Você está na página 1de 45

A REVIEW OF SOME DAMAGE TOLERANCE

DESIGN
APPROACHES
FOR AIRCRAFT STR~CT~RES~
PIR M. TGOR
Fracture Mechanics Section, Structures Engineering Division, Lockheed-Georgia Company,
Marietta, Georgia 30060, U.S.A.
Ah&aet-Damage tolerance design is becoming a necessity in the design of modern aircraft ~thou~ its
importance was recognized as long as four centuries ago by Leonardo da Vinci. Two decades ago structural
design e&neers and research workers felt the need of incorporating damage tolerance in the design of
aircraft structure. Due to a lack of comprehensive damage tolerance methodology large SC& component test
results were used to develop empirical damage tolerance methods. Recently, linear elastic Bacture mechanics
has been used in predicting residual strength and crack growth rates in damaged structure. As a result of these
efforts significant developments in cracked structure analytical methodology have been achieved.
The recent Au Force requirement to apply linear elastic fracture mechanics approach indamage tolerance
design of aircraft structures, warrants and critical review of various approaches. In this paper an attempt has
been made to criticalty review some damage tolerance design appro~hes and their a~I~c~ion to aircraft
structures.
The paper consists of three main sections: The first section reviews the residual strength aualysis methodology, assumptions and limitations of each method are discussed through a simple example. The second part
surveys the various crack propagation laws, including linear and non-linear ranges and spectrum loading
effects. In the third and last section, fracture mechanics methodology is applied to several types of built-up
st~ctural components under spectnun loadii conditions. The comparison of test results and analysis of
complex sect
indicate that simple methods of fracture mechanics can be applied to find the damage
tolerant strength and rate of crack grcwth.
The review presented in this paper indicates that the majority of work done in development of fracture
mechanics analytical methodology has been based on data obtained from small scale laboratory specimens
tested under closely controlled conditions of damage and environment. The validity of the methodology for
complex structure under complex loading conditions has not been established. Before the results of a fracture
mechanics analytical methodology can be accepted with a high degree of confidence many realism factors
must be properly accounted for in the analysis.

INTRODUCTION
design is becoming a necessity in the design of modern aircraft
although its importance was recognized as long as four centuries ago. Around the end
of the fifteenth century the first technical notes were written as what must have been the
first requirement for damage tolerant design. These were notebooks of Leonardo da
Vinci in which he discussed the physics of flight and the design of flying machines.
He wrote: In constructing wings one should make one card to bear the strain and a
looser one in the same position so that if the one breaks under strain the other is in
position to serve the same function.
More recently The United States Civil Aeronautics Board has defined the damage
tolerant structure as one in which: Cat~tr~phic failure or excessive structural deformation, which could adversely affect the flight characteristics of the airplane, are not
probable after fatigue damage or obvious partial damage of a single principal structuraI
element.
The object is to ensure that the structure will continue to sustain a high proportion
of its design load even after a damage has occurred.
The basic philosophy of damage tolerant design is ba%edon:
(1) The acceptance that damage will occur for one reason or another despite ah
precautions taken.
DAMAGE t&rat

tpresented at the Symposium on Fracture and Fatigue, George Washington University, May 3-5, 1972.
837

838

PIR M. TOOR

(2) An adequate system of inspection prescribed so that the damage may be


detected and repairs made at a proper time.
(3) An adequate residual strength maintained in the damaged structure so that,
during the period between inspection in which the damage lies undetected,
ultimate failure of the structure is not probable.
The importance for adopting damage tolerant design approaches to the design of
structures must be stressed due to the following:
(a) The extreme difficulty of predicting fatigue damage both in location and in time
of occurrence due to the variations in applied loads.
(b) After some years of service the structure itself will not be the same as that
tested in the design stage due to incorporation of repairs and modifications
incorporated by the operators.
(c) The fact that even if a large safety factor is applied to take into account the
scatter in fatigue endurance, there are other types of damage which can occur to
the structure while in service. Damage can arise as the result of corrosion,
runway debris and even mid-air collisions with birds.
Structural design engineers and research workers realizing the importance of
damage tolerant design have carried out extensive research to calculate residual
strength of damaged structure and rate of damage propagation. Important methods and
approaches are discussed in the following sections.
RESIDUAL STRENGTH

ANALYSIS

Inglis analysis
Inglis [ l] was the first to attempt the elastic stress analysis of cracks in an infinite
plate under various degrees of biaxial tension. He considered a centrally located
elliptical hole with major and minor axes, 2a and 26, respectively, subjected to tension
oO away from the crack and perpendicular to the major axis of the ellipse 2a. The major
and minor axes of the ellipse are collinear with the x and y-axes, respectively, and the
foci of the ellipse are symmetrically disposed on the x-axis at x = a and x = -a. He
concluded that the maximum tensile stress occurred at the tip of the major axis and
gave the following equation for maximum stress, (T,~
~~,,,,=a,[l+2(a/b)].

(1)

As the minor axes approaches zero, the ellipse shrinks to a line crack of length 2a.
Inglis gave a modified general form of the above equation which is applicable to this
case, i.e.

(2)
where p is the radius at the crack tip. Since p is normally very small in comparison to a,
the maximum stress is given approximately by

Grifiths energy approach


The above analysis is exact for brittle materials and it is applicable to limiting forms
of the ellipse, e.g. letting the minor axis approach to zero as a limit. It was this limiting

Damage tolerance design approaches for aircraB

839

approach which prompted Griffith[2] to develop his classical energy balance relationship for unstable crack growth. The condition for Griffiths criterion was that total
elastic energy remained constant as the crack grew and the stress becomes critical for
that condition, i.e.
~cri =
rr,ti =

2E(dWldA)
7ra

>

2EfdWldA)

7ra(l-4)

>

for plane stress


for plane strain,

Vb)

The above equations show the relationship between critical stress, geometry, and
elastic properties of a highly idealized plate. An interesting feature of the above
equations is that they can be interpreted as,
aCtit/;; = material constant.

(4)

The fact that makes Griffiths work so important is that he made independent measurements both of this predicted constant and of a,6,
by conducting two sets of experiments on glass. These results compare well.
In metallic materials creation of the new surface is not a complete source of conservation of the elastic strain energy. There is also plastic strain energy, prior to crack
growth, at the tip of the cracks. Therefore Griffiths criterion was modified by Orowan
[3] and Irwin[4] as:

(5)
where 2dWfdA was replaced by G,, energy release rate which was measured directly
from a fracture test.
Westergaards stress analysis approach

Westerg~rd[S] developed a method for solving a certain class of plane-strain or


plane stress problems. He showed that the stress dist~bution was influenced by bearing
pressures or by the presence of cracks. The limitation of Westergaards solutions is that
only those cases can be solved where shear stress, a,,, is zero along the x-axis.
In particular, if a large plate contains a single crack on the x-axis whose length is
small compared to the plate dimensions or a collinear series of such cracks as shown in
Fig. 1, and if the applied loads are such that cr,, is zero along the x-axis, then the stress
distribution is readily calculated with Westergaards equations.
An additional limitation is imposed because the mathematical model for finite plate
shown in Fig. 1 is obtained from Westergaards plate when divided along lines a-a and
b-b are constrained to remain straight, whereas the edges of the free plate can deform.
Also, if the crack moves rapidly, the determination of the stress distribution away
from the crack will require a dynamic stress analysis.
Irwin[4] suggested that the stress distribution near the end of the crack can be
expressed (1) independently of uncertainties of both magnitude of applied loads and of
the dynamic unloading influences, and (2) in such a way that strain gauges near the end
of the crack can determine the crack tip stress distribution.

840

PIR M. TOOR
(a)

WEST

ERcAAR

DS

CRACKED

PLATE

lb)

Fig. 1. (a) Westergaards cracked plate. (b) Finite plate with central crack.

Irwin also suggested that the stress components


written as,

near the crack tip Fig. l(c) may be

a,=

J@cosi(l-sinisin?)

(64

o,=

J(~)cos~(l+sin~sin~)

(6b)

EG
sinqcosicos?)
4 ZiT)(

(6~)

0 = ozz =

(64

~,,=
CT= =

In the above equations r and 8 are


tip of the crack, E = Youngs modulus,
rate associated with the unit extension
meter method of representing the above

CT,,

the usual polar co-ordinates measured from the


and G = is the magnitude of the energy release
of the crack. He further suggested a one paraequations by,

K= I
where K is the stress intensity factor and has the unit of ksi 6

(7)
and it depends on the

841

Damage tolerance design approaches for aircraft

geometry of a particular crack. Therefore equations (6a-c) become,

uz=i&~0s; ( l-sin;sinT

38
1

(84
@b)
(8~)

Taking the limiting case of the above equations when 8 = 0 and rearranging we get,
K = lim 0 --* OV(2flr)cr,.

(9)

The critical value of K called K, is associated with the instability of the crack, i.e.
specifically the critical crack length.
The basic understanding of the derivation of stress intensity factor is very important
if meaningful results are to be achieved. There is much confusion in the literature on
fracture mechanics about the value of K, e.g.
K = a,V(27rr)
K = aJ42r)

(Irwin[ 191)

(10)

(Paris[3 11).

(11)
Sometimes investigators in fracture mechanics use half of the value of crack length,
and in that case a factor of 2 is missing from the equation.
Thickness effect

The critical stress intensity factor, K,, is very much dependent on material thickness. Figure 2 shows the trend for 7075-T6 aluminum alloy. There are three distinct
regions which exhibit three different types of failure modes, namely, plane stress, mixed
mode, and plane strain. Therefore, to correctly apply the value of K, in the design of
structural components it is very important to properly account for the material thickness.
7075-T6

Plane
SO.

stress region

aluminium

Mixed

mode

-Plane

Longitudinal
Transverse
-

60

40.

L
20.

001

flnlshed thickness after machining t.

I0

In.

Fig. 2. Plot of K,vs material thickness.

EFM

Vol. 5.

No. 4-D

stroln region

842

PIR M. TOOR

Finite-width correction factor


Derivation of the stress intensity factor is based on the original elastic analysis of
Inglis and Westergaard for an infinite plate. It is obvious that if one wants to use this
factor for the analysis of finite plates, some sort of width correction factor is needed.
Paris[6] has suggested that to ensure applicability of the elastic analysis used in
determining K, the panel length should be at least twice its width, W, which in turn
should be at least twice the crack length, 2a.
Tangent correction
Irwin[7] originally suggested the width correction factor for a plate of width, W,
and having a central crack of length, 2a, under gross tension stress, cr,, as,
J(itan@)=

J(gtany)whereB=$?.

(124

This expression results from Westergaards analysis which gives an approximate


solution for finite width plates and does not properly satisfy the boundary conditions
for the specimen as discussed by Paris and Sih[7].
Secant correction
Recently Isida[8] and Forman and Kobayashi[9] have performed an analysis and
these results are in good agreement with each other. Isidas polynomial expression
gives excellent correlation with the experimental results over a large range of aspect
ratios (2a/W) . The ASTM committee on fracture mechanics recommends the use of
Isidas relation. Fedderson[ IO] has suggested that there exists a natural trigonometric
function which approaches that of Isidas relationship very closely and has suggested
representation by the secant expression
\i(secg).

(12b)

This expression has a similarity with Irwins tangent expression because


J(jtanB)

= J(+&jtanB)

= V&G

(12c)

since 8 = sin ~9for small values of 8.


Irwin et al. [ 1 l] and Bowie et al. [ 121 have given detailed discussion about various
finite width correction factor.
Corrections for non-linear eflects
It is apparent that application of the elastic parameter K to fracture problems is
justified if the yield zone accompanying the crack tip is small compared to the crack
length. However, virtually all materials exhibit some ability to deform plastically
without fracture. If the size of the plastic zone around the crack tip is very much smaller
than all other significant dimensions of the structure and the defect, the value of
parameter K is not significantly changed. When the plastic zone becomes larger, as in
a relatively ductile material, the value of parameter K becomes questionable. Several

Damage tolerance design approaches for aircrafk

843

plastic zone models have been proposed to predict the non-linear characteristics at
crack tips. Some of the important plastic zone models are described below;
lrwin correction

Irwin[7], recognizing the importance of plastic zone ahead of the crack tips and its
influence on fracture toughness, has given somewhat emperical and approximate in
nature a plastic zone size which has been used with some success in predicting actual
fracture criteria. However, Irwins correction factor does not represent a true physical
model of stress strain condition at the crack tip. Irwin suggested that an estimate of the
plastic zone can be obtained by considering a circular shape plastic zone in front of the
crack tip. He determined the radius of plastic zone from the tip of the crack at which
the stress, CT~,
is approaching the yield stress of the material. Hence:
1 K,2
rp= 5G ( uII.$
>

Wa)

where r, is the radius of plastic zone in front of the crack tip


= is the yield strength of the material.
+U.S
Dugdale model
The Dugdale model [ 131consists of a crack of length 2a, in an inlinite plate subjected

to a nominal stress. The shape of the crack and the area in front of the crack tip is
assumed to have the shape of the stressed slit. The stress within the plastic zone is
then replaced by an equivalent distribution of internal stress, acting on the slit interface, equivalent in the sense that it simulates the support that was derived from the
yielded material. This final step transforms the elastic-plastic problem into a purely
elastic one that can be solved by Muskhelishvili mapping procedure. For the case of
plane stress, the plastic zone expression is:

f-=5 secz-1.

b,

>

(13b)

Schijves approach
This approach [ 141is based on an equivalent circular plastic zone with the same area

as an actual plastic zone determined by a criteria of a critical shear stress loci around
the crack tip. Schijves approach is somewhat similar to that of Irwin%. Schijves
expression is:
(13c)
Liebowitz and EJkis approach

Reahzing the limitation of Irwins plastic zone approach (which does not include
the influence of the crack extension i.e. the influence of stable crack growth), Liebowitz
and Eftis [U-16] have generalized Irwins strain energy release rate by a simple
approach for incorporating the effects of both crack front yield and stable crack extension into a calculation of the energy release rate.
Liebowitz and Eftis suggests that an idealized fracture toughness test load dis-

844

FIR M. TOOR

placement curve (shown in Fig. 3) for a thin specimen carried out to the onset of fast
fracture can be empirically described by the simple three parameter relation:

where r = gage point displacement


F = applied load
M=tan8(c),here8referstocracklength(c).
Using the deformation theory of plasticity, they have proved that Equation (14a)
can be interpreted as an app~~a~
elastic strain hardening plastic uniaxial constitutive relation for a cracked specimen in a fracture toughness test. It is obvious that
the above relation is a load-displacement characterization of the test specimen, but,
due to its simplicity is very useful for direct experimental calculation of the energy
variation with both load increase and crack extension.
~onside~g the mechanical work (energy) per unit thickness corresponding to
displacement V,in the above equation, Liebowitz and Eftis derive the following relation:

(14b)
where iF2[ d/dc( l/M) ] is Irwins elastic energy release rate G, and where
--2nK F n-=p
n+l 0M

Fig. 3. Idealized fracture toughness test load-displacement curve of a thin specimen (Lieb~titz
and Effits model).

Damage tolerance design approaches for aircraft

845

writing equation (14b) as:

Here G is the total energy release rate per unit thickness.


Letting c = 1+jj = [ 1 + (2&/n + 1) ] (F/M) R-1.The above equation can be written
as:
c=(?G

(14d)

here c is regarded as a nonlinear correction factor to the Irwin elastic energy release
rate G.
The parameters n and K which determine c, are calculated as shown in Fig. 4 by
drawing two secant module a&l4and a,M to the load displacement curve. These are

E=

[1+(-$)(~)($1-

Liebowitz and Eftis compared the KC value for 707%T6 aluminum alloy sheet with
Irwins corrected valu,e. They observed that Irwins values were about 5 per cent
greater than the uncorrected KC values where theirs were about 17 per cent greater.
These results were corresponding to a net section stress of one-half the yield stress.
The above approach is very useful for engineering problems due to its directness
and simplicity.
In their recent publication, Liebowitz and Eftis [ 161have applied the above method,
for double and single edge cracked specimens, four point notched bend specimen,
notched round specimens and compact tension specimens.
Rices .I integral

The approaches described in the previous sections are somewhat empirical and
approximate in nature. Although they have been used with some success in predicting
actual fracture criteria, they do not represent a true physical model of stress-strain
condition at the crack tip.
In order to have a basic understanding of the complex problem of crack growth in
ductile materials, either under quasi-static or fatigue loading, knowledge of the stress
and strain ~s~bution in the region of the crack tip is required. The factors required
include such information as:
(a) Size and shape of the plastic zone,
(b) The strain distribution within the plastic zone,
(c) The effect of the plastic zone on the surrounding elastic stress distribution.

846

PIR M. TOOR

Rice [ 17-231,realizing the mathematical ditficulties accompanying the determination


of concentrated strain fields near notches and cracks, particularly in nonlinear materials
has proposed a generalized approach for the above problem. This approach first of all
identifies a line integral which has the same value for all integration paths surrounding
a class of notch tips in a two-dimensional deformation field of linear or nonlinear elastic
materials. Rice suggests that the choice of a near tip path directly relates the integral
to the locally concentrated strain field. He also points out that alternate choices for the
path will permit a direct evaluation of the integral.
Rices approach is called the path independent J integral. Below is a brief discription
of this integral. Consider a homogeneous body of linear or non-linear elastic material
free of body forces and subjected to two-dimensional deformation field (plane strain,
generalized plane stress, antiplane strain) so that all stresses (au) depend only on two
Cartesian coordinates x and y. Figure 4 shows the model used by Rice, with a general
notch having flat surfaces parallel to the x-axis and a rounded tip. Of course a straight
crack is a limiting case of this model.
The energy line integral is defined as follows:
J=

W(e) dy-T

$ds

where J = is Rices J integral


r = is a curve surrounding the notch tip
W(E) = is the strain energy density = foUdeu
T = is the traction vector acting on the outer-side of r
u = is the displacement vector
ds = is an element of arc length along r.

Flat surface notch In two dimemioml


defarmatlon field
r is any curve surmunding the notch tip
c is the curved notch tip

J%[

su
W(c) dy-T-ds
8x

Where J=Rices J integral


w(&the energy=f
u,,
dq,
u -the displacement vector
ds =an element of arc length along r

Fig. 4. Model

for Rices J integral.

(15)

847

Damage tolerance design approaches for aircraft

The details of the above integral are given in [23]. The J integral has been evaluated
for small scale yielding in plane strain as:
J++.

(16)

The above relation is Irwins energy release rate. J is always the energy release rate for
an elastic material specified by the energy density W(E) . And Equation (15), in relating
the near tip field to J, generalizes Irwins relation between linear stress intensity and
the energy rate.
Rice has established the validity of his integral, by including its value for Barenblatt-Dugdale crack model, strain concentration at smooth ended notch tips.
ALLENS METHOD
This method is based on the limitations of Irwins equation (10) which are:

(1) Equation is not accurate when applied to plates with small cracks.
(2) Mathematical plate model derived from Westergaards plate introduces

an
additional inaccuracy due to stress distribution.
(3) Irwins equation does not take into account the buckling of the plate which is
caused by the presence of cracks.
Allen[24] considers the Westergaard
section of the plate as follows:

equation for the stress distribution on the net

L( 11
. 7Ta
WI-

(+u=a,

l-

. TX

Sill-

2 -t/2

(Westergaards
equation).

(174

The stress at a small distance, r, in front of the crack tip is,

(1%)
if r is small compared to ~1,we get,

(17c)
K, described the stress field near the tip of the crack, as, for any value of r, the stress
will be proportional to K,. If r = 1/2m, a constant distance at the crack tip, then (+,,= KC.
Allen based the above argument on the experimental evidence that r remains a small
constant distance, replacing r by /27r in ( 16) we get,

(18)

848

PIR M. TOOR

Isida@] has given a width correction factor for Bnite plates as,
c=

1*0+0*3 g
0

=.

(19)

Allen uses this form in his equation as,

Buckling e$ect
Westergaard[5] in the solution of infinite plates considers the stress along the edge
of the crack in the x-direction equal to cr,. Redundant force analysis for finite plates
indicate that the stress along the edge of the crack is equal to 1.5 o& This causes
buckling and reduces the magnitude of stress that causes failure. Allen in making use
of the plate stability problems gives an expression for decrease in stress due to buckling
a%

(21)
This equation gives a good correlation with experimental data. Allen suggests that the
validity of the above equation needs further substantiation as it is based on a limited
amount of data, particularly since the range of (2alW) was only from O-3to 0.4. To use
the above equation Allen suggests to plot Au/oS vs ~r~(2a)~/Ete.By knowing cr, from
(20) one can fmd Au from this plot and the predicted failure stress would be @pRED.
=
ap- Au. This gives an average predicted stress if Kc used in the calculation of a, is
average. A factor of O-9is used to account for scatter.
BROEK APPROACH
In this analysis Broek [25] suggests that the classical fracture mechanics approaches
for calculating the residual strength of thin cracked sheets, namely,
(a) Grit&hs classical energy balance criterion, and
(b) Westergaards stress analysis approach,
can lead to a combined stress and energy criterion for crack growth and a modified
energy criterion for fracture from a phenomenological viewpoint.
In the intensive investi~tion of the residual strength properties of two light alloys,
Broek concluded that the relationship between initial crack length, 2ao, and the critical
crack length, 2a. at fracture is constant for the largest width tested (Fig. 5). He places
emphasis particularly on an infinitely wide sheet, i.e.

where y is a material constant.


It was further argued that the energy dissipated when the crack grows by an amount
dA is given by the function [(dW/d&M]. Borek further assumed that the function
dW/dA is the same for any value of the initial crack length. Therefore, equation (3a)

849

Ret-

Initial crack

length,

[3il

In.

Fig. 5. Relation between initial crack length and critical crack length 7075-T&

requires a distinct behavior of d W/dA, and this is given by the following equation:
(23)
Using Grit&hs energy balance criterion with equation (22), Broek gives the fracture
criterion as,
UC(crc)=: constant.

(24)

The above equation has certain limitations which are recognized by Broek, that the
above criterion cannot satisfactorily explain test results from specimens of small width
or specimens with large cracks.
KUHNS MWHOD
Kuhn[26] follows the Inglis general equation and bases his method on net section
stress rather than on gross section stress. He also uses the initial crack length in his
analysis instead of critical crack length. Derivation of Kuhns equation is as follows,

(25)
which is identical to Inglis ~lations~p (2).
Kuhn replaces gross section stress, o,, by trn, net section stress, and writes,
Kt = 2

= 1+ 2(;)?

(26)

PIR M. TOOR

850

Equation (26) applies to an infinite plate. Taking the width correction factor for a finite
plate as Kw, the equation for finite plates becomes,

0
a

112

&= l-4-2Kw ;

(27)

where Kw is defined by Dixons photoelastic analysis [27] as,


1

&7=-

--2a

w
1+$

(for central hole)

(for edge hole).


Kuhn, by using Neubers original formula for stress concentration and Stowel, Hardrath and Ohmans[28] equation for plastic stress concentrations, gives the following
equation for the plastic stress concentration factor considering a limiting value when
p+

0,

(30)
Where p is the material constant, E, and E are the secant modulus and tangent
modulus, respectively. In the above equation, Kuhn combines the constant term for
a particular material and calls it C,. Thus,

cm=-&(%.>

(31)

Rewriting, equation (30),


K, = 1+ C,K,vz

(32)

The net section stress at failure may be given by the expression,


CT,=---.

0;

Ku

The following example will give the details of calculating C,. Assume,
Width of the panel, W = 12-Oin.,
Thickness of panel, t = 0464 in.,
Crack length, 2a = 3.0 in. (located center of plate)
Failure load, P = 27.1 kips
Ultimate tensile stress, (+n= 7 1.0 ksi
Solve for C,
Net section area,& = (12-O-3.0) (0.064) = O-576in2

(33)

851

Damage tolerance design approaches for aircraft

P
27.1
u,=x=m=47*0ksi

= 0.775

Kw=

Ku+zm=
C,=

71-o

1.51

(Ku-l)&=
W

0.537 irP2.

WELBOURNS METHOD OF ANALYSIS


This semi-empirical method is based on an effective value of the crack tip strain
concentration factor as the criterion for stability. For unstable propagation of a crack
in any material Welbourn[29] suggests that the following two conditions must be
satisfied.
(1) The strain at the crack tip must be high enough to cause fracture of the material.
(2) The energy released during an incremental advance of the crack tip must just
exceed the energy absorbed in this advance.
Most of the materials experience a slow stable crack growth period, and if this is
the case, condition (1) is satisfied before condition (2). If the slow growth period is
neglected as most of the existing methods do, then the two conditions are assumed to
be satisfied at the same time. GritEths basic theory is based on condition (2) and the
large majority of the fracture mechanics investigations are directed toward improving
Griffiths theory. Welbourn suggested that the analysis approach based on satisfying
condition (1) is much simpler and less restricted in its application than satisfying
condition (2). Furthermore, he suggested that this analysis require only normal engineering concept of stress vs strain relationship, therefore avoiding the obscurity of
the assumptions used in an energy analysis.
The basic equation of an unstable crack propagation is given by

(34)
Two assumptions are made in Welboums analysis, i.e.
(a) Maximum stress and maximum strains are taken to be coincident on the stressstrain curve.
(b) Elastic stress concentration factor can be used as strain concentration factor up
to fracture as K values are determined empirically.
The only unknown in equation (34) is (K&.. For a central crack in a sheet of finite
width subjected to a uniform tensile stress normal to the crack, Welboum gives the
following relationship for calculating (K,) Eff..

(35)
The effective radius of the fatigue crack, r, is assumed to be constant for the material.

852

PIR M.

Effect of plate thickness is not considered in the above calculation; therefore,


results from thinner sheets where buckling of the crack edges occurs may be unconservative.
If the stress-strain curve of the material is available, then having found the appropriate value of K, the equation can be solved.
It is obvious from the above discussion that the difference in the relationship
between stress and unstable crack length for different alloys of the same basic group
results entirely from difference in their stress-strain behavior.
CRICHLOWS FOOD
Q-i&lows effective width methodf301 to calculate residual strength is well known
and given in detail in ER 11080. Therefore, the details of its development will not be
repeated here. The equation relating (TV,o#, and crack length, x, is given as
ah4
-=
@a

1+2&
(

-9

where We is the material property, considered to account for the plastic deformation.
For an infinite width panel, We is given by the following relation,

we= w()tanh

(F >*
1

Figure 6 shows the general trend of 2 W, vs. (W-x) for 707%T6 aluminum alloy.
Although the above equations are used in the industry they do not seem to answer
some basic questions. When x = 0, i.e. no crack aa = CT&which is correct. Also aa
decreases as x increases. On the other hand, when x = W, oa should become zero,
but it is not predicted by the above equation, as a matter of fact width of the plate does
not appearin the equation.
Denke and Christensen [3 l] empirically modified equation (36) to consider the width
effect by multiplying it by a factor of { l/[ 1 - (xl W)]}. Thus,
u
--g=
fit,,

L-L.
1-k

( 1+2

(38)

2 W,= k6 In.

7075T6
ALCLAD (Grain parmllel to
hxtdf r8f 1521

12

y6
(W-x),

20

24

28

32

36

in.

Fig. 6. Effective width vs distance of crack tip to free edge.

40

Damage tolerance design approaches for aircraft

To fit the experimental data, (38) was further mod&d


exponent, a: in the equation as follows,
%_
utu

853

by introducing an empirical

(1-G)
1+ $
u
e>
(

(39)

which resulted in two constants, (Yand W,.


In order to reduce the number of empirical constants, Denke and Christensen took
a new approach to the above problem. They assumed that the plastic zone of length
2W, ahead of the crack is stressed to the yield point of the material and the remainder
of the plate undergoes an elastic stress distribution derived theoretically by Westergaard. For equilibrium, the area under the plastic zone curve must be equal to the area
of the rectangle marked by the plastic zone in Fig. 7. Therefore,
w&T, = JI+wc UUdx .
Westergaard has given a&,

(40)

= (x/a) /d[ (x/a) 2- l]


a+w,

(x/a)co

weal =

(41)

d[(x/a)2-l]dxe

Integrating (41) and solving for as/u tU, we get by introducing a correction factor
similar to (38)
=3
(42)
fltu

( RP>
where R, = 3We is considered to be a material constant to be determined through tests.

1 I

I I

I WIG=;

area stress

Fig. 7. Cracked plate in equilibrium.

854

PIR M. TOOR

The question whether to use gross area stress or net area stress is also debated by
the residual strength investigators. Equation (42) is in terms of gross area stress. If the
concept of a stress concentration based on net section stress is used, then
K,= J[1+3@]

(43)

which is independent of specimen width and is no different from equation (36) proposed
originally by Crichlow.
Kuhn has made a comp~son between his formula (33) in]261 and Denke and
Christensens equation. Figure 8 shows a plot of K, vs crack length. It can be seen
that Denkes formula is only an average fit for various widths of specimens. It is also
obvious that once a value of R, has been chosen, (43) can only satisfy a single crack
length at any given width. This is shown in the upper part of Fig. 8 which shows the
ratio of calculated to experimental strength for W = 48. Christensen and Denke
formula is valid only for one crack length at each width. Thus, although Denke and
Christensen had tried to improve the Chrichlows formula, their equation seems to have
the same limitation as applicable to the ori8inal(3 6).
EXAMPLE AND DISCUSSION OF RESIDUAL STRENGTH METHODS
Material: 7075T6 ~uminum alloy.
Panel width = 24-Oin.
Panel thickness = 0.064 in.
Initial crack length = 0.5 in.
Tensile ultimate stress = 73.0 ksi
Yield stress = 64-Oksi
& from Fig. 2 = 72-Oksi 6
C, for thickness (0~09-0~1)= 1*4[24].
Various methods discussed in the previous sections are compared in this example.
Boreksmethod which has its limitation as discussed previously, is not carried out here.
The K, methods and Kuhn use initial crack length in the analysis, whereas Crichlow
used critical crack length. Therefore, to have consistent basis, results in Fig. 9 are
plotted as gross area stress vs initial crack length. Net section stress in Kuhns
method is converted to gross area stress. Only theoretical results are plotted on this
figure to 8nd the general trend of different methods. It is quite obvious firorn the plot
that Allen and Crichlows prediction agree reasonably throughout the crack length
range investigated. rU, results predict higher gross area stress for a given initial crack
length greater than 2 in. compared to Crichlows results and are in agreement for crack
length less than 2 in. On the other hand Kuhns prediction for gross area stress for a
given crack length are higher than either Allen or Crichlows predictions for crack
length greater than 4 in. and lower for crack length smaller than 4 in.
Kuhn compared his predictions with Welboums theoretical results and experimental data for Figge work Ref. (NASA-TN-D-26 13,1965), which is shown in Fig. 10.
Allen and Crichlows results were superimposed on Fig. 10, and as expected Born a
general trend shown in Fig 7, these results agreed excellently with the experimental
data, whereas both Kuhns and Welboum predictions give higher gross area stresses
for a given crack length. Figure 11 is a plot of gross area stress and critical crack length.

855

Damage tolerance design approaches for aircraft

and Cristensen

0 Denke

Q Kuhn

W=48 In.

Crack

length,

in.

v
'0

Crack

12

length,

16

20

24

in.

Fig. 8. Comparison of laws of Denke-Christensen and Kuhn for residual strength of 2219 sheet
(thickness = 0.1 in.).

Critical crack length was calculated using a linear relationship as given by equation
(37). Crichlow, from extensive amounts of data on 707%T6 aluminum alloy, has given
the value of y = 1*lo. Therefore, critical crack length was calculated by multiplying the
initial crack length by 1.1. Experimental data on Fig. 11 is from[30]. Kcs, Kuhns, and
Denke and Christensens theoretical results are compared in Fig. 11. General trend
of the various methods is the same as observed in Fig. 9.
Similarity of Allens and Crichlows predictions can be accounted for by the fact

Material 7075-T6
Thickness=OC64 in.
Width=24 in.

60.

richlows method
Allens method
0

Initial

crack length,

IO

12

I4

16

In.

Fig. 9. Comparison of Merent residual strength methods.

I6

856

PIR M. TOOR

Kuhn
Wel bourne

EXFM?mtoi
26134966)
i

(tVf.F@eNA6kTN-D-

wklnl* 6 in.
Thickness =0050 In.
a Allen and Crlchlow

lo.

I
0

0.3

0.2

01

a5

0.4

0.6

2 o/w

Fig. 10. Comparison between experimental for 2024-T6 sheet and prediction by two methods.

I
0

-1
-2

W
12

Critical

crack lenqth,

14

16

c
18

in.

Fig. 11.Comparison of experimental results and various residual strength methods.

that both methods consider a constant plastic zone ahead of the crack tip, Allen as
lj27r and Cri&low as 2W,. Also, width correction factor has a similar tangential form.
It can be concluded from the above discussion that K,, Kuhn and Welboums
predictions, although less empirical mathematically, are limited, in general use. On the
contrary, Allen and Crichlows methods have many empirical constants which are
based on large amounts of experimental data, are in general agreement with the
experimental results over large ranges of crack lengths.
CRACK

PROPAGATION

LAWS

To better study fatigue behavior and particularly low cycle fatigue, it is desirable to
divide the total fatigue life of a member into a crack initiation stage and a crack propa-

Damagetolerancedesignapproachesforaircraft

857

gation stage. Initiation is more likely to be influenced by the conditions near the point
of origin, while propagation is more likely to be afIected by the conditions throughout
the entire cross-section that the crack traverses.
Crack initiation

Ewing and Humphrey [32] were the first to make a major contribution to the theory
of crack initiation. At a point of high stress they observed the slip bands on a microscopic scale. They concluded that fatigue cracks, which were observed some time later,
were located on the original slip bands. Gough et al-[331 carried out fatigue experiments on single crystal of pure metals and concluded that although the initial slip occurred in the plane of maximum shear stress, the majority of the slip occurred in the
plane with the maximum range of shear stress and the fatigue cracks appeared in that
plane.
There are many theories based on the slip band cracking phenomenon. For instance,
Forsythe[34], while carrying out fatigue experiments on aunt
alloy specimens,
observed the appearance of a ribbon of metal which seems to have been extruded from
a slip band. Forsythe 1351also observed the inverse phenomenon of intrusion on silver
chloride. Cottrel and Hall [36] suggested that extrusion and intrusion can be produced
from the interaction between dislocations generated from two sources in opposite
directions.
Many more papers on crack vitiation are to be found in the Proceedings of the
International Conference on the Atomic Mechanism of Fracture, Swampcott,
Massachusetts, -1959, and in the Proceedings of the Crack Propagation Symposium,
Cranfield, England, 196 1.
As mentioned previously, there are many theories about the vitiation of the fatigue
crack, but there is still considerable uncertainty about the distinction between the final
stage of initiation and the start of steady crack propagation. There is common belief
among experimental observers that fatigue cracks start propagating very early in the
specimen life.
Thompsons work[37] is of particular importance here. He observed microcracks
very early (5 per cent of fatigue endurance), and 95 per cent of endurance was used by
the propagation of cracks.
Crack propagation laws given in the literature appear in many forms. In general,
they are based on cracks in infinite sheets subjected to a uniform stress perpendicular
to the crack, and they relate the crack length, 2a, to the number of cycles of applied
load, N, with the stress range, Au, and some material constant, C. Therefore, the single
form in which all the crack propagation may be written is,
g

=f(a, a, C).

(44)

HEADS LAW

Heads Law[38J was the first crack propagation law which drew wide attention. The
mechanism proposed by Head considered the plastic zone at the tip of the crack and the
elastic elements over the remainder of the finite sheet. He assumed that the plastic
zones remain initially rigid up to the yield stress u, and then work harden linearly until
EFM

vol.

5, No.

4-E

858

FIR M. TOOR

final fracture

occurs. Heads law may be written as:


da
dN = cl ~~~~~~~~

where 2a =
cr =
N=
W, =
a, =
C1 =

(45)

crackleugth
applied stress range
number of cycles
width of plastic zone at the tip of crack
yield strength
constant, which depends on the strain hardening modulus, the modulus of
elasticity, yield and fracture stress of the material.

The above equation gives a physical significance of the various parameters involved,
but its limitations remain.
(I) The assumption that W,,is constant at the tip of the crack during crack propagation
is unconservative. Frosts [39] conclusion from his experimental work is that W, is
directly proportional to crack length, Also, Irwin[7] analytically has given an approximate relation for W, = da. Therefore, Frost and Irwins observation is in
agreement.
Equation (45) becomes:
da

dN=

C2c?a212
(uy-a)a2a=

C&V%(a,-(r)

(46)

where Cz is a different constant than C,.


Yussuf 1401has suggested a relation for We = [ ua/4(a, - a) ] .
(47)
He conf%rmedthis equation by experiments. He suggests that o, is not necessarily the same as O-2per cent yield strength because of the restraint on yielding by
the surrounding materials.
(2) The second limitation of Heads formula is that it does not IWill the condition that
the rate of propagation is i&mite when the crack extends to critical length at fracture. Yussuf has suggested a non~~ension~
correction factor to fulfill this
condition in the form of
$=&whereK=ofi
c
(48)

1
= 1 - (a/a,) *I2

where a, is the critical half crack length.


Substituting equations 447) and (48) in Heads formula, the following relationship is obtained.

$i=[

(a&)

where C = 2C1.

l:Pn,
- (a/a,) 112

(49)

Damage tolerance design approachesfor aircraf%


Integrating

this

859

equation between the limits a, and a, we get,

For alummum alloy sheet, the value of C = 3.38 X 10e5 and for 2024T3, C =
7.2 x IO+.

FROSi AND DUGDALES LAW


Frost and Dugdale [29] stated that the plastic zone size is not constant as proposed
by Head. They suggested a new approach to a crack propagation law. It is obvious from
equation (46) that da/dN depends linearly on the crack length, a. They suggested a new
law which is independent of Heads law, i.e.
da/dN = Ba

(51)

where B is a function of applied stress. They experimentally found B = (@/Cd to fit the
data.
da

CT%

dN=C,

(52)

Liu [411restated the above law and suggested that B was, in general, a function of stress
range u (and mean stress), i.e.
B=B(a).

(53)

Liu argued that the mean stress is of secondary influence and, using a model of crack
extension employing an ideabzed elastic-plastic strain diagram and a concept of total
energy absorption to failure, he gives:
B(a)

= C&

(54)

Equation (5 1) becomes:
-$ = C&Pa

(55)

McEVILY AND ILLG APPROACH


This method I421 is based on the analysis technique used at NASA for calculating
the static strength of plates with cracks. They assumed a fictitious radius (pJ at the
crack tip and calculated a, in the element at the crack tip using the elastic stress concentration factor concept.

where & is stress ~n~en~tion factor and from the elastic solution of an elliptical hole
with a semi-major axis, LI,and end radius, pl, we get,
$&=1+2

a
J( Z*)

(57)

860

PIR M. TOOR

Replacing u,,,t by o, then,


(58)
They argued that under cyclic loading work hardening at the crack tip will increase the
local stress to a fracture stress. Hence, they concluded that crack propagation rate will
be a function of the stress, o0
(59)

McEvily and Illg [42] have empirically obtained the form of the function as,
= o*oo51K~a,,,-5*472-K

J434.
N A

(fm

It is interesting to note that the above equation is not to the special form as is the case
for the laws of Head and Frost.
McEvily and Illg checked their theory with extensive experimental data for R = 0.
All their results fell within a narrow scatter band around a single curve.
PARIS APPROACH
Paris [6] extended Irwins fracture mechanics concept. He stated that stress intensity
factor K reflects the effect of external load and configuration on the intensity of the
whole stress field in the vicinity of the crack tip. Paris rightly points out that for various
co~~tions
the crack tip stress field always has the same form. Therefore, he
suggested that the intensity of the crack tip stress field should control the rate of crack
propagation, i.e.
-$ = f( K) where K = afi

(61)

and (6 1) becomes,

Paris suggests that the laws of Head, Frost and Dugdale can all be approximated by
the form:
da -- uW
(63)
dNc
and equation (62) is a special form of (63) when m = n/2. It is also suggested that determination of m and n from a limited quantity of data leads to results which are questionable.
Paris solved (61) for an infinite plate subjected to uniaxial extension involving the
following two limitations:
(1) Effect of load ratio, R, is not taken into account.
(2) Cracks become unstable when K reaches the limiting value of K,.

Damagetolerancedesignapproachesforaircraft

861

From the data of 7075-T6 and 2024-T3 aluminum alloy plates, Paris suggested the
following equation for correct crack propagation law.

The above equation gives good correlation for the test data of Martin and Sinclair [43].
However, if comparisons are made with a large range of data including higher load
ratios and crack growth rates, the correlation is not good.
The above relation could be used for a structure if AK is available. In stiffened
structures, this will not be very simple and a rough estimate can be obtained at best. For
the theory of Paris to be accepted by designers, it will be necessary to accomplish the
following.
(1) To assess the stress intensity factor for configurations with the geometrical complexity as it occurs in actual aircraft structures.
(2) To show that such factors can be correlated with the crack rate in those structures
and in simple specimens made from the same material as the structure itself.

BROEK AND SCHIJVE


Since mean stress has a considerable effect on fatigue endurance, Broek and Schijve
[44] argued that the influence of mean stress on crack propagation must be important,
although there is some evidence in the literature that the effect of stress amplitude is
more important than mean stress. The knowledge gained from McEvily, Illg and
Pariss theories (they suggested a general formula for crack propagation) are as follows:

F
where

=f(Um,

w,t,4)ctnB(24

(65)

onI = mean stress,


W = testing frequency,
t = sheet thickness,
q= material parameter,
a and /3 = constants.

Broek and Schijve limited the above equation for its validity for crack length of lo-15
per cent of the sheet width. They plotted the experimental results for 2024-T3 and
7075-T6 material as crack rate versus stress amplitude and mean stress on a log-log
basis, and it was observed that such plots result in straight lines. Therefore, they suggested that (65) can be written as:
(66)
where a and /3 are the slopes of the log-log plots. They observed that p, in general, was
not constant but depends on crack length and mean stress. They made an interesting
observation that for 2024-T3 and 7075-T6 aluminum /3 seems to be almost the same.
After extensive data plotting, the authors developed a general equation applicable to

PIR M. TOOR

862

both materials. Thus,


(67)

where C1 and Cz are constants, which in principle can be determined for any material
by conducting two tests at different values of R.
Comparison of experimental and theoretical results reveals that (67) satisfies the
data for 2024-T3 very well. For 7075-T6 aluminum, the comparison is fairly good if the
crack length is not too large. For relatively large cracks, the above equation yields
unconservative results, especially for 7075-T6 material. The reason for this unconservative estimate is that the plastic zone at the tip of the crack becomes so large that a
complete redistribution of stresses takes place which was not properly accounted for
by a stress intensity factor based on elastic calculations.
It is interesting to observe that for small cracks if { 1 + lO[ (2a)/W]2} is ignored,
then (67) depends on cmax and a, which is similar to Pariss equation with the effect of
mean stress taken into account.
FORMANS EQUATION
Neither Pariss equation nor Broek and Schijves equation properly predicts the
crack propagation rates approaching fast fracture, or the instability of the crack growth.
Therefore, Forman[45] carried out an investigation to modify the exponential crack
propagation equations to cover the effects owing to the load ratio and the maximum
stress intensity factor. He formed an initial value problem with modified equation and
obtained a solution by direct numerical integration using the Runge-Kutta method.
Formans equation is based on the criterion that, assuming current fracture
mechanics theory is valid, a correct crack propagation law should have the form
d (2a) = m.
+ K,, do

limit K,,

From fracture mechanics theory it is known that,


AK= (r,,&(l

-R)

= J&,(1---R)

(69)
(70)

where R is the ratio of minimum to maximum stress intensity factor. Therefore a


general requirement for crack growth is
limitAK+

(l-R)K,,w=m.d (2a)

(71)

By assuming that the correct crack growth rate equation is of an exponential form and,
in addition, a singularity is considered to exist at ( 1 - R) K, - AK.
Forman gives an equation in the form of

d&-d _
dN

C(AK)
(l-R)K,-AK

(72)

Furthermore, with the comparison of extensive test data, he observed an interesting

Damage tolerance design approaches for aircraft

863

agreement between his theory and test results. By plotting the results on log-log paper
as AK vs [d(2a)/diVj[ (1 -R K, - AK] he found the value of constants C and n for
707%T6 and 2024T3 ~urn~urn alloys as follows
For 707%T6 aluminum alloy,
K,=68ksiV%
c = 5 x 10-13
n= 3.
For 2024T3 ~u~nurn alloy,
K, = 83 ksi V%
c = 3 x IO-13
n=3.
WALKERS METHOD
In this method [46] a single parameter (A@ which takes into account the variables;
maximum fatigue stress level, f~ma, crack size, a, and minimum to maximum stress
ratio, R, are used. This technique allows direct comparison of crack propagation data of
various maximum stresses and R values. Walker considered the general equation of
Broek and Schijve in the following form.
(73)
and suggested that it can be written as
(741
where l-m=&
m=&$
n=c+b.

Further simplification of the above equation was carried out by the introduction of
an effective stress which was defined as,
Effective stress, a = (a,,& Iem(Aa) m= cr,, ( 1 - R)m.

(75)

Therefore, (74) can be written

~=f(T31&ii)n=f(d%)

=f(A@

(76)

where AZ?is the effective stress intensity factor. The value of m was found from experiments and it is 0.50 for 2024T3 and O-425for 7075-T6.
Another similar approach based on the ~p~cation of an equation derived for sharp
notches in pure shear by Neuber [48],
K = ~(K&,)

(77)

which can reduce crack growth data at rauge ratios varying from -0.2 to +0*85 to a

864

PIR M. TOOR

single line log (daldn) vs AZ plot is illustrated by Walker[47]. The following assumptions were made to correlate stress intensity parameters and notch cyclic stresses and
strains.
(1) The strain near a sharp notch is proportional to the maximum stress intensity and is
primarily plastic strain.
(2) The stress near a sharp notch is primarily elastic under cyclic conditions and is
proportional either to K,,,, or Kmin.
(3) The effective stress intensity parameters, Kmaxor K,,, are proportional to Kt and
the effective stress intensity range AK to d(K, . Km).
Therefore, applying the above assumptions to (77) one obtains,
Knax = ~(Knax * Kmd

VW

Knin = wLx

VW

K = aGtmx(~In,

* Knin)
-K*)]

=&&(1-R).

(78c)

When R = 0, AK = AK; when R # 0, AZ? becomes an adjusted stress intensity


range corresponding to equivalent cycling at R = 0.
Walker observed that semi-log plots of AR vs d(u)/dN has a general shape of the
characteristic two slope form noted by Wilhen [5 11.The transition of slope is associated
with the change of a crack from a tensile mode of cracking to a shear mode of cracking.
The limitation of plotting Ax vs (log da/dN) is that the cracking rate does not seem
to approach infinity by deviation from the straight line relationship. This is an important
condition for any crack propagation equation to make physical significance. Recognizing the above limitation, Walker does suggest further investigations in his approach.
CRICHLOWS APPROACH
This method is based on the Griffith-Irwin criterion of energy balance and on the
shape of the curves of maximum load vs the crack length [Fig. 12(b)]. From extensive
experimental work Crichlow [30] makes the following observations.
(1) For material with reasonable ductility there exists a threshold stress level below
which no crack is evidenced [see Fig. 12(b)].
(2) Threshold stress, uO,is proportional to critical gross area stress, u,,, for stable crack
propagation at a particular initial crack length.
co = $((+cr)

(79)

(3) Over a substantial range of interest, initial crack length and critical crack length
have a linear relation.
da = +,(a,) = (UC,.-uo) = Ca,
or
a er =

a,(1 +y).

(go)

Crichlow has prepared a table which provides the value of y for different materials
derived from experimental data, and some of these are given in Table 1.
(4) The critical stress is also a function of the initial crack length and it can be arbitrarily delined as

865

Damage tolerance design approaches for aircraft

The above observation indicates that both threshold and critical stress are related
to the remote gross area stress by the same factor.
The shape of the stable crack growth vs gross stress is of an elliptical form which
can be written in cartesian coordinates:
(#!$z)m+ (q/b) = 1.0.
Above observations are the basis of Crichlows semi-empirical equation predicting
Table 1. Value of constant relating initialand critical crack
Iend [X,(1 -WI = Xc&
Material

No. of specimens

81
78
8
4

0.25
o-10
0.37
O-25

2024-T3
707S-T6
Ti 8-l-l MA (RT+550Ff
Ti8-l-l MA(-110F)
tcf. [30].

Three distinct zones of strain distri~tion


Zone
Zone
Zone

I =Fully elastic region


% * Elastic-plastic region
@* Fully plastic region

(of
tt

+T+t

r
L

i
111410
i----W4
Cracked

plate under

tension

Fastpmpagotionleadingto fracture

Ion

CfaCkgrowth between
ref

u,,,~,, and Q,,,,~


30

Fig. 12. (a) Cracked plate under tension. (b) Crack growth between u*bto umar.

866

PIR M. TOOR

the stable rate of crack growth. Semiempirical in a sense that the functions relating
gross area stress, threshold stress, critical stress, initial and final crack length, and
constants ~tzand n have to be determined from experiments for each material and
configuration.
For simple cyclic loading (0 to ornay) with R = 0, Crichlow after mathematical
manipulation gives a stable crack growth rate equation as

= caO&,ax,where &,, = l-[l-~;~~)n]lm.

(82)

When urnax= (T,,

-$ = Ca, = acr.
In an extreme case when the threshold stress is zero, to preserve the general form
of the stable crack growth curve, the exponent must be large and Crichlow gives a
general expression to predict the stable crack rate.
Fixing the variables as constants in the above derivation, (82) can be applied for
part-through cracks in monolithic materials, or through bracks in thin sheets. For partthrough cracks, A,, may be substituted for acr, A,, for a,, where A is the partially cracked
area.
The above relationship for rate of stable crack growth are based on a monotomically
increasing applied stress from 0 to urnax,i.e. a ratio R = (vmin/umax) = 0, but in actual
structure loadings are not so simple. Also, the above method is developed for positive
range ratios of stresses (a0 < oapplied< o,,.) and the rate of stable crack growth is given
by
(83)
where
(84)
and

where the prime values are not based on the actual instantaneous crack length, ao, upon
reducing stress from the previous maximum. Crichlow [30] has suggested a new propagation curve based on a new growth ellipse.
To determine the time, number of cycles or flight hours it is necessary to integrate
the rate of crack growth equation. It is stated that repetitive cyclic loading may be
considered to create a new static test with a new initial crack length which resulted from
the immediately preceding load cycle.
%r&
(86)
NC,=
7
I =rnin
where x = instantaneous crack length.

Damage tolerance design approaches for aircraft

867

CRACK PROPAGATION UNDER SPECTRUM LOADING


Crack growth studies have been carried out for constant stress amplitude, because
of simplicity in testing and because the data can be presented in a straightforward
manner. In practice engineering components are subjected to loads which are often
irregular and vary in a random manner. Fluctuating loads due to atmospheric turbulence
on aircrafts and the bending moments produced by waves on ships are typical examples
of random loadings. The difference between the conventional fatigue experiments and
the actual loading of the components has led to the development of various crack
analysis approaches.
Hardrath, Hudson and Schijve[53-551 have studied crack growth under spectrum
loadings and have observed that changes in loads can lead to accelerated or retarded
fatigue crack propagation.
The general conclusion from their study can be drawn that for a given second stress,
the delay in fatigue crack growth increased with increasing initial (first) stress. As soon
as the growth had started, it immediately reached that of constant amplitude fatigue
crack propagation at the second stress level. For low-high sequence of load application,
there is no conclusive evidence of the initial lower stress level on crack propagation at
the higher load.
Paris [56,571 and Irwin[58] have investigated the effect of single peak overloads.
Their observation is that single peak overloads also delay fatigue crack propagation. By
increasing the magnitude of the peak overloads, delay increased. They did not observe
noticeable effect due to negative peak location crack propagation.
Schijve et a14571 observed that under flight simulation the truncation of the gust
loads had a pronounced effect on crack propagation. They noticed that by applying
higher gust loads considerable decrease in crack propagation can be achieved. References [58-621 give detailed investigation in the field of spectrum loading. However, the
difference between the conventional fatigue experiments (on which crack propagation
laws are based) and the actual loading of the component has led to the development of
various crack analysis approaches. The crack propagation life for the structure has to be
determined by integrating (da/d/V) for the spectrum loads experienced by the component. These can be done as follows:
(a) Cycle by cycle integration. Starting with an initial crack length of (ai) for the
first layer of the spectrum where the stress range is known, stress intensity range can be
determined. The constants c and n used in the equation are determined from the du/dn
vs AK curve. This will give an extended crack length which will be the initial crack
length for the next stress level. The limitation of this procedure is that load sequence
cannot be accounted for.
(b) Block-integration.
rate is usually considered

Integrate blocks of cycles of the same load amplitude. Growth


constant, during growth of a small increment of crack length.

(c) Flight by jlight integration. Integration is done by flights of one peak to peak
load cycle per flight. The flight is considered as one stress cycle representing the maximum and minimum stresses for each flight.
(d) Integration per occurrence. In this procedure the total spectrum is divided into
a number of occurrences per flight, and analysis is performed according to occurrences.
Summation is made to obtain a crack propagation rate per flight.

868

PIR M. TOOR

CRACK CLOSUKE PI-IKNOMKNON


Elber[68,69] has investigated the phenomenon of crack closure in the study of
fatigue crack propagation. He suggests that fatigue fracture induced residual deformations along the crack front in the wake of the ~v~~~
crack tip and one should expect
partial crack closure after unloading the specimen. Elber, observed that a fatigue crack
propagating under zero-to-tension loading was fully closed at zero load due to internal
forces existing in the specimen. He argued that since crack propagation coot occur if
the crack is closed, only part of the applied load is effective in crack extension. He calls
this useful portion of the load cycle as the effective stress range. It is obvious from
this argument that in a high-low load app~cation, if the applied second load is lower
than the load required to open the crack tip, the crack will not propagate.
Therefore, in predicting crack growth under spectrum loading where high loads are
followed by low loads, Elber suggests that crack opening stress level should be used as a
reference stress level from which effective stress range is obtained. Therefore,
.Aueft=I amax

qop

(87)

where o,,, = is the crack opening stress


and
W= urnaxurnax

poor,

umln

=-Aaetl
AU

where U is the effective stress ratio. Then the crack propagation law becomes:
g

= C(M)&

= C( UAK).

(88)

CKACK ~T~ATION
P~NO~NON
Wheeler [70], recognizing the limitation of the crack growth laws which are based on
constant amplitude data and are used for spectrum loading with Minors cumulative
rate, has utilized a crack retardation model. He argues that the difference between
predicted and observed growth under spectrum loading condition is due to what
happens at the tip of the crack when load is applied. He takes into account the strain
history ahead of the crack tip.
Conventional analysis is based on the relationship,

where,
a, = is the initial crack length.
u, = is the total crack length after r load applications.
Ki = stress intensity range.
Wheeler has suggested a crack retardation model as shown in Fig. 13. He gives a
modified relation which considers the delay in crack growth after a high load application
as follows:

where Cpi = retardation parameter.

Damage tolerance design approaches for aircraft

869

.LGreat prior elastic-plastic


~-lnterfoce
Fig. 13. Crack tip plastic zones for Wheelers retardation model.

The general form of the retardation parameter is


CP=

m;a+R,<

a,

or
Cp= l;a+R,

3 ap

where
& = size of current plastic zone
(an - a) = is the distance from crack tip to elastic-plastic interface
m = shaping parameter.
In order to incorporate the retardation parameter into crack growth analysis, the
crack must be grown one load application at a time. Any computer program must keep
a record of loads, stress, crack tip location, plastic zone size and the elastic plastic
interface location.
Willenborg [7 11has given another approach to crack growth retardation.
EXAMPLE AND DISCUSSION
The following example is given to compare the results of the several approaches
previously presented.
Given: Material = 7075T6 Aluminum Alloy Sheet
Panel width = 12.0 in.
Panel thickness = 0.09 in.
Stress ratio, R = 0.20
Maximum stress = 30 Ksi
K,=40KsiG
Experimental data [ 181.
Comparison of Pariss approach, Broek and Schijves Method and Formans Equation is carried out to check the capabilities and limitations of each method for predicting
crack propagation behavior of flap sheet with a central crack. Walkers method is not
directly comparable with other methods as it is in terms of effective stress intensity
factor. Heads Law, Frost and Dugdales Law and McEvily and Illgs Approach all

870

PIR M. TOOR

lead to the one parameter equation of Paris; therefore, they will not be discussed here.
Crichlows method contains large numbers of constants which have to be known from
experimental data, which are not readily available. Therefore, only three methods which
to the author seem more important are discussed in this example. In Fig. 11, Ak vs
du/dN is plotted on log-log basis, and the experimental data is taken from Hudson and
Scardinas work. Results are calculated for one stress ratio (R = 0.2) only. Constants
used in the computation are given in Table 2. From Fig. 14, it can be observed that
Pariss equation seems to give reasonable correlation for the specific stress ratio.
Pariss equation is not an explicit function of stress ratios; therefore, constants in the
equation have to be calculated separately.
Broek and Schijves equation gives a reasonable fit at lower crack rates but seems to
become unconservative at higher crack rates.
Formans equation gives excellent correlation with experimental data, but it is very
sensitive to stress intensity factory. This is obvious from Fig. 15 where AK vs da/dN is
plotted for K, = 68 Ksi fi
and 40 Ksi 6
It is suggested that one should be very
careful in selecting the right value for this parameter if meaningful results are to be
achieved through the use of this equation.
Figure 16 illustrates the scatter associated with K, values, and Fig. 17 shows the
results of three analytical predictions discussed above on semi-log plots. The general
trend is the same as discussed above.
Saw cut and fatigue crack data is reproduced in Fig. 18 for 7075T6 aluminum alloy
from [24]. For a given (2a)/ W, saw cut notch gives higher net section stress compared to
a fatigue induced crack. This shows that a fatigue induced crack is more severe. It
should be pointed out here that the results are questionable for high values of (2a) / W.
Table 2. Constants used in computing (du)/dNt
Equation

Broek and Schijve


Forman

4.0
3.85 = C,
3.2

666 x 1O-21for R = 0.2


6.93 x 10-l = C
2.13 x lo-l3
1

Paris

tcf. Ref. [721.

lcpl
lo*

10-5

IO4
da/dN.

in.

10-3

10-2

&de

Fig. 14. Comparison of experimental and theoretical crack propagation rates in 7075-T6 plate
for R = 0.2.

871

Damage tolerance design approaches for aircraft


25-

0
lc6

IO4

10-s
do/dN,

to-'

in./cycle

Fig. 15. Effect of K, on Fornmns equation for 7075-T6 aluminum alloy.

Reference

m 6 In. wide panel


0 2hwkh
panel
A I in.WI& pmel
I
0

0.1

0.2

03

04

0.5

0.6

20/w

Fig. 16. Variation of K, with Wand 29/W for 7075-T6 aluminum alloy,

DAMAGE INVESTIGATION IN BUILT-UP STRUCTuREs

Two damage tolerant tests described in the next sections were carried out under
similar applied loading. Individual test loads were adjusted during the course of the test
for accurate load matching in the area under test, that is whether a beam web or a lower
cap was being tested.
Test specimen

The test specimen was a box beam composed of the rear beam, three adjacent upper
shin panels and three adjacent lower shin panels salvaged from the failed right hand
wing of fatigue test specimen and an added front beam, representative of a typical inner
wing box structure. To avoid stress concentration, aluminum straps were bonded to

872

PIR M. TOOR

30.

Paris
J

25.

IO+

104
do/dN. In./cycle

IO-

to-2

Fig. 17. Comparison of experimental and theoretical crack propagation rate in 7075T6 plate
for R = 0.2.

Fig. 18. Comparison of fatigue crack and saw cut, 7075-T6 aluminum alloy.

each end of the specimen and steel jig work was also attached to give uniform load
distribution during testing.
Loading arrangement
The inboard end of the test specimen was attached to pins to allow vertical movement only. The outboard end was placed on rollers to restrict vertical movement but to
have rotation only. Tension, vertical bending moment and tension at the outboard end
of the specimen were applied by hydraulic jacks acting through steel jig work. No
horizontal loads were applied.

Damage tokrance design appruaches for aircraft

873

From this arrangement, the loads and stresses predicted in a Unit Beam (statically
determined) analysis could be induced in the test specimen.
Experimental procedure
A saw cut 0.77 in. long was made in the skin of the rear spar cap, adjacent to and

forward of the riser attached to the rear beam web as shown in Fig. 19.
Cut was made with a saw blade. Width was 0.010 + 0.002 in. Loads were applied as
explained previously to give the appropriate stress distribution at an incremental step of
damage tolerant test load as follows:
0,20,40,60,80,100,110

per cent.

Strain gage readings were taken at each increment of load. Test loads were applied as
shown in Table 3. The saw cut did not extend until the final length was 2.019 in. At 10%
IWBRS 439

Rear

web

\-Lowar

cap

Panel NO. I

-sewcut
crack tznumatlon
Fig. 19. Experimental crack propagation
in a typical inner wing l&wer cap.
Table 3. Inner wiog rear beam lower cap crack growth

stress(KG)

No. of appiied
cycles (iV)

Maximum

::t

26.6
22.8

1.0
1-O

0.77

0.77
o-77

0.7701
o-7702

::;
3.8
3-8
3.8

41-8
41.8
41*8
41.8
41.8

1.0
1-o
1-o
::

O-77
1*006
1.256
1.505
2-019

0.77
1.006
l-256
l-505
848

o-7711
1.008
1.259
l-509

10
20
30
40
50

EFM

Vol. 5.

Crack length fin.)


X ES9
X,

Minimum

No. 4-F

X*

2-19
2.26
2-44
2-70
6.79

874

PIR M. TOOR

108 per cent a cracking noise was heard. At 110 per cent load the crack extended forward and outboard for 3.38 in. at an approximate angle of 25 deg., then turned outboard at approximately 45for 0.85 in., stopping at the tangency point of the fillet radius
of the skin panel riser. At the other end of the saw cut, the crack went through the riser
and stopped in the hrst lower taper-lok hole inboard of the skin panel saw cut. Overall
crack length rear beam to riser fillet became 6-48 in. After reaching the maximum load,
a check at 20 per cent load was carried out to detect any permanent set.
In the next run, ten load cycles (lo- 100 per cent) were applied, and the forward end
of the cracks grew forward and outboard 0.13 in. on the outer surface of lower skin
panel. No change occurred in forward end of cracks on inner surface. At the aft end of
cracks on the inside of box, the cracks came out of the lower taper-lok hole and proceeded straight out to the center line of the upper web to cap attachments and then
turned 90 inboard, stopping in the adjacent upper taper-lok hole.
A trace of the experimental crack propagation is shown in Fig. 19.
Numerical analysis and discussion of results
Using the physical models and computer program Misoni described in[84], crack
propagation was predicted for applied experimental stress levels and number of cycles.
Table 3 shows the minimum and maximum stress levels, number of cycles, initial crack
length, experimental crack length and predicted crack lengths.
For initial crack lengths 0.77-l 505 in. predicted values agree excellently with the
experimental value of the final crack length. For initial crack length of 2.019 in. experimental final crack length is 668 in. after 3 cycles under stress level of 10-l 10 per cent
of damage tolerant test load, whereas predicted crack length is only 2.089 in. In order to
achieve the crack length of 6.68 in., analysis was carried out by increasing the number
of cycles, in an incremental step of 10 cycles. It can be seen from the results that it took
50 cycles to achieve this. It is thought that this discrepancy in experimental and predicted values of number of cycles to give the same final crack length when the initial
crack length is 2.019 in. is due to the following reasons:
(1) Crack was extended by saw cutting during test - this is not a close representation of
a fatigue induced crack where a large number of cycles are required in the initiation
stage before propagation. Therefore, it is not surprising that there is a discrepancy
in the predicted and experimentally determined crack length.
(2) For a large crack length, and particularly near fast fracture, accurate measurement
of stress levels is very difficult and complicated, which can result in inaccurate
prediction.
(3) Due to complex stress distribution, the crack propagated at varying angles, but the
prediction is based on the fact that crack moves in a direction perpendicular to the
applied stress.
Test specimen
The specimen consisted of a portion of a typical inner wing lower surface panels 6,7
and 8. Loading arrangements were made in such a way that uniform stress distribution
was achieved at the test section. Rib caps were installed to simulate wing structure
conditions. The spanwise splices were the same as on the actual wing. A sketch of the
specimen is shown in Fig. 20.

Damage tolerance design approaches for aircraft


A

875

ILMN rib cap 4W23007-il6A

----__________________-______
_______ - _____ -- __________

-___

-4W23069-IOIA

Fig. 20. Test specimenfor a typical inner winglower surface, test 36.

Experimental procedure

The specimen was installed in the test fixture as shown in Fig. 21 and 35.1 ksi gross
area stress was incrementally applied and strain gage readings were taken. After removing the load, a fastener through rib clip and riser was removed.
Using a saw of 0.006 in. thickness, a cut 0.5 in. long was made in riser No. 2 1 extending from the open fastener hole in towards the skin. Loads were applied in increments of 20-60 per cent load, and then in increments of 10 per cent from 60 to 100 per
cent scheduled load (35.1 Ksi). Strain gage readings were taken at all increments. The
saw-cut did not extend under the above load conditions. Load was removed and the
saw-cut was extended as shown in Fig. 20, until a fast fracture occurred at 32.1 ksi
when the initial cut in the skin was 1.75 in.
Numerical analysis and discussion of results

The analysis was carried out as explained previously, using the appropriate stress
intensity factor (KC)values from (84). The general trend is the same as in all the previous damage tolerant tests, that the predicted values agree reasonably with the experimental values before the crack becomes unstable in the experiments, as shown in
Table 4. For an initial saw cut of 1.75 in., the predicted value of critical crack length is
again higher. It seems that prior to fast fracture, the state of stress in the component is
not exactly the same as that which is used in the analysis.
Progression offatigue crack in a typical rear beam lower cap

The ground study of fracture surfaces dictates that progression of fatigue crack in a
complex structure is a mixture of different crack growth phenomenon. In this example a

876

PIR An. TGGR


Table 4. Test loads and crack growth for Test 36

RUn

Stress

No. of

(Ksi)

cycles

35.1

;
4
5

35.1
35-l
35.1

f
:
1

Initial
saw cut
(W

Predicted crack length


& = 60.3 Ksi Uin.

Kc = 674il Ihi tin.

0.5 (Riser)

0.503

0,502

0.5 (Riser)
0.75
(Riser)
I.25 (Skin)
1a75(Skin)

0.505
0.755
1.295
1.75 (critical)

0*504
0.754
1.282
1.8 1 (critical)

phenomenological approach is taken to include a comer crack, part through crack and
edge crack. A typical rear beam lower cap is analyzed using a spectrum loading shown
in the Table 5. Figure 22 depicts a typical specimen and analysis.
The number of cycles in the spectrum were divided by 10, so that one pass is equal
to l/IO of the life of article. The method of analysis is the same as used in the previous
examples.
It is interesting to note that the article endured about 9 passes which is equal to 0.9
of the expected life. Unfortunately test results could not be compared at the time of
preparing this paper as the test is not yet complete.
CONCLUSIONS
(1) Linear elastic fracture mechanics can be used to predict the crack growth in aircraft structural components where the state of stress is accurately known.
(2) Principal stresses can be used to predict the approximate crack propagation behavior when the structural component is subject to complex state of stress, but to
have an accurate estimate, behavior of crack propagation under combined stresses
needs further ~vestiga~on.
(3) A saw cut as a crack starter is a rather unconservative replacement for a fatigue
_~

Table 5. Typical rear beam lower cap stresses

.~ ~~~~

Loadins

conditions
FlA
FIB
FIG
F2A
F2B
F2C
F3A
F3B
F3C
F4A
GlA
GiB
G2A
G2B
G3A
GAG 1
GAG2
GAG3

Max. stress

20,154
16.748
12.675
25-065
19.731
16.504
29.271
23.442
18.886
15.822
15*100
12.023
16402
13405
11.522
8.324
13.836
6.957

Min. stress
-5468
- 0.803
2.18
-0.244
4.288
8.063
2.321
7.352
12.041
10.521
-1.699
-3.854
-2.084
-4.533
-5.999
- 6,903
-11,620
- 2.647

Cy&@ife

Cycies/(l/lOof~e~
0.5

22,440
d
38,460
5
220
69,765
140,000
225
16,735
375
39,065
6%ooo
3206
4366
4428

26.0
2244.0
0.5
44-O
38&o
O-5
22.0
6976.5
14000*0
22.5
1673.5
37.5
3906.5
6000.0
320.6
436.6
442.8

Fig. 2 1. Typical inner wing lower surface specimen in the testing machine.

[Facing page 876]

877

Damage tolerance design approaches for aircraft

COVIW
I

PoasNo.
r6-4A2

02

kc

\64/12/6-t0/12

-%/4/hL64/12

requlmd fa cmck to that line

Fig. 22. Progression of fatiguecrack in a typical rear beam lower cap.

induced crack. This is possibly the reason for discrepancy in numbers of cycles to
fast fracture between the tests and the analysis.
(4) To make full use of the experimental data on damage tolerant component tests,
which are expensive to complete, it is very important to have close coordination
between the personnel responsible for analysis and testing.
(5) Substantial savings in test costs can be derived if the analysis is performed prior to
testing and much more meaningful results can be achieved to confirm accuracy of
analysis, which would eventually lead to improved damage tolerant design.
EXPERIMENTAL LIMITATIONS
(1) All the damage tolerant tests described in the previous sections were carried out at
different intervals. For example, when test on the beam web was carried out, the
lower cap and lower panels were experiencing certain loads. Similarly, the test on
the panel assembly was done after completing a fatigue test on the test specimen.
These loads, even if small on the surface, can produce plastic yielding at bolt holes
and other cutouts, which are most likely the critical sections to invite crack
initiation.
(2) The above mentioned prestressing, even though small, can produce work hardening
or work softening and may influence the final results.
(3) Prediction of crack growth using linear elastic fracture mechanics is based on the
applied stress and initial crack length. In these tests, the cracks were extended by
using saw cuts and, occasionally, by chisel and hammer (Test 1). These are very
different initial conditions from the ideal of fatigue induced cracks. Quite often only
a few stress cycles were applied and loads were removed and saw cuts were
extended. Therefore,
comparison of experimental and analytical prediction is
limited to only these few initial cycles.
(4) Strain gage readings prior to inserting a saw cut were taken only for Test 1; for other
damage tolerance tests this useful information is missing. For analytical evaluation,
it is essential to know the state of stresses before inserting a crack.

878

PIR M. TOOR

Acknowledgements-This
work was carried out at the Lockheed-Georgia Company under an independent
research and development program. The author would like to thank .I. A. Neilson, Jr., J. R. Roquemore and
F. J. Carter of Department 72-03 for the encouragement in preparin8 this paper. Dean H. Liebowitzs
(George Washington University) suggestion to include the spectrum ioading, non-linear effect and component
test data in the paper is greatly appreciated. The author is also indebted to D. P. William
of Dep~ent
72-03 for preparing the drawings and to Mrs. Kaye Fowler, Department 72-60, for typing this manuscript.

[l] C. E. In&s, Stresses in a plate due to the presence of cracks and sharp comers. Proc. Inst. NavalArch.
60,219..230(1913).
[2] A. A. Griffith, The phenomenon of rupture and ffow in solids. Trans. R. Sot. Wl, 163-198. (1920).
131 E. Orowan, Fun~en~s
of brittle behavior in metals. In Fatigue and Fracture technics
o~Meta~s,
M.I.T. and J. WiIey, New York. (1952).
[4] G. R. Irwin, Analysis of stresses and strains near the end of a crack transversing a plate. J. appl. Mech.
24: TransASME79,361-364
(1957).
[5] H.M. Westergaard, Bearing pressure and cracks. J. appl. Me&, Trans. ASME 61,49-53 (1939).
161 P. C. Paris and G. C. Sih. Stress analvsis of cracks. ASTM-STP381.410965).
i7] G. R. Irwin, Plastic zone near a crack and fracture toughness. Seienti Sagamore Ordance Materials
Research Confirence, Syracuse University Research Institute (1960).
[S] M. Isida, The effect of lon~tudin~ stiffeners in a cracked plate under tension. Proc. Fourth U.S. Cong.
appl. Mech. ASME(1962).
[9] A. S. Kobayashi and R. G. Forman, On the axial rigidity of a perforated strip and the strain energy
release rate in a centrally notched strip subjected to uniaxial tension, ASME, 63-WA (1963).
[lo] C. E. Fedderson. ASTM STP 410,77 (1966).
[l I] G. R. Irwin, H. Liebowitz and P. C. Paris, A mystery of fracture mechanics, Engng Fracture Mech. 1,
235-236 (1968).
1121
stressed strip, Engng Fracture
_ _ 0. L. Bowie and D. M. Neal, A note on the central crack in a unwon
Me&. 518 1-182 (1970).
1131 D. S. Dugdale, Yielding of steel sheets containing slits. J. Mech. Phys. Solids 8,100-104 (1960).
1141 J. Schiive, Analysis of the fatigue phenomenon in aluminum allows. NLR-TR-M, 2122 (1964).
il5] H. Liebowitz and J. Eftis, Correcting for nonlinear effects in fracture mechanics. Engng Fracture Mech.
3,267-281(1971).
[16] H. Liebowitz and J. Eftis, Correcting for nonlinear effects in fracture toughness testing. Nuclear Engng
Design 18, (1972).
1171 3. R. Rice, Plastic yielding at a crack tip. Prac. fnt. Conf. Fracture, Sendai, Japan (1965).
[18] J. R. Rice and Ducker, Energy changes in stressed bodies due to void and crack growth, ht. J. Fracture
Me& 2, (1971).
[19] J. R. Rice, Contained plastic deformation near cracks and notches under longitudinal shear. Int. 1.
Fracture Mech. (1966).
[20] J. R. Rice and G. F. Rosengren, Plane strain deformation near a crack tip in a power law hardening
material. J. Mech. Phys. Solids 16, (1968).
[21] J. R. Rice, A path independent integral and the approximate analysis of strain con~nt~tion
by notches
and cracks. J. appl. Mech. 35,379 (1968).
1221 J. R. Rice, Fatigue crack propagation. ASTM STP415,247 (1967).
[23] D. C. Drucker and J. R. Rice, Plastic deformation in brittle and ductile fracture. Engng Fracture Me&.
1,577-602(1970).
[24] F. C. Allen, Stress analysis of centrally cracked plates. Submitted to ASTM.
[25] D. Broek, The residual stren@h of light alloy sheets containing fatigue cracks. Fifth Con. ICAS, London,
(1966).
[26] P. Kuhn, Residual strength in the presence of fatigue crack. AGARD Report (1967).
1271 J.
Stress distribution around a central crack ln a plate loaded in tension. f. R. aeronaut. Sac.
.--.
-. R.
--- Dixon.
141-14s. (1959).
[28] H. F. Hardrath and L. Obman, A study of elastic and plastic stress concentration factors due to notches
and fillets inflat plates. NASA-TR-1117 (1953).
[29] E. R. Welboum, The correlation of unstable crack length data sheet materials. Aeronautical Quarterly,
Oct. 395-408 (1961).
[30] W. J. Crichlow, The material-structure
interface, Proc. AZAA/ASME Tenth Structures, Structural
Dynamics and Material ConJ, New 0rlea11s (1969).
[31] P. H. Denke and I. H. Christensen, Crack strength and propagation characteristics of high strength
metals. ASD-TR-61-207.
[32] J. A. EwiugandJ. C. Humphrey, Phil. Trans. R. Sot. (A) 200,241(1903).
[33] H. J. Goupb, Proc. R. Sot., Land. A104,538(1933).
[34] P. J. E. Forsythe, Proc. Crack Propagation Symp. Cranfield, England. 1,76(1961).

Damage tolerance design approaches for aircraft

879

[35] P. J. E. Forsythe, A two stage process of fatigue crack propagation, Proc. Crack Propagation Symp.
Crantield, England. 1,76-94 (1962).
WI A. H. Cottrel, Proc. R. Sot., Land. A242, (1957).
1371 N. Thompson and N. J. Wadsworth, Metal fatigue. Adu. Phys. 7,72-169 (1958).
1381 A. K. Head, The growth of fatigue crack. Phil. Mag. Ser. 7,44, ND356,925-938 (1953).
1391 N. E. Frost and D. S. Dugdale, The propagation of fatigue cracks in sheet specimens. J. Mech. Phys.
Solids 6,92 (1958).
1401 S. Yusef, Propagation laws of fatigue crack. Proc. AIAAIASME Tenrh Structures, Structural Dynamics
and Material Co&. New Orleans 429-437 (1969).
propagation in thin metalsheets under repeated loading. J. bas. Engng, ASME, D83,
1411 H. W. Liu, Crack
23 (1961).
[421 A. J. McEvily and W. Illg, The rate of crack propagation in two aluminum alloys. NACA-TN-4396
(1958).
[431 D. E. Martin and G. M. Sinclair, Crack propagation under repeated loading. Proc. Third U.S. Cong.
uppl. Mech. 595 (1958).
1441 D. Broek and J. Schijve, The influence of the mean stress on the propagation of fatigue cracks in aluminum alloy sheet. Nat. Aeronaut. Res. Inst., Amsterdam. NLR-TN-M 2111 (1963).
[451 R. G. Forman, V. E. Kearney and R. M. Engle, Numerical analysis of crack propagation in cyclic
loaded structures, J. bas. Engng (1967).
[461 E. K. Walker, The correlation of crack growth data at various range ratios. Lockheed Report 20686
(1967).
1471 E. K. Walker, Effect of stress ratio during crack propagation and fatigue lie for aluminum alloy.
Lockheed Report 22197 (1968).
..
[48] H. Neuber, Theory of stress concentration for shear strained prismatic bodies WIUIarbttrary nonlmear
stress-strain law. .I. appl. Mech. 83,544-550 (1961).
[49] H. Neuber, Kerlespannungslehre. Springer, Berlin (1937).
[50] E. A. Stowel, Stress and strain concentration at a circular hole in an infinite plate. NACA-TN-2073,
(1950).
[5 l] D. P. Wilhem, Investigation of cyclic crack growth transition behavior. ASTM preprint No. 38 (1966).
[52] F. M. Mueller, Failsafe criteria ultimate strength of damaged structures. Lockheed Report 11080 (1957).
1531 C. M. Hudson and H. F. Hardrath, Effects of changing stress amplitude on the rate of fatigue crack
propagation in two aluminum alloys. NASA-TN-D-960 (1960).
[54] C. M. Hudson and H. F. Hardrath, Effects of variable amplitude loading on fatigue crack propagation
patterns. NASA-TN-D-1803 (1963).
[55] J. Schijve, F. A. Jacobs and P. J. Tromp, Crack propagation in aluminum alloy sheet materials under
flight simulated loading. NLR-TR-68117U (AD692467)( 1968).
[56] P. C. Paris, The growth of crack due to variation in load. Ph.D. dissertation, Lehigh University (1962).
[57] P. C. Paris, The fracture mechanics approach to fatigue, an interdisciplinary approach. Syracuse University Press, N.Y. (1964).
[58] G. R. Irwin, Fracture mechanics. Structural mechanics. Proc. First Symp. Naval Structural Mech.
Pergamon Press, New York (1960).
[59] J. Schijve, Cumulative damage problems in aircract structures and materials. Eleventh Conference of
the International Committee on Aeronautical Fatigue, Stockholm (1969).
[60] S. H. Smith, Random loading fatigue crack growth behavior of some aluminum and titanium alloys.
ASTM STP404,74-100 (1966).
[61] E. E. Simodynes, Cycle by cycle simulation of fatigue crack growth for spectrum loading. General
Dynamics Corp. Report MRL- 127, Fort Worth, Texas (1970).
[62] H. A. Wood and T. L. Haglage, Test results and analysis of crack propagation under variable amplitude
spectrum loading for Db ac Steel. AFFDL-TM-FBR-71-2
(1971).
[63] C. M. Hudson and K. N. Raju, Invesigation of fatigue crack growth under simple variable-amplitude
loading. NASA-TN-D-5702 (1970).
[64] C. M. Hudson, Effects of stress ratio on fatigue crack growth in 7075-T6 and 2024-T3 aluminum alloy
specimens. NASA-TN-D-5390 (1969).
[65] J. C. Newman, Jr., W. Elber and M. Hudson, Fatigue crack propagation in surface cracked D6 ac steel
under block programmed loading. Langley Working Paper, LWP-930 (1970).
[66] D. Broek, Concepts in fail-safe design of aircraft structures. DMIC Memorandum 253 (197 1).
[67] E. F. J. Von Euw, Effects of single-peak over loading on fatigue crack propagation. Masters Thesis,
Lehigh University (1968).
[68] W. Elber, The significance of fatigue crack closure. ASTM STP 486,230-242 (1971).
[69] W. Elber, Fatigue crack closure under cyclic tension. Engng Fracture Mech. 2,37-45 (1970).
[70] 0. E. Wheeler, Crack propagation under spectrum loading. FZM-5602, General Dynamics, Fort Worth
Division (1970).
[71] J. D. Willenborg, R. M. Engle and H. A. Wood, A crack growth retardation model using an effective
stressconcept.TM-71-I-FBR(1971).

880

PIR M. TOOR

[72] C. M. Hudson and J. T. Scatdina, Effects of stress ratio on fatigde crack growth in 7075T6 aluminum
alloy. Engng Fracture Me&. 1,24-30 (1969).
[73] G. R. Irwin, Theoretical aspect of fracture failure analysis. Metals Engng Quarterly429-446 (1963).
[74] P. M. Toor, Damage tolerant design and strength of cracked or damaged structures. Lockheed-Georgia
Co. Rept. ER 10282, (1969).
[75] W. E. Anderson and P. C. Paris, Evaluation of aircraft material by fracture. Metals Engng Quarterly,
1,33-44 (1961).
[761 C. E. Feddenon, Plane-Strain crack toughness testing, ASTM STP 410.
[771P. V. Marcal, N. Levy, W. J. Ostergren and J. R. Rice, Small scale yielding near a crack tip in plane
strain, a finite element analysis. Znt.J. Fracture Me&s. 143- 156 (197 1).
[78] F. J. von Euw and R. W. Hertzberg, Delay in fatigue crack propagation. Private Communication.
[79] C. C. Poe, The effect of riveted and uniformly spaced stringers on the stress intensity factor of a cracked
sheet. NASA paper, Aircraft Conference on Fatigue and Fracture, Miami Beach (1969).
[SO] C. C. Poe, Fatigue crack propagation in stiffened panels. ASTM, Symposium on Crack Propagation,
Toronto (1970).
[S 11 C. E. Fedderson, Evaluation and prediction of the residual strength of center-cracked tension panels.
ASTM, Symposium on Crack Propagation, Toronto (1970).
[82] A. R. Jack and A. T. Price, The use of crack initiation and growth data in the calculation of fatigue lives
of specimens containing defects. Br. Weld. J. 416-419 (1971).
[83] J. R. Rice, Mathematic analysis in the mechanics of fracture. Fracture, An Advanced Treaties, Vol. II,
(Ed. Liebowitz), Chap. 3. Academic Press, New York (1968).
[84] P. M. Toor, SMN 358, Methodfor Crack Growth Predictions. Lockheed-Georgia Company (1970).
(Received 19 September 1972)

Você também pode gostar