Department of Mechanical, Materials & Manufacturing Engineering

Material Models and Modes of Failure MM4MMM

Convenor: Dr W Sun (Coates Building B68, [email protected])

FRACTURE
Consider the stress concentration factor for an elliptical hole in a large, linear-elastic
plate subjected to a remote, uniaxial stress:

kt =

2a
= 1+

as b 0, the elliptical hole degenerates

to a crack, i.e.

;
b

provided the behaviour remains linear elastic.

The root radius for an ellipse is given by
2

=b
a

a
Hence = 1+ 2

As 0,

a
2

Energy approach to fracture mechanics

The question arises:"If infinite stresses are created at sharp cracks in brittle materials, why can cracked
components sustain a load without failing?"
To answer this question, Griffith (Phil. Trans R. Soc. London, A, 221, 1921, pp.163197) adopted an energy approach to the problem. He reasoned that a crack will
propagate only if the energy released by an increment of crack growth, da, is
greater than the energy absorbed with the creation of the new surface.
Mode I
P
applied load
v
displacement

energy to create unit area of crack surface

a
crack length
da
increment of crack growth
B
thickness
S
2Bda (surface energy for da)

P, v

da
B

a
36

Constant Displacement

2a

Uo

Uo - U

(a)

(b)

Assuming a uniform tensile stress, , is applied to both ends of a flat plate (width L,
thickness t) and then the two ends are fixed (see Figure (a)). The strain energy
accumulated in the plate due to the stress and displacement is Uo. Now if we assume
that a through-thickness crack of length 2a (2a << L) (see Figure (b)) is made in the
plate shown in Figure (a), the strain energy in the plate will be reduced from Uo to Uo
U. However, a new surface energy S will be produced due to the creation of the
crack 2a. Therefore, relative to the original state (Figure (a)), the total potential
energy of the new system (Figure (b)) is
W = 0 U + S = -U + S
Constant Load

2a

Uo

Uo + U
vo

(c)

v (d)
P

In this case, for the same plate as shown in Figure (a), we assume that the top end of
the plate is fixed, but the bottom end is subjected to a uniformly distributed force,
which gives a total force of P. Under this load, the plate will have a uniform
37

displacement of vo, see Figure (c), and the strain energy accumulated in the plate due
to the stress and displacement is Uo. Now if we assume that a through-thickness crack
of length 2a (2a << L) (see Figure (d)) is made in the plate shown in Figure (c), due
to the stiffness reduction, the free end of the plate will have an additional
displacement v, and therefore, the strain energy in the plate will be increased from Uo
to Uo + U. Similar to the case shown in Figure (b), a new surface energy S will be
produced due to the crack 2a. Therefore, relative the original state (Figure (c), the
total potential energy of the new system (Figure (d)) is
W = -Pv +U+S
Since U = Pv, Pv = 2U, so
W = -2U +U+S = -U + S
which is the same as that from the Constant Displacement situation. These results
indicate that the condition for a crack to become unstable is independent of boundary
conditions.
Therefore, total potential energy:
W = -U + S
Unstable crack growth condition: W = Wmax,

2W
W
U S

0 ; and 2 0 . So
A
A A
A
U S

(units: N.m-1, for B = 1)

A A
where A is the surface area of the crack. Therefore,

U
- Driving force for per-unit length crack growth;
A
S
R
- Resistant force for per-unit length crack growth.
A

Example: Infinite Thin Plate (unit thickness, i.e. B = t = 1)

S 2a

1 2 2
a (from solution of elasticity theory). Therefore,
2E
1 2 2
W
a 2a
2E
S
U 2 a
2 ;

Therefore,
A
A
E

2W
2A

2
0 (i.e. Wmax exists).
E
38

Unstable crack growth condition:

a
2 . This leads to
E
2E
a cr
or
2
2

cr

2a

2E
a

Define G = the energy release rate (per length of new crack). Units kJ/m2.
i.e. G Gcrit
For the problem of an infinite plate with a crack of length 2a, it can be shown that
under plane stress conditions,

G=
and under plane strain conditions

2a
E

1 v a
G
2

39

Therefore, under plane stress, runaway crack growth (i.e. failure) occurs when

2a
E

or when

= Gcrit 2

a = EG crit .

Under plane strain, failure occurs when

a =

EG IC
(1 - 2 )

[Under plane strain Gcrit = GIC].

Thus for a given macroscopic stress level, , there is a critical crack length, acrit, when
fast failure starts.
Most materials are not linear elastic up to failure. However, the energy approach can
still be used if the plastic strains are restricted to a region very close to the crack tip;
small scale yielding. Under these conditions, the energy release rate can still be
reasonably accurately based on a linear elastic analysis. Also, Gcrit (or GIC) now
include a component associated with plastic deformation at the crack tip, as well as
the creation of new surface. Plasticity effects are considered later.
If G < GIC, the crack will not grow;
if G = GIC, stable crack growth occurs; and
if G > GIC, unstable crack growth occurs.
So far we have only considered the so-called Mode I loading case. There are three
modes:Mode I - opening mode
Mode II - shearing mode
Mode III - tearing mode
In general, Gtotal = GI + GII + GIII is the
energy release rate under mixed-mode
conditions.

Stress-intensity approach
40

Elastic crack tip stress fields

Mode I (Westergaard, 1939)

x=

3
KI
cos 1 - sin sin
2r
2
2
2

+ non - singular terms

y=

3
KI
cos 1 + sin sin + non- singular terms
2r
2
2 2

xy =

3
KI
sin cos cos + non- singular terms
2r
2
2
2

zx = zy = 0;

z = 0 (plane stress);

z = (x + y) (plane strain)

The equations show that the shape of the elastic stress fields is invariant in all
components subjected to a given mode of stress application. The magnitude of the
elastic field in the vicinity of the crack tip is described by KI , the Mode I stress
intensity factor, and will be a function of applied stress, crack size and crack shape.
KI has units N/m3/2 or MPa m.
Similar expressions exist for mode II and mode III loading situations. For mixed-mode
cases, the stress fields can be added.
It can be seen that KI, KII and KIII characterise the entire stress field (and hence strain
field) in the vicinity of the crack tip. It therefore seems reasonable to assume that, for
example, failure will occur when KI reaches a critical value Kcrit (or KIC under plane
strain conditions) under mode-I loading. KIC is the so-called fracture toughness of the
material and is the value generally used in design because it is the minimum,
thickness-independent value and is hence conservative.
41

For example, for an infinite plate with a crack of length 2a,

K I a

Comparing this with the energy release rate, GI, ( = a / E ), it can be seen that

K 2I
E
(1 2 )K 2I
GI =
E

GI =

for plane stress

for plane strain

Hence the energy and stress-intensity factor approaches are equivalent.

Solutions can be found in the literature for a wide range of geometries and loadings
(see sheet). For example:
1.
2.
3.

H. Tada, P. Paris and G. Irwin, 'The stress analysis of cracks handbook' DEL Research
Corporation, Hellertown, Pennyslvania, 1973.
D. P. Rooke and D. J. Cartwright, 'Compendium of stress intensity factors' HMSO 1975.
Y. Murakami (Editor), 'Stress-intensity factors handbook', Pergamon Press, Oxford,
1987 (2 volumes).

Crack tip plasticity zones

The critical stress intensity values (eg. KIC) of materials arise from a combination of
brittle crack surface creation and local plastic zone toughening.
Consider the Westergaard equations with = 0 under pure mode I loading.

x =

KI ;
y = KI (= x ); xy xy = 0
2 r
2 r

For an elastic-perfectly-plastic material with yield stress ys, obeying the von-Mises
yield criterion, yielding occurs when

(1 2 )2 + (2 3 )2 + (3 - 1 )2 = 2 ys2
In this case, 1, 2, 3, x, y, z, the principal stresses, so

1
2

( x - y )2 + ( y - z )2 + ( z - x )2 = ys

Therefore, at the crack tip, where = 0 and under plane stress (z = 0) and x = y,

ys

1
2

02 + y2 + x2

Substituting for the local stresses,

42

KI KI

+
= KI
ys =
2 2 r p * 2 r p *
2 r p *
1

A first approximation to the size of the plastic zone can be obtained, rp*, i.e.
2

1
r p*=
2

KI

ys

1
r p*=
18

KI

ys

Under plane strain conditions,

2

The actual plastic zones are larger because the load which would be carried with an
elastic material is shed onto the rest of the component, thus increasing the stress
outside the rp* region. Therefore, better estimates for the plastic zone sizes, rp, for
= 0 given by,

1 KI

r p
ys

for plane stress

43

1
rp
3

KI

ys

for plane strain

When rp is small (compared to crack and component dimensions), linear elastic

fracture mechanics can still be accurately applied. However, a plastic zone size
correction factor, ap, is applied to take the small-scale yielding into account, i.e. the
crack is assumed to be a length (a + ap) instead of 'a' when determining the stress
intensity factor, where

1 KI

=
ap
2 ys
1
ap =
6

KI

ys

for plane stress

2

for plane strain

By substituting other values of into the Westergaard equations, the first

approximation to the 'shape' of the plastic zone can be obtained, i.e. the variation of
rp* with .
NOTE: If rp is not small compared to the crack length then a linear elastic fracture
mechanics approach to the problem is inappropriate. Yielding fracture mechanism
and/or plastic collapse analyses may be more appropriate.
2

Thus (KIC/ys) , which is related to

rp, can be regarded as a defect
tolerance
parameter.
If
2
(KIC/ys) is 'small' failure will occur
2
due to fast fracture. If (KIC/y) is
large, failure occurs due to plastic
collapse. Intermediate behaviour
can be dealt with using yielding
fracture mechanisms. Likewise,
valid KIC values cannot be obtained
from test specimens unless rp is
small compared with crack and
test piece dimensions.

Schematic representation of plastic zone

ahead of crack tip (from Rolfe and Barson)

44

Fracture toughness determination [BS 5447 (now 7448)]

The aim of plane strain fracture toughness testing is to obtain reproducible values for
the lower limiting critical toughness of a material, tested in sufficiently thick section for
the shear lip contribution to the toughness to be neglected. The fracture must
proceed under quasi-elastic conditions (i.e. size of plastic zone at fracture must be
very much less than the cross-section of the specimen). In addition, the region of
interest at the crack tip must be very much smaller than the crack length, so that the
critical fracture event may be characterised by a single-valued K parameter.
Fracture toughness (KIC) tests are usually performed on one of two types of testpiece,
i.e. the single-edge-notched (SEN) bend specimen or the compact-tension (CT)
specimen.

Procedure
A fatigue crack is grown (at low loads compared with those to cause fast
fracture) from the notches to provide a sharp crack tip. The crack length (a/W)
must be within specified ranges (around 0.5), and the crack front should be
sufficiently straight.
Then load is applied until fracture occurs. The load-displacement curve is
recorded, and a preliminary toughness value is obtained from:

KQ

Y PQ
BW 1 / 2

where PQ is the load at failure (see later)

Y is the compliance coefficient which is a function of specimen type and a/W.

The Standard provides tables to look up Y. Then the validity of the test is
checked to see whether the KQ value can be equated to KIC.
45

Specimen thickness is
(a)
(b)
(c)

important as:
plane stress at surface;
mixed behaviour just inside the surface;
approx. plane strain conditions at centre.

Also, the plastic zone size (plane strain zone) must be small compared to other
important dimensions, effectively, a and B.
For 0.45 < a/W < 0.55

K
a, B 2.5 IC
ys

or else the test is not considered to be valid (note: this must be checked after the
test). This ensures that
a, B 25rp
It also means that for materials with a high toughness and low strength, testpieces
must be very large for the toughness values to be valid. Once validity is confirmed:

K IC =

PQ Y
BW 1/2

To determine PQ, the line with 5% reduction in gradient compared with the initial
(effectively elastic) load vs displacement gradient is used.
Load is measured by a
conventional load-cell. Displacement is measured using a clip-gauge mounted across
the open mouth of the notch.

46

Typical fracture toughness values (at 20oC)

Material

ys (MN/m2)

KIC (MN/M3/2)

Mild steel

220

140 to 200

Pressure vessel steel

(HY 130)

1700

170

Aluminium alloys

100 to 600

45 to 23

Cast iron

200 to 1000

20 to 6

Other KIC values (MN/m3/2)

Concrete

0.1 to 0.3

Glass

0.3 to 0.6

Engineering ceramics

1 to 8

Engineering polymers

0.5 to 10

Engineering composites

10 to 100

Titanium alloys

60 to 110

Nickel alloys

60 to 110

As a rule of thumb, avoid engineering materials with KIC less than 15 MN/m3/2.
Effect of temperature on fracture toughness
Within their practical range of use, the KIC value for most materials (metals and
metallic alloys) increases with temperature. At the same time, the yield stress, ys,
tends to decrease with temperature.

47

[NOTE: there are important exceptions to this general behaviour.]

Hence the defect tolerance parameter (KIC/ys)2 tends to increase with temperature
and the material becomes less susceptible to failure by fast fracture.
Effect of heat treatment on fracture toughness
In general, as strength decreases (for example a steel quenched and tempered to
different strengths), the fracture toughness increases.
Toughness can also be increased by removing inclusions (e.g. cleaner steels); or using
specially designed combinations of alloying additions together with heat treatments.
Aircraft undercarriages require very high strength and toughness this is achieved in
marageing steels in which complex, fine precipitates form within a martensitic
microstructure.
FATIGUE CRACK GROWTH
With smooth (plain) specimens without stress concentrations or surface defects the SN curve exhibits a combination of initiation-controlled behaviour (low stress range)
and propagation control (high stress range).
Sa
Sa2

Sa1

Ni2

Np2
NT2

NT1
Np1

Ni1

Log Nf
With many practical components, the design includes stress concentrations. We must
design against the possibility that surface defects are present with the potential for
initiating fatigue. Crack detection techniques can only find cracks greater than a
certain size. Fracture mechanics will determine the critical crack length, acrit , beyond
which fast failure occurs.
Calculations of crack growth are therefore essential, for
example to predict the minimum time before the critical crack length is reached.

48

The state of stress around a crack tip is expressed in terms of stress intensity factor,
K, where:
K = Y ( a)
where Y, sometimes called the K-compliance function, is a function of loading condition
and the relative geometry of the component and crack.
In cyclic loading, K varies over a stress intensity range K, where
K = (Kmax Kmin), or
K = Y (max - min ) ( a)
The stress intensity factor range K, causes fatigue crack growth at a rate da/dN
(distance per cycle). In general, crack growth becomes faster as the crack gets
longer.
a
P1
a

P2

da
dN
NT2

ai

N
If log (da/dN) is plotted against log (K), three regimes are identified.
log

da
dN

III

II
I

Kth

Log K

Region I contains Kth , a threshold value of K below which crack propagation does
not occur. Kth occurs at around da/dN 10-7 mm/cycle, i.e. 1 lattice spacing/cycle.
49

Beyond this is the linear regime (region II) about which most is known. At the other
extreme of the plot, region III describes rapid crack growth occurring due to the
presence of relatively large crack tip plastic zones ie. when Kmax KIC. At this crack
length, fast failure occurs.
To calculate crack growth behaviour in a general component, first it is determined
whether the threshold for crack growth has been exceeded. Then experimentally
derived crack growth data are used to predict crack growth in the linear region, using
data obtained under conditions as close to those in practice as possible. Various
expressions have been employed, but the most popular is the Paris-Erdogan equation:

da
C (K ) m
dN

the Paris-Erdogan equation

C and m are material constants. For most materials the constant C depends on R (=
Kmin/Kmax) and m is approximately constant, generally around 3.
Typical values of crack growth constants
Material

Kth
3

C
(x 10-11)

da
10 6 mm / cycle )
dN

3.3
3
3.1

0.24
4

6.2

2.6 3.9
3.9
4.0
4.0

3 - 19
0.34

(MN/m )
Mild steel
Structural steel
316
stainless
steel
Aluminium alloy
Copper
Brass
Nickel

4 to 7
2 to 5
4 to 6
1
1
2
4

to
to
to
to

2
3
4
8

K for

m
2

6.3

0.4

2.9
4.3
4.3 to 6.3
8.8

Kth and K (for da/dN = 10-6 mm/cycle) values depend on the R-value; units of C will
give da/dN in m/cycle when K is in MN/m3/2 (= MPa m).

50

Effect of mean stress

The R value is a measure of the mean stress, or mean stress intensity.

0<R<1
R=0
R = -1

log

da
dN

Higher R

Lower R

Similar slop

Log K
The dependence of crack growth rate on mean stress is complex, and is dealt with in
various ways. In some analyses, it is assumed that crack growth only occurs during
the part of the fatigue cycle when the crack experiences tensile loading, and a Parisequation is used with K as the tensile range only. In other applications or materials,
the full K range is used, including the compressive part because it gives a more
accurate answer. It should also be remembered that unless materials have been
stress relieved, local residual stresses can influence the crack growth rate.
Experimentally it is shown that in some materials (e.g. mild steel) crack growth rate is
largely independent of R-value; others such as high strength aluminium alloys are
very sensitive to mean stress. For design purposes it is clearly important to use crack
growth data obtained under conditions as close to the operating conditions as possible.
51

Determination of crack growth data

Experimental techniques for determining crack growth data need to control and
measure K, and measure crack growth rate under controlled temperature and
humidity conditions.
Usually SEN or CT specimens are used, with a constant applied load range. Crack
length can be measured by:
Potential drop methods change in electrical resistance due to crack growth;
Ultrasonic location of crack tip;
Visually at the surface, but this does not take account of curved crack front.
Plots of a versus N then enable da/dN to be derived for each value of crack length a.
Although load range is held constant, K values will change with crack length. These
K values are obtained using the standard solutions for given geometries, f(a/W).
Hence plots of log (da/dN) vs log (K) are produced and values of Kth, m and C
derived.
For more detailed determination of crack growth data, specimens designed to give
constant K values can be used.

52