FACTA UNIVERSITATIS

Series: Mechanics, Automatic Control and Robotics Vol.3, No 13, 2003, pp. 583 - 597

AN R-CURVE ASSESSMENT OF STABLE CRACK GROWTH

IN AN ALUMINIUM ALLOY
UDC 539.219.2 620.1

A. R. Luxmoore
Faculty of Engineering, University of Wales Swansea, UK

Abstract. An AlMgZn alloy has shown brittle behaviour when tested using standard
fracture specimens, but larger specimens with the same contour geometry have
produced stable fracture i.e. controlled crack propagation under either increasing load
or increasing displacement. An R-curve analysis produced a rather unusually shaped,
but unique, R-curve for the alloy, with the crack resistance increasing parabolically
with crack extension. In addition, crack initiation was indexed by the constraint of the
test specimen. Despite these complications, the R-curve analysis allowed a satisfactory
explanation for the different fracture behaviour between the small and large test
specimens. The parabolic nature of the R-curve, obtained from several different
geometries, suggested that a simple shear lip analysis, similar to that of Krafft, Boyle
and Sullivan, should explain the behaviour of the test pieces, but despite some
modifications to the original shear lip theory, the results are not conclusive.

Nomenclature
Abbreviations:
ASTM American Society for Testing and Materials
CCT Centre Cracked specimens in Tension
COD Crack Opening Displacement (at any point along crack flank)
CMOD Crack Mouth Opening Displacement
CT Compact Tension specimen
SSY Small Scale Yielding
2-D Two-dimensional
3-D Three-dimensional
3PB Three Point Bend specimen

Received October 20, 2002

584 A. R. LUXMOORE

Latin symbols:
a half-crack length (CCT specimen), crack length (3PB and CT specimens)
aeff a crack length assuming plasticity (in small scale yielding) at the crack tip
B specimen thickness
b ligament length
E Young's modulus
G the driving force
Geff the driving force for a effective crack length
J the J-integral
JC J value at instability
Je elastic component of J
JIC fracture toughness in terms of J
Jp plastic component of J
JR fracture resistance in terms of J
K stress intensity factor
KIC plane strain fracture toughness
Keff stress intensity factor with respect to the effective crack length
R fracture resistance
S the ratio of total width of shear lips to specimen thickness.
W specimen width
Wt total energy dissipated during crack extension
We energy dissipated due to crack extension alone
Wp plastic energy dissipated in shear lips during crack growth
Greek symbols:
∆a or da crack extension
∆atrans maximum crack extension for flat fracture zone

INTRODUCTION
A weldable AlMgZn alloy, similar in specification to 7019 alloy, has been developed
for lightweight structural applications. Standard fracture tests produced valid plane strain
KIC values of around 35MPa√m. A full scale fatigue test on a structural component by
Webber [1] showed an unexpected increase in fracture resistance compared with its plane
strain KIC. An early study [2] attributed this increase to the loss of in-plane constraint in
the structure.
The constraint effects were examined experimentally by Sumpter [3], using 3PB and
CCT specimens of the bridge alloy. Henry et al. [4] combined experimental and finite
element studies on the T and Q-stresses for these test specimens to show that the
variation of critical J values, JC, could be rationalised by an indexing parameter, i.e. using
either the T-stress [5] or the Q-values [6]. These results led to the application of practical
two-parameter failure criteria, in the form of JC-T and JC-Q loci. Henry and Luxmoore [2]
showed that the structural component was a low constraint geometry, and introduced a
fracture assessment scheme [7] using the JC-T locus, Figure 1, and the applied J and T-
stress for a given crack length. According to this scheme, a crack is unsafe when its
combination of J and T-stress exceeds the locus.
 An R-Curve Assessment of Stable Crack Growth in an Aluminium Alloy 585

Sumpter [3] noted that crack extension in a large (b >> B) fracture test specimen was
stable, whilst shallow cracks in small (b = B) specimens failed at maximum loads.
However, crack growth in all high constraint geometries, whether large or small, initiated
at a constant JIC value of 0.018MN/m (equivalent to KIC = 35MPa√m). In all specimens,
fracture surfaces consisted of a central flat fracture 'tunnel', with differing amounts of
slant fracture near the specimen surfaces. The slant fracture was associated with the
development of shear lips at the specimen surface. In the small specimens, the shear lips
at the edge of the growing crack (as indicated by the slant fracture) reached a maximum
width of 2-5 mm, but in the large specimens, they kept increasing until the fracture
surface became fully slant, and the central flat fracture reached a point. Further crack
growth was then fully slanted.
0.1

0.09
Upper boundary
0.08

0.07
experimental results[2] from CCT
0.06 and 3PB specimens
JC (MN/m) 0.05 Numerical result of CCT
specimen, a/W=0.5(b>>B)
Lower boundary
0.04

0.03

0.02

0.01

0
-1 -0.8 -0.6 -0.4 -0.2 0 0.2
T stress/σys

Fig. 1. JC-T locus [2] and indexed value of J from a CCT specimen.

Resistance (R) curves were determined for stable crack extension in two large
(b >> B) and a small (b = B) test specimens, and compared with R-curves from Sumpter.
Fracture resistance is represented by the J-integral, denoted by JR, where the subscript R
denotes values of J on the resistance curve. The analyses produced concave JR-curves,
where the fracture resistance rises increasingly with crack extension, corresponding to the
increasing width of the shear lips. This allows long cracks to be safe because, although
G > JR at initiation, the rate of change dG/da < dJR/da (the rate of change in fracture
resistance) at, and after, this point. Using the R-curve approach, it is shown that further
extension of long cracks is only possible by stable propagation under substantial load
increases, agreeing with the experimental results of Sumpter [8]. A JR-curve from a large
CCT specimen (low constraint geometry) showed that the JR values at initiation were
elevated, but the curve coincided with that from high constraint geometries after further
crack extension. Hence the fracture toughness (KIC), and the values of JR for low
constraint geometries, must be indexed by T-stresses to allow for losses of constraint.
For the high constraint specimens, a concave parabola could fit the R-curves quite
accurately, and a simple shear lip analysis was undertaken to explain the R-curve
behaviour. This has not produced an entirely satisfactory model.
586 A. R. LUXMOORE

NUMERICAL ANALYSIS OF STABLE CRACK EXTENSION

Sumpter [3] conducted fracture toughness tests on the AlMgZn alloy using standard
(b = B) 3PB specimens (for valid KIC test), with a range of a/W from 0.05 to 0.5. A large
3PB and a large CCT specimen (for the same thickness but with b >> B) were also tested
and these produced significant stable crack extension (see Table 1 for details). The effect of
ligament size was shown by the alteration of the fracture plane, which changed from
predominantly flat fracture in the square ligament (b = B) to slant fracture in the rectangular
ligament (b >> B) for the same thickness. The shear lips developed and caused the initial
flat fracture in the (b >> B) specimen to taper to a point (tunneling), whereas flat fracture
was dominant throughout crack extension in the (b = B) specimens [3].

Table 1. Details of the finite element models for 3PB and CCT specimens.
3PB specimens Ligament length, Crack length, Span, Thickness,
a/W b (mm) a (mm) S (mm) B (mm)
0.2 (b = B) 25 6.25 125 25
0.33 50 25 300 25
0.5 25 25 200 25
0.2(b >> B) 160 40 800 25
CCT specimen 2a/W
0.52 62.5 67.5 25

Stable crack growth in the 3PB specimens, with a/W = 0.5(b = B) and 0.2(b >> B), and
a CCT specimen with 2a/W = 0.52(b >> B) (see Table 1) was simulated numerically with
the ABAQUS 'debond' process (only available for 2-D meshes). A rectangular finite
element mesh at the crack tip region is required for incremental crack extension. Crack
propagation criteria can be specified in three ways in ABAQUS, namely critical crack
opening displacement, critical stress and crack length versus time (which is related to the
time period of the analysis). Elements and nodes associated with the two potential crack
surfaces are specified in pairs which are initially contacted, but can be separated using one
of the prescribed crack propagation criteria, under the debond operation.
The critical crack opening displacement (COD) criterion is used typically for crack
propagation in ductile materials, and is used in the present case. For this criterion, the
user inputs a relationship between the COD and the crack extension. A fixed distance,
measured from the current crack tip, locates the point where the COD is monitored,
which is not usually at the crack tip for reasons of precision. The crack propagates
progressively (from one node to another) as a function of the prescribed crack extension
and the ratio of current (computed) COD to the prescribed COD. In the debond process
used in the present case, stresses between the two initially contacted elements reduce to
zero as a step function when the separation begins.
No experimental crack extension data was available for the numerical analysis, but a
curve relating the load and crack mouth opening displacement (CMOD) was provided.
Using simple geometry, the COD values versus crack extension required for the
ABAQUS input could be estimated on a trial and error basis. The computation proceeded
by dividing the debond process into several trials which allowed the load versus CMOD
curve of the numerical result to match the experimental result when satisfactory crack
extensions had been selected (Figure 2). Within each load increment there were several
 An R-Curve Assessment of Stable Crack Growth in an Aluminium Alloy 587

increments of crack extension. This procedure was very time consuming, but provided
the required results and a good match between the experimental and numerical load-
CMOD curves, see Figure 2.

20

18
Plane strain
16
Plane stress
14 Experiment result

12

Load (KN) 10

0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
CMOD (mm)

a. a/W = 0.5 (b = B)
200

180

160

140

120 Plane stress result

Plane strain result
Load (KN) 100 Experimental result
80

60

40

20

0
0 0.2 0.4 0.6 0.8 1 1.2
CMOD (mm)

b. a/W = 0.2 (b >> B)

Fig. 2. The comparison of load vs. CMOD plots between numerical and experimental
results of 3PB specimens.

Meshes of half of the 3PB specimens and a quarter of the CCT specimen were
generated for the 2-D analysis. The three rollers in the 3PB experimental test were
modelled as rigid surfaces, which had no deformation when they contacted with the
588 A. R. LUXMOORE

model of the specimen. The roller positioned above the crack line loaded the model. The
use of applied loads was recommended by ABAQUS for accurate evaluation of the work
done to the model. The output of each increment of crack extension from ABAQUS was
calculated after the application of load. In this case the released energy due to crack
extension could not be separated from the energy change under increasing load.
Therefore the fracture resistance was studied in terms of a J-integral (JR) versus crack
extension (∆a).

OUT-OF-PLANE CONSTRAINT
The numerical analysis, being 2-D, does not model the actual physical separation of
the test piece, but gives a through-thickness average of the plasticity with increasing
crack extension. This average can be understood in terms of a mixture of 'slant' and 'flat'
plasticity at any particular average crack extension, with the slant component quickly
becoming dominant. This mixture must be consistent between different geometries in
order to obtain the same R-curve. In addition, there is the choice between either plane
strain or plane stress out-of-plane constraint, which effects the extent of in-plane
plasticity, and hence the JR value. Where possible, each specimen was analysed using
both constraints, and comparison with experimental data decided the most appropriate
constraint. For example, the plane strain JR-curve of the a/W = 0.2(b >> B) 3PB specimen
was not acceptable because the load-CMOD result (see Figure 2b) of the numerical test
did not coincide with the experimental record. This was evident before any crack growth,
which started at an applied load of 85KN. In addition, the 3-D analysis of this specimen
with a stationary crack showed close agreement between 3-D and plane stress load-
displacement curves.
Crack tip plastic zones remained small for the extent of the computed crack growth
(∆a < 10mm), whether plane strain or plane stress constraint was used. At crack initiation,
plane stress plastic zone ratios, ry/a, were typically 0.05, and at ∆a = 9mm, ry/(a + ∆a) was
around 0.26. Small scale yielding applied to this region of crack growth, and differences
between plane strain and plane stress computations at crack initiation were small (typically
10%). For a static crack, plane strain analysis gave the best estimate of near crack tip stress
and strain fields when compared to a 3-D analysis, but away from the near tip region, plane
stress conditions dominated. For a growing crack, the crack tip would quickly grow into a
plane stress field, and this was confirmed by examination of experimental data. It was clear
that shear lips developed at an early stage, so that the plasticity increased almost from the
beginning of crack propagation, resulting in a continuous increase of dJ/da. Hence, it was
considered that plane stress analysis gave a better estimate of the crack tip plastic work in
the early stages of crack propagation (∆a < 4mm) rather than plane strain conditions, which
existed around the crack tip prior to propagation.

COMPARISON OF JR CURVES
All the computed R-curves showed the same basic shape, whether computed using
plane stress or plane strain conditions, see Figures 3 and 4 for the plane stress curves. The
curves were 'concave', showing increasing values of dJ/da with crack extension, which is
 An R-Curve Assessment of Stable Crack Growth in an Aluminium Alloy 589

the opposite behaviour for more conventional R-curves. The shape corresponded to the
change from flat to slant fracture in the region of stable crack extension modelled
numerically. The mode of separation on both flat and slant plane remained ductile, and
showed the same microstructural characteristics.
0.5
0.45
0.4
0.35
0.3
J MN/m
0.25
0.2 3PB a/W=0.2
b>>B
0.15
3PB a/W=0.5
0.1 b=B

0.05 Experimental J
(CT)
0
0 5 10 15 20
Crack extension (mm)

Fig. 3. The JR-curves from numerical and experimental results.

0.1
CCT, a/W=0.5(plane
0.09 stress)
R curve, 3PB
0.08
R curve, CCT(Cubic fit)
0.07

0.06 R curve, CCT(Tstress-

index is applied to J1c)
J (MN/m) 0.05

0.04

0.03

0.02

0.01

0
0 2 4 6 8
Crack extension (mm)

Fig. 4. Comparison of plane stress JR-curve between 3PB and CCT specimens.

Crack growth in both a/W = 0.5(b = B) and the a/W = 0.2(b >> B) specimens was
stable [3], but the load behaviour was different after crack initiation (the specimens were
tested using displacement control). The load for the a/W = 0.5 specimen stayed constant
within ±10% of the maximum load, Figure 2a, whilst the load for the a/W = 0.2(b >> B)
3PB specimen rose with increasing CMOD, Figure 2b. Both specimens yielded similar R-
curves.
590 A. R. LUXMOORE

The form of the R-curve has since been confirmed experimentally by Sumpter[8],
Figure 3, after the numerical analysis was completed. He used a compliance unloading
technique (which gives a measure of the average crack extension) on an
a/W = 0.23(b >> B) CT specimen to obtain an R-curve for the material, using plane stress
calibrations. He used formulae from the ASTM standard [9] (which calculates the J-
integral for current crack lengths) to obtain Je and Jp, and hence JR = Je + Jp. He provided
an R-curve for a crack extension up to 18.5mm (the crack tunnelled, and tapered to a
point at ∆a = 18.5mm with a V-shaped cross-section) at which point the specimen failed
by unstable V-shaped fracture at 90° to the original notch direction. Another fracture test
by Sumpter [8] on an a/W = 0.5(b >> B) CT specimen provided a similar JR-curve for a
crack extension of 35mm. The numerical plane stress R-curve from the large ligament
3PB specimen agrees closely with the experimental result from the CT specimen, see
Figure 3. Both 3PB and CT specimens are high constraint geometries.
The higher J at initiation for the CCT test result (a low constraint geometry) in Figure
4 is correlated to a negative T-stress, which is rationalised by the JC-T locus, Figure 1. As
the JR-curve rises with crack extension, the curves from the CCT and 3PB (b >> B)
specimens converge, because the effects of in-plane constraint becomes insignificant due
to increasing amounts of slant fracture. The curve with the thicker line in Figure 4 is an
estimate of how the CCT curve would look when corrected for zero T-stress, hence
reducing J to 0.018MN/m at initiation. This estimated curve is close to the high constraint
JR-curve obtained from the 3PB specimen.

APPLICATION OF THE CONSTRAINT CORRECTED R-CURVE

For materials under elastic behaviour, the driving force (G) defines the change of
potential energy (Ue) per unit crack growth, (dUe/da). The crack propagates in a stable
fashion when [11]
G=R
dG dR
and ≤
da da
whereas for unstable crack growth;
dG dR
>
da da
where R is the fracture resistance.
Fracture tests [3] on the AlMgZn alloy showed that shallow cracks in small (b = B)
specimens failed at K1C, while in specimens with b >> B, failure was preceded by stable
crack growth. Instability prediction for three different 3PB specimens (all tested
experimentally), see Table 2, is assessed in this section to demonstrate fracture prediction
using the R-curve approach. Assuming the (experimental) plane stress JR curve and its JR
at initiation are independent of the crack tip condition and specimen geometry (i.e.
ignoring the T-stress effects), the invariant JR curve is reproduced for all crack
geometries. Instability of a crack geometry is determined graphically by comparing the
calculated driving force curve for the three geometries and the JR-curve. If the driving
force curve lies above the R-curve, then instability will occur at KIC, and vice-versa. If
the two curves cross, then instability can follow crack stability, or vice-versa.
 An R-Curve Assessment of Stable Crack Growth in an Aluminium Alloy 591

Table 2. Details of the 3PB and CT specimen.

3PB specimens Ligament length, Crack length, Span, Thickness,
a/W b (mm) a (mm) S (mm) B (mm)
0.1 25 2.5 110 25
0.5 25 25 200 25
0.2(b>>B) 160 40 800 25

In this report, the driving force, G, is calculated using the solution in BS 7448:Part 1
[11] with an effective crack length to produce a plasticity corrected Geff. Small scale
yielding is assumed because the finite element computations of the JR-curve have only
shown this behaviour. The plasticity correction makes a marginal correction to the elastic
value.
The Geff curves in Figure 5 are evaluated for constant load with increasing crack
length, i.e. increasing a/W ratios. The Geff is calculated for the maximum applied load in
the experimental test. An invariant (high constraint) R-curve is produced for each initial
crack length.
0.7
Effective G at P=53KN (for a/W=0.1)

0.6 Effective G at P=19.3KN (for a/W=0.5)

Effective G at P=250KN (for a/W=0.2)
0.5 R curve for given crack lengths

0.4
J (MN/m)
0.3

0.2

0.1

0
0 10 20 30 40 50 60 70
Crack length (mm)

Fig. 5. Instability prediction using R-curve approach for 3PB specimens of

a/W = 0.1(b = B), 0.5(b = B) and 0.2(b >> B).

For the specimen with a/W = 0.1(b = B), the Geff curve increases more rapidly than
the JR-curve, indicating that this specimen fails by unstable fracture at the applied load.
The Geff at initiation for this specimen was much larger than the JR value. This high G
value is associated with a negative T-stress, which raises the value of J at initiation. In
this case, T/σys = −0.35 [10], corresponding to JC = 0.03MN/m using the constraint index
[3], see Figure 1, while the computed Geff in Figure 5 is 0.028MN/m.
The Geff at initiation for the specimen with a/W = 0.5(b = B) is slightly larger than the
JR value, but dG/da is just less than dJR/da, and hence, initially, the crack extension can
be stable. However, Geff increases rapidly with ∆a, and soon dG/da > dJR/da, allowing the
Geff curve to cross the JR-curve. This is potentially an unstable situation in load control,
592 A. R. LUXMOORE

because small increases in load can increase Geff significantly. Cracks in specimens with
square ligaments (b = B) are known to be unstable under load control.
The Geff curve for the a/W = 0.2(b >> B) crosses the JR-curve (Figure 5), but dG/da is
much less than dJR/da at this point, and crack growth can only continue by increasing the
load. This coincides with the result of experimental test, which was stopped at a final
load of 247KN. The crack extension record of this test was not available. This specimen
is the closest to the structural situation.
All the above conclusions agree with the experimental observations [3].

SHEAR LIP MODEL

The parabolic nature of the R-curves suggested that a simple shear lip model might
explain the R-curve behaviour. Krafft et al [12] provided an empirical relationship
between the growth of shear lips and fracture resistance, on the basis of three
assumptions,
i. the shear lips were at 45° to the crack plane and the region bounded by these two
lines was a plastic deformation zone;
ii. this plastic zone was subjected to a constant plastic-work density, dWp/dV, for
all thicknesses of shear lips, at any stage of its formation;
iii. the energy to create new surfaces (dWe/dA) was supplied for the surface of flat
fracture in the central region and for the projected surface of slant shear lips, i.e.
the surface that was the product of the thickness and the extended crack length.
If S was the ratio of the total width of shear lips to the thickness, Figure 6, the total
plastic deformation zone on the crack flanks was (BS)2/2. An increment of work, dW, for
a growing crack was composed of an 'areal' (flat fracture) and volumetric (plastic
deformation) components,

dWe dWp (BS)2

dWt = Bda + da . (1)
dA dV 2
dWt
where = total energy consumption per unit crack extension,
da
dWe
= energy consumption per unit crack growth area,
dA
dWp
= energy consumption per unit volume of shear lip,
dV
Krafft et al[12] showed that the dWe/dA and dWp/dV could be determined from the
plot of R versus BS2/2 data from a test. From the results of their tests on 7075-T6 alloy,
they found dWe/dA and dWp/dV to be 68in.lbs/in2 and 8300in.lbs/in3 (i.e. 0.012MNm/m2
and 57.2MNm/m3) respectively.
 An R-Curve Assessment of Stable Crack Growth in an Aluminium Alloy 593

Half of specimens
Projected surfaces
thickness
of the shear lips

Shear lip width

Flat fracture
(1–S)B surface

∆a

Maximum flat
fracture extension

Fig. 6. The projected surface of the shear lips on the plane of flat fracture.

Green and Knott [13] suggested that the fracture of the flat and shear modes were
independent, but this made little practical difference to the analysis.
Previous test results [13-15] reported that the width of the shear lips reached a
maximum of between 2 to 3mm. It was also proportional to the initial ligament size, i.e.
S/b0 = 0.2 and 0.4, in tests on titanium alloy and HY130 steel respectively [14]. Results of
fracture testing on the AlMgZn alloy, presented in this paper, show that shear lips can
develop across the full 25mm thickness in the large specimens. The fracture mode
changed from flat to fully slant in specimens with rectangular (b >> B) ligaments, where
b is the ligament length.

TRANSITION FROM FLAT TO SLANT FRACTURE

For an initially flat (fatigue) crack, perpendicular to the direction of load in a parallel
sided test piece, slant fracture will start at the free surfaces and grow at approximately
45° to the flat fracture, which continues to tunnel in the middle of the specimen. For
small test pieces of the AlMgZn alloy, with approximately square ligaments, the shear
lips appear to reach some maximum value, as reported for other alloys [13-15]. In 10mm
thick CT specimens with square ligaments, Sumpter [8] observed shear lips reaching a
maximum width of 2mm, but in the 25mm thick CT specimens of the AlMgZn alloy, the
CT specimens with the crack in the LT orientation of the extrusion had maximum shear
lips of 5mm per surface.
For large ligament CT specimens, complete transitions from flat to slant fracture were
observed in 10 and 25mm thicknesses. After the transition was completed, crack growth
continued by slant fracture alone [8]. Complete transitions were also reported by Sumpter
[3] in 25mm thick large ligament 3PB and CCT test pieces. The main difference between
the transition in the three 25mm thick geometries was the amount of crack extension,
∆atrans, required to complete the transition: 18mm for the CT (a/W = 0.2), 28mm in the
594 A. R. LUXMOORE

3PB (a/W = 0.2), and 25mm in the CCT (a/W = 0.5). For the 10mm thick large ligament
specimen (a/W = 0.5), complete transition occurred after 6 mm of crack extension.
The reason for the transition from flat to slant fracture is not entirely clear. Shear lips
will saturate when the rate of change of the applied energy release rate exceeds the rate of
change of fracture resistance [16].

SHEAR LIP ANALYSIS

The JR curves for all the large ligament specimens used in this analysis are given in
Figures 3 and 4. Using the model due to Krafft et al, equation (1), the energy per unit
crack growth, i.e. BdWe/dA, is constant for the full range of crack extension, and J can be
plotted against BS2/2 to give a linear plot. The data used for the shear lip analysis was
taken from the curves fitted to the R-curve values by the least squares method.
For the low constraint CCT specimen, the JR curve is not parabolic, because of the
elevation of the initiation JR value by the loss of constraint. Figure 4 shows that this curve
can be fitted satisfactorily (using least squares) by a cubic polynomial with the following
equation:
J = 0.1558∆a3 – 0.46∆a2 + 0.4253∆a + 27.282 (N/mm) (2)
The first 2–3 millimetres of this curve are almost linear, and it is possible, by suitable
manipulation, to split equation (2) into an initial linear section, covering the first 2mm of
crack growth, and a parabolic curve covering the remaining crack growth. The
computation of the parabolic equation from (2) is very sensitive to the values of the first
two coefficients in equation (2). It was found more satisfactory to fit a parabola by least
squares in the region ∆a = 3 to 8mm, but including the JR initiation value obtained from
the J-T indexing curve, see Figure 4. This latter parabola was the JR curve used in the
shear lip analysis for the CCT specimen, and was very similar to the JR curves computed
directly from the measured data, see Figure 4.
The first 2-3mm of crack growth is covered by the pop-in observed in the load v. clip
gauge displacement curve.
The JR curves computed by the finite element method (3PB with a/W = 0.2 and 0.5,
and the CCT specimen) only covered the first 8-9 mm of crack extension. The
experimental curve for the CT specimen, provided by Sumpter [5], covered a crack
extension of 18mm, after which the specimen failed in the transverse direction. A
detailed photograph was available to the authors for this specimen, and the shear lip
width at several crack extensions could be measured with good accuracy. These
measurements were used to plot JR versus BS2/2, Figure 7, which shows a reasonable
linear variation, although with some deviation associated with the slight curvature of the
shear lip shape (this deviation could also be used to justify the Knott model [13]). For the
3PB and CCT specimens, the photographs could only be used to estimate the transition
range, ∆atrans, and a linear variation between S and the crack extension was assumed.
The results are given in Table 3, which gives the values of dWp/dV and dWe/da, as
well as the transition range, ∆atrans.
 An R-Curve Assessment of Stable Crack Growth in an Aluminium Alloy 595

0.5

0.45 2
JR = 36.7(BS /2) + 0.0153
0.4

0.35

0.3
J MN/m
0.25

0.2

0.15

0.1
Experimental
0.05 results

0
0 0.002 0.004 0.006 0.008 0.01 0.012 0.014
2
BS /2 (m)

Fig. 7. Plot of JR versus BS2/2 for CT specimen to determine elastic

and plastic work rates from the Krafft et al model.

Table 3. Elastic and plastic work rates from shear lip analysis
3PB CCT CT
Specimen
(a/W = 0.2) (a/W = 0.5) (a/W = 0.2)
Krafft dWP/dV (MNm/m3) 64.0 53.7 36.7
dWe/dA (MNm/m2) 0.017 0.016 0.015
∆atrans (mm) 28 25 18
(dWP/dV)/ ∆atrans (MN/m3) 2285 2148 2039
Cone dWp/dV (MNm/m3) 18.2 17.1 16.2

The range of values for dWP/dV scale with the transition distance, ∆atrans, which is not
unexpected. A constant dWP/dV for all geometries, plus a single, independent, R-curve
implies a constant ∆atrans. The Krafft model only considers the variation of the shear lips
through the thickness, and does not consider the effects of in-plane changes in plastic
zone size. The finite element analysis showed the plane stress plastic zones at initiation
were approximately circular, and centred on the crack tip. If we assume that the plastic
deformation at any crack extension is contained within two cones, of height BS/2 and
radius S∆atrans (which keeps one edge of the cones coincident with their starting point)
then the volume of the cone is given by (π/3)(BS/2)(S∆atrans)2. Differentiating, we have

 BS2   2
dV = π ∆atrans 2 dS = π BS ∆atrans da (3)
 2   2 
   
as dS = da/∆atrans. Substituting into equation (1), the coefficient for dWp/dV now becomes
(πS2∆atrans/2), instead of (BS2/2). Modifying the original Krafft analysis for the cone
assumption, we get the values for dWp/dV shown in line 5 of table 3, by multiplying line
4 by B/π.
596 A. R. LUXMOORE

This analysis gives reasonable constant estimates for dWp/dV, but loses the 45o
boundary for the shear lips. These boundaries now vary with geometry, as they are
dependent on ∆atrans, producing much steeper slant boundaries. A more detailed analysis,
assuming an elliptical plastic zone on the surface, did not give consistent results.

CONCLUSIONS
The increase of fracture resistance (compared with the plane strain KIC) during large-
scale fracture testing was due to a combination of low constraint and stable crack
extension. The elevated fracture toughness for low constraint geometry could be indexed
by negative T-stresses. Fracture assessment, using a T-stress indexed fracture toughness
alone, underestimates the critical crack length in structural components.
Numerical analysis of stable crack extension produced a concave JR-curve, and this
was confirmed later by experimental evaluations. After corrections for the T-stress, a
reasonably geometry independent R-curve was obtained. The concave shape was
associated with the developments of shear lips. Fracture assessment using the R-curve
approach showed that the extension of a crack in large ligament (b >> B) specimens of
the alloy are stable because dG/da < dJR/da, caused by dJR/da increasing with ∆a. For
shallow cracks in small (b = B) specimens, initial instability leads to final fracture. The
prediction of instability using the R-curve approach in small (b = B) and large (b >> B)
specimens agreed with experimental observations.
The difference of JR between 3PB (a high constraint geometry) and CCT (a low
constraint geometry) specimens at initiation is the result of in-plane constraint effects,
which can be represented by the T-stress in this study. The JR-curve for high and low
constraint geometries converged after 3mm or so of crack extension. The R-curve
approach must be corrected by T-stresses when it is applied to low constraint geometries.
The JC value is therefore correlated to ∆a and T-stresses.
A shear lip model was used to explain the 'concave' R-curve, based on the work of
Krafft et al [1]. The model gave a good fit to the R-curve, but the plastic work rate,
dWp/dV, was not constant for the different test piece geometry. It was found that this
work rate could be scaled by the transition crack extension, ∆atrans, i.e. the crack extension
in which the mode of fracture changed from fully flat to fully slant. A 'cone' model gave
consistent values for dWp/dV, but sacrificed the 45o shear lip assumption of Krafft et al.
More detailed analysis is required.

Acknowledgements. The author thanks Dr. J. D. G. Sumpter of the Defence and Evaluation
Research Agency, UK, for providing all the experimental data used in this paper, and the Agency
for permission to publish the data.

REFERENCES
1. Webber, D., (1985) Fatigue Design and Testing of an AlZnMg Alloy Bridge Girder, Royal Armament
Research and Development Establishment, Christchurch, Dorset, presented at the Third International
Conference on Aluminium Weldments, Munchen.
2. Henry, B.S., and Luxmoore, A.R., (1996) Two-Parameter Fracture Assessment of Through-Thickness
Cracks in an Aluminium Bridge Structure, Eleventh European Conference on Fracture, ECF-11,
Futuroscope, France, pp.2145-2150.
 An R-Curve Assessment of Stable Crack Growth in an Aluminium Alloy 597

3. Sumpter, J.D.G., (1996) Observations on Tearing Instability in an Aluminium Alloy, Mechanisms and
Mechanics of Damage and Failure, ECF-11, Vol. 2, pp.855-860.
4. Henry, B.S., Luxmoore, A.R. and Sumpter, J.D.G., (1996) Elastic-Plastic Fracture Mechanics Assessment of
Low Constraint Aluminium Test Specimens, International Journal of Fracture Vol. 81, , pp.217-234.
5. Rice, J.R., (1974) Limitation to Small Scale Yielding Approximation for Crack Tip Plasticity, Journal of the
Mechanics and Physics of Solids, Vol. 22, pp. 17-26.
6. O'Dowd, N.P. and Shih, C.F., (1992) Two-Parameter Fracture Mechanics: Theory and Applications, ASTM
24th National Symposium on Fracture Mechanics, Tennessee.
7. Henry, B.S., Constraint Based Fracture Assessment of Through-Thickness Cracks in a Bridge Girder
Structure, Ph.D. Thesis C/PH/194/96, Department of Civil Engineering, University of Wales Swansea,
March 1996.
8. Sumpter, J.D.G., Fracture and Fatigue Testing of a 7019 Aluminium Bridging Extrusion, Restricted
commercial report, Defence and Evaluation Research Agency, UK, November 2000.
9. ASTM E 1820-99a, (2000) Standard Test Method for Measurement of Fracture Toughness, American
Society for Testing and Materials,
10. Anderson, T.L., (1995) Fracture Mechanics-Fundamentals and Applications, Department of Mechanical
Engineering, Texas A&M University, College Station, Texas, CRC Press, Inc., Second Edition,
11. BS 7448 Part 1, (1991) Fracture Mechanics Toughness Tests, British Standard Institution.
12. Krafft, J.M., Sullivan, A.M. and Boyle, R.W., (1962) Effect of Dimensions on Fast Fracture Instability of
Notched Sheets, Proceedings of the Crack Propagation Symposium Cranfield, The College of Aeronautics,
Vol. 1, pp. 8-28.
13. Green, G. and Knott, J.F., (1975) On Effects of Thickness on Ductile Crack Growth in Mild Steel, Journal of
the Mechanics and Physics of Solids, Vol. 23, pp. 167-183.
14. Turner, C.E. and Braga, L., (1993) Energy Dissipation rate and Crack Opening Angle Analyses of Fully
Plastic Ductile Tearing, Constraint Effects in Fracture, ASTM STP 1171, American Society For Testing and
Materials, Philadelphia, pp. 158-175.
15. Gibson, G.P., Druce, S.G. and Turner, C.E., (1988) Effect of Specimen Size and Geometry on Ductile Crack
Growth Resistance in a C-Mn Steel, International Journal of Fracture Vol. 37, pp. 83-100.
16. Cheung, S. and Luxmoore, A.R., A Finite Element Analysis of Stable Crack Growth in an Aluminium
Alloy, to be published in Engineering Fracture Mechanics.

PROCENA NAPREDOVANJA STABILNE PRSLINE

U ALUMINIJUMSKOJ LEGURI PO R-KRIVULJI
A. R. Luxmoore
AlMgZn legura pokazala se krtom kada je testirana korišćenjem standardnih lomnih uzoraka,
ali su veći uzorci sa istom konturnom geometrijom dali stabilan lom, tj. kontrolisano prostiranje
prsline ili pod rastućim opterećenjem, ili pod rastućim pomeranjem. Analiza po R-krivulji dala je
R-krivulju za leguru koja je prilično neobičnog oblika, ali jedinstvena, kod koje se otpornost prsline
parabolično povećava sa prostiranjem prsline. Pored toga, zapodinjanje prsline bilo je indeksirano
ograničenjem testnog uzorka. Uprkos ovim komplikacijama, analiza po R-krivulji dala je
zadovoljavajuće objašnjenje za različito ponašanje pri lomu malih i velikih testnih uzoraka.
Parabolična priroda R-krivulje, koja je dobijena iz više različitih geometrija, sugeriše da bi
jednostavna analiza po smicajnoj ivici, slična Kraftovoj, Bojlovoj i Salivenovoj, trebalo da objasni
ponašanje testnih delova, ali uprkos odredjenim izmenama prvobitne teorije smicajne ivice,
rezultati nisu uverljivi.