Int. J. Fatigue Vol. 20, No. 9, pp.

641–649, 1998
 1998 Elsevier Science Ltd. All rights reserved
Printed in Great Britain
0142–1123/98/$—see front matter

PII: S0142-1123(98)00027-9

Energy criteria and cumulative damage

during fatigue crack growth
R.P. Skelton*, T. Vilhelmsen† and G.A. Webster*
*Department of Mechanical Engineering, Imperial College, Exhibition Road,
London SW7 2BX, UK
†Elsamprojekt, 7000 Fredericia, Denmark
(Received 17 December 1997; accepted 26 March 1998)

This paper considers the strain energy density generated ahead of a fatigue crack as predicted by elastic
and non-linear plastic theory. In both cases this decays inversely with distance ahead of the tip. The
process zone just ahead of the tip fails when the accumulated energy density reaches a critical value
after a certain number of fatigue cycles. A simple expression for LEFM/EPFM crack growth is derived,
demonstrating how appropriate energy values may be obtained from low cycle fatigue (LCF) tests,
where damage occurs in the bulk of the specimen. The analysis is extended to additional damage
accumulating further ahead of the crack during propagation, i.e. before the arrival of the process zone
itself. Predicted crack growth rates are compared with experimental LEFM/EPFM data, mostly at high
temperature where more energy data are available.  1998 Elsevier Science Ltd. All rights reserved

(Keywords: process zone; cyclic plastic zone; failure energy; damage accumulation)

NOMENCLATURE single cycle

Wc cumulative energy density
a, a0, af crack depth, initial crack depth, final
␤ exponent in stress–strain law
crack depth
⑀p plastic strain
a/W normalised crack depth
⑀t total strain
da/dN crack growth/cycle
␾ damage factor
f(a) damage multiplication factor
␯ Poisson’s ratio
A constant in stress–strain law
␳ size of process zone
C, C⬘ constants in cyclic crack growth laws
␴,␴y stress, yield stress
Ci idealised value for m = 2
⌬␴0 initial stress range
E Young’s modulus
G strain energy release rate
I␤ abbreviated constant
j, jmax no. of process zone INTRODUCTION
J, ⌬J contour integral, range of contour integral
K, ⌬K stress intensity factor, stress intensity Fatigue crack propagation rates in terms of linear
factor range elastic fracture mechanics (LEFM) are conventionally
⌬K0 threshold stress intensity factor range written1 as:
m, m⬘ exponent in crack growth laws da
N no. of cycles = C⌬K m (1)
N0 cycles to negotiate the first process zone dN
Nj cycles to negotiate the jth process zone where da/dN is crack growth/cycle, C and m are
r distance ahead of crack tip constants and ⌬K is the range of stress intensity factor.
rp, rc size of monotonic zone, size of cyclic In the elastic–plastic fracture mechanics (EPFM) range
plastic zone some investigators use the form:
R minimum stress/maximum stress
Y compliance factor da
= C⬘⌬Jm⬘ (2)
w energy density expended for a dN
where C⬘ and m⬘ are constants and ⌬J is the J integral
Corresponding author. Tel.: + 44-171-594-7098; Fax: + 44-171-823- range corresponding to Equation (1).
8845; E-mail: [Link]@[Link] From the available data it was originally suggested

641
642 R.P. Skelton et al.

that a value of m = 4 gave the best fit for engineering explicitly and still required the determination of numer-
applications1. Equation (1) has attracted extensive com- ous ‘LCF’ coefficients. In the present work, the term
ment, for example as to its dimensional status, the ␳ will be implicit in the final result, but will neverthe-
relationship between the terms C and m, the influence less be defined by the user and be associated with a
of minimum to maximum load ratio, R, the value of corresponding failure energy. This has the virtue of
⌬K above a threshold level, ⌬K0, below which fatigue extreme simplicity.
crack growth is not observed and the theoretical value The aim of the paper is therefore to derive an
of m when predicted by basic models. Reviews of equation for engineering use in the LEFM/EPFM range
fatigue crack growth laws2–4 have discussed various from a knowledge of energy expenditure values in the
modifications to Equation (1), e.g. the incorporation of LCF range taken from the literature.
intrinsic parameters such as yield stress, ␴y, and
Young’s modulus, E. Prior to Equation (1), treatment STRAIN ENERGY DENSITY AT A CRACK TIP
of fatigue crack propagation was generally in terms
of extrinsic parameters such as crack depth, a, and Elastic theory
mean stress5. The stress at a point (r, ␪) ahead of a sharp crack
Baı̈lon and Antolovich5 have classed crack propa- is given6,17,18 by the expression:
gation theories as those based on (i) crack tip opening
displacement, (ii) dislocation (i.e. micromechanistic) K
␴hi = f (␪ ) (4)
models and (iii) low cycle fatigue (LCF) processes in √2␲r hi
the bulk related to processes occurring near the crack
where the subscripts h, i refer to the components of
tip. Our paper is of type (iii) and includes cumulative
stress and f(␪) depends on the angular disposition of
damage in the same way as has been applied to creep
r. In this paper we consider conditions directly ahead
crack growth6,7. Here, a point some distance ahead of
of the crack only for which ␪ = 0. The corresponding
the crack tip amasses damage until the arrival of the
local energy density, w, at that point is given17 by:
tip itself. This is in contrast to cumulative damage
evolving at a single location at the crack tip, for K2
example during application of fatigue cycles of vary- w= (1 − ␯) (5)
2␲rE
ing magnitude.8,9
It is often stated10 that crack growth theories invok- for plane stress conditions where ␯ is Poisson’s ratio.
ing damage, strain, or energy accumulation at a crack For plane strain conditions the bracketed term is
tip lead to a value of m = 4 in Equation (1), whereas replaced by (1 − ␯ − 2␯2). Equation (5) thus shows
theories based on crack opening displacement lead to that the energy density is inversely proportional to the
m = 2. In fact, everything hinges4 on the size of the distance ahead of a crack tip. As the crack is
assumed ‘process zone’ with respect to the plastic zone approached, a singularity is avoided by the creation of
size, rp. The first attributed dislocation theory of growth a plastic zone.
due to Weertman11 uses the whole of rp and predicts In the presence of localised small scale plastic yield-
a value of m = 4. It can similarly be shown12 that if ing, gross linear elastic behaviour of the surrounding
the Griffith energy relieved above a crack is all trans- material controls the deformation in the plastic zone.
ferred into a plastic zone whose radius is proportional Thus the energy density in the zone should approximate
to (K/␴y)2 where K is the stress intensity factor, then to that calculated on the basis of the elastic solution.
the ratio of stored energy density in the zone to that
remote in the specimen is proportional to K4. In this Non-linear theory
paper an energy criterion together with a process zone In order to calculate the stress–strain distribution
␳¿rp is used, leading to a square power dependency ahead of a crack in a non-linear elastic–plastic material,
on ⌬K in Equation (1). a deformation law is required. In early analyses19–21
In early models relating LCF properties and micro- the Ramberg–Osgood law was used, which may be
structure with LEFM growth rates the cyclic plastic expressed as:

冉冊
zone rc (defined later) was divided into small micro-
␴ ␴ 1/␤
structural deformation zones within which the cyclic ⑀t = + (6)
deformation was considered uniform13,14. The dimen- E A
sion of each element along the crack path was taken as where ⑀t is total strain and A and ␤ are material
the amount of crack extension per cycle. The resulting constants. If the elastic strains are negligible compared
expression for da/dN was complex13 and contained the with the plastic strains near the crack tip, then only
term ␳ explicitly, besides many LCF constants the power term applies:
(‘strength’ and ‘ductility’ coefficients, etc.) requiring
separate determination. The term C (which depended ␴ = A⑀␤p (7)
on cyclic stress–strain properties) in Equation (1) was
where ⑀p is the plastic strain. From the original analy-
a function of ⌬K, thus giving an apparent increase in
ses19–21, the stress field analogous to Equation (4) may
m above the basic value of 2. In other models15,16 the
be written, for ␪ = 0:
crack is allowed to advance (under a local strain) a
distance ␳ in a finite number of cycles N. Thus:
da ␳
= (3)
␴hi = 冉 冊
JA1/␤
I␤ r
␤/(1 + ␤)
(8)
dN N where I␤ is a constant depending on the value of ␤.
A very similar relation applies to the strain field ahead
The final expressions again contained the term ␳ of the crack tip, except that the term r is raised to the
 Energy criteria and cumulative damage during fatigue crack growth 643

power −1/(1 + ␤). Over the years, the notation and From an argument involving the threshold stress inten-
presentation of the original equations (known as the sity range, ⌬K0, (and therefore independent of the level
HRR fields after the authors19–21) have been simplified, of ⌬K administered during the test) it was deduced8,9
and since energy density is a product of stress and that ␳ lies in the range 0.005–0.02 mm at room tem-
strain, the analogous relation to Equation (5) is: perature, depending on the alloy concerned.

wp =
1
冉 冊
1 + ␤ I␤r
J
f(␤) (9) ENERGY FAILURE CRITERION
where f(␤) = 1 for plane stress. The calculation is The model
more complex for plane strain where for example, f(␤) It is now assumed that the process zone depth is
takes the values 0.25, 0.5 and 0.7 for ␤ values of constant for all conditions, the peak energy density
0.33, 0.2 and 0.1, respectively22. The factor I␤ depends therein for a single loading cycle is uniform and is
on the strain hardening exponent, ␤, in the following equal to the value given by Equation (5) or Equation
manner6: (9) as appropriate and that the zone fails when the
For plane stress: accumulated energy density after N applied cycles
I␤ = 7.2(0.12 + ␤)1/2 − 2.9␤ (10a) reaches a critical value, Wc, the crack then moving
forward by an amount ␳ whereupon energy begins to
while for plane strain: build up in a new process zone. The scheme is shown
I␤ = 10.3(0.13 + ␤)1/2 − 4.6␤ (10b) in Figure 2. The dissipated energy in each zone (a, b,
c, etc.) for a complete cycle is given by Equation (5)
Comparing Equation (5) with Equation (9) it is seen or Equation (9) as appropriate with the term r replaced
that energy density is again inversely proportional to by ␳, 2␳, 3␳, etc. Thus if energy expenditure causing
the distance from the crack tip. early cracking during LCF tests can be identified, we
may thus derive an expression for fatigue crack growth
Reversed load cycling via an energy consumption criterion as follows.
During fatigue cycling, a contained yield region at
the crack tip is forced into alternate Prediction of fatigue crack growth from LCF data
tension/compressive yield by the surrounding elastic During LCF tests at constant total strain range at
matrix. The application of the energy gradient relations elevated temperature it has been shown that the cumu-
may be summarised as follows8,9, working towards the lative (hysteresis) energy absorbed to ‘failure’, Wc, is
crack tip. Outside the ‘monotonic’ plastic zone, rp, constant, effectively independent of strain range23. It
Equation (5) applies in an entirely elastic region. is also independent of cycle order under block loading
Between the boundary of rp and a ‘cyclic’ plastic zone, conditions24,25 and, for a given total strain range, the
rc (where rc = 0.25rp corresponding to an elastic stress hysteresis energy per cycle, w, (shown as loop area in
range of 2␴y) neither Equation (4) and Equation (5) Figure 1) is almost independent of the stage of harden-
nor Equation (8) and Equation (9) are valid, and an ing or softening23,24. Such attributes make the energy
empirical transition relation is required. The HRR criterion a worthy candidate for describing crack tip
relations (Equation (8) and Equation (9)) are valid conditions. Furthermore, the failure condition may be
throughout the cyclic plastic zone given by r ⬍ rc. chosen as early as practicable24,26 and it has been
Very close to the crack tip there exists a ‘process shown that a typical crack initiation size in a ferritic
zone’ of depth ␳ (shown as an inset in Figure 1). and an austenitic alloy at 550°C at the end of cyclic
hardening/softening is 0.05 mm or less26. If this depth
is identified with the zone size ␳, then the number of
cycles to fail the zone is given by:

Figure 1 Schematic, showing process zone embedded within cyclic Figure 2 Model indicating each process zone associated with a
plastic zone and associated energy expenditure ‘bar specimen’: crack propagates as each successive bar fails
644 R.P. Skelton et al.

Wc the end of hardening or rapid softening periods where

N= (11) crack depths are typically 0.05 mm. This provides a
w
lower bound on the size of the process zone. When
Combining Equation (3), Equation (9) and Equation this method was not possible, a later ‘failure’ criterion,
(11)) we have for the general case: i.e. 1–5% load drop, was used which corresponds23,26
da
dN
=␳
w
=
1 ⌬J
Wc 1 + ␤ I␤Wc冉 冊
f(␤) (12a)
to a typical crack depth of 0.5–0.8 mm. This gives an
upper bound on the size of the process zone.
The energy values calculated in Table 1 and Table
which translates in the elastic case to: 2 thus demonstrate that a useful connection can be
made between bulk behaviour in LCF and near-crack-
da ⌬K 2(1 − ␯) tip processes responsible for LEFM crack propagation.
= (12b)
dN 2␲EWc The method may also be applied to single crystal
materials. A survey has shown that an average energy
for plane stress from Equation (5). Equation (12b) expenditure to a 0.3 mm crack depth criterion for
demonstrates a square dependency on stress intensity nickel-based superalloys may be taken42 as 0.25 J mm−3
for crack growth. Equation (12a) and (12b) thus pro- in the temperature range 700–1000°C. This value was
vide a simple link between events in the bulk of an effectively independent of crystal orientation to the
LCF specimen and those occurring at a crack tip. As stress axis. In the case of the single crystal alloy
noted in the Introduction, the value of ␳ does not SRR99, the energy for complete failure with the stress
appear explicitly in Equation (12a) and (12b) but is axis in the [001] direction is calculated43 to be
associated via the term Wc which characterises the 0.85 J mm−3 at 950°C. An example44 of LEFM crack
number of cycles to produce an embryo crack of growth for SRR99 at 900°C is given in Figure 3.
specified depth. Taking E = 1 × 105 MPa for this orientation and using
the two energy values above, predicted crack growth
PRELIMINARY EXAMPLES rates from Equation (12b) are also shown plotted in
the figure and are seen to bound the experimental data.
In their survey of crack growth in a variety of steels, It is noteworthy that since values of Young’s modulus
Tanaka and Matsuoka27,28 determined a best-fit relation are much higher in the [111] direction for single
C = 2.35 × 10−4/(35.96)m for room temperature growth crystals, typically 2.5 × 105 MPa at these tempera-
expressed in mm/cycle and MPa√m. Thus when m = tures,43 Equation (12b) would therefore predict that for
2 and taking a value E = 2 × 105 MPa it is deduced the same values of stress intensity factor range and
from Equation (1) and Equation (12b) that Wc = energy, crack growth rates would be about 2.5 times
3.0 J mm−3. This calculated value for a small crack is lower.
encouraging, since it is much less than the energy to
complete failure in an LCF test at room temperature
for a ferritic steel25,29, which is 22 J mm−3. CUMULATIVE DAMAGE CORRECTION
From Equation (1) and Equation (12b) we thus have For those alloys which demonstrate an enhanced crack
a relation between C and Wc when m = 2, viz. C = growth rate, that is where m > 2 in Equation (1),
(1 − ␯)/2␲EWc. Corresponding values of ␳ must be Equation (12a) and (12b) may be modified to allow
obtained by appeal to experiment. Quantitative infor- for possible damage accumulation ahead of the crack
mation is sparse26 but examination of the data of before a given location has been embraced by the
Fournier and Pineau30 on alloy 718 for example, shows process zone. The inverse relation between energy
that at a plastic strain range of 0.84%, their observed density and distance (Equation (5) and Equation (9))
crack ‘lengths’ of 70 ␮m at 550°C and 100 ␮m at room allows a simple assessment of such ‘spatial’ damage.
temperature were obtained after energy expenditures of A similar scheme was attempted by Glinka16 but the
4.8 J mm−3 and 22.5 J mm−3, respectively. This energy resulting expression, which contained the term ␳
(almost all of which is dissipated as heat per cycle) is explicitly, was rather complex.
not yet directly linked with any specific mechanism of Suppose in Figure 4 that a grid comprising of
crack tip fracture. equidistant elements of width ␳ is set up ahead of the
crack tip. If N0 cycles are consumed in breaking the
APPLICATION OF EQUATION (12A) AND (12B) first zone (a), this may be identified with a fatigue
damage ␾ of unity. While the N0 cycles were operating,
The crack growth literature was surveyed for examples from Equation (5) or Equation (9) the relative damage
of Equation (1) and Equation (2) where m = 2 and m⬘ inserted into zone (b) at distance 2␳, (c) at distance
= 1 (or very nearly), respectively. Corresponding values 3␳, (d) at distance 4␳, (e) . . ., is 1/2N, 1/3N, 1/4N,
of C (or C⬘) were also determined. The data are 1/5N and so forth, i.e. ␾ = 0.5, 0.33, 0.25, 0.20, etc.
summarised10,31–35, in Table 1. Cyclic test frequencies In breaking zone (b), only N/2 cycles are now required,
were generally 苲 0.1 Hz or less and ⌬K ranged from thereby inserting damage 1/4N, 1/6N, . . ., into zones
20–40 MPa m1/2; details of Young’s modulus etc. being (c), (d), . . ., so that the total damage now in zone
given in the original sources. Values of the parameter (c) is 1/3N + 1/4N = 7/12N which in turn only requires
Wc which satisfy Equation (12a) and (12b) are also damage of 5/12N (␾ = 0.417) to rupture it. Equation
included in Table 1. (3) may therefore be written in general for the jth
Corresponding values of Wc calculated from LCF zone to give failure in Nj cycles:
tests for the nearest equivalent conditions23,36–41 are
given in Table 2. These accumulated ‘failure energies’ da ␳
=
have been calculated23,26 from the number of cycles to dN Nj
 Energy criteria and cumulative damage during fatigue crack growth 645

Table 1 Calculated energies from LEFM data for ⌬K in MPa√m

Alloy Temperature (°C) m C (mm/cycle) (苲 Ci) Wc (J mm−3) Reference

2.25Cr1Mo 510 1.8 1.7 × 10−6 0.42 31

1CrMoV 427 2.3 1.6 × 10−7 4.2 32
1CrMoV 538 2.5 3.8 × 10−7 1.9 32
AISI 316 625 2.0 1.6 × 10−6 0.48 10
718 650 2.3 1.1 x 10−6 0.40 33
Waspaloy 650 1.8 6.7 × 10−7 0.90 34
Astroloy 650 2.3 6.7 × 10−7 0.90 34
HastelloyX 760 1.0a 2.5 × 10−1a 0.95 35

a
Equation (2), with ⌬J expressed in MJ m−2.

Table 2 Calculated energies from LCF data

Alloy Temperature (°C) Wc (J mm−3) Limiting bound Reference

2.25Cr1Mo 480–530 2.6 Lower 23, 36

1CrMoV 480 2.9 Lower 23, 36
1CrMoV 538 1.8 Lower 23, 36
AISI 316 650 5.1 Lower 38
718 600 5.1 Upper 38
Waspaloy 650 1.3 Upper 39
Astroloy 650 1.0 Upper 40
HastelloyX 760 1.3 Upper 41

which tends to zero at large distances from the crack

tip. The inverse of Equation (13) gives a ‘damage’
multiplication factor, f(a), which increases with increas-
ing crack depth. For values of j up to 100, this may
be written as:
f(a) = 1 + ln(j) + (j)−1/2 (14)
see Appendix A for more exact solutions.
Equation (14) is shown plotted as the bold line in
Figure 5 for values of j up to 50. The function f(a)
is thus a multiplier upon the crack growth rate in
Equation (12a) and (12b) according to the distance
ahead of the crack, j␳, where advanced damage has
accumulated. Equation (12b), for example becomes:
da ⌬K 2(1 − ␯)
= f(a) (15)
dN 2␲EWc

Whereas the advantage of the stress intensity factor

range ⌬K is to incorporate the externally applied stress
and crack depth into a single parameter, application of
Figure 3 Typical crack growth rates in single crystal material44 the term f(a) now requires distances ahead of the crack
together with calculated upper and lower bound failure energies in
the process zone
to be re-introduced. Unfortunately, studies employing
Equation (1) or Equation (2) seldom report correspond-
ing crack depths at each stage of ⌬K or ⌬J (there are
exceptions45) but in standard compact tension (CT)
where specimens, for example, measurements are usually

冘 taken at normalised crack depth ratios (a/W) between

j
Nj Nj − i
=1− (13) 0.4 and 0.6, i.e. over the range 20–30 mm.
N0 i=1
(i + 1)N0

COMPARISON WITH EXPERIMENT

Proceeding to build up the grid, it may be shown
that successive terms of Equation (13) are represented As noted in the Introduction, damage accumulating
by the series: ahead of the crack tip is expected to increase the
exponent in Equation (1) above a value of 2, see the
Nj/N0 = 1;1/2,5/12;9/24;251/720;95/288;
data in Figure 3 for example. A crack growth test has
19 087/60 480;. . . been performed46,47 on a rigidly-gripped single edge
646 R.P. Skelton et al.

Figure 4 Damage grid assuming energy density decaying inversely with distance ahead of the crack tip

Figure 6 Experimental crack growth data and comparison with

damage models assuming a reference square power dependency in
Figure 5 Damage multiplication factor indicated by bold line. Cal- Paris law, Equation (1)
culated points based on ‘restricted damage’ model, see text

notch (SEN) specimen of 0.5CrMoV steel at ambient with E = 2 × 105 MPa predicts an energy at failure,
temperature under conditions of R = 0 and frequency Wc, of 7.5 J mm−3 if the first observable crack growth
of 1–5 × 10−1 Hz. The specimen cross-section was increment is identified with rupture of the first element.
25 mm × 25 mm and propagation rates were determined The number of elements involved in crack growth
over the depth range 5.5–17.5 mm, i.e. 12 mm of across the section is given by:
growth. The resulting Paris plot is given in Figure 6 af − a0
(closed circles) and best-fit values for the constants in jmax = (16)
␳
Equation (1) are C = 1.62 × 10−10 and m = 3.96 for
growth expressed in mm/cycle. In Figure 6 a broken where af is the crack depth at the end of the test and
line of slope 2, which represents an idealised Paris a0 is the initial crack depth. Taking jmax = 100, the
law (i.e. from Equation (1) Ci = 7.39 × 10−8) has been process zone size is thus 0.12 mm. Assuming that the
constructed to pass through the first crack growth damage grid of Figure 4 initially operates over the
observation. Taking values of cyclic growth rate and whole region (af – a0), then the enhanced crack growth
⌬K at this point and substituting into Equation (12b) rate is obtained from Equation (15). Details are pro-
 Energy criteria and cumulative damage during fatigue crack growth 647

Table 3 Damage correction factors for crack propagation test

Original damage model Restricted damage model

Average crack ⌬K (MPa m1/2) Idealised da/dN j value f(a) rc(mm)a rc/␳ f(a)
depth (mm) (mm/cycle)
(units of 10−5)

5.79 27.6 5.63 2 2.4 1.5 13 2.4

6.73 30.8 7.0 10 3.6 1.9 16 3.3
7.98 30.5 6.8 21 4.2 1.9 16 3.3
9.25 32.9 8.0 31 4.5 2.2 18 3.5
10.41 34.3 8.7 41 4.8 2.3 20 3.6
11.28 36.5 9.8 48 4.9 2.7 22 3.7
12.37 37.4 10.3 57 5.1 2.8 23 3.7
13.64 37.0 10.1 68 5.2 2.7 23 3.7
14.91 36.7 10.0 79 5.4 2.7 22 3.7
16.13 34.8 9.0 89 5.5 2.4 20 3.3
17.06 37.5 10.4 97 5.6 2.8 24 3.8

a
Assumes ␴y = 200 MPa in Equation (17).

Table 4 Calculated energies from LEFM data assuming m = 2

Alloy Temperature (°C) Observed m Ci value (mm/cycle) Wc (J mm−3) Reference

if m = 2 (Equation (15))

AISI 304 538 3.86 1 × 10−7 7.0 48

AISI 304 550–700 1.5a 2 × 10−1a 1.3 49
IN100 650 2.7 6.7 × 10−7 0.91 34
René 95 650 2.82 1.0 × 10−6 0.60 50
Udimet 700 850 2.9 1.2 × 10−6 0.57 51

a
Equation (12a) for ⌬J expressed in MJ m−2.

vided in Table 3 as j increases towards its maximum DISCUSSION: RESTRICTED DAMAGE MODEL
value, and predicted growth rates for each experimen- Up till now it has been assumed pessimistically that
tally determined ⌬K value are plotted in Figure 6 as the damage grid of Figure 4 applies across the whole
open triangles: they generally overestimate observed projected crack growth region, which embraces the
rates by a factor of two. An alternative calculation, process zone, the cyclic plastic zone (Figure 1 and
where damage is limited by the plastic zone size, is Figure 2) and the elastic region beyond. Although the
provided in the Discussion section. energy density gradient obeys an r−1 relation in all
This type of calculation, which attempts to separate cases, energy in the elastic region is recovered during
the ‘masking’ effect of cumulative damage, can be unloading from tension and/or compression. There is
performed on other data where m > 2. Values of Ci thus a ‘cut-off’, whereby cyclic energy is only dissi-
have been calculated at the first measured value of pated within the plastic zone boundary. This boundary
da/dN in published data34,48–51 assuming m = 2 in moves forward as crack growth proceeds and Figure
Equation (1) and corresponding ‘failure’ energies calcu- 4 shows the case for rc/␳ = 5, i.e. for 5 process zones
lated from Equation (12b). The results are given in in the plastic zone.
Table 4 while energies for these materials determined In order to calculate the restricted damage contri-
from the nearest equivalent LCF tests52–56 taking the bution, all zones to the right of the bold ‘staircase’ in
earliest reported failure criterion are provided in Table Figure 4 for example must be discounted. With the
5. Agreement between the energies appears to corre- assistance of a computer spreadsheet, revised damage
spond more closely at higher temperatures and this factors were calculated for the case of rc/␳ = 5, 10,
comparison depends on establishing the first measured 20, and 30, respectively. The results are shown as
crack growth rate accurately. discrete points in Figure 5. An initial sharp drop in
f(a) is apparent, as the crack encounters the first cut-
Table 5 Nearest equivalent LCF energies for comparison with off zone where full damage did not accrue, shown for
Table 4 example as the box ‘f’ in Figure 4. Thereafter, in each
case there is a slight increase in f(a) during a transition
Alloy Temperature W(J mm−3) Reference period to a secondary maximum. The number of zones
(°C) negotiated in each transition period (where only partial
damage accumulates) always corresponds to the
AISI 304 500 0.40 52
600 0.15
respective number in the original plastic zone. Finally,
IN100 650 0.10 53 a state of dynamic equilibrium ensues and the damage
René 80 870 0.64 54 factor becomes constant in each case.
Udimet 700 760 0.17 55, 56 Using the revised contours it is then possible to re-
927 0.58 evaluate crack growth rates, taking into account the
plastic zone size:
648 R.P. Skelton et al.

⌬K 2 CONCLUSIONS
rc = (17)
4␲␴2y
1. When an energy value for crack initiation is known
at each stage. Details for the test on 0.5CrMoV steel from LCF tests, an estimate of LEFM crack growth
are provided in Table 3. The correction predicts a rates may be obtained from Equation (12a) or Equ-
reduction in crack growth rate, compared with the full ation (12b) which are based on the energy to fail
damage model, see Figure 6. a process zone at the crack tip.
When LEFM tests are conducted under displacement 2. This growth rate is increased when damage occurs
control, the opportunity again arises for testing the ahead of the crack tip (by incorporating an energy
energy models, since crack depth appears explicitly in density gradient accounted for by Equation (14)).
the expressions. In a test between constant displace- This in turn may be attenuated by constraining the
ment limits on a SEN specimen of 0.5CrMoV steel at accumulated energy to occur within the cyclic plas-
550°C (a0 = 2.97 mm; af = 15.1 mm) it has been tic zone only.
demonstrated46 that the crack growth rate accelerates
at first, passes through a maximum, then decreases.
The experimental data are given in Figure 7. This
effect arises because of an increasing specimen com- REFERENCES
pliance function, Y, and a decreasing load range 1 Paris, P. C. and Erdogan, F., J. Basic Eng., Trans ASME, 1963,
coupled with crack closure effects due to a negative 85, 528–534.
2 Hoeppner, D. W. and Krupp, W. E., Eng. Fract Mech., 1974,
R ratio46. Combining these effects, it can be shown 6, 47–70.
that Equation (15) becomes in this case: 3 Schwalbe, K.-H., Eng. Fract. Mech., 1974, 6, 325–421.

冋 冉 冊册
4 Irving, P. E. and McCartney, L. N., Metal Sci., 1977, 11,
da a(1 − ␯) a 2
= f(a) ⌬␴0Y 1 − (18) 5
351–361.
Baı̈lon, J.-P. and Antolovich, S. D. In Fatigue Mechanisms:
dN 2EWc W
Advances in Quantitative Measurement of Physical Damage, eds
The term ⌬␴0 is an effective stress range for zero J. Lankford et al., ASTM STP 811, Philadelphia, 1983, pp.
crack depth. Assuming E = 1.6 × 105 MPa, ⌬␴0 = 313–349.
6 Webster, G. A. and Ainsworth, R. A., High Temperature Compo-
163 MPa (extrapolated from experimental data), it is nent Life Assessment, Chapman and Hall, London, 1994.
found that Equation (18) passes through a maximum 7 Vilhelmsen, T. and Webster, G. A., Service Experience and
in Figure 7. This has been illustrated for several cases, Reliability Improvement: Nuclear, Fossil and Petroleum Plants,
viz. no damage f(a) = 1, full damage and restricted PVP-Vol. 288, ed. W. H. Bamford, ASME, New York, 1994,
pp. 291–295.
damage. In the latter model the term rc (Equation (17)) 8 Glinka, G., Int. J. Fatigue, 1982, 4, 59–67.
varied from 1.1 to 3.6 mm and a discontinuity is again 9 Glinka, G., Robin, C., Pluvinage, G. and Chehimi, C., Int. J.
evident as the plastic zone is propelled into less dam- Fatigue, 1984, 4, 37–47.
aged regions. The plots in Figure 7 indicate how 10 Lloyd, G. J. In Fatigue at High Temperature, ed. R. P. Skelton,
agreement with experiment may be obtained by Applied Science Publishers, London, 1983, pp. 187–258.
11 Weertman, J., Int. J. Fract. Mech., 1966, 2, 460–467.
assuming either (a) no damage and a low failure energy 12 Fine, M. E. and Davidson, D. L. In Fatigue Mechanisms:
or (b) damage accumulation associated with a high Advances in Quantitative Measurement of Physical Damage, ed.
failure energy. J. Lankford et al., ASTM STP 811, Philadelphia, 1983, pp.
350–370.
13 Majumdar, S. and Morrow, J., Fracture Toughness and Slow–
Stable Cracking. ASTM STP 559, Philadelphia, 1974, pp.
159–182.
14 Chakrabortty, S. B., Fatigue Eng. Mater. Struct., 1979, 2,
331–344.
15 Antolovich, S. D., Saxena, A. and Chanani, G. R., Eng. Fract.
Mech., 1974, 7, 649–652.
16 Glinka, G., Eng. Fract. Mech., 1985, 21, 245–262.
17 Glinka, G., Low Cycle Fatigue, ASTM STP 942, ed. H. Solomon
et al., Philadelphia, 1988, pp. 1022–1047.
18 Knott, J. F., Fundamentals of Fracture Mechanics, Butterworth,
London, 1973.
19 Hutchinson, J. W., J. Mech. Phys. Solids, 1968, 16, 13–31.
20 Hutchinson, J. W., J. Mech. Phys. Solids, 1968, 16, 337–347.
21 Rice, J. R. and Rosengren, G. F., J. Mech. Phys. Solids, 1968,
16, 1–12.
22 Riedel, H., Fracture at High Temperatures, Springer-Verlag,
Berlin, 1987, p. 37.
23 Skelton, R. P., Mater. Sci. Technol., 1991, 7, 427–439.
24 Skelton, R. P., Rees, C. J. and Webster, G. A., Fatigue Fract.
Eng. Mater. Struct., 1996, 19, 287–297.
25 Kliman, V. and Bily, M., Mater. Sci. Eng., 1984, 68, 11–18.
26 Skelton, R. P., Mater. Sci. Eng., 1994, 9, 1001–1008.
27 Tanaka, K. and Matsuoka, S., Int. J. Fract., 1977, 13, 563–583.
28 Tanaka, K., Int. J. Fract., 1979, 15, 57–68.
29 LeFebvre, D. and Ellyin, F., Int. J. Fatigue, 1984, 6, 9–15.
30 Fournier, D. and Pineau, A., Metall. Trans., 1977, 8A, 1095–
1105.
31 Corwin, W. R., Booker, M. K., Booker, B. L. P. and Brinkman,
C. R., ORNL Report No. 5355, 1978.
Figure 7 Crack growth data in displacement-controlled test com- 32 Saxena, A., Gieseke, B., Yoon, K. B., Norris, R. H. and Banerji,
pared with predictions from various damage models. Note maxima K., Elevated Temperature Crack Growth, Vol. 18, Materials
in growth rates Division, ASME, New York, 1990, pp. 113–121.
 Energy criteria and cumulative damage during fatigue crack growth 649

33 Pedron, J. P. and Pineau, A., Mater. Sci Eng., 1982, 56, 143–156. 53 Reger, M. and Rémy, L., Mater. Sci Eng. A, 1988, 101, 47–54.
34 Cowles, B. A., Sims, D. L., Warren, J. R. and Miner, R. V., 54 Antolovich, S. D., Liu, S. and Baur, R., Metall. Trans., 1981,
J. Eng. Mater. Technol. (Trans ASME), 1980, 102, 356–363. 12A, 473–481.
35 Huang, J. S. and Pelloux, R. M., Metall. Trans, 1980, 11A, 55 Wells, C. H. and Sullivan, C. P., Trans. ASM, 1967, 60,
899–904. 217–222.
36 Jaske, C. E., Frey, N. D. and Utah, D. A. In Application of 56 Wells, C. H. and Sullivan, C. P., Trans. ASM, 1968, 61,
Materials for Pressure Vessels and Piping, ed. G. V. Smith, 149–155.
MPC-10, ASME, New York, 1979, pp. 175–210.
37 Conway, J. B., Stentz, R. H. and Berling, J. T., Fatigue, Tensile
and Relaxation Behaviour of Stainless Steels, TID-26135, US APPENDIX A
Atomic Energy Commission, 1975.
38 Clavel, M., Levaillant, C. and Pineau, A. In Creep-Fatigue Sensitivity analyses for damage function
Environment Interactions, eds R. M. Pelloux and N. S. Stoloff, (1) A more exact curve for Equation (14) over 100
Proc. TMS-AIME Symp. Fall Meeting, Milwaukee, 1980, pp. terms gives:
24–45.
39 Bressers, J. and Verhegghe, B., Project CCR-2, COST 50, f(a) = 1.054 + 0.961 ln(j) + 0.955(j)−1/2 (A1)
P/COST/11/2.1, 1980.
40 Bressers, J., Roth, M., Tambuyser, P. and Fenske, E., Final while a fit over 1000 terms gives:
Report, Project CCR-2, COST 50, Round II, Report EUR
8162, 1982. f(a) = 0.952 + 0.978 ln(j) + 1.207(j)−1/2 − 0.161j−1
41 Castelli, M. G., Miner, R. V. and Robinson, D. N. In Thermome-
chanical Fatigue Behaviour of Materials, ed. H. Sehitoglu, (A2)
ASTM STP 1186, Philadelphia, 1993, pp. 106–125.
42 Skelton, R. P. and Webster, G. A., Mater. Sci. Engng., 1996,
A216, 139–154. Both Equation A(1) and Equation A(2) have a
43 Li, S. X., High temperature creep-fatigue and crack growth maximum error of about 0.12% over their respective
behaviour of a single crystal nickel base superalloy, Ph.D. whole ranges. Equation A(2) may be used, for example,
Thesis, University of Bristol, 1993. for smaller values of the process zone size, ␳, than
44 Affeldt, E. E., Timm, J. and Bennett, A. In Fatigue under
Thermal and Mechanical Loading, eds J. Bressers et al., Kluwer
has been considered in this paper.
Academic Publishers, Dordrecht, 1996, pp. 159–168. (2) It is remarkable that a very close approximation
45 Miller, M. S. and Gallagher, J. P. in Fatigue Crack Growth may also be obtained by considering f(a) to be half the
Measurement and Data Analysis, eds S. J. Hudak et al., ASTM number of terms in the well known Fibonacci series:
STP 738, Philadelphia, 1981, pp. 205–251.
46 Skelton, R. P., Beech, S. M., Holdsworth, S. R., Neate, G. J., 1; 1; 2; 3; 5; 8; 13; 21; 34; 65;. . .; j
Miller, D. A. and Priest, R. H. In Behaviour of Defects at High
Temperatures, ESIS 15, eds R. A. Ainsworth and R. P. Skelton, such that:
MEP, London, 1993, pp. 299–325.
ln(j√5)
47 Skelton, R. P. In Behaviour of Defects at High Temperatures, f(a) = 0.5 × (A3)
ESIS 15, eds R. A. Ainsworth and R. P. Skelton, MEP, London, ln[(1 + √5)/2]
1993, pp. 191–218.
48 James, L. A., Stress Analysis and Growth of Cracks, ASTM
STP 513, Philadelphia, 1972, pp. 218–229. (3) In the ‘restricted damage’ model, calculation of
49 Ohtani, R., Kitamura, T., Nitta, A. and Kuwabara, K. In Low the ‘steady state’ values of f(a) for a given plastic
Cycle Fatigue, eds H. Solomon et al., ASTM STP 942, Philadel-
phia, 1988, pp. 1163–1180. zone size is not simple, even using a spreadsheet,
50 Gayda, J., Gabb, T. P. and Miner, R. V. In Low Cycle Fatigue, owing to the ‘triangular’ shape of the damage matrix,
eds H. Solomon et al., ASTM STP 942, Philadelphia, 1988, pp. see Figure 4. The plateau (constant) values for f(a) up
293–309. to a maximum value of j = 30 in Figure 5 have been
51 Shahinian, P. and Sadananda, K., Proc. ASME-MPC-3 Sym- verified by a simple modification of Equation A(1):
posium Creep-Fatigue Interaction, ASME, New York, 1976, pp.
365–390. f(a) = 1.054 + 0.961 ln(j) + 0.955(j)−1/2 − j−0.2 (A4)
52 Tsuzaki, K., Hori, T., Maki, T. and Tamura, I., Mater. Sci.
Eng., 1983, 61, 247–260.