Vous êtes sur la page 1sur 14

International Journal of Solids and Structures 48 (2011) 19912004

Contents lists available at ScienceDirect

International Journal of Solids and Structures


journal homepage: www.elsevier.com/locate/ijsolstr

Analysis of stress concentrations in plates with rectangular openings


by a combined conformal mapping Finite element approach
A. Louhghalam a,, T. Igusa a, C. Park b, S. Choi b, K. Kim b
a
Department of Civil Engineering, Johns Hopkins University, United States
b
High-Speed Rail Research Center, Korea Railroad Research Institute, Uiwang City 437-757, Republic of Korea

a r t i c l e i n f o a b s t r a c t

Article history: Plates with rectangular openings develop stress concentrations under bending. While these stresses can
Received 11 October 2010 be determined using nite elements, in many problems this would be difcult because a high density
Received in revised form 18 February 2011 mesh would be needed in the neighborhood of every opening corner. In this paper, it is shown how a
Available online 16 March 2011
complex-variable conformal mapping approach can be numerically coupled with the nite element
method to analyze these corner stresses. This approach can be used even with relatively coarse meshes
Keywords: where the nite element results, by themselves, do not resolve the stress concentrations. In essence, the
Stress concentration
method relies on nite element analysis to obtain information on the stress eld in a region surrounding
Plate bending
Rectangular hole
the plate opening; this information is subsequently used to set up the parameters of the conformal map-
Conformal mapping ping approach to obtain the near eld stresses at the opening corners.
Complex variable 2011 Elsevier Ltd. All rights reserved.
Regression

1. Introduction the stress concentration factor around large holes in tensile strips
in terms of the size of the openings. Durelli et al. (1970) experi-
There are a variety of methods for evaluating stress concentra- mentally evaluated the large strains around elliptical holes and
tions in plates due to geometrical irregularities such as openings used photoelasticity to determine the stresses at those locations.
and cracks. In design it is common to use the stress concentration The method of complex variables (Muskhelishvili, 1975) pro-
factor, dened as the ratio of maximum local stress to the nominal vides a powerful approach to solve elasticity problems with geo-
body or far-eld stress. The stress concentration factor can be eval- metric irregularities. The method uses a complex representation
uated by using computational techniques, elasticity theory and for the displacement eld and maps the physical space of the plate
experimental stress analysis such as photoelasticity. Peterson with an irregularity to a unit disk via a conformal transformation.
(1961) has provided stress concentration factors for different geo- The boundary conditions are then expressed in the unit disk space
metric irregularities under various types of loadings. Wu and Mu by contour integrals. Researchers have used this method to solve a
(2003) developed simple methods to obtain stress concentration wide variety of elasticity problems. Savin (1961) was one of the
factors around circular holes located in nite anisotropic plates rst to apply this method to the plate problem. He calculated the
and cylinders due to in-plane loads. stress concentrations around openings of different shapes under
Using edge function analysis, Haani and Dwyer (1999) studied several types of far-eld loads.
the effect of the opening shape and the anisotropy of laminated More recent work on this problem using complex variables
composite plates under plane stress conditions. Henshaw et al. include papers by Ukadgaonker and Rao (2000), Xiwu et al.
(1996) used nite element analysis to study composite laminates (1995), Wu and Cheng (1999), Chen and Hsu (1996), Bryukhanova
with multiple openings under in-plane loads to demonstrate the (1967), Wang and Hasebe (2000), Tsukrov and Novak (2002),
increase in the stress concentration at the corners of an opening Datsyshin and Marchenko (1985), Vigdergauz (1993), Exadaktylos
when another opening is added in its vicinity. Using integral equa- et al. (2003), Cherkaev and Grabovsky (1998). Ukadgaonker and
tions, Hu et al. (1993) studied the interactions between openings Rao (2000) solved the bending problem for laminated composite
and cracks in different domains by decomposing the problem into plates with openings of different shapes and Xiwu et al. (1995)
single hole and crack problems using the superposition principle. considered a nite laminate with elliptical hole under in-plane
Chong and Pinter (1984) used nite element analysis to determine extensional and shear loads. Datsyshin and Marchenko (1985)
solved for the stress concentration around curvilinear cracks in a
Corresponding author. half-plane problem by satisfying the boundary conditions numeri-
E-mail address: alouhgh1@jhu.edu (A. Louhghalam). cally using Gauss quadrature formulas. Wu and Cheng (1999)

0020-7683/$ - see front matter 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ijsolstr.2011.03.005
1992 A. Louhghalam et al. / International Journal of Solids and Structures 48 (2011) 19912004

considered a circular hole in a laminated composite material under surface on the top and bottom faces to convert the problem to a
extensional loading. Chen and Hsu (1996) studied the stress con- two-dimensional formulation to avoid discretization in the plate
centrations around undulating cracks located at the interface of thickness direction. In another study, Richards and Daniels
the two materials. Two sets of analytic functions were used for (1986) used the conventional eight-node nite element to deter-
the two materials and the unevenness of the cracks was modeled mine displacements, and then utilized a complete quadratic func-
using perturbation. Conformal transformation was used to nd tion for redening the displacement eld for subsequent
the optimal opening shape for plates with multiple holes under differentiation. The coefcients of the quadratic function were ob-
far-eld uniform tension in Vigdergauz (1993) and for plates with tained by least-squares t of the displacements and stresses on the
a single opening under shear load in Cherkaev and Grabovsky boundary.
(1998). Wang and Hasebe (2000) studied bending of a plate with Smoothing techniques have also been used in problems where
an inclusion and a crack. The method of complex variables was strain measurements are available. Rowlands et al. (1978) and
used in Exadaktylos and Stavropoulou (2002) to obtain a closed- Feng and Rowlands (1991) combined smoothing with nonlinear
form solution for displacements and stresses around D-shape tun- regression to examine a plate with a hole under in-plane extension
nels located in prestressed soil/rock masses and in Exadaktylos and bending. The strain measurements were obtained using photo-
et al. (2003) to present a semi-analytical stressdisplacement solu- mechanical techniques. The noise presented in the measured
tion for circular holes with notches. Jasiuk (1995) also used this deformation leads to erratic derivatives of displacement and there-
method to nd the stresses around rigid polygonal inclusions to fore noisy stresses and strains. The smoothing techniques provide
determine the effective elastic material properties of the composite an approach to calculate more accurate estimates of these
containing those inclusions. The SchwarzChristoffel transforma- derivatives.
tion was used by Tiwary et al. (2007) to generate stress functions This paper focuses on stress concentrations around openings
for heterogeneities of arbitrary shapes. Ishikawa and Kohno with 90 corners in plates under different types of loads. These
(1993) used this transformation to determine the stresses around stresses are usually determined using nite element calculations
square openings and inclusions for plates under in-plane with high-density meshes in the vicinity of the corners. In this pa-
extension. per a new method is developed that would not require such high-
While the complex variable method is very powerful in solving density meshes. The idea here is to couple the conformal mapping
a wide variety of elasticity problems, it is complicated and must be approach with nite element results for resolving stress concentra-
reformulated for each type of far-eld load. The other main class of tions. As explained below, this method can be used even with
methods for evaluating stress concentration factors in plates are relatively coarse meshes where the nite element results, by them-
based on the nite element method. A special nite element is of- selves, do not resolve the stress concentrations. Fig. 1 is used to
ten introduced for plane stress problems. For instance, Chen (1993) explain this idea more precisely. On the left side of the gure is a
developed an element with a circular hole and Piltner (1985) intro- plate with four square openings and on the right side, one of the
duced one with circular or elliptical holes. These special elements corners is shown in detail. It can be seen that the radius of curva-
are used at the hole locations at the region near the hole boundary ture rc of the corner is small compared to the nite element mesh
to resolve the stress concentrations while regular elements are size smesh. To accurately determine the stress distribution in the
used at other locations. The stresses in these elements are deter- vicinity of the corner, a much smaller mesh size is typically needed.
mined using a numerical implementation of the aforementioned In this paper it is shown that the nite element results obtained by
complex variable method. The stress concentration factors for a coarse meshes with sizes smesh of the order of two or three times
crack are also determined by these special nite elements. To study larger than the radius of curvature rc as indicated in the gure pro-
a composite plate under in-plane extension and shear, Nishioka vide sufcient information to determine the magnitude of the
and Atluri (1982) modied the complimentary energy principle stress concentration. This approach would be practical for complex
and developed a special plane-stress nite element for laminated structures, where a high-density nite element mesh cannot be
composite plates with a hole. Pan et al. (2001) developed a 3D used at the corners of every rectangular opening. The key and
boundary element formulation for the analysis of composite lami- novel aspect of this method is in extending the complex variable
nates with holes. They used a special Greens function which satis- method so that it can be effectively coupled with coarse nite
es the continuity equation between the laminae and the free element results to evaluate these stress concentrations. In essence,

Fig. 1. Position of openings in four-hole plate.


A. Louhghalam et al. / International Journal of Solids and Structures 48 (2011) 19912004 1993

the coarse nite element results are used to provide moment and unity. Then the moments and shear forces shown in Fig. 2 can be
shear information at intermediate regions away from the plate written as (Timoshenko and Woinowsky-Krieger, 1959):
openings while the extended complex variable expressions are ! !
subsequently used to obtain the stress eld in the neighborhood @2w @2w @2w @2w
M x D m 2 ; M y D m 2 ;
of each corner. @x2 @y @y2 @x
The paper begins with a brief review of plate bending and the
@2w 1
conformal transformation method. Then, an extension of the com- M xy D1  m
@x @y
plex variable method is developed, with derivations given in the
appendix. These new derivations are used for satisfying the bound- @ r2 w @ r2 w
Nx D ; Ny D
ary conditions via standard regression rather than contour integra- @x @y
tion. It is then shown how these results can be used in a semi- Eh 3
where D 121 m2 is the exural rigidity. The bending moments and
analytical approach that combines the conformal transformation
shear forces in curvilinear coordinates with axes n and t as shown in
with coarse-mesh nite element results to obtain predictions for
Fig. 2(right) are related to those in Cartesian coordinates through
the stress concentration at opening corners. A series of examples
the following transformations (Savin, 1961):
are examined to illustrate the range of applicability of this semi-
analytical approach. Mx My Mn Mt My  M x 2iMxy e2iun M t  M n 2iMnt
Nx  iNy eiun Nn  iN t 2
2. Analytical results
where un is the angle to the horizontal axis x.
2.1. Plate bending equations The biharmonic bending equation is obtained by writing the
equilibrium equations of moments and forces and eliminating
the shear terms (Timoshenko and Woinowsky-Krieger, 1959).
Consider a plate element in the x, y plane as shown in
Fig. 2(left). Let w(x, y) be the out of plane deection of the midplane @4w @4w @4w
in the vertical direction. In thin-plate bending theory, one can use 4
2 2 2 4 0 3
@x @x @y @y
the Kirchhoff assumption, which states that straight lines normal
to the midsurface before bending remain straight and normal to In the conformal mapping approach to this problem, the x, y coordi-
the midplane; the deection of the midsurface is small compared nates are expressed by the complex variable z = x + iy. A conformal
to the plate thickness and the slope of the surface is much less than transformation z(f) is used to establish a one-to-one mapping

Fig. 2. Internal forces and moments in a plate element.

Fig. 3. Conformal transformation to a unit circle.


1994 A. Louhghalam et al. / International Journal of Solids and Structures 48 (2011) 19912004

between the plate in physical space to a unit disk R with coordinate considered in order to map a square hole with sharper corners to
f shown in Fig. 3. Muskhelishvili (1975) showed that the solution to a unit disk. In the semi-analytical method developed in Section 3,
this biharmonic equation can be expressed as: however, the above transformation is sufcient for numerical anal-
ysis of the stress concentration at opening corners, even for corners
wx; y Rez/p z vp z 4
with relatively small radii of curvature.
where / and v are analytic functions determined by the boundary Using the CauchyRiemann equations for analytic functions, the
conditions. The subscript p indicates that the functions are in terms Jacobian of the mapping can be represented as:
of the variable in physical space z. The far-eld boundary conditions  @u 
are evaluated directly in physical space, but the boundary condi-  @u 
1
 @x @y 
J  @v  0 Aeih
tions at the interior plate opening are more easily evaluated with  @x @v  z f
@y
the help of the aforementioned conformal mapping.
The moments and shear forces in a plate with hole can be where A is the ratio of area of an innitesimal element in the unit
decomposed into two components: disk to its corresponding area in the physical space and h is the an-
M M0 M 5 gle of rotation of the conformal transformation as shown in Fig. 3.
For the conformal mapping in Eq. (7) and shown in Fig. 3 the Jaco-
N N0 N 6 bian is:
where the superscript 0 indicates that the solutions are for a plate 2Rf2
without hole. The rst components M0 and N0 are the bending mo- J
2 f4
ment and shear force in an innite, unweakened plate which can be
evaluated using the basic bending equations. For instance for a plate
To solve the plate bending problem for plates with a hole, the con-
under a constant moment Mx = M at innity shown in Fig. 3, it can
formal transformation is used to write the displacements in Eq. (4)
be observed that M0x M and N 0x N 0y 0. The second components
in terms of the variable f in the unit disk. Therefore, the moments
M and N are associated with the existence of the hole and ac-
and shear forces are all in terms of the analytic functions /(f) and
counts for the redistribution of stress. The analytic functions can
v(f), which are decomposed as:
also be decomposed into two components accordingly:
/p z /f /0 f / f
/p z /0p z /p z
vp z vf v0 f v f
vp z v0p z vp z
In the above equations, /0(f) and v0(f) are analytic functions asso-
For the case where Mx = M at innity, it can be shown using Eq. (1) ciated with the bending of the unweakened plate and can be deter-
for plate bending that: mined by substituting z(f) into /p(z) and vp(z). For the plate with a
Mz Mz2 square hole with rounded corners corresponding to the conformal
/0p z  ; v0p z  transformation in Eq. (7) the results are:
4D1 m 4D1  m
M ! !2
w0 x; y  x2  my2 MR 1 f3 MR2 1 f3
2D1  m2 0
/ f   ; 0
v f  
4D1 m f 6 4D1  m f 6
Since the components of the moments M and shear force V atten-
uate rapidly with increasing distance from the hole, the following
The functions /(f) and v(f) are the analytic functions correspond-
form for the analytic functions / and v are typically used (Savin,
ing to the existence of the hole and are chosen to attenuate rapidly
1961)
outside the hole:
a1 a2
f z a0 2 
z z X
1 X
1
/ f aj fj ; w f bj fj 8
where ajs are constant coefcients. j1 j0
The conformal transformation technique explained in the next
section is used to nd these coefcients and the components of The moments and shear forces normal to the boundary of the hole
the moments and shear forces M and V. must vanish. It has been shown that these boundary conditions are
expressed as two contour integral equations in terms of /0(f), v0(f),
2.2. Review of analytical solution using conformal mapping /(f) and v(f) inside the unit disk (Savin, 1961). When these inte-
gral equations are solved the unknown coefcients of the expan-
2.2.1. Conformal transformation denition and properties sions in (8) are obtained and the midplane deection, moments
The function z(f) that maps f in the unit disk R to z in physical and shear forces in the plate are then determined accordingly. For
space S is called a conformal transformation if it preserves angles example, the analytic functions for the square hole with rounded
(Muskhelishvili, 1975). The Riemman mapping theorem states that corners are (Savin, 1961):
any simply connected region can be mapped to a unit circle by
 
such a conformal mapping. Therefore a plate with a hole can be MR 6 1
/ f  f f3 9
mapped to a unit disk as shown in Fig. 3. Savin (1961) explains 2D 17 7m 63 m
how a conformal mapping can be found that maps a plate with a
hole of an arbitrary shape to a unit disk. For example the conformal 
MR2 1 1
transformation: v f f6 f4
2D 721  m 21  m3 m
! 
1 f3 27  4m 35 13m
zf R  7 f2 ln f 10
f 6 31  m17 7m 121  m3 m

maps the physical space of an innite plate with a square hole with These results are substituted into Eq. (4) and the conformal trans-
rounded corners to a unit disk. More terms in Eq. (7) needs to be formation in Eq. (7) to obtain the midplane deection:
A. Louhghalam et al. / International Journal of Solids and Structures 48 (2011) 19912004 1995


MR2 35 13m 1 1 6 explained in Appendix A. Rearranging the above equations one
w ln q   cos2u can write:
2D 121  m3 m 21 m q2 17 7m
2m 1 Mxx M yy A2 M uu M vv 12
q2 cos4u q4 cos2u
61 m3 m 17 7m
1 1 cos2u Myy  M xx 2iMxy A2 e2ih M vv  Muu 2iMuv
 q6   
723 m 21  m q2 @w @w @Aeih
  2D1  m  i 13
35  m 1 @v
q2 cos2u q4 cos4u @u @x
21  m17 7m 21  m3 m
11 @Aeih 2
Nx  iNy A3 eih Nu  iNv  2DAeih ru;v w 14
@x
where f = qeiu.
It is noted that this conformal mapping technique has some In the above equations it is noted that the terms Muu, Mvv, Muv, Nu
limitations in practical applications. It is fairly complicated, requir- and Nv have the same mathematical form as moments and shear
ing the solution of a contour integral equation. In addition it is lim- forces but do not physically represent these quantities. To get the
ited to relatively simple expressions for the far-eld moment. In moments and shear forces normal to the boundary, one needs to
more general congurations such as a plate with four interacting transform the above relations to curvilinear coordinates using Eq.
holes as shown in Fig. 1, the conformal mapping technique be- (2) as follows:
comes analytically intractable. Finally, in practice it is only neces-
sary to examine the stress concentrations, which occur at the Mtt M nn A2 M uu M qq
corners, but the purely analytical approach requires an analysis
of the entire near eld surrounding all openings. The proposed M tt  M nn 2iMnt e2iun A2 e2ih M uu  M qq 2iMqu e2iu
semi-analytical method presented in the next sections was devel-  
@w @w @Aeih
oped specically to address these practical limitations.  2D1  meiu  i
@q q@ u @x
2.3. Moment and shear relations
@Aeih 2
Nn  iNt eiun A3 eih Nq  iNu eiu  2DAeih rq;u w
The rst contribution of this paper is in deriving closed-form @x
analytical expressions for the moments and shear forces in the z where Mn, Mnt and Nn are the physical bending moment, twisting
plane in physical space in terms of /, v and the variable f inside moment and shear force per unit length along the contour with
the unit disk. These expressions are needed to develop the pro- the outward normal n and Mtt and Nt are the bending moment
posed semi-analytical method, as shown in Section 3. and shearing force per unit length along the tangential direction.
The derivation begins with the relations between the deriva- In the above equation, un = h + u is the angle between the horizon-
tives in physical space and in the unit disk as summarized in tal axis and the normal to the boundary and the terms Mqq, Mqu,
Appendix A. When these derivatives are substituted into the plate Muu and Nq again have the same mathematical form as the mo-
bending Eq. (1), the moment and shear forces in the physical space ments and shear forces in polar coordinates q, u in the unit disk
can be written in terms of the rectangular coordinates u and v in and are therefore named, accordingly.
the unit disk: The moments and shear forces in curvilinear coordinates in
" physical space can be written as:
@2 @2 @2   00 
M xx  D a2 mb
2 2
b ma2 2  2ab1  m z f iu
@u 2 @v @u@ v M nn A2 M qq  D1  mRe q 0 e
z f
       00 
@a @b @ @a @b @ 2 z f iu
m m  w M tt A M uu D1  mRe q 0 e
@x @y @u @y @x @ v z f
  00 
2 z f iu
" M nt A M qq D1  mIm q 0 e
@2 2 2 @2 @2 z f
M yy  D ma2 b mb a2 2 2ab1  m   00 
@u 2 @v @u@ v z f
     Nn A3 Nq 2Dr2q;u wRe 0 eiu
m@a @b @ @a @b @ z f
m w
@x @y @u @y @x @ v
where
@w @w
" q i
2 2 2 @q q@ u
@ @ 2 @
M xy  D1  m ab  ab 2 a2  b
@u2 @v @u@ v
 The effective transverse force per unit length along the contour with
@a @ @b @
 w normal n is:
@y @u @y @ v
      @M nt
@a @b @ @ V n Nn
Nx D 2 a b r2u;v w a2 b2 a  b r2u;v w @s
@x @x @u @v
(  00 
      3 @M qu 1 @A2 z f
@a @b @ @ V n A Nq 2 M qu 2Dr2q;u wRe 0 eiu
Ny D 2 a b r2u;v w a2 b2 b a r2u;v w q@ u A q@ u z f
@y @y @u @v "
2
!!#)
00 000 0 002
z @q 1 @A i iu z z  z
D1  mIm eiu 0 q 2 ie
where a and b are partial derivatives of u and v in the unit z q@ u A q@ u q z0 z00
disk with respect to physical space coordinates x and y as
1996 A. Louhghalam et al. / International Journal of Solids and Structures 48 (2011) 19912004

The moments and shear forces above are then written in terms of The following expansions are then chosen for the analytic func-
the conformal transformation and their derivatives in Appendix B. tions / and v:
Previous researchers in conformal mapping techniques for plate X
m
problems did not consider the derivation of these equations be- / bj f2j1 17
cause they are not needed in evaluating the contour integrals asso- j1
ciated with boundary conditions. In the semi-analytical approach m0
X
developed in the present paper an alternate approach to solving v b00 ln f 0
bk f2k 18
the boundary value problem is used and the new analytical results k1

in this section are essential in this formulation. These equations are compatible with the analytic functions /0 and
There are two basic parts to the semi-analytical method. The v0 since they can cancel the moments and shear forces of unweak-
rst part is in developing series expansions for the analytic func- ened plate at the near-eld.
tions / and v for various types of far-eld moments and shears. To formulate the linear regression problem, the analytic func-
The analytical results of the preceding section are needed to deter- tions can be written in the following matrix form:
mine these expansions. The second part is in applying these expan-
sions to specic plate problems involving one or more rectangular /f /0 U 1 b
19
openings. Here, results from a nite element analysis are coupled vf v0 U 2 b
with the series expansions to evaluate the stress concentrations
at the interior corners. The nite element analysis is used only to where b is a vector of unknown coefcients and U1 and U2 are the
obtain stresses at regions away from the corners, and hence it does covariate matrices corresponding to / and v. For the square hole
not need to have a ne mesh to resolve the relatively high stresses with rounded corners example, the analytical solution for bj can be
in the neighborhood of the corners. This is because the series expressed as the following vector:
expansions are used to evaluate the stress concentration. The MR h 6 1 3513m 274m 1 1
iT
method is semi-analytical in that it is explicitly in terms of the b R 121m3m R 31m177m R 21m3m R 721m
2D 177m 63m

functions / and v, but with coefcients that are partly derived


from the general analytical results in Appendix B and partly de- where
rived from numerical results from nite element analysis of the    
U 1 f1 f3 0 0 0 0 ; U 2 0 0 ln f f2 f4 f6
specic plate of interest.
are single-row covariate matrices evaluated at an arbitrary point in
3. Semi-analytical method the unit disk. It is noted that the above results for b are valid only
for a plate with a square hole with rounded corners under constant
3.1. Regression approach for evaluating / and v far-eld moment M (Savin, 1961). They were obtained through a
contour integration approach. Below it is shown how these coef-
As stated above, the rst step in the semi-analytical method is cients can be obtained without contour integration for more general
in developing the series expansions for / and v. The standard ap- hole geometries and loads through the use of the analytical results
proach for this is through evaluation of the boundary conditions in Section 2.3. The appropriate covariates, as explained earlier in
using contour integrals (Muskhelishvili, 1975). Herein, a novel ap- this section and discussed in more detail in the following, are placed
proach based on regression is developed instead. Specically, the in matrices U1 and U2 corresponding to / and v, respectively.
form and coefcients of the analytic functions / and w in Eq. These matrices are arranged by inserting submatrices of zeros such
(8) are determined by regression using the results of the previous that only a single vector of unknown coefcients is needed for rep-
section. The form of the analytic functions /0 and v0 for an resenting / and v. For the case of constant moments at far-eld
unweakened plate provides guidelines on the form of the expan- these matrices can be written as:
2 3
sions for / and v. This means the covariates (essentially basis f11 f31    f12m1 0  0
functions) used in the regression analysis of / and v should in- 6 . .. .. .. .. .. 7
clude functions that are compatible with the form of /0 and v0. U1 6
4 .. . . . .
7
. 5;
In other words, using those covariates, one must be able to cancel f1n f3n    f12m1 0  0
the moments and shear forces of the unweakened plate at the 2 0 3
0    0 ln f1 f21    f2m
near-eld due to the functions /0 and v0. To choose these covari- 6.
1

ates properly we need to investigate the form of the Eq. (4) for 6
U 2 4 .. .. .. .. .. .. 7 7
. . . . . 5
w. A close look at this equation reveals that the power of v in z 0
0    0 ln fn f2n    f2m
m
should be one plus the power of / in z. On the other hand, one no-
tices that the conformal transformations for square holes with The rows of the above matrices are the covariates evaluated at dif-
rounded corners in Eq. (7) consists of odd powers of f. These facts ferent points on the boundary.
imply that if / can be written in terms of odd powers of f then v To determine the unknown coefcients b of the analytic func-
can only include the even powers and vice versa. To clarify this tions, the moments and shear forces at the boundary of the hole
even further let us consider the case of a plate under constant mo- where q = 1 are evaluated in the direction normal to the boundary
ment at far-eld. The analytic functions /0 and v0 for this case using Eqs. (25) and (29). These moments and shear forces are then
have the following form: set to zero using regression to determine b. One can arrange the
equations for the bending moment and shear forces normal to
X
n
cj X
m
/0 f
0
bj f2j1 15 the boundary as follows:
2j1
j1 f j1
M nn M 0nn Mn M 0nn U Mnn b
n0
X m0
X
c0k
v0 f
00
bk f2k 16 V nn V 0nn V n V 0n U V n b
k1 f2k k1
where M0nn and V 0n are the moment and shear components associ-
where for the example of an innite plate with a square hole ated with the unweakened plate and can be evaluated by substitut-
with rounded corners we have n = 1, m = 2, n0 = 1 and m0 = 3. ing the analytic functions /0 and v0 in Eqs. (25) and (29). In the
A. Louhghalam et al. / International Journal of Solids and Structures 48 (2011) 19912004 1997

above equations, U Mnn and U V n are matrices of covariates corre- with the analytic functions /0 and v0 corresponding to the unweak-
sponding to the bending moment and shear forces normal to the ened plate. For constant moment at innity, six unknown coef-
boundary due to the existence of the hole. The matrices of moment cients are used to determine the analytic functions / and v. For
covariates are constructed by arranging the different derivatives of the case of linearly varying moment at far-eld, the powers of z
the components of U1 and U2 according to Eqs. (25) and (29). Then, in /0 and v0 are two and three. Substituting the conformal transfor-
following standard regression procedure the following least- mation for a square hole with rounded corners in Eq. (7), /0 consists
squares problem is solved to determine the unknown vector b so of even powers of f to the maximum of six and the terms in v0 have
that the moments and shear forces normal to the boundary will odd powers of f to f9. Therefore the following covariates are used in
vanish. the case of linear far-eld moments:
 
minM 0nn U Mnn bT M 0nn U Mnn b V 0n U V n bT V 0n U V n b U 1 f2 f4 f4 0 0 0 0 0 0 ;
b  
U 2 0 0 0 ln f f f3 f5 f7 f9
This can be rewritten as:
Hence, a total of nine unknowns appear in vector b in Eq. (19).
minU n b F 0 T U n b F 0 20 In the nal step, linear regression is used to obtain the unknown
b
coefcients b by enforcing the boundary conditions. The coef-
where: cients of the analytic functions for the case of constant moment
  " #
U Mnn M0nn Mx = 1 is determined using the formulations derived in Section
Un ; F0 2.3 and the procedure explained in Section 3.1. The results for,
UV n V 0n D = 1,R = 1 and m = 0.3 are shown in the rst column of Table 2.
The unknown coefcients b are readily determined by the standard The coefcients obtained here are equal to the exact analytical
regression equation: coefcients in Eq. (9) up to sixteen signicant digits. The unknown

1 coefcients, evaluated using linear regression are shown in Table 2.
b U Tn U n U Tn F 0 In the cases of non-constant moments, it is cumbersome to obtain
the exact analytical results using contour integration, and this
In the following, a small library of series expansions for the analytic analysis was not carried out herein. Instead, the proposed semi-
functions is developed for the basic cases of constant and linearly analytical approach was veried by examining the magnitudes of
varying moments. The rst step in formulating the linear regression the moments and shear forces normal to the boundary. According
problem is to nd the analytic functions /0 and v0 associated with to Eq. (20) these non-zero moments and shear forces are the resid-
the unweakened innite plate under the specied loads. Using basic ual errors in the regression analysis. For the case of a constant unit
plate bending equations in (1), the analytic functions, displacement moment Mx = 1, the maximum absolute value of the moment at the
eld, and the moments and shear forces distributions can be boundary is 1.5e14 which is of the order of machine round-off er-
determined. A set of results for constant and linearly varying ror, indicating the high accuracy of regression analysis in nding
moments is summarized in Table 1. Using the results in this table the coefcients of analytic functions. Similar small residual errors
and substituting the conformal transformation in Eq. (7) into the were found for all of the other moment loads shown in Table 2.
functions /0 and v0, the powers of f in /0 and v0 are evaluated. Con- It is noted that these moment residuals at the boundary are errors
sequently the regression covariates are chosen to be compatible associated with the rst step of the semi-analytical approach in
which the regression analysis rather than contour integral equa-
tions is used for satisfying the boundary conditions.
Table 1
Relations between the moments and shear forces, the displacement eld w(x, y) and
the analytic functions /p(z) and vp(z) for a plate without hole. 3.2. Coupling with coarse nite element results
Mx My Mxy Nx Ny w(x, y) /p(z) vp(z)
m1 z
The second part of the semi-analytical method is in coupling the
m1 0 0 0 0 m1 2 2 m1 z 2
m2 x  my
 2D1  4D1 m 4D1m series expansions for the analytic functions with intermediate
m2 2 2 m2 z
0 m2 0 0 0 m2 z 2
m2 x  my
 2D1  4D1 m stress results obtained from nite element analysis. As noted ear-
4D1m
m3 xy
0 0 m3 0 0  D1 m 0 im3 z2
2D1m
lier, the nite element analysis can be relatively coarse, and does
m4x 0  m1
m4 y m4
0 m4
 2D1 x3
 mxy 2 m4 z2 4 13mz
 m24D1
3 not need to be accurate near the interior corners. In design appli-
m 1m m2 3  8D1 m m2
0 m5x m5 y m5
0 m5 2 x3 m5 z2 m5 3m z3
cations, it is useful to know the relationship between the radius
1m 1m  2D1 m2 xy  m 3  8D1m 24D1m2
m6 x m6
of curvature and the maximum stress at the corners. The proposed
m6y 0 0 m6
 2D1 2
m y3 im6 z2 im6 3mz3
1m 1m m2 yx 3 8D1m 24D1m2 approach can be used to estimate this relationship for a range of
0 m7y  m1
m7 x
0 m7 m7 y3
2 im7 z2
 im7 13mz
3
radii using the results of a single coarse-mesh nite element anal-
m 1m  2D1 m2 3  mx y 8D1m 24D1m2
ysis evaluated for a single radius of curvature. The basic idea is to

Table 2
Coefcients bj of the analytic functions for various far-eld moments.

Far-eld Far-eld moment


moment
Mx = 1 My = 1 Mxy = 1 Mx = x Mx = y My = x My = y
b1 0.157 0.157 0.293 0.0874 0.0641 0.0641i 0.0874i
b2 0.0253 0.0253 0 0.00680 0.00680 0.00680i 0.00680i
b3 0.702 0.702 0 0.00267 0.00267 0.00267i 0.00267i
b4 0.145 0.145 0.0395 0 0 0 0
b5 0.108 0.108 0 0.545 0.307 0.307i 0.545i
b6 0.00992 0.00992 0.0198 0.0747 0.0347 0.0347i 0.0747i
b7    0.0471 0.0273 0.0273i 0.0471i
b8    0 0 0 0
b9    0.000403 0.000700 0.000700i 0.000403i
1998 A. Louhghalam et al. / International Journal of Solids and Structures 48 (2011) 19912004

use the nite element results to nd the stresses SFEM in the inter- relations in Eq. (22). In short the radius of curvature of the corner
mediate eld away from the vicinity of the corner where the coarse dictates the locations of regression points used to nd the far-eld
mesh can provide results with good accuracy. Next, regression is moments. As noted in the example studies below, the regression re-
used to obtain the far-eld moments Mfar-eld corresponding to sults are not sensitive to their locations.
these intermediate-eld stresses. Once these far-eld moments The unknown components of Mfar-eld are then evaluated by
are determined, the stress concentration at the corner can be com- inverting Eq. (21) using standard regression:
puted using Table 1. The details of these calculations are shown
Mfar-field U Tm U m 1 U Tm SFEM 23
below.
The relationship between the nite element stresses and the Once the far-eld moments are determined, the stresses in the
far-eld moments can be written as: vicinity of the corner can be evaluated by an expression similar to
Eq. (21):
SFEM U m M far-field 21
Sx; y Ux; yM far-field 24
where SFEM is the vector of stresses evaluated at points (xk,yk) for
k = 1, 2, . . . , n using a coarse-mesh nite element analysis. The ma- The difference between these two equations is that the stresses SFEM
trix Um is the matrix of covariates evaluated at (xk, yk) using the in Eq. (21) are obtained from the coarse nite element mesh at
expressions in Eqs. (25)(29) and points (xk, yk) in the intermediate eld as shown in Fig. 4(b) so that
the far-eld moment Mfar-eld can be computed by regression. The
M far-field m1 m2    m7 0 predicted stresses S(x, y) are determined at arbitrary points (x, y)
is the vector of unknown coefcients for the far-eld moment. in the vicinity of the corner by the matrix multiplication in Eq. (24).
The value of R in the conformal transformation in Eq. (7) is
6.32rc. For the case where the corner of interest is at the upper 3.2.1. Illustrative examples
right of the rectangular opening, the points (xk, yk) for obtaining To clarify the applicability of this method it is useful to compare
the far-eld stresses are chosen as shown in Fig. 4 (b). The opening this analysis with the following experiment. A plate with four
shape associated with the conformal transformation, shown by the rounded square holes as shown in Fig. 4(a) is under far-eld load
dashed curve in the gure, is positioned such that the upper right and the stress concentration at the magnied corner is of interest.
corner is tangent to the corner of the actual plate opening. To The strain gauges are placed at the intermediate region away from
match the curvature of the corner of the actual plate with the aver- this corner. In this paper the experiments are simulated using -
age curvature of the corner of the conformally mapped shape, the nite element results where the ne-mesh nite element results
size s of the conformally mapped shape is set to 10.4rc, in which rc provide the real stress distribution in the domain and coarse-mesh
is the radius of curvature of the actual plate. The origin of XY-plane nite element results determine the stresses which are accurate
in the conformal mapping approach is then located accordingly. only at the location of regression points (strain gages) away from
Using the middle of the conformally mapped shape as the origin, the corners. It is noted that in this method one can focus on the
the points shown in the gure are along the following two line stress concentration at only one corner of a hole at a time. The con-
segments: formal mapping is used to transform an innite plate with a square
hole with rounded corners to a unit disk. The size of the square
s hole with rounded corner and the location of regression points
< xk < s; yk s
4 22 shown in Fig. 4(b) are determined according to Section 3.2. The
s
< yk < s; xk s interaction effects of the other neighboring holes are taken into ac-
4
count using the results of coarse-mesh nite element analysis SFEM
For a corner with specic radii of curvature, the size of the square at the location of regression points where the coarse mesh can pro-
hole with rounded corners associated with the conformal mapping vide the stress distribution with good accuracy. The far-eld mo-
can be obtained as s = 10.4rc. Once the size of the square hole with ments producing these intermediate stress eld in an innite
rounded corners is known, the location of regression points to eval- plate with only one square hole with rounded corner are then ob-
uate SFEM as shown in Fig. 4(b) can be determined using the above tained using standard regression in Eqs. (21) and (23). Once these

(a) (b)

Fig. 4. Location of regression points for the semi-analytical approach for evaluating stress concentrations: (a) conformally mapped shapes and regression points shown for
one of the sixteen opening corners and (b) close up at one corner.
A. Louhghalam et al. / International Journal of Solids and Structures 48 (2011) 19912004 1999

far-eld moments are evaluated, the extended version of complex


variable method in Eq. (24) is used to nd the stress concentration
at the corner of interest.
In the following, the accuracy and applicability of the proposed
semi-analytical approach is examined through a series of exam-
ples. The rst example is a plate with a square hole of unit length
with rounded corners; the load is a constant moment Mx = 1 at the
far-eld. Fig. 6(a) shows how the maximum von Mises stress (nor-
malized with respect to the far-eld von Mises stress) varies with
respect to the radius of curvature of the corners.
A set of reference results are obtained by repeated nite ele-
ment analyses of this plate using ABAQUSs S4R general purpose
four-node shell element models (ABAQUS, 2007) for radii of curva-
ture ranging from approximately 0.001 to 0.1, where in each anal-
ysis a sufciently small mesh size as illustrated in Fig. 5 was used
to produce accurate values for the maximum von Mises stress.
These reference results are shown by the circles in the gure. For
the coarse nite element analyses, a simple square mesh was used Fig. 5. An example of nite element models with ne mesh around corner of
where the corner was modeled as a square, exactly as shown in the interest used for evaluation of reference stresses.
right side of Fig. 1. This was done twice, for mesh sizes of 0.04 and
0.008. Since there was no attempt to model the curvature of the
corner, such meshes would naturally produce poor stress results mesh size is used for the horizontal axis ordinate. This inaccuracy
at the corner. For sake of reference, the maximum stresses are plot- of the corner stress results is inconsequential in the proposed
ted in Fig. 6(a) by the square and triangular symbols, in which the semi-analytical method because only the stresses in the

(a) (b)

(c) (d)

Fig. 6. Normalized maximum von Mises stress versus the radius of curvature, computed by: ( ) ne-mesh nite element analysis; () semi-analytical approach based on the
coarse-mesh nite element analysis with smaller mesh size; ( ) semi-analytical approach based on the coarse-mesh nite element analysis with larger mesh size; (M)
coarse-mesh nite element analysis with smaller mesh size; ( ) coarse-mesh nite element analysis with larger mesh size. (a) Plate with one hole under unit constant Mx at
far-eld, (b) plate with one hole under unit Mxy at far-eld, (c) plate with four holes as shown in Fig. 1 with simply supported left and right edges under uniform transverse
body force, (d) plate with four holes as shown in Fig. 1 under transverse displacement proportional to xy at the edges.
2000 A. Louhghalam et al. / International Journal of Solids and Structures 48 (2011) 19912004

intermediate eld, as indicated in Fig. 4(b), are needed for the maximum stresses from the ne- and coarse-mesh nite element
stress calculations. Following the procedure in the preceding calculations are plotted along with the semi-analytical results.
section, these intermediate-eld stresses, SFEM, are used in Eqs. Here, the mesh sizes for the two coarse-mesh nite element
(21)(24) to obtain the nal predictions for the maximum von models are 0.02 and 0.077.
Mises stress. There are two sets of results here, one based on the The maximum von Mises stresses are plotted in the same
nite element results with the 0.04 mesh size and the second manner as in the preceding gures. It can be seen that the semi-
based on the results with the 0.008 mesh size; the results are plot- analytical results can provide accurate predictions for radii of
ted by the dashed and solid lines in Fig. 6(a). It can be seen that the curvature that are substantially smaller (1/4 to 1/3) than the mesh
proposed method can produce accurate results for the stress con- sizes for the coarse nite element models that are used to supply
centration for radii of curvature that are substantially smaller than the intermediate stresses, SFEM. The inaccuracy at the larger radii
the size of the mesh. For the 0.04 mesh size results, the stress con- of curvature is due to the fact that the regression points shown
centration predictions are accurate for radii of curvature of approx- in Fig. 4 actually approach the adjacent hole of the four-hole plate.
imately 0.008 while for the 0.008 mesh size results, the predictions Hence, the primary application of the method would be to investi-
are accurate for radii of approximately 0.003; hence, the mesh gate small radii of curvature rather than large ones. The method re-
sizes that are three to ve times larger than the radius of curvature mains practical because the stress analysis of plates with holes
of the interior corners. It is noted that to produce similarly accurate with small radii of curvature is far more computationally challeng-
results using only a ne nite element mesh, mesh sizes of approx- ing than cases where the radii are large.
imately one tenth of the radius of curvature is typically required. For the forth example, the loading produces a predominantly
In essence, the semi-analytical method leverages standard nite torsional moment. The results for this case, shown in Fig. 6(d),
element analyses by providing more detailed information about are similar to those shown in Fig. 6(b). In particular, there is rela-
the corner stress concentration. The end result is that it is possible tively little interaction between stresses generated by adjacent
to obtain accurate predictions for the maximum stresses using a corners. This remains true even with the four closely spaced
mesh size that is well over a magnitude larger than meshes used corners shown near the middle of the four-hole plate.
in standard nite element analysis of such plate problems. Further- It is noted that the results of the proposed method is not
more, it can be seen that, by using the results from a single nite sensitive to the location of regression points used to evaluate SFEM
element analysis, it is possible to predict the maximum stresses as shown in Fig. 4(b). A sensitivity analysis is performed by varying
for a range of radii of curvature. With standard nite element anal- the lower bound of the inequality in Eq. (22) that species the
ysis, such calculations would require repeated remeshing and locations of the regression points from s/4 to 0. This variation in
reanalysis of the plate. The semi-analytical method, however, does the location of the regression points resulted in at most a 6%
have some limits. Fig. 6(a) shows some inaccuracy at larger radii of change in the stress predictions for both plate examples and for
curvature. The reason for this is that, at such radii, the intermediate a wide range of radii of curvature.
eld, as dened in Fig. 4, extend far into the plate. In these ex- The results shown in Fig. 6(a)(d) are all for the maximum von
tended regions, there is some interference arising from stress pat- Mises stress, plotted with respect to the radius of curvature of the
terns from neighboring corners. However, as the radius of corner. It is also useful to examine and compare the spatial stress
curvature decreases, the intermediate eld also reduces to rela- eld obtained by the various methods. In Fig. 7, the third example
tively small regions in the neighborhood of the corner. In such with radius of curvature rc = 0.012 is chosen to compare the von
neighborhoods, the stress eld is dominated by the stress patterns Mises stress elds. The nite element results with coarse 0.02
associated with the nearby corner, and the stress patterns from mesh size are shown in Fig. 7(a) using a relatively large 1  1 win-
neighboring corners are of negligible magnitudes, leading to more dow so that the overall von Mises stress distribution can be seen
accurate results. At very small radii of curvature, the nite element and again in 7(b) in a smaller 0.12  0.12 window centered at
analysis would be unable to produce accurate stress results at the the upper right corner to get a closer view of the stress concentra-
intermediate locations shown in Fig. 4. To obtain accurate stress tion. As expected, the coarse mesh results do not resolve the stress
concentration results at such small radii, ner meshes are needed concentration at the corner. The reference von Mises stress eld,
in the nite element analysis. This is reected in the results in obtained using the nite element model in Fig. 5 where the mesh
Fig. 6(a), where the semi-analytical results based on a mesh size size in the neighborhood of the corner is approximately 0.001 is
of 0.04 is accurate down to a radius of approximately 0.008 and shown using the smaller window. In this case, the stress concentra-
that more accurate results for such small radii were obtained by tion is clearly seen at the upper side of the rounded corner. In the
using the smaller mesh size of 0.008. coarse mesh analysis, there is no attempt to model the curvature at
The second example is similar to the rst except that the far- the corner given that the radius of curvature is over 50% of the
eld moment is a constant unit torsion, Mxy = 1. The maximum mesh size. Nevertheless, the coarse mesh results are still useful be-
von Mises stresses obtained using the ne-mesh and coarse-mesh cause the intermediate stress eld SFEM at the points indicated in
nite element methods and by the semi-analytical approach are the gure can be used in the semi-analytical approach. The results
plotted in Fig. 6(b). The comparison of results is very similar to of the semi-analytical approach, shown in Fig. 7(d) are in very good
the comparison in Fig. 6(a) for the constant unit-moment case. agreement with the results of the reference ne-mesh nite ele-
The main difference is that at larger radii of curvature, the semi- ment analysis in Fig. 7(c), particularly when compared with the
analytical approach gives relatively accurate results. This is be- coarse mesh results in Fig. 7(b). It was found that the spatial stress
cause the interference of stress patterns from neighboring corners pattern for the fourth example (four-hole plate under torsional
is not as signicant for the torsion loading case. load) has some differences with the stress patterns obtained using
For the next two examples, a plate with four openings with the the ne-mesh nite elements. Nevertheless, the predicted magni-
corner of interest shown in Fig. 1 is analyzed. For the third exam- tude of the stress concentration is close to the nite element result,
ple, the plate is simply supported at the left and right edges and a as shown in Fig. 7(d).
uniform transverse body force is applied. For the fourth example, To obtain more accurate results in the spatial stress patterns as
all four edges are simply supported, and the transverse displace- well as in the maximum stress predictions, it is necessary to in-
ment of these edges are proportional to xy, where the origin is at clude more general far-eld moment distributions in Table 2. This
the center of the plate; this displacement produces a predomi- would increase the number of expressions in the functional
nately torsional load on the plate. As in Figs. 6(a) and (b), the expressions for the stress distribution, leading to better matches
A. Louhghalam et al. / International Journal of Solids and Structures 48 (2011) 19912004 2001

(a) (b)

(c) (d)
Fig. 7. Distribution of von Mises stresses in the plate with four holes shown in Fig. 1 with simply supported left and right edges under uniform transverse body force
computed by: (a) coarse-mesh nite element analysis for a wide region; (b) coarse-mesh nite element analysis for a narrow region with location of intermediate stress eld
shown with (+) symbols; (c) ne-mesh nite element analysis; (d) semi-analytical approach.

to more complex patterns in the intermediate-eld spatial stress. nite-element results. It is noted that the stress values are up to
Such an analysis is straightforward, but lengthy, and is beyond 52% higher than those in the four-hole example shown in
the scope of the present paper, in which the emphasis is in the Fig. 6(c). This indicates the sensitivity of the corner stress concen-
introduction of the basic semi-analytical method. tration to the opening conguration, and it also indicates that the
The semi-analytical approach is robust with respect to the con- proposed method can capture this sensitivity.
guration of the openings. This is because it uses the local stress For the nal series of examples, the radii of curvature of the
eld in a neighborhood around the particular opening corner of corner of interest is held xed at rc = 0.012, while the distance d
interest to predict the stress concentration at this corner. Thus, if
the conguration of openings change, resulting in changes in this
local stress eld, then the semi-analytical approach would give
new stress concentration predictions at the corner. To illustrate
this robustness of the method, two more sets of examples, with
the opening conguration shown in Fig. 8, are examined. The plate
has two openings, one angled at 45 with respect to the other. It is
loaded by a transverse body force, and the stress concentration of
interest is in the corner enclosed by the dashed lines.
In the rst set of examples, the diagonal spacing d between the
corner of interest and the adjacent hole, as indicated in Fig. 8, is
held constant. A coarse nite-element mesh with mesh size of
approximately 0.02 is used as input to the semi-analytical method.
As in the results in Fig. 6, the maximum von Mises stress at the cor-
ner is evaluated for different radii of curvature using the semi-ana-
lytical approach and compared with ne-mesh nite element
results. The comparison, shown in Fig. 9(a), shows high accuracy
for a relatively wide range of radii of curvature. As in Fig. 6, the
accuracy decreases as the radii of curvature becomes over an order
of magnitude smaller than the mesh size used in the coarse Fig. 8. Plate with two holes used for fth and sixth examples.
2002 A. Louhghalam et al. / International Journal of Solids and Structures 48 (2011) 19912004

(a) (b)
Fig. 9. Normalized maximum von Mises stress computed by () semi-analytical approach based on a coarse-mesh nite element analysis with mesh size 0.02; ( ) ne-mesh
nite element analysis for the plate with: (a) xed opening distance d = 0.57 and variable corner radius of curvature; (b) xed radius of curvature rc = 0.012 and variable
opening distance d.

between the two openings is varied. The comparison of von Mises @u @ v @u @v


a;  b
stress results, shown in Fig. 9(b), shows a dramatic increase of the @x @y @y @x
stress concentration, as the distance d decreases. This increased
stress, which is well over triple the stress results shown in the @a @b @a @b
0;  0
four-hole example, is accurately predicted by the semi-analytical @x @y @y @x
results for the entire range of distances shown in the gure.
Using the chain rule, the rst derivatives of a function with respect
to the variables in physical space can be related to the derivatives
4. Conclusion with respect to the variables in the unit disk:

The method of complex variables (Muskhelishvili, 1975) for


@ @ @ @ @ @
plates with a rectangular opening was analytically extended so a b ; b a
@x @u @v @y @u @v
that regression, instead of contour integration, can be used to sat-
isfy the boundary conditions. It was shown how these analytical
results can be used in a semi-analytical method for evaluating The relation between the second derivatives are as follows:
the stresses at the corners of the openings. The method combines  
stress results from a single, relatively coarsely meshed nite ele- @2 @ @ @ @2 2 @
2
a  b a2 2 b
ment model with the complex variable expressions to predict the @x2 @x @u @v @u @v 2
magnitude of the stress concentration for a range of radii of curva-
ture of the interior corners. Accurate results were obtained even @2 @a @ @b @
 2ab 
when the nite element model itself is too coarse to resolve the @u @ v @x @u @x @ v
stress concentration. Specically, the method can predict maxi-
mum corner stresses for radii of curvature that are of the order  
of the nite element mesh and smaller. @2 @ @ @ 2 @
2
@2
b a b a2 2
A series of examples were used to assess the accuracy and appli- @y2 @y @u @v @u 2 @v
cability of the proposed approach. It was found that the magnitude
of the stress concentrations can be determined with very good @2 @b @ @a @
2ab
accuracy for a wide range of radii of curvature and for plates with @u @y @y @u @y @ v
multiple rectangular holes. For simpler bending deformations, the  
spatial stress distributions away from the corner are fairly accu- @2 @ @ @ @2 @2
a b ab 2  ab 2
rately reproduced, but it was found that these distributions are less @y @x @y @u @v @u @v
accurate when the plate is under twisting moments. This is due to
the fact that the functional expansion developed herein do not ac- 2 @2 @a @ @b @
a2  b 
count for the spatial complexity of the stress patterns. In future @u @ v @y @u @y @ v
work, the library of expressions used in the functional expansion
will be enlarged so that the proposed method would be applicable
Appendix B. Moments inside the unit disk in terms of the
to a wider range of loads and plate congurations. This work would
analytic functions
not require any fundamental changes in the analytical results or
in the semi-analytical method presented herein; it would simply
Using the complex representation for the solution to the bihar-
require the determination of more terms in the functional
monic equation we have:
expansion.
w Rez/ v
Appendix A. Relation between the derivatives in physical space
and unit disk where / and v are analytic functions. Using the conformal proper-
ties of these functions, the rst and second derivatives of the deec-
For a conformal function Cauchy Reimann relation holds: tion w can be determined as follows:
A. Louhghalam et al. / International Journal of Solids and Structures 48 (2011) 19912004 2003

(  00 
@w @M qu 1 @A2 z
Rez0 / /0 z v0 V n A3 Nq 2 Mqu 8DRez0 /0 Re 0 eiu
@u q @u A q @u z
 00
@w z 2iu  v00 i/0 z0 z0 /0
Reiz0 / i/0 z iv0 D1  mIm eiu 0 ie /00 z z00 /
@v z
iu
ie
@2w   v0  eiu /0z z0 /
/0z z0 /  v0
Rez00 / 2z0 /0 z/00 v00 q
@u2 !!#)
1 @A2 i 000 0
iu z z  z
002
 ie 29
@2w A2 q @ u q z0 z00
Reiz00 / iz/00 iv00
@u@ v
In the above equations the terms Mqq, Mq/, M// and Nq can also be
@2w written in terms of the analytic functions and the conformal
Rez00 / 2z0 /0  z/00  v00
@v 2 transformation:
!
The derivatives of deection used in the moments and shear rela- @2 m @ m @2
Mqq D w
tions are evaluated in terms of the analytic functions as shown in @ q2 q @ q q2 @ u2
the following:

DRe e2iu /00 z z00 /


 v00 1  m /0 z0 z0 /0 1 m
@w @w @w  v0
c s Reeiu /0 z z0 / 30
@q @u @v
!
D1  m @2 @
Mqu  w
@w @w @w q @ q@ u q@ u
s c  v0
Imeiu /0 z z0 /
q @u @u @v D1  mIm e2iu /00z z00 /
 v00 31

@2w !
 v00 /0 z0 z0 /0
Ree2iu /00 z z00 / @2 1 @ 1 @2
@ q2 Muu D m 2 w
@q q @ q q2 @ u2

iu  v00 1  m /0 z0 z0 /0 1 m
@2w ie 2iu DRe e2iu /00 z z00 /
Re /0 z z0 / v0 ie /00 z z00 / v00
q @u @q q 32
!
i/0 z0 z0 / 0
@ 2

Nq D r w 4DRe eiu /00 z0 /0 z00 33


@q
iu
@2w ie  v0 ie2iu /00 z z00 /
 v00
 Im /0 z z0 / References
q2 @ u2 q
!
ABAQUS, ABAQUS version 6.7, ABAQUS Inc., 2007.
i/0 z0 z0 /0 Bryukhanova, E.N., 1967. Thermal stress in a circular cylinder with regularly
arranged circular cavities. International Journal of Applied Mechanics 5 (4),
380384.
and Chen, C., Hsu, J., 1996. The stress intensity factors of slightly undulating interface
cracks of bimaterials. International Journal of Fracture 80, 277293.
Chen, H.C., 1993. Special nite element including stress concentration effects of a
@w @w hole. Journal of Finite Element Analysis and Design 13, 249258.
 i eiu /0 z z0 /
 v0
@q q@ u Cherkaev, A.V., Grabovsky, Y., 1998. The cavity of the optimal shape under the shear
stresses. International Journal of Solids and Structures 35 (33), 43914410.
Chong, K.P., Pinter, W.J., 1984. Stress concentrations of tensile strips with large
holes. Journal of Computers and Structures 19 (4), 583589.
@2w @2w 2iu
Datsyshin, A.P., Marchenko, G.P., 1985. Interaction of curvllinear cracks with the
 i  ie /00 z z00 / v00 i/0 z0 z0 /0 boundary of an elastic half-plane. Journal of Material Science 20 (5), 466473.
q @ qu q2 @ u2 Durelli, A.J., Parks, V.J., Lopardo, V.J., 1970. Stresses and nite strains around an
iu elliptic hole in nite plates subjected to uniform load. International Journal of
ie  v0
 /0 z z0 / Nonlinear Mechanics 5, 397411.
q Exadaktylos, G.E., Liolios, P.A., Stavropoulou, M.C., 2003. A semi-analytical elastic
stressdisplacement solution for notched circular openings in rocks.
International Journal of Solids and Structures 40 (5), 11651187.
The moments and shear in physical space are: Exadaktylos, G.E., Stavropoulou, M.C., 2002. A closed-form elastic solution for
   stresses and displacements around tunnels. International Journal of Rock
00
 v0 z eiu
Mechanics and Mining Sciences 39, 905916.
M nn A2 Mqq  D1  mRe eiu /0 z z0 / 25 Feng, Z., Rowlands, R.E., 1991. Smoothing nite element and experimental hybrid
z0 technique for stress analyzing composites. Journal of Computers and Structures
   39 (6), 631639.
00
 v0 z eiu
Haani, H., Dwyer, J.F., 1999. Edge function analysis of anisotropic materials with
M tt A2 M uu D1  mRe eiu /0 z z0 / 0
26 holes and cracks. Journal of Computers and Structures 72, 779791.
z Henshaw, J.M., Sorem Jr., J.R., Glaessgen, E.H., 1996. Finite element analysis of ply-
   by-ply and equivalent stress concentrations in composite plates with multiple
00
 v0 z eiu
M nt A2 Mqu D1  mIm eiu /0 z z0 / 27
holes under tensile and shear loading. Journal of Composite Structures 36, 45
58.
z 0
Hu, K.X., Chandra, A., Huang, Y., 1993. Multiple void-crack interaction. International
  00  Journal of Solids and Structures 30 (11), 14731489.
z Ishikawa, H., Kohno, Y., 1993. Analysis of stress singularities at the corner point of
Nn A3 Nq 8DRe z0 /0 Re 0 eiu 28 square hole and rigid square inclusion in elastic plates by conformal mapping.
z
International Journal of Engineering Science 31 (8), 11971213.
2004 A. Louhghalam et al. / International Journal of Solids and Structures 48 (2011) 19912004

Jasiuk, I., 1995. Cavities vis-s-vis inclusions: Elastic moduli of materials with Timoshenko, S., Woinowsky-Krieger, S., 1959. Theory of Plates and Shells, second
polygonal inclusions. International Journal of Solids and Structures 32 (34), ed. McGraw Hill.
407422. Tiwary, A., Hu, C., Ghosh, S., 2007. Numerical conformal mapping method based
Muskhelishvili, N.I., 1975. Some Basic Problems of the Mathematical Theory of Voronoi cell nite element model for analyzing microstructures with irregular
Elasticity, fourth ed. Noordhoff International Publishing. heterogeneities. Journal of Finite Elements in Analysis and Design 43, 504520.
Nishioka, T., Atluri, S.N., 1982. Stress analysis of holes in angle-ply laminated: an Tsukrov, I., Novak, J., 2002. Effective elastic properties of solids with defects of
efcient assumed stress special-hole-element approach and a simple irregular shapes. International Journal of Solids and Structures 39, 15391555.
estimation method. Journal of Computers and Structures 15 (2), 135147. Ukadgaonker, V.G., Rao, D.K.N., 2000. A general solution for moments around holes
Pan, E., Yang, B., Cai, G., Yuan, F.G., 2001. Stress analyses around holes in composite in symmetric laminates. Journal of Composite Structures 49, 4154.
laminates using boundary element method. Journal of Engineering Analysis Vigdergauz, S., 1993. optimal stiffening of holes, under equibiaxial tension.
with Boundary Elements 25, 3140. International Journal of Solids and Structures 30 (4), 569577.
Peterson, R.E., 1961. Petersons Stress Concentration Factors. John Wiley and Sons Wang, X.F., Hasebe, N., 2000. Bending of a thin plate containing a rigid inclusion and
Inc.. a crack. Engineering Analysis with Boundary Elements 24, 145153.
Piltner, R., 1985. Special nite elements with holes and internal cracks. Wu, H.C., Mu, B., 2003. On stress concentrations for isotropic/orthotropic plates and
International Journal for Numerical Methods in Engineering 21, 14711485. cylinders with a circular hole. Journal of Composites Part B: Engineering 34,
Richards, T.H., Daniels, M.J., 1986. Enhancing nite element boundary stress 127134.
predictions for plane and axisymmetric situations. Journal of Strain Analysis Wu, X.J., Cheng, S.M., 1999. A higher order theory for plane stress conditions of
21 (1), 3344. laminates consisting of isotropic layers. Journal of Applied Mechanics 66,
Rowlands, R.E., Winters, K.D., Jensen, J.A., 1978. Full-eld numerical differentiation. 95100.
Journal of Strain Analysis 13 (3), 177183. Xiwu, X., Liangxin, S., Xuqi, F., 1995. Stress concentration of nite composite
Savin, G.N., 1961. Stress concentration around holes. International Series of laminates with elliptical hole. Journal of Computers and Structures 57 (1),
Monographs in Aeronautics and Astronautics. Pergamon Press. 2934.

Vous aimerez peut-être aussi