Professional Documents
Culture Documents
Kc
2
Fig. 2. Variation of fracture toughness versus the plate thickness (several fracture toughness tests are needed to establish the trend).
B. Farahmand, K. Nikbin / Engineering Fracture Mechanics 75 (2008) 21442155 2147
4. Example of fracture toughness determination (FTD) for a range of alloys
Three sets of alloys of varying properties and ductilites were investigated in this exercise. These were alu-
minum, titanium and steel alloys. Table 1 shows the range of alloys used with available failure elongation and
fracture properties. In all cases data on alloys with full stressstrain curves were unavailable to establish frac-
ture toughness through the proposed technique to identify the K
Ic
/DK
th
relationships needed for deriving the
fatigue crack growth properties. As an example analysis for an aluminum alloy is presented to highlight the
FTD methodology. Fig. 3 shows typical examples of uniaxial tensile behavior for two aluminum alloys.
Material fracture toughness as a function of part thickness was calculated for a range of aluminum alloys
using the FTD software [3]. Eect of plate width and crack length on fracture toughness (narrow and wide
plates) was compared with the data from the NASGRO database [1]. Fig. 4 shows an example of fracture
toughness versus plate thickness for two aluminum alloys compared to NASGRO data. It should be noted
Table 1
Tensile and fracture properties for a range of alloys used in the analysis (K units in MPa
p
m)
Materials K
Ic
K
th
K
Ic
/K
th
% Elong e
f
AISI 304, Ann Plt & Sht, Cast; 550F Air 137 5.5 25 25
AISI 304, Ann Plt & Sht, Cast; 800F Air 91 6.4 14 25
AISI 316, Ann Plt & Sht, Cast; 600F Air 137 5.46 25 25
AISI 316, Ann Plt & Sht, Cast; 800F Air 91 7.3 12 25
PH13-8Mo ( H1000; Plt, Forg, Extr) 91 4.55 20 16
15-5PH (H1100; Rnd, C-R) 73 4.6 16 16
Inconel 706 (Forg and extrusion) 85 12 7 27
280 Maraging steel 64 2.73 23 11
AF1410 100 3.1 32 12
4340 steel (Ftu = 1518 Mpa) 72.8 4 18 13
17-4PH, H1100 (Ftu = 1035 Mpa) 82 3.6 23 16
2014-T6 (Plt & Sht, L-T) 24.5 2.5 10 10
7075-T7351 (Extr; L-T) 30 2.7 11 12
Ti-55 ( Plt & Sht) 45.5 4.5 10 18
Ti-70 ( Plt & Sht) 45.5 4.5 10 16
Ti5Al2.5Sn 59 4.5 13 13
Ti2.5Cu STA 45.5 4.6 9.9 15.5
2024-T3 (Clad, Plt & Sht, L-T) 30 2.6 12 10
Ti8Al1Mo1V 50 3.1 13 13
2014-T6 (Plt & Sht, L-T) 30 2.6 12 11.5
2124-T851 27 2.7 10 8
6061-T651 (Plt; L-T & T-L) 24.6 3.2 8 14
7075-T7351 (Extr; L-T) 30 2.7 11 12
0
100
200
300
400
500
600
0 0.025 0.05 0.075 0.1 0.125 0.15
S
t
r
e
s
s
,
M
P
a
S
t
r
e
s
s
,
M
P
a
2014-T6 Aluminum -RT
(Longitudinal)
0
100
200
300
400
500
0 0.025 0.05 0.075 0.1 0.125
Strain, mm/mm Strain, mm/mm
2219-T87 Aluminum - RT
(Long Transverse)
2014-T6 Aluminum -RT
(Longitudinal)
2014-T6 Aluminum -RT
(Longitudinal)
2219-T87 Aluminum - RT
(Long Transverse)
Fig. 3. Typical full stressstrain curve for 2219-T87 and 2014-T6 aluminums.
2148 B. Farahmand, K. Nikbin / Engineering Fracture Mechanics 75 (2008) 21442155
that the NASGRO curve for a single material are typically not based on numerous tests at dierent thicknesses
through which a smooth curve passes, assuming a relationship between K
c
and thickness observed from multi-
ple materials.
In all cases the FTD analysis assumes no net-section yielding across the specimen width under monotonic
load. Note that the fracture toughness values, K
c
, in the NASGRO material library are available via an empir-
ical equation proposed by Vromen [10] as a function of K
Ic
and material yield value, which represent the lower
bound of K
c
values for a given part thickness. Fig. 5 shows the fracture toughness versus thickness data and
the curve t plot obtained from NASGRO manual. This gure includes numerous test data which umbrellas
whole range of plate width and crack lengths, a. The estimated empirical NASGRO curve t seems to repre-
sents the typical values of K
c
. The upper bound value of data shown in Fig. 5 must be associated with larger
cracks and wider plates. For this reason the FTD has the option of plotting the fracture toughness versus
thickness variations for plates of dierent width. The narrow plate represents the fracture toughness associated
with small cracks where the residual strength can be as high as 75% of material yield value. In Fig. 5 it is seen
that a relatively good agreement of a factor of about two dierence in K
Ic
/K
c
exists between the physical test-
ing taken from the NASGRO best t [1] and the present analysis shown in Fig. 4.
5. Estimation of threshold K
th
using predicted K
Ic
values
Fig. 4 shows that the K
Ic
from NASGRO and predictions are close to each other for the aluminum data con-
sidered. Therefore if K
Ic
is not available through the ASTM testing standards, it can be derived conservatively
2014-T6 Aluminum Alloy
0
10
20
30
40
50
60
70
0 10 20 30 40 50 60 70 80 90
Thickness - mm.
FTD (Wide Plate)
FTD (narrow Plate)
NASGRO
2219-T87 Aluminum Alloy
0
10
20
30
40
50
60
70
0 10 20 30 40 50 60 70 80 90
Thickness (mm)
F
r
a
c
t
u
r
e
T
o
u
g
h
n
e
s
s
-
M
P
a
(
m
)
0
.
5
F
r
a
c
t
u
r
e
T
o
u
g
h
n
e
s
s
-
M
P
a
(
m
)
0
.
5
FTD (Wide Plate)
FTD (Narrow Plate)
NASGRO
Fig. 4. Example of fracture toughness versus plate thickness for two aluminum alloys compared to NASGRO data [1].
Fig. 5. Measured fracture toughness versus thickness for several crack length of 2219-T87 aluminum [1].
B. Farahmand, K. Nikbin / Engineering Fracture Mechanics 75 (2008) 21442155 2149
through this analysis by using data from the stressstrain curve. Studying the threshold values of more than
hundred tests from dierent metallic alloys, the quantity K
th
was found to be related to the material plane
strain fracture toughness, where K
Ic
. = aK
th
. Fig. 6 shows the NASGRO value of K
th
for several aluminum,
titanium and steel alloys.
From Figs. 6 and 7 it can be seen that the NASGRO value of K
th
is material dependent and have varying
ranges for aluminum, titanium and steel alloys. From the experimental K
th
values in Fig. 6 mean values of
DK
th
= 2.5 MPa
p
m for aluminum alloys, DK
th
= 4 MPa
p
m for the titanium alloys and DK
th
= 5 MPa
p
m
for steels are found. The scatter for the steels are seen to be highest (because of dierent types of steel with
wide range of tensile properties) and for Al the lowest.
Looking at the stressstrain curves for several aluminums, the nal elongation for most aluminums fall
between 8% and 12% with K
Ic
values range from 20 to 27 MPa
p
m. This is one reason for aluminums alloys
having the minimum amount of scatter as indicated in Fig. 6 when compared with the steel. The range of a for
aluminums and titanium falls approximately between 8 and 14 and 9 and 16, respectively (see Fig. 7). The
range of a for variety of steels is shown in Fig. 8 and their lower and upper values are 5 and 25, respectively.
To understand the physical relationship that may exist between tensile elongation and fracture behavior the
correlations between them have been established for the dierent alloys shown in Table 1. It should be noted
that Table 1 has narrow range of ductility due to the lack of available K
th
and e
f
values for more ductile or
brittle alloys. However it is important to understand that the failure elongation values presented are within
a practical range for engineering alloys. Invariably any material outside this range could render the alloy either
very brittle or too ductile.
0.0
2.0
4.0
6.0
8.0
10.0
12.0
0 20 40 60 80 100
Number of Materials
K
t
h
(
M
p
a
m
0
.
5
)
Steel Titanium Aluminum
Fig. 6. NASGRO K
th
values for a range of steel, Al and Ti alloys.
0.0
2.0
4.0
6.0
8.0
10.0
12.0
14.0
16.0
18.0
0 10 20 30 40 50 60 70
Number of Tests
K
I
c
/
K
t
h
(
)
Titanium
Aluminum
Fig. 7. a Values for a range of Al and Ti alloys.
2150 B. Farahmand, K. Nikbin / Engineering Fracture Mechanics 75 (2008) 21442155
Fig. 9 shows the relationship between tensile elongation and K
th
for the data of dierent alloys. In addition,
Fig. 10 shows the relationship of a to tensile elongation for the dierent alloys in Table 1. It can be observed
that fairly good correlation for these properties with respect to tensile elongation exist in Fig. 9 which shows a
linear relation between K
th
and tensile elongation (taken from standard gauge lengths [1]). This suggests that
this relationship is material independent within the range examined. It should be noted that a necking ratio at
failure could be a less geometry dependent than elongation. However as data for necking is not always avail-
able it would be dicult to make direct comparisons at this stage. In Fig. 10 the relationship between a and
failure elongation is investigated. Given the range of failure strains of a factor of over three the best t equa-
tion shown in Fig. 10 gives a very good indication that a is relatively insensitive to failure elongation. As an
engineering method it can be used to identify an a lower limit even though there is a scatter of a factor of 3 in
0
5
10
15
20
25
30
0 20 40 60 80 100
Number of tests
K
I
c
/
K
t
h
(
)
Steel
Fig. 8. Predicted a values for a range of steel alloys containing a wide scatters.
1.0
2.0
3.0
4.0
5.0
6.0
7.0
8.0
5.0 10.0 15.0 20.0 25.0 30.0
elongation %
K
t
h
(
M
P
a
m
0
.
5
)
.
Steel
Aluminum
Titanium
Best fit
Kth = - 0.6337 + 0.3143
Fig. 9. Relationship of K
th
data and tensile elongation from Table 1.
1
10
100
5 10 15 20 25 30
elongation %
K
I
c
/
K
t
h
(
)
Steel
Aluminum
Titanium
Best Fit
= 16.348 -0.0685
f
Fig. 10. Relationship of a versus tensile elongation for data in Table 1.
B. Farahmand, K. Nikbin / Engineering Fracture Mechanics 75 (2008) 21442155 2151
a. This scatter can be reduced if individual alloys are considered. For example the Al and Ti alloys exhibit a
lower range than the maximum seen in the steel. Hence correlations in Figs. 9 and 10 suggest that the tensile
elongation could be an important factor in trying to understanding the physical fracture properties in these
alloys as well as them being used as a practical engineering tool to improving predictions for fatigue and
fracture.
Using the best t equations shown in the Figs. 9 and 10 it would possible to get a material independent
mean values of a and K
th
using available tensile ductility. The best t equations give
K
th
0:6337 0:3143e
f
7
a 16:348 0:0685e
f
8
Eq. (7) gives a very good indication of a linear relationship between K
th
and e
f
. This will allow FCG to use
estimated K
th
from Eq. (7) to generate the total da/dN curve.
6. Fatigue crack growth rate (FCG) estimation
Typically in a test a complete fatigue crack growth rate curve contains three regions. The primary and ter-
tiary regions (Fig. 11) are sensitive to the stress ratio, R and material micro-structural dependence. The sec-
ondary region is not and is invariably described by the Paris equation, where C and n are material constants
and DK is the stress intensity factor range (Eq. (9))
da=dN CDK
n
9
Appropriate material properties taken from experimental tests of the aluminum alloys are then introduced for
use in the deterministic analyses. The method later was extended to other aerospace alloys. The methodology
used to generate the fatigue crack growth curve for the threshold and Paris regions are described below. As
was mentioned earlier the DK value in the accelerated region is related to the critical value of the stress inten-
sity factor, K
c
, and is plate thickness and width dependent. Moreover, from the derivation of K
Ic
by the ex-
tended Grith theory, the lower portion of fatigue crack growth rate curve, K
th
, can be estimated for the stress
ratio, R = 0. Because of the diculty of obtaining full stressstrain curves for several materials, it was decided
to use the value of K
Ic
from the NASGRO database for calculating the K
th
value.
7. Analysis to predict steady state paris region for FCG
To estimate the Paris region it is possible to do it by using the FTD code where full uniaxial data are avail-
able or from a pedigree fatigue data set. It can be shown that for estimating the Paris region the two quantities
K
c
& K
Ic
must be available. These two quantities will be helpful to establish two points in the Paris region. The
Accelerated Region
(Thickness dependent)
K
da/dN
(1)
(2)
Kth
K K
Paris Region
Threshold
Region
Accelerated Region
(Thickness dependent)
K
da/dN
(1)
(2)
Kth
K K
Accelerated Region
(Thickness dependent)
K
da/dN
(1)
(2)
Kth
K K
Paris Region
Threshold
Region
Fig. 11. Three regions of the da/dN curve (threshold, Paris and accelerated regions).
2152 B. Farahmand, K. Nikbin / Engineering Fracture Mechanics 75 (2008) 21442155
two points in Fig. 11 (the lower and upper points of Paris region corresponding to points 1 and 2) have unique
properties which are common among many aluminum alloys. The lower point in the steady crack growth rate
region, just before getting into the threshold zone of the fatigue curves (point 1 of Fig. 11), has a material
independent property so that the ratio of the stress intensity factor, K
L
, at the mean lower bound point
and the threshold value, K
th
, (DK
L
/DK
th
for R = 0) is 1.125 for the crack growth rate per cycle,
da/dN 2.54E-6 mm/cycle (1.0E7 in./cycle) in aluminum. In the upper region of the da/dN curve (at the
end of the steady crack growth Paris region, point 2 of Fig. 11), the ratio of the upper bound stress intensity
factor, K
U
, and its critical value, K
c
, (DK
U
/K
c
for R = 0) is found to be 0.9 for the da/dN 0.127 mm/cycle
(0.005 in./cycle). Having the two quantities K
c
and K
Ic
available (either through the extended Grith theory
or NASGRO database) the two points in the Paris region can be generated. Hence, the C and n of Eq. (9) can
be determined.
Note that the K
Ic
value is used to estimate the K
th
value. The above assumptions used for establishing the
Paris region is also applicable to Titanium and Steel alloys. The total predicted fatigue crack growth curve can
then be plotted using Eq. (10), where the fracture parameters and constants are taken from the estimated K
Ic
,
K
c
, K
th
, and the Paris constants C and n values. For any other range of R-ratios the Newman closure equation,
f, [11] can be used to establish the full fatigue crack growth rate curve when R 50
da
dN
C1 f
n
DK
n
1
DK
th
DK
_ _
p
1 R
n
1
DK
1RK
c
_ _
q
10
In all cases the constants p and q of Eq. (10) were taken as 0.5 and 1, respectively.
8. Comparison of predicted bounds with experimental fatigue crack growth data
The computer program [3] which can run the simulations of the model and also on the world wide web
(www.alphastarcorp.com) was made available in order to verify a number of test cases. The input of data
can be treated as deterministic data or as probabilistic bounds of the data. Based on the above-mentioned
assumptions (construction of accelerated, Paris, and threshold regions), the fatigue crack growth curves for
three alloys are established and then compared with fatigue crack growth test data in NASGRO database.
The use of statistical methods give further condence to the methodology and is therefore crucial to any
sensitivity analysis that would be needed in design and life estimation methods. In obtaining material prop-
erties through physical or virtual testing, it is always expected to observe some amount of scatter on fracture
toughness and fatigue crack growth values due to material variations that can vary through heat lots when the
material is processed. Lesser amount of variability can also be observed in test coupons that have been
machined from a given plate of a given manufacturer by a specied heat lot. Figs. 7 and 8 give the range
of scatter that could exist in the parameter a for the dierent alloys. It should be noted that in the present case
the comparison is made with a wide variation of alloys of dierent ductilities. It is expected that in case specic
1.E-09
1.E-08
1.E-07
1.E-06
1.E-05
1.E-04
1.E-03
1.E-02
1.E-01
1.E+00
1 10 100
K - MPa (m)
0.5
d
a
/
d
N
-
m
m
/
c
y
c
l
e
NASGRO
Alpha=8
Alpha=14
Fig. 12. FCG and NASGRO da/dN curves for 6061-T62 aluminum.
B. Farahmand, K. Nikbin / Engineering Fracture Mechanics 75 (2008) 21442155 2153
single alloy conditions the scatter in a will be much less. However for the present example the upper lower
bounds will be considered.
Hence using the upper and lower bounds for alpha (a) for each alloy taken from Figs. 7 and 8 the Paris
curve can be constructed and compared with experimental mean data from NASGRO [1]. Figs. 1214 show
the predicted bounds of fatigue crack growth curves for the Al, Ti and the steel family of alloys shown in Table
1. The mean experimental data in all cases were compared with the upper and lower bound cases of a. It can be
seen that in the Paris region there is good agreement between the da/dN data provided by the FCG and the
NASGRO database. In the threshold region the predictions give the widest variation in the analysis which
assumes a xed K
Ic
. The increase in variation of a in the steel by a factor of 4 gives the biggest dierence
in the fatigue prediction as shown Fig. 14. For life assessment analysis of structural components it will be
appropriate to use the lower bound da/dN data corresponding to the upper bound a values. The upper bound
values of a (lower K
th
value) can provide conservative life assessment. It may be possible to avoid unnecessary
conservatism by tuning the threshold value to the material ductility using the correlations shown in Figs. 9 and
10 in order to obtain better agreement with the test data.
1.E-08
1.E-07
1.E-06
1.E-05
1.E-04
1.E-03
1.E-02
1.E-01
1.E+00
1 10 100
K - MPa (m)
0.5
d
a
/
d
N
-
m
m
/
c
y
c
l
e
NASGRO
Alpha=9
Alpha=16
Fig. 13. FCG and NASGRO da/dN curves for Ti2.5 Cu Titanium.
Table 2
Mean fatigue data form NASGRO fatigue database
Material K
Ic
(MPa
p
m) K
th
(MPa
p
m) C n
6061-T62 Al 25.5 2.7 5.5E10 2.8
Ti2.5CuSTA 45.5 4.6 1.4E10 2.9
Inconel 706 80.1 10.9 2.9E7 4.0
1.E-08
1.E-07
1.E-06
1.E-05
1.E-04
1.E-03
1.E-02
1.E-01
1.E+00
1.E+01
1 10 100 1000
K - MPa (m)
0.5
d
a
/
d
N
-
m
m
/
c
y
c
l
e
NASGRO
Alpha=7
Alpha=25
Fig. 14. FCG and NASGRO da/dN curves for Inconel 706 steel.
2154 B. Farahmand, K. Nikbin / Engineering Fracture Mechanics 75 (2008) 21442155
The eect of the upper/lower bound a on the prediction of threshold and Paris region, mainly C and n, were
compared with the NASGRO data [1]. The details are shown in Tables 2 and 3. For those cases where the full
stressstrain curves were available (Table 2), the range of predicted values of K
th
, C and n were falling well
within the mean of the parameters obtained through experimental data shown in Table 3.
9. Conclusion
A parametric analysis has been carried out to quantify the eects of static tensile properties on the predic-
tion of fracture and fatigue properties of three dierent classes of alloys, namely Al, Ti and steel alloys which
had dierent tensile failure strains. The proposed analytical/empirical approach, using an extended Grith
method to evaluate the energy dissipation at the crack tip using simple stress/strain data, can provide a useful
tool for engineers to derive fracture toughness and fatigue crack growth data for classical metal alloys used in
the aerospace industry where only a few or no test data are available. Two methods have been developed that
can estimate material fracture allowables; (1) facture toughness determination (FTD), which can estimate
material fracture toughness and also generates the fracture toughness versus part thickness and (2) fatigue
crack growth rate (FCG) that can generate the whole regions of da/dN curve. It has been shown that the pre-
dictions compare well with the test data of the same materials and geometries used in ASTM testing standards
for fracture toughness and fatigue crack growth. In addition, a sensitivity analysis using the variation in the
empirically derived parameter a = K
Ic
/K
th
was carried out to see the eects of material variability on the pre-
diction of fatigue crack growth of the alloys considered. It has been found that, for the range considered, a has
a lower bound with respect to failure elongation and the fatigue threshold and da/dN predictions can vary by
at most a factor of three using the upper/lower bounds of the parameter a which is found to exhibit a range of
525 for the present set of alloys. It has also been found that K
th
is, within the range of experimental scatter
and specimen sizes, directly proportional to material tensile elongation and could be used as a material inde-
pendent property for the range of alloys investigated. These eects need further investigations in order to
understand the physical reasons for the correlations between tensile ductility and fracture properties and to
validate these relationships for a wider range of alloys.
References
[1] Fatigue Crack Growth Computer Program NASGRO 4.0, JSC, SRI, ESA, and FAA, January 2002.
[2] Farahmand B. Virtual testing versus physical testing for material characterization. In: 45th AIAA/ASME/ASCE/AHS/ASC, April
1922, Palm Springs, California; 2003.
[3] Virtual Testing NASA/LaRC Contract # NAS-01067, NASA Langley Research Center, Alpha STAR Corporation and Boeing
Aerospace Company.
[4] Grith AA. The phenomenon of rupture and ow in solids. Philos Trans R Soc London, Ser A 1920;221.
[5] Tetelman AS, McEvily Jr AJ. Fracture of structural materials. John Wiley and Sons; 1967.
[6] Farahmand B. Fatigue and fracture mechanics of high risk parts. Chapman and Hall; 1997 [chapter 5].
[7] Farahmand B. Fracture mechanics of metals, composites, welds, and bolted joints. Kluwer Acadamic Publisher; 2000. now Springer-
Verlag Publishers [chapter 5].
[8] Nikbin KM, Smith DJ, Webster GA. Prediction of creep crack growth from uniaxial data. Proc R Soc London, Ser A
1984;396:18397.
[9] Rice JR, Rosengren GF. Plane strain deformation near a crack tip in a power-law hardening material. J Mech Phys Solids 1968;16:1.
[10] Vroman GA. Material thickness eect on critical stress intensity factor. Monograph #16, TRW Space and Technology Group; 1983
[February].
[11] Newman Jr JC. A crack opening stress equation for fatigue crack growth. Int J Fract 1984;24(3).
Table 3
Range of prediction, lower and upper alpha (a) values (MPa
p
m)
Material Predicted K
Ic
Predicted K
th
-range Predicted C Predicted n
6061-T62 Al 26.9 1.93.4 2.6E84.7E7 3.24.2
Ti2.5CuSTA 43 2.74.8 3.0E81.1E7 3.14.1
Inconel 706 77.5 3.011.0 4.6E111.5E7 2.84.5
B. Farahmand, K. Nikbin / Engineering Fracture Mechanics 75 (2008) 21442155 2155