Accepted Manuscript

Approximate stress intensity factor solutions for semi-elliptical cracks with

large a/W and c/B under tension and bending

Stefan Kolitsch, Hans-Peter Gänser, Reinhard Pippan

PII: S0167-8442(17)30017-4
DOI: http://dx.doi.org/10.1016/j.tafmec.2017.06.010
Reference: TAFMEC 1892

To appear in: Theoretical and Applied Fracture Mechanics

Received Date: 10 January 2017

Revised Date: 11 June 2017
Accepted Date: 12 June 2017

Please cite this article as: S. Kolitsch, H-P. Gänser, R. Pippan, Approximate stress intensity factor solutions for
semi-elliptical cracks with large a/W and c/B under tension and bending, Theoretical and Applied Fracture
Mechanics (2017), doi: http://dx.doi.org/10.1016/j.tafmec.2017.06.010

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers
we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and
review of the resulting proof before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Approximate stress intensity factor solutions for semi-elliptical cracks with
large a/W and c/B under tension and bending
Stefan Kolitscha,b*, Hans-Peter Gänsera, Reinhard Pippanb
a
Materials Center Leoben Forschung GmbH, Roseggerstraße 17, A-8700 Leoben
b
Erich Schmid Institute of Materials Science, Austrian Academy of Sciences, Jahnstraße 12, A-8700 Leoben

Abstract

Analytical calculations of the stress intensity factor offer a big time benefit compared to finite element simulations. In the present
work, new approximate solutions for the stress intensity factor under tension and bending around two axes are proposed.
Compared to already available solutions (e.g., by Newman and Raju), the proposed formulas extend the range of validity and
comprise cases where the crack width 2·c and the crack depth a are close to the specimen width 2·B and the specimen thickness
W, respectively.

Keywords: stress intensity factor SIF, semi-elliptical crack, geometry factor

1. Introduction

In linear-elastic fracture mechanics, the stress intensity factor is calculated by Irwin’s formula as

a a c 
KI   Y  , ,     a , (1)
c W B

where KI is the stress intensity factor for crack opening Mode I, σ is the stress of the un-cracked specimen and a is the crack
length. Y represents the geometry factor depending on the crack depth a, the crack width 2·c, the specimen thickness W and the
specimen width 2·B. In Fig. 1 the dimensions of a specimen with a semi-elliptical surface crack are illustrated. Additionally, the
applied loads on the specimen are sketched, which are the tensile force Fz, bending around the x-axis with the moment Mx and,
additionally in the present work, bending around the y-axis with the moment My. With these load types all possible external
loadings with respect to fracture mode I occurring in the structural stress concept can be covered using superposition, i.e.,
unsymmetrical bending (Mx and My) and membrane (tensile) loading (Fz).

Fig. 1. Dimensions of the specimen and the semi-elliptical surface crack.

* Corresponding author. Tel.: +43-664-2164744

E-mail address: stefan.kolitsch@gmx.at
 Theoretical and Applied Fracture Mechanics 00 (2016) 000–000

The geometry factor Y depends only on the geometry of the specimen and the geometry of the crack front. If, for given crack
and specimen dimensions a, c, W, B and applied stress  the stress intensity factor KI has been computed by a finite element (FE)
simulation, the geometry factor Y can subsequently be calculated by transforming Eq. 1 into

a a c  KI
Y , ,   . (2)
 c W B     a

Early works [1] represent simple analytical equations of the geometry factor in 2D boundary problems and were extended in
3D geometries with a satisfying accuracy. Newman and Raju (NR) [2-4] published approximate solutions for the geometry factor
of semi-elliptical cracks under tension and bending (Mx), which are nowadays commonly used. For these calculations, a
maximum difference to the FE results of ±5% is mentioned and is valid for ratios 0.2 ≤ a/c ≤ 1.0, c/B < 0.5 and a/W ≤ 1.0.
Additionally, Newman [5] provided an enhanced formulation for higher c/B ratios (c/B ≤ 0.8) under tensile loading with an
accuracy of 5 %.

Based on the work by Newman and Raju, several investigations have been performed. Isida [6] uses the body force method to
provide weight functions for a/W ratios ≤ 0.6 and based on this work, Fett et. al. [7] improves the angular dependency for the near
surface points. Furthermore, Wang and Lambert [8] provide stress intensity functions depending on the stress distribution for
ratios a/W ≤ 0.8 and a comparison with the investigations from Shiratori [9], which can be found in the stress intensity factor
handbook by Murakami [10]. Vainshtock [11] pointed out the different validities and accuracies of commonly used geometry
factor formulations for center, semi-elliptical and corner edge cracks by several FE simulations. While commonly available
results [12] assume a Poisson’s ratio of  = 0.3 typical for metals, Strobl [13] provides an analytical function for different  valid
for a/W and c/B ≤ 0.5.

These works consider different crack shapes and specimen dimensions with straight boundaries (prism). Additionally several
papers provide stress intensity factors for cylindrical specimens like pipes or shafts [14-15] and also for the consideration of the
root stress concentration when a crack starts from a notch [16-17]. For welded joints different solutions depending on the crack
shape and boundary dimensions are available [18-20].

The goal of the present work is to provide geometry functions for the deepest point and the surface points of semi-elliptical
cracks in prismatic beams for large ratios of crack to specimen width (c/B) and crack length to specimen thickness (a/W) for
tension (Fz) and bending around the horizontal axis (Mx). This will be achieved by enhancing the Newman-Raju (NR) formulas
while paying special attention to deriving a formulation which defaults to the NR formulas in their original range of validity. This
will ensure full compatibility to existing analyses using the NR solutions by offering the option to extend them to larger crack
lengths – which is expected to be useful in some applications dealing with the lifetime extension of existing components.

In addition, a geometry factor solution for bending around the vertical axis (My) has been developed.

Referring to Fig. 1, the new empirical solution for the geometry factor is developed for the following ranges of validity:
 ratio of crack semi-axes a vs. c: 0.2 ≤ a/c ≤ 1.0
 ratio of crack length a vs. specimen thickness W: 0.01 ≤ a/W ≤ 0.9
 ratio of crack width 2·c vs. specimen width 2·B: 0.01 ≤ c/B ≤ 0.9.

Nomenclature

a crack depth (minor semi-axis of the semi-elliptical crack) [mm]

B half of the specimen width [mm]
c half of the crack width (major semi-axis of the semi-elliptical crack) [mm]
E Young’s modulus [MPa]
ex distance between outer fiber and neutral x-axis [mm]
ey distance between outer fiber and neutral y-axis [mm]
Fz tension force [N]
Ix axial moment of inertia around the x-axis [mm4]
Iy axial moment of inertia around the y-axis [mm4]
J J-integral [J/m²]
KI stress intensity factor for mode I [MPa√m]
Mx bending moment around x-axis [Nmm]
My bending moment around y-axis [Nmm]
2
 Theoretical and Applied Fracture Mechanics 00 (2016) 000–000

W specimen thickness [mm]

x characterization variable for 0.5 < c/B < 1.0 enhancement [-]
Y(a/c,a/W,c/B) geometry factor, analytical approximation [-]
YA,C-a/W-Fz geometry factor for point A or C and pure tension, enhanced for 0.5 < a/W ≤ 0.9 [-]
YA,C-a/W-Mx geometry factor for point A or C and bending around the x-axis, enhanced for 0.5 < a/W ≤ 0.9 [-]
YA,C-c/B-Fz geometry factor point A or C and pure tension, enhanced for 0.5 < c/B < 1.0 [-]
YA,C-c/B-Mx geometry factor point A or C and bending around the x-axis, enhanced for 0.5 < c/B < 1.0 [-]
YA,C-Fz (a/W = 0.9) geometry factor for point A or C and pure tension for the ratio a/W = 0.9, determined from the FE simulations
[-]
YA,C-Fz (c/B = 0.9) geometry factor for point A or C and pure tension for the ratio c/B = 0.9, determined from the FE simulations
[-]
YA,C-lin-a/W-Fz linear approximation for point A or C and pure tension between NR and a/W = 0.9, enhanced for 0.5 < a/W ≤
0.9 [-]
YA,C-lin-a/W-Mx linear approximation for point A or C and bending around the x-axis between NR and a/W = 0.9, enhanced
for 0.5 < a/W ≤ 0.9 [-]
YA,C-lin-c/B-Fz linear approximation for point A or C and pure tension between NR and c/B = 0.9, enhanced for
0.5 < c/B < 1.0 [-]
YA,C-lin-c/B-Mx linear combination for point A or C and bending around the x-axis between NR and c/B = 0.9, enhanced for
0.5 < c/B < 1.0 [-]
YA,C-Mx (a/W = 0.9) geometry factor for point A or C and bending around the x-axis for the ratio a/W = 0.9, determined from the
FE simulations [-]
YA,C-Mx (c/B = 0.9) geometry factor for point A or C and bending around the x-axis for the ratio c/B = 0.9, determined from the
FE simulations [-]
YA,C-NR-Fz geometry factor by Newman and Raju for pure tension [-]
YA,C-NR-Mx geometry factor by Newman and Raju for bending around x- axis [-]
YFE(a/c,a/W,c/B) geometry factor extracted from FE simulations [-]
z characterization variable for 0.5 < a/W < 0.9 enhancement [-]
ΔYA,C -a/W-Fz alternative geometry factor for point A or C and pure tension, varying around the linear combination between
the boundaries, enhanced for 0.5 < a/W ≤ 0.9 [-]
ΔYA,C -a/W-Mx alternative geometry factor for point A or C and bending around the x-axis, varying around the linear
combination between the boundaries, enhanced for 0.5 < a/W ≤ 0.9 [-]
ΔYA,C-c/B-Fz alternative geometry factor for point A or C and pure tension, varying around the linear combination between
the boundaries, enhanced for 0.5 < c/B < 1.0 [-]
ΔYA,C-c/B-Mx alternative geometry factor for point A or C and bending around the x-axis, varying around the linear
combination between the boundaries, enhanced for 0.5 < c/B < 1.0 [-]
 Poisson’s ratio [-]
σ stress of an un-cracked specimen [MPa]
σbending-x-x stress due to bending around the x-axis [MPa]
σbending-y-y stress due to bending around the y-axis [MPa]
σtension stress due to tension load [MPa]

2. Determination of the geometry factor by finite element simulations

The nominal stress for an un-cracked specimen can be calculated using the conventional equations for a beam with rectangular
cross section (2∙B)×W (see Fig. 2) under tension and bending:

Fz
 tension  (3)
2  B W

Mx Mx W 3 M x
 bendingx x   ey    (4)
Ix 2  B W 3 2 B W 2
12

3
 Theoretical and Applied Fracture Mechanics 00 (2016) 000–000

My My 2 B 3 M y
 bending y y   ex    (5)
Iy W  2  B 
3 2 2 W  B 2
12

The mode I stress intensity factor KI is calculated by using the FE package ABAQUS assuming linear-elastic material behavior
for steel (E = 210000 MPa, = 0.3). For the crack front, the J-integral is computed via the virtual crack extension method [21].
From this information, the stress intensity KI can be calculated, assuming plane strain conditions [22], as

J E
KI 
1  2
(6)

KI is calculated at the deepest point and at the surface points of the crack front as shown in Fig. 2, i.e., A (, C1
(≈and C2 (≈. For accuracy reasons discussed in detail in [23], points C1 and C2 are positioned not directly at the
free surface of the specimen but as close as possible, which means  ≈ 1.40° for point C1 and  ≈ 178.60° for point C2.

Fig. 2. Section view for the dimensions of specimen, crack front and related positions of the deepest and surface points of the crack.

The geometry factor Y depending on the crack shape can then be calculated by using Eqns. 2-6 and the results of the FE
simulations.

The accuracy of the geometry factor depends crucially on the mesh quality. Therefore obtaining a high quality mesh has to be
the main focus of FE model preparation. To calculate the contour integral, the semi-elliptical crack is modelled with a fan-shaped
structured mesh swept along the crack front and discretized with 35 elements in radial direction. The semi-ellipse is subdivided
into 150 elements along the crack front. The crack front is modeled with 8-node hexahedral elements and in the center with 6-
node wedge elements.
In Fig. 3 an example of the mesh for a/c = 0.8, a/W = 0.9 and c/B = 0.9 is shown. The challenge of generating an accurate
mesh for such high ratios a/W and c/B is easily seen in this figure.

a) b)

Fig. 3. Example of the used FE-mesh for a/c = 0.8, a/W = 0.9 and c/B = 0.9. a) front view of the whole specimen and crack front; b) detailed view of the meshed
crack front.

At points A, C1 and C2, 20 contour integrals were calculated for each load. The J-integral was regarded as correct when the
minimum error of 5 successive contours was within a difference of ±0.1 % related to the maximum value.
In Tab. 1 all considered geometry parameters are listed; FE simulations were performed for all 240 parameter combinations.
The focus on high a/W and c/B ratios is evident.
4
 Theoretical and Applied Fracture Mechanics 00 (2016) 000–000

Table 1: Considered geometry parameters

a/c a/W c/B

0.2 0.01 0.01

0.3 0.3 0.5
0.5 0.5 0.6
0.6 0.7 0.7
0.7 0.9 0.8
0.8 0.9
0.9
1.0

3. Results and comparison with the Newman-Raju approximation

The improvement of calculation capacity in the last 30 years and resulting higher mesh quality will improve the accuracy of
FE results. Therefore the validity of the NR formulations is compared with the actual FE results of the geometry combinations
from Table 1. In Fig. 4 and Fig. 5 the results of the FE simulations (horizontal axis) are compared with the Newman-Raju
approximate solution [1,2,4] (vertical axis) for each combination from Table 1, for loading in pure tension and by a bending
moment Mx at points A ( = 90°) and C (in these cases, C1 at  ≈ 0° and C2 at  ≈ 180° give identical results). The full line
represents the exact solution, which means that the analytical approximation is equal to the FE result. The dashed and dotted lines
represent 10 % and 20% over- and under-prediction with respect to the FE solution, respectively.

a) b)

Fig. 4. Comparison of the FE results with the approximate solution by Newman and Raju for pure tension Fz a) all results for point A, color/symbol distinction of
a/W; b) all results for point C (C = C1 = C2), color/symbol distinction of c/B.

In Fig. 4a the comparison of the Y factors for point A is plotted for all parameter combinations. The different marks represent
different a/W ratios, as the a/W ratio will have the highest influence on the geometry factor at point A. It can be seen that the NR
solution fits almost perfectly to the FE results for a/W ≤ 0.5; however, for higher a/W the deviation differs compared to the FE
results.
Fig. 4b shows the comparison for point C. In this figure, the different marks represent different c/B ratios indicating how
closely point C approaches the specimen boundary. It can be observed that the NR approach again fits perfectly inside the
original range of validity, but under-predicts the FE results for c/B ≥ 0.5, i.e. outside the original range of validity.

In Fig. 5, the results for the specimen loaded by a bending moment Mx exhibit good agreement, with the NR formulation
fitting acceptably up to a/W = 0.5 for point A and c/B = 0.5 for point C.

5
 Theoretical and Applied Fracture Mechanics 00 (2016) 000–000

a) b)

Fig. 5. Comparison of the FE results with the approximate solution by Newman and Raju for bending MX a) all results for point A, color/symbol distinction of
a/W; b) all results for point C (C = C1 = C2), color/symbol distinction of c/B.

In general, the results of the FE simulations compared to the NR approach for tension and bending around the x-axis Mx show
an acceptable prediction for all a/c ratios and up to moderate a/W and c/B ratios of 0.5. For higher values of a/W and c/B the NR
approximation tends to over-predicting YA and under-predicting YC. For this reason the development of an approximate solution
addressing especially higher a/W and c/B ratios is required.
For a bending moment Mx, the stress intensity factor at point A becomes negative for a/W ~ 0.8 (which was excluded in the
original Newman-Raju approximation), slightly depending on the a/c and c/B ratios, due to the bending stress distribution (Fig.
6a).

For a bending moment My, the existing solution for a quarter-elliptical crack [10,22] is not usable due to the anti-symmetric
stress distribution. For this bending load, due to symmetry, YA is zero as point A marks the neutral axis (shown in Fig. 6b).
Another result due to symmetry is YC1 = – YC2, where YC1 is positive regarding the convention of a positive moment My around
the y-axis (Fig. 6b).

a) b)

Fig. 6. a) Compression at point A due to stress distribution for a bending moment Mx and high a/W ratio; b) tension resp. compression; depending on the bending
stress distribution for YC at point C1 resp. C2 due to the stress distribution for a bending moment My.

6
 Theoretical and Applied Fracture Mechanics 00 (2016) 000–000

4. Development of a new prediction for the geometry function Y

Based on the good agreement between the NR equations and the FE simulations for the ratios a/W ≤ 0.5, c/B ≤ 0.5 and
a/c ≤ 1.0, new equations have only to be developed for higher ratios in the deepest point A and the near surface points C1 and C2.
This is illustrated in Fig. 7a, where the grey area represents the range of validity of the NR approach; the blue area and green
areas display the regions for which enhanced formulations for the geometry function are developed in the following sections.
To this purpose, two auxiliary variables x and z are introduced for simplicity, each ranging from -1 to 1 for the relevant c/B
and a/W ranges, cf. Fig. 7b.

a) b)

Fig. 7. Area of validity and enhancement of the geometry function depending on the a/c, a/W and c/B ratios and introduction of the auxiliary variables x and z
characterizing the boundaries.

In Fig. 7b the geometry function Y depends on the ratios a/W and c/B. The extension of c/B (blue area) follows the
characterization variable x which is -1 for c/B = 0.5 at the boundary of the NR equation and 1 for c/B = 0.9 (the highest c/B value
for which FE results are available).
To consider high a/W ratios (green area), the characterization variable z is introduced, equaling -1 at a/W = 0.5 and 1 at a/W =
0.9.

To generate a new prediction for the geometry function Y for different loadings, depending on all geometry ratios shown in
Table 1, a statistical analysis with a linear regression model [24] and the method of least squares is conducted. The results of the
empirical equations for different loadings are specified below (Sections 4.1- 4.3, Eqns 7 - 35).

4.1. Tension Fz

4.1.1. Enhancement for high c/B ratios (0.5 ≤ c/B ≤ 0.9)

The geometry function is decomposed into a linear interpolation YA,C-lin-c/B-Fz and a higher order term YA,C-c/B-Fz for
0.5 < c/B < 0.9,

YA,Cc/BFz  YA,C-lin-c/BFz  YA,C-c/B-Fz . (7)

For the linear interpolation for 0.5 ≤ c/B ≤ 0.9, the auxiliary variable x (Eq. 9) is used:

1 x  c  1 x 
YA,Clinc/BFz  YA,C-NR -Fz     YA,C-Fz   0.9     (8)
 2   B   2 

c 7
x  4  (9)
B 2

7
 Theoretical and Applied Fracture Mechanics 00 (2016) 000–000

Here, the solutions at the boundaries are given by the NR solution YA,C-NR-Fz at x = -1(i.e., c/B = 0.5) and by the following
approximation functions determined by linear regression from the FE results at x = -1 (i.e., c/B = 0.9):

2 2 3 2 2
c  a a a a a  a 
YA-Fz   0.9   1.31  1.42  + 0.94    + 4.9     3.73     4.66      
 B  c  
c  
W  
W  c  W 
3 3
(10)
a  a 
+ 3.77      
 c  W 

2 2 2 2
c  a a a a a a  a 
YC-Fz   0.9   0.25 + 1.47   0.72    + 2.21   + 1.04    1.76       (11)
B  c c W  c W  c  W 

The higher order terms YA-c/B-Fz and YC-c/B-Fz for point A (Eq. 12) and C (Eq. 13) are fitted by polynomial functions whose
parameters are determined by means of linear regression from the difference between FE results and linear approximation for
0.5 < c/B < 0.9.

 
2
a
ΔYA  c/B- Fz  x 2  1     0.15 (12)
W 

   
2 2
a a
ΔYC  c/B- Fz  x 2  1     0.12 + x 2  1     0.59 (13)
c W 

4.1.2. Enhancement for high a/W ratios (0.5 ≤ a/W ≤ 0.9)

Similarly to the enhanced formulation for higher c/B ratios, the extension is here done for the range 0.5 < a/W ≤ 0.9. Using the
values from Eqns 7 -13 for a/W = 0.5 and a statistically determined function for YA,C-Fz (a/W = 0.9) extracted from the FE results,
the geometry function can then be calculated as follows:

YA,Ca/WFz  YA,C-lin-a/WFz  YA,C-a/W-Fz (14)

1 z  a  1 z 
YA,Clina/WFz  YA,C-c/B-Fz     YA,C-Fz   0.9     (15)
 2  W   2 

with the variable z

a 7
z  5  (16)
W 2

and the statistically determined functions at a/W = 0.9 for points A (Eq. 17) and C (Eq. 18)

2 3 3 2 2
a  a a a c a  c 
YA  Fz   0.9   2.8  6.35  + 7.28     2.95    + 0.56     0.46       , (17)
 W  c  
c  
c  
B c B

2 2 3 3 4
a  a a c a c c
YCFz   0.9   0.96 + 2.34   4.62    + 2.8    + 2.34     7.7    + 7.23    . (18)
W  c c B c B B

Finally, the higher order terms for the ranges 0.5 < a/W ≤ 0.9, 0 < c/B ≤ 0.9 and 0 < a/c ≤ 1.0 can then be described for points
A (Eq. 19) and C (Eq. 20) as

     
2 3 3
a a c
ΔYA  a/W - Fz  z 2  1     0.18  z 2  1     0.19  z 2  1     0.06 , (19)
 
c  
c B

8
 Theoretical and Applied Fracture Mechanics 00 (2016) 000–000

       
2 3 3
a a c
ΔYC  a/W - Fz  z 2  1  0.19  z 2  1     0.72 + z 2 -1     0.52  z 2  1     0.29 . (20)
 
c  
c B

Regarding pure tension and symmetry, points C1 and C2 have the same applied stress and hence the same value for the
geometry factor.

4.2. Bending Mx

The same approach as above is used for bending around the x-axis for 0.5 < c/B ≤ 0.9 and 0.5 < a/W ≤ 0.9:

4.2.1. Enhancement of c/B ratio (0.5 ≤ c/B ≤ 0.9)

YA,Cc/BMx  YA,C-lin-c/BMx  YA,C-c/B-Mx (21)

1 x  c  1 x 
YA,Clinc/BMx  YA,C-NR -Mx     YA,C-Mx   0.9     (22)
 2  B   2 

2 2 3 2 2
c  a a a a a a  a 
YA-Mx   0.9   1.23  0.81  1.13  + 0.37    + 2.91    2.88     1.79      
 B  c W  
c  
W  
W  c  W 
3 3
(23)
a  a 
+ 1.81     
 c  W 

2 2 3 3
c  a a a a a a
YC-Mx   0.9   0.12 + 2.77   0.28   3.07    + 2.54    + 1.3     1.34    
 B  c W  
c  
W  
c W 
2 2
(24)
a  a 
0.65      
 c  W 

     
2 2 2
a a c
ΔYAc/B-Mx   x 2  1     0.02 + x 2  1     0.1 + x 2  1     0.04 (25)
 
c  
W B

   
2 2
a a
ΔYCc/B-Mx  x 2  1     0.07 + x 2  1     0.33 (26)
c W 

4.2.2. Enhancement of a/W ratio (0.5 ≤ a/W ≤ 0.9)

YA,Ca/WMx  YA,C-lin-a/WMx  YA,C-a/W-Mx (27)

1 z  a  1 z 
YAlina/WMx  YA,C-c/B-Mx     YA,C-Mx   0.9     (28)
 2   W   2 

2 2 2 2 2
a  a c a c a c a  c 
YAMx   0.9   0.11  0.8   0.09  + 0.47    + 0.55     0.89     + 0.49       (29)
W  c B c B c B c B

2 3 3
a  c a c
YCMx   0.9   0.8  0.66     0.13    + 1.69    (30)
 W   
B  
c B

9
 Theoretical and Applied Fracture Mechanics 00 (2016) 000–000

   
2
a
ΔYAa/W-Mx   z 2  1  0.27 + z 2  1     0.13 (31)
c

   
2 3
a c
ΔYCa/W-Mx   z 2  1     0.08 + z 2  1     0.14 (32)
c B

4.3. Bending My

For bending around the y-axis, a new function for the geometry factor has to be developed. Based on the symmetry shown in
Fig. 6b, YA is situated on the neutral axis under the bending load My and therefore is consequently zero, independent of the
geometry ratio:

YA - My  0 . (33)

Assuming a positive bending moment My, point C1 is under tensile and point C2 is under compressive stress, shown in Fig. 6b.
The distance of both points to the neutral axis is equal; therefore the magnitude is also the same, only the sign changes from
positive to negative.

YC2 - My  YC1- My (34)

2 2 3 3 2 2
a c a c a c
YC1-My  0.128    + 1.165     0.126     0.587    + 0.307       
 
c  
B  
c  
B W   B 
2 2 2
(35)
a  a   c 
0.303         
 c  W   B 

The residual standard errors for the new empirical equations are typically less than ±10% for tension and bending around the
x-axis and ±5% for bending around the y-axis. On this basis, the new geometry functions are now compared with all FE results in
order to verify their range of validity.

5. Comparison of results for the new approximate geometry factor solutions

In Fig. 8- Fig. 10, the new approximate solutions for Y are compared to the FE results. The different marks denote different
ratios a/W for YA and c/B for YC the same way as in section 3. The full lines represent again the exact prediction, the dashed lines
10% and 20% over-prediction and the dotted lines 10% resp. 20% under-prediction, respectively.

a) b)

Fig. 8. Comparison of the FE results with the new approximate solution for pure tension Fz a) all results for point A and color/symbol distinction of a/W; b) all
results for point C (C = C1 = C2) and color/symbol distinction of c/B.

10
 Theoretical and Applied Fracture Mechanics 00 (2016) 000–000

Due to the region-wise approximation, the prediction in the original region of validity (0 ≤ c/B ≤ 0.5 and 0 ≤ a/W ≤ 0.5) is
given unchanged by the NR equations. For higher c/B and a/W ratios the prediction is now more precise and lies within an error
band of ±20% for all and ±10% for most cases.

Fig. 8 shows that the new formulas give a very satisfactory estimate of the FE results for a bending moment Mx. Most
importantly, the new prediction fits well for high a/W ratios, where point A is beyond the neutral axis.

a) b)

Fig. 9. Comparison of the FE results with the new approximate solution for bending Mx a) all results for point A and color/symbol distinction of a/W; b) all
results for point C (C = C1 = C2) and color/symbol distinction of c/B.

In Fig. 9 it can be observed that for bending with the moment My, the prediction for point C fits within a minimum error of
±5% to the FE results. As mentioned, YA is constantly zero due to the position of point A on the neutral axis (Fig. 6b).

Fig. 10. Comparison of the FE results with the new approximate solution for bending My for point C (C = C1 = C2) and color/symbol distinction of c/B.

6. Comparison of the new approximate geometry factor solutions with existing formulations

In the Fig. 11 new enhanced equation of the geometry factor for tension is plotted as a function of the a/W ratio
(0.2 ≤ a/W ≤ 0.9), with the different colours denoting different c/B ratios (0.6 ≤ c/B ≤ 0.9). The new equation is compared with
existing solutions of the NR formulation [2-5] represented by dashed lines, the approach from Strobl [13] for c/B = 0.6 (black
dashed line), the function of Isida [6] (black dash-dotted line) and the formulation from Wang [8] (black dotted line).
The different diagrams represent different a/c ratios: a) and b) a/c = 0.2, c) and d) a/c = 0.6, e) and f) a/c = 1.0. Furthermore
the left side a), c) and e) represent the results for the deepest point A and the right side b), d) and f) the near surface points C1 =
C2 = C.

11
 Theoretical and Applied Fracture Mechanics 00 (2016) 000–000

a) b)

c) d)

e) f)

Fig. 11. Comparison of different geometry functions of the new investigation, NR [2-5], Strobl [13], Isida [6] and Wang [8] with the FE results for tension over
the ratio a/W (0.2 ≤ a/W ≤ 0.9) for high c/B ratios (0.6 ≤ c/B ≤ 0.9) and different a/c ratios, in a) for point A and b) for point C for a/c = 0.2, in c) for point A and
d) for point C for a/c = 0.6 and in e) for point A and f) for point C for a/c = 1.0.

The comparison of the geometry factor equations for tension in Fig. 11 represents an acceptable agreement of the new
enhanced formulation with the FE results for point A and perfect fit for point C. The NR equations also provide a satisfying
approximation of the FE results for small ratios, but for higher a/c and c/B ratios the NR formulation highly underestimates the
FE results in point A, also overestimating for high ratios in point C.
The approach from Strobl [13] is valid for a/W and c/B ratios up to 0.5; for higher ratios the functions become unstable and
eventually negative. Isida [6] provided a polynominal function for point A which provides a good estimation for small ratios.
Nevertheless for high ratios the function underestimates the FE results due to the validity of c/B → 0 and neglecting c/B ratio.
Similar to Isida, Wang and Lambert [8] give a polynominal equation for c/B → 0 underestimating the FE results at higher c/B
ratios.

12
 Theoretical and Applied Fracture Mechanics 00 (2016) 000–000

In Fig. 12, the new equation of the geometry factor for bending around the x-axis is again plotted as a function of the a/W
ratio (0.2 ≤ a/W ≤ 0.9), with the different colours denoting different c/B ratios (0.6 ≤ c/B ≤ 0.9). The new formulation is
compared with the NR approach [2-5] represented by dashed lines, the approach from Strobl [13] for c/B = 0.6 (black dashed
line), the function of Isida [6] (black dash-dotted line) and the formulation of Wang [8] (black dotted line).
The different diagrams represent different a/c ratios: a) and b) a/c = 0.2, c) and d) a/c = 0.6, e) and f) a/c = 1.0. Furthermore,
the left side (Figures 12 a, c and e) represent the results for the deepest point A and the right side (Figures 12 b, d and f) the near
surface points C1 = C2 = C.

a) b)

c) d)

e) f)

Fig. 12. Comparison of different geometry functions of the new investigation, NR [2-5], Strobl [13] and Wang [8] with the FE results for bending around the x-
axis over the ratio a/W (0.2 ≤ a/W ≤ 0.9) for high c/B ratios (0.6 ≤ c/B ≤ 0.9) and different a/c ratios, in a) for point A and b) for point C for a/c = 0.2, in c) for
point A and d) for point C for a/c = 0.6 and in e) for point A and f) for point C for a/c = 1.0.

In Fig. 12, a good agreement between the new enhanced equation and the FE results can be observed for points A and C. The
differences of the geometry factor equations are not as big for point A as for point C. Nevertheless, the NR formulation for the
deepest point A, underestimates the FE results for high a/W and c/B ratios. For point C, the prediction of NR highly
underestimates the FE results. In general the approaches from Strobl and Isida provide a satisfying prediction for point A and
13
 Theoretical and Applied Fracture Mechanics 00 (2016) 000–000

high ratios. Nevertheless, for the near surface point C, the approach proposed by Strobl [13] is unstable for higher a/W ratios.
The predictions of Wang and Lambert [8] and Isida [6] cannot be used for varying c/B ratios due to the missing c/B dependency
in their functions.

In Fig. 13, the new equation of the geometry factor for bending around the y-axis is plotted as a function of the a/W ratio (0.01
≤ a/W ≤ 0.9), with the different colours denoting different c/B ratios (0.01 ≤ c/B ≤ 0.9). Existing solutions [10,11,22] for corner
cracks cannot be used due to the different stress distributions.
Here, the different diagrams represent different a/c ratios: a) a/c = 0.2, b) a/c = 0.5, b), a/c = 0.8 and d) a/c = 1.0. As already
mentioned, the deepest point A is on the neutral axis and, therefore, the YA is equal to zero and YC1 = -YC2.

a) b)

c) d)

Fig. 13. Comparison of the new geometry function and the FE results for point C and bending around the y-axis over the ratio a/W (0.01 ≤ a/W ≤ 0.9) for high c/B
ratios (0.01 ≤ c/B ≤ 0.9) and different a/c ratios, in a) for a/c = 0.2, in b) for a/c = 0.5, in c) for a/c = 0.8 and in d) for a/c = 1.0.

The new enhanced equation for bending around the y-axis shows an acceptably accurate prediction of the FE results for all
geometric ratios. The higher the a/c ratio is, the smaller is the gradient with respect to the a/W ratio and the geometry function is
almost constant, independent of the ratio c/B.

Conclusion

A new solution for the geometry factor of a semi-elliptical crack for the deepest point A and the near surface points C1 and C2,
for high a/W and c/B ratios by enhancing the range of validity compared to the original solution of Newman and Raju (NR) is
presented in this work. Additionally, solutions for bending around the x-axis, where the crack tip crosses the neutral bending
axis, as well as for bending around the y-axis are provided in order to describe all possible loadings in Mode I occurring in the
structural stress concept.
In comparison with existing solutions, the new enhanced equations show a good agreement with the FE results, where existing
equations underestimate the FE results or present an unstable behaviour for higher ratios.
The proposed formulas combining the existing solutions from NR and the new enhanced formulas are valid in the range
0.2 ≤ a/c ≤ 1.0, 0.01 ≤ a/W ≤ 0.9, 0.01 ≤ c/B ≤ 0.9 with an error typically less than ±10% for tension and bending around the x-
14
 Theoretical and Applied Fracture Mechanics 00 (2016) 000–000

axis and ±5% for bending around the y-axis. In summary, the new approximate formulas for the geometry factor are expected to
allow an acceptable estimation of the crack tip loading even if the crack depth and width are close to the specimen thickness and
width, respectively.

Acknowledgements

Financial support by the Austrian Federal Government (in particular from Bundesministerium für Verkehr, Innovation und
Technologie and Bundesministerium für Wirtschaft, Familie und Jugend) represented by Österreichische
Forschungsförderungsgesellschaft mbH and the Styrian and the Tyrolean Provincial Government, represented by Steirische
Wirtschaftsförderungsgesellschaft mbH and Standortagentur Tirol, within the framework of the COMET Funding Programme is
gratefully acknowledged.

References

[1] H. Nisitani, Y. Murakami, Stress intensity factors of an elliptical crack or a semi-elliptical crack subject to tension, International Journal of Fracture 10
(1974) 353–368.
[2] I.S. Raju, J.C. Newman Jr., Improved stress-intensity factors for semi-elliptical cracks in finite-thickness plates, NASA Report TM-X-72825 (1977).
[3] J.C. Newman Jr., I.S. Raju, An Empirical Stress-Intensity Factor Equation for Surface Cracks, Engineering Fracture Mechanics 15 (1981) 185-192.
[4] J.C. Newman Jr., I.S. Raju, Analysis of surface cracks in finite plates under tension or bending loads, NASA Report TP-1578 (1979).
[5] J.C. Newman Jr., W.G. Reuter, C.R. Aveline Jr., Stress and fracture analyses of semi-elliptical surface cracks, ASTM Special Technical Publication
1360 (1999) 403-423.
[6] M. Isida, H. Noguchi, T. Yoshida, Tension and bending of finite thickness plates with a semi-elliptical surface crack, International Journal of Fracture
26 (1984) 157-188.
[7] T. Fett, Stress intensity factors for semi-elliptical surface cracks in a plate under tension based on the Isida's solution, International Journal of Fracture
48 (1991) 139-151.
[8] X. Wang, S.B. Lambert, Stress intensity factors and weight functions for high aspect ratio semi-elliptical surface cracks in finite-thickness plates,
Engineering Fracture Mechanics 57/1 (1997) 13-24.
[9] M. Shiratori, T. Niyoshi, K. Tanikawa, Analysis of stress intensity factors surface cracks subjected to arbitrarily distributed surface stresses. Stress
Intensity Factors Handbook Vol. 2, ed. Y. Murakami (1987) 698-705.
[10] Y. Murakami, Stress Intensity Factors Handbook, 3rd edition, Pergamon (1986).
[11] V. A. Vainshtok, I. V. Varfolomeyev, Stress intensity factor equations for part-elliptical cracks and their verification, Engineering Fracture Mechanics
34/1 (1989) 125-136
[12] Y. Murakami, Stress intensity factors and weight functions, Computational Mechanics Publications (1997).
[13] S. Strobl, P. Supancic, T. Lube, R. Danzer, Surface crack in tension or in bending: A reassessment of the Newman and Raju formula in respect to
fracture toughness measurements in brittle materials, Journal of the European Ceramic Society 32 (2012) 1491-1501.
[14] A. Carpinteri, R. Brighenti, S. Vantadori, Circumferentially notched pipe with an external surface crack under complex loading, International Journal
of Mechanical Sciences 45/12 (2003) 1929-1947.
[15] C.S. Shin, C.Q. Cai, Experimental and finite element analyses on stress intensity factors of an elliptical surface crack in a circular shaft under tension
and bending, International Journal of Fracture 129/3 (2004) 239-264.
[16] A. Carpinteri, R. Brighenti, S. Vantadori, Notched shells with surface cracks under complex loading, International Journal of Mechanical Sciences
48/6 (2006) 638-649.
[17] A. Carpinteri, S. Vantadori, Sickle-shaped surface crack in a notched round bar under cyclic tension and bending, Fatigue and Fracture of Engineering
Materials and Structures 32/3 (2009) 223-232.
[18] A. Carpinteri, R. Brighenti, H.-J. Huth, S. Vantadori, Fatigue growth of a surface crack in a welded T-joint, International Journal of Fatigue 27/1
(2005) 59-69.
[19] A. Carpinteri, C. Ronchei, D. Scorza, S. Vantadori, Fracture mechanics based approach to fatigue analysis of welded joints, Engineering Failure
Analysis 49 (2015) 67-78.
[20] S. Vantadori, A. Carpinteri, D. Scorza, Simplified analysis of fracture behaviour of a Francis hydraulic turbine runner blade, Fatigue and Fracture of
Engineering Materials and Structures 36/7 (2013) 679-688.
[21] W. Brocks, I. Schneider, Numerical Aspects of the Path-Dependence of the J-Integral in Incremental Plasticity, Technical Note GKSS/WMS/01/08,
(2001)
[22] TL. Anderson, Fracture Mechanics – Fundamentals and Applications, 3 rd edition, CRC Press (2005)
[23] I. Varfolomeev, M. Busch, M. Petersilge, Characterization of the computational accuracy in surface crack problems, Internatio nal Journal of
Numerical Methods in Engineering 41 (1998) 721-738.
[24] JP. Marques de Sá, Applied Statistics: Using SPSS, STATISTICA, MATLAB and R, Springer (2007).

Appendix A

Below, the equations for the NR solution [1,2,4] with  = 0° for point A and  = 90° for point C are summarized:
a a c 
YA,C-NR -Fz   tension  Q  F  , , ,   (A1)
W c B 

15
 Theoretical and Applied Fracture Mechanics 00 (2016) 000–000

a a c 
YA,C-NR -Mx   bendingx x  H  Q  F  , , ,   (A2)
W c B 

1,65
a
Q  1  1.464    (A3)
c

 a
2
a 
4
F  M 1  M 2     M 3      f  f W  g (A4)
 W   W  

a
M 1  1.13  0.09  (A5)
c

0.89
M 2  0.54  (A6)
a
0.2 
c

24
1.0  a
M 2  0.5   14  1.0   (A7)
0.65 
a  c
c

0.25
 a  2 
f     cos 2   sin 2   (A8)
 c  

 a 
 
2
g  1  0.1  0.35      1  sin 2  (A9)
  W  

H  H1  H 2  H1   sin   p (A10)

a a
p  0.2   0.6  (A11)
c W

a a a
H1  1  0.34   0.11   (A12)
W c W

2
a a
H 2  1  G1   G2    (A13)
W W 

a
G1  1.22  0.12  (A14)
c
0.75 1.5
a a
G2  0.55  1.05     0.47    (A15)
c c

With the factor fw for tension

1
   c a   2

f w  f b  sec    1  0.6  sin    (A16)

  2 B t  

16
 Theoretical and Applied Fracture Mechanics 00 (2016) 000–000

where fb is calculated with

2
a a c
f b  1  0.38      (A17)
c W B

The factor fw for bending around the x-axis can be calculated as follows:

0.5
  c a 
f w  sec  

(A18)
  2  B W 

17
Scientific Highlights

of Manuscript
“Approximate stress intensity factor solutions for semi-elliptical cracks with large a/W and c/B under
tension and bending”.

• approximate stress intensity factor solutions for semi-elliptical cracks with large a/W and c/B
• includes solutions for tension and bending around both major axes
• detailed assessment of the predictive accuracy of the approximate solutions