Engineering Fracture Mechanics 66 (2000) 41±49

www.elsevier.com/locate/engfracmech

A note on a simple recategorisation of cracks in fatigue

analysis
F.D. Fischer a,*, F.G. Rammerstorfer b
a
Institute of Mechanics, MontanuniversitaÈt Leoben, Franz-Josef-Straûe 18, A-8700 Leoben, Austria
b
Institute of Lightweight Structures and Aerospace Engineering, Vienna University of Technology, Guûhausstraûe 27-29, A-1040
Vienna, Austria
Received 26 October 1999; received in revised form 16 December 1999; accepted 3 January 2000

Abstract

Two procedures for the fatigue analysis of elliptic and semi-elliptic cracks proposed in engineering guidelines are
compared. The ``original'' procedure follows the growth of an actual crack, the ``simpli®ed'' one considers the
growth of a recategorised strip (tunnel) crack. The conservatism of the ``simpli®ed'' procedure is investigated in an
extensive computer study. 7 2000 Elsevier Science Ltd. All rights reserved.

Keywords: Fracture mechanics; Fatigue analysis; Engineering guidelines; Elliptical crack; Semi-elliptical crack; Recategorisation;
Conservatism

1. Introduction

When investigating the residual life of a construction or a workpiece containing a detected ¯aw,
usually several simpli®cations are performed to ®nd a prediction. These simpli®cations are mainly
concerned with the guidelines which are applied for estimating the residual life. Nowadays, several
guidelines dealing with fracture mechanics application exist. The authors refer here mainly to the British
guidance PD 6493 [1] and an Austrian guideline [2] which is based on [1], but has been signi®cantly
modi®ed and extended in part Ð speci®cally with respect to fatigue analysis, see also [3]. Under these
aspects, the following simpli®cations are necessary:
. The existing ¯aw must be modelled by a plane crack, if the ¯aw is ``voluminous''.
. The plane crack must be projected into the direction of the proper principal stress axis. The

* Corresponding author. Tel.: +43-3842-402-475; fax: +43-3842-46048.

E-mail addresses: ®scher@unileoben.ac.at (F.D. Fischer), ra@ilfb.tuwien.ac.at (F.G. Rammerstorfer).

0013-7944/00/$ - see front matter 7 2000 Elsevier Science Ltd. All rights reserved.
PII: S 0 0 1 3 - 7 9 4 4 ( 0 0 ) 0 0 0 0 8 - 4
42 F.D. Fischer, F.G. Rammerstorfer / Engineering Fracture Mechanics 66 (2000) 41±49

conservatism of this simpli®cation is dealt with in detail by the authors in [4].

. In conjunction with the second step, the shape of the crack must be categorised. Here, the guidelines
provide the user with several recommendations. If the considered crack does not penetrate completely
through the wall (a ``trough crack'' can be represented by a line crack), the user has to deal with a
full elliptical crack within a workpiece or the wall thickness of a plate or shell (e.g. of a vessel) or a
semi-elliptical crack representing a typical surface crack.
. Since the stress intensity factor KI possesses a distribution dependent on the polar angle, j, KI j†, the
evaluation of the KI-distribution would, in general, be necessary. If crack propagation has to be
considered, as in the case of fatigue analysis, the crack con®guration changes gradually and, as a
consequence, such an evaluation must be performed numerous times. Although proposals exist in the
literature to automate the procedure, see e.g. [5], the procedure is usually very time-consuming,
speci®cally if the elasto-plastic behaviour and the crack closure behaviour are taken into account, see
e.g. [6]. Therefore, guidelines, such as [1,2], provide rules for a recategorisation of the elliptical crack
by a ``tunnel crack'' with the crack fronts being parallel to the (plane) contour of the wall, or a
recategorisation of a semi-elliptical (surface) crack by an in®nitely long surface crack. Appropriate
methods of recategorisation will be dealt with in the next section.
Now, the main aim of this paper is to check the conservatism of this type of recategorisation with
respect to a fatigue analysis.

2. Fatigue analysis and recategorisation

This section explains de®nitions and the way of recategorisation given in [1,2].
Two types of cracks are dealt with, see Fig. 1 and the corresponding notation. Moreover, a linear
stress distribution over the cross-section is assumed.
The Erdogan±Paris±Sih relation is taken as crack propagation law,
da m
ˆ C DK † : 1†
dN
DK is the di€erence between the maximum and the minimum stress intensity factor KI in a point on the
crack front during a load cycle,
DK ˆ KI, max ÿ KI, min , 2†

KI, min must be positive.

s1 t† is the time history of the longitudinal stress on the lower surface of a plate, s2 t† is that for the
upper surface of a plate. Here, a proportional time history of the stress state is assumed leading to a
constant ratio s1 =s2 with time.
No threshold value DK0 is assumed for the sake of simplicity. N stands for the number of cycles. C
and m are the usual material constants for crack propagation. Relation (1) can be reformulated as
ÿ
m
Dai j † ˆ C DKi j † , 3†

for the crack growth during the ith cycle. Dai means the crack propagation increment. Only two
``simple'' locations exist, where the crack propagation increment Dai can simply be added: the point B,
where Dai is added to the current crack length ai in x-direction and the points A or A1 and A2,
respectively, where Dai points in the y-direction. The well-known fact that in reality, the crack looses its
elliptical shape should be noted. However, the crack pro®les are often Ð speci®cally in the case of
 F.D. Fischer, F.G. Rammerstorfer / Engineering Fracture Mechanics 66 (2000) 41±49 43

tension and bending loading Ð truly very close to being semi-elliptical or elliptical as observed from
experiments and numerical simulations [5,7±9]. Therefore, within the context of [1,2], the deviation from
the elliptical or semi-elliptical crack shape, respectively, needs not to be taken into account. This means
a0 and 2a0 for the surface crack and for the internal ¯aw, respectively, grows to a and 2a, respectively,

X
N X
N X
N
a ˆ a0 ‡ Dai , 2a ˆ 2a0 ‡ D 1 ai ‡ D2 ai : 4†
iˆ1 iˆ1 iˆ1

The left superscripts ``1'' or ``2'' refer to the position A1 or A2.

Fig. 1. (a) Geometry of a surface crack (solid line: initial, dotted lines: during crack growth). (b) Geometry of an internal (central
or subsurface) crack (solid line: initial, dotted lines: during crack growth).
44 F.D. Fischer, F.G. Rammerstorfer / Engineering Fracture Mechanics 66 (2000) 41±49

For the sake of simpli®cation, we assume in this paper that D1 ai is substituted by D2 ai which is
conservative.
Similarly, `0 grows to

X
N
` ˆ `0 ‡ 2 D B ai : 5†
iˆ1

DB ai refers to the crack propagation increment in the point B. If in the following context, the growth of
an elliptical or semi-elliptical crack is treated, ellipses with a horizontal total length ` and a vertical half
length a are considered.
It should be mentioned that this procedure can not be directly applied to more complicated crack
con®gurations, for example, interacting and coalescing cracks, see e.g. [10±12], or cracks near holes, see
e.g. [13].
If the internal crack approaches the surface and the original distance 2p0 decreases to 2 p  T,

X
N
2
p ˆ 2 p0 ÿ D2 ai , 6†
iˆ1

it may happen that the stress intensity factor KI in A2 increases very rapidly. Finally, KI goes to in®nity,
if A2 reaches the surface. This is pointed out in a series of papers by Mayrhofer et al. [14±17], a paper
by Belytschko et al. [18] applying the self-similar crack expansion method, a recent paper by Smith et al.
[19], and speci®cally with respect to its consequence on fatigue analysis in [20]. Therefore, the rapid
growth of the crack near the surface is treated in such a way that the elliptical crack is reshaped for
consideration of subsequent crack growth by a semi-elliptical surface crack. Diagrams for the stress
intensity factor in A2 are provided in the open literature mostly for a= a ‡ 2 p† < 0:8: The authors
extended the diagrams based on their own research [14±17] to a range of validity of a= a ‡ 2 p† < 0:9:
For values of a= a ‡ 2 p† > 0:9, a break-through to the surface is assumed, forming a semi-elliptical
surface crack with the depth
a
as0 ˆ 2a ‡ 2 pf , 2
pf ˆ : 7†
9

The major axis `s0 is assumed to be the length, maximum of the two values `, see Eq. (5), and 2as0
ÿ

`s0 ˆ max `, 2as0 : 8†

This means that the crack reshapes for ` < 2as0 to a semi-circular surface crack. This practice is justi®ed
by a recent study by Dai et al. [21]. They could show that in the ®rst phase, an internal crack grows
nearly una€ected by the surface. After touching the surface, it breaks through and grows within a
certain transient period of only few cycles to a nearly semi-circular con®guration. Omitting this transient
interim period leads to a conservative procedure which is also followed in this paper.
The recategorisation is now performed by equalising the (largest) stress intensity factor di€erence DK
of the actual crack with the corresponding value DK of a tunnel or in®nitely long surface crack,
respectively. Since DK follows as
p
DK ˆ DSY a, `† pa, 9†

the width a of the recategorised crack is found by DK ˆ DK from

 F.D. Fischer, F.G. Rammerstorfer / Engineering Fracture Mechanics 66 (2000) 41±49 45
p p p
Y a, `† a ˆ Y a,
 ` 41† a ˆ Y a † a 10†

DS is the nominal stress range and is assumed to be the same for the original and the recategorised case.
The guidelines provide diagrams for a as a function of a and ` of the actual crack, see [1], Sections
19, 20 and Figs. 18, 20±22 therein, and [2], Sections 12, 13 and Figs. 17, 18 therein. In the case of an
internal crack, the recategorised tunnel crack must have one of the two crack contours in the point,
where for the original crack, the maximum of DK appears (which is point A2 in the study below, see
also Fig. 1b). The advantage of the recategorisation lies in the easy handling of such types of cracks in
fatigue analyses, since (more or less) simple analytical formulae exist expressing DK in the positions A
Å 1,
A2 and A,Å respectively, in dependence of the wall-thickness T and the y-distance of A Å 1 and A2 and A,
Å
respectively. Diagrams are provided, e.g. in [1], in dependence on quality categories and in [2], Figs.
23a±h therein, to perform the fatigue analysis without any signi®cant analytical work. However, the
main question remains: How conservative is such a procedure?

3. Conservatism study

In the following context, the results of an extended computer study are reported. Nine di€erent crack
shapes and three possible positions of the cracks are studied. The cracks themselves are characterised
with respect to their length as SHORT, INTERMEDIATE, LONG and with respect to their size (area)
as SMALL, MIDDLE, LARGE. With respect to their positions, the following types are distinguished:
SURFACE, SUBSURFACE (with the crack centre at y ˆ T=4† and CENTRAL (with crack centre at
y ˆ T=2).
It is assumed that in the initial stage, types SHORT, INTERMEDIATE and LONG possess the same
area, independent of their position. The dimensions are chosen as below:

SURFACE SUBSURFACE, CENTRAL

SHORT a0:l0 = 1:2 a0:l0 = 1:4
INTERMEDIATE a0:l0 = 1:4 a0:l0 = 1:8
LONG a0:l0 = 1:6 a0:l0 = 1:12

With respect to the size, the relative crack areas are chosen in relation to the area A of the smallest
crack and tabulated below:

SMALL MIDDLE LARGE

SHORT A 2.25A 6.25A
INTERMEDIATE 2A 4.5A 12.50A
LONG 3A 6.75A 18.75A

Since concrete data must be used, the thickness of the wall is assumed as T = 20 mm and the area A ˆ
2p mm2.
With respect to the fatigue life, two limits for the SURFACE and CENTRAL cracks are checked,
N50 and N90. N50 stands for the number of cycles if the crack extends over 50% of the wall thickness
along the y-line; in the case of N90, 90% are cracked. The highest values of KI are approximately
1600 N mmÿ3/2, which is for a typical steel reasonably below KIC which is of the magnitude of
2500 Nmmÿ3/2.
In the case of a SUBSURFACE crack, a break-through to the surface is assumed for a= a ‡ 2 p† ˆ
0:9: That means, a semi-elliptical crack has formed with a depth of 9.5 mm extending over 47.5% of the
46 F.D. Fischer, F.G. Rammerstorfer / Engineering Fracture Mechanics 66 (2000) 41±49

wall thickness. Therefore, the indication N47.5 is introduced. The study is now continued with a semi-
elliptical SURFACE crack which may grow up to 50 or 90%, respectively, of the wall thickness. This
case is again denominated as N90, N 50 , N 90 and N 47:5 , N 50 , N 90 , respectively, corresponding to the
number of cycles for the recategorised cracks. Since the diagrams become more and more inaccurate for
the crack approaching the surface, a footnote is attached to those values of N which are in¯uenced by
this inaccuracy.
Two types of stress states are selected, a TENSION state with s1 ˆ s2 ˆ sT and TENSION +
BENDING state with s1 ‡ s2 †=2 ˆ sT and s2 ÿ s1 ˆ 0:1818sT , sT is assumed to be 330 MPa.
First, a cycle-by-cycle crack growth analysis was performed on the basis of elliptical or semi-elliptical
cracks, i.e. the ``original'' procedure.
The actual dimensions of the crack, which change per load cycle, are established according to Eqs.
(4)±(6). Very often, millions of changes were calculated during the simulations.
The KI-values are calculated according to Section 9.2 of [1] or 7.2 of [2] as
 
Mm sm ‡ Mb sb p
KI ˆ pa, 11†
f

where
s2 ÿ s1
sm ˆ sT ˆ 330 MPa, and sb ˆ ˆ 0:0909sT ˆ 30 MPa;
2
Mm and Mb are data from existing diagrams depending on the crack shape and the position A1, A2 or
A, respectively, and B. f represents the elliptic integral and exists both in closed form and in the form
of a diagram.
It should be mentioned that a collection of stress intensity factors, i.e. explicit formulae for KI in A1
and A2, for easy programming was recently published by the GKSS group, see [22].
The following concrete data have been chosen:

C ˆ 10ÿ15 in N mm system† and m ˆ 3:

Furthermore, the ``simpli®ed'' procedure is performed for recategorised cracks with crack width a and
 respectively, as discussed in the last section. The corresponding number of cycles for the
2a,
recategorised crack growing to 50 and 90% of the wall thickness are denominated as N 50 and N 90 :
The simpli®ed procedure can be performed by using only diagrams provided in Sections 12, 13 of [2].
The change of the KI-values due to growing of a is taken into account, of course, in these diagrams.
The interested reader is also referred to a paper by BruÈckner-Foit et al. [23], where a procedure

Table 1
Conservatism ratios for 50 and 90% cracks of various size and length for the crack position: SURFACE and loading: TENSION

Size Length

SHORT INTERMEDIATE LONG

N50 =N 50 SMALL 2.54 3.29 3.41

MIDDLE 2.47 4.25 5.20
LARGE 2.73 19.86 8.78
N90 =N 90 SMALL 2.72 3.54 3.69
MIDDLE 2.74 4.52 5.18
LARGE 3.26 9.24 13.37
 F.D. Fischer, F.G. Rammerstorfer / Engineering Fracture Mechanics 66 (2000) 41±49 47

Table 2
Conservatism ratios for 50 and 90% cracks of various size and length for the crack position: SURFACE and loading: TENSION
+ BENDING

Size Length

SHORT INTERMEDIATE LONG

N50 =N 50 SMALL 2.79 3.69 3.87

MIDDLE 2.76 4.88 6.04
LARGE 3.13 23.44 10.42
N90 =N 90 SMALL 3.05 4.06 4.30
MIDDLE 3.13 5.32 6.20
LARGE 3.82 11.24 16.58

Table 3
Conservatism ratios for 50 and 90% cracks of various size and length for the crack position: CENTRAL and loading: TENSION

Size Length

SHORT INTERMEDIATE LONG

N50 =N 50 SMALL 3.14 2.75 1.93

MIDDLE 3.30 2.40 1.62
LARGE 5.95 2.12 1.76
N90 =N 90 SMALL 3.24 2.96 2.54
MIDDLE 3.41 3.26 2.72
LARGE 4.77 4.47 2.87

Table 4
Conservatism ratios for 50 and 90% cracks of various size and length for the crack position: CENTRAL and loading: TENSION
+ BENDING

Size Length

SHORT INTERMEDIATE LONG

N50 =N 50 SMALL 4.33 3.82 2.68

MIDDLE 4.59 3.34 2.24
LARGE 8.27 2.94 2.45
N90 =N 90 SMALL 4.54 4.19 3.60
MIDDLE 4.82 4.64 3.88
LARGE 6.79 6.38 4.08

similar to the ``original'' procedure is outlined in detail, e.g. with analytical expressions for the stress
intensity factors in A1, A2.
The ratios N50 =N 50 and N90 =N 90 indicating conservatism (>1) or non-conservatism (<1) are
evaluated and reported in Tables 1±6.
48 F.D. Fischer, F.G. Rammerstorfer / Engineering Fracture Mechanics 66 (2000) 41±49

Table 5
Conservatism ratios for 50 and 90% cracks of various size and length for the crack position: SUBSURFACE and loading: TEN-
SION

Size Length

SHORT INTERMEDIATE LONG

N47:5 =N 47:5 SMALL 3.82 3.73 2.29a

MIDDLE 5.34 3.55 2.21a
LARGE 24.15a 4.00a 2.88a
N90 =N 90 SMALL 4.01 4.02 2.66a
MIDDLE 5.52 4.06 2.85a
LARGE 11.97a 5.45a 5.44a

Table 6
Conservatism ratios for 50 and 90% cracks of various size and length for the crack position: SUBSURFACE and loading: TEN-
SION + BENDING

Size Length

SHORT INTERMEDIATE LONG

N47:5 =N 47:5 SMALL 4.72 4.64 2.85a

MIDDLE 6.57 4.39 2.75a
LARGE 28.63a 4.81a 3.50a
N90 =N 90 SMALL 4.69 5.00 3.33a
MIDDLE 6.77 5.06 3.59a
LARGE 13.42a 6.74a 7.22a

a
Achieved from extrapolated data.

It is important to note that the results N=N are independent of the actual value of sT (330 MPa) if
KI, max does not reach KIC. Both N and N include a factor sT †m , which cancels out if we consider the
ratio N=N or N=N:


4. Conclusions and recommendations

Tables 1±6 stemming from an extremely extensive computer study show that the ratios N50 =N 50
N50 =N 47:5 for SUBSURFACE cracks) and N90 =N 90 are signi®cantly larger than 1.0. In most cases, the
values for N50 =N 50 N47:5 =N 47:5 for SUBSURFACE cracks) are even larger than those for N90 =N 90 :
This means that the ``simpli®ed'' procedure with initially recategorised cracks generally appears to be
conservative. However, sometimes even the degree of conservatism is so high that an ``original''
procedure considering the actual growing crack should be performed if the ``simpli®ed'' procedure leads
to an insucient life time. The degree of conservatism would even be increased if the ``original''
procedure was performed by an extra calculation of D2 ai , see Eq. (3) and the remark below.
 F.D. Fischer, F.G. Rammerstorfer / Engineering Fracture Mechanics 66 (2000) 41±49 49

Acknowledgements

The authors are indebted to D.I.W. Posch, former M.Sc. student at the Institute of Mechanics,
Leoben, who performed the programming and prepared the results of the computer study.

References

[1] British Standards 1991 PD 6493. Guidance on methods for assessing the acceptability of ¯aws in fusion welded structures. 2nd
ed. Welding Standards Policy Committee, London: British Standards Institution, August 1991.
[2] Empfehlungen zur bruchmechanischen Bewertung von Fehlern in Konstruktionen aus metallischen Werksto€en, 2. Ausgabe,
Juni 1992. Wien: OÈsterreichischer Stahlbauverband.
[3] Feigl G, Fischer FD, Maurer KL, Rammerstorfer FG. Bruchmechanik-Empfehlungen der Arbeitsgruppe ``Bruchmechanik''
des OÈsterreichischen Stahlbauverbandes (OÈSTV). OÈsterrreichische Ingenieur- und Architekten-Zeitschrift (OÈIAZ)
1992;137:257±8.
[4] Fischer FD, Rammerstorfer FG. Evaluation of mode-I crack projection procedures using the strain energy criterion.
Theoretical and Appl Fract Mech 1994;20:67±74.
[5] Lin XB, Smith RA. An improved numerical technique for simulating the growth of planar fatigue cracks. Fatigue Fract
Engng Mater Struct 1997;20:1363±73.
[6] Chermahini RG, Palmberg B, Blom AF. Fatigue crack growth and closure behaviour of semicircular and semielliptical surface
¯aws. Int J Fatigue 1993;15:259±63.
[7] Lin XB, Smith RA. Numerical analysis of fatigue growth of external surface cracks in pressurized cyclinders. Int J Pres Ves
and Piping 1997;71:293±300.
[8] Lin XB, Smith RA. Fatigue growth prediction of internal surface cracks in pressure vessels. J Pressure Vessel Technology
1998;120:17±23.
[9] Lu Y-L. Crack aspect development curves and fatigue life prediction for surface cracks at weld toes in the presence of residual
stress. Int J Fatigue 1995;17:551±7.
[10] Lin XB, Smith RA. Fatigue growth analysis of interacting and coalescing surface defects. Int J Fracture 1997;85:283±99.
[11] Pitt S, Jones R, Atluri SN. Further studies into interacting 3D cracks. Computers and Structures 1999;70:583±97.
[12] Bayley CJ, Bell R. Parametric investigation into the coalescence of coplanar fatigue cracks. Int J Fatigue 1999;21:355±60.
[13] Lin XB, Smith RA. Fatigue shape analysis for corner cracks at fastener holes. Engng Fracture Mech 1998;59:73±87.
[14] Fischer FD, Mayrhofer K. The behaviour of a perpendicular penny-shaped crack situated closely to the free surface of a
halfspace. In: Czoboly E et al., editors. ECF7: failure analysis Ð theory and practice, vol. 1. 1988, pp. 146±54.
[15] Mayrhofer K, Fischer FD. Stress intensity factor variation of a penny-shaped crack situated close to the free surface of a
halfspace. In: Salama K, Ravi-Chandar K, Taplin DMR, Rama Rao P, editors. ICF7, advances in fracture research, vol. 1.
1989, pp. 83±9.
[16] Mayrhofer K, Fischer FD. The stress-intensity factor distribution for an inclined elliptically shaped crack embedded near the
surface of a halfspace: a solution for 3-D fatigue analysis. In: Luxmoore AR, Owen RJ, editors. 22nd Nat. Symp. Fracture
Mech.: Numerical Methods in Fracture Mechanics. Swansea: Pineridge, 1990. p. 657±68.
[17] Mayrhofer K, Fischer FD. A singular integral equation solution for an elliptical crack subjected to a triaxial stress state. In:
Rossmanith HP, Miller KJ, editors. Mixed mode fatigue and fracture. London: Mechanical Engineering Publications, 1993. p.
245±53.
[18] Xu Y, Moran B, Belytschko T. Analysis of three-dimensional edge cracks under tensile loading. Acta Mechanica Solida Sinica
1999;12:174±87.
[19] Noguchi H, Smith RA, Carruthers JJ, Gilchrist MC. Stress intensity factors of embedded elliptical cracks and an assessment
of the ASME X/defect recharacterisation criteria. Int J Pres Ves and Piping 1997;70:69±76.
[20] Fischer FD, Mayrhofer K, Parteder E. Elliptical subsurface cracks under a normal stress and a residual stress ®eld. Fatigue
Fract Engng Mater Struct 1996;19:129±39.
[21] Dai DN, Hills DA, HaÈrkegard G, Pross J. Simulation of the growth of near-surface defects. Engng Fracture Mech
1998;59:415±24.
[22] Schwalbe K-H, Zerbst U, Kim Y-J, Brocks W, Cornec A, Heerens J, Amstutz H. EFAM ETM 97 Ð the ETM method for
assessing the signi®cance of crack-like defects in engineering structures, comprising the versions ETM 97/1 and ETM 97/2.
GKSS, Geesthacht, 1998 [Appendix 1].
[23] BruÈckner-Foit A, Krause A, Munz D. Extension of embedded cracks by fatigue. Mat.-wiss. u. Werksto€tech. 1988;219±23.