, o projectista já não assume um componente

perfeito, mas antes que existe um dano inicial que se vai propagar, sendo este detectado e reparado

dentro dos limites de segurança impostos.

Esta tese tem como objectivo a definição e uniformização de procedimentos a serem

seguidos por uma empresa de manutenção de aeronaves, OGMA, de modo a implementar uma

filosofia de Tolerância ao Dano. Quando aprovado pelas autoridades competentes, o manual dará um

importante contributo para redução de custos, sem comprometer a segurança das manutenções.

A tese foca-se principalmente sobre estruturas em asas, em particular juntas rebitadas. Para

tal, tem de ser feita uma análise, determinação e definição detalhada de concentração de tensões,

determinação de resistência residual, análise de propagação de fendas e intervalos de inspecção.

Mais, o dano inicial existente na estrutura também tem de ser correctamente assumido, em termos de

tamanho, forma, direcção e quantidade. Os métodos computacionais assumem grande relevância na

determinação de alguns destes parâmetros.

Para concretizar o procedimento criado, foi estudado um painel de asa, usando os princípios

de Tolerância ao Dano.

Palavras-chave: Tolerância ao Dano, Propagação de Fendas, Junta Rebitada, Carta de Inspecção.

vi

vii

Contents

Acknowledgements ................................................................................................................................... i

Abstract .................................................................................................................................................... iii

Resumo .................................................................................................................................................... v

Contents ................................................................................................................................................. vii

List of Figures .......................................................................................................................................... ix

List of Tables ........................................................................................................................................... xi

Abbreviations, Acronyms and Nomenclature ........................................................................................ xiii

1. Introduction ....................................................................................................................................... 1

2. Theoretical Background of Fundamental Concepts ......................................................................... 3

2.1. Damage Tolerance Design ....................................................................................................... 3

2.2. Fracture Mechanics Design ...................................................................................................... 4

2.2.1. Energy Methods................................................................................................................ 7

2.3. Fatigue ...................................................................................................................................... 9

2.3.1. S–N Curves .................................................................................................................... 11

2.3.2. Crack Growth Rate ......................................................................................................... 11

3. Airworthiness Requirements ........................................................................................................... 15

3.1. General ................................................................................................................................... 15

3.2. Fail Safe Evaluation ................................................................................................................ 16

3.3. Safe Life Evaluation ................................................................................................................ 18

3.4. Sonic Fatigue .......................................................................................................................... 18

3.5. Damage Tolerance Evaluation ............................................................................................... 18

4. Stress Concentration Factor ........................................................................................................... 21

4.1. Rivet – State of the Art ........................................................................................................... 21

4.2. Riveted Joints ......................................................................................................................... 23

4.3. Correlation Method ................................................................................................................. 25

4.3.1. System Construction and Definition ............................................................................... 26

4.4. Finite Element Method (FEM) ................................................................................................. 27

5. Initial Damage Characterization ..................................................................................................... 29

5.1. Non Destructive Inspection Methods ...................................................................................... 29

5.2. Initial Damage Size Assumption ............................................................................................. 32

5.3. Damage Shape and Direction ................................................................................................ 35

5.4. Damage Disposition ............................................................................................................... 36

5.5. Damage Quantity .................................................................................................................... 37

5.6. Damage Location ................................................................................................................... 38

6. Load Spectrum ............................................................................................................................... 39

6.1. Wing Spectrums ..................................................................................................................... 40

6.1.1. TWIST Spectrum ............................................................................................................ 40

6.1.2. FALSTAFF Spectrum ..................................................................................................... 40

viii

6.2. Normalized Spectrum ............................................................................................................. 41

6.2.1. Complete Spectrum Normalization ................................................................................. 42

6.2.2. Spectrum Normalization per Flight Phase ...................................................................... 43

7. Residual Strength ........................................................................................................................... 45

7.1. Residual Strength on Wing Skins ........................................................................................... 46

7.1.1. Simplifications and Assumptions .................................................................................... 46

7.1.2. Requirements Application ............................................................................................... 47

7.2. Residual Strength Determination ............................................................................................ 48

8. Crack Growth Analysis ................................................................................................................... 51

8.1. Crack Retardation ................................................................................................................... 51

8.2. Stress Intensity Factor ............................................................................................................ 53

8.2.1. Stress Intensity Factor Determination in Crack Growth Analysis ................................... 54

8.3. Crack Growth Rate Determination using an Analytical Procedure ......................................... 55

8.4. Crack Growth Rate Determination using AFGROW ............................................................... 57

9. Inspection Requirements ................................................................................................................ 61

9.1. Inspection Type and Crack Detection .................................................................................... 62

9.2. Scatter Factor ......................................................................................................................... 64

9.2.1. Scatter Factor for Complete Life ..................................................................................... 65

9.2.2. Scatter Factor after First Inspection ............................................................................... 65

9.3. Initial Inspection Requirement ................................................................................................ 66

9.4. Recurrent Inspection Requirement ......................................................................................... 66

10. Example of a Damage Tolerance Analysis of a Wing Panel .......................................................... 69

11. Concluding Remarks and Future Developments ........................................................................... 75

References ............................................................................................................................................. 77

Attachments ............................................................................................................................................ 81

Attachment 1. – Utilities for Stress Concentration Factor Calculi [6]...................................................... 81

Attachment 1.1. –Stress Concentration Factor for Bearing Stress .................................................... 81

Attachment 1.2. – Stress Concentration Factor for Bypass Gross Area Stress ................................. 81

Attachment 1.3. – Bearing Distribution Factor .................................................................................... 82

Attachment 2. – Variation of Crack Detection with the Inspection‟s Conditions [45].............................. 83

Attachment 2.1. – Prior information on crack location influence on crack detection .......................... 83

Attachment 2.2. – Structural area influence on crack detection ......................................................... 83

Attachment 2.3. – Crack location influence on crack detection .......................................................... 84

Attachment 2.4. – Surface condition influence on crack detection ..................................................... 84

Attachment 2.5. – Surface condition influence on crack detection ..................................................... 84

Attachment 3. – Characteristics of the Lockheed C-130A [49] [50] ....................................................... 85

ix

List of Figures

Figure 1 – a) Crack Growth; b) Residual Strength [4] ............................................................................. 4

Figure 2 – The Three Different Stress Modes ......................................................................................... 5

Figure 3 – Crack‟s Tip Tensile Field Model Description [1] ..................................................................... 5

Figure 4 – Plastic Zone [7] ....................................................................................................................... 6

Figure 5 – Energy Release Rate for Plain Strain Cases [5] .................................................................... 7

Figure 6 – Energy Release Rate for Plain Stress Cases [5] ................................................................... 8

Figure 7 – Fatigue Process [1] ................................................................................................................ 9

Figure 8 – Fatigue Striations [1] ............................................................................................................ 10

Figure 9 – a) S-N Curve; b) Goodman Diagram [1] .............................................................................. 11

Figure 10 – a) Crack Growth Rate versus Crack Length; b) Crack Growth Rate versus Stress Intensity

Factor Range [1] .................................................................................................................................... 12

Figure 11 – Crack Growth Rate Regions [1] ......................................................................................... 12

Figure 12 – Flight Envelope................................................................................................................... 15

Figure 13 – Example of Stress Concentration near a hole [1] .............................................................. 21

Figure 14 – Different Rivet Shape [21] .................................................................................................. 22

Figure 15 – a) Normal Row; b) Staggered Row .................................................................................... 23

Figure 16 – a) Doublers; b) Splices ....................................................................................................... 24

Figure 17 – Secondary Bending [1] ....................................................................................................... 24

Figure 18 – Splice Spring System ......................................................................................................... 26

Figure 19 – Doubler Spring System ...................................................................................................... 26

Figure 20 – Finite Element Model for Splices [23] ................................................................................ 28

Figure 21 – Crack lengths and quantity [4] ............................................................................................ 29

Figure 22 – Geometries for Cracks (to be used along with Table 2) [25] ............................................. 32

Figure 23 – Alternative Initial Damage Location for Countersunk Rivet Holes [28] .............................. 35

Figure 24 – Position of the crack around the fastener hole, for axial loading ....................................... 35

Figure 25 – Position of the crack around the fastener hole, for biaxial loading .................................... 36

Figure 26 – Multiple Site Damage Impact on Fatigue Life [28] ............................................................. 37

Figure 27 – Typical Manufacturing Hole Quality Damage [4] ................................................................ 38

Figure 28 – a) Normal Flight Mission; b) Typical Military Flight Mission ............................................... 39

Figure 29 – TWIST Spectrum: a) simplified; b) detailed ....................................................................... 40

Figure 30 – FALSTAFF Spectrum: a) simplified; b) detailed ................................................................. 41

Figure 31 – Normalized Spectrum per Flight Phase [1] ........................................................................ 42

Figure 32 – TWIST Spectrum after Normalization ................................................................................ 43

Figure 33 – Residual Strength Definition [4] .......................................................................................... 45

Figure 34 – Box Beam Section, Idealized using Ten Booms ................................................................ 46

Figure 35 – Wing Model for Residual Strength Determination .............................................................. 47

Figure 36 – Determination of in-section Loads...................................................................................... 49

Figure 37 – a) Cycles vs. Crack Growth; b) Stress Intensity Factor vs. Growth Rate [1] ..................... 51

x

Figure 38 – Crack Retardation Process [40] ......................................................................................... 52

Figure 39 – Crack Retardation Effects on Crack Growth [1] ................................................................. 52

Figure 40 – Superposition Method Example [1] .................................................................................... 53

Figure 41 – Fatigue Life Variation with the Loading Sequence [1]........................................................ 54

Figure 42 – Geometries Considered for Shape Factor Determination .................................................. 56

Figure 43 – Inspection‟s Detection Interval [43] .................................................................................... 61

Figure 44 – Inspections‟ Influence on Failure [44] ................................................................................ 61

Figure 45 – Crack Detection Capabilities for Different Inspection Methods [47] ................................... 63

Figure 46 – Crack Detection Interval [49] .............................................................................................. 63

Figure 47 – Scatter Factor – j2 [50]........................................................................................................ 65

Figure 48 – Recurrent Inspection Interval Definition [3] ........................................................................ 67

Figure 49 – Structure of the Lockheed C-130A Wing ........................................................................... 69

Figure 50 – Scheme for the Analyzed Wing (without Externally Mounted Probes) .............................. 69

Figure 51 – Crack Growth Prediction using AFGROW ......................................................................... 71

Figure 52 – Initial Damage Location Influence on the Panel Life .......................................................... 72

Figure 53 – Reparation to Conduct ....................................................................................................... 73

Figure 54 – Inspection Chart for Wing Root Damage ........................................................................... 73

xi

List of Tables

Table 1 – Minimum Distance between Rivets – [6] (in inches) ............................................................. 23

Table 2 – Minimum Detectable Crack Sizes – [23] (in inches) .............................................................. 31

Table 3 – Initial Crack Size Assumption – [4][22] (in inches) ................................................................ 33

Table 4 – Initial Crack Size Assumption – [25] (in inches) .................................................................... 34

Table 5 – Crack Disposition around the Hole ........................................................................................ 36

Table 6 – Number of Cracked Holes to be Introduced .......................................................................... 38

Table 7 – Spectrum‟s Main Properties .................................................................................................. 42

Table 8 – Spectrum‟s Properties per Flight Phase ................................................................................ 43

Table 9 – Geometric Models Built for Insertion on AFGROW ............................................................... 58

Table 10 – Visual Inspections‟ Characteristics [4] [43] [44] ................................................................... 62

Table 11 – Life Reduction Factors [44] ................................................................................................. 64

Table 12 – Damage and Residual Strength Data ................................................................................. 70

Table 13 – Results for Crack Growth Prediction using AFGROW ........................................................ 71

Table 14 – Inspection Requirements using a Detailed Visual Inspection ............................................. 74

xii

xiii

Abbreviations, Acronyms and Nomenclature

In this document, the Imperial System is used, as required by OGMA – Indústria Aeronáutica

de Portugal, S.A.

a - Crack Length

a*

- Acceleration

acrit - Critical Crack Length

a0 - Initial Damage Length

A - Area

B - Thickness

B*

- Boom Area

C - Paris Growth Law Constant

C*

- Fastener Spring Constant

C’ - Compliance

CF - Forman Growth Law Constant

CP - Priddle Growth Law Constant

CS - Certification Specifications

(EASA Regulations)

d - Distance Between Two Booms

D - Diameter

da/dN - Crack Growth Rate

DTD - Damage Tolerance Design

E - Young‟s Modulus of Elasticity

EASA - European Aviation Safety Agency

f - Fraction of the Load

Absorbed by a Doubler

FAA - Federal Aviation Administration

FAR - Federal Aviation Regulations

(FAA Regulations)

FEM - Finite Element Method

Fg - Flight Profile Alleviation Factor

g - Acceleration of Gravity

(assumed as 32.174 ft/s2)

G - Energy Release Rate

h - Altitude

hceiling - Maximum Operational

Altitude (Ceiling)

H - Distance Parallel to the Aircraft‟s

Flight Path until Gust Peak Velocity

H(Ω) - Frequency Response Function

I - Moment of Inertia

I* - Inspection Interval

j1 - Inspection Scatter Factor

for Complete Life

j2 - Inspection Scatter Factor after

the first Inspection

K - Spring Constant for a Plate

KI - Stress Intensity Factor for mode I (IC

– critical value – fracture toughness)

Kt - Stress Concentration Factor

Ktb - Stress Concentration Factor

for Bearing Stress

Ktg - Stress Concentration Factor for

Bypass Gross Area Stress

K1, K2,

K3, K4, -

Life Reduction Factors for Scatter

Factor Determination

L - Length

L* - Scale of Turbulence

Lfast - Length of the Fastener

m - Paris Growth Law Exponent

M - Moment

mF - Forman Growth Law Exponent

mP - Priddle Growth Law Exponent

MSD - Multiple Site Damage

MTOW - Maximum Take-Off Weight

mwing - Wing Mass

n - Load Factor

N - Number of Cycles

NDE Non-Destructive Evaluation

NDI - Non-Destructive Inspection

p - Pressure

P - Load

P1g - Load for Normal 1g Flight

POD - Probability of Detection

xiv

q - Shear Flow Distribution

R - Stress Ratio

R* - Crack Growth Resistance

s - Distance Penetrated in the Gust

S - In Plane Loads Applied in a Beam

Cross-section

t - Thickness

T - Tension vector

U - Elastic Energy

Udes - Design Gust Velocity

Uref - Reference Gust Velocity

USAF - United States Air Force

Uσ - Limit Turbulence Intensity

V - Velocity

VA - Maneuvering Velocity

VC - Cruise Velocity

VD - Dive Velocity

VMC - Calibrated Air Velocity

w - Plate Width

W - Weight

W *

- Crack Formation Energy

β - Shape Factor

ΔKth - Stress Intensity Factor Range in

Threshold Region

θ - Bearing Distribution Factor

θ*

Angle with the crack direction

μ - Friction Factor

ν - Poison Coefficient

ξ - Rigid Body Damping Factor

σ - Stress

σa - Alternating Stress

σmax - Maximum Stress

σm - Mean Stress

σmin - Minimum Stress

σres - Residual Strength

σys - Yield Stress

τ - Shear Stress

Ω - Reduced Frequency

1

1. Introduction

The purpose of this thesis was the development of a Damage Tolerance Design procedure, to

be used by OGMA – Indústria Aeronáutica de Portugal, S.A., one of the main aircraft maintenance

contractors in Portugal. Located in Alverca do Ribatejo, near Lisbon, the company was founded in

1918 and currently is owned by a consortium led by EADS and Embraer, with a workforce of about

1600 employees. The most important roles played by OGMA are concerned with engine and structural

maintenance and its most important maintenance contracts include several aircrafts, such as the

Lockheed C-130, P-3 Orion, F-16 Fighting Falcon, Embraer ERJ 145 family, Airbus A320 family,

among others.

Safety is the major concern for all aircraft related subjects, since the design stage through the

entire duration of the service life. Aircraft maintenance and modification must be made using certified

procedures, which guarantee safety. This procedure will allow the company to make quick estimations

on the damage tolerance capability of a structure, in accordance with EASA specifications. The thesis

will emphasize riveted joints of wings.

The certification of the procedure must be requested by OGMA to the responsible authorities,

such as EASA, which will execute a thorough evaluation in order to guarantee the safety of the

procedures there inscribed.

Since the end of the Second World War, with the increasing operational life of the aircrafts and

the introduction of jet engines, fatigue related failures started to occur. As a consequence, the study of

varied themes, like Fracture Mechanics, Fatigue, Residual Strength, Stress Concentration and Stress

Intensity Factors, was intensified. The understanding of these subjects is crucial for a good

comprehension of Damage Tolerant Design principles.

Furthermore, Damage Tolerance calculation has a great dependence on crack growth

estimations and initial crack assumptions which will display a major role during this thesis, requiring

the use of computational methods, such as FEM and AFGROW, for its determination.

Therefore, using the procedure proposed and its outputs, safe ways to ensure damage

tolerant capabilities can be applied to newly designed or repaired structures. Furthermore the costs

will be reduced, which is the primary objective for every company, as the procedure will allow the

calculus of more accurate inspection charts, assuring that damage won‟t lead to the catastrophic

failure of the structure.

2

3

2. Theoretical Background of Fundamental Concepts

2.1. Damage Tolerance Design

Fatigue related cracks are the greatest responsible for component collapse in engineering.

Thus, methods to prevent fatigue cracks from evolving to catastrophic failure where developed [1].

Safe Life was one of the first methods to be used, and it states that once a component

reaches a specified number of cycles it is replaced with a new one. This method only takes into

account fatigue life issues, and has severe economic implications, as a component will be used for a

number of cycles inferior to the one it can withstand. Nowadays, it‟s only used in critical components

of an aircraft, such as the landing gear [2].

In order to obtain a safe but economically viable component, a different approach was needed,

and thus, in the 70‟s, Damage Tolerance Design (DTD) was created.

Damage Tolerance Design is a relatively recent philosophy in structural design, usually

described as “the ability of aircraft structure to sustain anticipated loads in the presence of fatigue,

corrosion or accidental damage until such damage is detected through inspections or malfunctions

and repaired” [3].

Using DTD, the structural engineer no longer assumes a perfect structural part, like for a safe

life component, but rather assumes that the new part already has a defect that will eventually evolve

leading to the catastrophic failure of said component [3] [4].

At first sight, this theory might seem too conservative, but analysis of in service fractures and

cracking instances have indicated that a major source of cracks is the occurrence of initial

manufacturing defects, as well as service induced damage (like corrosion). As so, the consideration of

initial damage in the form of cracks or equivalent damage is absolutely necessary to ensure structural

safety.

This gives DTD a great advantage over fatigue life tests, as a much more realistic component

life prediction will be obtained, thus allowing a much more time accurate inspections‟ program.

Damage tolerant structures can be divided into two major groups:

1. Slow Crack Growth;

This category includes all types of structures, single and multiple load paths which are

designed such that initial damage will grow at a stable, slow rate and does not achieve a size large

enough to fail the structure for a specified slow crack growth period. Safety is assured by the slow rate

of growth [4].

2. Fail Safe;

Usually are structures comprised of multiple elements or load paths such that damage can be

safely contained by a failing load path or by the arrestment of a rapidly running crack at a tear strap or

other deliberate design feature [4].

Fail safe structures must meet specific residual strength requirements following the failure of

the load path or the arrestment of a running crack. Safety will be assured by the allowance of a partial

failure of the structure, the residual strength and a period of usage during which the partial failure will

4

be found.

Following this philosophy, and in order to achieve safe structures, there are several aspects

that should be taken into consideration, such as [5]:

Obtaining the residual strength as function of the crack dimension;

The allowed crack dimension;

The crack growth time;

The dimension of the pre-existent crack permitted in the structure;

The time gap between inspections, replacements or proof testing.

Some of the properties referred will be subject of analysis in the subsequent sections, but

through them, there are two types of graph that can be built, providing the needed information to

proceed with a damage tolerant design, in accordance with the specifications. Figure 1 presents

typical graphs that describe these properties.

Figure 1 – a) Crack Growth; b) Residual Strength [4]

In order to obtain an economically viable solution for the component, one has to ensure long

inspection periods and/or long replacement intervals. To do so the designer has several different tools

and has to take them into account for a damage tolerant design [6].

Using a different material, with better mechanical properties, with a handicap related to

the increased cost;

Improved inspection procedure;

Redesign in order to lower the stresses, as the effect of stress on crack growth is

enormous;

Use of redundant systems, in order to prevent catastrophic component failure.

2.2. Fracture Mechanics Design

Fracture mechanics has always been an important study theme in Engineering, in order to

prevent catastrophic structural failures. Its two biggest branches are Linear-Elastic and Elastic-Plastic.

This work will be mostly focused on Linear-Elastic Fracture Mechanics, which is simpler and produces

good and accurate results. One of its most important study themes is related with cracks, their sizes,

a) b)

5

shapes and growth rates. Using damage tolerance design, the existence of a crack, with

characteristics that will allow propagation, is assumed [5] [7].

A crack in a component can be stressed in three different modes, described in Figure 2, being

mode I the most critical.

Figure 2 – The Three Different Stress Modes

As so, for mode I, the tensile field on the crack tip can be described by equations 1 to 4, for

the simple case of an infinite plate model with a remote applied stress [7]:

Where, σ is the applied stress, r is the distance between the crack tip and an arbitrary

element, θ* is the angle in respect to the crack plane and a is the semi-crack length. Figure 3

illustrates the situation described.

Figure 3 – Crack’s Tip Tensile Field Model Description [1]

The equations represent the first term of a series, and therefore give a good approximation of

6

the crack‟s tip tensile field. As so, they can be written as:

The parameter KI is the stress intensity factor for mode I. It‟s an essentially elastic concept

that gives an indication of the stress intensity and severity near the crack‟s tip. This factor and its

variation are used to describe fatigue crack growth resistance of materials. Also the limiting value, KIC

is a property of the material, occurring when the crack reaches its critical size, known as Fracture

Toughness [1].

The critical stress intensity factor cannot be affected by different crack or component

geometries, once its limiting value defines a material property. As so the shape factor, β, is introduced,

dependent of the component, crack geometry, crack length and load applied. [7]

For an infinite plate, the shape factor will have a value near the unit, as the crack will have

neglectable size compared to the component.

For different geometries, a different β factor will be used. For simpler geometries, this factor

can be found on tables and graphs, in reference [8], but for complex geometries, its determination is

difficult and is made recurring to computational methods.

As the distance to the crack‟s tip decreases, the tensions will rise, but, as it‟s impossible to

reach an infinite value, at the location where the yield stress is reached, plastic deformation will begin

to occur, thus maintaining the stress level at the yield stress value. This distance defines the crack‟s

tip plastic zone. Figure 4 presents the plastic zone.

Figure 4 – Plastic Zone [7]

This plastic zone is very important because it behaves like a part of the crack, as Irwin noticed

[5] [9]. The region‟s length was later corrected from rp to rp* that is doubled from the previous one. This

happened because the uncorrected plastic zone didn‟t carry out the entire load applied, thus an

7

increased plastic zone would be needed [5].

The stress intensity factor determines what happens near the crack tip, particularly inside the

plastic zone, as different cracks with the same KI have similar tensile fields. This has important

consequences on crack retardation and will be discussed later on this thesis [7].

2.2.1. Energy Methods

Crack growth can occur if the system is able to provide the required energy to form an addition

to the crack length, da. This is called the Energy Release Rate, G, and can be calculated from the

elastic energy, U, and the crack resistance, R*, can be calculated from the crack formation energy, W

[5] [7].

G R*

As so, it is possible to obtain the expressions for this parameter, G, for both plain strain and

stress, considering only mode I:

Where KI is the stress intensity factor for mode I, E is the Young Modulus and υ is the Poisson

coefficient. In addition, if the component is under effects from more than one mode, the energy release

rate will be equal to the sum of the energy release of each mode.

For metals in plane strain cases, crack tip plasticity promotes the crack growth, and energy is

used to expand the plastic zone, thus propagating the crack. As the plastic energy, R*, required is the

same for every increment in the crack size; components fail for the same value of G. Figure 5 clearly

illustrates the critical value for Energy Release Rate.

Figure 5 – Energy Release Rate for Plain Strain Cases [5]

8

Unlike for plane strain, for plane stress it must be taken into account that the crack resistance,

R*, varies with the crack growth rate. As so, and as Figure 6 illustrates, the method to obtain the

critical energy release rate is different and varies according to the crack size and applied stress. This

method is known as the R-curve.

Figure 6 – Energy Release Rate for Plain Stress Cases [5]

It is important to notice that the stress must vary with the crack growth; otherwise this case

would be similar to plain strain.

There are also other ways to determine the critical energy release rate, such as the

compliance method. The compliance is defined as:

Other important method is the J integral. It has particular relevance if the crack‟s tip plasticity

effects cannot be ignored. As the energy release rate is defined through the elastic field, plasticity will

have an effect that G won‟t account for. Thus, the J integral is introduced, providing means to obtain

results in plasticity conditions, but with limitations [7].

Through the theorem of energy conservation, Eshelby, [7] [10], achieved the definition of J,

stated through equation 12, and with this it can be shown that J = 0 for any contour, Γ. Thus, J is the

potential energy variation for a virtual crack extension, where T is the tension vector.

9

As so, if the material behaviour is linear elastic, J = G, as expected; but in other cases a

solution is also possible. Therefore the J integral method is a more universal failure criterion.

For non linear materials the method presents limitations, as it considers component unloading

linear, fact that won‟t lead to an accurate value on the recovery energy. In spite of this limitation, this

method accounts for plasticity effects [7].

2.3. Fatigue

As mentioned, fatigue is the primary collapse reason for structural components, as so its study

and comprehension is of great importance.

This phenomenon was first acknowledged in the 19th century as a consequence of railroad

accidents, and ever since is one the most studied subjects in Engineering, based on concepts from

fracture mechanics.

Fatigue is the consequence of cyclic loads, and has two important and distinct phases, crack

initiation and crack growth, as shown in Figure 7. Usually this phenomenon starts from the material

surface, as in one side there is no material, which leads to a weaker constrain to the material grain. As

a consequence only some material grains are under plasticity effects and this can occur for lower

stresses, below the yield stress. Combined with surface roughness, this explains why fatigue cracks

are usually originated in the surface, being crack initiation a material surface phenomenon [1].

Cyclic slipCrack

Nucleation

Micro Crack

Growth

Macro Crack

GrowthFinal Failure

Crack Initiation Period Crack Growth Period

Stress Concentration Factor Kt

Stress Intensity

Factor KIKIC

Figure 7 – Fatigue Process [1]

The presence of a crack will cause a faulty stress distribution, leading to stress concentration

on the crack‟s tip, thus promoting crack growth and propagation. This propagation is only dependent

on the material, not on the surface conditions.

The transition between crack initiation and crack growth happens as soon as micro-crack

growth doesn‟t depend on the surface conditions. In most of the cases, crack initiation phase

represents a significant part of the fatigue life of a component.

Fatigue originated cracks usually propagate in a direction perpendicular to the main stress, in

mode I.

As was stated before, fatigue is responsible for most of the failures in mechanical components

and its effects on material surface are easy to notice, as they appear as striations, usually similar to

load cycle applied on the component. As so, detailed microscopic analysis may be used to obtain the

10

detailed loading history of the component. Figure 8 is a good example on visible fatigue striations.

Figure 8 – Fatigue Striations [1]

There are two important fatigue regimes: low and high cycle. The difference is not clearly

defined by a certain number of cycles, but mostly for the reasons described in the next paragraphs.

In low cycle fatigue, the component‟s life is of about 104 cycles, and it‟s associated with

macroplastic deformations in every cycle.

As for high cycle fatigue, the relation is more similar to an elastic process, and therefore the

component‟s life is extended to over 105 cycles. This case has more practical applications. Also stress

amplitude has more effect on fatigue than the mean stress, mostly for high fatigue life components.

To predict component failure due to fatigue, several methods are used, but the Palmgren-

-Miner Linear Cumulative Fatigue Damage Theory, or simply the Miner‟s Rule, is the most common,

due to its simplicity and good results. It states that a component‟s capability to absorb damage is

cumulative, as so, expression 14 defines the rule:

For different stress levels, n is the number of cycles that the component carried out and N is

the maximum number of cycles that the component can withstand, until failure, and is defined using

the material‟s S-N curve. For design purposes this sum is often chosen as 1, but its value can vary

from 0.61 to 1.45.

This rule is a very good first approximation, due to its simplicity, but has several issues that

contribute to make it inaccurate. One of them is that the Miner‟s Rule ignores the fatigue contribution

of cycles below the fatigue limit to crack growth, as N would be infinite, but these cycles could

contribute to increase an existing damage [1].

Another fault is its inability to account for the order of the cycles applied on the component.

Tests have proved that the order in which the loads are applied has a significant importance on fatigue

life. For example, it is expected that a Hi-Lo loading sequence will induce a significantly greater fatigue

life to a component than a Lo-Hi loading sequence [1].

11

2.3.1. S–N Curves

In order to predict the fatigue life of a material a diagram must be determined, called S-N

curve or Wöhler curve, in honour of one of the pioneers in fatigue tests. The line drawn in the graph

indicates critical stress level and number of cycles for a material to fail due to fatigue effects [1].

This curve is determined by fixating a stress level, and applying it in a cyclic slip until rupture

occurs or more than 10 million cycles are achieved. If this high number of cycles is achieved, the

material is said to have infinite life for the imposed stress level; this number of cycles defines the flat

zone of the curve, where fatigue life is infinite.

These curves are a material characteristic, and its determination takes a lot of time, as for

each stress level, at least two specimens are required to validate the result. Also the fatigue life

depends on the values of the mean stress and on the stress concentration factor, which establishes a

relationship between the fatigue life and the stress concentration factor that will be discussed in

Chapter 4. Still, a great compilation of S-N curves, for different material, stress ratios and stress

concentration factors, has already been created – Metallic Material Properties Development and

Standardization (MMPDS) [1] [11] [12].

The values stipulated in this curve are the fatigue limits of a material, which are used in the

Miner‟s Rule, N, representing the maximum damage allowed for the applied stress. Additionally, they

indicate the maximum expectable service period for a Safe Life component.

It is also possible to convert the S-N curve into a stresses diagram that takes into account the

effect of the mean stress, through the fixation of the number of cycles. This diagram is known as the

Goodman diagram. In Figure 9, it is possible to observe the relationship between the two diagrams.

Figure 9 – a) S-N Curve; b) Goodman Diagram [1]

2.3.2. Crack Growth Rate

During the study of the fatigue phenomenon it is very important to clearly define and determine

the crack growth rate, da/dN. The evolution of this parameter with the crack length is of great

significance. Still, for different stresses these curves are different, but Paris, [1] [13], noticed an

overlap between the several plots, and therefore proposed a different solution, plotting the range of

12

the stress intensity factor, ΔK, with the growth rate [1]. The overlap noticed by Paris is illustrated in

Figure 10.

Figure 10 – a) Crack Growth Rate versus Crack Length; b) Crack Growth Rate versus Stress Intensity Factor Range [1]

With this information, and a logarithmic plot, it‟s possible to notice that there are three different

regions, limited by two asymptotes, one where Kmax = KIC, indicating the stress intensity for fracture

and other to indicate a value of KI below which no propagation occurs. These regions are clearly

visible in Figure 11.

Figure 11 – Crack Growth Rate Regions [1]

The first region is known as the threshold region and experimental studies have demonstrated

that ΔKth is not a material defined constant, but rather dependent on the ratio between minimum and

maximum stresses, R. This region is associated with macro crack growth to values above ΔKth. If a

crack slows its growth rate when ΔK applied decreases it‟s assumed that this value for ΔK is below

ΔKth and so the region is limited.

Knowing this, one could state that any crack growth would occur for values of ΔK inferior to

ΔKth, but this is not necessarily true, as there is no account for material interruption that it‟s much more

important than micro cracks generated in the material. Thus, an extrapolation from the Paris region is

a) b)

13

used to define crack growth rates for low values of ΔK, like is shown in Figure 11 [1].

The second region, known as the Paris region, describes a linear relation, in the logarithmic

plot, between the stress intensity factor range and the growth rate.

The third region is defined by high growth rates, leading to a fast final rupture of the

component. Normally this rupture occurs for the critical value of the stress intensity factor. However,

for very ductile materials, rupture occurs due to plasticity yielding, where the stress intensity factor is

meaningless, as it was defined as an elastic concept.

Several equations were introduced to define the relationship between the stress intensity

factor and the crack growth rate. The first one was developed by Paris and Erdogan, and accurately

defines the second growth region (Paris region) and is given by [14] [15]:

C and m are the Paris material constants.

As there were many limitations to this formula, other ones where proposed, that took into

account the effect of stress ratio, R, and could include the asymptotes already mentioned.

Forman proposed equation 16 to account for the stress ratio, R, and also include the

asymptote for KIC, using the material constants mF and CF [1] [14] [16]:

Priddle, [1] [14], suggested an alternative solution that included both asymptotes, being the

value of ΔKth determined through a formula proposed by Klesnil and Lukáš.

Once again A, CP, mP and γ are tabulated material constants.

Several other models exist, with increasing complexity, mostly used for computer software like

AFGROW or NASGRO. The material constants are different for each growth law, even if they cover

the same data set for the material. As so the use of the parameters of one equation on another may

lead to dramatic errors [5] [14].

It is also important to notice that all of these formulas are built in order to adapt to trends, and

have no physical value, being experimental testing the only way to accurately determine stress

intensity factor versus crack growth rate [1].

14

15

3. Airworthiness Requirements

In order to develop a damage tolerant structure, certain laws and specifications must be met.

All aircraft related laws and authority requirements can be found in FAR and CS documents; FAR for

the USA, determined by the FAA and CS for Europe, regulated by EASA. For any aircraft with a

maximum take-off weight (MTOW) higher than 12500 lbs, the documents needed to check are FAR-25

or CS-25. [17] [18]

Damage Tolerance requirements appear in sections FAR-25.571 and CS-25.571, where the

same goals are stated in both documents to achieve a damage tolerant design. Some subchapters of

this section may require the consult of other subsections in these documents.

The following subchapters state the CS-25 requirements. These requirements will define the

limiting loads that can be applied to the structure, in order to obtain a safe operation. Figure 12

represents the flight envelope, and can be used to facilitate the identification of the airspeeds that will

be referred to in this chapter.

Figure 12 – Flight Envelope

3.1. General

All evaluations must show that catastrophic failure induced by fatigue, corrosion or accidental

damage won‟t happen during the aircraft‟s expected lifetime. Alongside with subchapters, 3.2, 3.4, and

3.5, the following is applied:

16

a) Evaluations must include;

i. Load spectra, temperature and humidity in flight conditions;

ii. Identification of principal structural elements, with detailed design;

iii. Detailed analysis and testing of the elements defined in 3.1.a) i..

b) Historical data from similar aircrafts, with alterations due to flight conditions;

c) Inspections must be established to prevent catastrophic failure.

3.2. Fail Safe Evaluation

The evaluation must include the determination of location and damage mode due to fatigue,

corrosion and accidental damage. This must be supported by testing, for either static and repeated

loads, or service experience. Locations exposed to prior fatigue must be included if damage is

expected to occur. For residual strength evaluation, damage extension must be consistent with the

initial detection and growth rates. Also, the evaluation must prove that the structure resists to static

ultimate loads in these conditions:

a) Normal flight envelope conditions:

Additionally

b) Gust envelope conditions:

The limit load for gust conditions has three different approaches; one for a single gust,

other for continuous turbulence and other for a gust pair.

i. For a single gust, dynamic analysis of every structural part is needed, considering

a gust with the following shape:

17

ii. In the case of continuous turbulence, the limit load is given by:

If the velocity is VD, Uσ is half of the upper value

iii. For a gust pair, one vertical and one lateral the following is applied:

Being LV and LL the loads induced by the vertical and lateral gusts, respectively,

determined using 3.2.b) i.

c) For roll maneuvering:

i. a load factor of 2/3 of the design load factor in normal flight conditions is used;

ii. Loads resulting from engine failure due to fuel flow interruption are considered

limit loads;

iii. Loads resulting from engine failure due to turbine blade loss or disconnection

between compressor and turbine are considered ultimate loads;

iv. For horizontal tails in slipstream cases are considered maximum loadings from

symmetrical conditions, plus vertical gust conditions on one side and 80% of this

load on the other side.

d) For yaw maneuvering, limited pilot force on rudder deflection of:

i. 1335 N for speeds between VMC and VA;

ii. 890 N for speeds between VC and VD;

iii. Linear variation between the upper values.

e) For pressurized fuselages, the worst case is chosen:

18

f) For landing gears:

a) A maximum vertical load factor of 1.2 for the design weight.

b) During taxi the structure is assumed to be operating under the worst ground

expected, as so, the structure is considered to be under a load factor of 2 for one

axle gears and 1.7 for multi axle gears;

3.3. Safe Life Evaluation

Compliance with subchapter 3.2 is not needed if the structure application is proven to be

unpractical. This structure must be the subject of accurate analysis and testing, proving it able to

withstand the loads expected during the predictable lifetime.

3.4. Sonic Fatigue

Analysis supported on testing or similar aircraft service history must show that:

a) Sonic fatigue cracks are not probable to appear in any structural part subjected to sonic

excitation;

b) Catastrophic failure of any structural part doesn‟t occur assuming loads as prescribed in

subchapter 3.2 applied in crack affected areas.

3.5. Damage Tolerance Evaluation

The structure must be capable of finishing a flight where structural damage occurs due to:

a) Bird strike;

The aircraft must withstand the impact of a 4 lbs bird if the relative velocity between the

aircraft and the bird doesn‟t surpass a critical value for velocity:

i. The cruise speed at sea level;

ii. 85% of the cruise speed for 8000 ft.

b) Sudden decompression of any compartment;

Every component must be designed to withstand a sudden loss of pressure, at any flight

altitude, resulting from such conditions as:

i. Engine part penetration, resulting from engine disintegration;

ii. The maximum opening caused by airplane or equipment failure that are not

extremely improbable;

iii. Any opening to the size of H0:

The maximum cross section of the pressurized compartment is given by As.

19

The fail safe features of the design should be used to determine the probability of failure

or penetration and possible opening dimensions. Improper operations on closure

devices and inadvertent door openings must be also considered. Any loads created by

the depressurization must be considered ultimate loads and be, rational and

conservatively, combined with 1g flight loads.

The main difference between FAR and CS documents, in respect to Damage Tolerance, is

that CS-25 documents make a clear definition of continued turbulence, unlike FAR-25.

Also CS-25 is more specific on gust pairs, and thus should be used to achieve a more reliable

result in such conditions, especially since this condition will most likely represent the worst case

scenario.

These reasons justify the presentation of CS-25 requirements instead of FAR-25. Even more,

this document is to be used by a European company, under EASA regulations.

20

21

4. Stress Concentration Factor

Mechanical components should maintain constant section, or its alteration should be very

smooth, otherwise stress concentration will occur. The stress concentration factor is an essentially

elastic and dimensionless parameter that relates the nominal applied tension and the local tension [1]

[19]:

Stress concentration factors, Kt, can be determined by theoretical formulas, testing or

computational methods. The index near K stands for theoretical, because it is determined recurring to

the elastic theory. Figure 13 is a good example for stress concentration and its malignant effects.

Usually, theoretical modes of obtaining Kt are through the Elastic Theory, computational ones

through the Finite Element Method and testing is made using photo-elasticity or strain gauges. There

are only few examples for which a theoretical solution exists, and it is always a complex one [1].

Figure 13 – Example of Stress Concentration near a hole [1]

This parameter is also very important for crack initiation and propagation, as a crack is

normally formed due to stress concentration on the micro crack‟s tip. As so, designers should try to

avoid stress concentration on the components in order to prevent fatigue.

Additionally, geometrically similar components have the same Kt, but different stress gradients

will be found in the two components. This occurs because the stress concentration factor is a

dimensionless parameter. As a consequence, the biggest components will have higher areas and

volumes where there will be highly stressed material, thus contributing to an increase in fatigue

effects. This is known as the fatigue size effect.

Stress concentration around fastener holes is one of the most critical aspects leading to

fatigue in aircrafts. As so, its comprehension and determination is of great importance.

4.1. Rivet – State of the Art

Connectivity and load transfer between structural components is an important topic during

design. Several design features are used, such as rivets, bolts – nuts, welding or adhesive bonding.

For the Aerospace Industry, weight is one of the biggest issues. Rivets are much lighter than

22

conventional bolts; in the other hand its installation process can be more complex and thus expensive,

unlike bolting. Also, riveted structures cannot be disassembled without destroying the fastener.

Although adhesive bonding or welding would be the lightest, the quality of the adhesive or of the weld

is a big concern and so, for safety issues, they are not used [1].

Together with adhesive bonding, rivets are one of the most ancient joining techniques known

to Mankind. They are used since Ancient Greece, as a method to join bronze parts [20] [21].

Riveting was the most important joining technique until the appearance of welding in the 19th

Century. As welding techniques produce modifications in the material atomic structure, unexpected

behaviours in fatigue processes often occur. Thus, nowadays the fields and applications for rivets

have been increasing, being the Aerospace Industry the greatest responsible.

Most rivets are installed in a predrilled hole and the tail section (opposite to the head) is then

deformed until reaching 1.5 times its original diameter to assure the rivet stays in place. In some

cases, in order to increase the joining capabilities, the rivet is inserted at a high temperature, so that

after cooling, the length reduction induced compresses the plates. The process to remove a rivet is

irreversible, as the rivet must be cut off in order to disassembly the component. This is a significant

flaw of this technique versus bolting, and introduces additional costs.

There are many different kinds of rivets in the market, Figure 14, each with different properties

and purposes. The most common and important types are:

Solid rivet – the simplest ones, consisting of a head and shaft that is deformed with a hammer

or rivet gun, in order to fasten the plates.

Tubular rivet – similar to the solid rivet, but with a partial hole at the tip that allows a reduction

in the amount of force needed to deform it.

Blind rivet – it‟s comparatively new, its patent was registered in 1916, and its main advantage

is that access is only needed from one side. This type of rivet is similar to the tubular but also

supplied with a central mandrel that is pulled into the rivet, expanding the tip and assuring the

connection; after that, the mandrel snaps off. In the down side, this type of rivet should not be

used in critical structural components, as the mandrel may fall, leaving a hollow rivet that won‟t

transfer the same load as a solid rivet. Also, issues due to corrosion and vibration are more

likely to appear.

The shape of the rivet is also an important issue. For the Aerospace Industry a rivet with a

proeminent head cannot be used in the outer part of a fuselage and wing, as it would increase drag.

Figure 14 – Different Rivet Shape [21]

23

As so, countersunk rivets are the primary choice. For this rivet shape, the hole will have to

accommodate the head, thus increasing the production costs. In spite of this fact, some aircraft use

button head rivets in the rear part of the fuselage because the airflow has so much turbulence that this

type of rivet wouldn‟t influence the aerodynamical behavouir of the structure.

Rivets can be made from several metals and different alloys. The most commonly used are

made from steel or aluminium, depending on the application and its design purpose. More recently

titanium rivets started to be introduced in aerospace applications due to its reduced weight and

superior mechanical properties. However, such fasteners still have an increased cost which leads to a

reduced market penetration.

Rivet placement on the structure is also a major concern for a designer, as the distances to

use between the rivets are limited by specifications; most of them are contained in ASTM Standards.

Therefore, the most important distances are between rivets and to the margin of the plate. Also rivets

placed in normal parallel rows have different behaviours than staggered ones (Figure 15). The

distance between the rivet and the margin is an important parameter as well, because it affects stress

distribution in the cross section of the plate where the first row of rivets is installed. These parameters

are mainly defined through the fastener diameter and can be found in tables, such as Table 1 [6].

Fastener Diameter Pattern A B C

Normal Rows 0.63 0.55 _______ 0.34

Staggered Rows 1.0 0.39 0.63 0.34

Normal Rows 0.75 0.66 _______ 0.41

Staggered Rows 1.18 0.47 0.75 0.41

Normal Rows 1.0 0.9 _______ 0.53

Staggered Rows 1.56 0.63 1.0 0.53

Table 1 – Minimum Distance between Rivets – [6] (in inches)

Figure 15 – a) Normal Row; b) Staggered Row

4.2. Riveted Joints

Fatigue in fastener holes is responsible for 50 to 90% of aircrafts‟ components failures.

Therefore, in the interest of an accurate fatigue life calculus, stress concentration determination is very

a) b)

24

important in joints [22].

There are two important types of components that would need to be riveted to an aircraft skin

panel, splices and doublers (see Figure 16). Splices allow load transfer between components, as

doublers pick up load in order to relieve stresses on other component [23].

Usually, in doublers the first row absorbs most of the load, as for splices the first and final rows

absorb most of the load. This has impact in the stress concentration factors of these rows that are

often the most critical ones.

Figure 16 – a) Doublers; b) Splices

As so, for these structures two types of loads must be considered. Transfer loads, which are

loads transmitted through the fastener. Bypass load, on the other hand, is the load still carried by the

skin after a row of fasteners. Both these loads will contribute to determine the stress concentration

factor, thus allowing accurate calculations of the fatigue life [2].

Using γ as the percentage of bypass load, the stress concentration can be obtained using:

This concept is valid for both butt and lap joints. Nevertheless, lap joints have other important

issues to fatigue life, as the secondary bending effect, represented by the third factor in equation 38.

Secondary bending is originated by the misalignment in the forces applied due to thickness and its

non symmetrical effects (eccentricity) – Figure 17. Thus, the bending factor, k, is defined, as a ratio

between the bending stress and the tension stress [1].

Figure 17 – Secondary Bending [1]

a)

b)

25

An alternative formula to obtain Kt, used in this thesis, will be discussed in detail in subchapter

4.3.

There are features that can be used to increase fatigue life in joints. One of them is pre-

tension of the fastener, which will lead to better results on fatigue. Other method is substituting

fasteners for adhesive; it can lead to longer life, as this type of connection eliminates fretting corrosion

and stress concentration. However, durability and quality of the adhesive material must be taken into

account. In spite of these methods, the best way to improve fatigue life is avoiding stress

concentration areas during the design.

Additionally, the effect of Kt on the joint can be minored through the introduction of steps,

tapering, in the connection area that will allow a more balanced load distribution, thus leading to lower

stress concentration factors [23].

Also, crack detection is an important issue, as most of the times visual inspection is impossible

without disassembly. Pre-tension often produces a shift in the crack initiation location from the hole to

the contact point of both plates [1].

In order to determine the stress concentration factor for riveted joints there are two important

methods: an adaptation of the FEM and a correlation developed by Tom Swift in 1990. Both of these

methods must take into account fastener deflection and bearing, and also plate bearing [2] [6] [23].

4.3. Correlation Method

The correlation method relates test data through equations and is proven to have good results.

It models both fasteners and plates with springs, and through a displacement analysis the bypass and

transfer loads are determined. In order to do so, Swift proposed equation 39 to describe the fastener

spring constant: [2]

One can observe that this formula has many limitations, such as the materials that can be

used for both fasteners and plates. Other formulas, improved and more accurate, to determine the

spring constant are used by Boeing and Airbus, but they are protected by copyright and thus cannot

be used by others. In spite of that, Swift‟s formula provides a conservative value for this constant, and

thus safety is assured [2].

For the plates (skin and splice/doubler), using the normal force equation, where A is the cross

sectional area, the spring constant is determined through:

26

4.3.1. System Construction and Definition

Knowing the fastener constant, C*, and using for the panel springs, K, it is possible to build a

matrix that relates fastener and plate displacement, and thus create a system of equations where the

only unknowns are the transfer loads in each fastener. This matrix is generated through the spring

system displacement analysis.

It is now important to differentiate splices from doubler, as they serve different purposes they

have different behaviours. This way, the systems that describe them will also be distinct. Thus,

considering n fasteners, the systems will be modelled as Figures 17 and 18 illustrate, whether a splice

or a doubler.

Figure 18 – Splice Spring System

Figure 19 – Doubler Spring System

The spring constant values are calculated as previously defined and f is the fraction of the load

that the doubler absorbs, in order to relieve the skin; its value is obtained using:

Then for n fasteners, the systems‟ transfer loads (ΔP) can be calculated using equation

systems 42 and 43 (first system for splice [sp] and second for doubler [d]):

27

The transfer loads, ΔP, calculated using systems 42 and 43, appear in a geometrically

coherent order, and will allow a simple and easy determination for the bypass load, P, through the

total load applied. As a consequence, one can obtain the maximum stress knowing the transfer and

bypass loads, using [2] [23]:

The parameters Ktb, Ktg and θ can be obtained by graphics or formulas through the

geometrical data of the plates and hole. Some of these factors account for the bearing effects

referenced earlier, and the graphs that allow the determination of each one are presented in

Attachment 1.

Knowing the maximum tension, Kt can be determined by its usual formula, the ratio between

the maximum and the nominal stress, for every fastener. The most important value is obviously the

higher, and its location is the most critical for fatigue life purposes [2].

4.4. Finite Element Method (FEM)

A different way to determine the transfer loads and thus the stress concentration factors is

through a FEM approach. Unlike the previous method, that could have slow calculations due to the

system size, FEM produces good results; it is also very simple and fast to implement. Its use is made

by modelling a fastener as a circular beam, cantilevered in both extremities, with a diameter given by:

[2]

The fastener constant, C*, is determined using Swift‟s equation (39), and the length of the

28

fastener is assumed unitary in order to improve calculations.

The construction of the finite element model is very simple, the fasteners are modelled as

circular cantilevered beams and linked to each other by springs, with a spring constant equal to the

one presented previously for plates (eq. 40). This way, it‟s possible to determine the bypass and

transfer loads and so, the stress concentration factor, using the same procedure that was described in

subchapter 4.3.

Figure 20 illustrates a model for splices. For doublers some obvious and simple alterations

would be needed, like an application of a total constrain after the last fastener row, both in the skin

and doubler. As a consequence, in a FEM analysis, the software automatically determines the

percentage of load that the doubler will absorb.

Figure 20 – Finite Element Model for Splices [23]

29

5. Initial Damage Characterization

As stated previously, DTD assumes that the structure has initial defects. These are assumed

to be located in the most critical area, in the most critical quantity, and in the most critical direction

(according to the load applied). Figure 21 illustrates cracks‟ lengths that can be found in a component

and its quantity. Also, one can confirm that DTD is not as conservative as it might seem, and although

most of the damage occurs during fabrication or is inherent to material, a great part will not grow and

lead to the component‟s catastrophic failure [4] [24].

Figure 21 – Crack lengths and quantity [4]

First of all, one has to characterize and define the methods used to determine the crack

length, their capabilities, advantages and disadvantages. This will set the minimum detectable value

for the crack‟s length, in accordance with the non destructive method (NDI) used – aNDI. Yet, it is

expected that the initial damage has an inferior value impeding its detection by NDI methods. As so,

this value must be assumed.

5.1. Non Destructive Inspection Methods

Understanding the mechanisms of crack initiation and propagation is very important, but crack

detection and the methods used also have an important role, in order to avoid complete component

failure [5] [25] [26].

There are two main ways to proceed to crack detection, destructive methods or non

destructive method. For almost all industries only non destructive methods have relevance, as they do

not imply the component‟s destruction.

There are several types of non destructive methods, but they can be divided into two main

groups, direct and indirect [5].

30

1. Direct Methods

1.1. Visual

Visual inspections aided with lights, magnifying glasses and mirrors. Only possible for

components which are easy to access, performed by experienced technicians so that

small cracks aren‟t missed;

1.2. Liquid Penetrant

Coloured liquids are brushed on the component and are allowed to penetrate into

cracks. The liquid is then washed off and chalk is applied, revealing the crack location

and shape. It is only possible for easy access components, and has a similar failure rate

as visual inspections;

1.3. Magnetic particle

A fluorescent liquid with iron particles covers the component, which is then subjected to

an intense electromagnetic field. Cracks will disturb the magnetic field lines. It is a very

precise method, only optional for magnetic materials, but implies disassemble of the

component;

1.4. X-ray (Radiography)

Using a portable X-ray machine the component is inspected. Cracks absorb less

radiation than the surrounding areas, thus becoming delineated in the film. This method

is very versatile and efficient. On the other hand, it is expensive and time consuming.

2. Indirect Methods

2.1. Ultrasonic

A probe device sends a high frequency wave through the material. Upon contact with

the defect, that might not be a crack, the wave is reflected and a receiver is able to

determine its position. Although the position is determined, crack size and shape cannot

be obtained. This method has applications on every component due to different wave

impulses and probes that can be used, and is one of the most widely used;

2.2. Eddy current

Using a coil to induce Eddy currents on a metallic component, one can verify its

integrity, as cracks alter the induction from the metal on the coil. This is a cheap

method, with good results when done by trained technicians. Although almost every

component can be examined, as coils can be made of different sizes, information on the

defect‟s nature and size cannot be obtained;

2.3. Acoustic emission

Measures of the intensity of the stress waves emitted due to plasticity effects on crack‟s

propagation can be obtained. This method allows continuum surveillance on a loaded

component, but it‟s rather expensive and hard to implement, due to difficulties on proper

signal reading.

31

Although all these methods are applied in today‟s industry, naked eye inspection still

represents the greatest percentage of inspections made, usually assisted with magnifying glasses and

lanterns, as it‟s the most economical one. Still, visual inspections have zero relevance when dealing

with initial damage detection because the crack is too small to be detected, even by experienced

technicians. The use of this kind of inspection will be discussed in detail further on, when dealing with

in-service inspections, in chapter 9.

It is also important to notice that there are more methods than the ones referenced, but these

ones are the most important and more used, and consequently more economical.

Table 2 indicates, for the most common NDI methods, the damage‟s size that a particular

method is able to determine, according to the type of crack indicated in Figure 22. This will allow a

correct choice of the method to employ for the most expectable crack type [25].

NDI Method Crack

Location

Component

Thickness Crack Type

Crack

Dimension, a

Crack

Dimension, c

Eddy Current

Open

surface

t≤0.050

t≥0.050

Through

Partly Through

t

0.020

0.050

0.050

0.100

0.050

Edge or

Hole

t≤0.075

t≥0.075

Through

Corner

t

0.075

0.100

0.075

Liquid

Penetrant

Open

surface

t≤0.050

0.05<t<0.075

t≥0.075

Through

Through

Partly Through

t

t

0.025

0.075

0.100

0.150 – t

0.125

0.075

Edge or

Hole

t≤0.100

t≥0.100

Through

Corner

t

0.100

0.150

0.150

Magnetic

Particle

Open

surface

t≤0.075

t≥0.075

Through

Partly Through

t

0.038

0.075

0.125

0.188

0.125

Edge or

Hole

t≤0.075

t≥0.075

Through

Corner

t

0.075

0.250

0.250

Radiography Open

surface

t≤0.107

t≥0.107

Partly Through

Partly Through

Embedded

0.7 t

0.7 t

0.35 t

0.075

0.7 t

0.7 t

Ultrasonic Open

surface

t≥0.100 Partly Through

Embedded

0.030

0.065

0.017

0.039

0.150

0.065

0.087

0.039

Table 2 – Minimum Detectable Crack Sizes – [25] (in inches)

32

Figure 22 – Geometries for Cracks (to be used along with Table 2) [25]

In spite of the limits defined in Table 2, no inspection program will have a 100% detection rate.

So, a statistical procedure is adopted, where the probability of detection (POD) is the main concern,

with a confidence level placed on the estimate of the crack size, usually from 90 to 95%.

5.2. Initial Damage Size Assumption

As mentioned earlier, DTD assumes that the component already possesses an initial damage,

a0, such as a crack, but the length of said defect must be assumed, as NDI methods able to determine

it would have increased costs.

Such assumption must distinguish the position of the crack, whether it‟s near a hole or on an

open surface. The values to be assumed were determined by the USAF, using highly accurate NDI

methods solely developed for this process, and are not used in industry mostly due to economic

reasons [4] [24].

Different solutions and assumptions can be adopted. The main aircraft manufacturers, Boeing

and Airbus, assume different values for the initial flaw, but their data are protected by copyright issues,

and are not accessible.

Table 3 indicates the initial crack dimension to be assumed, for metallic damage tolerant

structures (a different crack shape can be used if the stress intensity factor remains constant).

33

Holes and Cutouts

Open Surface

Table 3 – Initial Crack Size Assumption – [4][24] (in inches)

Still, other bibliography suggests an alternate compilation of values for the initial damage,

where a stronger differentiation between Slow Crack Growth and Fail Safe Structures is made. This

data is presented in Table 4 [27].

Table 4 provides the most varied results, in accordance with the type of structure. Therefore, it

should be the designer‟s primary choice, even more for Fail Safe structures, where the initial damage

assumed is much smaller than for Slow Crack Growth structures, ultimately leading to longer service

lives.

Yet, smaller initial primary damage sizes may be used if a successful NDI demonstration is

conducted for the case. It must be demonstrated that the method has a POD of 90% for a confidence

level of 95%. Also, if proof tests are made and the calculated critical crack size is smaller than the

ones tabulated, it may also be assumed [27].

34

Slow Crack

Growth Structures

Holes and

Cutouts

Open Surface

Fail Safe

Structures

Holes and

Cutouts

Open Surface

Table 4 – Initial Crack Size Assumption – [27] (in inches)

In addition, it‟s important to notice that, for holes, the cracks usually begin in the faying

surfaces, which means that this surface condition, along with rivet hole quality, are very important.

Also, for countersunk rivet holes, that have a more complex shape, a different position of the crack

35

may be adopted, like described in Figure 23. This has particular impact on aerospace components

where this type of rivet is most commonly used [28].

Figure 23 – Alternative Initial Damage Location for Countersunk Rivet Holes [28]

Still the value for the initial damage length, corner crack radius, is assumed using tables 3 and

4.

5.3. Damage Shape and Direction

Although the assumption of the crack size is an extremely important issue, so is the growth

direction of the crack around the fastener hole and its shape. In damage tolerance, the worst case

scenario is always adopted, as so, the determination of this direction and the worst crack shape for

different kinds of loadings must be specified.

Assuming the coordinate system of Figure 24, as well as the remote applied loading, for the

crack location, several works demonstrate that the impact of the defect, on Kt and thus fatigue life,

only starts for angles superior to approximately 55 degrees. Still, the worst crack direction is obtained

for a 90 degree angle [22].

Figure 24 – Position of the crack around the fastener hole, for axial loading

Assuming a different layout for the loading, such as the one in Figure 25, a shift in the worst

initial position of a crack happens. The worst direction is no longer at a 90 degree angle, but rather at

45 degrees. This situation is consistent with a shear-like system, like the case described [1].

36

Figure 25 – Position of the crack around the fastener hole, for biaxial loading

For holes, like for an open surface, it was already expected that the worst crack growth

direction would be perpendicular to the direction of the applied loads.

In terms of shape, circular cavity-like initial damage induce the highest stress concentration

factor versus corner/through cracks, as this case resembles a superposition of notches, where the

final value of the stress concentration factor is obtained by multiplying the values of Kt for both

notches. Yet, this cracking instance occurrence is very rare [1] [22].

5.4. Damage Disposition

In order to correctly introduce initial damage for the damage tolerance analysis, one has to

define the disposition of this damage, with particular emphasis towards fastener holes.

Cracks can have different dispositions around the hole, as described in Table 5, which

indicates the most common cracking instances, obtained through experience.

Half of the initial damage size assumed

for each side of the hole

All the damage assumed on one side of

the hole

Half of the damage adjacent to the hole

and the other half in the plate border

Initial damage size assumed on one side

and half of it on the other

Table 5 – Crack Disposition around the Hole

37

Obviously, the last hypothesis is the most conservative one, as the initial damage assumed is

greater than the one stipulated in the tables 3 and 4. Yet, the most common cracking instance on

holes is the first one described.

Studies in order to determine which disposition is the most critical must be made and which is

most likely to occur. Still, the direction in which the crack grows is defined as presented in the previous

subsection.

5.5. Damage Quantity

The quantity of cracks to be introduced has to be correctly defined, in order to ensure safety,

but without creating an overestimated structure.

So, another important concept to introduce is Multiple Site Damage (MSD). This phenomenon

occurs when dealing with multiple cracks on a row of fasteners and is of vital importance as it shortens

the fatigue life of the structure, as these cracks are assumed to be growing in multiples holes of the

structure, and will eventually link-up, leading to an early failure. Figure 26 indicates the impact of this

concept on the fatigue life of a component.

Figure 26 – Multiple Site Damage Impact on Fatigue Life [28]

So, a division must be made between „new‟ holes and „old‟ holes before applying MSD. A new

hole is made on a new plate that was never drilled before, as old holes are the ones still in use after

service time and an inspection. It is natural to assume that old holes are more likely to have small

undetectable cracks, with lengths as stipulated in the initial damage tables – Table 3 or Table 4 [29].

The cracks to be introduced must have the length, direction, shape and dispositions

prescribed in this chapter that will lead to a worst case scenario. Furthermore, for new holes, one

crack should be introduced every 10 holes drilled, which is a conservative approach, based on

experience. It accounts for accidental manufacture errors and malpractice, and like most fatigue

related aspects is made recurring to a statistical approach. This data is summarized in Table 6.

38

New Holes Introduction of one cracked hole

for every 10 new holes drilled

Old Holes Introduction of cracks in

every hole

Table 6 – Number of Cracked Holes to be Introduced

Furthermore, for new holes, it is assumed that every hole, where no primary initial flaw was

introduced, posseses a corner damage of 0.005 in, as shown in Figure 27, to account for the typical

manufacturing hole quality. Even more, the interference between this damage and the primary must

be properly determined. [4]

Figure 27 – Typical Manufacturing Hole Quality Damage [4]

5.6. Damage Location

The initial damage must be introduced in the worst location, in order to obtain a worst case

scenario, ensuring safety.

As mentioned in the previous chapter, for the most common systems, the first and last rows of

a splice absorb most of the load, thus being the most critical area, in terms of tension. For doublers,

the same applies for the first row, which absorbs most of the load.

Still, the area where the maximum tension is reached may not be the most critical. For

sandwich-like systems, cracks in the central skin cannot be seen by visual inspection, and therefore

may grow until obtaining critical sizes. As a consequence, cracks should be introduced in the most

critical rows of a component, whether due to the lack of inspecting capability or the existence of a

maximum stress region.

Additionally, if the cracks in the different fasteners of a row have different sizes, a “catch up”

phenomenon occurs if the ratio between the crack length, a, and the hole radius, r, is in a certain

range, determined for the problem conditions. This effect will allow many cases in which the cracks

around every fastener hole will have the same length, thus facilitating their detection by the

appropriate NDI method chosen, with smaller cracks producing a reduction in the growth rates of

larger ones [30].

Also, a cracking instance may occur due to accidental damage on a free surface, like a panel.

If so, a lead crack with the prescribed sizes and directions must be introduced in the most critical

region of said panel. Furthermore, if this lead crack has sufficient size and is formed between two

fastener holes, it will contribute to create a critical area, where failure will most likely occur [31].

39

6. Load Spectrum

The loads applied to a component in service usually have a significant variation through time;

as so, they are referred to as spectrum. This spectrum must be obtained before making any fatigue

calculations, and has different shapes and properties according to the structural part from where it

belongs (wing, tail or fuselage).

Spectrums for wings and tails have a rather chaotic shape, unlike fuselage ones. This

happens because fuselages are only subjected to a pressurization/depressurization cycles, unlike

wings and tails, which suffer the effects from gusts and turbulence, inherent to the airflow. This fact

has particular relevance on fatigue life calculations, as fuselages‟ lives will be calculated under a

constant amplitude loading, unlike the wing and tail. Still, it isn‟t true that fuselages have longer fatigue

lives.

Additionally, the aircraft type has great influence on the spectrum type, due to the different

mission types of different aircraft, as illustrated by Figure 28. Still, a fighter aircraft‟s spectrum may be

used on a transport aircraft, obtaining overly conservative structures. The inverse shouldn‟t be done,

as safety will not be assured.

Figure 28 – a) Normal Flight Mission; b) Typical Military Flight Mission

In order to determine the main properties of the spectrum, one has to obtain the load variation

on time, to determine its maximum and minimum. When these values are reached, cyclic slip inversion

occurs somewhere on the component, fact of great relevance on fatigue damage accumulation.

As so, there are five important parameters for a load spectrum: alternating, maximum, mean

and minimum stresses and the stress ratio, R, but only two are needed, as the others may be

determined using:

a) b)

40

6.1. Wing Spectrums

Load spectrum must be obtained for the specific part of the aircraft in order to achieve

accurate fatigue results. Load spectrums may be obtained by two different ways: either from data

collected from an actual aircraft or through computer algorithms. The two wing spectrums presented in

this thesis belong to the AFGROW software, and are algorithm based.

Data collection spectrums are very expensive and time consuming, as the full instrumentation

of the area of interest is needed. The instrumentation is made using extensometers or accelerometers.

If necessary, through FEM software, it is possible to estimate the whole structure behaviour using the

extensions/accelerations collected.

6.1.1. TWIST Spectrum

The first spectrum is TWIST, Transport Wing Standard Load, developed in 1973, and is

representative of a lower wing root panel of a transport aircraft. This spectrum stipulates that, in

average, 100 cycles are equivalent to one flight and consists of nearly 40 000 flights, considered to be

an average design life [32]. Figure 29 illustrates this spectrum.

The Nationaal Lucht-en Ruimtevaartlaboratorium (NLR), in the Netherlands, and the

Laboratorium für Betriebsfestigkeit (LBF), in Germany, developed, a few years later (1979), a

shortened version of TWIST, the MiniTWIST. This version is a concentration of the first one that

removes part of cycles repeated in the initial version; it has only 15% of the initial version‟s size. This

has significant testing time gains, but the fatigue life will be overestimated, leading to a more

conservative structure [33].

Figure 29 – TWIST Spectrum: a) simplified; b) detailed

The variation on the normal cycles appears mostly due to gust events, as for transport aircraft

aggressive maneuvering is very rare.

6.1.2. FALSTAFF Spectrum

The second spectrum presented in AFGROW is the FALSTAFF, Fighter Aircraft Loading

a) b)

41

Standard for Fatigue, released in 1975, and represented in Figure 30. It represents the load spectrum

of the lower wing root panel for a combat aircraft. This kind of spectrum has a widespread application

and represents about 200 flights with about 36 000 cycles – an average of 180 cycles per flight. [34]

As expected, the FALSTAFF spectrum that belongs to a fighter aircraft has much more

pronounced peaks, and they present a greater variation, due to the much more aggressive

manoeuvres of this kind of aircraft. Still the FALSTAFF spectrum is used for fatigue evaluations on

many aircraft, not only fighters, as it overestimates the fatigue life, thus promoting safety [33] [34].

Figure 30 – FALSTAFF Spectrum: a) simplified; b) detailed

In conclusion, TWIST has a larger quantity of peaks, which will require additional computing

time when calculations are made.

In addition, both spectrums are presented, in AFGROW, in percentage of a characteristic

stress; for TWIST is used the mean stress for the cruise flight condition, and for FALSTAFF the

maximum stress. Furthermore they are a standardization of the loads to apply for fatigue testing.

6.2. Normalized Spectrum

As shown in the spectrums of Figures 28 and 29, a chaotic disposition is present, leading to

the need of transforming them into simpler ones, which will, not only allow faster calculations, but also

to compare different spectrums. The transformed spectrum will have a sinusoidal shape, where the

medium and alternate stresses will be determined from the initial spectrum.

Using the initial spectrums, a medium of the peaks and of the valleys is determined. These

two mediums will set the maximum and minimum stresses to be used to create the normalized

spectrum.

Additionally, the spectrum can be divided in blocks, according to flight phase. This will

originate different sinusoids for each of the phases, which will allow the application of the Miner‟s Rule

to estimate the damaged accumulated by the component.

a) b)

42

The division can be made through the observation of the initial spectrum. Taxi phases (ground

roll) are usually characterized by compression on the lower wing skin, and pronounced peaks are

almost always significant manoeuvres or very strong gusts.

Still, it is important to notice, as mentioned previously, that the order in which the loads are

applied matters, even if that isn‟t translated into the Miner‟s Rule. As so, a proper coefficient of safety

must be introduced when using the normalized spectrum. To do so, the limiting damage of the Miner‟s

Rule is often decreased to values inferior to one [1]. Figure 31 is a good example for spectrums

organized per flight phase.

Figure 31 – Normalized Spectrum per Flight Phase [1]

6.2.1. Complete Spectrum Normalization

Considering a complete normalization of both spectrums, in percentage of the cruise flight

condition stress, the most important parameters are referenced in Table 7:

TWIST Spectrum FALSTAFF Spectrum

Absolute Maximum Stress

σmax 2.600 6.158

Absolute Minimum Stress

σmin – 0.600 – 1.644

Mean Stress

σm 0.993 1.379

Alternating Stress

σa 0.245 0.618

Stress Ratio

R 0.604 0.381

Table 7 – Spectrum’s Main Properties

As expected, FALSTAFF is more aggressive than TWIST, as it belongs to fighters. Still the

43

maximum stress cannot be reached very often or a quick deterioration and failure of the component

will occur.

Figure 32 illustrates the normalization of the TWIST spectrum. FALSTAFF has a similar

shape, but obviously different values for maximum and minimum stresses. Yet, it is important to

observe the transformation between the initial chaotic spectrum and the normalized one.

Figure 32 – TWIST Spectrum after Normalization

6.2.2. Spectrum Normalization per Flight Phase

The flight phase blocks are built from direct observation of the initial spectrum, limited for a

specific condition, like stormy weather or ground roll, and their defining characteristics, as mentioned

previously. Therefore, one cannot use the entire spectrum, but rather build a new one using the mean

and alternating stresses inscribed in Table 8 for the specific flight condition. Still, the number of cycles

for each phase may vary according to the level of conservatism that the designer wants, the

characteristics of the aircraft and the surrounding environment.

Flight Condition TWIST Spectrum FALSTAFF Spectrum

σm σa R σm σa R

Cruise Flight 1.0000 0.2220 0.6367 1.000 0.3775 0.4519

Ground Roll 0.2500 0.7500 – 0.5000 – 0.038 0.4718 – 1.1752

Maneuvering –––––– –––––– –––––– 1.7549 1.5098 0.0751

Light Gust 0.9973 0.2685 0.5758 1.1915 0.4729 0.4318

Stormy Weather 1.0256 0.3897 0.4493 1.5776 0.7562 0.3520

Table 8 – Spectrum’s Properties per Flight Phase

TWIST is the spectrum for a transport aircraft, for which no significant aggressive manoeuvres

44

exist, thus justifying the empty cells, which leads to a worst operating condition on stormy weather.

Also for FALSTAFF, the maneuvering condition presents itself as the most critical, with higher mean

and alternating stresses, according to what was expected. Furthermore, it is expected that the aircraft

has a more intensive and thorough maintenance program after long combat/maneuvering periods

occur.

Even more, for both spectrums, long periods on stormy weather may require shorter

inspection intervals, in order to ensure that safety has not been compromised.

Both normalizations demonstrate that FALSTAFF is much more aggressive than TWIST, as

expected, once it belongs to a fighter.

The difference between these two spectrums has severe consequences on fatigue life, as the

same component under FALSTAFF spectrum is expected to have less life time until failure. Yet, and

as referenced, its use for non-fighter aircraft will consist on a conservative approach, leading to a safe,

but overestimated structure. This type of reasoning is often used as it consists on the introduction of a

safety factor.

45

7. Residual Strength

The existence of cracks has a great influence on the ability of a structure to transfer loads, and

as long as the crack grows this capacity starts to decrease, as shown in Figure 33. Thus, the concept

of residual strength was defined as the limiting load transfer capability of the structure.

Residual strength analysis is used to determine a previously damaged component‟s ability to

withstand an ultimate static load without failure. In an analysis such as this one, safety is achieved by

designing a component where initial damage is not allowed to grow beyond a prescribed value for

residual static strength. This way, the residual strength is always kept above a critical value that, for

safety issues, must have a comfortable safety margin towards the failure moment [4].

The residual strength criterion also defines the type of damage tolerant for the structure, either

Fail Safe or Slow Crack Growth. For a Slow Crack Growth structure, the inspections program must

detect the decrease in the residual strength that is an indication of damage propagation, while for a

Fail Safe structure, the decrease in its value indicates a partial failure of the structure that must be

repaired [4].

The value for residual strength can be obtained through the stress intensity factor critical value

(fracture toughness), KIC, for unstable conditions:

Figure 33 – Residual Strength Definition [4]

In spite of that residual strength determination is usually made using the airworthiness

requirements. The CS-25 and FAR-25 documents state the main residual strength requirements on

section 25.571.b, already indicated in this document (subchapter 3.2.).

Before introducing any formulas it is important to clearly define ultimate load and design limit

load. The design limit load specifies the load that the structure is intended to support throughout its

life. The ultimate load is defined applying a safety margin of 1.5 times the design limit load, and

defines the maximum load that structure is intended support in extreme conditions, such as

malfunction or malpractice.

46

7.1. Residual Strength on Wing Skins

For wing skins, the regulations state that ultimate loads always define the static residual

strength of the structure. There are four different cases to define the residual strength, and the most

critical one must be used [17] [18]:

Normal Flight Conditions;

Gust Conditions;

Roll Maneuvering;

Continued Turbulence Conditions.

Still, the direct application of the formulas contained in the requirements‟ documents doesn‟t

provide the residual strength value but rather a group of different parameters such as load factors,

gust velocities or turbulent force [35].

7.1.1. Simplifications and Assumptions

In order to determine the residual strength some simplifications where considered. The first

one was to consider a box section beam for the internal structure of the aircraft. This type of structure

is very simple to analyze and as wide range of applications on wings. Next was used an idealization of

the structure, composed by ten booms, as Figure 34 illustrates:

Figure 34 – Box Beam Section, Idealized using Ten Booms

In an idealized structure, the booms, concentrations of area representing stringers (circles)

and spar flanges (black squares), will account for the axial load to which the wing is subjected, and the

skins will account for the shear loads. For a wing with one symmetry axis and only one load applied

(Ixy = 0 and Mx or My = 0), the normal stress varies linearly with one of the coordinates (x or y), and

thus the area of the boom 1 can be calculated using expression 50 [36]:

47

A1 is the spar flange area considered. A similar reasoning can be applied to the remaining

booms in order to determine their areas. Furthermore, the stress ratio can be reduced to a simple

distance ratio for linear stress distributions, consistent with the application of a single load on a section

with at least one symmetry axis. For a non-symmetric section the boom areas must be determined

using different methods [36].

This type of structure idealization was chosen because it enables the simulation of a wing with

a central structure composed of two spars and a few stringers, six in this case, consistent with some

in-service wings. Stringers are only introduced in the upper and lower skin in order to help the spars to

carry out the axial load.

After defining the section properties, it is also important to notice that the wing can be

modelled as a cantilevered beam with a load applied at half the semi span, as described in Figure 35.

Figure 35 – Wing Model for Residual Strength Determination

7.1.2. Requirements Application

A different concern is related to the determination of the load, whether force or moment, since

the requirements don‟t provide it. As so, if the requirements return a load factor, like for the normal

flight conditions and roll maneuvering cases, the force may be calculated using equations 51 and 52:

Wing mass must be either known or estimated. Several estimation procedures exist, such as

the one referenced in [37].

For gust cases, the requirement provides the maximum gust velocity. The requirements state

that the distance for the gust peak and the aircraft‟s penetration in the gust must be known. A range of

gust lengths is given, and, in order to be conservative, the largest gust and maximum penetration

should be considered.

Through the distance penetrated and the cruising speed, the time duration of the gust is

48

obtained. This time can be used to determine the acceleration that the gust induces on the aircraft,

through the maximum gust velocity obtained by the requirement document and considering an initial

vertical velocity equal to zero. Knowing the acceleration, a similar reasoning to normal flight conditions

can be applied.

For the gust pair case scenario, it‟s assumed that two gusts, determined like described above,

act simultaneously in two different directions on the aircraft. A specific method to combine the gusts is

suggested in the requirements. Still, another mode may be adopted if the procedure is proven safe

and conservative.

Even more, for this case, if one of the gusts is very small compared to other, the direct

utilization of the formula inscribed in the requirements returns a value inferior to the one obtained for a

single gust. As so the designer must be cautious and do both calculations in order to determine the

most critical value for the residual strength.

For continued turbulence the requirements provide the maximum load on the wing. This

turbulent force will be also applied at half the semi span. It‟s significant to notice that the turbulent

force greatly depends on the reduced frequency and its response function, as so, their accurate

determination, through a dynamic analysis, is of great importance.

The maximum moment to which the wing model is subjected can be obtained through the

multiplication of the force determined via requirements with the arm – half the semi span, for wing root

panels (the most critical part on the wing).

As final remark, and restating chapter 3, turbulent force calculations along with gust pairs mark

the greatest difference between FAR and CS documents, so, whether to use these parameters on

residual strength calculations is up to the requirements that the designer is following. Still, it is advised

to follow the CS-25, which provide more accurate definitions on these conditions and therefore will

allow the designer to build a more secure structure.

7.2. Residual Strength Determination

With the assumptions defined, the next step would be the determination of the moments of

inertia and with them one can obtain the shear flow distribution and the normal stress, using the usual

equations for n booms [36] [38] [39]:

49

The determination of qs,0 is made through a moment equilibrium towards a point of the

structure.

Furthermore, using the assumptions proposed in the previous subchapter, one can observe

that several simplifications to equations 53 and 54 will arise. The most important ones are that the

shear flow will be incremented after every boom and the elimination of the integral, since only the

booms support the axial force and the skins the shear. Moreover, the shear flow will be kept constant

between two adjacent booms.

A more realistic case scenario was also considered, in order to increase the accuracy of the

calculations – the introduction of tapering in the beams. To do so, the effects of forces and moments in

x and y directions must be accounted. These loads are generated due to the tapering of the structure

and have a direct influence in the value of qs,0. Still, no major transformation of equations 53 and 54 is

needed, and the process described in Figure 36 should be used to determine the loads Px and Py for

every boom [36]:

Determination of

σz

Determination of

Pz = σz B

Determination of

dy/dz

Determination of

Py = Pz (dy/dz)

Determination of

Px = Pz (dx/dz)

Determination of

dx/dz

Figure 36 – Determination of in-section Loads

The terms dx/dz and dy/dz represent the variation of the first coordinate along the semi span,

and can be associated with taper ratios in both directions.

Finally, the forces determined for each boom must be accounted in the moment equilibrium

used to determine qs,0. Besides the taper ratios, these forces greatly depend on the symmetry of the

section. For a section with two symmetry axis, these forces exist but their effects will compensate for

each other, leading to a null effect on the shear flow distribution.

After obtaining the normal stresses and the shear flow distribution expressions, one must

appoint the maximum shear and normal stresses. For a wing, the maximum normal stress is expected

to occur on the inferior root panel of the wing, where the moment is maximized, as it was pointed out

before.

This location is also consistent with the maximum shear flow zone, which will lead to the

maximum shear stress area, for the conditions assumed (single vertical load pointing upwards).

Therefore, the wing‟s inferior root panel can be assumed as the critical maximum stress region, as it

was previously expected.

Normal stresses are often much higher than shear ones; still, the final step will be to obtain the

combined effects of both normal and shear stresses, thus obtaining the residual stress value, to be

used further on, particularly on AFGROW calculations (failure criterion). This is made using the von

50

Mises criterion, which states that [36]:

It is expected that a gust pair of maximum intensity or very strong continued turbulence define

the conditions for the most critical value for residual strength.

51

8. Crack Growth Analysis

The subjects addressed in this chapter were already defined and presented in chapter 2,

stress intensity factors and crack growth rates. Still, a much more detailed analysis as well as

determination procedures for these parameters will be conducted.

Furthermore, important characteristics already discussed in this thesis will now be gathered.

The highest stress concentration area will be chosen and the initial damage will be applied there, for

the residual strength value already stipulated.

The residual strength value along with KIC will define the most critical option that will lead to

component failure.

The definition of the crack growth rate through the stress intensity factor will enable the correct

determination of the number of cycles that a component can withstand until achieving a critical crack

length, thus allowing the design of a damage tolerant component. This can be observed in Figure 37.

Figure 37 – a) Cycles vs. Crack Growth; b) Stress Intensity Factor vs. Growth Rate [1]

There are two main paths to follow when determining the crack growth: an analytical

approach, based on the formulas already presented, or a computational approach, through software

like AFGROW or NASGRO.

8.1. Crack Retardation

Before defining the calculus for crack growth rate it is important to introduce the concept of

crack retardation. This phenomenon appears, most of the times, when an overload occurs. The

occurrence of overloads will create a greater plastic zone, like shown in Figure 38, which will slow

down crack progression due to residual compression stresses in the overload‟s plastic zone. This type

of phenomenon is very common and can lead to an increased number of cycles that a component can

withstand. [5]

a) b)

52

Figure 38 – Crack Retardation Process [40]

Figure 39 illustrates the described effects of crack retardation on the fatigue life of a

component.

Figure 39 – Crack Retardation Effects on Crack Growth [1]

Several authors developed models to describe this process thus enabling calculations of

fatigue life:

Closure Model – a fairly simple, single input, model based on early studies of Fracture

Mechanics developed by Erdogan and Elber [41];

FASTRAN Model – based on an improved closure model developed by James Newman

[41];

Wheeler Model – is the most empirical method and modifies crack growth rate through the

use of a “knock down” factor [41];

Hsu Model – uses an effective stress combined with the closure model and is able to

account for the compression effects on tension-compression load cycles [41];

Willenborg Model – one of the most common load interaction models, based on early

Fracture Mechanics works developed at Wright-Patterson [41].

53

Each of these methods presents its own set of advantages and disadvantages, and the

designer must be cautious when selecting one of them. For more detailed information on each of

these methods reference [41] should be consulted.

In order to be conservative, crack retardation is often ignored. This way the structure will be

overestimated and more expensive, but safety will be most likely assured. For this thesis purposes no

crack retardation was considered.

Furthermore, an opposite effect, crack acceleration, exists in compression-compression load

cycles when a compressive overload occurs, but such phenomenon will not be discussed in the

present document [41].

8.2. Stress Intensity Factor

During crack growth, the stress intensity factor is the most important parameter, as it

characterizes the tensile field of the crack‟s tip, and its critical value defines the moment in which the

component will fail.

As so, its determination is very important, but isn‟t always easy, especially for complex

geometries or tensile fields because of the shape factor, β. The determination can be made through

several different processes as:

Table and graphic compilations, usually available for simple cases, in reference [8];

Superposition method, for more complex geometries. As the tensile field equations are

the same for all mode I cases, the stress intensity factor can be determined as a

combination of the load cases presented in the component [1]. Figure 40 provides a good

illustration of the applications for this method

Figure 40 – Superposition Method Example [1]

54

Computational methods, such as the Finite Element Method, FEM.

Still, nowadays, computational methods represent the most common form of determining this

parameter for their simplicity, variety and celerity.

8.2.1. Stress Intensity Factor Determination in Crack Growth Analysis

The stress intensity factor is calculated using the loading spectrum chosen. As the main

interest falls on the determination of stress intensity factor range (ΔK), one is obliged to consider also

a stress range (Δσ). The stress range is defined by the difference between the maximum and

minimum stresses.

The spectrum chosen has severe consequences, particularly when dealing with flight blocks

spectrums. As stated before, the order in which the loads are applied is relevant. A Lo-Hi loading

sequence will induce a significantly shorter fatigue life to a component than a Hi-Lo loading sequence.

The occurrence of higher stresses in the beginning of the component‟s life induces the creation of

residual stresses, which allow it to have a longer life, once subjected to lower stresses. Yet, these

residual stresses depend of the signal of the last peak in the spectrum; if this peak is positive, its

effects expand the fatigue life of the component, otherwise they tend to shorten it [1].

Figure 41 clearly illustrates the relation between the loading events sequence and the

consequent fatigue life.

Figure 41 – Fatigue Life Variation with the Loading Sequence [1]

55

Also, the stress intensity factor greatly depends on the geometry of the component under

study due to the shape factor. Such problem will be discussed in the next subchapter.

8.3. Crack Growth Rate Determination using an Analytical

Procedure

The determination of the crack growth rate is made using the growth laws presented in chapter

2, developed by Paris, Forman, etc [14]. This will allow a comparison between the results proposed by

each formula, in order to determine the most conservative, and in case of real tests, a comparison to

determine which one suits best with reality.

The calculi will follow a very simple process [42]:

1. Initial damage, a0, assumed with the dimensions and shapes presented in chapter 5;

2. Determination of ΔK using the chosen spectrum‟s stress range (Δσ):

3. Determination of da/dN applying the Growth Law in use (Paris, Forman, etc.);

4. Obtaining a medium (arithmetic or geometrical) of two consecutive values of da/dN –

;

5. Determination of number of cycles for the current increase in crack length:

6. The number of cycles, N, is obtained adding the value for ΔN determined.

7. Adding an increment to the crack size – Δa;

8. The process is then repeated from point 2 until a failure criterion is reached, whether based

on Residual Strength, acrit, or based on the fracture toughness of the material, KIC.

Furthermore, it is important to be attentive on the stress intensity factor because of the shape

factor, which may also depend on the crack growth increment. Even more, the choice of medium to

use greatly depends on the growth law that is being used. For the Paris law, an arithmetic medium

would be sufficient, as this law only defines with good approximation the second growth region. For

the remaining growth laws, a geometrical medium may return more accurate results.

The shape factor is determined using tables and graphs from reference [8], as this method is

purely analytical and no complex software was used. Thus, it presents several limitations, particularly

for the geometries that can be analyzed. Following the purposes of this thesis and its main application

to riveted joints, only such geometries were considered. Still, different loading and cracks disposition

were considered. Figure 42 illustrates the geometries and loadings considered for shape factor

calculation. [8]

56

Figure 42 – Geometries Considered for Shape Factor Determination

a) b)

c) d)

e) f)

g)

57

Figure 42 illustrates the geometries and load configurations chosen. Only one of the

geometries considers a finite plate model, while others consider an infinite plate with remote applied

stresses (tension, compression or a combination of both).

Through this data, one is able to plot the two characteristic graphs for crack growth presented

before – number of cycles vs. crack length and stress intensity factor vs. crack growth rate, Figure 37.

Analyzing the two graphs, one can determine which law describes the process better, and make

comparisons between the several propagation models.

8.4. Crack Growth Rate Determination using AFGROW

In order to obtain the crack growth rate, without destructive testing, computer aided simulation

methods are often used. Using crack growth software, such as AFGROW or NASGRO, it is possible to

obtain the estimation to the fatigue life of the component. In some software, for more complex

geometries, there is the possibility of importing a finite element model, geometry and load disposition

into the crack growth software and make the estimation for the fatigue life of said part.

For this thesis purposes, AFGROW was the software chosen to be used, as it is freeware, has

an incorporated material database and possesses a tool that allows simple model‟s construction. This

software is one of most commonly used worldwide.

Software used for determining crack growth computes it through a similar procedure to one

presented in the previous subchapter. Furthermore, the software automatically determines the shape

factor and uses much more complex crack growth laws, but if necessary allows its manual introduction

for simple geometries, in table form.

One of the most used growth laws is the NASGRO Equation, which allows the introduction of

the higher number of constants, indicating that this law is expected to have the more accurate results.

As mentioned, the software already includes a material database, which contains the constants

needed for the NASGRO Equation growth law. Still, if the user needs to alter some parameters,

manual introduction is available [14] [41].

So the process to use for crack growth determination on AFGROW will follow these stages:

1. Introduction of the Geometry (model), whether user inputted or software incorporated (if

needed the shape factor table may be also inserted);

2. Crack Growth Law selection, usually NASGRO Equation for the reasons explained

previously. The component‟s material will be also selected in this step;

3. Introduction of a Retardation Model (by default no retardation is chosen leading to a more

conservative structure and its use is recurrent);

4. Spectrum insertion, with the definition of the Residual Strength value;

5. Run the software to determine the two plots (ΔK vs. da/dN and N vs. a).

It is very important to be cautious on spectrum selection, when the stress multiplication factor

is introduced. In AFGROW, the two spectrums are in percentage of different stresses – TWIST is in

percentage of the cruise flight stress and FALSTAFF is in percentage of the maximum stress.

Furthermore, if using the proposed normalized spectrums notice that they were built in percentage of

58

the cruise flight stress too.

Ten geometric models were created using the advanced model construction tool from

AFGROW. These models were created using standard hole, skin and crack sizes, which must be

altered to meet the problem conditions. Furthermore, steps 2 to 5 still must be conducted. Table 9

contains the models built.

a) Two holes and a central through crack

b) Two holes with a through crack on each one

c) Two holes with a corner crack on each one

d) Two holes with a through crack on one of them

e) Two holes with a corner crack on one of them

f) One hole with one through

crack

g) One hole with one corner

crack

h) One hole with two through

crack

i) One hole with one through

crack and one corner crack

j) One hole with two corner

crack

Table 9 – Geometric Models Built for Insertion on AFGROW

59

Some of the models created are consistent with the geometries considered in the previous

subchapter, Figure 42, for crack growth analytical determination. This will allow a comparison between

the results obtained using the software and the ones obtained using the analytical method.

If the results obtained are inconsistent with expectations, one can correct them by using the

beta correction tool from AFGROW. With this tool, one can set a correction for the shape factor and is

often used to account for the interferences from one crack tip on another. The correction is introduced

in a table form, and software automatically interpolates solutions between two introduced points if

needed [41].

In spite of the several methods that can be used, it is important to emphasize that only real

fatigue life testing can give the designer a solid design point. Still, computational and analytical

methods such as these ones are a good starting point to a first estimation and sizing for the fatigue life

of the component under study. Safety factors are often used to ensure safety due to the reliability of

the results obtained.

60

61

9. Inspection Requirements

The goal of an inspection is the detection of a growing crack; therefore, the time definition of

the first and subsequent inspections is very important. To do so, are used some of the NDI methods

described earlier. Once the crack is detected, a repair procedure is adopted, in order to prevent

catastrophic failure, thus extending the operational life of the aircraft.

In order to properly create inspections charts, two types of plots, already introduced, may be

used – the evolution of the crack size with the number of cycles or the residual strength reduction, like

shown in Figure 43.

Figure 43 – Inspection’s Detection Interval [43]

If an inspection allows a crack detection and repair, the probability of occurring component

failure diminishes. Figure 44 illustrates this influence.

Figure 44 – Inspections’ Influence on Failure [44]

62

9.1. Inspection Type and Crack Detection

The type of inspection to conduct is defined by the methods used to determine the crack‟s

length and the degree of inspectability of the structure. Different NDI methods can be used to detect a

crack, still visual inspections, whether general or detailed, represent the greatest part of inspections

made, more than 80%, because they are cheap, fast and simple. [45]

The minimum capability that each NDI method has to detect the crack length was defined in

Table 2, in chapter 5. Still, visual inspections were left apart, as its capability greatly depends on the

skill and experience of the technician, and had no relevance when dealing with small cracks, such as

the initial damages assumed.

Yet, in-service inspections are quite different from manufacturing inspections because

structure disassembly is not always an option and also cracks have greater lengths. This leads to a

greater importance for visual inspections.

As defined previously, visual inspections should be made by experienced technicians aided

with lights, mirrors and magnifying glasses. Still, they can be divided into two major groups: detailed

visual inspections and general visual inspections. [46]

While general visual inspections are made over the entire structure, the detailed ones are

focused on a particular detail of the structure, like for example rivet holes or edges, usually using the

optical aids described. Table 10 summarizes the information on visual inspections:

General Visual Detailed Visual

Defects Detectable Cracks, holes, corrosion,

blisters, impact damage

Defect Length Larger than 1 inch near a hole,

larger than 2 inches otherwise

Approximately 0.25

inches and larger

Size of the System to

Inspect Any

Inspector Training Highly recommended

Equipment None

Optical Aids. For some internal areas

the use of a borescope or cameras

may be considered

Restrictions Surface areas only. Internal and

inaccessible areas cannot be inspected

Relative Cost Low

Table 10 – Visual Inspections’ Characteristics [4] [45] [46]

Furthermore, reference [47] provides graphics that list the influence of several parameters on

the different inspections capability for detection of a growing crack. Such parameters include prior

63

information of the expected crack location, structural part to be inspected, surface treatment and

condition, origin and location (internal vs. external). These graphs are presented in Attachment 2.

Figure 45 illustrates the capabilities of different inspections methods for crack detection,

comparing results between FAA data (visual) and the study made (present visual): [47]

Figure 45 – Crack Detection Capabilities for Different Inspection Methods [47]

The inspection type to employ depends of many factors. Detailed description of the area to

inspect must be always provided, as well as the NDI method‟s detailed description, if needed and

used. [48]

The location of the damage has a great influence on the inspection method to use.

Inaccessible locations must be inspected recurring to NDI procedures, which are only used during a

base level inspection. This may have important consequences if the most critical fatigue location is

coincident. As so, special precautions to avoid such situations should be made.

Even more, the designer must guarantee that the inspections procedures coincide with a

specific time range, where the crack is detectable but doesn‟t present any danger towards the

structure. This interval can be observed in Figure 46. The designer‟s goal would be to design such a

component where the placement of the greater number of inspections is in this interval, in order to

assure that the crack is detected and repaired, thus ensuring safety.

Figure 46 – Crack Detection Interval [49]

64

The degree of inspectability refers methods, techniques and equipment to conduct inspections

and is often used to define design goals for aircraft structures. Consequently, it may be used to define

the type of inspection and intervals for inspections. [4]

For Fail-Safe structures the degree of inspectability is also the main requirement to determine

the maximum time that an aircraft can be in service after the primary load path failure. [4]

A structure may have six different degrees of inspectability: [4]

In-flight evident – the nature and extent of the damage is sufficient to make the flight crew

aware of its existence, leading to an abortion of the mission;

Ground evident – the damage extension and nature is obvious to the ground personnel

without a specific inspection;

Walk-around visual – the damage is clearly visible in a routine general visual inspection,

with no particular aids to the technician;

Special visual inspection – the damage is located in a detailed visual inspection;

Depot or base level – the damage is found by inspection through NDI procedures, like

Eddy current, X-ray or ultrasonic. Removal of components for accessibility issues may be

executed;

In-service non-inspectable structure – if damage size or accessibility impedes detection for

the above inspections.

9.2. Scatter Factor

In an inspection, the main purpose of the scatter factor is to concede several crack detection

opportunities, thus preventing catastrophic failure of the structure. To do so, it divides the crack growth

time into smaller intervals, where inspections will be made. This factor‟s definition greatly depends on

the component to inspect. Most sensitive parts should have higher scatter factors.

The scatter factor can be obtained considering a multiplication of several life reduction factors,

which are shown in Table 11. One of these factors even accounts for the malignant effects of

corrosion. Fatigue on corrosive environments is characterized by shorter expected lives. However, this

theme, fatigue under corrosion, will not be approached in this thesis. [46]

K1, K2 Defined as 2, to imposing inspections at least on half

the component‟s life

K3

1 for low humidity environment

1.5 for medium humidity environment

2 for high humidity environment

K4

Special factor to account for unknowns (to be

discussed and approved by the Aeronautical

Authorities to ensure safety)

Table 11 – Life Reduction Factors [46]

65

Increased scatter factors will lead to greater maintenance costs, but will be created more

chances for a damage to be detected. As so, a compromise solution must be adopted when defining

this parameter.

On multiple cases, the scatter factor can be altered by the designer in order to make an

inspection coincide with the aircraft‟s maintenance program agenda, especially if a change to an

existing aircraft is made, avoiding the creation of a new inspection procedure. Still, great caution is

needed, particularly if this factor is reduced, as Aeronautical Authorities may not allow such

amendment prior to confirmation that safety was not compromised.

There are two types of scatter factors: one concerning the entire life of the component, and

other to define the number of recurrent inspections to be made after the first one.

9.2.1. Scatter Factor for Complete Life

In order to determine the moment for the first inspection, one has to define a scatter factor for

the entire expected life of the component (j1). Values between 2 and 4 are often obtained, usually to

coincide with a major depot level inspection, such as a C-check or a D-check.

A scatter factor of 2 is commonly used so that the crack has enough time to grow beyond a

value which allows its detection or close to it, using a specific inspection procedure. This way, in the

first inspection or with few recurrent inspections the damage will be rapidly detected.

Still, its determination can be made using the life reduction factors defined:

9.2.2. Scatter Factor after First Inspection

After the first inspection, a new scatter factor must be defined (j2), for the recurrent

inspections. Usually the second factor is expected to be between 3 and 8. It defines the number of

inspections between the moment the damage is expected to be detected and the moment of failure.

Figure 47 illustrates the interval where the second scatter factor will be applied.

Figure 47 – Scatter Factor – j2 [50]

66

Higher factors will create more opportunities to detect the damage, particularly if combined

with more sensitive inspection programs, mostly NDI based, with a disadvantage on the increased

costs.

9.3. Initial Inspection Requirement

Slow Crack Growth structures must be designed to withstand a period of two times the

component‟s service design life, in order to cover different uncertainties related with crack growth,

such as variability of the material‟s properties, manufacturing quality or inspection reliability. [4]

Yet, the timing to make the first inspection, also known as inspection threshold, may vary

according to the component‟s geometry, its location on the structure and the degree of inspectability.

Still, it must never be greater than half of the expected time until failure. [4]

The inspection threshold is usually determined using the total life the component is expected

to have divided by the total life scatter factor.

Furthermore, a compromise between the several components growing cracks should be made

so that fewer inspections are needed. This may have significant relevance on the design of every

component. This compromise is mostly achieved through the scatter factor.

In some occasions, the inspection program defines the inspection threshold for values of

expected crack length inferior to the ones that can be detected using a specific visual or NDI

procedure. This introduces a safety margin on the inspecting process.

9.4. Recurrent Inspection Requirement

Recurrent inspections are made after the first inspection, until the damage is detected and

repaired. In order to prevent catastrophic failure of the structure, the recurrent inspections‟ time gap is

much tighter than the initial inspection as the damage is now assumed to have grown.

The definition of the recurrent inspection interval is made considering the number of cycles

between the length in which the crack is detectable (for a particular NDI method) and the expected

moment of failure, available inspection time, divided by the appropriate scatter factor (j2). [3]

67

The number of cycles for which the crack is detectable obviously depends on the type of

inspection being conducted, whether NDI based or visual. Its determination is made using the

previously determined plot relating the number of cycles with the crack growth, knowing the methods‟

detectable crack length, as demonstrated in Figure 48.

Figure 48 – Recurrent Inspection Interval Definition [3]

The recurrent inspections would be made since the first inspection, in the remaining life time,

until the crack is detected. Once the detection is made a repair procedure is executed.

Different components will have growing cracks that might have inspection moments spaced in

time; as so, a compromise for the inspections interval should be made in order to reduce the costs,

allowing the fewer inspections needed; also, during one inspection, as many parts as possible should

be inspected. This can be made through component design and proper scatter factor choice.

68

69

10. Example of a Damage Tolerance Analysis of a Wing Panel

In this chapter will be conducted a study to determine the damage tolerance capabilities of a

repair made on a lower wing panel, following the procedure developed, together with the Excel files

and AFGROW models created.

The wing under study belongs to a Lockheed C-130A, with an internal structure, as presented

in Figure 49. Three damaged locations were considered in the highlighted panel (see Figure 50): one

on the root, other on the mid semi span and another at the wing tip. This panel is a machinated semi

span long panel, with non linear variable thickness, built in aluminium alloy 7075-T6.

Figure 49 – Structure of the Lockheed C-130A Wing

Figure 50 – Scheme for the Analyzed Wing (without Externally Mounted Probes)

For the damages considered near the wing root and tip, two different approaches can be

made: either considering 10 inches towards the margin or that the damage starts in the margin. The

following calculations were done considering the 10 inches margin.

Stress concentration is not an issue for this example once the panel doesn‟t contain any

fastener holes or particular geometrical feature that would introduce such concentration. Still, the

70

repair procedure adopted, as will be seen, will need the application of fasteners and so, after a first

repair, stress concentration will become an important issue, reducing fatigue life.

The damages were introduced with the length prescribed in Table 3, and they vary along the

panel once the thickness also varies. These damages were all assumed in the spanwise direction

once propagation is more critical in this direction.

Next, the residual strength was calculated, using the procedure described in this thesis. The

aircraft analyzed is a good example for a box beam cross section, thus a safety factor was not applied.

Some data relative to the aircraft had to be introduced, such as MTOW, cruise speed, ceiling altitude,

among others, collected through reference [51], and presented in Attachment 3. The turbulence

requirement for residual strength wasn‟t considered, as frequency response functions weren‟t

available. The wing weight was assumed to be 7% of the aircraft empty weight, as suggested in

reference [37].

Table 12 summarizes the information determined, where the residual strength value was

maximum for a peak intensity gust pair, as expected:

Lower Surface

Thickness [in]

Crack Length

and Shape [in]

Upper Surface

Thickness [in]

Residual

Strength [ksi]

Wing

Root 0.21

0.16 24.29

Half Semi

Span 0.125

0.12 21.23

Wing

Tip 0.09

0.055 28.19

Table 12 – Damage and Residual Strength Data

This data will now be introduced into AFGROW, in order to determine the damage growth.

Two failure criterions where considered, the residual strength and the fracture toughness of the

material, in order to obtain accurate results for the fatigue life of the panel under study.

Once AFGROW doesn‟t allow the introduction of the panel‟s variable thickness, the

geometrical model adopted assumes a panel with a constant thickness, consistent with the location of

the damage.

Furthermore, AFGROW only allows the placement of a central corner crack. As so, for the

wing root damage case, it was considered a through crack. This will induce an overestimation of the

structure, which will ensure the safety of the calculations, acting as a safety factor.

The spectrum considered was FALSTAFF, once it is the most aggressive, and the aircraft in

consideration is military. Thus, the life prediction will most likely be overestimated, once more assuring

safety. Furthermore, a stress multiplication factor of 29 was considered. This means that the maximum

71

stress expected is of 29 ksi (200 MPa).

The propagation law chosen was the NASGRO equation, for the aluminium alloy 7075-T6,

which enables the largest quantity of parameters to introduce. These parameters are already a part of

the software.

Although some of the considerations made might seem too much conservative, it is important

to observe that AFGROW doesn‟t enable the introduction of the panel‟s variable thickness. Even

more, for safety purposes, it was crucial to introduce a safety margin, like through a stronger spectrum

than the one that the aircraft will be subjected to.

Table 13 resumes the data collected for crack growth predictions using AFGROW. It is

relevant to notice that, for FALSTAFF, one flight is simulated using an average of 180 cycles, and

approximation by defect was considered.

Wing Root

Damage

Mid Semi Span

Damage

Wing Tip

Damage

Maximum

Crack Length [in] 10.797 4.976 12.567

Maximum

Number of Cycles 554 247 667 729 677 126

Flights

until Failure 3079 3709 3761

Table 13 – Results for Crack Growth Prediction using AFGROW

Figure 51 illustrates the evolution of the crack growth.

Figure 51 – Crack Growth Prediction using AFGROW

72

Central and tip damages have longer lives, as expected. Rupture is originated, for all three

cases, by the fracture toughness failure criterion.

Still, it is important to emphasize that, for the wing root and tip damages, AFGROW expands

the crack in the moment of failure till hitting the panel margin, which can be seen as the vertical graph

segments in Figure 51. As so, the crack length in these locations is higher.

In order to properly comprehend this phenomenon, an alternate case scenario was

considered, placing the initial damage at the panel margin. For this case, Table 3 indicates a different

initial damage assumption – a corner crack with a 0.05 inch radius. Figure 52 illustrates the influence

of the initial damage location on the panel fatigue life.

Figure 52 – Initial Damage Location Influence on the Panel Life

The differences are considerable, and are justified mostly due to the distinct initial damages

considered, which lead towards important changes in the shape factor, β. Even more, the failure in the

second case studied is also determined by the fracture toughness. Still, the case scenario where the

initial damage is placed at 10 inches from the margin is the worst case, and therefore will be used

further on.

To finalize, the inspections program is presented, considering the damages at 10 inches from

the wing tip and root, as well as, the central damage, for the reasons already explained. In it, the

inspection threshold must be defined, alongside with the recurrent inspections interval.

Once an inspection detects the damage, it will be corrected using a flush skin repair, like

Figure 53 illustrates. It is important to notice that the cleaned up damage area may not surpass 8

73

inches for this repair procedure. Through the maximum repairable damage length and the minimum

detectable damage length, one can define the inspection interval.

Figure 53 – Reparation to Conduct

For all three cases, the crack growth prediction allows the choice of a scatter factor (j1) of 2 for

the inspection threshold, as the crack still has small sizes at half the expected life. Figure 54 illustrates

the inspection chart determined for the wing root damage.

Figure 54 – Inspection Chart for Wing Root Damage

As the cracks are expected to grow to easily visible sizes, a detailed visual inspection is hoped

to be sufficient. With this type of inspection, it is possible to detect a crack with at least 0.25 inches, as

presented in chapter 9.

The recurrent inspection interval is defined knowing the crack‟s length for detectability and the

maximum damage length allowed – available number of flights between detection and repair divided

by an appropriate scatter factor (j2). Recurrent inspections are made since the first inspection, and

Table 14 contains the data for their intervals.

74

Wing Root Mid Semi Span Wing Tip

Inspection Threshold

[flights] 1539 1854 1880

Available Flights between

Detection and Repair 3079 3709 3761

Scatter

Factor (j2) 4

Recurrent Inspection

Interval [flights] 769 927 940

Table 14 – Inspection Requirements using a Detailed Visual Inspection

The second scatter factor was chosen equal for all three damages because the available

inspection intervals were similar. It was determined considering K2 = 2, K3 = 1 and K4 = 2 to account

for uncertainties and to place a safety margin (if this wasn‟t an example but a real case, the value

adopted for K4 would need approval from the Aeronautical Authorities).

As mentioned previously, and in order to reduce costs, all three inspections could be made at

the same time. The wing root case, the most critical, should be adopted to do so, defining the

inspection threshold as well as the recurrent inspection interval. Furthermore, if the inspections

program of the aircraft was available, integration would be possible, and the scatter factors might need

modifications to meet the inspection program schedule.

In conclusion, one can state that the inspection procedure of this wing panel would need either

a detailed visual inspection or an NDI method for the damage to be detected. Once the visual

inspections are much cheaper they are most likely to be used.

It is also important to remind that the results obtained, in spite of providing a good starting

design point, can only be confirmed with actual fatigue tests.

75

11. Concluding Remarks and Future Developments

Damage Tolerance Design analysis of structures is an everyday procedure for maintenance

companies all over the world. As so, this thesis contains a standardized procedure to make

estimations on damage tolerance of a structure as thus promoting more accurate inspection timing,

which will ultimately lead to smaller costs for the company.

Damage tolerance is estimated over several steps:

1. Determination of the most critical point in terms of stress concentration (crack initiation and

propagation is easier at these locations);

2. Definition of the initial damage (length, direction, shape and quantity);

3. Definition of the spectrum to use, whether from a real aircraft or algorithm based, usually

selected concerning the type of aircraft that is being analysed;

4. Residual strength limits, calculated from the Airworthiness Requirements, defining the

maximum allowed crack length on the structure;

5. Crack growth determination will enable to compute the number of cycles for the structure to

have grown a crack with a prescribed dimension;

6. Inspection procedure definition, emphasizing the definition of the inspection threshold and

inspection interval. The construction of the inspection charts is made through these

parameters.

When these steps are completed, and a Damage Tolerance Analysis has been conducted, the

following must be provided:

The residual strength as a function of the crack size;

The crack growth, emphasizing the number of cycles till failure and critical crack length;

The initial damage assumed;

The inspection interval.

Future developments should be concerned with the introduction of more accurate methods to

determine residual strength and crack growth rate, for which several simplifications were made in the

interest of aiding the calculations. Also, in the interest of obtaining more accurate calculations, in-flight

spectrums of the aircraft analysed should be used. This will allow a reduction of the overestimation of

the structures thus reducing costs.

The methods used to determine stress concentration, calculate the fastener constant using

Tom Swift‟s formula. Yet, other formulas, more precise and accurate, exist. The development of such

formulas will allow the designers not to build overestimated structures.

AFGROW provides acceptable results, but its use should be made with special attention, as

only real fatigue testing can accurately estimate the fatigue life. Even more, the software doesn‟t allow

the introduction of more than two cracks, which may lead to the utilization of other software, that isn‟t

freeware, such as NASGRO. Nevertheless the software AFGROW provides a good and reliable

starting point.

Procedures that allow a more accurate determination for the inspection threshold and inspection interval may be considered.

76

77

References

1. SCHIJVE, J. Fatigue of Structures and Materials. [S.l.]: Kluwer Academic Publishers, 2004.

2. Aircraft Structural Repair for Engineers, Part III. OGMA. [S.l.].

3. HEIDA, J. H.; GROOTEMAN, F. P. Airframe Inspection Reliability using Field Inspection

Data. Nationaal Lucht-en Ruimtevaartlaboratorium (NLR). [S.l.]. 1998.

4. Damage Tolerance Design Handbook. USAF. [S.l.]. 1979.

5. BROEK, D. The Practical Use of Fracture Mechanics. [S.l.]: Kluwer Academic Publishers, 1988.

6. NIU, M. C.-Y. Airframe Stress Analysis and Sizing. [S.l.]: Hong Kong Conmilit Press, 1997.

7. BROEK, D. Elementary Engineering Fracture Mechanics. [S.l.]: Martinus Nijhoff Publishers,

1982.

8. ROOKE, D. P.; CARTWRIGHT, D. J. Compendium of Stress Intensity Factors. [S.l.]:

Procurement Executive, United Kingdom's Ministry of Defense, 1974.

9. IRWIN, G. R. Plastic Zone near a Crack and Fracture Toughness. 7th Sagamore Conference.

[S.l.]. 1960.

10. ESHELBY, J. D. Calculation of Energy Release Rate. Prospects of Fracture Mechanics, p. 69-

84, 1974.

11. RICE, R. C. et al. Metallic Materials Properties Development and Standardization (MMPDS).

FAA. [S.l.]. 2003.

12. Metallic Materials and Elements for Aerospace Vehicle Structures. Department of Defense. [S.l.].

2003.

13. PARIS, P. C.; GOMEZ, M. P.; ANDERSON, W. E. A Rational Analytical Theory of Fatigue. The

Trend of Engineering, v. 13, p. 9-14, 1961.

14. BEDEN, S. M.; ABDULLAH, S.; ARIFFIN, A. K. Review of Fatigue Crack Propagation Models

for Metallic Components. European Journal of Scientific Research. [S.l.]. 2009.

15. PARIS, P. C.; ERDOGAN, F. A Critical Analysis of Crack Propagation Laws. Trans. ASME,

Series D, v. 85, p. 528-535, 1963.

16. FORMAN, R. G.; KEARNEY, V. E.; ENGLE, R. M. Numerical Analysis of Crack Propagation in

Cyclic-loaded Structures. Journal of Basic Engineering, v. D89, p. 459-464, 1967.

17. Federal Aviation Regulations for Large Aeroplanes (FAR-25). FAA. [S.l.]. 2009.

18. Certification Specifications for Large Aeroplanes (CS-25). EASA. [S.l.]. 2009.

19. PILKEY, W. D. Peterson's Stress Concentration Factors. [S.l.]: Wiley Interscience, 1997.

20. SINGH, S.; KHALIL, Z. Simulation-aided Development of New Blind Rivets and Lock Bolts

78

Enables Form-fit Joints With Superior Mechanical Properties. [S.l.].

21. REICHLE, E. E. Rivet Replacement Analysis. [S.l.]. 1999.

22. LIU, Y. S. et al. Effect of Defects on Fatigue Life of Plates with Fastener Holes. [S.l.]. 2010.

23. NIU, M. C.-Y. Airframe Structural Design. [S.l.]: Conmilit Press, 1988.

24. Joint Service Specification Guide, Aircraft Structures (JSSG-2006). Department of Defense. [S.l.].

1998.

25. Non-Destructive Evaluation Requirements For Fracture - Critical Metallic Components. NASA.

[S.l.]. 2008.

26. Non-Destructive Inspection Methods, Basic Theory. [S.l.]. 2005.

27. CHANG, J. B.; RUDD, J. L. Damage Tolerance of Metallic Structures. [S.l.]: ASTM Committee

E-24 on Fracture Testing, 1984.

28. WANHILL, R. J. H.; KOOLLOOS, M. F. J. Fatigue and Corrosion in Aircraft Cabin Pressure

Lap Splices. Nationaal Lucht-en Ruimtevaartlaboratorium (NLR). [S.l.]. 2000.

29. WANG, X.; MODARRES, M.; HOFFMAN, P. Analysis of Crack Interactions at Adjacent Holes

and Onset of Multi-site Fatigue Damage in Aging Airframes. [S.l.]. 2009.

30. PARK, J. H.; ATLURI, S. N. Fatigue Growth of Multiple Cracks near a Row of Fastener Holes

in a Fuselage Lap Joint. [S.l.]. 1993.

31. JEONG, D.; TONG, P. Onset of Multiple Site Damage and Widespread Fatigue Damage in Aging

Airplanes. International Journal of Fracture, v. 85, p. 185-200, 1997.

32. DE JONGE, J. B. et al. A Standardized Load Sequence for Flight Simulation Tests on

Transport Aircraft Wing Structures. Laboratorium für Betriebsfestigkeit (LBF) and Nationaal

Lucht-en Ruimtevaartlaboratorium (NLR). [S.l.]. 1973.

33. LOWAK, H. et al. MiniTWIST a Shortened Version of TWIST. Laboratorium für Betriebsfestigkeit

(LBF) and Nationaal Lucht-en Ruimtevaartlaboratorium (NLR). [S.l.]. 1979.

34. DE JONGE, J. B. Additional Information About FALSTAFF. Nationaal Lucht-en

Ruimtevaartlaboratorium (NLR). [S.l.]. 1979.

35. LOMAX, T. L. Structural Loads Analysis for Commercial Transport Aircraft - Theory and

Practice. [S.l.]: AIAA Education Series, 1996.

36. MEGSON, T. H. G. Aircraft Structures for Engineering Students. 4th. ed. [S.l.]: Elsevier

Aerospace Engineering Series, 2007.

37. CORKE, T. C. Design of Aircraft. [S.l.]: Prentice Hall, 2003.

38. BARRADAS CARDOSO, J. Apontamentos de Mecânica dos Sólidos - Flexão e Corte de

Vigas. [S.l.]: Secção de Folhas, 2008.

79

39. YOUNG, W. C.; BUDYNAS, R. G. Roark's Formulas for Stress and Strain. 7th. ed. [S.l.]:

McGraw-Hill, 2002.

40. Engrasp - Worldwide Engineering Solutions. Available at: <http://www.engrasp.com/

doc/etb/mod/fm1/paris/Figure06a.gif>. Acessed in: August 2010.

41. AFGROW Manual. [S.l.]. 2007.

42. ZHANG, X. Aircraft Fatigue and Damage Tolerance. [S.l.]: Cranfield University - College of

Aeronautics, 2002.

43. VAN BEIJNEN, M. N. et al. Damage Tolerance Aspects of a Full Composite Airplane

Fuselage. Delft University of Technology. [S.l.]. 1993.

44. LINCOLN, J. W. Role of Non-destructive Inspections in Airworthiness Assurance. USAF.

[S.l.]. 1998.

45. MATZKANIN, G. A. Selecting a Non-destructive Testing Method, Part II: Visual Inspection.

AMMTIAC Quarterly - Tech Solution 2, v. 1.

46. Approval Procedures For Modifications and Repairs to Damage Tolerant Aircraft Structures.

Transport Canada. [S.l.]. 2004.

47. ASADA, H. et al. Pratical Evaluation of Crack Detection Capability for Visual Inspection in

Japan. National Aerospace Laboratory and Civil Aviation Bureau. [S.l.]. 1993.

48. SCHMIDT, H. J.; SCHMIDT-BRANDECKER, B.; TOBER, G. Design of Modern Aircraft

Structure and the Role of NDI. Daimler Benz Aerospace Airbus. [S.l.]. 1999.

49. SWIFT, S. A Collective Approach to Aircraft Structural Maintenance Programs. Civil Aviation

Safety Authority (CASA). [S.l.]. 2008.

50. ASSLER, H.; TELGKAMP, J. Design of Aircraft Structures under Special Consideration of

NDT. Airbus Deutschland. [S.l.].

51. JUSTINO, L. M. Systematic Procedure for Design and Certification of Structural Repairs.

[S.l.]. 2008.

52. ASTM Standards, Sections 1.8, 3.1 and 3.3. [S.l.]. 2004.

53. WANHILL, R. J. H. Milestone Case Histories in Aircraft Structural Integrity. [S.l.]. 2002.

54. JONES, R.; MOLENT, L.; PITT, S. Understanding Crack Growth on Fuselage Lap Joints.

Elsevier - Theoretical and Applied Fracture Mechanics. [S.l.]. 2007.

55. ANSYS Manual. [S.l.]. 2007.

80

81

Attachments

Attachment 1. – Utilities for Stress Concentration Factor Calculi [6]

Attachment 1.1. –Stress Concentration Factor for Bearing Stress

Attachment 1.2. – Stress Concentration Factor for Bypass Gross Area

Stress

82

Attachment 1.3. – Bearing Distribution Factor

83

Attachment 2. – Variation of Crack Detection with the Inspection’s

Conditions [47]

Attachment 2.1. – Prior information on crack location influence on crack

detection

Attachment 2.2. – Structural area influence on crack detection

84

Attachment 2.3. – Crack location influence on crack detection

Attachment 2.4. – Surface condition influence on crack detection

Attachment 2.5. – Surface condition influence on crack detection

85

Attachment 3. – Characteristics of the Lockheed C-130A [51]

External Dimensions Value

Wing Span 132.6 ft / 40.41 m

Aspect Ratio 10.1

Overall Length 97.75 ft / 29.79 m

Overall Height 38.8 ft / 11.84 m

Tailplane Span 52.7 ft /16.05 m

Wheel Track 14.25 ft / 4.34 m

Propeller Diameter 13.5 ft / 4.11 m

Wing Area (Gross) 1745 ft2 / 162.12 m

2

Weights and Loadings Value

Empty Weight 75562 lbs / 34274 kg

MTOW 155 000 lbs / 70305 kg

Maximum Landing Weight 130 000 lbs / 58965 kg

Wing Loading 88.83 lb/ft2 / 433.7 kg/m

2

Power Loading 8.44 lbs/shp / 5.14 kg/kW

MTOW Overload 175 000 lbs / 79380 kg

Maximum Landing Weight Overload 155 000 lbs / 70305 kg