Вы находитесь на странице: 1из 22

Department of Mechanical, Materials & Manufacturing Engineering

Material Models and Modes of Failure MM4MMM

Convenor: Wei Sun (Coates Building B68, w.sun@nottingham.ac.uk)

CREEP 1
Creep is the time-dependent deformation which occurs at elevated temperature
when materials are subjected to stress.
Most engineering components are designed on the basis that strength has a constant
value for a particular material, and that the strain is instantaneous and remains
constant with time on the application of a given load. As operating temperature
increases, the tensile properties (ultimate tensile strength and yield strength)
generally decrease but strain continues to be essentially instantaneous when load is
applied. At higher temperatures materials show an additional strain the creep strain
- which gradually increases with time even under constant load. Ultimately, the creep
deformation will lead to fracture of the component. A related phenomenon is stress
relaxation. This is the reduction in load which occurs with time when the strain of
certain components is kept constant. An example is bolts holding together high
temperature components which need to be periodically retightened because of load
relaxation.
Main engineering areas where creep can be significant are fossil-fuel and nuclear
power plant (boilers, pipework and steam turbines), chemical plant, and aero-engines.
Since thermodynamic efficiencies increase with increasing temperature, technological
advances in these and other areas often involve higher operating temperatures.
Furthermore, operating lifetimes are continually being increased. Detailed knowledge
of creep behaviour is therefore essential for safe and economic design of critical
components and structures.
Comparison of metals
Creep behaviour is categorised according to the temperature relative to the melting
point of a metal: Th = T / Tm (in K) - the homologous temperature.
Material

Melting temp, Tm
C

0.3 TmC

0.5 TmC

Th (=T/Tm) achieved
in alloys

lead (Pb)

320

600

- 93

27

aluminium (Al)

660

933

194

copper (Cu)

1083

1356

134

405

steels

~1400

~1673

229

564

0.47 (ferritic steel at 575C)

stainless steels

~1450

~1723

244

589

0.57 (austenitic steel at 750C)

nickel (Ni)

1453

1726

245

590

0.74 (Nimonic 115 at 980C)

titanium (Ti)

1668

1941

309

698

~0.4

0.56 (RR58 at 250C)

50

tungsten (W)

3380

3653

823

1554

0.76 (light filament at 2500C)

Different metals have similar creep characteristics when tested at the same Th.
general, creep deformation is not significant at temperatures below ~0.3 Tm.

In

Plastics also exhibit creep, which may be significant at ambient temperatures. Creep
of plastics is not covered in this module.
Types of creep failures
excessive deformation - pipework, gas or steam turbine blades;
creep buckling - deformation under compressive load;
creep cracking - pipework and pipework welds;
creep rupture - rupture of tungsten filament in an incandescent light bulb;
interactions - between creep and fatigue;
- between creep and environmental degradation; etc
Creep and stress-rupture testing
In metals, the majority of creep data are
collected

using

tensile

specimens

subjected to a uniaxial state of stress.


Usually

constant

load

tests

(dead-

weight) are employed using a simple


lever arrangement. In stress-rupture or
creep-rupture

testing

the

only

data

collected are the time-to-fracture, tf,


and the strain at failure, f, at a fixed
load and temperature.

In standard

creep testing, an extensometer is fitted


precisely to ridges at either end of the
gauge length, to enable strain values to
be recorded throughout the test.

from Introduction to Creep by R W Evans & B Wilshire

It should be noted that the popular constant-load test does not give precisely
constant-stress conditions, and for the most detailed understanding and verification of
creep theories truly constant-stress equipment is used. These use devices (e.g.
profiled cam) to adjust the load on the specimen as a function of measured strain,
assuming uniform deformation and constant volume. However, provided the creep
51

ductility of the material is low, say, only a few percent total creep strain, the results
from constant-load and constant-stress test methods are practically identical.
A typical creep curve of strain against time (linear scales) can be divided into three
regions:

III
D

II
I

c
B

t
A

After the initial immediate elastic extension i, there is primary creep BC (or transient
creep) in which the strain rate decreases with time. This is followed by secondary
creep CD in which the creep rate is constant with time. This is the steady-state strain
rate. Finally there may be tertiary, or accelerating, creep leading to fracture at E,
otherwise known as creep rupture or stress rupture. The total strain at rupture is
typically only a small fraction of the elongation to fracture in a tensile test at the same
temperature. The strain due to creep is c. If the specimen is unloaded at any point on
the curve there is an immediate recovery of strain, according to Hookes Law, but
negligible subsequent recovery.
The details of the c vs t curve are sensitive to alloying composition, including trace
elements in some cases; grain size; heat treatment; previous stress history; and
environment as well as temperature; applied stress and stress state.
1 < 2 < 3

= Fracture

3
2

2
1

(a)

(b)

52

Physical mechanisms of creep


The atoms (or molecules) in a solid vibrate; the crystal lattice defines the mean
position. The amplitude of vibration increases as temperature increases. As they
vibrate, the atoms collide. The result of the collisions is to increase the energy, and
hence the amplitude of vibration, of some atoms at the expense of others. The
Vacancy
increased amplitude of vibration can result in the possibility
of an atom being able to
diffusion
move from one atomic site to another ie diffusion can occur. Examples of diffusion
in a typical metal include:(a)

Interstitial diffusion
Small atoms (e.g. C and N in steel)
can diffuse by squeezing through
the space between larger atoms.
Vacancy diffusion

Interstitial
diffusion

(b)

When there are different species of


atoms of similar size, such as the one
in the figure, diffusion occurs by atoms
moving into vacancies. Self-diffusion
refers to the movement of atoms of a
(a)
single species by means of vacancies.

Vacancy
diffusion

To enable the diffusion to occur a certain activation energy is required.


Interstitial
diffusion

(b)

53

Diffusion along preferential paths grain boundaries and dislocation cores


Diffusion at grain boundaries or down the core of a dislocation is much easier than
within a crystal; it may be 106 times as fast. This is because of the larger and / or
irregular atomic spacings present in these regions.
The activation energies for these processes are about half the magnitude of those for
diffusion through the lattice bulk.

Gra
in

bou

nda

Diagrammatic representation

ry

of a boundary between two


grains of a metal. Grain
boundary acts as both source
Gra
in

and sink for vacancies


bo u

n da

ry

In the absence of stress, the direction in which atoms (or vacancies) diffuse is
governed by their concentration. They move from the more concentrated regions to
the less concentrated regions, or may be chemically attracted by other species to form
precipitates (e.g. Ti + C). The stress system also influences the direction of the
movement of the atoms or vacancies. Time dependent deformation (i.e. creep)
therefore occurs.
This is called diffusion creep and it is the dominant creep
mechanism at high temperatures (T/Tm > 0.4) and low stresses.
There are two main categories of diffusion creep: at the highest temperatures
diffusion in the bulk of the lattice, known as Nabarro-Herring creep, is rapid and
diffusion along the grain boundaries and dislocations makes negligible contribution.
However, as temperatures decrease towards 0.4Tm, diffusion along the easy paths
becomes progressively more significant because of their lower activation energies.
Dominance by grain boundary diffusion is called Coble creep.
Dislocation creep
At any temperature, creep also occurs by the movement of
dislocations provided the stresses are sufficiently high. This can be by the same
mechanism that plastic flow occurs under high load at ambient temperature (see
sketch). Other processes of dislocation movement are also activated at elevated
temperature.
54

(a)

(b)

(c)

A mechanism map indicates schematically the different dominating regimes as a


function of homologous temperature.

Note: the three regions of the typical creep curve do not represent three different
mechanisms, but rather the result of competing effects:
Strain hardening caused by dislocation movement results in strain hardening
(multiplication and tangling of dislocations which make subsequent dislocation
movement more difficult);
Recovery enabled by diffusion processes (loss of dislocations making subsequent
deformation easier).

55

The high temperature creep curve thus shows a linear portion where rates of
hardening and recovery balance.
At low temperatures, the creep curve may show a continuously decreasing gradient,
because of strain hardening which is not removed by recovery. This is also called
logarithmic creep.
Creep resistant alloys
For a material to withstand creep at high temperatures, a high melting point is
desired. Ceramics are sometimes used for this reason, but some of the most
refractory metals (e.g. tungsten and niobium) have limited use because they are
manufactured by a powder route. Oxidation resistance is another requirement.
The most commonly used creep resistant alloys are based on iron or nickel. Increases
to creep resistance are achieved by alloying additions and heat treatments to provide
extra strengthening: in particular fine dispersions of precipitates are effective in
restricting dislocation movement.
The precipitates must be stable at high
temperature, with minimal tendency to coarsen. The Nimonic alloys are some of the
most advanced creep-resisting alloys they contain chromium for oxidation
resistance, and complex precipitates based on Ni3 (Ti, Al).
Further improvements to creep resistance are achieved by using directional
solidification or single crystal components, to minimise the number of grain
boundaries lying normal to the load. This reduces the degree of grain boundary creep.
For example: Ni-based gas turbine blades.
Representations of creep behaviour

high , high T
Low , high T

Low T (< 0.3Tm)

t
56

For design purposes, predictions of the long-term creep strain under the operating
conditions of stress and temperature are required. A typical design standard requires
the stress to give minimum creep rates of 1% strain in 10,000 hours, or 1% strain in
100,000 hours. It is impractical to derive direct data from all possible combinations of
loading parameters, and to carry out numerous tests lasting these times. In other
applications predictions at even greater times are needed for example 300,000
hours for parts of power station boilers, pipework and steam turbines. Accelerated
tests at either higher stress or higher temperature must therefore be performed and
the results extrapolated to longer times. This must be done with caution!
For low-temperature creep, occurring at <0.3Tm, it is found that the transient creep
strain vs time can be expressed by

c 1 ln( 2 t 1) ,

c 1 2
2t 1

i.e. logarithmic creep

where 1 and 2 depend only weakly on and T.


For high temperature, high stress conditions the material fails rapidly by stress
rupture where the applied stress approaches the high temperature yield stress.
The most attention is paid to high temperature, low stress regimes exhibiting the
three-stage creep curve, concentrating on the secondary creep regime which takes up
the majority of the creep life.
The first step is to find mathematical expressions
which describe the experimentally measured behaviour.
It is assumed that creep strain c = f1 ().f2 (T).f3 (t), where the three functions are
separate.
Temperature dependence
Theoretical considerations and experimental evidence show that the temperature
dependence of the creep strain rate follows a form of Arrhenius Law:

Q
f 2 (T) exp

RT
where

Q is the activation energy


R is universal gas constant
T is absolute temperature
57

For T >> 0.4Tm, Q is found experimentally to be close to the activation energy for
lattice self-diffusion expected.

At test temperatures approaching 0.4Tm, Q is lower

because diffusion along preferred paths dominates.


Time dependence
The time dependence of strain has been represented by many mathematical models.
For example:
f3 (t) = t

(steady-state creep)

f3 (t) = b tm

(Bailey)

0.3 m 0.5 usually

f3 (t) = (1 + b t1/3)ekt -1

(Andrade)

f3 (t) b t1/3 at small t

mj
f3 (t) = a j t

(Graham and Walles)

A combination of the Bailey model and steady state is also often used, i.e.
f3 (t) = t + b tm
this being capable of representing primary and steady-state creep.
NB: Primary creep is sometimes included as part of the initial strain.
Stress dependence
By plotting a series of creep curves at the same temperature but with different
stresses, the relationship between strain rate and stress can be explored. Various
equations have been used to represent the stress dependence, f1 (), of the secondary
creep behaviour.

SS

= An

(Norton)

SS

= B sinh ()

(Prandtl)

SS

= C [sinh ()]p

(Garofalo)

SS

= D ( o)q

(friction stress)

where A, B, C, D etc are temperature dependent.


58

Because of its simplicity in mathematical analysis, the Norton equation is the most
commonly used.

SS

= An

(Norton)

It can be rewritten:
log (

SS

) = log A + n log ()

Nortons Law provides the basis for the power law relationships which have been
widely used to describe high-temperature creep behaviour. The stress exponent, n, is
an important creep property.
To obtain n values, creep curves are recorded at the same temperature but at
ss
different stresses. The secondary creep rates, , are plotted against on a log-log
scale. This can be repeated for tests at different temperatures. It is found that n
varies between 3 and 10 under the most common conditions used for creep tests but
at low stress n ~ 1.
A simple overall equation for the rate of steady state (secondary) creep can therefore
be of the form:

d css
d ( f 3 (t ))
f1 ( ). f 2 (T ).
dt
dt
For example, the power law expression often fits experimental data quite well and is
commonly used:
SS = C. n. exp[ Q ]
RT
This is likely to be valid within limited ranges of stress and temperature when n and Q
are constant i.e. the same physical mechanisms are occurring.
Tertiary creep and fracture
Creep at low temperatures very rarely leads to failure, and little change in
microstructure is evident apart from plastic strain. Creep at high temperatures almost
always leads to fracture, and the onset of fracture is indicated by the acceleration of
creep strain rate in the tertiary region.
Creep fracture data are usually presented as log () vs. log (tf) and

f vs log (tf).
59

Ductility usually tends to decrease with increasing test duration (although this is not
always the case). The log () vs log (tf) and f vs log (tf) data are usually of the form:

Log
Ductile failure
Load

Intergranular crack

Brittle failure
Material
Grain

Grain boundary

Log tf

(a)
Transgranular means the fracture path runs across
the bulk of grains; intergranular
means it runs along grain boundaries.
Load

Load

Intergranular crack
Grain boundary

Grain boundary

Material
Grain

Material
Grain

Transgranular crack

(b)

(a)

Mechanisms of failure
Load

(i) Ductile failure


In ductile rupture, as the tensile specimen elongates, the cross-sectional area
Grain boundary
decreases; creep deformation occurs
at constant volume. As the cross-sectional area
decreases, the stress increases because the load is constant, hence the elongation will
increase at an ever increasing rate.
Material
Grain

Transgranular crack

60
(b)

o,

lo , l

Ao, A
F
Failure life for ductile creep
0 0

Constant volume:

= =

Equilibrium:

= =

Nortons law:

0 0

0 0

= = (

(1)

True strain increment:

= (

0 0

= (
) =

(2)

(1) = (2)

1

1
0

= , so

1 = 0
0

0 0

1
0

= (

]
[

1(

(3)

when 0, , So

=
i.e.

(4)

0
1

= ( ) (slope = - 1/n)

61

Failure mechanisms:
Transgranular cracking;
Necking (area reduction);
Voids nucleation is weak.
(ii) Brittle-failure
Brittle creep fracture occurs because voids nucleate and grow (mainly at the grain
boundaries), and grain boundaries crack and slide due to the increases in stress in
those regions as creep occurs.

Grain boundary sliding and formation


of wedge-cracks at grain triple points.

Nucleation and growth of voids


(cavities) on grain boundaries

Voids and cracks tend to grow on boundaries which, are perpendicular to the stress
axis. These effectively reduce the load-bearing cross-sectional area.
The damage, , is defined as the ratio of the area of voids, Av, at any cross sectional
area to the overall cross-sectional area, Ao, i.e.

Av
A
or 1
Ao
Ao

where A is the current, un-damaged cross-sectional area. As increases during the


creep test, increases and hence the creep strain rate increases (tertiary creep).
Most commonly used creep damage constitutive equations are (Kachanov). Assume
creep strain rate and damage rate are:

= (

= (

)
1

)
1

(1)
(2)
62

Failure life for Brittle Creep

Av (voids)

Integrating (2) between limits

0 (1 ) = 0

i.e.

Ao

(1)+1

+1

] = ,
0

(1 ) = [1 ( + 1) ]+1

(3)

when 1,
therefore,

(4)

(+1)

Creep strain
Substituting (3) (1)

0 = 0 [1 ( + 1) ]

()

(+1)

{1 [1 ( + 1)

+1

+1

] +1

From (4)

) (slope = - 1/)
(+1)

= (

63

Appearance of creep damage and fracture


Inspection during the lifetime of a component, using surface replication techniques,
can reveal changes in microstructure such as changes in precipitate size and shape,
the presence of voids along grain boundaries, and microcracks.

After creep failure has occurred, fracture surfaces often show grain facets, sometimes
with dimples, and they may be oxidised from exposure to high temperature.

Extrapolation of rupture data


Despite their inherent uncertainties, extrapolation methods are used to predict both
rupture and creep strain at much longer times, and lower stresses and temperatures.
The prediction of rupture have received more attention because the consequences of
brittle rupture are generally more catastrophic than is excess strain, and because of
the larger amount of rupture data available.
64

For planning creep tests, a simple relationship can be used:


SS . t f = constant (over a limited stress range)
For more accurate predictions, one of the most used is the Larson-Miller method. Their
theory is derived from the Arrhenius equation but it is found to hold empirically. They
showed that families of creep rupture curves all showed a similar shape, and that by
using a time-temperature equivalence parameter they could be superimposed.
Log

A1

A3

A2

600oC

550oC

500oC
Log tf

Log

Master Curve

600oC

550oC
PLM (T, tf)

The Larson-Miller parameter


PLM
T
C
tf

is
is
is
is

PLM ()

T (C + log10 tf), where

constant for a given material and stress level


absolute temperature (K)
constant, usually = 20; sometimes in range 17 23
time in hours to rupture (or to reach a given value of creep strain)

Equivalent conditions using P = T ( 20 + log10 t )


Service conditions
time, h
temp, C
10 000
400

Test conditions
time, h
temp, C
8
500

PLM
~ 16152
65

10 000
1 000

400
500

219
40

450
550

~ 16152
~ 17779

A variation of the Larson-Miller parameter (putting C = 23.3 for better fit) employed to
superimpose results for a low alloy steel used for steam pipework (from Evans &
Wilshire Introduction to Creep). Test temperatures from 500 to 600C are
superimposed.
Use of the Larson-Miller parameter to compare qualitatively the performance of
different materials.

66

Multi-axial creep behaviour


The development of multi-axial creep models is usually a compromise between trying
to fit the complex behaviour observed in experimental tests and in producing relatively
simple mathematical models. The models are usually developed to fit the main
experimental observations. Only one model is developed here. However, this is the
most commonly used model and has been shown to fit (reasonably well) most
experimental data.
2
2

Consider the behaviour of an element with the


normals to its faces set up in the principal stress
directions (i.e. 1, 2 and 3 directions).

3
Firstly, experimental results indicate that creep is a constant volume process, i.e.

1c 2c 3c 0
The condition also implies that hydrostatic stress states (i.e. 1 = 2 = 3) do not
cause any creep strains to occur. If 1 = 2 = 3 , from symmetry and assuming the
material is isotropic,

1c 2c 3c ,

1c 0 , hence 2c 3c 0

1c 2c 3c 0 ,

and since,

then 3 1

0,

i.e.

too.

The results of experimental tests also indicate that the superposition of a hydrostatic
stress state onto any other stress state does not influence the creep behaviour.

Consider a general stress state


(1, 2, 3) or represented by

O (Hydrostatic
axis)

P (, , )

vector P. The line 00' represents


hydrostatic stress states, and 00''

O (m, m, m)

is the projection of 0P onto 00'. In

the vector triangle 00''P, 00'' is the

hydrostatic component (m , m ,
O

m ) of and 0''P is the effective


(deviatoric) stress which governs
the creep behaviour.

2
m

m
67

It can be shown that


hydrostatic stress, m
and effective stress,

( 1 2 3 )
3

eff K1 ( 1 2 ) 2 ( 2 3 ) 2 ( 3 1 ) 2

The value of K1 can be determined by considering the uniaxial stress state, i.e. 1 0,
2 = 3 = 0. Under these conditions, 1 = eff
Therefore, K1 ( 1 0) 0 (0 1 )
2

Thus K1

K1 2. 1 = eff

1
2

eff

and the effective stress

1
2

( 1 2 ) 2 ( 2 3 ) 2 ( 3 1 ) 2

The effective creep strain rate can be defined in a similar way:

ceff K 2 ( 1c c2 ) 2 ( c2 3c ) 2 ( 3c 1c ) 2
This must also reduce to the uniaxial form, i.e.

Because of the constant volume condition,

K 2 ( (
c
1

c
1

1c
2

)) 0 (
2

1c
2

c
eff
1c when 1 0 but 2 = 3 = 0.

2c 3c 1c / 2

when 2 = 3 = 0,

1c ) 2

ie K2 = 2/3

ceff

2
( 1c c2 ) 2 ( c2 3c ) 2 ( 3c 1c ) 2
3

Therefore if the uniaxial creep behaviour of a material is of the form


then under multiaxial conditions we have

= f1 () f2 (t),

c eff = f1 (eff) f2 (t)

68

c
It is also necessary to define a flow rule, ie we need to be able to split eff into its

component parts. The flow rule must contain the constant volume condition and the
fact that the hydrostatic stress component does not affect creep behaviour.
If we subtract the hydrostatic stress components from each of the stress components
(1, 2 and 3) we obtain the components of 0''P in the 1, 2 and 3 directions, ie ( 1m), (2- m) and (3- m). If the creep strain rates in the 1, 2 and 3 directions are
proportional to these components of 0''P, the sum of the components must be zero, ie
K3 (1- m) + K3 (2- m) + K3 (3- m) must be zero.

Now , m

(1 2 3 )
3

Thus K3 { (1- m) + (2- m) + (3- m) }


= K3 (1 + 2 + 3 - 3m)
=0
Therefore the sum of the creep strain rate components obtained using the above flow
rule satisfies the constant volume conditions.

and

= K3 (1- m)

= K3 (2- m)

= K3 (3- m)

Again the value of K3 can be obtained by considering the behaviour under a uniaxial
stress condition: 1 0 but 2 = 3 =0, hence eff = 1 and eff

eff

= f1(eff ). f2(t) = K3 (1 - m) = K3 (1-

= 1 . So,
c

1
2
1) =
K3 1
3
3

69

but for uniaxial conditions 1 = eff


K3 =

3 ceff
.
2 eff

Therefore the multiaxial creep law is

1
2

3 ceff
ceff
1
( 1 m ) =
(1 (2 3 ))
=
2 eff
eff
2

ceff
1
(

(3 1 ))
=
2
eff
2
ceff
1
(

(1 2 ))
=
3
eff
2
c
eff

Substituting

where eff

= f1(eff ).f2(t) gives

f1 ( eff )

eff

( 1

f1 ( eff )

eff

1
( 2 ( 3 1 )). f 2 (t )
2

f1 ( eff )

eff
1
2

1
( 2 3 )). f 2 (t )
2

1
( 3 ( 1 2 )). f 2 (t )
2

( 1 2 ) 2 ( 2 3 ) 2 ( 3 1 ) 2

The multiaxial creep law can be further generalised to include shear stresses and
strains.

f1 ( eff )
1
( x ( y z )).f 2 ( t )
eff
2
f1 ( eff )

eff
f1 ( eff )

eff

( y

1
( z x )). f 2 (t )
2

( z

1
( x y )). f 2 (t )
2
70

f1 ( eff )

xz 3
.

yz

( eff )

xy 3

f1 ( eff )
( eff )
f1 ( eff )
( eff )

. xz . f 2 (t )

. yz . f 2 (t )

. xy . f 2 (t )

where

eff

1
2

( x y ) 2 ( y z ) 2 ( z x ) 2 6( 2xy 2yz 2zx )

zz
zy
zx
xz

yz
xy

yy
yx

xx

71

Вам также может понравиться