Documente Academic
Documente Profesional
Documente Cultură
15
16
Young modulus and the Poisson ratio. In this chapter the rigid solid motion and
deformation are described, with emphasis on rotation, which plays an important
role in nonlinear continuum mechanics. Also the concepts of strain and stress in
nonlinear mechanics are introduced. The equilibrium and the constitutive equations are presented afterwards.
2.1.1 Kinematics
The general motion of a deformable body is represented in Fig. 2.1. The body, in
the initial position t 0, is considered to be an assemblage of material particles,
labelled by the coordinates X, with respect to the Cartesian basis e. The current
position of a particle is defined at time t by the coordinates x.
The motion can be mathematically described by a mapping function / between
initial and current particle positions,
x /X; t
2:1
o/
r/
oX
2:2
2:3
ox
oX
2:4
17
ox
6 oX
oy
4 oX
oz
oX
ox
oY
oy
oY
oz
oY
ox
oZ
oy
oZ
oz
oZ
3
7
5
2:5
2:6
The Jacobian determinant can be used to relate the integral of a given functional
f in the current and in the initial configuration by,
Z
Z
f x; t dX
f /X; t; t J dX0 :
2:7
X
X0
2.1.1.2 Strain
Consider the change of the scalar product of the two elemental vectors from
dX 1 Q1 P and dX 2 Q2 P, initial configuration, to dx1 q1 p and
dx2 q2 p, current configuration, as a general measure of deformation. Where
Q1 and Q2 are two material particles in the neighbourhood of a material particle P
for the initial configuration and q1 and q2 and p the same respective material
18
particles in the current configuration. Equation (2.4) permits the following relations dx1 F dX 1 and dx2 F dX2 , and the spatial scalar product dx1 dx2 can
be found in terms of the material vectors dX 1 and dX2 as,
dx1 dx2 dX1 FT FdX 2
2:8
2:9
2:10
The change in scalar product can now be found in terms of the material vectors
dX 1 and dX2 and the Lagrange or Green strain tensor E can be defined as,
1
dx1 dx2 dX 1 dX 2 dX1 E dX2
2
2:11
2:12
2:13
where
RT R I
and U UT
2:14
2:15
In order to actually obtain U from Eq. (2.15) it is first necessary to evaluate the
principal directions of C, represented by the eigenvectors set W i and the
19
3
X
k2i W i W i
2:16
i1
3
X
ki W i W i
2:17
i1
Once the stretch tensor U is known, the rotation tensor R can be obtained
without difficulty from Eq. (2.13).
R F U 1 :
2:18
2.1.1.4 Stress
In a large deformation analysis a body can experience a large rotation and/or a
large strain. The defined stress terms together with the obtained strain terms
enables to express the virtual work as an integral over the known body volume,
expressing in this manner the change in the body configuration. Both strain tensor
and stress tensor are referred to the same deformed state. The Cauchy stress tensor,
here defined as K, is a symmetric tensor and it represents the stresses of the current
configuration. For the three-dimensional case it can be defined as,
2
rxx
K 4 ryx
rzx
rxy
ryy
rzy
3
rxz
ryz 5
rzz
2:19
In this work it is used the Voigt notation, since the development of fourth order
tensors is less practical. In Voigt notation the tensors are expressed in column
vectors, so the stress tensor K is reduced to the stress vector r,
r f rxx
ryy
rzz
rxy
ryz
rzx gT
2:20
exx
eyy
ezz
exy
eyz
ezx
T
2:21
20
3 2
rxx
t^e1
K 4 t^e2 5 4 ryx
rzx
t^e3
rxy
ryy
rzy
3
rxz
ryz 5
rzz
2:22
where ^e1 , ^e2 and ^e3 are the versors of the coordinate system and t^ei is the stress
vector on a plane normal to ^e1 passing through the interest point, Fig. 2.2(a).
Following Cauchys stress theorem, if the stress vectors of three orthogonal planes,
with a common point, are known, then the stress vector on any other plane passing
through that point can be found through the coordinate transformation equations
[5]. Thus, the stress vector tn in a point belonging to an inclined plane,
Fig. 2.2(b), can be defined by,
2
tn n rij n1
n2
rxx
n3 4 ryx
rzx
rxy
ryy
rzy
3
rxz
ryz 5
rzz
2:23
where n is the inclined plane normal vector. The relation in Eq. (2.23) leads to the
transformation rule of the stress tensor. The initial stress tensor rij , defined in the
xi coordinate system, can be transformed in a new stress tensor r0ij , defined in
another x0i coordinate system by the relation,
K0 A K AT
2:24
rxx
6 0
4 ryx
0
rzx
rxy
0
ryy
0
rzy
0 3
2
rxz
a11
0 7
ryz 5 4 a21
0
a31
rzz
a12
a22
a32
32
rxx
a13
5
4
ryx
a23
rzx
a33
rxy
ryy
rzy
32
rxz
a11
ryz 54 a12
rzz
a13
a21
a22
a23
3
a31
a32 5
a33
2:25
The aij coefficients can be understood as the projection of the x0i coordinate
system versors in the xi coordinate system versors. Therefore, the angle between
the versors of each coordinate system can be defined as,
cij cos1 aij
2:26
Through Eq. (2.26) and Fig. 2.3 it is possible to comprehend better the physical
meaning of the aij coefficients and the respective angles.
21
Fig. 2.2 a Three-dimensional stress components. b Stress vector acting on a plane with normal
vector n
Let P be an interest point of a considered stressed body. There are at least three
planes, orthogonal with each other, crossing P where the corresponding stress
vector is normal to the plane. These planes are called principal planes and the
22
normal vectors of each plane are called principal directions. The stress vectors are
parallel to the plane normal vectors and are called principal stresses.
The stress tensor is a physical quantity, independent of the coordinate system
chosen to represent it. Therefore, there are certain invariants associated with it
which are also independent of the coordinate system. Being a second order tensor,
the stress tensor has associated three independent invariant quantities. One set of
such invariants are the principal stresses of the stress tensor, which are just the
eigenvalues of the stress tensor. Their direction vectors are the principal directions
or eigenvectors. A stress vector parallel to the normal vector n is given by,
tn kn rn n
2:27
ti
kni ) rij nj kdij nj ) rij kdij nj 0
2:28
rxz
ryz 0
rzz k
rxy
ryy k
rzy
2:29
2:30
1
r r
2 ii jj
rij rji
I3 det rij
2:31
2:32
2:33
The three roots k1 r1 ; k2 r2 and k3 r3 of Eq. (2.30) are the eigenvalues or principal stresses, which are unique. Therefore the stress invariants have
always the same value regardless of the orientation of the chosen coordinate
system. For each eigenvalue k, exists a non-trivial solution n on Eq. (2.28). These
nj solutions, called eigenvectors, are the principal directions, which defines the
plane where the respective stress acts. Applying Eq. (2.25),
r1
4 0
0
0
r2
0
3 2
0
n11
0 5 4 n21
r3
n31
23
32
rxx
n13
n23 54 ryx
rzx
n33
n12
n22
n32
32
rxz
n11
ryz 54 n12
rzz
n13
rxy
ryy
rzy
n21
n22
n23
3
n31
n32 5
n33
2:34
The principal stresses and principal directions characterize the stress in P and
are independent of the orientation of the coordinate system.
2:35
2:36
1
E
6 xxtxy
6 Exx
6 t
6 xz
6 Exx
s6
6 0
6
6 0
4
0
Eyxyy
1
Eyy
t
Eyzyy
0
0
0
Etzxzz
t
Ezyzz
1
Ezz
0
0
0
0
0
0
1
Gxy
0
0
0
0
0
0
1
Gyz
0
0
0
0
0
1
Gzx
3
7
7
7
7
7
7
7
7
7
5
2:37
The material constitutive matrix c is obtained by inverting the material compliance matrix s, which is here defined for an three-dimensional anisotropic
material. The elements on matrix s are obtained experimentally. Eii is the Young
modulus in direction i, tij is the Poisson ratio which characterizes the deformation
rate in direction j when a force is applied in direction i, Gij is the shear modulus
which characterizes the variation angle between directions i and j. Due to symmetry the following relation can be established,
Ei tji Ej tij
2:38
For the two-dimensional case the plane stress and plane strain [5] deformation
theory assumptions can be presumed. Considering the plane stress assumptions,
24
rzx rzy rzz 0, the material compliance matrix s is obtained directly from
the three-dimensional compliance matrix s,
2
s
1
Exx
6 txy
4 Exx
Eyxyy
1
Eyy
3
0
0 7
5
2:39
1
Gxy
For the plane strain deformation theory it is considered ezx ezy ezz 0 and
the material compliance matrix s is defined as,
2
s
tzx txz
1
Exx Exx
6 txy tzy txz
4 Exx
Exx
t t
Eyxyy zxEyyyz
tzy tyz
1
Eyy Eyy
0
3
0
0 7
5
2:40
1
Gxy
1
n n
oxy ox
noxy knox k
!
and
x cos
1
n n
oxy
noxy knk
!
2:41
With the obtained angle information it is now possible to rotate the material
matrix using the rotational transformation matrix and therefore align the material
ox axis with the known vector n. The rotational transformation matrix that permits
an anticlockwise rotation along the ox axis of a known angle b can be defined as,
T ox
1
60
6
60
6
60
6
40
0
25
0
cos2 b
sin2 b
0
sin b cos b
0
0
0
sin2 b
0
cos2 b
0
0
cos b
sin b cos b
0
0
sin b
0
sin 2b
sin 2b
0
cos2 b sin2 b
0
3
0
0 7
7
0 7
7
sin b 7
7
0 5
cos b
2:42
T oy
cos2 b
6
0
6
2
6
sin
b
6
6
0
6
4
0
sin b cos b
0
1
0
0
0
0
sin2 b
0
0
0
0
0
cos2 b
0
0
0
cos b sin b
0
sin b cos b
sin b cos b
0
0
3
sin 2b
7
0
7
sin 2b 7
7
7
0
7
5
0
cos2 b sin2 b
2:43
cos2 b
6 sin2 b
6
6
0
T oz 6
6 sin b cos b
6
4
0
0
sin2 b
cos2 b
0
sin b cos b
0
0
0
0
1
0
0
0
sin 2b
sin 2b
0
cos2 b sin2 b
0
0
0
0
0
0
cos b
sin b
3
0
0 7
7
0 7
7
0 7
7
sin b 5
cos b
2:44
2:45
26
accurate results. These are the reasons why so many prefer the weak form to obtain
the approximated solution.
In this work the discrete equation system is obtained using the Galerkin weak
form, which is a variational method [6]. For meshless methods used in this book
the discrete system of equations is obtained similarly with the FEM, with some
differences inherent to the meshless approach. The discrete equations for the static
and the dynamic approach are developed and shown for the basic three-dimensional deformation theory.
2:46
being T the kinetic energy, U is the strain energy and Wf is the work produced by
the external forces. The kinetic energy is defined by,
T
1
2
qu_ T u_ dX
2:47
where the solid volume is defined by X and u_ is the displacement first derivative
with respect to time, i.e., the velocity. The solid mass density is defined by q. The
strain energy, for elastic materials, is defined as,
Z
1
eT r dX
2:48
U
2
being e the strain vector and r the stress vector. The work produced by the external
forces can be expressed as,
Wf
Z
X
uT b dX
Z
Ct
uT t dC
2:49
27
where u represents the displacement, b the body forces and Ct the traction
boundary where the external forces t are applied. By substitution the Lagrangian
functional L can be rewritten as,
L
1
2
qu_ u_
1
dX
2
e r dX
u b dX
uT t dC
2:50
Ct
2
61
42
t1
qu_ T u_ dX
1
2
eT r dX
uT b dX
3
7
uT t dC5 dt 0
2:51
Ct
2
61
42
1
d qu_ T u_ dX
d eT r dX
duT b dX
3
7
duT t dC5 dt 0
Ct
2:52
Since all operations are linear, changing the order of operation does not affect
the result. In the first term of Eq. (2.52) the time integral can be moved inside the
spatial integral,
2
3
Zt2
Z
Z Zt2
1
1
4 d qu_ T u_ dt5 dX
d qu_ T u_ dX dt
2:53
2 X
2
t1
t1
Using the chain rule of variation and then the scalar property, the integral can
be rewritten as,
28
Zt2
d qu_ u_ dt q
T
t1
Zt2
du_ u_ u_ du_ dt 2q
t1
Zt2
du_ T u_ dt
2:54
t1
du_ T u_ dt
t1
Zt2
oduT ou
dt
ot ot
2:55
t1
t2
Zt2
2
oduT ou
o
u
ou
duT 2 dt duT
dt
ot
ot
ot ot
t1
2:56
t1
Notice that u satisfies, by imposition, the conditions at the initial time, t1 , and
final time, t2 , leading to a null du at t1 and t2 . Therefore the last term in Eq. (2.56)
vanishes. Considering the last development and switching the integration order
again, Eq. (2.53) becomes,
Zt2
Z
t1
1
d qu_ T u_ dX
2 X
Zt2
Z
T
dt
q
du u
dt
t1
2:57
being
u = o2 u ot2 the acceleration. The second term on Eq. (2.52) can also be
developed. The integrand function in the second integral term can be written as
follows,
d eT r deT r + eT dr
2:58
as the two terms in Eq. (2.58) are in fact scalars, the transpose does not affect the
result, as so,
T
eT dr eT dr drT e
2:59
2:60
29
2:61
2
41
2
2
3
Zt2 Z
d eT r dX5dt 4 deT rdX5dt
t1
2:62
2
6
4q
t1
dX
duT u
Z
X
deT r dX
Z
X
duT b dX
3
7
duT t dC 5dt 0 2:63
Ct
To satisfy Eq. (2.63) for all possible choices of the integrand of the time
integration has to be null, leading to the following expression,
Z
Z
Z
Z
T
dX
q
du u
deT r dX duT b dX
duT t dC 0
2:64
X
Ct
This last equation is known as the Galerkin weak form, which can also be
viewed as the principle of virtual work. The principle of virtual work states that if
a solid body is in equilibrium, the virtual work produced by the body inner stresses
and the body applied external forces should vanish when the body experiments a
virtual displacement. Considering the stress-strain relation, r c e, and the straindisplacement relation, e L u, Eq. (2.64) can be rearranged in the following
expression,
Z
Z
Z
Z
T
T
T
dX 0 2:65
dL u cL udX du b dX
du t dC q duT u
X
Ct
which is the generic Galerkin weak form written in terms of displacement, very
useful in solid mechanical problems. In static problems the fourth term of
Eq. (2.65) disappears.
30
influence-domain, which imposes the nodal connectivity between the neighbouring nodes. The meshless trial function uxI is given by,
uxI
n
X
ui xI ui
2:66
i1
2:67
n
X
ui xI dui
2:68
i1
where dui are the nodal values for the test function.
Z
X
r de dX
b du dX
t du dC q
dX 0
duT u
2:69
2:70
where B is the deformation matrix. Thus, the virtual work of the first term in
Eq. (2.69), using Eq. (2.70), can be expressed as,
L1
duT B r dX
2:71
The strain vector can be divided in two parts, the linear part and the nonlinear
part,
31
e e0 + eNL
2:72
1
1
e |{z}
Lh Ah L A h
2
2
|{z}
e0
2:73
eNL
eT1
6
6 0
6 13
6
6 0
6
6 13
L6 T
6 e2
6
6
6 0
6 13
4 T
e3
13
eT2
13
eT1
eT3
0
13
13 7
2
0 7
1
7
13 7
6
0
7
eT3 7 6
0
7 6
76
0 7 6
60
13 7
40
7
eT2 7
0
7
5
T
e1
0
0
0
1
0
0
0
0
0
0
0
1
0
0
0
1
0
0
0
1
0
0
0
0
0
0
0
0
1
0
0
0
0
0
0
1
0
0
0
0
1
0
3
0
07
7
17
7
07
7
05
0
2:74
e2
e3
2:75
2:76
2 ou
ou
oz
6
GT 4 0
ou
oy
ou
ox
ou
oy
ou
oz
ou
ox
ou
oy
7
05
ox
2:77
ou
oz
2 ou 3
on
being
6 ov 7
hn 4 on
5
2:78
ow
on
The current configuration displacement is considered in matrix A, which corresponds in Eq. (2.73) to the actualized component.
32
hTx
6
6 0
6 1x3
6
6 0
6
6 1x3
A6 T
6 hy
6
6
6 0
6 1x3
4
hTz
1x3
hTy
1x3
hTx
hTz
0
1x3
7 2 ou
0 7
7 6 ox
1x3 7
60
hTz 7
7 6
0
7 6
76
ou
6
0 7 6 oy
1x3 7
7 6
40
hTy 7
7
ou
5
oz
T
hx
1x3
ov
ox
ow
ox
0
0
0
ow
oy
0
0
0
0
0
0
ou
oy
ov
oy
ov
ox
ov
oz
ow
ox
ow
oz
ov
oz
ou
ox
ou
oz
ou
oz
ov
oy
ow
oy
ow
oz
ou
oy
ou
ox
ov
oy
ov
ox
ov
oz
3
0
07
7
ow 7
7
oz 7
07
7
ow 7
oy 5
2:79
ow
ox
2:80
since it varies with the deformation of the solid. The linear part of the deformation
matrix is represented by B0 and the nonlinear contribution by BNL . For the threedimensional case,
2 ou
6
BT0 4 0
ou
oy
ou
oz
ox
ou
oy
ou
ox
0
ou
oz
ou
oy
ou 3
oz
7
05
2:81
ou
ox
and
BNL A G
2:82
2:83
33
2:84
2:85
2:87
Being,
Z
Kr
dBTNL r dX
2:88
BT0 c B0 dX
2:89
K0
Z
X
K NL
BT0 c BNL BTNL c BNL BTNL c B0 dX
2:90
GT dAT r dX
2:91
dhx
6
6 0
dAT 6 31
4
0
31
31
31
dhy
dhy
dhx
dhz
dhy
31
31
dhz
31
31
dhz
7
0 7
31 7
5
dhx
2:92
34
rxx
33
6
6
I
dAT r 6 syx 33
4
szx I
33
sxy
sxz
syz
szy I
rzz
33
ryy
33
33
33 7
I 7
dh
33 7
5
Z dh ZG du
2:93
33
GT Z G dX
2:94
Therefore, the initial stress matrix K r takes into consideration the actualized
stress field.
dX5 M d
duT u
u
2:95
HT q H dX
2:96
being H the interpolation function matrix for the interest point i defined as,
H i ui I
2:97
Where ui is the interpolation function for interest node i and I is the identity
matrix defined in Eq. (2.75). The density diagonal matrix can be defined as,
q qI
being q the solid material density.
2:98
35
3
b du dX5 f b
2:99
and
2
dL3 d4
3
t du dC5 f t
2:100
2:101
2:102
2:103
K cc
K dc
K cd
K dd
uc
M cc
ud
M dc
M cd
M dd
uc
f
c
ud
fd
2:104
where uc are the unknown displacements and ud the known, or prescribed, displacements. The vectors f c and f d correspond respectively to the known applied
36
Fig. 2.6 Essential boundary condition nonaligned with the global axis
loads (external and body forces) and to the unknown reactions due the imposed
displacement constrains. With the Eq. (2.104) it is assumed that the displacement
components considered are axial aligned with the prescribed displacements. If this
is not the case it is required the identification of all prescribed displacement
orientations and transform locally the discrete equilibrium equations to correspond
to the global axis. Thus,
u Tu
2:105
where u is the vector of nodal point degrees of freedom in the required directions.
The transformation matrix T is defined by Eq. (2.106) and Fig. 2.6, which is a
typical representation of the constrained displacements in 2D and 3D analysis.
T 2D
u
x
uy
vx
vy
and
T 3D
ux
4 uy
uz
vx
vy
vz
3
wx
wy 5
wz
2:106
2:107
M T TM T
2:108
K T TK T
2:109
f T Tf
2:110
where,
Notice that the matrix multiplications in Eqs. (2.108), (2.109) and (2.110)
involve changes only in those columns and rows of M, K and f that are actually
37
2:111
2:112
being / the vector of order m, t the time variable, the constant initial time is
defined by t0 and x is the vibration frequency vector. Substituting Eqs. (2.112) into
(2.103) the generalized eigenproblem is obtained, from which / and x must be
determined,
K / x2 M /
2:113
K /1 x21 M /1
6 K /2 x2 M /2
2
6
4:
K /m x2m M /m
2:114
The vector / i is called the ith mode shape vector and xi is the corresponding
frequency of vibration. Defining a matrix U whose columns are the eigenvectors / i ,
U /1
/2
. . . /m
2:115
38
x21
6 0
6
X6 .
4 ..
0
x22
..
.
..
.
x2m
0
0
..
.
3
7
7
7
5
2:116
2:117
It is required that the space functions satisfy the following stiffness and mass
orthogonality conditions,
UT K U X
2:118
UT M U I
2:119
and
After substituting Eq. (2.111) and its time derivatives into Eq. (2.103) and premultiplying it by UT , the equilibrium equation that corresponds to the modal
generalized displacement is obtained,
xt X xt UT Ft
2:120
The initial conditions on xt are obtained using Eq. (2.111) and considering
the the M-orthonormality of UT at time t 0,
x0 UT Mu0
x_ 0 UT M u_ 0
2:121
xi t x2i xi t fi t
fi t / Ti Ft
2:122
/ Ti Mu0
xt0
i
t0
x_ i / Ti M u_ 0
2:123
For the complete response, the solution to all m equations in Eq. (2.122) must
be calculated and then the modal point displacements are obtained by superposition of the response in each mode.
39
ut
m
X
/ i xi t
2:124
i1
Therefore the response analysis requires, first, the solution of the eigenvalues
and eigenvectors of the problem, Eq. (2.113), then the solution of the decoupled
equilibrium equations in Eq. (2.122) and, finally, the superposition of the response
in each eigenvector as expressed in Eq. (2.124).
2:125
gB t 1 if
gB t 0 if
t ti
t [ ti
2:126
2:127
The solution of each equation in Eq. (2.123) can be calculated using the
Duhamel integral,
1
xi t
xi
Zt
2:128
where ai and bi are determined from the initial conditions: Eq. (2.123) and
fi t / Ti f t. For load case A and load case B the obtained solution is defined as,
xi t
fi t
x_ t0
1 cosxi t i sinxi t xt0
cosxi t
i
2
xi
xi
2:129
fi t
c
sin
c
t
sin
x
t
i
xi
x2i c2
2:130
40
References
1. Fung YC (1965) Foundations of solid mechanics. Englwood Cliffs, Prentice-Hall, New Jersey,
USA
2. Malvern LE (1969) Introduction of the Mechanics of a Continuous Medium. Englwood Cliffs,
Prentice-Hall, New Jersey, USA
3. Hodge PG (1970) Continuum mechanics. Mc Graw-Hill, New York
4. Lekhnitskii SG (1968) Anisotropic Plates. Gordon and Breach Science Publishers, New YorkLondon-Paris
5. Timoshenko S, Goodier JN (1970) Theory of Elasticity. 3rd ed. Singapore, McGraw Hill
6. Reddy JN (1986) Applied functional analysis and variational methods in engineering.
McGraw-Hill, Singapore
7. Bathe KJ (1996) Finite element procedures. Prentice-Hall, Englewood Cliffs
http://www.springer.com/978-3-319-06399-7