EFFECT OF SPECIMEN THICKNESS AND STRESS

RATIO ON FATIGUE CRACK GROWTH AFTER

A SINGLE OVERLOAD CYCLE
ON STRUCTURAL STEEL

M.Skorupa1, A. Skorupa1, J. Schijve2, T. Machniewicz1 and P. Korbut1

1
University of Mining & Metallurgy,
Al. Mickiewicza 30, 30-059 Kraków, Poland,
2
Delft University of Technology,
Kluyverweg 1, P.O.Box 5058, 2600 GB Delft, The Netherlands,

ABSTRACT

Effects of specimen thickness and stress ratio on fatigue crack growth transients after a single overload
applied among constant amplitude cycles of a smaller amplitude have been studied experimentally for a low
carbon structural steel. The corresponding crack growth data from the fatigue tests have been presented and
evaluated. The overload cycles introduced crack growth retardation in all cases and the transient crack
growth response depended on a specific combination of specimen thickness and stress ratio. The retardation
effect at a given stress ratio decreased as specimen thickness was increased. For a given specimen thickness,
the benefit from the overload application was larger at a higher stress ratio than at a lower stress ratio. The
test variables predominantly affected crack growth retardation during the period when the crack growth rate
was recovering from the minimum value, whilst their effects prior to this minimum (i.e. in the delayed
retardation period) were negligible. The observed experimental trends were compared to those reported in
the literature for other structural steels.

INTRODUCTION

Since the use of damage tolerance designs became established in many industries, much attention has been
devoted to load interaction phenomena in fatigue crack growth which typically occur under variable
amplitude load histories. The transient crack growth behaviour following imposition of a single overload
(OL) during constant amplitude (CA) cycling of a smaller amplitude is among the most documented load
interaction effects. However, as recognized in a recent review study [1], the corresponding data bank for
structural steels remains by far more meager than in the case of materials used in the aircraft industry.
Especially, the knowledge of the influence of various factors related to loading, material and specimen
geometry on OL-induced retardation in crack growth is far from being complete.

This paper examines the effects of specimen thickness (t) and stress ratio (R) on post-OL fatigue crack
growth in a low carbon structural steel. The corresponding results of the fatigue crack growth tests are
presented and evaluated. A comparison is made between the observed experimental trends and those
reported in the literature for various steels. The work was intended to create an experimental data bank for
structural steel required to improve a crack growth prediction model developed by the authors.

EXPERIMENTAL TECHNIQUES

The material used is 18G2A low carbon structural steel, composition 0.14C-1.36Mn-0.21Si-0.020P-0.020S-
0.11Cu-0.11Cr-0.05Ni. Mechanical properties in the rolling direction are: yield stress = 392 to 402 MPa,
tensile strength = 536 to 544 MPa, and elongation to failure = 22 to 28%. Central crack tension specimens of
various thicknesses (Fig. 1) were machined from a 20 mm thick plate with the specimen axis in the rolling
direction. Prior to testing, all specimens were stress-relieved.

Figure 1: Fatigue specimen geometry, dimensions in mm

All fatigue crack growth tests were performed under load control at 15 to 30 Hz using a Dartec computer-
controlled closed-loop servohydraulic fatigue machine with load capacity of 250 kN. A survey of the
experiments is given in Table 1. A single 100% OL was always applied at a frequency of 0.1 Hz at the half
crack length aOL = 13 mm, namely when the fatigue crack advanced 3 mm after precracking had been
completed.
TABLE 1
DESIGN OF THE FATIGUE CRACK GROWTH TESTS

SOL
Smax
Smin

Specimen Stress ratio Thickness Test Stress levels (MPa)

No. R t (mm) type Smin Smax SOL
0205 0 4 CA 4.3 84.3
0206 0.5 4 CA 80 160
0211 0.7 4 CA 116.67 166.67
0202 0 12 CA 4.3 84.3
0205 0.5 12 CA 80 160
0212 0.7 12 CA 116.67 166.67
0213 0 18 CA 4.06 79.62
0210 0 4 VA 4.3 84.3 164.3
0209 0.5 4 VA 80 160 240
0208 0 12 VA 4.3 84.3 164.3
0207 0.5 12 VA 80 160 240
0214 0 18 VA 4.06 79.62 155.17
The crack length was monitored with an accuracy of ±0.02 mm by a direct current potential drop method
and visually, employing a travelling microscope. Fatigue fracture surfaces were observed macroscopically
and microscopically in SEM in order to see the crack front marking of the OL and to reveal possible
striations.
FATIGUE TEST RESULTS

The crack growth curves from all tests are plotted in Fig 2 where N denotes number of cycles after
precracking. It is clear that the OL cycles introduced crack growth retardation in all tests and that the
retardation effect at a given R-ratio decreased as specimen thickness was increased. Under CA loading
conditions, however, the influence of t was negligible.

Figure 2: Crack growth curves from all fatigue tests

An incremental polynomial method using a 7-point fit was employed to compute the fatigue crack growth
rate (FCGR) from the data presented in Fig. 2. The results are shown in Figs 3 and 4 by plotting the FCGR
(da/dN) as a function of crack length (a). As seen in Figs 3a and b, a thinner specimen always shows more
retardation than a thicker specimen at the same R-ratio. The above trend is more pronounced when specimen
thicknesses of 4 mm and 12 mm are considered whereas only minor differences can be noted between the
behaviour of the 12 mm thick and 18 mm thick specimens. The thickness effect is more significant at R =
0.5 (Fig. 3b) than at R = 0 (Fig. 3a).

It is apparent in Fig. 4a that the retarded FCGRs observed in the R = 0 and R = 0.5 test on the 4 mm thick
specimens coincide, except at highest values of a. This behaviour is in contrast to that shown in Fig. 4b by
the 12 mm thick specimens. Here the FCGRs recover slower from the minimum da/dN value at R = 0 than at
R = 0.5.

The magnitude of the instantaneous load interaction effect after an OL can be represented by the FCGR
observed in the OL test, (da/dN)OL, normalized by the FCGR measured in the CA test at the same crack
length for the same specimen thickness and R-ratio, (da/dN)CA. It can be seen in Fig. 5 where the
(da/dN)OL/(da/dN)CA ratios from all the tests are plotted against the post-OL crack increment (a - aOL) that
the initial acceleration following an OL application only occurred in the R = 0 tests on the 4 mm thick and 12
mm thick specimen. This is also evident in Figs 3a where the da/dN values measured in the above mentioned
two tests jump immediately after the OL above the level observed in the corresponding CA tests. Fig. 5
further shows that the minimum in FCGRs is always attained when the crack has grown some distance after
an OL. This implies that the so called delayed retardation occurred under all test conditions. The a-aOL crack
growth increment corresponding to the minimum post-OL da/dN value was nearly the same in all tests (0.38
- 0.51 mm), as also was the minimum (da/dN)OL/(da/dN)CA ratio (0.08 - 0.12 except of a higher value of 0.33
for the 18 mm thick specimen). From Fig. 5, for specimen thickness of 4 mm, more retardation occurred at
R = 0.5 than at R = 0. A conspicuous feature observed in Fig. 5 for the larger thicknesses and also apparent in
Figs 3 and 4 is accelerated crack growth following the retardation phase. This effect is most pronounced for
the 18 mm thick specimen.

The overall retardation effect induced by an OL can be quantified using the measures shown in Fig. 6,
namely the delay period (defined alternatively as ND’ or ND’’) and the delay distance also referred to as the
OL-affected zone (∆aOL). The normalized ND’’/NCA parameter seems to be a more realistic measure of the
amount of retardation than ND’ or ND’’, as it quantifies the benefit from applying an OL under given test
conditions.

(a)
10 -4
(mm/cycle)
da/dN

R=0
10 -5

a (mm)

(b)
10 -4
(mm/cycle)
da/dN

10 -5 R = 0.5

a (mm)

Figure 3: The effect of specimen thickness on post-OL crack growth rates: (a) at R = 0;
(b) at R = 0.5

The observed effects of t and R on the retardation parameters depicted in Fig. 6 are summarized in Figs 7a
and b. The ND’ vs. t data in Fig. 7a confirm the trend also visible in Fig. 2: increasing specimen thickness
reduces the delay period. When the test results are viewed in terms of ND’’/NCA vs. t it is apparent that larger
profits from the OL application have been gained at R = 0.5 than at R = 0. At R = 0 the OLs brought about
equally beneficial results for the 4 mm thick and the 12 mm thick specimen (same ND’’/NCA values), whilst
for the 18 mm thick specimen the OL effect on crack growth was dramatically smaller.

(a)
10 -4

(mm/cycle)
da/dN

t = 4 mm
10 -5

a (mm)

(b)
-4
10
(mm/cycle)
da/dN

t = 12 mm
10 -5

a (mm)

Figure 4: The effect of stress ratio on post-OL crack growth rates: (a) in 4 mm thick
specimens; (b) in 12 mm thick specimens

Figure 5: Variations of normalized crack growth rates throughout the OL tests

 Figure 6: Definition of parameters to quantify the retardation effect

(a) (b)

Normalized delay distance, ∆aOL/rOL

Normalized delay period, ND”/NCA
Delay period, ND’ (kcycles)

Delay distance, ∆aOL (mm)

Thickness, t (mm) Thickness, t (mm)

Figure 7: The effects of specimen thickness and stress ratio on: (a) delay period;
(b) OL-affected zone

In terms of ∆aOL, the most effective was an OL applied on the 4 mm thick specimen in the R = 0 test, Fig. 7b.
The ∆aOL vs. t data further show that delay distances observed at R = 0 for 12 mm and 18 mm specimen
thickness are similar. A normalized parameter often cited in the literature, e.g.[2-4], is ∆aOL/rOL, where rOL is
the OL plastic zone size. The ∆aOL/rOL vs. t data with rOL computed for plane stress conditions show similar
tendencies as the ∆aOL vs. t data, Fig. 7b. Note that at R = 0 the delay distances exceed the corresponding
plastic zone sizes, i.e. ∆aOL/rOL > 1, even for the thicker specimens.

Results generated on structural steels show that an increase of specimen thickness reduces the retardation
phenomenon, both in terms of the delay period and the OL-affected zone [2,3], which is in agreement with
the present work (see Fig. 7). However, Shuter and Geary [3] obtained a linear relationship between log ND’’
and t. It implies a more significant effect of thickness on the delay period than seen in Fig. 7a. As recognized
by Fleck [2], ∆aOL measured on the surface of thinner specimens at R = 0 can significantly exceed rOL,
whereas ∆aOL measured on the surface of thick specimens is always smaller than rOL [2,3]. In the present
tests, however, ∆aOL at R = 0 was in all cases greater than rOL (Fig. 7b). Investigations [2,3] indicated that an
increase of specimen thickness caused the minimum in da/dN to occur earlier and become less deep. In
contrast, Figs 3 and 5 demonstrate that both the minimum da/dN value and the distance at which this
minimum is reached do not depend on thickness. It is apparent that specimen thickness affects retardation
only in the stage when FCGRs are recovering from the minimum value.

Systematic investigations on R-ratio effects on post-OL crack growth transients in structural steels are
lacking in the literature. Shin and Hsu [4] noted the normalized minimum FCGR in stainless steel to become
higher and ∆aOL to decrease as the R-ratio was increased. Only the latter trend is in accordance with the
present results, Fig. 7b. The ratio of the minimum da/dN position from point of OL to OL-affected zone
ranges in our tests from 0.05 to 0.15 which is much lower than most of the values which can be estimated
from test data presented in other works [2-4].

In contrast to the present work, all literature results cited above have been obtained from K-controlled tests
on CT specimens. Hence, the differences between the post-OL crack growth behaviour observed here and
that reported by others may stem not only from material related aspects but also from the different loading
conditions and specimen geometry.

FRACTOGRAPHY

The macroscopic examination of the fatigue fracture surfaces indicated that each OL cycle produced a
marking line which was vaguely visible only. It was slightly curved, i.e. the crack length at mid thickness
was larger than at the specimen surface, a difference in the order of 0.5 mm. Curved crack fronts
corresponding to OL cycles were also reported by Matsuoka and Tanaka [5]. Shear lips, typically observed
for aluminium alloys [6] and also for a high strength steel [5], were not formed.

An OL marking line observed in the SEM was rather tortuous, and at a high magnification could not always
be followed continuously over the full thickness of the specimen, Fig. 8. Close to the specimen surface, the
OL marker lines could not be observed any more. A tentative explanation is that these lines may have been
erased due to enhanced closure in the „plane stress” regions near the specimen surfaces. The fatigue fracture
surfaces showed a rather irregular pattern with tortuous lines, as also evident for a low carbon steel (0.18C,
0.92Mn) from a SEM examination by Shuter and Geary [3]. No striations of the base line fatigue load cycles
were found in the SEM. Apparently, the corresponding FCGR (in the range of 0.01 – 0.1 µm/cycle) was too
low to lead to an easy detection in the SEM for this type of steel. In this respect, fatigue fracture surfaces of
cracks in Al-alloys can be more instructive because they usually show many striations. Also, the correlation
between a variable amplitude load history and the striation pattern has been found to be most useful [6].

a) b)

Figure 8: SEM micrographs of the fracture surface of Specimen 0207 at the OL position:
(a) at a lower and (b) at a higher magnification. Vertically, in the centre of both
pictures the crack front of the OL
ACKNOWLEDGEMENTS

The authors would like to acknowledge the partial financial support of the State Committee for Scientific
Research via Grant No. PB 1459/T07/97/12.

REFERENCES

1. Skorupa M. (1998) Fatigue Fract. Engng Mater. Struct. 21, 987.

2. Fleck N.A. (1988) ASTM STP 924, vol. 1, 157.
3. Shuter D.M. and Geary W. (1996) Fatigue Fract. Engng Mater. Struct. 19, 185.
4. Shin C.S. and Hsu S.H.( 1993) Int. J. Fatigue 15, 181.
5. Matsuoka S. and Tanaka K. (1980) Engng Fracture Mech. 13, 293.
6. Schijve J. (1999) Fatigue Fract. Engng Mater. Struct. 22, 87.