Griffith's Analysis + Fatigue + Fracture

Natural Sciences Tripos Part II
MATERIALS SCIENCE
C15: Fracture, Fatigue and Creep Deformation
Dr C. Rae
Michaelmas Term 2010-11
II
Part II Materials C15 Michaelmas 2010

1
C15: FRACTURE FATIGUE AND CREEP DEFORMATION

Catherine Rae 12 Lectures
Synopsis

Introduction: This course examines the use of fracture mechanics in the
prediction of mechanical failure. We explore the range of macroscopic failure
modes; brittle and ductile behaviour. We take a closer look at fast fracture in
brittle and ductile materials characteristics of fracture surfaces; inter-granular
and intra-granular failure, cleavage and micro-ductility. We describe the range of
fatigue failure and apply fracture mechanics to the growth of fatigue cracks.
Finally we look at the processes of creep and how it combines with fatigue.

Griffiths analysis: Revision of concept of energy release rate, G, and fracture
energy, R. Obreimoffs experiment. Timeline for developments.

Linear Elastic Fracture Mechanics, (LEFM). We look at the three loading
modes and hence the state of stress ahead of the crack tip. This leads to the
definition of the stress concentration factor, stress intensity factor and the
material parameter the critical stress intensity factor.

Superposition principle, prediction of crack growth direction, failure of thin
films.

Plasticity at the crack tip and the principles behind the approximate derivation
of plastic zone shape and size. Limits on the applicability of LEFM. The effect of
Constraint, definition of plane stress and plane strain and the effect of component
thickness.

Concept of G - R curves, measuring G and K.

Elastic-Plastic Fracture Mechanics; (EPFM). The definition of alternative
failure prediction parameters, Crack Tip Opening Displacement, and the J
integral. Measurement of parameters and examples of use.

The effect of Microstructure on fracture mechanism and path, cleavage and
ductile failure, factors improving toughness,

Fatigue: definition of terms used to describe fatigue cycles, High Cycle Fatigue,
Low Cycle Fatigue, mean stress R ratio, strain and load control. S-N curves.

Adapting data to real conditions: Goodmans rule and Miners rule. Micro-
mechanisms of fatigue damage, fatigue limits and initiation and propagation
control, leading to a consideration of factors enhancing fatigue resistance.

Total life and damage tolerant approaches to life prediction, Paris law.

Creep deformation: the evolution of creep damage, primary, secondary and
tertiary creep. The use of Larson-Miller parameters. Micro-mechanisms of creep
in materials and the role of diffusion.

Ashby creep deformation maps. Stress dependence of creep power law
dependence. Comparison of creep performance under different conditions
extrapolation and. Creep-fatigue interactions.


2
Booklist:

T.L. Anderson, Fracture Mechanics Fundamentals and Applications, 2
nd
Ed. CRC
press, (1995) (Fracture mechanics and its application to fatigue, very thorough
and readable)

B. Lawn, Fracture of Brittle Solids, Cambridge Solid State Science Series 2
nd
ed
1993. (Exactly as it says on the label very good on LEFM)

J.F. Knott, P Withey, Worked examples in Fracture Mechanics, Institute of
Materials. (Excellent short summary of fracture mechanics and good worked
examples)

H.L. Ewald and R.J.H. Wanhill Fracture Mechanics, Edward Arnold, (1984).
(Provides very clear explanations different perspective from Anderson)

S. Suresh, Fatigue of Materials, Cambridge University Press, (1998)
(Excellent on fatigue but not very readable)

L.B. Freund and S. Suresh, Thin Film Materials Cambridge University Press,
(2003) Chapter 4 for a very thorough description of failure of thin films (even
less readable than the above).

G. E. Dieter, Mechanical Metallurgy, McGraw Hill, (1988)
(Good entry-level text on mechanical properties)

D.C. Stouffer and L.T. Dame, Inelastic Deformation of Metals, Wiley (1996)
(Particularly chapters 2 and 3 for creep and fatigue)

R.C Reed, The Superalloys, CUP (2006). Particularly Chapters 2 and 3 for creep
and fatigue in superalloys and Chapter 4 for lifing strategies.

3
FRACTURE, FATIGUE AND CREEP DEFORMATION

SYNOPSIS

This course examines the use of fracture mechanics in the prediction of
mechanical failure. We explore macroscopic failure modes; brittle and ductile
behaviour, and take a closer look at fast fracture in brittle and ductile materials
characteristics of fracture surfaces; inter-granular and intra-granular failure,
cleavage and micro-ductility.

Fatigue causes 90% of engineering failures: we examine how we characterise the
susceptibility of materials to fatigue and estimate lifetimes.

At high temperatures time-dependent plastic deformation occurs: we describe the
mechanisms of creep and how it can both exacerbate and mitigate the effects of
fatigue.

GRIFFITHS THEORY, REVISION FROM 1B COURSE.

Griffiths Theory provides the thermodynamic or energetic criterion for failure: it
does not consider the mechanism by which failure occurs.

The basic premise is that a crack will propagate in a material when the elastic
energy released as a result of that propagation exceeds the energy required to
propagate the crack. In the first instance just the surface energy needed to
create two new surfaces was considered, but this applies only to ideal brittle
solids i.e. those where fracture occurs without any plastic deformation.
Subsequently this was widened to include the work required to perform the
plastic deformation associated with ductile failure and, in principle, can include
any work necessary such as de-cohesion on composites phase changes etc.

o
o
2a

If we introduce a crack of length 2a into an infinite plate of thickness B under a
uniform stress o, the elastic stresses relax around the crack and reduce the
elastic potential energy U
E
stored in the plate. Extra surface is created at the
crack, U
S
, and, if the grips are fixed, no external work, U
F
, is done by the applied
force, U
F
= 0.


4
( )
s E F
U U U a U + + =
At equilibrium:

dU
da
=
dU
E
da
+
dU
S
da
= 0

The change in the potential energy is estimated from an elastic analysis of the
stresses around the crack:

U
E
~
to
2
a
2
B
E

And the work done to propagate the crack is:

s S
aB 4 U =
Where the area of the crack is 2aB, the surface area is 4aB and the surface
energy is
s
.

Thus:

d(U
E
)
da
=
2Bto
2
a
E
and

dU
S
da
= 4B
s
:
hence:
E
a
2
2
s
to
=

Rearranging:
a
E 2
s
t
= o Griffiths Equation

This is for an ideal brittle solid; for a ductile material the plastic work of
deformation
p
, is introduced:

a
E ) 2 (
p s
t
+
= o

Modification of the fracture criterion to include plastic work leads to the more
general definition of the energy release rate or the crack extension force: G.
This is the change in the potential energy, U, of the system per unit increase in
crack area, A, and has the dimensions of force/length.

Energy Release Rate:

G =
dU
dA
=
dU
2Bda
=
to
2
a
E

According to Griffiths crack extension occurs when this equals the work to
fracture, 2
s
+
p
.

p s c
2 G G + = =
G
c
is a material constant and a measure of the fracture toughness.

The rhs is the resistance to crack growth is termed R where R = 2
s
+
p
.


5
OBREIMOFFS EXPERIMENT

A real example illustrates two important points: firstly that brittle fracture is
reversible under the right circumstances and secondly, that whether it occurs or
not is governed by balancing stored elastic energy with the work of fracture.

In 1930 Obreimoff split a thin sheet of mica off a larger piece by inserting a
wedge of thickness h beween the layers. The crystal cleaves along the weak
interfaces between the layers to give a thin upper fillet and a thick lower section.
As the wedge is driven into the crack the crack grows to keep the length
constant. The elastic energy stored as the wedge is forced into the open crack is
principally in the thin upper fillet, and is balanced by the cohesive forces at the
crack tip. The crack opens until these are balanced. The energy is calculated
easily from the elastic properties of the mica, and the geometry of the set-up.

The elastic strain in the cantilever is given by beam theory:

U= U
E
=
Ed
3
h
2
8a
3
where the constants are given in the diagram.

The surface energy needed to grow the crack is

U
S
= 2a where is the surface energy.
Equating the elastic energy to the surface energy gives an equilibrium crack
length a
o
of:

a
o
= 3Ed
3
h
2
/16
4

As the wedge is withdrawn the crack closes and the damage is pretty much
repaired if the process is done in vacuum. This can be shown by reopening the
crack and noting that the value of a
o
for the re-opened crack is almost the same.
As air and moisture are introduced, the quality of the repair deteriorates and the
equilibrium length a
o
increases.

6
TIME LINE

Fatigue Fracture
~1500 - Leonardo da Vinci failure stress of iron wires
depends on length i.e. on probability of flaw
1842 - Railway accident Versailles - failure of axle
1843 - significance of fatigue striations recognized
WJM Rankin
1852-1869 - Wohler systematic experiments on bending
and torsion development of S-N curves
1874 & 1899 Gerber and Goodman life prediction
methodologies
1886 Baushinger effect noted
1900 Ewing and Rosenberg recognition of persistent
slip bands extrusions and intrusions
1913 Inglis elastic stress field around elliptical hole
1920 Griffiths equation for brittle materials
1930
1938
Obreimoffs experiment
Westergaarde elastic solution of the stress
distribution at a sharp crack
1945 Constance Tipper and the Liberty ships -
Recognition of the Ductile Brittle transition
Tipper test and the role of crystal structure in
failure
1945 Minor accumulation of fatigue damage
1953 -54 Comet airliner losses due to fatigue failure
1954 Coffin Manson empirical laws for HCF and LCF
1956 1956 Wells applies fracture mechanics to fatigue to
explain the Comet fatigue fractures
1956 Irwin development of the concept of energy
release rate based on Westergardes work
1956 Demonstration of the role of PSB in initiating
fatigue failure
1957 Fracture mechanics predicts disc failures for GE
1960 1960 Paris law relating the crack growth rate to the
stress intensity factor
1960-61 Irwin/Dugdale/Wells development of LEFM and
effect of plastic zone size and shape
1968 Proposal of the J integral by rice and the CTOD
by Wells to cope with the failure of ductile
materials
1976 Shih and Hutchinson establish the theoretical
basis of the J-Integral and link it to the CTOD
1980 Chaboche Development of time dependant
fracture interactions between creep and
fatigue.

7
WHAT IS A BRITTLE FRACTURE?

Brittle
Brittle
Ductile
Ductile

Very few fractures are truly brittle i.e. have no permanent deformation.

But fracture is still determined by the energy balance and the energy driving the
cracking process is still the elastic energy stored in the cracked body. Fast
fracture is a more accurate term than brittle fracture to use for rapid failure.

Where local deformation occurs the cracking process is not reversible as it was in
the case of Mica.

Can deal with a great many materials and situations using simple elastic
assumptions. This is known as linear elastic fracture mechanics.

[There is a fundamental flaw inherent in LEFM the calculations assume elastic
behaviour but we know that for the crack to have any chance of growing the
stresses at the tip must vastly exceed the yield stress: yet we carry on anyway!
The point is that in many materials the contribution to the energy balance from
the non-elastic part is a tiny fraction of the total equation. We can put this to one
side for the time being, but will examine this later.]

8
LINEAR ELASTIC FRACTURE MECHANICS

When a crack occurs in a material the local stress around the crack is raised.
LEFM relies on the sufficient of the specimen/component being elastic such that
the energy release rate can be calculated from the elastic displacements around
the crack tip. Hence if you can solve for the elastic stress in any configuration
you can (in principle) calculate G from dU
E
/da.

STRESS CONCENTRATION AT FEATURES

In some simple situations the equations governing elastic deformation can be
solved analytically:

i. Expressing the stresses in terms of complex potentials
ii. Specifying the boundary conditions
iii. Finding functions to satisfy the above

Or, more generally, solving the problem using finite element analysis. One
problem for which there is a solution is that of a circular hole in an infinite thin
plate subject to a stress o
o
.

In polar co-ordinates the stresses
are given by:

o
rr
=
o
o
2
1+
r
o
2
r
2
+ 1+ 3
r
o
4
r
4
4
r
o
2
r
2
|
\

|
.
|
|
cos2u

`

)

o
uu
=
o
o
2
1+
r
o
2
r
2
1+ 3
r
o
4
r
4
|
\

|
.
|
|
cos2u

`

)

o
ru
=
o
o
2
1 3
r
o
4
r
4
+ 2
r
o
2
r
2
|
\

|
.
|
|
sin2u

`

)

Substituting r = r
o
and u = 90 and 0: gives the maximum and minimum hoop
stresses o
uu
, at the edge of the notch as 3o
o
and -o
o
. Thus the presence of a
round hole in the plate increases the tensile stress by a factor of three in one
direction and introduces a compressive stress at the top of the hole equal to the
distant tensile stress.


9
Because all the stresses are elastic and therefore small, the imposed stress fields,
and the solutions for those stress fields, can be added: this is known as the
PRINCIPLE OF SUPERPOSITION.
Hence, adding two stresses o
o
at right angles to each other to produce a 2D
hydrostatic tension and the stresses around the hole in the plate are now:

3o
o
- o
o
= 2o
o
.

Another important situation for which an exact solution exists is that of an
elliptical hole, semi-axes a and b, in a plate, subject to a distant stress o
o
. In this
case the maximum stress is at the tip of the ellipse:

2a
2b
2
oo

o
max
= o
o
1+
2a
b
|
\

|
.
| or

o
max
= o
o
1+2
a
|
\

|
.
|
|

where
a
b
2
= the radius tangential at the tip.

Hence for a long thin crack where a >>b,

o
max
= o
o
2
a
|
\

|
.
|
|

This is slightly modified for a half crack at the edge of a plate by the factor 1.12
because the free surface (zero stress) allows the ellipse to open rather wider than
for the embedded crack.

The factor o
max
/o
o
by which the elastic stress is raised by a feature such as a
crack or a hole is the stress concentration factor k
t
. This is dimensionless.

10
SHARP CRACKS

The above is very useful for finding the effect of features (intended or
unintended) in the structure, but most cracks are long and have sharp tips.
These can be of atomic dimensions in brittle materials.

In 1939 Westergaard solved the stress field for an infinitely sharp crack in an
infinite plate. The elastic stresses were given by the equations;

o
yy
=
o
o
ta
2tr
cos
u
2

`
)
1+sin
u
2

`
)
sin
3u
2

`
)

(

(

o
xx
=
o
o
ta
2tr
cos
u
2

`
)
1sin
u
2

`
)
sin
3u
2

`
)

(

(

t
xy
=
o
o
ta
2tr
sin
u
2

`
)
cos
u
2

`
)
cos
3u
2

`
)

+ similar expressions for displacements u

[Equations for the polar stresses as a function of r and u are in the data-book.]

All the equations separate into a geometrical factor and the stress intensity
factor:

K = o
o
ta

K determines the amplitude of the additional stress due to the crack over the
whole specimen, but particularly at the crack tip where growth has to occur.

When u = 0 the stress opening the crack has the value :

o
yy
=
o
o
ta
2tr
=
K
2tr

The value of K at which fracture occurs is the material-dependant

Fracture Toughness:

K
Ic
= o
f
ta

For a fixed stress this defines the maximum stable crack length or for a fixed
crack length the maximum stress.

r

11
You have come across K in 1A and 1B: Be careful, there are a number of
parameters K:

k
t
=
o
max
o
o
stress concentration factor (dimensionless)

K = o ta stress intensity factor Pa m

K
Ic
= o
f
ta critical stress intensity factor Pa m
or Fracture Toughness

The equations indicate an infinite stress at the crack tip when r = 0. This is not a
problem as the stored elastic energy forms a finite interval. A small volume at
the crack tip will be above the yield stress and thus in a plastic state.

OTHER MODES OF FAILURE PRINCIPLE OF SUPERPOSITION

The above equations considered only a stress normal to the crack surface but
much more complex states of stress will exist at cracks. These can be resolved in
to three distinct crack opening modes, termed with extraordinary imagination,
modes I II and III. Combinations of these can describe any state of stress and
the stresses are additive as they remain elastic.
For example the mode II stress equations include the factor

o = K
II
tr , for any
location around the crack tip since the stresses are additive, the values of K from
the separate crack modes are also additive.

Crack opening modes I, II and III.

The energy release rate is given by integrating stress strain with respect to r,
and has the value:

G =
K
2
E
For plain stress, or

G =
K
2
E
(1u
2
) for plane strain.

Hence, because the values of K for each opening mode can be assessed
independently and then added, it is possible to assess complex multimode
cracking modes.

The total change in energy in the body as a whole can be expressed directly in
terms of the individual stress intensities which characterise the crack tip stress
and displacement fields. The total energy release rate is given by the expression:

12

EG=K
I
2
+K
II
2
+(1+u)K
III
2
For plane stress or:

EG =(1u
2
)K
I
2
+(1u
2
)K
II
2
+(1+u)K
III
2
For plane strain.

Note: These equations do not include the background stress which must be
added.
oys
oo
K dominated
Overall stress
r
Plastic zone

Diagram showing the net stress resulting from the remote stress and the stress
intensity . For o
o
<< o
ys
the plastic zone is dominated by the stress concentration
effect of the crack.

ANGLED CRACKS USE OF THE SUPERPOSITION PRINCIPLE

So far all the cracks we have looked at lie perpendicular to the principal stress. A
simple example of a crack lying at an angle to the principal stress illustrates the
use of the superposition principle. A crack lies at | = 45 to the principal stress
o
o
in an effectively infinite thin sheet (i.e. in plane stress). We want to find the
angle u at which the crack will propagate under a sufficiently high stress o
o
.

[Note that the stress equations as a
function of u relative to the frame of
reference of the main crack are being
used to calculate the local stresses at a
tiny crack taking off at an angle u from
the end of the main crack. The stress
intensity K for a crack continuing in the
same direction is not a function of the
angle u.]

13
The stress on the crack inclined at 45 can be resolved into a component acting
perpendicular to the crack, i.e. in Mode I, and a component acting parallel to the
crack plane, i.e. in Mode II. The stress fields from each can be considered
separately and then combined to give the overall energy release rate for the new
crack. The path the crack takes in propagating further will be that which
maximizes the total energy released. We can find this by differentiating the
energy release rate with respect to the angle u.

The radial stresses for Mode I and Mode II growth are given below using polar co-
ordinates because we will be looking to define the angle of propagation of the
crack. (Note the trig. expressions below are given in the data book)

Mode I:

o
rr
o
uu
o
ru
|
\

|
.
|
|
|
=
K
I
2tr
( )
1 2
cos(u / 2)[1+sin
2
(u / 2)]
cos
3
(u / 2)
sin(u / 2) cos
2
(u / 2)
|
\

|
.
|
|
|

Mode II:

o
rr
o
uu
o
ru
|
\

|
.
|
|
|
=
K
II
2tr
( )
1 2
sin(u / 2)[13sin
2
(u / 2)]
3sin(u / 2) cos
2
(u / 2)
cos(u / 2)[13sin
2
(u / 2)]
|
\

|
.
|
|
|

The axes for the above equations are located in line with the existing crack. We
have two independent stress fields from the mode I and II stresses on this crack.
We use these stresses to work out what the energy release rate for a small
(virtual) crack taking off at an angle u from the end of the main crack. For the
crack continuing in the same direction u would be zero etc, see diagram above.

We extract the stresses which will cause mode I opening of the virtual crack;
these are the o
uu
values from each of the stress fields.

From perpendicular stress:

o
uu
=
o
o
2
ta
2tr
cos
3
u
2

where the factor 1/2 = cos |
From parallel stress:

o
uu
=
o
o
2
ta
2tr
3sin
u
2
cos
2
u
2

(

(
Total stress opening crack in mode I by adding the above:

o
uu
=
o
o
2
ta
2tr
cos
2
u
2
cos
u
2
3sin
u
2

(

(
Similarly the stress to cause mode II opening comes from the o
ru
components:

o
ru
=
o
o
2
ta
2tr
sin
u
2
cos
2
u
2
+ cos
u
2
13sin
2
u
2

(

(

(

(


14
These stresses convert into K values: K
I
(u)= o
uu
a and K
II
(u) = o
ru
a

The K relates to the very small new crack growing at the end of the main crack.
We now have to find the value of u for which the energy release rate will be a
maximum, and do this by adding the G values for each of the two modes of
opening:

G(u)=(K
I
)
2
/E +(K
II
)
2
/E = o
uu
2
a/E + o
ru
2
a/E

We are only concerned with the angle u and can plot the normalised contributions
for Mode I and Mode II opening combined, normalised by the values at u = 0

Plot of normalised energy release rate for propagation of a crack angles at 45 to
the principal stress direction.

These are plotted above, and it can be seen that the mode I crack opening mode
has a very strong maximum at ~-55 corresponding to a minimum in the Mode II
crack. Nevertheless, the sum of the two, denoted by the bold line, is dominated
by the energy released from Mode I (as is nearly always the case).

It should be stressed that K still remains oa: the inclusion of the angular
function in calculating K is a result of using the stress field from the main crack to
generate the energy release rate of the new crack going off at an angle u.

This illustrates how the principle of superposition works both Mode I and Mode
II cracks could grow given sufficient stress. The K
IC
and K
IIC
values for a
particular material are different and characteristic of that material. In practice
nearly all cracks grow in Mode I this normally generating the highest energy
release rate.


15
FRACTURE OF THIN FILMS

Fracture mechanics is increasingly being employed to very small scale problems
such as the spallation of thin films or the delamination of layered structures. This
is relevant to issues such as oxidation, coatings, composites etc.

Generally a film bonded to a substrate will have a different thermal expansion
coefficient, with the result that there is a residual stress in the coating due to the
coating process, or that one develops during thermal cycling. Assuming that the
misfit between the two does not exceed the yield strain of the material and that
the film remains bonded to the substrate, this can be accommodated elastically.
Far from the free edges the film stress is constant and in two-dimensional plane
strain i.e. there is no traction force between the film and the substrate.

But either at the edge of the film or if the film breaks creating an edge, a shear
stress will develop between the film and substrate which transfers the biaxial
stress from the film to the substrate so that the stress in the film diminishes to
zero at the free edge. This gives rise to a stress state where a Mode II type crack
could propagate along the (weak) interface.

For points close to the edge of the film (relative to the thickness of the film d
f
)
there is also Mode I stress as the edge of the film curls up in response the
varying tension through the thickness of the film. These forces gradually diminish
as the distance from the film edge increases and reach about 90% of the remote
value in a distance comparable to the film thickness (this depends on the relative
elastic properties etc see Suresh: Thin film Materials Ch. 4).

For a film delaminating from a substrate at steady state the energy release rate is
proportional to the film thickness d
f
, and the remote misfit stress squared, o
m
2
:

G =
1u
f
2
2E
f
o
m
2
d
f
where

o
m
=
E
f
1u
f
( )
c
m
for plane strain

Note that the crack length a does not matter if the stress in the film is relaxed as
the film delaminates. However, if some stress is retained, for example by a
continuous film maintaining the stress or a curved front, the driving force will
eventually drop below that needed for propagation of the crack and it will stop.


16

17
In most realistic problems not only is the geometry the elastic properties
of the two materials differ and this greatly complicates calculating the size
and distributions of the stresses. Such problems are best addressed using
finite element modeling.

The mode I stresses can make cut edges particularly vulnerable to spallation but
the relaxation of the substrate can also reduce the value of G.

Breaks or cuts in the film surface lead to a greater stress as the substrate cannot
strain locally to reduce the strain in the film.

If the film delaminates but remains intact there will in principle be no traction
forces at the crack edges unless and until the film actually breaks.

If the film is in compression during some part of the cycle there is a risk of
buckling and thus breaking (this typically occurs in protective oxide films).

18
PLASTIC ZONE SIZE

The equations above indicate an infinite stress at the crack tip when r=0. Thus a
small volume at the crack tip will be above the yield stress and thus in a plastic
state. This has two effects:

1. The deformation occurring in the plastic zone as the crack grows greatly
increases R, the work to propagate the crack.

2. The nominal elastic energy stored in the plastic zone is not released as the
crack grows, but, provided the plastic zone remains small, this is a small
proportion of the integral evaluating the energy release rate. Hence, for
small plastic zone size linear elastic fracture mechanics can be applied to
ductile failure.

How big does the plastic zone size need to be before we need to modify the
energy release rate equation? This occurs when the elastic energy not stored in
the plastic zone represents a sizeable proportion of the total energy release rate
G. Calculating the plastic zone size is not easy, and we rely on a couple of
approximations (Dugdale and Irwin, see Ewalds page 56) to estimate the effect.
They give similar results and so we will look briefly at only one method, that due
to Irwin.

The simplest estimate is made by assuming that the area ahead of the crack tip
where the stress exceeds the yield stress is plastic; (see previous diagram). Thus
ignoring the remote stress, the size of the plastic zone r
p
is:

o
ys
=
K
I
2tr
y
hence

r
y
=
1
2t
K
I
o
ys
|
\

|
.
|
|
2

This, however, takes no account of the redistribution of the stress which would
have been carried by the material at the crack tip which has yielded and can only
carry the yield stress.

We can estimate the error by assuming a plastic zone, width 2r
y
ahead of the
crack tip. The effect of the plastic flow is to open the crack more widely than the
purely elastic response would predict, thus the elastic field of the crack behaves
as if it were Aa longer than it really is. The tip of the virtual crack acts as the
nominal centre for the stress and strain fields resulting from the crack and for
the associated plastic zone.

Diagram showing elastic
stress redistribution as a
result of yielding Irwin
model.
A

19

The extent of the extended plastic zone is defined by the yield stress.

r
y
=
1
2t
K
I
o
ys
|
\

|
.
|
|
2
=
o
2
2o
ys
2
a+ Aa
( )

Where

K
I
= o t a+ Aa
( )
and the new plastic zone size is:

r
p
= Aa+ r
y

Irwin determined Aa on the basis that the average of the nominal stress in the
plastic zone in the plane perpendicular to the stress axis should equal the real
stress, i.e. the yield stress. Then the load is being supported by the cracked
component remains the same with and without the plastic zone. In effect the
area under the stress graph, A, is set equal to o
ys
Aa.

o
ys
Aa =
o t a+ Aa
( )
2tr
dr o
y
r
y
0
r
y
}

o
ys
Aa+ r
y
( )
=
o t a+ Aa
( )
2tr
dr
0
r
y
}

o
ys
Aa+ r
y
( )
=
2o a+ Aa
2
r
y
but

o
ys
2r
y
= o a+ Aa
( )
from above

o
ys
Aa+ r
y
( )
=
2o
ys
2r
y
2
r
y

Aa = r
y
and

Aa =
1
2t
K
I
o
ys
|
\

|
.
|
|
2

and

r
p
=
1
t
K
I
o
ys
|
\

|
.
|
|
2
= 2r
y

Thus the virtual crack tip determining the elastic stress/strain field ends at the
centre of the plastic zone.

Dugdales analysis is rather more sophisticated but also assumes that the crack is
longer than it really is and superimposes point closure forces onto each end of the
crack onto the overall elastic solution for the enlarged crack. The criterion for the
imposed closure stress is that the sum of the closure and remote stresses cancel
at the crack tip removing the singularity. (see Anderson page 77)

Dugdales analysis gives a slightly larger plastic zone size:

r
p
= 0.392
K
I
o
ys
|
\

|
.
|
|
2
instead of

r
p
= 0.318
K
I
o
ys
|
\

|
.
|
|
2
from Irwin.


20
It is not worth worrying too much about these factors as both analysis are
predicated on perfect plastic behavior, i.e no work hardening. In fact materials
will work harden to different extents and would thus be able to sustain higher
loads in the plastic zone than these analyses predict. FE analysis provides a
better method of assessing the plastic zone size for each material from its
particular plasticity characteristics.

REAL PLASTIC ZONE SIZES

We can use this to estimate the error introduced by the plasticity at various
ratios of the stress to the yield stress.

o
ys
MPa K
IC
MPam
1/2
ASM
Crit.
r
p
plane
stress
r
p
Plane
strain
High strength Steel 1200 60
Structural steel

400 150
Alumina

5000 1

Perspex

30 1

For most components the size of the plastic zone is fairly small but concerns must
be raised for the validity of LEFM in the case of structural steels. In practice the
ASM standard requires that the crack length a, the specimen thickness B, and the
residual specimen width of a test-piece are all greater than

2.5
K
I
o
ys
|
\

|
.
|
|
2
.

This means that, in effect, r
p
< a/8 for LEFM to apply. The plastic zone should be
less than 20% of the area dominated by the crack tip stresses (rather than the
remote stresses) which is about 10% of the crack length.

Alternatively we can look at the effect of the plastic zone on the fracture stress

o
f
=
EG
crit
t a+ r
y
( )

(

(
(
or

o
f
=
EG
crit
t a+o
f
2
a / 2o
ys
2
( )

(

(
(

The plastic zone has the effect of dividing by the factor

1+
o
f
2
2o
ys
2
|
\

|
.
|
|

For 4 . 0
ys
f
=
o
o
the error is 4%; for 0.6 the error is 8.5% and for 0.8 the error
reaches 15%. Hence the closer the fracture stress gets to the yield stress the
more ductile the failure and the greater the influence of the plastic zone.


21
REAL SHAPE OF PLASTIC ZONE

The plastic zone is not going to be circular since the largest shear stresses occur
at 45 to the crack (equations page 10). The exact shape is tricky to calculate
and depends on the yield criterion used. Using the Von Mises criterion for yield :

o
ys
=
1
2
o
1
o
2
( )
2
+ o
1
o
3
( )
2
+ o
2
o
3
( )
2
| |
1
2

and substituting the Mode I principal stresses in polar co-ordinates:

o
1
=
K
I
2tr
cos
u
2
|
\

|
.
| 1+sin
u
2
|
\

|
.
|

(

(

o
2
=
K
I
2tr
cos
u
2
|
\

|
.
| 1sin
u
2
|
\

|
.
|

(

(

0
3
= o for plane stress, and

o
3
=
2vK
I
2tr
cos
u
2
|
\

|
.
| for plane strain

we are able to solve for r
p
and obtain the limits of the plastic zone:

r
p
u
( )
=
1
4t
K
I
o
ys
|
\

|
.
|
|
2
1+cosu +
3
2
sin
2
u

(

(
For plane stress

r
p
u
( )
=
1
4t
K
I
o
ys
|
\

|
.
|
|
2
12v
( )
2
1+cosu
( )
+
3
2
sin
2
u

(

(
For plane strain

plotting this gives the shapes for the plastic zone. Note the value for plane strain
will be smaller by some (1-2v)
2
which is 0.16 for v = 0.3. Thus the plastic zone
is of a slightly different shape and smaller in size for the constrained central part
of the crack.

Diagram of the plastic zone and the effect of through thickness crack.
Plane stress at outside edge

Plane strain in centre

22

Plastic Zone shape for Mode I, II and III crack opening, calculated from
von Mises yield criterion.

Similarly the plastic zone size and shape can be derived for the other crack
opening modes and these are shown in the above Figure. In general the most
likely cause of crack growth is mode I opening, and consideration of this is able to
solve most problems.

Again it must be emphasized that the exact solution depends on the plasticity of
the material and that there is a gradual transition from plane stress to plane
strain. A high work-hardening rate reduces the plastic zone size as more stress
can be sustained by the plastic material. When the plastic zone size becomes
comparable with the thickness of the specimen, plain strain is not achieved at the
centre of the crack. However, provided the plastic zone size is small compared to
the thickness the stress intensity factor K
Ic
provides a reasonable fracture
criterion.

As the thickness decreases the measured K
Ic
increases from a plane strain
plateau value to a higher value characteristic of plane stress. Thus to define K
Ic
a
small plastic zone size and plane strain conditions are required. But use can be
made of LEFM in situations of plane stress i.e. thin plates, provided the values of
K
Ic
that are used are found in material of similar thickness, In these
circumstances K
Ic
is not a material constant as it varies with the dimensions of
the specimen.


23
KIc
Specimen Thickness
Plane strain Plane stress

Diagram showing the effect of specimen thickness on the critical stress intensity.

The constraint at the centre of a thick sample causes the crack to progress the
furthest at the centre of the crack and the sides fail by plastic shear forming two
lips which will point up or down randomly as in the cup and cone fracture. The
centre part of the crack will be normal to the tensile axis on average, (this masks
valleys and ridges on a smaller scale). As the load on the sample increases the
plastic zone size increases and the width in plane strain decreases. Eventually
the plane stress conditions extend across the sample and a diagonal shear failure
results.

This leads to the kind of fracture surface seen below where the crack starts at a
notch propagating by ductile cleavage at right angles to the stress but as the
stress increases the area of plastic shear failure gradually takes over.

Notch
Shear Failure - 45
Square failure - plane strain

24
R AND G CURVES:

The material resistance to crack extension, R, consists of the energy to create
two new surfaces, 2
s
together with any mechanism which absorbs energy as the
crack grows. In the case of brittle fracture R does not depend on the size of the
crack, but where plastic work is done developing a plastic zone R may well vary
with the crack size, increasing or decreasing. The increase could result from an
increase in the plastic zone size as we saw on the previous page. Initially the
constraint due to the thickness of the specimen inhibits plastic flow, restricts the
size of the plastic zone and keeps R low. As a plastic zone develops at the sides
of the sample R increases reducing the area of ductile cleavage until the entire
crack fails by shear. At this point R reaches a maximum value.

[Alternatively, a decrease could result from the strain rate sensitivity of the flow
stress reducing the plastic zone size as the crack grows faster.]

G varies with the size of the crack and the geometry of loading. For fixed grips
the load drops as the crack extends and thus the energy release rate, G, will
drop. But for the same specimen at fixed load, G increases as the crack grows.

G
a
LOAD CONTROL
STRAIN CONTROL
FIXED GRIPS


25
MEASURING G:

Consider two simple situations, a fixed strain where a growing crack reduces the
load (strain control) and a fixed load where the crack growth increases the length
of the specimen (load control).

G =
1
B
dU
da
|
\

|
.
|
u
for strain control and

G =
1
B
dU
da
|
\

|
.
|
P
for load control.

*Note U = potential energy and u = displacement and P = load.

Consider a plate, thickness B, loaded with a force P. This contains a crack length
a and as a result of the crack the plate has extended a distance u. The crack
extends by oa. Under load control the specimen lengthens by ou, and the work
done by the external force is oU
F
= - Pou. The extra work stored elastically by
virtue of the change in crack length and the consequent change in specimen
length oU
E
= 1/2Pou. Thus half the work done is stored in the regular way as in
an un-cracked body and the rest is released as the elastic response of the body
changes as a result of the crack growth. Under strain control the load is reduced
by oP and the energy released: oU
F
= -1/2uoP as no external work is done (oP is
negative).

LC:

dU
E
=
1
2
Pdu Pdu =
1
2
Pdu SC:

dU
E
=
1
2
udP
LC:

GBoa = +
1
2
Pou SC:

GBoa =
1
2
uoP

We now introduce the Compliance: the inverse stiffness C = u/P.

LC:

G = +
P
2B
du
da
|
\

|
.
|
P
= +
P
2B
du
dC
|
\

|
.
|
P
dC
da
|
\

|
.
|
p
=
P
2
2B
dC
da
|
\

|
.
|
p

SC:

G =
u
2B
dP
da
|
\

|
.
|
u
=
u
2B
dP
dC
|
\

|
.
|
u
dC
da
|
\

|
.
|
u
=
P
2
2B
dC
da
|
\

|
.
|
u

The expression for G is the same in both cases.

The compliance depends on the specimen shape, in particular on the crack
geometry and length, remember the sample is assumed to be elastic at all points.

By measuring the compliance as a function of the crack length the energy release
rate can be calculated from the load P.


26
Lets look at this graphically: for a specimen under strain control (the grips are
fixed) the crack growth causes a fall in the external force P which is equal to the
energy released by the crack in growing oa. This is equal to the area of the
shaded triangle OAC.

a
P
P
u
Pdu
du
dUE = 1/2Pdu
a
a+da
Fixed Load
A B
O
C

For Load control, the specimen extends at fixed load and the energy released is
the area of the triangle OAB. Thus the only difference between the two cases is
the area of the triangle ABC which is of the order 1/2oPou and approaches zero in
the limit.

Thus the value of G depends only on the geometry of the sample: shape, crack
length etc, and the loading, P.
-
-

27
MEASURING R:

For brittle materials R does not change as the crack grows and failure
occurs when the stress rises to the point where G equals R.

The R curve can be measured from a plot of load P against extension u, using the
gradient of the unloading line at any point to give the compliance as the crack
extends.

G =
P
2
2B
dC
da
|
\

|
.
|
u

For a rising R curve G must exceed R at any crack length, but as the crack grows
R can exceed G. Hence, for fast fracture, G must increase with the crack length
faster than the resistance to crack growth. Fast fracture will occur when dG/da >
dR/da. If dG/da = dR/da the crack will continue growing in a controlled manner
(so-called stable crack growth).


28
MEASURING K
Ic

In principle k can be measured from the load at failure and the crack length in a
standard sized specimen containing a sharp crack grown usually by fatigue.
However, for the test to be valid three criteria must be satisfied:

- the specimen must be large enough for the plastic zone size to be a small
proportion of the sample and we have the criterion for the dimensions a, B
and W discussed earlier: a, B and W

2.5
K
I
o
ys
|
\

|
.
|
|
2

- The maximum fatigue stress intensity K is less than 80% of K
Ic

- the crack is still roughly in the middle of the sample, 0.45 a/W 0.55.

If the testpiece were entirely elastic and the load displacement curve would be
linear, it is generally not as the tip of the crack begins to yield. The value of the
load, P
Q
, to be used to assess K
IC
is taken as the point at which the curve crosses
a line drawn with a gradient 95% of the initial tangeant. Sometimes there is a
small amount of unstable crack growth prior to failure at a higher load, pop-in
behaviour. In this case or if the sample fails before a 5% deviation from linearity,
the pop-in stress or the ultimate stress prior to failure are used.

The provisional value of K
Ic
, K
Q
can then be calculated from the equation:

K
Q
=
P
Q
B W
f a / W
( )

where f(a/W) is a dimensionless function of the specimen dimensions specific to
the testpiece design. These are all set out in the ASTM standard E399. As an
example, for the most common compact specimen testpiece the equation is:

f a / W
( )
=
2+a / W
1a / W
( )
3 / 2
0.866+4.64 a / W
( )
13.32 a / W
( )
2
+14.72 a / W
( )
3
5.6 a / W
( )
4

(

(


29
ELASTIC PLASTIC FRACTURE MECHANICS

The requirements for the minimum specimen test-piece size for LEFM to be valid
are very stringent for ductile materials. In fact the size of test-piece needed to
produce a valid and representative value of K
Ic
are such that large amounts of
material and huge machines are required for testing. More importantly, the scale
could well exceed the size of the component the results are to be applied to.
Under these circumstances we still need a measure of the fracture toughness of
these materials in order to predict and avoid possible failure. Two methods have
been developed which enable small scale testing to be applied to the failure of
ductile materials. These are the Crack Opening Displacement and the J Integral
method.

CRACK TIP OPENING DISPLACEMENT

Back in 1961 Wells had been trying unsuccessfully to obtain reliable K
Ic

measurements for ductile steels, when he noticed that the crack tips showed
considerable blunting which increased with the toughness of the material. He
proposed measuring the critical diameter of the crack tip and using this directly
as a measure of the toughness. We will see that for limited plastic zone size the
crack tip opening is related directly and simply to the LEFM energy release rate,
but the really useful extension of this to a much larger plastic zone size was at
that point purely empirical. It has since been demonstrated rigorously that the
use of the CTOD is valid even for very extensive plasticity and the method is now
widely used to test and design components.

Additional crack opening as a result of plasticity at crack tip.

We saw earlier that the effect of a plastic zone at the crack tip is to extend the
effective length of the crack by r
y
~ half the diameter of the plastic zone. Hence
the opening of the crack at its real tip can be approximated from the calculated
elastic displacements of the virtual (extended) crack evaluated at a point some r
y

from the virtual crack tip. See Figure above.


30
The CTOD o is given by double the displacement u
yy
in the tensile direction, for
plane stress this is given by the equation:

u
yy
=
K
I
2
r
2t
sin
u
2
|
\

|
.
| k +12cos
2
u
2
|
\

|
.
|

(

( where

k =
3 v
1+v
for plane stress

evaluating this at r
y
from the crack tip u = 180:

u
yy
=
K
I
k +1
( )
2
r
y
2t

and substituting for the plastic zone size from the Irwin value (second estimate,
page 18):

r
y
=
1
2t
K
I
o
ys
|
\

|
.
|
|
2
gives:

u
yy
=
k +1
( )
2
K
I
2
2to
ys

where

k +1
( )
2
=
4
1+v
( )
1+v
( )
E
=
4
E

and hence

o = 2u
yy
=
4
E
K
I
2
to
ys
=
4
t
G
o
ys
where G is the energy release rate.

Again the Dugdale model gives a similar result:

o =
G
mo
ys

where m is a constant 1 for plane stress and 2 for plane strain.

Remember that this is all derived from the elastic solution surrounding a small
plastic zone (page 10) but it has since been demonstrated from plasticity theory
that this is generally true even if the plastic zone is extensive. The critical value
of the CTOD thus gives a reliable measure of the fracture toughness of the
material. Clearly this will be a function of the specimen thickness but provided
the thickness of the test-piece is similar to the component the test result can be
used.


31
MEASURING CTOD

This is very difficult to measure directly and is usually inferred from the width of
the crack opening V of a three point bending specimen. It is assumed that the
specimen behaves as a rigid hinge pivoting about some point in the uncracked
ligament of the specimen the displacement o is then proportional to V:

o
W a
( )
=
V
W a
( )
+a

where is a dimensionless constant between 0 and 1.

CTOD measured from a three point bend specimen.

Painstaking experiments measuring the value of V and then o by sectioning the
crack established this relationship. But beware - it depends on the specimen
thickness and the width of the slot and the length of the crack.

There are four values of o recognised by the ASTM standards:

- o
i
the CTOD at the onset of stable ductile crack growth.
- o
c
the CTOD at the onset of unstable cleavage failure,
- o
u
the CTOD at the onset of unstable crack growth following extensive
ductile stable crack growth
- o
m
the CTOD at maximum load where the specimen does not break.

The first is hard to detect; the only clue in the load curve being a slight change in
gradient. The next two are identified by the failure of the sample and the last by
a maximum in the load curve without the failure of the sample.

W a


32
V
LOAD
P
Vc Vi
Vm
Vi
Vu
cleavage
stable crack
growth +
cleavage
stable crack
growth +
plastic
collapse

Mouth Opening Displacement v Load curves.

J INTEGRALS

The J integral is the equivalent of the G for the elastic-plastic case. It is the rate
of energy absorbed per unit area as the crack grows; it is not however the energy
release rate because the plastic energy is not recoverable as it would be in the
elastic case. The definition is:

J =
dU
dA

where U is the potential energy of the system and A the area of the crack.

Energy release rate for non-linear deformation.

An analogy with the Linear elastic case can be made; compare the Figure above
with those on page 25. The stress strain curve is no longer linear, but the area
under the curve represents the work done in extending the cracked body (without
extending the crack).


33
Plotting two curves for specimens differing only in the length of the crack, a and
a+Aa, the energy required to grow the crack is the difference in the areas under
the two graphs shaded in the Figures on page 25. Since the area decreases as
the crack grows dU/da is negative and J =-dU/da at unit thickness. Although this
is the same as the definition of the energy release rate we used earlier, the J
integral for the plastic case does not represent the energy released as the crack
grows because much of the energy used performs plastic deformation. This is
fine so long as you are just loading the specimen but becomes tricky if you try
and reverse the stress.

The term J integral comes from the property of J which can be expressed and
evaluated as a closed line integral around the crack tip. J is the strain energy
density within the line minus the surface integral of the normal traction stress
forces normal to the surface defined and is independent of the path the integral
takes.

y
x

Diagram showing the line integral around the crack tip J integral.

It can be evaluated experimentally by measuring the stress strain curves for a
number of identical specimens containing cracks of different lengths and plotting
the area under the graph U for each specimen as a function of the crack length
and thus evaluating dU/dA and hence J. There are also specific specimen
geometries (deeply double notched and nothed three point bending specimens)
that allow J to be measured from a single specimen.

These experiments allow J to be plotted as a function of the crack extension.
Thus although J is defined in similar terms to the energy release rate G, and
indeed reduces to G for linear elastic behavior, J for elastic-plastic materials is
closer to R, the resistance to crack growth, in both interpretation and form. The
curve plotted against the crack growth from the original crack length a, shows
three distinct regions; an initial zone where the original crack blunts but does not
grow and the curve rises steeply, a secondary region initiating at J
Ic
, where a new
crack nucleates and grows developing the elastic-plastic zone at the crack tip,
until finally steady state crack tip conditions are achieved and the crack
propagates at a constant value of the J resistance J
R
.


34
JR
Crack
blunti ng
Fracture
Initiation
Steady
state
crack
growth
Aa
A
C
B
A B C

Diagram indicating the J curve during crack growth.

The validity of this approach has limits, just as the LEFM has. These are reached,
in general terms, when the extent of plastic yielding becomes a large proportion
of the remaining ligament length. At this point a single parameter for crack
growth is not sufficient and even more complicated analysis is necessary.


35
FRACTURE MORPHOLOGY

DUCTILE FAILURE:

What do we mean by ductile failure? You are familiar with ductile failure in uni-
axial specimen characterised macroscopically by cup and cone failure and on a
microscopic scale by the formation and coalescence of voids generally nucleated
at second phase particles. This occurs after the point of plastic instability has
been reached when the rate of work hardening can no longer compensate for the
increase in the stress as the section decreases. Voids nucleate and grow most
rapidly in the centre of the sample where the state of triaxial stress exists. These
grow and coalesce to produce a circular internal crack which grows, and finally
fails by shear in the plane stress outer regions of the sample. Where void
formation is difficult, (for example in pure metals) much more ductility is
observed and the sample can thin almost to a point before failure occurs.

Diagram showing cup and cone failure in tensile specimen

Voids almost always nucleate at second-phase particles either by decohesion at
the interface or by fracture of the second phase or inclusion. A number of models
have been developed which look at the effect of dislocation pile-ups at second-
phase precipitates formed during plastic flow as the trigger to void nucleation but
fail to predict the observation that voids appear to nucleate most readily at larger
particles. This is not entirely surprising because the largest precipitates are likely
to be those with the highest interface energy and thus the largest incentive to
reduce surface to volume ratio, and, in addition, are also those most likely to
crack under extensive plastic flow in the surrounding matrix. This latter process
is the most likely to occur where large precipitates are present and can be readily
observed.

The 45 sides of the cone fail last as the central crack propagates outwards. In
the absence of general yielding across the full remaining section of the sample
the progress of a crack by ductile means relies upon the nucleation and growth of
voids ahead of the crack tip. The stress ahead of the crack tip is raised to about
4 times the stress at approximately two times the crack tip opening
displacement or CTOD from the tip. Voids form in this area of raised stress
ahead of the crack tip.


36

Once formed, the voids grow, becoming elliptical and undergoing extensive
plastic flow at the sides. The ligaments between the voids fail by shear on the
plane of highest shear stress at 45 to the tensile axis.

CLEAVAGE FRACTURE IN DUCTILE MATERIALS.

The cleavage fracture surface is characterised by a planar inter-granular crack
which changes plane by the formation of discrete steps. Facets correspond to the
individual grains and in single crystals an entire slip plane can consist of one
facet.

Facetted brittle failure showing river lines.

The steps or river lines on the facets converge and eventually disappear in the
direction of crack growth. They are formed at a grain boundary where the
cleavage plane in one grain is not parallel to the plane in the adjacent grain; the
difference being accommodated by a series of steps. These gradually diminish as
the crack propagates adopting the cleavage plane of the new grain before being
re-formed at the next grain boundary.


37
If a cleavage crack is to propagate across a grain boundary distinct new cracks
must be nucleated ahead of the interface before sufficient plasticity in the
material is achieved to relieve those stresses.

Conditions favoring brittle fracture are:
- high yield stress,
- reduced slip systems (HCP and BCC metals, low temperature),
- high constraint (plane strain) and rapid deformation.

However, for metals, in particular for iron , it has been shown that the fracture
stress follows the value of yield stress measured in compression (even though in
tension the material demonstrates brittle failure). For small grains sizes yielding
precedes failure, at larger grains sizes the two occur together.

At the tip, the crack becomes blunted through plasticity and thus the potentially
very high stresses are reduced (see next section). As a result the stresses
achieved ahead of the crack tip do not in effect exceed 3-4 times the yield stress.
This is way below the theoretical strength of most materials:

t
~ o
E
c

Hence the crack cannot simply propagate as it would in a brittle ceramic. (e.g.
the wedging discussed on page 5. There must be a crack or defect ahead of the
crack to further raise the stress and propagate the crack if cleavage is to occur.
Under conditions of plane strain i.e. constraint, the critical length for a crack from
the Griffiths criterion is:

a
crit
=
2E
s
t 1v
2
( )
o
f
2
= 0.3m

where, for example in iron, o
f
= 1GNm
-2
and E = 200GNm
-2
, and
s
= 2Jm
-2
.

Hence some plasticity at the crack tip is necessary to form cracks of roughly this
size in order to propagate the crack further. A number of mechanisms by which
micro-cracks can form have been proposed and are illustrated on the next page.

The micro-crack is limited to a single grain due to the difficulty in propagating
across the boundary. Hence the stress intensity ahead of a micro-crack is limited
by the (grain size), this limits the stress to nucleate further cracks and
propagate the failure. This results in a Hall-Petch type relationship between the
failure stress and the grain size:

o
f
~
tE
gb
1v
2
| |
d
|
\

|
.
|
|
1
2

where
gb
is the plastic work to propagate across the grain boundary and
generally exceeds the usual
p
term.

There are other mechanisms by which grain refinement to affect the fracture
stress; in mild steels the cleavage fracture is controlled by the fracture of grain
boundary carbides, and an increase in the overall grain boundary area with
smaller grain size leads to smaller carbides and thus a higher fracture stress.
Grain size is hence the one of the best strengthening mechanisms as it increases
both strength and ductility.

38


39
BRITTLE DUCTILE TRANSITION.

Macroscale:
The brittle ductile transition represents the change from general plastic yielding
to the propagation of a distinct crack this so-called brittle failure can be very
ductile and the fracture surface show evidence of extensive plasticity.

The brittle ductile transition is governed by the macroscopic yield in the
specimen, not what is going on at the crack tip. Hence values depend, within
limits, on the particular geometry of the specimens. Tests such as the impact
test of which there are several standards (Charpy, Izod etc) provide relative
rather than quantitative data. They are nevertheless extremely useful as they
are quick and simple to perform can be compared with reference data to provide
excellent quality control.

If the energy absorbed by rapid failure is plotted against the temperature for
steels a transition is observed from a high to a low value over a limited
temperature range.
Temperature
E
n
e
r
g
y

a
b
s
o
r
b
e
d

% Cleavage failure
Energy absorbed
FATT
NDT

Two of the transition temperature defined are: the nil ductility temperature where
the curve just begins to rise, and the fracture-surface appearance transition
temperature, FATT, based on 50% of the surface being cleavage failure. The
former corresponds to the point at which general yield occurs throughout the
remaining width of the sample.

Factors promoting cleavage failure are:
1 high yield stress large amount of stored elastic energy
2 large grain size large build up of stress from pile-ups
3 coarse carbides can crack
4 deep notches - constraint
5 thick specimens (plane strain).

40
At the nano-micro scale:

Fracture of very small components is crucial to the development of small devices
and it is here that much interest in fracture is currently focused. Here plasticity is
also crucial, particularly in materials with limited dislocation mobility (Si, Ge, Fe,
Cr, Al
2
O
3
, and inter-metallics) essentially everything other than fcc metals. All
these materials display very brittle behaviour at low temperatures and a
transition to a more ductile behaviour as temperature rises.

Rice introduced the concept that brittleness was determined by the competition at
the crack tip between the generation of dislocations in the very high stress field
at the crack tip and cleavage. His paper of 1974 explains the issue very lucidly
(skip the mathematics in the middle) J.R. Rice and R Thomson, Phil Mag 29, 1,
p73, (1974), with a more modern interpretation given by J.R. Rice, Journal of the
Mechanics and Physics of Solids, V.40, Iss.2 p.239-271 (1992).

This is demonstrated by a series of experiments performed by Prof Steve Roberts
on pure iron single crystals. (Acta. Mat. 56 (2008) 5123)

- 4Pt bending with pre-cracked single crystals of specific orientation (2 slip
planes at 45 to the crack tip)
- Strain rate varied from 4 x 10
-3
to 4 x 10
-5
s
-1

- K
Ic
calculated from failure stress and geometrical factors
- DBT indentified from examination of the fracture surface and evidence of
slip bands
- Plotting 1/T
DBT
against strain rate shows an Arhenius relationship
- Activation energy correlates very well with that for dislocation movement
The DBT decreases from 130K at the lowest strain rate to 154K at the highest.
The observed behaviour can be modeled very accurately by dislocation
dynamics. This means calculating the distribution and movement of dislocations
during the test from their initial positions, the complete stress field and an
exponential equation for dislocation velocity.
Essentially the DBT occurs when the shielding effect of the dislocations on the
two slip planes (i.e. the elastic stress fields from those generated) reduces the
stress at the crack tip sufficiently rapidly to prevent the stress at the tip reaching
the cleavage stress.
Part II Materials C15 Michaelmas 2010 41

FATIGUE

Fatigue is damage (usually failure) caused by oscillating stress below the fracture
stress. 90% of all mechanical failures can be attributed to fatigue. Paradoxically,
although the stress is below the yield stress, fatigue is essentially concerned with
the generation of defects by plastic flow and the movement of dislocations.

Time
Stress
omax
om
omin
0
Ao
om
oa

The diagram above defines some of the variables used to describe a fatigue test
run under stress control: the stress range A o
a
, mean stress
o
m
. the R ratio R = o
min
/o
max
.

Similar definitions apply to tests where the strain on the sample is controlled and
the maximum stress may vary through the test.

Real fatigue situations cover a baffling range of variables; examples include high
frequency mechanical fatigue for example in a crankshaft, to low frequency
pounding of a north-sea oil rig structure in a highly corrosive environment, to
thermal fatigue caused by the periodic heating and cooling in the turbine of a
transatlantic jet engine. We need to understand fatigue so that we are able to:

i) predict the engineering life of these components,
ii) design structures and materials which maximise economic life.

Factors affecting fatigue which we will consider in varying degrees of detail are:

- Mean stress o
m

- Stress amplitude o
- Frequency
- Waveform
- Temperature
- Temperature variation
- Environment corrosion and oxidation
- Surface finish
- Coatings
- Microstructure


Test procedures have been developed which address these variables and by the
use of a number of mostly empirical laws these are able to provide some degree
of predictability in most situations. Fatigue conditions fall into a number of
regimes:

High Cycle Fatigue HCF: Low amplitude stresses induce primarily elastic strains
which results in long life, i.e. endurance in excess of 10,000cycles

Low Cycle Fatigue LCF: Considerable plastic deformation during cyclic loading
results in an endurance limit below 10,000 cycles and behavior dominated by
plastic deformation.

Thermo-mechanical Fatigue TMF: varying both stress and temperature to
give strain cycles in phase, out of phase (and all things in between) with the
temperature cycle.

APPROACHES TO FATIGUE

We can break Fatigue in ductile materials into several stages:

1. Initial micro-structural changes leading to the nucleation of permanent
damage
2. Nucleation of the first micro-cracks
3. Growth and coalescence of these flaws to produce a dominant crack.
4. Stable propagation of the dominant crack.
5. Failure

Macroscopically there are ambiguities in defining the initiation and growth stages
of cracks depending on the resolution of the techniques being used to
investigate. Generally stages 1-3 constitute crack initiation and stages 4-5 crack
growth.

Depending on the conditions, these stages occupy widely differing fractions of the
sample life and thus require different strategies to determine life. The method
adopted also depends on the consequences of failure.

TOTAL-LIFE OR SAFE-LIFE:

This strategy is to predict the total life and retire the component at a fixed
proportion of this, to include a considerable margin for error. The aim is to retire
the component before a crack forms and it is used where fatigue failure would
result in component failure. Total-life can be wasteful as much useful life remains
unused where the scatter in the data is large.

This approach focuses on predicting the number of cycles to failure, N for an
initially un-cracked specimen. This is most appropriate where the initiation of the
dominant crack occupies the majority of the total life (as much as 90%). For HCF
where the stress range is low and the stresses principally elastic, the stress range
is used to characterise the component and produce a reference S-N curve. For
higher stresses resulting in LCF plastic strain is extensive and the strain range is
typically (but not always) used.

DAMAGETOLERANT OR FAIL-SAFE:

This approach recognises that all structures contain defects and that these grow
at stable and predictable rates. The strategy involves periodic inspection of the
structure and repairs or replaces components as cracks are found. This is
generally used where failure would not result in component failure due to
structural redundancy. A greater proportion of the useful life is used and the risk
of wrong assumptions in the predictive process are dimished.

Thus if the maximum size of the initial defects in the structure is known (a
max
) the
interval between inspections is determined by the time predicted for this crack to
achieve critical size (t
1
), (we will quantify this later) The component may survive
several iterations (two in the case below) before being replaced.

LEAK BEFORE BREAK:

A special case of the fail-safe approach widely used for pressure vessels and
pipes. The thickness and properties of the vessel are arranged so that a through-
thickness crack does not propagate catastrophically. This means that the crack
will be below the critical size for the stress on the vessel. Such a leak can be
detected and repaired without the severe consequences of the rupture of the
vessel.

TOTAL LIFE APPROACH

If we perform a series of tests at varying stress ranges and plot the number of
cycles to failure the life increases as the stress range decreases. Some materials
(typically low alloy steels and Titanium alloys) show an asymptote to a fatigue
limit, otherwise (high alloy steels and aluminium), an endurance limit is set.

Ao
ln N
Fatigue limit
10 Endurance limit

S-N curve
7

BASQUINS LAW

The curve can be approximated by an empirical expression due to Basquin:

( )
b
f f a
N 2
2
o' = o =
o A
N
f
is the number of complete cycles to failure.
where o
f
is the fatigue strength coefficient o
f
the static fracture strength and b
takes the value 0.05 to 0.12 for metals.

COFFIN MANSON LAW.

Under conditions of high plastic deformation we have low cycle fatigue conditions
and for strain controlled tests, Coffin and Manson independently noted an
empirical relation very similar to Basquins law.

The total strain amplitude can be split into plastic and elastic components:

2 2 2
p
e
c A
+
c A
=
c A

where the plastic component is linear when plotted against the log (number of
load reversals), 2N
f
:

( )
c
f f
p
N 2
2
c' =
c A

Here c
f
is the fatigue ductility component and roughly equal to the failure ductility
in tension, and c takes the value 0.5 to 0.7 for metals.

Adding in the Basquins law for the elastic (high cycle fatigue) component we
have:
( ) ( )
c
f f
b
f
f
N 2 N 2
E 2
c' +
o'
=
c A


Plotting log(Ac) against log (2N
f
) gives two distinct regimes, at low strain and
long life the gradient b (-0.1) dominates, HCF conditions, and at high strain and
short life the gradient is c (-0.5). The transition is gradual but extrapolating the
asymptotes allows a transition number of cycles, 2N
t
, to be identified.
log 2Nf
logAc
c
b
cf
2Nt

Note: fatigue is inherently variable variation in life of 100% is not unusual for
nominally the same test. This is masked by the widespread use of log plots.

The intercepts of the two parts of the curve correspond roughly to:

1. LCF: the total strain, plastic and elastic, at failure.
2. HCF: the elastic component of the strain at failure

Lets put some figures in here:

E o
f
c
f
b c
Aluminium 7075 72GPa 193MPa 1.8 -0.106 -0.690
Steel 0.15%C 210GPa 827MPa 0.95 -0.110 -0.640

Aluminium: ( ) ( )
69 . 0
f
106 . 0
f
N 2 8 . 1 N 2
72000
193
2

+ =
c A

(HCF intercept 666 times less than the LCF intercept - note log scale)

Steel: ( ) ( )
64 . 0
f
11 . 0
f
N 2 95 . 0 N 2
210000
827
2

+ =
c A

(HCF intercept 240 times less than LCF)


TOTAL LIFE APPROACH - COPING WITH FATIGUE VARIABLES

There is a huge number of variables in fatigue far to many to construct S/N
curves for all combinations even if they did not change during the lifetime of the
component. The challenge is to understand how the damage produced by fatigue
varies with these parameters and adds together over a complex life cycle.

The effect of increasing the mean stress is to decrease the fatigue life. Several
relations exist to link the stress range and the mean stress for a given life. The
simplest are linear extrapolations indicating that the sample will fail at the static
yield stress in the absence of a stress range and at the fatigue strain at zero
mean strain.

o
a
= o
a
|
o
m
=0
1
o
m
o
y

`
)
Soderberg: original and most conservative

o
a
= o
a
|
o
m
=0
1
o
m
o
TS

`
)
Goodman relation: good for Brittle materials
conservative for metals

(Other expressions exist giving non-linear extrapolations


GOODMAN DIAGRAM

The effect of mean stress and R value can be expressed on a Goodman diagram
shown below:

TEMPERATURE

It is possible to adjust for temperature where the nature of Fatigue does not
change by normalizing with the yield stress. By plotting o/o
y
fatigue curves from
different temperatures can become very similar.

COPING WITH VARIABLE STRESS - MINERS LAW:

In real situations components very rarely experience constant regular damage.
The level of stress or strain can vary throughout life and the simplest way of
dealing with this is by the use of Miners law. This proposes that the life of a
component experiencing fatigue at various stress amplitudes can be assessed by
expressing the number of cycles at each amplitude as a proportion of total life
and summing the fractions. When the fraction reaches 1, the fatigue life is
exhausted. The order of exposure is not taken into account.

o
time

=
=
m
1 i fi
i
1
N
n
Miners Law

This is useful as a first approximation but has serious shortcomings. The most
important being that no account can be taken of the impact of prior damage on
the later exposure at a different stress (or strain) amplitude. In particular the
balance between crack initiation and crack growth can vary considerably with
stress, thus brief exposure to high amplitude may nucleate damage which at a
lower stress would not occur until a much later stage and thus accelerate the
damage rate at a subsequent lower stress. Conversely, early exposure to low
stress amplitude may strain harden the material and thus prolong life during later
high amplitude exposure. This emphasizes the importance of looking at the
specific mechanisms of damage and how it accumulates in the material.

DEVELPOMENT OF MICROSTRUCTURE DURING FATIGUE

For some situations the loading conditions are controlled by the amplitude of the
strain rather than the stress. This is reflected in the tests which are done under
strain control. These are also most likely to be the conditions where plastic
deformation forms a considerable proportion of the strain, LCF.

Where the strain is kept constant the stress can either increase (cyclic hardening)
decrease (cyclic softening) or stay the same.

Typically materials harden if 4 . 1
ys
UTS
>
o
o
and soften if 2 . 1
ys
UTS
<
o
o

To understand why this occurs we need to consider dislocation microstructure of
the material.

- From the above materials where the initial state is highly work hardened
the dislocation density is high, the effect of the cyclic strain is to allow the
rearrangement of the dislocations into stable networks, reducing the stress
at which the plastic component occurs, and thus the effective stress.

- Conversely where the initial dislocation density is low, (soft material) the
cyclic strain increases the dislocation density increasing the amount of
elastic strain and the stress on the material.

For a given alloy both hard and soft materials tend to a stable dislocation
configuration. For example, detailed work on the development of dislocation
configurations in copper and shows that a stable Laberynth structure develops
(see the figure on page 52 from Suresh, chapter 2)

This results in a hysteresis loop which becomes stable at some point during the
test. When this stable loop is plotted as a function of increasing stress (or
strain) the locus of the maximum values from a series of tests defines a cyclic
stress strain curve.



BAUSCHINGER EFFECT

Strain control first cycle showing the Bauschinger effect.

During cyclic deformation the material can retain a memory of the initial plastic
strain which reduces the stress at which plastic yield occurs in the reverse cycle.

This effect can persist for many cycles and is known as the BAUSCHINGER
EFFECT. This reversible but plastic deformation can occur by dislocation pile-ups
at, for example, incoherent or semi-coherent precipitates exerting a back-stress
which assists plastic yield in compression. This reduces the yield stress in
compression.


The picture above shows a single crystal superalloy CMSX-4 fatigued in LCF at
750C and interrupted at about half the expected life. In this section cut on the
primary slip plane dislocation loop enter the phase precipitates trailing Anti-
Phase Boundary faults under max stress the loops expand contracting as the
stress decreases.


THE EFFECT OF STACKING FAULT ENERGY:

The material response is closely linked to the stacking fault energy since this
governs the ability of the dislocation to cross-slip between planes and thus form
stable cell structures.

High SFE easy cross-slip rapid formation of stable cell structure.

For high SFE materials the cell size is a decreasing function of the strain range
and ultimately does not depend on the starting microstructure. Very low SFE
e.g. Cu 7.5% Al, materials do not form stable cell structures, the highly
dissociated dislocations being arranged in planar arrays where the spacing
depends on the initial state. The microstructures and lives are thus very sensitive
to the prior deformation state.


CYCLIC SOFTENING OF PRECIPITATION HARDENED ALLOYS:

Another cause of cyclic softening of major importance is the cutting of coherent
precipitates in precipitation-hardened alloys. Small coherent precipitates provide
very effective hardening in aluminium alloys and in nickel based superalloys. The
small size maximises the cutting/bowing stress for dislocations in the matrix and
the coherency enhances the stability of these small precipitates. However when a
dislocation does cut the precipitate the fault produced in the precipitates
decreases the stress for the following dislocation since the energy penalty of the
fault no longer applies and indeed may be negative. Thus slip is concentrated in
narrow slip bands cutting the precipitates in two. These smaller precipitates may
dissolve in these areas leaving the un-strengthened matrix vulnerable to high
plastic deformation and early crack formation.

Here cutting of precipitates early in this test (TMF of Nimonic 90) has caused the
to dissolve leaving precipitate free channels in the alloy. The precipitates are
visible from the dislocations wrapped around them
(a) OP (0.4)

STRUCTURAL FEATURES OF FATIGUE CRACK INITIATION:

Here we look briefly at the effect of surface condition, the evolution of damage,
and the effect of coatings and surface treatments on the initiation and
propagation of fatigue cracks.

INITIATION BY DEFORMATION

Fatigue failures can occur at stress of 1/3 the tensile yield stress, yet the
nucleation of cracks requires that there be local yielding. We thus need
heterogeneous nucleation sites for cracks within the structure.

- Pre-existing defects such as inclusions, porosity, surface damage
- Defects generated during cyclic straining for example at stress
concentrations: persistent slip bands, Fracture of carbides, oxidation of
carbides.

In pure metals the major source of fatigue cracks is the persistent slip bands or
PSBs: so called because traces of the bands persist even after surface damage is
polished away. The plastic strain in the PSB is 100 times greater than that in the
surrounding material and results from specific arrangements of dislocations as
parallel walls with relatively low dislocation density between. An equilibrium is
maintained between nucleation and annihilation of mobile edge dislocations
bowing out from the walls. Thus the cyclic strain is concentrated in these zones
leading to their persistence. Although the strain is reversed within the PSB the
distribution is not even and this leads to the formation of intrusions and
extrusions where PSBs intersect the surface. These may act as nucleation sites
for cracks. The initial stages of crack growth therefore often follow the slip planes
and lie at 45 to the tensile axis (see later).

Diagram showing the formation of intrusions and extrusions at a persistent slip band.

In the vast majority of cases fatigue initiates on the surface, however cracks can
sometimes initiate internally at defects cracked precipitates of internal porosity.
The crack then grows under vacuum until it reaches an external surface. This
gives a characteristic circular area on the fracture surface. Once the crack
becomes a surface crack and air is admitted and the stress intensity increases
and the growth rate increases. This sometimes leads to immediate failure.


EFFECTS OF OXIDATION AND CORROSION

The absence of a fatigue limit is normally and indication of material which is
immune from corrosion or oxidation effects otherwise the mere passage of time
will eventually allow the initiation and propagation of cracks even at very low
stress.

At high temperatures oxidation at grain boundaries, Carbides or as in the single
crystal above, in areas between the dendrites where the composition varies

slightly, results in crack initiation. The brittle oxide layer cracks and raises the
stress at the tip sufficiently for the crack to propagate into the substrate thus
allowing further oxidation.

Coatings with different mechanical properties to the substrate can accelerate
cracking by promoting rapid initiation.

CoNiCrAlY coating on IN738 Tested in
TMF 300C 850C

DESIGN AGAINST FAILURE

Where initiation occupies most of the life of the sample, initiation control, any
measure which reduces crack nucleation will extend life.

Surface damage, even scratches can act as stress concentrators and lead to local
plastic deformation and crack initiation. High cycle fatigue is very sensitive to
surface finish and can be extended by polished surface finish.

Corrosion protection to suppress the formation of cracks at interfaces by
preferential attack

Treatments which induce a residual compressive stress in the surface layer will
extend life by reducing the mean stress at the surface and delaying the onset of
cracking. Carburising, Nitriding, shot peening.

Coatings have different stress and/or thermal response to the imposed stress
may crack. The crack can act as a stress concentrator and promote early
cracking. Thermal barrier coatings accelerate HCF failure.

Pores act as stress concentrators and a major source of fatigue cracks in single
crystal superalloys: remove by HIPING.

As with fracture plastic deformation in the matrix can cause cracking of carbides
and the nucleation of cracks.

Crack interface bridging

DAMAGE TOLERANT APPROACH

The damage tolerant approach recognises that crack initiation may occur early in
the life of the sample or that there are pre-existing cracks which grow in a stable
manner through the life of the component.

Following the development of fracture mechanics for monotonic deformation Paris
recognised in the 1960s that the same concepts could be applied to fatigue to
estimate the fatigue crack growth rate and thus predict the time taken for the
crack to reach an unstable size.

If the rate of crack growth is measured and plotted against the AK on a log-log
plot the curve takes the general sigmoidal form shown below.

I: Crack initiation,
crack at 45
following slip planes

II: Crack propagates
at 90 to tensile
axis, striations
formed

III: Final rupture

There are three
distinct regions, an initial stage usually showing a threshold value for AK, a 2nd
stage where the crack growth rate shows a power law dependence on AK only;
and a final stage where the crack growth rate approaches infinity as the AK
reaches AK
Ic
. The central region is the most useful as it allows the CGR for the
major part of the life to be predicted from a knowledge of the conditions at the
crack tip. This equation is known as the Paris Equation.

m
K C
dN
da
A =

where m ~ 4 but can vary from 2-7 for various materials.

This implies that da/dN does not depend on the value of R. This is not strictly the
case particularly for low values of R where the crack closes during the cycle (see
p59).

Note: Minors Law follows directly from the Paris Law see question sheet 2

l
o
g
(
d
a
/
d
N
)

log AK
Fracture

I II III
m
K
Ic

CRACK MORPHOLOGY

The three regions can be identified from the morphology of the fracture surface.

I. Initiation; crack initiates at intrusions and follows slip plane at
approximately 45 to principle stress direction. When the length is
sufficient for the stress field at the tip to become dominant the overall
crack plane becomes perpendicular to the principle stress and the
crack enters stage II.

II. Growth typically showing striations for each cycle and beach marks at
points where conditions changed. Striations may be obscured by
closure damage or by oxide formation at high temperatures.

III. Final failure ductile or brittle rupture associated with fast fracture.

Intrusions and
extrusions
Crack -
Stage I
Crack
Stage II

Crack stage I growth at 45 to stress axis, following persistent slip bands,
during stage II turns to growth normal to stress axis.

Formation of striations by ductile flow during crack growth

WARNING: Crack morphology can differ from this simple formula having more or
fewer stages. Not all fatigue failures show striations or a clear stage I. There
may also be different morphologies as AK or the microstructure change.

CRACK CLOSURE

The plastic deformation at the fracture surface, distortion or oxide growth during
exposure can all act to cause the crack faces to touch before the elastic
deformations of the sample would predict. This means that the stress intensity
range is effectively reduced. Clearly this will have a greater effect the closer the
minimum stress intensity is to zero.

Kmax
Kop
Kmi n
Time
AKeff
Closure

Crack Closure: Effective stress range where crack does not fully close.

The Paris equation can be modified to use a reduced value of the stress intensity
factor AK
eff
.

( )
m
m
eff
K U C K C
dN
da
A = A = where
K
K
U
eff
A
A
=


Mechanisms causing crack closure:

Plasticity Stress induced transformation
Mode II displacement Oxide debris

See 504 Suresh TRIP Steels and Martensitic transformations in Zirconia
containing ceramics.


CREEP
WHAT IS CREEP?

Creep is time-dependant plasticity occurring at stresses below the flow
stress and at temperatures in excess of T
m
/3. Creep occurs by a number
of mechanisms, some well characterised others not, but all of which have
an element of thermal activation which enables plastic deformation at
stresses below those needed to deform the lattice without thermal
activation.

Creep is critical in a number of applications: steels used in power and
chemical plants where the service temperature is high, turbine
components, ice and lead at ambient temperatures.

When loaded with, for example, a constant load a material will stretch
elastically but will also gradually extend plastically. The main concern in
applications is dimensional stability, although prelonged creep will also
lead to rupture. Creep extension plotted against time produces a graph
with a characteristic shape consisting of three distinct regions, we shall
see later however that the particular deformation mechanisms can modify
that shape considerably:

Time
S
t
r
a
i
n
Secondary (or steady
state) creep
Tertiary
Creep
Primary Creep

Primary creep decreasing creep rate as dislocation microstructure
develops to reduce strain rate

Secondary creep - equilibrium is established between deformation and
recovery mechanisms to maintain a steady state strain rate.
(Minimum creep rate may not be entirely constant but occurs in this
region.)

Tertiary creep increasing creep rate as the effective cross section
reduces leading to failure typically logarithmic curve. (Note tertiary creep
can refer to both the final failure stage of creep or to the whole curve
where this has a logarithmic curve.)

CREEP DATA AND ITS USES

Creep tests are conducted at constant load or at constant stress. In the
latter the creep rig is designed to reduce the load on the test-piece in
proportion to the extension, this keeps the load constant provided the
strain is homogenous. Since engineering exposure usually occurs at
constant load most creep tests are done at constant load. For research
purposes (e.g. investigating the effects of stress) there is however much
to be said for working at constant stress.

There are several ways of quantifying material resistance to creep; the
most appropriate depends on the use and loading of the component and
what needs to happen for the component to fail.

Creep strength: - stress to produce nominal strain (normally 1.0 %) in
given time typically 10,000 h (1 year), for aeronautical industry and
100,000 h (11 years), for power plant. Ideally timescale of tests should
be comparable with the timescale of service exposure.

Creep life: - Time to nominal 1% or 2% stress at given stress and
temperature.

Minimum creep rate: measured as a function of stress and temperature.
Tests expensive, focus on accurate value for the minimum creep rate
need to maintain constant stress and temperature over very long periods
of time. Tests can last 100,000 h.

Stress Rupture life: the time to break the specimen at a particular stress
and temperature. Use of the word rupture implies tests done at high
stress these are quick and cheap but the strain is not measured
accurately.

MONKMAN-GRANT RELATIONSHIP

For many materials the steady state regime dominates the creep life and
the creep life is thus mainly dependent on the minimum creep rate, c .
This relationship is expressed as the MonkmanGrant relationship:

dc
min
dt
t
R
=
Eq. 3.1

where is a constant and t
R
is the life to rupture.


TEMPERATURE AND STRESS DEPENDENCE

Empirically the minimum creep rate is well described by the expression:

|
.
|
\
|
|
|
.
|
\
|
o
o
c = c
RT
E
exp
a
n
o
o
Eq. 3.2

where o
o
and
o
c , are creep constants, n takes values from 3-8 at high
stress and 1 at low stress, and E
A
is the activation energy for creep which
is usually the same as the diffusive process by which the creep occurs,
e.g. self diffusion.

The temperature dependence can be explored experimentally using
temperature dip tests. Once the steady state regime has been reached
and the creep rate is stable, the temperature is changed by a small
amount (10-20) and the creep rate re-measured once this has reached a
new steady-state. The activation energy is found in the normal way
taking logs of both sides of equation 3.2 above:

( )
|
|
.
|
\
|
c c
=
2 2
2 1
A
T
1
T
1
/ ln
R E


The value for E
A
correlates well with the self-diffusion coefficient for the
bulk material.

At lower temperatures it correlates with the lower value of the self-
diffusion coefficient for grain boundary diffusion.

LARSON MILLER PARAMETER

If new materials are to be brought into service within a reasonable
timescale it is necessary to estimate the creep life by extrapolating higher
temperature data to much longer times at lower temperatures.
Essentially, the creep test is accelerated by raising the temperature, and
the time difference is accounted for by using the activation energy,
assuming this to be the same as at the service temperature. To do this a
considerable amount of data at a range of stresses and temperatures are
required. To plot this data on a single graph a number of parameters
combining time to failure and temperature have been proposed. We will
look at only one, the Larson-Miller parameter, P.

( ) R Q C t ln T P = + = Eq. 3.5

Where t is the time to rupture (or to some defined strain e.g. 1%) Q is the
activation energy for creep, R is the gas constant and C is a constant for a
particular material and has a value typically between 30-65. A value of 46
can be used for most metals where a specific value is not determined, (a
value of 20 will be quoted if log
10
are used).


If the Larson Miller parameter P is plotted against the stress for all the
tests and the results should lie on a single line reflecting the effect of the
stress of lowering the activation energy for the creep process. (Details of
the derivation are discussed in Dieter section 13-12 p. 459.) Tests
conducted at the same stress but at different temperatures should give
similar LM parameters within the experimental spread. The parameter is
used to present data from a range of tests even if the above assumptions
are not entirely valid.

0
100
200
300
400
500
600
700
800
54000 59000 64000 69000
LM parameter
S
t
r
e
s
s

M
P
a
CMSX-4
RR3000
H
D

Larson Miller parameter against stress for two single crystal superalloys
CMSX-4 and RR3000 together with some experimental alloys H and D.

MECHANISMS FOR CREEP:

Remember that creep is essentially plastic deformation at stresses below
the flow stress. The key is that thermal activation processes allow the
physical barriers to deformation (work-hardening, precipitates etc.) to be
overcome. For this to happen takes time; the higher the temperature and
the closer to the flow stress the more rapid this process will be, hence the
increase in creep rate with stress and temperature mentioned above. The
details of the creep mechanism are as variable as the microstructure of
the material and the nature of the dislocations; nevertheless strong
patterns emerge which are common to many materials.

Creep deformation occurs by the activated movement of dislocations or
directly by the diffusion of atoms to change the shape of individual grains
and thus the macroscopic material (diffusion creep).

We will go through most of these main mechanisms distinguishing
between the dislocation mechanisms and those involving diffusion alone.
For some of them we will derive the temperature and stress dependence.
Creep deformation maps are constructed from these expressions plus a
large amount of experimental data.

Dislocation
plasticity
Atom/vacancy
Diffusion
Herring Nabarro creep
bulk diffusion
Coble Creep grain
boundary diffusion
Thermally
controlled glide
Climb controlled
dislocation glide
CREEP PLASTICITY
Bulk Diffusion
Pipe Diffusion

DIFFUSION CREEP

Herring-Nabarro Creep:

Deformation occurs by the movement of atoms between differently
oriented interfaces under the influence of an imposed stress to produce a
macroscopic shape change. It is intrinsically a very slow process but
comes into its own at stresses too low for dislocation motion to be
activated and at relatively high temperatures where the diffusion is fast
enough to produce a measurable creep rate.

Grains undergoing diffusion creep under tensile stress the material
within the dotted boundaries has come from the predominantly vertical
boundaries. The arrows represent the flux of atoms.

The creep rate can be derived without resorting to arbitrary constants as
follows:

The equilibrium concentration of vacancies C
o
depends on the energy to
form a vacancy E
V
:
|
.
|
\
|
=
kT
E
exp C
V
o
the effect of a tensile stress is to
reduce the energy for the formation of a vacancy by ob
3
hence:

|
|
.
|
\
| o
=
|
|
.
|
\
| o
|
.
|
\
|
=
|
|
.
|
\
| o
=
kT
b
exp C
kT
b
exp
kT
E
exp
kT
b E
exp C
3
o
3
V
3
V


The excess volume fraction of vacancies at the horizontal boundaries is
given by:
(
|
|
.
|
\
| o
= 1
kT
b
exp C C C
3
o o
which for ob
3
<< kT
kT
b C
C C
3
o
o
o
=

(The concentrations C and C
o
are dimensionless)

The flux of vacancies from the horizontal to the vertical interfaces is given
by Ficks 1
st
law:
x
C
D J
c
c
= , hence:
kTL
b C D
J
3
o v
o
=

Where L is the diffusion distance approximated to the grain size L.
This gives the thickness of matter removed from the vertical boundaries
and plated out on the horizontal boundaries per second as a result of the
stress. Hence, the resulting strain rate is given by dividing this flux by the
distance between the boundaries, L, and replacing D
v
C
o
by the self-
diffusion coefficient D:

2
3
kTL
b Do
= c Eq. 3.5

The strain rate is proportional to the applied stress and inversely
proportional to the grain size squared. A more careful analysis of the
diffusion paths leads to an expression where the strain rate is 12 times
that of equation 3.5. This makes sense because L represents the
maximum distance any vacancy would have to travel and the equation
estimates the minimum concentration gradient that would be achieved.
The agreement between this expression and the data is extraordinarily
good considering the expression is derived from first principles and
contains no fitted parameters.

Linear dependence on stress and inverse square dependence on grain size for
creep in alumina fibres tested at 1700C. Note: second graph plots stress for
constant strain rate at each temp.

GRAIN BOUNDARY SLIDING

As the grains change shape relative movement of the grain centres is
geometrically necessary to maintain continuity at the grain boundaries.
This has been demonstrated by looking at displacements of scratches and
fiducial grids following creep. It can be inhibited by precipitates at the
grain boundaries and this may become rate limiting.

**Note the observation of grain boundary sliding does not imply that the
creep mechanism is diffusion creep, it can occur to adjust grain shape
whatever the deformation mechanism of the grains.**

Diagram showing the displacement of a line starting at the same position
relative to the grain centre following diffusion creep.

COBLE CREEP

As diffusion is very sensitive to temperature, at lower temperatures the
main diffusion path is along the grain boundaries since the activation
energy for grain boundary diffusion is considerably less than that for bulk
diffusion. The effect on the expression is to replace the self-diffusion
coefficient by that for grain boundary diffusion and to reduce the cross-
sectional area through which diffusion occurs from L
2
to obL where ob is
the thickness of the grain boundary for diffusion purposes, and o is a
dimensionless constant ~ 1. Hence the creep rate for Coble Creep is:

3
4
GB
kTL
b D o o
= c Eq. 3.6.

Both these mechanisms of creep occur simultaneously at all temperatures
but, as E
A
reaches ~ kT, the high sensitivity of the diffusion rate to the
activation energy means that there is an abrupt change from Coble creep
to Herring-Nabarro creep kinetics.

Apart from the evidence of the kinetics of creep in both metals and
ceramics, very direct evidence of the mechanism comes from the
formation of precipitate-free material on the horizontal boundaries and the
accumulation of precipitates on vertical boundaries as a result of the
process.

Herring-Nabarro creep of Mg 0.55% Zr alloy showing precipitate-free
horizontal boundaries and precipitate enrichment on vertical boundaries.


DISLOCATION MECHANISMS:

POWER LAW CREEP

For the practically important range of higher stresses and at temperatures
in excess of 0.5 T
m
, the strain rate is related to the stress by an empirical
equation of the form:

n
|
|
.
|
\
|
o
c Eq. 3.7

Where the exponent n takes values between 3 and 10 and thus the creep
rate is very sensitive to the stress. Specifically n ~ 5 for a wide range of
metals and alloys. (Taken together with the Arhenius equation, this gives
rise to the equation 3.2)

Power law creep essentially describes deformation produced by the
movement (glide) of dislocations which is itself limited by the climb of
those same dislocations around obstacles substantial enough to prevent
plastic flow.

These can be precipitates or hardening interactions with other
dislocations. The o
5
dependence is difficult to rationalise but the following
argument in R.C Reed (pages 92-94) and based on work by Brehm and
Daehn (Met. Trans. 33A, p.363) fits the experimental findings well.


As creep proceeds the dislocations accumulate and form networks, often,
but not necessarily, arranged in a cellular structure. These result in work
hardening and will be called hardening dislocation density,
h
. The
density of dislocations builds up with the strain so that:

d
h
dt
= M
dc
dt

where M is a constant. This results in softening processes in which
dislocations annihilate reducing the density establishing steady state
creep. The rate of generation equals the rate of loss and the net
dislocation concentration and creep rate is constant. The derivation
depends on the rate of coarsening of the dislocations to produce
softening d
s
, to produce this steady state.

d
h
dt
+
d
s
dt
= 0 (at steady state)

The dislocation network is assumed to coarsen by diffusion controlled
dislocation climb in much the same way as an array of precipitates
coarsen and obeys the following rate equation:

3

o
3
= KDt

where
o
is the initial dislocation spacing, D is the self diffusion
coefficient, t is time taken to expand the network to the scale , and K a
constant.

Differentiating this expression tells us the rate at which any particular
network can coarsen and hence the rate, at steady-state, at which
hardening dislocations can accumulate.

d
dt
= 3
2
d
3
( )
dt
= 3
2
KD hence

d
s
dt
=
d
s
d
d
dt
=
2
3

KD
3
2
=
2
3
KD
5

As at steady state the rate at which the materials can deform is controlled
by the rate at which it can recover then:

d
h
dt
= M
dc
dt
=
d
s
dt
=
2
3
KD
5

We need only now relate the stress o, to the spacing , using the Orowan
equation for a dislocation bowing out between pinning points distance
apart:

o =
Gb
to give:

dc
dt
= =
2KD
3Mb
5
o
G
|
\

|
.
|
5
= A
o
G
|
\

|
.
|
5
exp
Q
RT
|
\

|
.
|

where A= 2KD
o
/3Mb
5
and Q is the activation energy for self diffusion.


PIPE DIFFUSION AT LOW TEMPERATURES

At low temperatures it is found that the measured activation energy for
creep can fall and that the stress exponent rises. The explanation for this
is that as lattice diffusion is frozen out, diffusion along the dislocation
cores, or pipe diffusion, becomes the dominant mechanism. The
activation energy for pipe diffusion is pretty difficult to measure although
some values have been determined and the value is close to that for grain
boundary diffusion. The effective relative cross section through which
diffusion can occur is reduced from one, for lattice diffusion, to the
dislocation density times the cross-sectional area of the dislocation core.
Since we are assuming that the dislocation density is proportional to the
stress squared, this will raise the stress exponent by 2 for pipe diffusion.

2 n
DC
2
kT
b D
A
+
|
|
.
|
\
|
o
= Eq. 3.12

Where D
DC
is the diffusion coefficient for pipe diffusion. On the Ashby
diagrams this is labelled LT creep to distinguish it from the HT creep
controlled by lattice diffusion.

POWER LAW BREAKDOWN

At very high stresses the power law begins to break down and the strain
rates increase rapidly. The stress dependence becomes closer to an
exponential. Empirically the creep rate can be expressed as the function:

n
2
sinh
kT
b D
A
'
(
|
|
.
|
\
|
o o'
' = Eq. 3.13.

where A
2
o
n
= A
2
in 3.12. At stresses below o
s
~ /o this equation
reduces to 3.12 and at high stresses the stress dependence becomes
exponential. The diffusion coefficient can be larger than the self-diffusion
coefficient but this is in part due to the fall in the shear modulus with
temperature which has the effect of apparently increasing the stress.

The deformation process is activated glide rather than distinct processes
of climb which are rate controlling followed by glide which largely
produces the strain. The stress has reached the point where the
activation barriers are reduced to the level where they can be surmounted
directly and physically we are moving into the realm of plasticity.
However then sinh formulation for the stress dependence is widely used in
some important applications of creep modelling for systems operating at
high stress, e.g Turbine blades.


CREEP DEFORMATION MAPS PUTTING IT ALL TOGETHER

In principle all of the above processes can occur simultaneously but at any
combination of the stress and the temperature the kinetics of one of the
processes will dominate the rest, with the boundaries between the
different regimes being reasonably sharp. This enables Ashby to produce
the creep deformation maps which you will have come across at the end
of the 1st year.

Each map applies to a particular material and microstructure with the
composition and grain size being important parameters. They are
produced by analysing a very large amount of creep data from which it is
possible to identify the stress and temperature dependence and to fit to
the rate equations outlined above. The strain rate is then plotted against
the dimensionless axes, T/T
m
and o/ to give the familiar diagram.

The shape and position of the various regions is consistent from alloy to
alloy. Microstructure plays an important role in shifting the relative
positions and indeed occurrence of the mechanisms. For example
compare the creep diagrams for nickel at two grain sizes, 1mm and
0.1mm; in the smaller grain size the material is also work-hardened.

The effect of grain size and work-hardening on the deformation mechanism map
for nickel. Taken from the previous two graphs.

The effects are:

- The yield stress is higher for the work-hardened material.

- Diffusional creep dominates to higher stresses in the small-grained
material.

- Power law creep dotted interface does not move despite change in
dislocation density.

- The vertical line between diffusion creep occurring by lattice
diffusion and by boundary diffusion moves to the right for the
smaller-grained material.


Creep deformation maps for Nickel. Top: work-hardened 1mm grain size,
lower: annealed Ni grain size 0.1mm.


Strain rate contours:

Within each of the regions the strain rate contours are roughly parallel.
The values are determined by the rate equations and the spacing and
gradients are found from these equations.

Stress:
For power rate dependence:
n
o c , taking logs of both sides: o = c lg n lg .
Since the stress is plotted on a log scale the exponent n is just the
gradient of the strain rate and the stress measured at constant
temperature i.e. straight up and down.

For example for the power-law HT creep in the 1mm Nickel diagram the
strain rate changes by 4.5 orders of magnitude whilst the stress changes
by one order of magnitude so the value of n is 4.5. Looking at the
diffusion creep the ratio is very close to one corresponding to the linear
relationship between the strain rate and stress.

At constant stress the separation of the contours with respect to the
temperature depends on 1/T and hence the contours become more widely
spaced towards the right hand side of any creep region (or regime).

Creep Failure

In a constant stress test the processes described might well continue
indefinitely with the sample becoming thinner but carrying a
proportionately reduced load. However, eventually failure occurs, almost
always by the formation growth and coalescence of voids and cracks
formed throughout the component of test-piece. (This is essentially the
process which is taking over during the tertiary creep stage.) The
diffusion-driven deformation processes occurring during creep lead to high
fluxes of vacancies. The obvious sites for these to accumulate are at
horizontal grain boundaries or at precipitate interfaces. Once they have
formed and reached a critical size, the voids can continue to grow
mopping up excess vacancies as the increase in surface area per vacancy
consumed diminishes rapidly as the size of the void grows (as in
nucleation and growth). Ductile plastic damage can also result in the
formation an accumulation of micro-cracks and voids for example at non-
deforming precipitates or inclusions.

The gradual loss of section which results from this creep damage
effectively raises the stress on the remaining cross section and it is
possible to define a Damage Parameter e which is the ratio of the loss of
CSA to the nominal CSA. We will use e later to look at the combined
effects of creep and fatigue.


Single crystal superalloy crept to failure
at 950C 275MPa. Voids have formed
in the material at precipitates spreading
to form cracks and reduce the effective
cross-section of the material.

Schematic showing the formation of voids on
horizontal grain boundaries during creep leading to
eventual failure.

ADDING CREEP AT DIFFERENT CONDITIONS

We came across a similar challenge with fatigue real exposure is not at
the same stress nor even at the same temperature and as we have seen,
as these change so does the mechanism of deformation and also the rate
of damage accumulation.

The simplest way to add creep exposure is to look at the total strain as
the measure of damage and to use this to locate the equivalent position of
the creep curve corresponding to the new conditions. Clearly this is not
very satisfactory as the strain to failure can vary considerably with the
conditions and the damage is not necessarily proportional to the strain.

Robinsons law

This is the equivalent of Minors law in fatigue; the exposure is broken
down into intervals at a particular stress and temperature for which the
rupture time is known from tests directly or from the extrapolation of test
data. A fraction of the total life is calculated for each segment and failure
is anticipated when the sum of these exposure fractions is 1.

t
i
t
ri
=1
i=1
m
Robinsons law 3.14


In general Robinsons approach works reasonably well, but it assumes
that damage accumulates evenly with time and that the creep curves
obtained at constant strain and temperature are applicable to more
complex histories.

The graph below shows two identical creep testpieces cut from the same
bar, one of which had a very small strain at a higher temperature prior to
deformation at the test temperature shown of 750C. Prior exposure has
more than halved the amount of primary creep and extended the ductility
but the secondary creep behaviour is almost unaffected.

Creep curves from two stage tests D and D* showing the decrease in primary
creep rate and strain, but that secondary creep is virtually unaffected. The arrow
shows where the creep rate equals the maximum value in D*.

Use of Damage Parameters

Life fraction is an example of a damage parameter. It is an attempt to
quantify the micro-structural state of the material and transfer this from
one stress and temperature to another. In modelling creep exposure
under changing conditions it allows the microstructure to be transferred
to the new conditions at each increment. When the damage parameter
reaches a critical value associated with failure, the life is assumed to
finished. In the case of life fraction this terminal value is 1.

One very widely used parameter is e. The easiest way to visualise this is
as the reduction in the effective (load-bearing) section of the crept
specimen as the voids and crack grow during exposure. Where the failure
mechanism is the growth of voids this has a very direct physical
interpretation and can even be quantified experimentally. If the cross
section is reduced, the effective stress on the remaining material
increases and leads to an acceleration in the creep rate. This is in
addition to the reduction in macroscopic section which is easier to
quantify. This leads to the accelerating creep curves typical at the end of
a creep test.

The amount of damage can, in principle, be found by comparing the
modulus of the virgin material with that crept to close to the final state.
In practice this is extremely difficult to do. But the important point is that
e is a variable which affects all the mechanical properties in a similar way
and can be transferred between different states recording the damage like
a baton in a relay race.

e =
A A
eff
A
3.15

where A
eff
is the effective area of the sample

hence:

o
eff
=
o
1e
3.16

and

o = E 1e
( )
c
el
3.17

where c
el
is the elastic strain and E(1e) is effective modulus of the crept
material containing the voids.

Assume that at the start of steady state creep e = 0 and that the steady
state creep follows the equation:

dc
ss
dt
= Ao
n

and that during secondary creep damage begins to accumulate with the
effect of accelerating the creep rate to give:

c
ter
= A
o
1e
|
\

|
.
|
n

where the strain rate

c
ter
begins to accelerate away from the steady state
value c
ss
.

Combining these and compensating for the true strain gives:

e =1
1+ c
tot
1+ c
ss
c
ss
c
ter
|
\

|
.
|
1 n
3.18

allowing the damage to be plotted from the creep strain curve.

The creep can be modelled by assuming that damage accumulates as a
function of stress temperature etc using appropriate equations (i.e. one
which work!)

Typical damage evolution curve for high temperature creep at a loss of
section of about 0.5 the sample fails. Note the damage does not
accumulate evenly throughout the life.

COMBINING CREEP AND FATIGUE

In predicting the life of a component at high temperature we have to
consider the possibility that the material is subject to both creep and
fatigue. Damage from one process may well contribute to the failure by
the other process, or alternatively repair and mitigate the effects of the
other process.

One approach is to use life fraction as a damage parameter; in effect
combining Miners and Robinsons laws. For fatigue this will be the number
of cycles normalised by the number of cycles to failure: N/N
f
. For creep
this will be the time exposed normalised by the rupture life, t/t
r
. There is
a problem in assessing what the creep exposure has been in a fatigue test
where the stress is constantly varying this is a matter of judging the
total time spent at the highest (tensile) stress during a fatigue cycle.

To simplify this in tests investigating these effects, test cycles with a dwell
time at maximum and or minimum stress are devised. Obviously the
higher the test frequency the smaller the possible contribution from creep
will be and a strong dependence of life on the frequency does indicate
some creep contribution. (This will show up as a mismatch between HCF
and LCF data when plotted on the same graph as on page 44.)

Graph of a constant strain fatigue test with a dwell at both the upper and
lower temperatures. During the dwell the specimen creeps to reduce the
stress.

The diagram below plots the fractions of creep and fatigue life
expired for a variety of tests and shows three possible scenarios.

1. NO INTERACTION: creep and fatigue have absolutely no effect on
each other and the lifetimes are unchanged by damage of the other
sort.
2. LINEAR ADDITION: creep and fatigue damage contribute in a linear
manner; this results in a halving of life in a case where creep and
fatigue would be expected to result in failure at the same number of
cycles.
3. STRONG INTERACTION: the lifetimes for both creep and fatigue are
much shorter as a result of the interaction of the two.

Clearly if life is being predicted on the basis of the creep and fatigue data
separately there will be a serious problem for the 3
rd
case.

Application of Miners and Robinsons leads to a linear expression for life
of the component of the form where the creep and fatigue contribute
equally to the life:

N
i
N
f i
+
t
i
t
r i
<1 3.19


In situations where creep and fatigue interact this can be fitted empirically
to an expression of the type:

N
i
N
f i
|
\

|
.
|
|
n
+
t
i
t
r i
|
\

|
.
|
m
<1 3.20

where n and m can be less than or greater than 1 to give positive and
negative interactions.

We can also use the damage parameter e applied to fatigue in the
simplest linear case

Ae =
AN
N
f
which is equivalent to Miners law. More
complicated relations ships can be used.

Mechanisms of interaction:

An increase in dislocation density due to fatigue can increase creep rate
through dislocation movement and /or enhanced diffusion.

Increasing dislocation density can also decrease primary creep (see
earlier).

Cutting of grain boundary porosity increasing the growth rate of the pores

Blunting of crack tip by creep can reduce crack growth rate and hence
fatigue damage rate.


DESIGNING CREEP RESISTANT MATERIALS

- Use FCC alloys

- High melting point material

- Increase grain size or elongate grains in direction of principle
tensile stress.

- Add grain boundary precipitates to prevent sliding

- Dispersion hardening

Creep rate versus reduced stress for FCC BCC and HCP alloys


NICKEL-BASED SUPERALLOYS

Nickel based superalloys form a large proportion of the gas turbine for
both aeroengine and power generation purposes. The creep performance
of these single cyrstal superalloys used in the most demanding
component, the high-pressure turbine blades, ultimately determines the
operating parameters for the engine (gas temperatures and component
stresses) and thus the performance and efficiency of the engine. They are
not only of enormous economic importance in transport and power
generation, but provide an excellent example of the strategies that can be
used to design and optimise a material for creep resistance.

Requirements of High Pressure turbine blade material:

- Operate at gas stream temperatures ~1600C;
- Pressure to increase the TET: 200C rise increases efficiency by
5%.
- Rotates at 10,000 rev per min to produce an (average) stresses at
root of 300MPa.
- Max stress during takeoff time at stress will increase greatly with
the new A380 aircraft
- Life of 10,000 hours 3 years at 9 hours per day

Achieve this by using a nickel based (FCC) alloy with ~ 10+ additions.

Typical alloy in use at the moment: CMSX-4 Composition: 61.7% Ni +
9.5% Cr + 6.5% Co + 6.4% W + 3.0% Re + 5.7%Al + 1.0% Ti + 6.5%
Ta + 0.1% Ta +0.1%Hf

Al, Ti & Ta - formers

Cr, Al, Hf oxidation and corrosion resistance

W, Mo, Re solid solution hardening; reduction in diffusion in matrix;
manipulating misfit.

Diffusion creep is reduced by producing elongated grains with minimal
transverse boundaries, or eliminated by casting as a single crystal. This is
achieved by directional solidification combined with a spiral grain selector.
The most rapidly dendrite growth direction, [001], is selected as the long
axis for the blades. This is fortunate because the alloy also has a high
elastic anisotropy with the minimum value along this axis. This minimises
the strain produced during thermal cycling.


Effect of ordered intermetallic precipitates

The elements Al, Ti and Ta all form ordered precipitates of the form Ni
3
X.
Where Al is the main element these precipitates are cubic and termed the
phase. This has the same basic close-packed structure, and very
nearly the same lattice parameter, as the main Nickel matrix and hence
form as coherent precipitates. Intermetallics form 70% of the most
advanced superalloys, and are the major contributor to the high
temperature creep resistance. They are themselves, very creep-resistant
materials and have been the subject of exhaustive (and expensive)
research as low-density superalloy replacements, usage, however,
remains limited because of the lack of low temperature ductility.

Y
i
e
l
d

s
t
r
e
s
s

Temperature 700C
/

The really unique element of the ordered phase is that the resistance to
dislocation motion, effectively the yield stress, does not fall with
temperature as is normally the case, but rises. The rise in yield stress
occurs because the energy of the APB fault on the slip plane {111} is
much higher than the energy of the fault when it lies on the {100} plane.
Hence as the temperature rises the dislocations are able to lower the
energy of the configuration by cross-slip (for screw dislocations) or climb
(edge and mixed dislocations) and reducing the fault energy between the
pair. Unfortunately glide in this configuration leads to two parallel APB
faults.

At higher temperatures still, short-range diffusion behind the moving
dislocation is able to rapidly resolve the APB faults into ordered crystal,
and the critical resolved shear stress drops again. Ordering also
diminishes with temperature lowering the APB energy.

Griffith's Analysis + Fatigue + Fracture

Uploaded by

Copyright:

Available Formats

Griffith's Analysis + Fatigue + Fracture

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Griffith's Analysis + Fatigue + Fracture

Uploaded by

Copyright:

Available Formats

Natural Sciences Tripos Part II

Robinsons law 3.14

You might also like