Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
Altair Engineering, Inc., World Headquarters: 1820 E. Big Beaver Rd., Troy MI 48083-2031 USA
Phone: +1.248.614.2400 • Fax: +1.248.614.2411 • www.altair.com • info@altair.com
RADIOSS THEORY Version 10.0 CONTENTS
CONTENTS
1.0 INTRODUCTION 6
1.1 INTRODUCTION 6
1.2 NOTATION 7
1.3 KINEMATICS OF PARTICLES 8
20-jan-2009 1
RADIOSS THEORY Version 10.0 CONTENTS
20-jan-2009 2
RADIOSS THEORY Version 10.0 CONTENTS
20-jan-2009 3
RADIOSS THEORY Version 10.0 CONTENTS
20-jan-2009 4
Chapter
INTRODUCTION
20-jan-2009 5
RADIOSS THEORY Version 10.0 INTRODUCTION
1.0 INTRODUCTION
1.1 Introduction
Nonlinear finite element analyses confront users with many choices. An understanding of the fundamental
concepts of non linear finite element analysis is necessary if the user does not want to use the finite element
program as a black box. The purpose of this manual is to describe the numerical methods included in RADIOSS.
RADIOSS belongs to the family of hydro-codes, in which the material is considered as a non viscous fluid.
These hydro-codes find their origin in the work supported by the American Department of Energy at the end of
the 70's and which lead to software like DYNA2D/3D, HEMP, PRONTO, STEALTH, HONDO and WHAM.
RADIOSS main features are:
• a 3D Lagrangian formulation for mesh description
• an explicit time integration scheme, leading to small time steps
• simplicity, under integrated finite element models
• element by element assembly of nodal forces leading to in memory codes and low I/O requirements
as compared to implicit approaches where matrix assembly and inversion is required every time step
• non-iterative approaches
• penalty methods based contact
• highly vectorized implementation.
This first chapter introduces the notations which will be used throughout the document. An introduction to
kinematics is also given.
The second chapter recalls the basic equations in non linear dynamics. Different aspects are covered:
• Material and spatial coordinates
• Mesh description
• Kinematic and kinetic descriptions
• Stress rates and stresses in solids
• Updated and total Lagrangian formulations
• Equations of equilibrium
• Principle of virtual power and the physical names of power terms.
The small strain formulation is also introduced.
The finite element formulation of the virtual power principle is introduced in chapter 3, leading to the discretized
equations of equilibrium.
Chapter 4 deals with time discretization and the integration schemes. Stability and time step concepts are also
discussed.
Different finite element models are presented in chapter 5. Tetrahedral solid elements, hexahedral solid and
solid-shell elements, 3 and 4-node shell elements, 2-node truss and beam elements and spring elements are
successively presented.
Chapter 6 deals with kinematic constraints, i.e. constraints placed on nodal velocities.
The DYREL and DAMP options are introduced in chapter 7.
20-jan-2009 6
RADIOSS THEORY Version 10.0 INTRODUCTION
The very important concept of interfaces is considered in chapter 8. Interfaces allow the solution of contact and
impact conditions between two parts of a model. The different interface types available in RADIOSS are
presented.
Material laws are discussed in chapter 9.
In chapter 10, the formulations of different kinds of monitored volumes are presented in detail. Airbag theory is
also developed.
Chapter 11 deals with the use of explicit algorithms to model quasi-static or static problems. Different
approaches are discussed: slow dynamic computation, dynamic relaxation, viscous relaxation and energy
discrete relaxation. The dynamic relaxation approach is developed.
Chapter 12 concerns the presentation of the fundamentals in RADIOSS parallelization.
In the ALE, CFD and SPH Theory Manual, the ALE formulation is presented in chapter 1.
Finally, chapters 2 and 3 are respectively dedicated to the Computational Aero-Acoustic and the Smooth Particle
Hydrodynamics formulations.
1.2 Notation
Two types of notation are used:
• Indicial notation: Equations of continuum mechanics are usually written in this form.
• Matrix notation: Used for equations pertinent to the finite element implementation.
To avoid confusion with nodal values, coordinates will be written as x, y or z rather than using subscripts.
Similarly, for a vector such as the velocity vi , numerical subscripts are avoided so as to avoid confusion with
node numbers. So, x1 = x, x2 = y , x3 = z and v1 = v x , v2 = v y and v3 = v2 .
Indices repeated twice in a list are summed. Indices which refer to components of tensors are always written in
lower case. Nodal indices are always indicated by upper case Latin letters. For instance, viI is the i-component
of the velocity vector at node I. Upper case indices repeated twice are summed over their range.
A second order tensor is indicated by two subscripts. For example, Eij is a second order tensor whose
components are E xx , E xy , ...
r 2 = xi ⋅ xi = x1 ⋅ x1 + x2 ⋅ x2 + x3 ⋅ x3 EQ. 1.2.2.1
r 2 = xT x EQ. 1.2.2.2
All vectors such as the velocity vector v will be denoted by lower case letters. Rectangular matrices will be
denoted by upper case letters.
20-jan-2009 7
RADIOSS THEORY Version 10.0 INTRODUCTION
The motion of a particle along a curved path in space is called space curvilinear motion. The position vector R,
the velocity v, and the acceleration of a particle along a curve are:
R = x î + y ĵ + z k̂ EQ. 1.3.1.1
where x, y and z are the coordinates of the particle and î , ĵ and k̂ the unit vectors in the rectangular
reference. In the cylindrical reference (r, θ , z), the description of space motion calls merely for the polar
coordinate expression:
v = v r + vθ + v z EQ. 1.3.1.4
where:
vr = r&
vθ = rθ&
vz = z&
Also, for acceleration:
a = a r + aθ + a z EQ. 1.3.1.5
where:
ar = &r& − rθ& 2
aθ = rθ&& + 2r&θ&
a z = &z&
The vector location of a particle may also be described by spherical coordinates as shown in Figure 1.3.1.
v = v R + vθ + vφ EQ. 1.3.1.6
where:
vR = R&
vθ = Rθ& cos φ
20-jan-2009 8
RADIOSS THEORY Version 10.0 INTRODUCTION
vφ = Rφ&
Using the previous expressions, the acceleration and its components can be computed:
a = a R + aθ + aφ EQ. 1.3.1.7
where:
&& − Rφ& 2 − Rθ& 2 cos 2 φ
aR = R
cos φ d 2 &
aθ =
R dt
( )
R θ − 2 Rθ&φ& sin φ
aφ =
R dt
( )
1 d 2&
R φ + Rθ& 2 sin φ cos φ
The choice of the coordinate system simplifies the measurement and the understanding of the problem.
Figure 1.3.1 Vector location of a particle in rectangular, cylindrical and spherical coordinates
θ
z φ
dvφ dvθ R
vφ vθ
dvR
vR
R
y
φ
θ
x
20-jan-2009 9
RADIOSS THEORY Version 10.0 INTRODUCTION
Direct transfer from rectangular to spherical coordinates may be accomplished by combining EQ. 1.3.2.1 and
EQ. 1.3.2.2:
{V } = [T ][T ]{V }
Rθφ φ θ xyz EQ. 1.3.2.3
Consider now the curvilinear motion of two particles A and B in space. Let’s study at first the translation of a
reference without rotation. The motion of A is observed from a translating frame of reference x-y-z moving with
the origin B (Figure 1.3.2). The position vector of A relative to B is:
rA / B = x î + y ˆj + z k̂ EQ. 1.3.3.1
where î , ĵ and k̂ are the unit vectors in the moving x-y-z system. As there is no change of unit vectors in
time, the velocity and the acceleration are derived as:
v A = v B + v A/B
a A = a B + a A/B
Z
z
rA
rA/B
k Y
y
O j
rB
B
X i x
20-jan-2009 10
RADIOSS THEORY Version 10.0 INTRODUCTION
In the case of rotation reference, it is proved that the angular velocity of the reference axes x-y-z may be
represented by the vector:
ω = ω x î + ω y ĵ + ω z k̂ EQ. 1.3.3.5
The time derivatives of the unit vectors î , ĵ and k̂ due to the rotation of reference axes x-y-z about ω , can be
studied by applying an infinitesimal rotation ωdt . We can write:
d
dt
()
î = ω × î ;
d
dt
()
ĵ = ω × ˆj ;
d
dt
()
k̂ = ω × k̂ EQ. 1.3.3.6
Attention should be turned to the meaning of the time derivatives of any vector quantity V=Vxi+ Vyj+ Vzk in the
rotating system. The derivative of V with respect to time as measured in the fixed frame X-Y-Z is:
⎛ dV ⎞
⎜ ⎟ =
d
⎝ dt ⎠ XYZ dt
(
Vx î + V y ĵ + Vz k̂ ) EQ. 1.3.3.7
⎛ d
= ⎜ Vx
⎝ dt
()
î + V y
d
dt
ĵ + Vz
d
()
dt
⎞
() (
k̂ ⎟ + V&x î + V&y ˆj + V&z k̂
⎠
)
With the substitution of EQ. 1.3.3.6, the terms in the first parentheses become ω × V . The terms in the second
⎛ dV ⎞
parentheses represent the components of time derivatives ⎜ ⎟ as measured relative to the moving x-y-z
⎝ dt ⎠ xyz
reference axes. Thus:
⎛ dV ⎞ ⎛ dV ⎞
⎜ ⎟ = ω ×V + ⎜ ⎟ EQ. 1.3.3.8
⎝ dt ⎠ XYZ ⎝ dt ⎠ xyz
This equation establishes the relation between the time derivative of a vector quantity in a fixed system and the
time derivative of the vector as observed in the rotating system.
Consider now the space motion of a particle A, as observed both from a rotating system x-y-z and a fixed system
X-Y-Z (Figure 1.3.3).
z
ω A y
j
r=rA/B
rA k Y
O rB B
i x
X
20-jan-2009 11
RADIOSS THEORY Version 10.0 INTRODUCTION
The origin of the rotating system coincides with the position of a second reference particle B, and the system has
an angular velocity ω . Standing r for rA / B , the time derivative of the vector position gives :
⎛ dr ⎞
r& = ⎜ ⎟ = ω × r + v rel EQ. 1.3.3.10
⎝ dt ⎠ XYZ
The relative acceleration equation is the time derivative of EQ. 1.3.3.12 which gives:
⎛ dv ⎞
v& rel = ⎜ rel ⎟ = ω × v rel + a rel EQ. 1.3.3.14
⎝ dt ⎠ XYZ
⎛ dv rel ⎞
and a rel = ⎜ ⎟ = &x& î + &y& ĵ + &z& k̂ EQ. 1.3.3.15
⎝ dt ⎠ xyz
20-jan-2009 12
Chapter
BASIC EQUATIONS
20-jan-2009 13
RADIOSS THEORY Version 10.0 BASIC EQUATIONS
The motion or deformation of a body can thus be described by a function ϕ (X , t ) where the material
coordinates X and the time t are considered as independent variables:
x = ϕ (X , t ) EQ. 2.1.0.1
The function ϕ gives the spatial positions of material points as a function of time.
The displacement of a material point is the difference between its original and final positions:
u( X , t ) = ϕ ( X , t ) − X EQ. 2.1.0.2
20-jan-2009 14
RADIOSS THEORY Version 10.0 BASIC EQUATIONS
∂
Dj = EQ. 2.3.0.2
∂X j
The transformation is fully described by the elements of the Jacobian matrix F:
dΩ = F dΩ 0 EQ. 2.3.0.5
Physically, the value of the Jacobian cannot take the zero value without cancelling the volume of a set of
material points. So the Jacobian must not become negative whatever the final configuration. This property
insures the existence and uniqueness of the inverse transformation:
At a regular point whereby definition of the field u ( X ) is differentiable, the vicinity transformation is defined
by:
20-jan-2009 15
RADIOSS THEORY Version 10.0 BASIC EQUATIONS
or in matrix form:
F=1+A EQ. 2.3.0.8
So, the Jacobian matrix F can be obtained from the matrix of gradients of displacements:
A ≡ D j ui EQ. 2.3.0.9
1 ⎛ ∂v ∂v j ⎞⎟
Dij = ⎜ i + EQ. 2.4.0.1
2 ⎜⎝ ∂x j ∂xi ⎟⎠
• the Green strain tensor (Lagrangian strain tensor) measured with respect to initial configuration
1 ⎛ ∂u ∂u j ∂u k ∂u k ⎞
Eij = ⎜ i + + ⎟ EQ. 2.4.0.2
2 ⎜⎝ ∂X j ∂X i ∂X i ∂X j ⎟
⎠
• the Almansi strain tensor (Eulerian strain tensor) measured with respect to deformed configuration
1 ⎛ ∂u ∂u j ∂uk ∂uk ⎞⎟
E Aij = ⎜ i + − EQ. 2.4.0.3
2 ⎜⎝ ∂x j ∂xi ∂xi ∂x j ⎟⎠
All these three measures of strains can be related to each other and can be used with any type of mesh.
dε ij 1 ⎛⎜ ∂vi ∂v j ⎞⎟
ε&ij = = Dij = + EQ. 2.4.1.1
dt 2 ⎜⎝ ∂x j ∂xi ⎟⎠
or in matrix form:
ε& = D =
1
2
(L + LT ) EQ. 2.4.1.2
where:
∂vi
Lij = EQ. 2.4.1.3
∂x j
is the velocity gradient in the current configuration.
The velocity of a material particle is:
∂xi
vi = EQ. 2.4.1.4
∂t
where the partial differentiation with respect to time t means the rate of change of the spatial position x of a
given particle. The velocity difference between two particles in the current configuration is given by:
20-jan-2009 16
RADIOSS THEORY Version 10.0 BASIC EQUATIONS
∂vi
dvi = dx j = Lij dx j = Lij F jk dX k EQ. 2.4.1.5
∂x j
In matrix form:
dv = Ldx = LFdX EQ. 2.4.1.6
On the other hand, it is possible to write the velocity difference directly as:
∂
dv = (FdX ) = F&dX EQ. 2.4.1.7
∂t
where:
∂F
F& = EQ. 2.4.1.8
∂t
One has as a result:
Since these are rate quantities, the spin can be treated as a vector. It is thus possible to decompose L into a
symmetric strain rate matrix and an anti symmetric rotation rate matrix just as in the small motion theory the
infinitesimal displacement gradient is decomposed into an infinitesimal strain and an infinitesimal rotation. The
symmetric part of the decomposition is the strain rate or the rate of deformation and is:
ε& = D =
2
(
1 & −1
FF + F −T F& T ) EQ. 2.4.1.11
Ω=
2
(FF − F −T F& T )
1 & −1
EQ. 2.4.1.12
The velocity-strain measures the current rate of deformation, but it gives no information about the total
deformation of the continuum. In general, EQ. 2.4.1.10 is not integrable analytically except in the
unidimensional case, where one obtains the true strain:
l and L are respectively the dimensions in the deformed and initial configurations. Futhermore, the integral in
time for a material point does not yield a well-defined, path-independent tensor so that information about
phenomena such as total stretching is not available in an algorithm that employs only the strain velocity.
Therefore, to obtain a measure of total deformation, the strain velocity has to be transformed to some other strain
rate.
The volumetric strain is calculated from density. For one dimensional deformation:
ρ − δΩ − δl
μ= −1 = = EQ. 2.4.1.14
ρ0 Ω l
20-jan-2009 17
RADIOSS THEORY Version 10.0 BASIC EQUATIONS
E=
2
(F F − 1)
1 T
EQ. 2.4.2.3
E=
1
2
(
A + AT + AT A ) EQ. 2.4.2.4
1 ⎛ ∂u ∂u j ∂uk ∂uk ⎞
Eij = ⎜ i + + ⎟ EQ. 2.4.2.5
⎜
2 ⎝ ∂X j ∂X i ∂X i ∂X j ⎟
⎠
In the unidimensional case, the value of the strain is:
E = (l 2 − L2 ) / (2 L2 ) EQ. 2.4.2.6
where l and L are respectively the dimensions in the deformed and initial configurations.
It can be shown that any motion F can always be represented as a pure rigid body rotation followed by a pure
stretch of three orthogonal directions:
F = RU = R(I + H) EQ. 2.4.2.7
with the rotation matrix R satisfying the orthogonality condition:
RT R = I EQ. 2.4.2.8
and H symmetric.
The polar decomposition theorem is important because it will allow to distinguish the straining part of the
motion from the rigid body rotation.
One has from EQS. 2.4.2.3 and 2.4.2.7:
E=H+
H2 1 T
2
= F F−I
2
( ) EQ. 2.4.2.9
R = F (I + H )
−1
EQ. 2.4.2.10
R T R = (I + Ω ) (I + Ω )
T
EQ. 2.4.2.12
ΩT + Ω = 0 EQ. 2.4.2.13
if second order terms are neglected.
As a result, one has for the Jacobian matrix:
leading, if the second order terms are neglected, to the classical linear relationships:
A=Ω+H EQ. 2.4.2.15
20-jan-2009 18
RADIOSS THEORY Version 10.0 BASIC EQUATIONS
H=
2
( A + A)
1 T
EQ. 2.4.2.16
Ω=
1
2
(
A − AT ) EQ. 2.4.2.17
So from EQ. 2.4.2.15. and 2.4.2.16., when rigid body rotations are large, the linear strain tensor becomes non-
zero even in the absence of deformation.
The preceding developments show that the linear strain measure is appropriate if rotations can be neglected; that
−2
means if they are of the same magnitude as the strains and if these are of the order of 10 or less. It is also
−1
worth noticing that linear strains can be used for moderately large strains of the order of 10 provided that the
rotations are small. On the other hand, for slender structures which are quite in extensible, non-linear kinematics
−2 −3
must be used even when the rotations are order of 10 because, if we are interested in strains of 10 − 10 −4 ,
using linear strain the error due to the rotations would be greater than the error due to the strains.
Large deformation problems in which non-linear kinematics is necessary, are those in which rigid body rotation
and deformation are large.
E=
2
(
1 T
F F−I ) EQ. 2.5.0.1
(F& T
) ( )
F + F T F S = F&F −1 + F −T F& T σ F EQ. 2.5.0.5
Second Piola-Kirchhoff stresses have a simple physical interpretation. They correspond to a decomposition of
forces in the frame coordinate systems convected by the deformation of the body. However, the stress measure is
still performed with respect to the initial surface.
20-jan-2009 19
RADIOSS THEORY Version 10.0 BASIC EQUATIONS
σ& ij is not simply the time derivative of the Cauchy stress tensor as Cauchy stress components are associated with
spatial directions in the current configuration. So, the derivatives will be nonzero in the case of a pure rigid body
rotation, even if from the constitutive point of view the material is unchanged. The stress rate is a function of
element average rigid body rotation and of strain rate.
For this reason, it is necessary to separate σ& ij in to two parts; one related to the rigid body motion and the
remainder associated with the rate form of the stress-strain law. Objective stress rate is used, meaning that the
stress tensor follows the rigid body rotation of the material [14].
A stress law will be objective if it is independent of the space frame. To each definition of the rigid body rotation
corresponds a definition of the objective stress rate. The Jaumann objective stress tensor derivative will be
associated with the rigid body rotation defined in EQ. 2.4.1.11:
where:
⎡σ 1 0 0 ⎤
R σR = ⎢⎢ 0 σ 2 0 ⎥⎥
T
EQ. 2.7.1.1
⎢⎣ 0 0 σ 3 ⎥⎦
The diagonal components are called the principal stresses and allow a 3D representation of the state of stress at a
point.
σ xx + σ yy + σ zz
p=− EQ. 2.7.2.1
3
20-jan-2009 20
RADIOSS THEORY Version 10.0 BASIC EQUATIONS
σ vm =
3
2
(
(σ xx + p )2 + (σ yy + p )2 + (σ zz + p )2 + 2σ 2 xy + 2σ 2 zy + 2σ 2 xz ) EQ. 2.7.2.2
called pressure and von Mises stress after Richard von Mises.
The values of these functions remain invariant under transformation by a proper orthogonal matrix.
If
σ = R tσ 0 R ,
then :
p = p0
σ vm = σ vm 0
The horizontal axis corresponds to the hydrostatic loading, the vertical axes to pure shear. The line with tangent
1/3 is uniaxial compression. The line with tangent -1/3 is uniaxial tension.
σ vm =
2
(
3 2
s xx + s 2 yy + s 2 zz + 2 s 2 xy + 2 s 2 yz + 2 s 2 xz ) EQ. 2.7.4.1
A surface of constant von Mises stress in deviatoric space or principal deviatoric space is a sphere (in stress
space it is a cylinder).
20-jan-2009 21
RADIOSS THEORY Version 10.0 BASIC EQUATIONS
z0
x0 xn xn+1
Initial undeformed configuration C0 Following configuration Cn+1
zn
CR Current
corotated
z0
xn
0
x Corotational configuration (CR)
Initial undeformed configuration C0
By default, RADIOSS uses a large strain, large displacement formulation with explicit time integration. The
large displacement formulation is obtained by computing the derivative of the shape functions at each cycle. The
20-jan-2009 22
RADIOSS THEORY Version 10.0 BASIC EQUATIONS
large strain formulation is derived from the incremental strain computation. Hence, stress and strains are true
stresses and true strains.
In the updated Lagrangian formulation, the Lagrangian coordinates are considered instantaneously coincident
with the Eulerian spatial x coordinates. This leads to the following simplifications:
∂xi ∂X j
= = δ ij EQ. 2.8.0.1
∂X j ∂xi
dΩ = dΩ 0 EQ. 2.8.0.2
The derivatives are with respect to the spatial (Eulerian) coordinates. The weak form involves integrals over the
deformed or current configuration. In the total Lagrangian formulation, the weak form involves integrals over the
initial (reference) configuration and derivatives are taken with respect to the material coordinates.
The corotational kinematic description is the most recent of the formulations in geometrically nonlinear
structural analysis. It decouples small strain material nonlinearities from geometric nonlinearities and handles
naturally the question of frame indifference of anisotropic behavior due to fabrication or material nonlinearities.
Several important works outline the various versions of CR formulation [7], [50], [51], [52], [53].
Some new generation of RADIOSS elements are based on this approach. The reader is invited to refer to
“Element Library” chapter for more details.
REMARK:
A similar approach to CR description using convected-coordinates is used in some branches of fluid mechanics
and rheology. However, the CR description maintains orthogonality of the moving frames. This allows to
achieve an exact decomposition of rigid body motion and deformational modes. On the other hand, convected
coordinates form a curvilinear system that fits the change of metric as the body deforms. The difference tends to
disappear as the mesh becomes finer. However, in general case the CR approach is more convenient in structural
mechanics.
∂vi
∫ τ dΓ + ∫ ρb dΩ = ∫ ρ
Γ
i
Ω
i
Ω
∂t
dΩ EQ. 2.9.0.1
τ i = n jσ ji EQ. 2.9.0.2
where n is the outward normal to Γ at that point. Using this definition, EQ 2.9.0.1 is written:
∂vi
∫n σ
Γ
j ji dΓ + ∫ ρbi dΩ = ∫ ρ
Ω Ω
∂t
dΩ EQ. 2.9.0.3
20-jan-2009 23
RADIOSS THEORY Version 10.0 BASIC EQUATIONS
Gauss' theorem allows the rewrite of the surface integral as a volume integral so that:
∂σ ij
∫n σ
Γ
j ji dΓ = ∫
Ω
∂x j
dΩ EQ. 2.9.0.4
As the volume is arbitrary, the expression can be applied at any point in the body providing the differential
equation of translation equilibrium:
∂σ ij ∂vi
+ ρbi = ρ EQ. 2.9.0.5
∂x j ∂t
Use of Gauss' theorem with this equation leads to the result that the true Cauchy stress matrix must be
symmetric:
⎛ ⎛ ∂σ ji ⎞⎞
⎜ δv ⎜ ⎟ ⎟dΩ = 0
∫Ω ⎜ ⎜ ∂x j
i + ρbi − ρv&i ⎟⎟
EQ. 2.10.0.1
⎝ ⎝ ⎠⎠
The first term in EQ 2.10.0.1 is then expanded:
⎛ ∂ ⎞
⎜
∫Ω ⎜ ∂x j ((δvi )σ ji )⎟ ∫ [ ]
⎟dΩ = (δvi )n jσ ji dΓ EQ. 2.10.0.3
⎝ ⎠ Γσ
taking into account that stresses vanish on the complement of the traction boundaries.
Replacing EQ 2.10.0.3. in EQ. 2.10.0.2 gives:
⎛ ∂σ ji ⎞ ∂ (δvi )
∫Ω ⎜ ∂x j ⎟⎟dΩ = ∫Γ (δvi )τ i dΓ − Ω∫ ∂x j σ ji dΩ
⎜ δvi EQ. 2.10.0.4
⎝ ⎠
If this last equation is then substituted in 2.10.0.1, one obtains:
⎛ ∂ (δvi ) ⎞
∫Ω ⎜⎜ ∂x j ⎟⎟σ ji dΩ − Ω∫ δvi ρbi dΩ − ∫Γ (δvi )τ i dΓ + Ω∫ δvi ρv&i dΩ = 0 EQ. 2.10.0.5
⎝ ⎠
The preceding expression is the weak form for the equilibrium equations, traction boundary conditions and
interior continuity conditions. It is known as the principle of virtual power.
20-jan-2009 24
RADIOSS THEORY Version 10.0 BASIC EQUATIONS
∂ (δvi )
δP int = ∫ δDijσ ji dΩ = ∫ σ ji dΩ ≡ ∫ δLijσ ji dΩ EQ. 2.11.0.2
Ω Ω
δx j Ω
The second and third terms in EQ. 2.10.0.5 are the virtual external power:
Inserting EQS 2.11.0.2., 2.11.0.3 and 2.11.0.4 into EQ. 2.10.0.5, the principle of virtual power can be written as :
We can show that virtual power principle implies strong equations of equilibrium. So it is possible to use the
virtual power principle with a suitable test function as a statement of equilibrium.
The virtual power principle has a simple physical interpretation. The rate of work done by the external forces
subjected to any virtual velocity field is equal to the rate of work done by the equilibrating stresses on the rate of
deformation of the same virtual velocity field. The principle is the weak form of the equilibrium equations and is
used as the basic equilibrium statement for the finite element formulation. Its advantage in this regard is that it
can be stated in the form of an integral over the volume of the body. It is possible to introduce approximations by
choosing test functions for the virtual velocity field whose variation is restricted to a few nodal values.
20-jan-2009 25
RADIOSS THEORY Version 10.0 BASIC EQUATIONS
negative volume, the computation will halt or the element will have to be removed. Using small strain can
eliminate these problems.
Using a small strain formulation for part of a large deformation process introduces of course errors. These errors
depend on the specific case, but there can provide a better solution than element deletion.
On the other side, materials like honeycomb, which have no Poisson's effect, can have the small strain
limitations corrected by using adjusted stress-strain curves.
A small strain, small displacement formulation can thus be specified for some specific material behavior, like
honeycomb, or can be implemented when the time step with a large strain formulation reaches a minimum value
that is defined by the user. This allows the computation to proceed at an acceptable rate.
The small displacement formulation is, however, not recommended for some simulations, e.g. crash analysis.
⎛ ∂Φ ⎞
ε&ij = ⎜⎜ I ⎟⎟ viI EQ. 2.12.1.2
⎝ ∂x j ⎠ t =0
with Φ I the interpolating shape functions and viI the components of velocity at node I.
⎛ Δδx ⎞ Δx
ε x = ∑ ε&x dt = ∑ ⎜⎜ ⎟⎟ = EQ. 2.12.1.3
⎝ x0 ⎠ x0
Thus, the strain is the engineering strain.
The stress is calculated using the strain rate and the material law provided by the user. The later is integrated
over the element volume to produce the internal force vector, which is summed over the elements to obtain the
overall force vector:
20-jan-2009 26
RADIOSS THEORY Version 10.0 BASIC EQUATIONS
⎛ ∂φ ⎞
f iIint = ∫ σ ij ⎜ I ⎟ dΩ EQ. 2.12.1.4
⎜ ∂x ⎟
Ω ⎝ j ⎠ t =0
The stress is the engineering stress.
The volumetric strain using the small strain formulation is independent of density. For one dimensional
deformation, one has:
δl
μ = −(ε xx + ε yy + ε zz ) = − EQ. 2.12.1.5
l0
The small strain formulation for solid elements was developed for specific material like honeycomb. In the
crushing direction, honeycomb has no Poisson's effect and stress integration over the initial surface is acceptable.
The effect on strain is small during elastic deformation and can be corrected in the plastic phase by using a
modified engineering stress-engineering strain material curve.
For materials like crushable foam, with a small Poisson's ratio, this formulation can be applied successfully in
certain situations. However, this formulation has to be used very carefully for other materials.
Shell elements have fewer limitations than solid elements. For crash applications, the main shell deformation is
bending. The small strain formulation has no effect on the bending description if membrane deformation is
small.
The small strain formulation can be applied to some elements for which the time step is reaching a user specified
value.
If the critical time step is small, compared to the initial one, this formulation gives acceptable results and is more
accurate than removing the deformed elements.
By default RADIOSS uses a large strain large displacement formulation with explicit time integration. By
computing the derivative of shape functions at each cycle, large displacement formulation is obtained. The large
strain formulation results from incremental strain computation. Stresses and strains are therefore true stresses and
true strains.
The spatial derivatives of isoparametric brick shape functions are given by:
∂Φ I −1 ∂Φ I
= F (t ) EQ. 2.12.2.1
∂x j ∂r
Where F(t) is the Jacobian matrix.
For each element the internal forces are integrated over the volume with one integration point:
⎛ ⎛ ∂Φ I ⎞ ∂Φ I ⎞⎟
= ∫ ⎜ σ ij ⎜⎜ ⎟⎟ dΩ = σ ij
int
f iI Ω EQ. 2.12.2.2
⎜ ⎝ ∂xi ⎠ t =0 ∂xi ⎟⎠
Ω⎝
dσ ij (t )
σ ij (t + δt ) = σ ij (t ) + dt EQ. 2.12.2.3
dt
uses objective stress rate, meaning that the stress tensor follows the rigid body rotation of the material. Stress
rate is a function of element average rigid body rotation and of strain rate. Strain rate is obtained from spatial
velocity derivative:
dε ij 1 ⎡ ∂vi ∂v j ⎤
= ⎢ + ⎥ EQ. 2.12.2.4
dt 2 ⎣⎢ ∂x j ∂xi ⎥⎦
20-jan-2009 27
RADIOSS THEORY Version 10.0 BASIC EQUATIONS
where:
∂vi ∂Φ I
= EQ. 2.12.2.5
∂x j ∂x j vi
Stability of explicit scheme is given by the Courant condition:
l
Δt < EQ. 2.12.2.6
c
with l the element characteristic length and c the sound speed.
The time step is computed at each cycle.
Large element deformation can give a large time step decrease. For overly large deformations a negative volume
can be reached and it then becomes impossible to invert the Jacobian matrix and to integrate the stresses over the
volume.
⎛ ρ ⎞ ΔV Δl
μ = ⎜⎜ − 1⎟⎟ = = EQ. 2.12.3.5
⎝ ρ0 ⎠ V l
For small strain we have:
Δl
μ = −(ε x + ε y + ε z ) = − EQ. 2.12.3.6
l0
20-jan-2009 28
Chapter
20-jan-2009 29
RADIOSS THEORY Version 10.0 FINITE ELEMENT FORMULATION
where Φ I ( X ) are the interpolating shape functions and xiI is the position vector of node I. Summation over
repeated indices is implied. In the case of lower indices, summation is over the number of space dimensions. For
upper case indices, summation is over the number of nodes. The nodes in the sum depend on the type of entity
considered. When the volume is considered, the summation is over all the nodes in the domain. When an element
is considered, the sum is over the nodes of the element.
Similarly, nodal displacements are defined using EQ. 2.1.0.2 at nodes:
The velocities are obtained by taking the material time derivative of the displacement giving:
∂ui ( X , t )
vi ( X , t ) = = Φ I ( X )viI (t ) EQ. 3.1.0.4
∂t
It is worth pointing out that the velocity is a material time derivative of displacements, i.e. the partial derivative
with respect to time with the material coordinate fixed.
Finally, accelerations are similarly given by the material time derivative of velocities:
Emphasis is placed on the fact that shapes functions are functions of the material coordinates whatever the
updated or the total Lagrangian formulation is used. All the dependency in the finite element approximation of
the motion is taken into account in the values of the nodal variables.
From EQ. 2.4.1.5, the velocity gradient is given by:
∂vi ∂Φ I
Lij = = viI = viI Φ I , j EQ. 3.1.0.6
∂x j ∂x j
and the rate of deformation (EQ. 2.4.1.1) by:
Dij =
1
(Lij + L ji ) = 1 (viI Φ I , j + v jI Φ I ,i ) EQ. 3.1.0.7
2 2
Similarly, the test functions are approximated as:
The test functions are next substituted into the principle of virtual power (EQ. 2.10.0.5) giving:
∂Φ I
δviI ∫ σ ji dΩ − δviI ∫ Φ I ρbi dΩ − δviI ∫ Φ Iτ&i dΓ + δviI ∫ Φ I ρv&i dΩ = 0 EQ. 3.1.0.9
Ω
∂x j Ω Γσ Ω
20-jan-2009 30
RADIOSS THEORY Version 10.0 FINITE ELEMENT FORMULATION
The virtual velocities must be kinematically admissible, i.e. satisfy boundary conditions on Γu , the part of the
boundary where kinematical conditions are specified. Using the arbitrariness of the virtual nodal velocities
everywhere except on Γu , the weak form of the momentum equation is:
∂Φ I
∫ ∂x
Ω j
σ ji dΩ − ∫ Φ I ρbi dΩ − ∫ Φ Iτ&i dΓ + ∫ Φ I ρv&i dΩ = 0
Ω Γσ Ω
EQ. 3.1.0.10
with Γσ the part of the boundary where traction loads are imposed.
⎛ ∂Φ I ⎞
= ∫ σ ji ⎜ ⎟dΩ
int
f iI EQ. 3.2.0.2
⎜ ∂x ⎟
Ω ⎝ j ⎠
These nodal forces are called internal because they represent the stresses in the body. The expression applies to
both the complete mesh or to any element. It is pointed out that derivatives are taken with respect to spatial
coordinates and that integration is taken over the current deformed configuration.
The external forces are similarly defined in terms of the virtual external power:
= ∫ Φ I ρbi dΩ + ∫ Φ Iτ&i dΓ
ext
f iI EQ. 3.2.0.4
Ω Γσ
= ∫ Φ I ρv&i dΩ
inert
f iI EQ. 3.3.0.2
Ω
= ∫ ρΦ I Φ J dΩv&iJ
inert
f iI EQ. 3.3.0.3
Ω
It is usual to define the inertial nodal forces as the product of a mass matrix and the nodal accelerations. Defining
the mass matrix as:
20-jan-2009 31
RADIOSS THEORY Version 10.0 FINITE ELEMENT FORMULATION
(
δviI M ijIJ v& jJ + f iI int − f iI ext = 0 ) EQ. 3.4.0.6
l1α1 + (I 3 − I 2 )ω2ω3 = m1 − m1
ext int
EQ. 3.6.0.1
20-jan-2009 32
RADIOSS THEORY Version 10.0 FINITE ELEMENT FORMULATION
l3α 3 + (I 2 − I1 )ω1ω2 = m3
ext int
− m3 EQ. 3.6.0.3
where
• I1 , I 2 , I 3 are the principal moments of inertia about the x, y and z axes respectively,
• α 1, α 2 , α 3 are the angular accelerations expressed in the principal reference frame,
• ω 1, ω 2 , ω 3 are the angular velocities,
• m1ext , m2ext , m3ext are the principal externally applied moments,
• I= ∑I
elements
e is the diagonal inertia matrix,
• M ext = ∑m
elements
ext
is the externally applied moment vector,
• M int = ∑m
elements
int
is the internal moment vector,
• M hgr = ∑m
shells
hgr
is the anti-hourglass shell moment vector.
20-jan-2009 33
RADIOSS THEORY Version 10.0 FINITE ELEMENT FORMULATION
The motion in each element can thus be described by the composition of three maps (the reasoning is described
only for the direction u):
• the map from the intrinsic coordinates system to the initial configuration:
X (ξ ) EQ. 3.7.0.1
• the map from the intrinsic coordinates system to the current configuration:
x = ϕ (X , t ) EQ. 3.7.0.3
Shape functions Φ I (ξ ) have no dimensions. They simply relate coordinates in the physical world to the
intrinsic coordinates system. Writing EQ. 3.7.0.4 at t=0 , we obtain:
So, it can be seen from the last equation that the material coordinates system and the intrinsic coordinates system
are invariant in a Lagrangian element. As a result, as intrinsic coordinates are time invariant and it is possible to
write displacements, velocities and accelerations in terms of intrinsic coordinates (one coordinate system, the
two other coordinates have similar shape functions):
Isoparametric elements use the same shape functions for the interpolation of x, u , u& and v& .
∫ g (x )dΩ = ∫ g (x ) F dΩ
Ω
0 = ∫ g (ξ ) Fξ dΔ
Δ
EQ. 3.8.0.1
Ω0
and
∫ g ( X )dΩ = ∫ g (ξ ) Fξ dΔ
0
0 EQ. 3.8.0.2
Ω0 Δ
where:
F EQ. 3.8.0.3
20-jan-2009 34
RADIOSS THEORY Version 10.0 FINITE ELEMENT FORMULATION
is the Jacobian determinant of the transformation between the current and the initial configuration,
Fξ EQ. 3.8.0.4
is the Jacobian determinant of the transformation between the current configuration and the domain in the
intrinsic coordinate system and,
0
Fξ EQ. 3.8.0.5
is the Jacobian determinant of the transformation between the reference configuration and the intrinsic
coordinate system.
On the other hand, it comes from EQS 2.3.0.2. and 3.7.0.4:
∂xk ∂Φ I (ξ )
Fξkj = = xkl EQ. 3.8.0.6
∂ξ j ∂ξ j
So, using EQ. 3.8.0.1, internal forces computed by integration over the current domain will be obtained by the
following quadrature:
∂Φ I ∂Φ I
f iIint = ∫ σ ij dΩ = ∫ σ ij Fξ dΔ EQ. 3.8.0.7
Ω
∂x j Δ
∂x j
External forces and the mass matrix can similarly be integrated over the domain in the intrinsic coordinate
system.
20-jan-2009 35
RADIOSS THEORY Version 10.0 FINITE ELEMENT FORMULATION
f (ξ )dξ = ∑ w j f (ξ j )
n
∫ j =1
EQ. 3.10.0.1
where n is the number of integration points in the element and w j is the weight associated to the integration
point j. Values of w j and locations of ξ j are given in tables according to the numerical quadrature approach.
RADIOSS uses either full or reduced integration schemes.
For full integration, the number of integration points is sufficient for the exact integration of the virtual work
expression. The full integration scheme is often used in programs for static or dynamic problems with implicit
time integration. It presents no problem for stability, but sometimes involves "locking" and the computation is
often expensive.
Reduced integration can also be used. In this case, the number of integration points is sufficient for the exact
integration of the contributions of the strain field that are one order less than the order of the shape functions.
The incomplete higher order contributions to the strain field present in these elements are not integrated.
The reduced integration scheme, especially with one-point quadrature is widely used in programs with explicit
time integration to compute the force vectors. It drastically decreases the computation time, and is very
competitive if the spurious singular modes (often called “hourglass modes” which result from the reduced
integration scheme) are properly stabilized. In two dimensions, a one point integration scheme will be almost
four times less expensive than a four point integration scheme. The savings are even greater in three dimensions.
The use of one integration point is recommended to save CPU time, but also to avoid "locking" problems, e.g.
shear locking or volume locking.
Shear locking is related to bending behavior. In the stress analysis of relatively thin members subjected to
bending, the strain variation through the thickness must be at least linear, so constant strain first order elements
are not well suited to represent this variation, leading to shear locking. Fully integrated first-order isoparametric
elements (tetrahedron) also suffer from shear locking in the geometries where they cannot provide the pure
bending solution because they must shear at the numerical integration points to represent the bending kinematic
behavior. This shearing then locks the element, i.e. the response is far too stiff.
On the other hand, most fully integrated solid elements are unsuitable for the analysis of approximately
incompressible material behavior (volume locking). The reason for this is that the material behavior forces the
material to deform approximately without volume changes. Fully integrated solid elements, and in particular
low-order elements do not allow such deformations. This is another reason for using selectively reduced
integration. Reduced integration is used for volume strain and full integration is used for the deviatoric strains.
However, as mentioned above, the disadvantage of reduced integration is that the element can admit deformation
modes that are not causing stresses at the integration points. These zero-energy modes make the element rank-
deficient and cause a phenomenon called hour-glassing: the zero-energy modes start propagating through the
mesh, leading to inaccurate solutions. This problem is particularly severe in first-order quadrilaterals and
hexahedra.
To prevent these excessive deformations, a small artificial stiffness or viscosity associated with the zero-energy
deformation modes is added, leading in EQS 3.5.0.1. and. 3.6.0.4.to anti-hourglass force and moment vectors:
dv
M = f ext − f int + f hgr EQ. 3.10.0.2
dt
dω
I = M ext − M int + M hgr EQ. 3.10.0.3
dt
Zero energy or hourglass modes are controlled using a perturbation stabilization as described by Flanagan-
Belytschko [12], or physical stabilization as described in [15] (chapter 5).
20-jan-2009 36
RADIOSS THEORY Version 10.0 FINITE ELEMENT FORMULATION
So, for isoparametric elements, reduced integration allows simple and cost effective computation of the volume
integrals, in particular on vectorized supercomputers, and furnishes a simple way to cope with locking aspects,
but at the cost of allowing hourglassing.
1. For the displacement, velocity and acceleration at a particular time step, the external force vector is
constructed and applied.
2. A loop over elements is performed, in which the internal and hourglass forces are computed, along with
the size of the next time step. The procedure for this loop is:
2a. The Jacobian matrix is used to relate displacements in the intrinsic coordinates system to the physical
space:
∂Φ ∂Φ
= Fξ−1 EQ. 3.11.0.4
∂x j ∂ξ t
t
⎛ ∂Φ ⎞ 1 ∂v ⎛ ∂v ⎞
ε&ij = ⎜⎜ I ⎟⎟ x& = ⎜⎜ i + j ⎟⎟ EQ. 3.11.0.5
⎝ ∂x j ⎠ 2 ⎝ ∂x j ∂xi ⎠
2c. The stress rate is calculated:
3. After the internal and hourglass forces are calculated for each element, the algorithm proceeds by
computing the contact forces between any interfaces.
4. With all forces known, the new accelerations are calculated using the mass matrix and the external and
internal force vectors:
(
v&i = M −1 f exti − f int i ) EQ. 3.11.0.8
5. Finally, time integration of velocity and displacement is performed using the new value.
20-jan-2009 37
RADIOSS THEORY Version 10.0 FINITE ELEMENT FORMULATION
1
Figure 3.11.1 Numerical Procedure
t = t + Δt
ext
Compute f
Loop over
1 ⎛ ∂v ∂v ⎞
ε&ij = ⎜⎜ i + j ⎟⎟
2 ⎝ ∂x j ∂xi ⎠
Compute f cont
v&i = ∑ Fi / mi
Time integration
20-jan-2009 38
Chapter
DYNAMIC ANALYSIS
20-jan-2009 39
RADIOSS THEORY Version 10.0 DYNAMIC ANALYSIS
h 2 (2 ) h s (s )
f (t n + h ) = f (t n ) + hf ′(t n ) + f (t n ) + ... + f (t n ) + Rs EQ. 4.1.1.1
2 s!
t +h
1 n ( s +1)
Rs = ∫ f (τ )[t n + h − τ ]s dτ EQ. 4.1.1.2
s! tn
The preceding formula allows the computation of displacements and velocities of the system at time t n + 1 :
t n +1
20-jan-2009 40
RADIOSS THEORY Version 10.0 DYNAMIC ANALYSIS
t n+1
un +1 = un + hu&n + ∫ (t
tn
n +1 − τ )u&&(τ )dτ EQ. 4.1.1.4
The approximation consists in computing the integrals for acceleration in 4.1.1.3 and in 4.1.1.4 by numerical
quadrature:
t n +1
t n +1
⎛1 ⎞
∫ (t
tn
n +1 − τ )u&&(τ )dτ = ⎜ − β ⎟h 2u&&n + βh 2u&&n+1 + rn′
⎝2 ⎠
EQ. 4.1.1.6
⎛1 ⎞
un +1 = un + hu&n + ⎜ − β ⎟h 2u&&n + βh 2u&&n +1 + rn′ EQ. 4.1.1.8
⎝ 2 ⎠
1
u&n +1 = u&n + hn +1 (u&&n + u&&n +1 ) EQ. 4.1.2.1
2
1 2
u n +1 = u n + hn +1u& n + hn +1 + u&&n EQ. 4.1.2.2
2
20-jan-2009 41
RADIOSS THEORY Version 10.0 DYNAMIC ANALYSIS
un +1 − 2un + un −1
u&&n = EQ. 4.1.2.3
h2
if the time step is constant.
From the algorithmic point of view, it is, however, more efficient to use velocities at half of the time step:
⎛ ⎞ 1
u& 1 = u&⎜⎜ t 1 ⎟⎟ = (un+1 − un ) EQ. 4.1.2.4
n+
2 ⎝ 2 ⎠ hn +1
n+
so that:
1 ⎛ ⎞
u&&n = ⎜ u& 1 − u& 1 ⎟ EQ. 4.1.2.5
h 1 ⎜⎝ n + 2 n− ⎟
2 ⎠
n+
2
2
The accuracy of the scheme is of h order, i.e. if the time step is halved, the amount of error in the calculation is
one quarter of the original. The time step h may be variable from one cycle to another. It is recalculated after
internal forces have been computed.
h1
h0 = 0 ; h1 = EQ. 4.1.3.1
2 2
20-jan-2009 42
RADIOSS THEORY Version 10.0 DYNAMIC ANALYSIS
20-jan-2009 43
RADIOSS THEORY Version 10.0 DYNAMIC ANALYSIS
h
m
x& = 2 gh
k x
In this equation the damping effects are neglected to simplify the solution. The general solution of the
differential equation is written as :
k
ω is the natural frequency of the system : ω=
m
Introducing these initial solutions into 4.1.5.13, we obtain the following result:
2 gh g
x= Sinωt + (1 − Cosωt ) EQ. 4.1.5.4
ω ω2
The same problem can be resolved by the numerical procedure explained in this chapter. Considering at first the
following numerical values for the mass, the stiffness, the dropping height and the gravity:
From EQ. 4.1.5.12, the dynamic equilibrium equation or equation of motion is obtained as :
Using a step-by-step time discretization method with a central difference algorithm, for a given known step tn the
unknown kinematic variables for the next step are given by EQ. 4.1.5.6, EQ. 4.1.2.4 and EQ. 4.1.2.5:
20-jan-2009 44
RADIOSS THEORY Version 10.0 DYNAMIC ANALYSIS
&x&n = 10 − 20 xn
x& n+1 = &x&n Δt + x& n EQ. 4.1.5.7
xn+1 = x& n Δt + xn
For the first time step the initial conditions are defined by EQ. 4.1.5.3. Using a constant time step Δt = 0.1 the
mass motion can be computed. It is compared to the analytical solution given by EQ. 4.1.5.4 in Figure 4.1.3.
The difference between the two results shows the time discretization error.
1,5
1
Displacement
Analytical
0,5
Numerical
0
0 0,5 1 1,5 2 2,5
-0,5
-1
Time
On the other hand, the numerical stability of time integrators discussed in the literature concerns generally linear
systems and extrapolated to nonlinear cases by examining linearized models of nonlinear systems. The
philosophy is the following: if a numerical method is unstable for a linear system, it will be certainly unstable
for nonlinear systems as linear cases are subsets of the nonlinear cases. Therefore, the stability of numerical
procedures for linear systems provides a useful guide to explore their behavior in a general nonlinear case.
To study the stability of the central difference time integration scheme, we establish the necessary conditions to
ensure that the solution of equations are not amplified artificially during the step-by-step procedure. Stability
also means that the errors due to round-off in the computer, do not grow in the integration. It is assured if the
time step is small enough to integrate accurately the response in the highest frequency component.
20-jan-2009 45
RADIOSS THEORY Version 10.0 DYNAMIC ANALYSIS
1
u& n +1 = u& n +
Δt n
2
( ) n+ Δt
&u& + &u& n +1 = u& 2 + &u& n +1
2 EQ. 4.1.7.1
1
n +1 n Δt 2 n
n
n+
u = u + Δt.u& + .&u& = u n + Δt.u& 2
2
For stability studies, we aim to establish a recursive relationship to link the displacements at three consecutive
time steps:
⎡u n+1 ⎤ ⎡ un ⎤
⎢ n ⎥ = [A ]⎢ n−1 ⎥ + [L] EQ. 4.1.7.2
⎣u ⎦ ⎣u ⎦
[ ]
where A is called amplification matrix. A spectral analysis of this matrix highlights the stability of the
integration scheme:
[ ]
The numerical algorithm is stable if and only if the radius spectral of A is less than unity. In the other words
[ ]
when the module of all eigen values of A are smaller than unity the numerical stability is ensured.
[ ]
The stability of a numerical scheme can be studied using the general form of the 2x2 matrix A :
[A] = ⎡⎢
A11 A12 ⎤
EQ. 4.1.7.3
⎣A 21 A 22 ⎥⎦
Then, the equations are developed for the systems with or without damping [55].
[ ]
The eigen values of A are computed from the characteristic polynomial equation:
λ2 − 2A1λ + A 2 = 0
1 1
where A1 = tr[A ] = (A11 + A 22 ) ,
2 2
A2 = det[A ] = A11 A22 − A12 A21 .
The eigen values are then obtained as :
If A12 < A 2 , eigen values are complex conjugate; if A12 = A 2 , they are real and identical; if A12 > A 2 , they
are real and distinct. We intend to define a stability domain in the (A1 , A 2 ) -space, where the spectral radius
20-jan-2009 46
RADIOSS THEORY Version 10.0 DYNAMIC ANALYSIS
We then have:
1 − 2A 1 + A 2 = 0 EQ. 4.1.7.6
This yields:
A 2 = 2A1 − 1
λ1 = 1 EQ.4.1.7.7
λ2 = A2
The corresponding part of the boundary of the stability domain is the segment analytically defined by
1 − 2A1 + A 2 = 0 and − 1 ≤ A 2 ≤ 1 .
We then have:
1 + 2A1 + A 2 = 0 EQ.4.1.7.8
This yields:
A 2 = −2A1 − 1
λ1 = −1 EQ.4.1.7.9
λ 2 = −A 2
In this case, the corresponding part of the boundary is the segment given by 1 + 2A1 + A 2 = 0 and
−1 ≤ A2 ≤ 1.
0 = e 2iα − 2 A1e iα + A2
= (cos 2α − 2 A1 cos α + A2 ) + i(sin 2α − 2 A1 sin α ) EQ.4.1.7.10
= (2 cos α (cos α − A1 ) + A2 − 1) + i (2 sin α (cosα − A1 ))
This yields:
2 cos α (cos α − A1 ) + A2 − 1 = 0
EQ.4.1.7.11
2 sin α (cos α − A1 ) = 0
20-jan-2009 47
RADIOSS THEORY Version 10.0 DYNAMIC ANALYSIS
The 3 segments introduced above define a closed contour. Point A1 = A 2 = 0 is located inside this contour et
in this case, ρ( [A ]) = 0 . Since ρ( [A ]) varies continuously with respect to A1 and A2, we can conclude that the
stability domain corresponds to the interior of the contour. To precisely define the stability domain, we must also
have points leading to double eigen value of modulus 1, i.e. the intersections between the parabola A 1 = A 2
2
and the boundary of the domain. This corresponds to Points (A1 , A 2 ) = (± 1,1) .
A
A2 = 1
STABILITY
Complex conjugate
eigenvalues
A
Real
A 12 = A 2 eigenvalues
Double eigenvalue
1 − 2A 1 + A 2 = 0
1 + 2A 1 + A 2 = 0
We can analytically summarize the description of the stability by means of the two following sets of conditions:
(1) −
(A 2 + 1) ≤ A ≤
(A 2 + 1) , −1 ≤ A2 < 1
1
2 2 EQ.4.1.7.13
(2) − 1 < A1 < 1 , A 2 = 1
mu&&n + ku n = f n EQ.4.1.7.14
n
where m and k are respectively the nodal mass and stiffness. f is the external force at time tn. Rewriting the
central difference time integration equations from EQ. 4.1.7.1, we obtain :
1 1
n+ n−
u n +1 = u n + Δt.u& 2 = u n + Δt.u& 2 + Δt 2 .&u& n
EQ.4.1.7.15
1
n−
n n −1
u =u + Δt.u& 2
20-jan-2009 48
RADIOSS THEORY Version 10.0 DYNAMIC ANALYSIS
and:
u n+1 − 2u n + u n−1
u&&n = EQ.4.1.7.16
Δt 2
Substituting these equations into EQ.4.1.7.14, it yields:
u n+1 − 2u n + u n −1
m + ku n = f n EQ.4.1.7.17
Δt 2
This equation can be written as EQ. 4.1.7.2. Then the amplification matrix takes the expression:
⎡ − 1⎤
[A] = ⎢2 − ω Δt 2
2
⎥ EQ.4.1.7.18
⎣ 1 0⎦
k
where ω= is the pulsation of the considered mode.
m
ω 2 Δt 2
Comparing with EQ. 4.1.7.3, we have, A1 = 1 − and A 2 = 1 . Stability is then given by:
2
ω 2 Δt 2
−1 < 1− <1 EQ.4.1.7.19
2
The right inequality is always true if ω ≠0. For, the particular case of ω =0, the scheme is unstable. However,
the analytical solution for a system with ω =0 leads to an unbounded solution. The left inequality implies:
2
Δt < EQ.4.1.7.20
ω
1
n+
u n +1 = u n + Δt.u& 2
EQ.4.1.7.22
1
n−
u n = u n −1 + Δt.u& 2
we deduce:
20-jan-2009 49
RADIOSS THEORY Version 10.0 DYNAMIC ANALYSIS
⎛ n+ 12 n− ⎞
1
u n +1
−u n −1
⎜
= Δt ⎜ u + u 2 ⎟⎟
& & EQ.4.1.7.23
⎝ ⎠
Substituting these equations into EQ.4.1.7.21, the recurring continuation equation on the displacement is written
in the form:
u n+1 − 2u n + u n−1
m
Δt 2
+
c
2Δt
( )
u n +1 − u n −1 + ku n = f n EQ.4.1.7.26
The equation can be rearranged to obtain the expression of the amplification matrix:
⎡ cΔt ⎤
⎢ 2 − ω 2 Δt 2 −1+
2m ⎥
⎢ cΔt ⎥
[A] = ⎢ 1 + cΔt 1+ ⎥ EQ.4.1.7.27
⎢ 2m 2m ⎥
⎢ ⎥
⎢ 1 0 ⎥
⎣ ⎦
ω 2 Δt 2 cΔt
1− 1−
This yields A1 = 2 and A 2 = 2m . Stability is given by the set of conditions from EQ.4.1.7.13:
cΔt cΔt
1+ 1+
2m 2m
ω 2 Δt 2
1−
1 2 ≤ 1
− ≤ EQ.4.1.7.28
cΔt cΔt cΔt
1+ 1+ 1+
2m 2m 2m
cΔt
1−
−1 ≤ 2m < 1
cΔt
1+
2m
The second expression is always verified for c > 0. It is the same for the right inequality of the first expression.
The left inequality of the first expression leads to the condition on the time step:
20-jan-2009 50
RADIOSS THEORY Version 10.0 DYNAMIC ANALYSIS
2
Δt ≤ EQ.4.1.7.29
ω
We find the same condition as in the undamped case, which echoes a conclusion given in [55]. We may yet
remark that damping has changed the strict inequality into a large inequality, preventing from weak instability
due to a double eigen value of modulus unity.
It is important to note that the relation EQ.4.1.7.29 is obtained by using the expression EQ.4.1.7.25 to compute
1 1
n+ n−
nodal velocities at time steps. However, in an explicit scheme generally the mid-step velocities u& 2
and u& 2
integration algorithm requires the matrix [M ] + Δt [C] to be inverted, which can harm the numerical
2
performances. We therefore often compute the viscous forces using the velocities at the preceding mid-step,
which are explicit. This leads to an equilibrium at step n in the form:
1
n−
mu&&n + c.u& 2
+ ku n = f n EQ.4.1.7.30
u&
n−
2
=
1 n
Δt
(
u − u n −1 ) EQ.4.1.7.31
u n+1 − 2u n + u n−1 c n
m
Δt 2
+
Δt
(
u − u n−1 + ku n = f n ) EQ.4.1.7.32
⎡ cΔt cΔt ⎤
⎢2 − ω Δt − m −1+
2 2
[A] = ⎢ m ⎥
⎥ EQ.4.1.7.33
⎢ ⎥
⎢⎣ 1 0 ⎥⎦
ω 2 Δt 2 cΔt cΔt
We have in this case A1 = 1 − − and A 2 = 1 − . Stability is again given by the set of
2 2m m
conditions EQ.4.1.7.13:
20-jan-2009 51
RADIOSS THEORY Version 10.0 DYNAMIC ANALYSIS
cΔt
−1 ≤ 1− <1
m
Right inequalities are always verified in both preceding expressions. Left inequalities now lead to two conditions
on the time step:
c c2
− + + 4ω 2
m m2 2m
Δt ≤ , Δt ≤ EQ.4.1.7.35
ω 2
c
Therefore, the critical time step depends not only to ω but also to the mass and the damping . However, the
critical time step depends only to ω when using the exact velocities to compute the viscous forces as described
in the previous section.
In the case of a direct step-by-step time integration, it is necessary to evaluate the damping matrix [C] explicitly.
The Rayleigh damping method assumes that the matrix [C] is computed by the following equation:
⎧ 1
⎫
[M ]{u&&n }+ α [M ]{u& n }+ β [K ]⎨u& { } { }
⎬ + [K ] u = f
n−
2 n n
EQ.4.1.7.38
⎩ ⎭
Studying the equilibrium of a node to obtain a one dimensional equation of motion, we can write:
20-jan-2009 52
RADIOSS THEORY Version 10.0 DYNAMIC ANALYSIS
ω 2 Δt 2 βω 2 Δt αΔt
1− − 1− − βω 2 Δt
In this case, A1 = 2 2 et A 2 = 2 . Stability is obtained as before by means
αΔt αΔt
1+ 1+
2 2
of the set of conditions from EQ.4.1.7.13:
βω 2 Δt ω 2 Δt 2 βω 2 Δt βω 2 Δt
−1+ 1− − 1−
2 ≤ 2 2 ≤ 2
αΔt αΔt αΔt
1+ 1+ 1+
2 2 2 EQ.4.1.7.42
αΔt
1− − βω Δt 2
−1 ≤ 2 <1
αΔt
1+
2
This yields again two conditions on the time step, coming from the left inequalities in both expressions:
− βω + β 2ω 2 + 4 2
Δt ≤ , Δt ≤ EQ.4.1.7.43
ω βω 2
It is equivalent to consider only the β[K ] contribution in the damping for the computation of the time step,
which appears to be logical since the α[M ] contribution is used with the exact velocity. It is advantageous to
separate the two contributions, restrictions of the time step then becoming lighter. It can be shown that for the
complete treatment of the Rayleigh damping using mid-step velocities, the stability conditions can be given by
the following expressions:
Δt ≤
− α − βω 2 + (α + βω ) 2 2
+ 4ω 2
, Δt ≤
2
EQ.4.1.7.44
ω α + βω 2
20-jan-2009 53
RADIOSS THEORY Version 10.0 DYNAMIC ANALYSIS
MX&& + KX = 0 (a)
1 − 1⎤
[M ] = m ⎡⎢
1 0⎤
⎥; [K ] = k ⎡⎢ ⎥
⎣0 1 ⎦ ⎣− 1 1 ⎦
where the angular frequency ω and the phase angle α are common for all Xi. α and Ψi are the constants of
integration to be determined from the initial conditions of the motion and ω is a characteristic value (eigen
value) of the system. Substituting (b) into (a) yields:
(K − ω M )ΨCos(ωt − α ) = 0
2
(c)
det (K − ω 2 M ) = 0
2k
=> ω2 = (d)
m
2k
Assuming the following numerical values m = 1 and k =10 , we have ω= = 4.472136 .The critical
m
time step of the system is given by the EQ.4.1.7.20:
2 2
Δt ≤ ⇒ Δt ≤ ⇒ Δt ≤ 0.4472
ω 4.472136
2
Δt cr = m k
ω max
k 2
ω= = 20 ⇒ Δt cr = ⇒ Δt cr = 0.4472
m 4.472136
20-jan-2009 54
RADIOSS THEORY Version 10.0 DYNAMIC ANALYSIS
m k k m
20-jan-2009 55
RADIOSS THEORY Version 10.0 DYNAMIC ANALYSIS
where
ωmax is the highest natural frequency in the system:
where K and M are respectively the stiffness and the mass matrices of the system.
The time step restriction given by EQ. 4.1.8.1 was derived considering a linear system (see section 4.1.7), but the
result is also applicable to nonlinear analysis since on a given step the resolution is linear. However, in nonlinear
analysis the stiffness properties change during the response calculation. These changes in the material and the
l
Δt ≤ c EQ.4.1.8.3
c
where:
Figure 4.1.6 Bar element
lc is the characteristic element length, representing the
shortest road for a wave arriving on a node to cross the
element,
1 − 1⎤
[K ] = k ⎡⎢ ⎥ EQ. 4.1.8.6
⎣− 1 1 ⎦
20-jan-2009 56
RADIOSS THEORY Version 10.0 DYNAMIC ANALYSIS
[M ] = m ⎡⎢
1 0⎤
⎥ EQ. 4.1.8.7
⎣0 1 ⎦
where m and k are respectively the nodal mass and stiffness of the bar:
EA Aρl
k= ; m= EQ. 4.1.8.8
l 2
Equation 4.1.8.5 yields:
(k − mω ) 2 2
−k2 = 0 EQ. 4.1.8.9
then :
2k
ω= EQ. 4.1.8.10
m
which can be simplified with EQ. 4.1.8.8 to obtain :
2 E 2c
ω= = EQ.4.1.8.11
l ρ l
where c is the speed of sound in the material and its expression is given as :
E
c=
ρ EQ. 4.1.8.12
with ρ the material density and E the Young’s modulus. Combining EQ.4.1.8.11 and EQ. 4.1.8.1, we obtain:
l
Δt ≤
c EQ. 4.1.8.13
20-jan-2009 57
RADIOSS THEORY Version 10.0 DYNAMIC ANALYSIS
This relation is that of EQ. 4.1.8.3 and shows that the critical time step value in the explicit time integration of
dynamic equation of motion can be carried out by the interpretation of a shock wave propagation in the material.
This is shown for the first time by Courant et al. in 1928 [56]. In spite of their works limited to simple cases, the
same procedure can be used for different kinds of finite elements. The characteristic lengths of the elements are
found and EQ. 4.1.8.3 is written for all elements to find the most critical time step over a mesh. Regarding to the
type (shape) of element, the expression of characteristic length is different. Figure 4.1.7 shows some typical
cases for elements with one integration point.
⎛l⎞
Δt = min Elements ⎜ ⎟ EQ. 4.1.9.1
⎝c⎠
where:
• l is the element characteristic length,
• c is the speed of sound in the material.
20-jan-2009 58
RADIOSS THEORY Version 10.0 DYNAMIC ANALYSIS
• Implement small strain formulation using a constant time step. This only works for shell and brick
elements.
These options are defined using a third keyword: either STOP, DEL or CST.
2m
Δt = min Nodes EQ. 4.1.9.2
k
where m is the nodal mass and k an equivalent nodal stiffness.
The nodal stiffness is one half of eigen value from element stiffness matrix; for a truss element this value is
equal to the diagonal term of the stiffness matrix. It is computed from the accumulation of element and interface
stiffnesses. These stiffnesses are obtained during internal force computation.
For a regular mesh, the element time step and nodal time step conditions are identical. For example, take a truss
element, Figure 4.5.2. Using an element time step condition:
l l
Δt element = = EQ. 4.1.9.3
c E
ρ
2m
Δt nodal = EQ. 4.1.9.4
k
1 EA
with m= ρAl and k = EQ. 4.1.9.5
2 l
Therefore:
ρAl l l
. Δt nodal = = = = Δt element EQ. 4.1.9.6
EA E c
l
ρ
To select the nodal time step when running RADIOSS the option /DT/NODA has to be used.
As for the element time step, minimum time step and scale factors are required. The default value is for the scale
factor is 0.9. If the minimum time step is reached, the analysis can either be stopped or a mass scaling
formulation can be applied. In this latter case, mass is added to the affected nodes so that the time step remains
constant at the minimum value. This option can be enabled using the same third keyword as used in the element
time step control. It must be checked that added masses do not affect the accuracy of results. If one uses the
nodal time step, the element time step is ignored.
20-jan-2009 59
RADIOSS THEORY Version 10.0 DYNAMIC ANALYSIS
Option /DT/NODA is used with or without interface Nodal time step is computed and a scaling factor of 0.9
types 7 and/or 10, 11 (default) is applied.
If the deletion option is applied with the /DT/INTER/DEL interface time step control, the node controlling the
minimum time step is deleted from the interface.
Mass scaling, where mass is added to an interface node, can be enabled using the option: /DT/INTER/CST
20-jan-2009 60
RADIOSS THEORY Version 10.0 DYNAMIC ANALYSIS
When we substrate EQ. 4.1.10.6 to EQ. 4.1.10.5 and we use the EQ. 4.1.10.4 :
20-jan-2009 61
RADIOSS THEORY Version 10.0 DYNAMIC ANALYSIS
−1
Where A = 2 I + dt M
2
K
For non divergence of EQ. 4.1.10.7.:
=> λ : largest eigen value of A-I is smaller than 1
=> det (λI + I + dt M K ) = 0
2 −1
Application
ke1 ke 2
1 1 1 1
me1 me 2 me1 me 2
2 2 2 2
2mel
dt1 =
kel
2me 2
dt 2 =
ke 2
me1 + me 2
dt n =
kel + ke 2
Interface
2m
dt n = k = ∑ k interface + ∑ k elements
k
Kinematic time step (interface type 7):
dt < (gap-pene)/2 V
V: impact speed
20-jan-2009 62
RADIOSS THEORY Version 10.0 DYNAMIC ANALYSIS
n
U1 (t ) = ∑ α1i (t )Φ1i = Φ1α1 (t ) EQ. 4.3.1.1
1
U & = Φ α&
1 1 1
EQ. 4.3.1.2
&U& = Φ α&&
1 1 1
20-jan-2009 63
RADIOSS THEORY Version 10.0 DYNAMIC ANALYSIS
The number of modal unknowns α1 chosen is much smaller than the original number of degrees of freedom of
Sub-domain 1.
In order to obtain the new coupled system, the dynamic equilibrium of sub-domain 1 must be projected onto the
reduction basis and the velocities involved in the kinematic relations must be expressed in terms of the modal
coordinates. Thus, we write the new matrix system for a single time scale as:
⎡ Δt Δt T ⎤ ⎡ Δt n +1 ⎤
⎢ 2 M̂ 1 0 -
2
Ĉ1 ⎥ n +1
⎡ &α&1 ⎤ ⎢ 2
F̂1 ⎥
⎢ Δt Δt T ⎥ ⎢ && n +1 ⎥ ⎢ Δt n +1 ⎥
⎢ 0 M2 - C 2 ⎥ ⎢U 2 ⎥ = ⎢ F2 ⎥ EQ. 4.3.1.3
⎢ 2 2 ⎥⎢ ⎢ 2 ⎥
⎢- Δt Ĉ Δt ⎥ ⎣ Λ ⎥⎦ ⎢Ĉ P α& n +1 + C P U& n +1 ⎥
- C2 0 1 1 2 2
⎢⎣ 2 1 2 ⎥⎦ ⎢⎣ ⎥⎦
T
M̂1 = Φ1 M1Φ1
T
F̂1 = Φ1 Fext1 − K̂1α1
where
T
K̂1 = Φ1 K1Φ1
Ĉ1 = C1Φ1
The structure of this system is strictly identical to that which existed before reduction. Therefore, we can use
exactly the same resolution process and apply the multi-time-step algorithm.
The time step for a reduced sub-domain is deduced from the highest eigen frequency of the projected system in
order to preserve the stability of the explicit time integration. This time step is often larger than that given by the
Courant condition with the finite element model before reduction.
Four points (O, A, B, C ) in space are arbitrarily chosen to represent the position of a local frame attached to the
sub-domain. In order to simplify the local equations, we choose these points so that they constitute an ortho-
normal frame.
Remark: The four points defining the local frame do not have to coincide with nodes of the mesh.
20-jan-2009 64
RADIOSS THEORY Version 10.0 DYNAMIC ANALYSIS
u = Xu A + Yu B + Zu C + (1 − X − Y − Z ) u O + Pu L = u E + Pu L EQ. 4.3.2.1
( X
) (
Y
)Z
( ) X
[ ( Y
) (
Z
u& = X + u L u& A + Y + u L u& B + Z + u L u& C + 1 − X + u L − Y + u L − Z + u L u& O ) ( )]
+ Pu& L
( ) ( ) ( ) [ ( ) ( ) ( )]
&u& = X + u L X &u& A + Y + u LY &u& B + Z + u L Z &u&C + 1 − X + u L X − Y + u LY − Z + u L Z &u&O
+ P&u& + 2P& u&
L L EQ. 4.3.2.4
X Y Z
where u L , u L , u L are the components of the local displacement in the local frame. The assumption of small
perturbations in the local frame enables us to consider that the rigid and the deformed configurations are the
X Y Z
same, i.e. we can neglect u L , u L , u L compared to the local coordinates X , Y , Z . Thus, we get simplified
expressions of the velocity and acceleration fields:
To express the weak form of the dynamic equilibrium, we also need the variation δu of the displacement field:
δP = [δu A − δu O ; δu B − δu O ; δu C − δu O ]
where
20-jan-2009 65
RADIOSS THEORY Version 10.0 DYNAMIC ANALYSIS
The principle of virtual work yields a weak form of this equilibrium, taking into account, naturally, Dirichlet-
type boundary conditions:
∫ ρδu
T
&u&dΩ − ∫ δu T f int (u )dΩ = 0 EQ.4.3.2.9
Ω Ω
where index L expresses that the coordinates and the spatial derivatives are taken in the local frame.
The virtual work of the internal forces is then:
The integration by parts in the local frame introduces external surface forces f ext :
( )
δWint = − σ L (u L ) : ε L P T δu dΩ + δu T Pσ L (u L ).n L dΩ
∫ ∫
Ω Γ
∫ ( T
)
= − σ L (u L ) : ε L P δu dΩ + δu T f ext dS ∫
Ω Γ EQ.4.3.2.12
To compute forces associated to the rigid links, first, we introduce new Lagrange multipliers to express the
energy of a link:
Wlinks = ∑Λ D (u I , u J )
IJ IJ EQ.4.3.2.13
( I , J )∈(O , A, B ,C )2
J >I
where ( ) ( )
D IJ = x 0I + u I − x 0J + u J − x 0I − x 0J and x 0I are the initial coordinates of point I and the rigid
link between points I and J is given by: D IJ (u I , u J ) = 0 .
20-jan-2009 66
RADIOSS THEORY Version 10.0 DYNAMIC ANALYSIS
Then, the differentiation of this energy is used to obtain the virtual work to be introduced into the weak form of
the equilibrium:
D IJ (u I , u j )
δWlinks = − ∑ δΛ IJ D IJ (u I , u j ) − ∑ δu I ∑ δΛ IJ EQ.4.3.2.14
( I , J )∈(O , A, B ,C )2 I ∈ O , A , B ,C )
( J ∈ O , A , B ,C )
( ∂u I
J >I J ≠I
∀ δu I I ∈ (O, A , B, C ) ,
∂D IJ (u I , u J )
Remark: The quantity ∑
J ∈ O , A, B ,C )
(
Λ IJ
∂u I
= Flinks IJ can be viewed as the resisting force applied to
J ≠I
point I to preserve the distances from this point to the other points of the local frame.
3
u= ∑u i =1
i i
AφX + u iB φ iY + u iC φ iZ + u iO φ 1i − X − Y − Z + y i Pφ iL
ˆ
= ΦPU EQ.4.3.2.15
where φ iX = Xei , φYi = Yei , φ iZ = Zei , φ1i − X −Y −Z = (1 − X − Y − Z )ei , (e1 , e 2 , e3 ) is a basis of the
{ }
i
global frame, φ L is a basis of local Ritz vectors obtained, for example, by finite element discretization or by
modal analysis, Û is the vector of the discrete unknowns:
⎡ [u ] ⎤⎥
i
⎢ A
⎡ [u i
] ⎤⎥
⎢ [u ] ⎥
i
⎢ A
⎢
B
⎡ ⎤ ⎢ [u i
] ⎥ and Û = [y ] ,
Û = ⎢ [u ] ⎥⎥ = ⎢ÛÛ
i E
⎥ , with Û E = ⎢
B i
C
[u i
] ⎥⎥ L
⎢
[u ] ⎥⎥ ⎣
i L⎦ ⎢ C
⎢ O ⎢
[u i
] ⎥⎦⎥
⎢
⎣ [y ] ⎥⎦
i ⎣⎢ O
[
Φ P is the projection basis: Φ P = {φiX }, {φYi }, {φiZ }, {φ1i − X −Y −Z }, {PφiL } = [Φ E , PΦ L ] . ]
EQ.4.3.2.5 and EQ.4.3.2.7 yield:
&
u& = Φ P Û
&& &
&u& = Φ P Û + G⎛⎜ Û ⎞⎟ EQ.4.3.2.16
⎝ ⎠
δu = Φ PδÛ
&
where G ⎛⎜ Û ⎞⎟ is the gyroscopic contribution to the acceleration, given by:
⎝ ⎠
&
([ ] [ ] [ ] [ ]) &
G⎛⎜ Û ⎞⎟ = 2P& u& iA , u& iB , u& iC , u& iO Φ L Û L
⎝ ⎠
20-jan-2009 67
RADIOSS THEORY Version 10.0 DYNAMIC ANALYSIS
The final expression of the complete weak form of the dynamic equilibrium is obtained as:
∫ δÛ (Φ )Φ ( ) ( ) ( )
&& &
ÛρdΩ + ∫ δÛT Φ P G⎛⎜ Û ⎞⎟ρdΩ + ∫ σ L Φ L Û L : ε L ΨPδÛ dΩ
T T T
Ω
P P
Ω
⎝ ⎠ Ω
EQ.4.3.2.17
− δΛ D(Û ) − δÛ ( ) = ∫ δÛ (Φ )f
T T
T
E E Flinks Λ, Û E T
P ext dS
Γ
where { }
Φ L = ϕ Li ,
ψ P = [{PT ϕ Xi }{
, P T ϕYi }{
, P T ϕ Zi }{ , ϕ Li }] = [P T Φ E , Φ L ]
, P T ϕ1i− X −Y − Z }{
D is the vector formed by the 6 relations preserving the relative distances of points (O, A, B, C ) ,
Λ is the vector of the Lagrange multipliers corresponding to each rigid link,
Flinks is the vector of the link forces given by Equation (31).
EQ. 4.3.2.17 can be rewritten using classical matrix and vector operators obtained by finite element
discretization:
&& &
( ) (
δÛ T M P Û + δÛ T Fgyr ⎛⎜ Û ⎞⎟ + δÛ T K L Û L − δΛT D Û E − δÛ E T Flinks Λ, Û E = δÛ T Fext P
⎝ ⎠
) EQ.4.3.2.18
T
where M P = Φ P MΦ P , with M being the classical mass matrix of sub-domain and Φ P the projection
matrix consisting of vectors of Φ P discretized on the nodes of the mesh,
T
K L = ΨP K Φ L , with K being the sub-domain’s local stiffness matrix and ΨP and Φ L deduced (as was
T
Φ L ) from ΨP , Φ L and the mesh, Fext P = Φ P Fext , with Fext being the classical vector of the external forces
assembled on the sub-domain.
Now, we are able to reduce the number of unknowns on the sub-domain drastically by choosing as the Ritz
vectors, instead of classical finite element shape functions, an appropriate (and small) family of local reduction
vectors. The modal vibration problem is purely local and guidelines found in the literature for the proper choice
of the projection basis apply here.
Remark 1
As far as inertia coupling with local vibration and overall large motion is concerned, two separate contributions
must be considered. The first contribution appears in the projected mass matrix, which as now the following
form:
⎡M M TC ⎤
MP = ⎢ E ⎥ EQ.4.3.2.20
⎣M C MV ⎦
where ME is the constant mass matrix corresponding only to the global displacement field given by
Xu& A + Yu& B + Zu& C + (1 − X − Y − Z )u& O , MV is the constant mass matrix corresponding to the local vibration
given by uL, MC is a coupling matrix, variable with overall rotation, arising from the interaction between the
20-jan-2009 68
RADIOSS THEORY Version 10.0 DYNAMIC ANALYSIS
local vibratory acceleration field expressed in the global frame P&u& L and the overall virtual displacement field
XδuA + YδuB + ZδuC + (1 − X − Y − Z )δuO ; M TC naturally comes from the symmetric interaction between
virtual local displacement field and the overall acceleration field.
Remark 2
In RADIOSS a special procedure is used to linearize the rigid links. The method is fully explained in [109].
Remark 3
The rigid body motion component of the displacement increment is computed in unconditionally stable way by
the use of Lagrange Multiplier to impose the rigid links. The deforming part is generated by the local vibration
modes retained in the reduction basis. Therefore, we can conclude that the stability condition is the same as that
given by the local vibrating system. The critical time step is constant throughout the calculation and can be
derived from the highest eigen frequency of the local reduced stiffness matrix with respect to the local reduced
mass matrix.
Φ TL KΦ LV = ω 2 Φ TL MΦ LV EQ.4.3.2.21
ω
where { }
Φ L = ϕ Li and f =
2π
.
Having determined f max , the maximum time step which can be used on the reduced sub-domain while ensuring
the stability of the time integration is:
1
Δt max = EQ.4.3.2.22
π f max
20-jan-2009 69
Chapter
ELEMENT LIBRARY
20-jan-2009 70
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
The RADIOSS finite element library can be classified into the following categories of elements:
20-jan-2009 71
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
Where: r ≡ ξ , s ≡ η , t ≡ ζ
The shape functions of an 8 node brick element, shown in Figure 5.1.1, are given by:
1
Φ1 = (1 − ξ )(1 − η )(1 − ζ ) EQ. 5.1.1.1
8
1
Φ2 = (1 − ξ )(1 − η )(1 + ζ ) EQ. 5.1.1.2
8
1
Φ3 = (1 + ξ )(1 − η )(1 + ζ ) EQ. 5.1.1.3
8
1
Φ4 = (1 + ξ )(1 − η )(1 − ζ ) EQ. 5.1.1.4
8
1
Φ5 = (1 − ξ )(1 + η )(1 − ζ ) EQ. 5.1.1.5
8
1
Φ6 = (1 − ξ )(1 + η )(1 + ζ ) EQ. 5.1.1.6
8
1
Φ7 = (1 + ξ )(1 + η )(1 + ζ ) EQ. 5.1.1.7
8
1
Φ8 = (1 + ξ )(1 + η )(1 − ζ ) EQ. 5.1.1.8
8
The element velocity field is related by:
8
vi = ∑ Φ I .viI EQ. 5.1.1.9
I =1
⎡ ∂Φ I ⎤ ⎡ ∂x ∂y ∂z ⎤ ⎡ ∂Φ I ⎤ ⎡ ∂Φ I ⎤
⎢ ∂ξ ⎥ ⎢ ∂ξ ∂ξ ⎥ ⎢
∂ξ ⎥ ∂x ⎥ ⎢ ∂x ⎥
⎢ ⎥ ⎢ ⎢ ⎥ ⎢ ∂Φ ⎥
⎢ ∂Φ I ⎥ = ⎢ ∂x ∂y ∂z ⎥ ⎢ ∂Φ I ⎥
. = Fξ .⎢ I ⎥ EQ. 5.1.2.1
⎢ ∂η ⎥ ⎢ ∂η ∂η ∂η ⎥ ⎢ ∂y ⎥ ⎢ ∂y ⎥
⎢ ∂Φ ⎥ ⎢ ∂x ∂y ∂z ⎥ ⎢ ∂Φ I ⎥ ⎢ ∂Φ I ⎥
⎢ I⎥ ⎢ ⎥⎢ ⎥ ⎢ ⎥
⎢⎣ ∂ζ ⎥⎦ ⎢⎣ ∂ζ ∂ζ ∂ζ ⎥⎦ ⎣ ∂z ⎦ ⎣ ∂z ⎦
20-jan-2009 72
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
Hence:
⎡ ∂Φ I ⎤ −1 ⎡ ∂Φ I ⎤
⎢ ⎥ = Fξ .⎢ ⎥ EQ. 5.1.2.2
⎣ ∂xi ⎦ ⎣ ∂ξ ⎦
1 ⎛ ∂v ∂v ⎞
ε&ij = ⎜⎜ i + j ⎟⎟ EQ. 5.1.2.3
2 ⎝ ∂x j ∂xi ⎠
∂vi 8
∂Φ I
=∑ ⋅ viI EQ. 5.1.2.4
∂x j I =1 ∂x j
Hence, the strain rate can be described directly in terms of the shape function:
⎛
1 ∂v ∂vj ⎞ 8 ∂ΦI
ε&ij = ⎜⎜ i + ⎟ =∑ ⋅ viI EQ. 5.1.2.5
2 ⎝ ∂xj ∂xi ⎟⎠ I =1 ∂xj
As was seen in section 2.4.1, volumetric strain rate is calculated separately by volume variation.
For one integration point:
∂Φ1 ∂Φ 7 ∂Φ 2 ∂Φ 8 ∂Φ 3 ∂Φ 5 ∂Φ 4 ∂Φ 6
=− ; =− ; =− ; =− EQ. 5.1.2.6
∂x j ∂x j ∂x j ∂x j ∂x j ∂x j ∂x j ∂x j
F.E Method is used only for deviatoric strain rate calculation in A.L.E and Euler formulation.
Volumetric strain rate is computed separately by transport of density and volume variation.
8
{ε&} = [B]{v} = ∑[BI ]{vI } EQ. 5.1.3.1
I =1
with
{ε&} = ε&xx
t
ε&yy ε&zz 2ε&xy 2ε&yz 2ε&xz
t
⎡∂ΦI ∂ΦI ∂ΦI ⎤
⎢ 0 0 0 ⎥
⎢ ∂x ∂x ∂y ⎥
[BI ] = ⎢⎢ 0 ∂ΦI 0 ∂ΦI 0 ∂ΦI ⎥⎥
∂y ∂x ∂z
⎢ ∂ΦI ∂ΦI ∂ΦI ⎥
⎢ 0 0 0 ⎥
⎢⎣ ∂z ∂y ∂z ⎥⎦
20-jan-2009 73
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
It is useful to take the Belytschko-Bachrach's mix form [27] of the shape functions written by:
4
Φ I ( x, y, z, ξ ,η , ζ ) = Δ I + bxI ⋅ x + byI ⋅ y + bzI ⋅ z + ∑ γ Iα φα EQ. 5.1.3.2
α =1
Where
∂Φ I
biI = (ξ = η = ζ = 0);
∂xi
1⎡ ⎛ 8
⎞ ⎛ 8
⎞ ⎛ 8
⎞ ⎤
γ Iα = ⎢ΓIα − ⎜ ∑ ΓJα x J ⎟bxI − ⎜ ∑ ΓJα y J ⎟b yI − ⎜ ∑ ΓJα z J ⎟bzI ⎥;
8⎣ ⎝ J =1 ⎠ ⎝ J =1 ⎠ ⎝ J =1 ⎠ ⎦
φ = ηζ ξζ ξη ξηζ
∂Φ I 4
= biI + ∑ γ Iα ∂φα
∂xi EQ. 5.1.3.3
∂xi α =1
It is decomposed by a constant part which is directly formulated with the Cartesian coordinates, and a non-
constant part which is to be approached separately. For the strain rate, only the non constant part is modified by
the assumed strain. We can see in following that the non constant part or the high order part is just the hourglass
terms.
We have now the decomposition of the strain rate:
with:
⎡bxI 0 0 ⎤ t
⎡ 4 α ∂φα 4 4
⎤
⎢∑γ I ∂x ∑γ ∑
⎢0 b 0 ⎥ 0 0 0 α ∂φα
I ∂x
α ∂φα
γI ∂y ⎥
⎢ yxI ⎥
⎢ 0 0 bzI ⎥ ⎢α=1 4 4
α=1 α=1
4 ⎥
[BI ] = ⎢ ⎥ ; [BI ] = ⎢ 0
α ∂φα ⎥
0
b b 0
H
⎢ ∑γIα ∂∂φyα 0 ∑ γα φ ∂α
I ∂x 0 ∑γI ∂z ⎥
⎢ yxI xI ⎥ α=1 α=1 α=1
⎢ bzI 0 bxI ⎥ ⎢ 4 4 4 ⎥
⎢ ⎥ ⎢ 0
⎢⎣
0 ∑γIα ∂∂φzα ∑γIα ∂∂φyα ∑ γα φ ∂α
I ∂z 0 ⎥
⎥⎦
⎣⎢ 0 bzI b ⎥
yxI⎦
α=1 α=1 α=1
with
20-jan-2009 74
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
also removed since this mode is non-physical and is stabilized by other terms in BI [ ]
H
.
The terms with Poisson coefficient are added to obtain an isochoric assumed strain field when the nodal velocity
is equivoluminal. This avoids volumetric locking as ν = 0.5 . In addition, these terms enable the element to
capture transverse strains which occurs in a beam or plate in bending. The plane strain expressions are used since
this prevents incompatibility of the velocity associated with the assumed strains.
∂Φ I
f iIint = ∫ σ ij dΩ EQ. 5.1.4.1
Ω
∂x j
However, to increase the computational speed of the process, some simplifications are applied.
20-jan-2009 75
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
∂Φ1 ∂Φ 7 ∂Φ 2 ∂Φ 8 ∂Φ 3 ∂Φ 5 ∂Φ 4 ∂Φ 6
=− ; =− ; =− ; =− EQ. 5.1.4.2
∂x j ∂x j ∂x j ∂x j ∂x j ∂x j ∂x j ∂x j
∂Φ I
Hence, the value is taken at the integration point and the internal force is computed using the relation:
∂x j
⎛ ∂Φ I ⎞
FiI = σ ij ⎜ ⎟ Ω EQ. 5.1.4.3
⎜ ∂x ⎟
⎝ j ⎠0
The force calculation is exact for the special case of the element being a parallelepiped.
Γ1 = (+ 1,−1,+1,−1,+1,−1,+1,−1)
Γ 2 = (+ 1,+1,−1,−1,−1,−1,+1,+1)
20-jan-2009 76
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
Γ 3 = (+ 1,−1,−1,+1,−1,+1,+1,−1)
Γ 4 = (+ 1,−1,+1,−1,−1,+1,−1,+1)
To correct this phenomenon, it is necessary to introduce anti-hourglass forces and moments. Two possible
formulations are presented hereafter.
∂qiα 8
= ∑ ΓIα ⋅ viI EQ. 5.1.5.1
∂t I =1
where:
• Γ is the non-orthogonal hourglass mode shape vector,
• v is the node velocity vector,
• i is the direction index, running from 1 to 3,
• I is the node index, from 1 to 8,
• α is the hourglass mode index, from 1 to 4.
This vector is not perfectly orthogonal to the rigid body and deformation modes.
All hourglass formulations except the physical stabilization formulation for solid elements in RADIOSS use a
viscous damping technique. This allows the hourglass resisting forces to be given by:
( ) ∑ ∂∂qt
α
1 2
f iIhgr = ρch 3 Ω i
⋅ ΓIα EQ. 5.1.5.2
4 α
where:
• ρ is the material density,
• c is the sound speed,
• h is a dimensional scaling coefficient defined in the input,
• Ω is the volume.
20-jan-2009 77
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
⎛ ∂v
viIHour = viI − ⎜ viI + iI ⋅ x j − x j
⎜ ∂x j
( )⎞⎟⎟ EQ. 5.1.5.4
⎝ ⎠
Decomposition on the hourglass vectors base gives [12]:
∂qiα ⎛ ∂v ⎞
= ΓIα ⋅ viIHour = ⎜ viI − il ⋅ x j ⎟ ⋅ ΓIα EQ. 5.1.5.5
∂t ⎜ ∂x j ⎟
⎝ ⎠
where:
∂qiα
• are the hourglass modal velocities,
∂t
• ΓIα is the hourglass vectors base.
∂vi ∂Φ j
Remembering that = ⋅ viJ and factorizing EQ. 5.1.5.5 gives:
∂x j ∂x j
∂qiα ⎛ ∂Φ j ⎞
= viI ⋅ ⎜ ΓIα − x j ΓIα ⎟ EQ. 5.1.5.6
∂t ⎜ ∂x j ⎟
⎝ ⎠
∂Φ j
γ Iα = ΓIα − x j ΓJα EQ. 5.1.5.7
∂x j
α
is the hourglass shape vector used in place of ΓI in EQ. 5.1.5.2.
{f } = {( f ) }+ {( f )
I
int
I
int 0
I
int H
} EQ. 5.1.5.8
20-jan-2009 78
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
{f } = ∫ [B ] [C ]∑ [B ]{v }dΩ t 8
int J
I I J
Ω
j =1
EQ. 5.1.5.9
(
= ∫ [BI ] + [B I ] ) [C ]∑ ([B ] + [B ] ){v }dΩ
8
0 H t 0 H J
J J
Ω j =1
The constant part {( f ) }= ∫ ([B ] ) [C ]∑ [B ] {v }dΩ is evaluated at the quadrature point just like
I
int 0
I
0 t
8
j =1
J
0 J
Ω
other one-point integration formulations mentioned before, and the non constant part (Hourglass) will be
calculated as following:
∂xi
Taking the simplification of = 0; (i ≠ j ) (i.e. the Jacobian matrix of EQ. 5.1.2.1 is diagonal), we have
∂ξ j
(f )
4
int H
iI = ∑ Qiα γ Iα EQ. 5.1.5.10
α =1
.
with 12 generalized hourglass stress rates Q iα calculated by:
[
Q ii = μ (H jj + H kk )q& ii + H ij q& jj + H ik q& kk ]
.
.
⎡ 1 ⎤
Q jj = μ ⎢ H ii q& ij + ν H ij q& ij ⎥ EQ. 5.1.5.11
⎣1 − ν ⎦
. 1 +ν
Q i 4 = 2μ H ii q& 4j
3
and
2 2 2
⎛ ∂φ j ⎞ ⎛ ∂φ ⎞ ⎛ ∂φ ⎞
H ii = ∫ ⎜⎜ ⎟⎟ dΩ = ∫ ⎜⎜ k ⎟⎟ dΩ = 3∫ ⎜⎜ 4 ⎟⎟ dΩ
Ω⎝
∂xi ⎠ Ω⎝
∂xi ⎠ Ω⎝
∂xi ⎠ EQ. 5.1.5.12
∂φ ∂φ j
H ij = ∫ i dΩ
Ω
∂x j ∂xi
α
Where i,j,k are permuted between 1 to 3 and q& i has the same definition than in EQ. 5.1.5.6.
Extension to non-linear materials has been done simply by replacing shear modulus μ by its effective tangent
values which is evaluated at the quadrature point. For the usual elastoplastic materials, we use a more sophistic
procedure which is described in the following section.
With one-point integration formulation, if the non-constant part follows exactly the state of constant part for the
case of elasto-plastic calculation, the plasticity will be under-estimated due to the fact that the constant
equivalent stress is often the smallest one in the element and element will be stiffer. Therefore, defining a yield
criterion for the non-constant part seems to be a good ideal to overcome this drawback.
Plastic yield criterion: The Von Mises type of criterion is written by:
20-jan-2009 79
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
for any point in the solid element, where σ y is evaluated at the quadrature point.
As only one criterion is used for the non-constant part, two choices are possible:
2. taking the value by some representative points, e.g. eight Gausse points
The second choice has been used in this element.
Elastro-plastic hourglass stress calculation:
The incremental hourglass stress is computed by:
• Elastic increment
(σ i )trH = (σ i )n + [C ]{ε&} Δt
H H
n +1
5.1.6 Stability
The stability of the numerical algorithm depends on the size of the time step used for time integration (section
4.5). For brick elements, RADIOSS uses the following equation to calculate the size of the time step:
l
h≤k
(
c α + α 2 +1 ) EQ. 5.1.6.1
This is the same form as the Courant condition for damped materials. The characteristic length of a particular
element is computed using:
Element Volume
l= EQ. 5.1.6.2
Largest Side Surface
For a 6-sided brick, this length is equal to the smallest distance between two opposite faces.
The terms inside the parentheses in the denominator are specific values for the damping of the material:
2v
• α=
ρcl
• v effective kinematic viscosity,
1∂p
• c= for fluid materials,
ρ∂ρ
K 4μ λ + 2μ
• c= + = for a solid elastic material,
ρ 3ρ ρ
• K is the bulk modulus,
• λ , μ are Lame moduli.
20-jan-2009 80
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
The scaling factor k=0.90, is used to prevent strange results that may occur when the time step is equal to the
Courant condition. This value can be altered by the user.
2
⎛ ∂ε ⎞ ∂ε
q = qa ρl 2 ⎜ kk ⎟ − qb ρlc kk
2
EQ. 5.1.7.1
⎝ ∂t ⎠ ∂t
where
• l is equal to 3
Ω or to the characteristic length,
• Ω is the volume,
∂ε kk
• is the volumetric compression strain rate tensor,
∂t
• c is the speed of sound in the medium.
• qa is a scalar factor on the quadratic viscosity to be adjusted so that the Hugoniot equations are
verified. This value is defined by the user. The default value is 1.10.
• qb is a scalar factor on the linear viscosity that damps out the oscillations behind the shock. This is
user specified. The default value is 0.05.
Default values are adapted for velocities lower than Mach 2. However for viscoelastic materials (law 34, 35, 38)
or honeycomb (law 28), very small values are recommended, i.e. 10-20.
20-jan-2009 81
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
Recommended degeneration
Connectivity: 1 2 3 4 5 5 5 5
Connectivity : 1 2 3 4 5 6 6 5
20-jan-2009 82
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
where
• σ& ijr is the stress rate due to the rigid body rotational velocity,
The correction for stress rotation from time t to time t+ Δt is given by [2]:
effects. The constitutive law is directly applied to the Jaumann stress rate tensor.
Deviatoric stresses and pressure (see section 2.7) are computed separately and related by:
where
So, when co-rotational formulation is used, EQ. 5.1.10.2 σ& ij = σ& ij + σ& ij reduces to:
v r
where σ& ij is the Jaumann objective stress tensor derivative expressed in the co-rotational frame.
v
The following illustrates orthogonalization, when one of the r, s, t directions is orthogonal to the two other
directions.
20-jan-2009 83
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
⎡(∂v x / ∂y − ∂v y / ∂x ) = 0
⎢
ΔΩ = Δt / 2 ⋅ ⎢(∂v x / ∂z − ∂vz / ∂x ) = ∂v x / ∂z EQ. 5.1.9.6
⎢(∂v y / ∂x − ∂vx / ∂y ) = 0
⎣
So, ΔΩ y = αΔT / 2 where α = v / h equals the imposed velocity on the top of the brick divided by the height
of the brick (constant value).
Due to first order approximation, the increment of stress σ xx due to the rigid body motion is:
Δσ xxr = ΔΩ y (τ xz + τ zx ) = 2ΔΩ yτ xz = αΔTτ xz EQ. 5.1.9.7
20-jan-2009 84
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
Δτ xz / ΔT 2 = −α 2τ xz EQ. 5.1.9.14
So, it is recommended to use co-rotational formulation especially for visco-elastic materials such as foams, even
if this formulation is more time consuming than the global one.
20-jan-2009 85
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
On the other hand, when the co-rotational formulation is used, the orthotropic frame keeps the same orientation
with respect to the local (co-rotating) frame, and is therefore also co-rotating (see below).
⎛ 1 ⎞
sijel (t + Δt ) = sij (t ) + s&ijr Δt + 2G⎜ ε&ij − ε&kk δ ij ⎟Δt EQ. 5.1.10.1
⎝ 3 ⎠
where G is the shear modulus.
This relationship is Hooke's Law, where the strain rate is multiplied by time to give strain.
STEP 2: Compute Von Mises Equivalent Stress and Current Yield Stress
Depending on the type of material being modeled, the method by which yielding or failure is determined will
vary. The following explanation relates to an elastoplastic material (law 2).
The Von Mises Equivalent stress relates a three dimensional state of stress back to a simple case of uniaxial
tension where material properties for yield and plasticity are well known and easily computed.
The Von Mises stress, which is strain rate dependent, is calculated using the equation:
3 el el
σ vm
e
= sij sij EQ. 5.1.10.2
2
The flow stress is calculated from the previous plastic strain:
σ y (t ) = a + bε p (t )
n
EQ. 5.1.10.3
20-jan-2009 86
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
For material types 3, 4, 10, 21, 22, 23 and 36, EQ. 5.1.11.3 is modified according to the different modeling of the
material curves.
STEP 3: Plasticity Check
The state of the deformation must be checked.
σ vm
e
−σ y ≤ 0
If this equation is satisfied, the state of stress is elastic. Otherwise, the flow stress has been exceeded and a
plasticity rule must be used. This is shown in Figure 5.1.2.
Figure 5.1.2 - Plasticity Check
dσ y
H= EQ. 5.1.10.4
dε p
This is used to compute the plastic strain at time t:
σ vm − σ y
ε& p Δt = EQ. 5.1.10.5
3G + H
This plastic strain is time integrated to determine the plastic strain at time t + Δt :
σ y (t + Δt ) = a + bε p (t + Δt )
n
EQ. 5.1.10.7
σ y el
sijpa = sij EQ. 5.1.10.8
σ vm
20-jan-2009 87
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
20-jan-2009 88
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
L1 = r
L2 = s
L3 = t
20-jan-2009 89
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
L4 = 1 − L1 − L2 − L3
Φ 5 = 4 L1 L2 EQ. 5.2.2.5
Φ 6 = 4 L2 L3 EQ. 5.2.2.6
Φ 7 = 4 L3 L1 EQ. 5.2.2.7
Φ 8 = 4 L1 L4 EQ. 5.2.2.8
Φ 9 = 4 L2 L4 EQ. 5.2.2.9
With,
α = 0 ⋅ 58541020 and β = 0 ⋅13819660
a, b, c, d are the 4 integration points.
20-jan-2009 90
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
2
Lc = a ; Lc = 0.816a EQ. 5.2.2.12
3
Figure 5.2.2 - 4 nodes tetra
5/ 2
Lc = a ; Lc = 0.264a EQ. 5.2.2.13
6
Figure 5.2.3 - 10 nodes tetra
For another regular tetra obtained by the assemblage of four hexa as shown in Figure 5.2.4, the characteristic
length is:
2/3
Lc = a ; Lc = 0.204a EQ. 5.2.2.14
4
Figure 5.2.4 - Other regular tetra
20-jan-2009 91
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
20-jan-2009 92
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
5.3.1 Introduction
The historical shell element in RADIOSS is a simple bilinear Mindlin plate element coupled with a reduced
integration scheme using one integration point. It is applicable in a reliable manner to both thin and moderately
thick shells.
This element is very efficient if the spurious singular modes, called “hourglass modes”, which result from the
reduced integration are stabilized.
The stabilization approach consists of providing additional stiffness so that the spurious singular modes are
suppressed. Also, it offers the possibility of avoiding some locking problems. One of the first solutions was to
generalize the formulation of Kosloff and Frazier [10] for brick element to shell element. It can be shown that
the element produces accurate flexural response (thus, free from the membrane shear locking) and is equivalent
to the incompatible model element of Wilson et al. [21] without the static condensation procedure. Taylor [47]
extended this work to shell elements. Hughes and Liu [22] employed a similar approach and extended it to non
linear problems.
Belytschko and Tsay [23] developed a stabilized flat element based on the γ -projections developed by Flanagan
and Belytschko [12]. Its essential feature is that hourglass control is orthogonal to any linear field, thus
preserving consistency. The stabilized stiffness is approached by a diagonal matrix and scaled by the
perturbation parameters hi which are introduced as a regulator of the stiffness for nonlinear problems. The
parameters hi are generally chosen to be as small as possible, so this approach is often called a perturbation
stabilization.
The elements with perturbation stabilization have two major drawbacks:
20-jan-2009 93
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
which the strain field is expressed explicitly in terms of natural coordinates by a Taylor-series expansion. A
remarkable improvement in the one-point quadrature shell element with physical stabilization has been
performed by Belytschko and Leviathan [18]. The element performs superbly for both flat and warped elements
especially in linear cases, even in comparison with a similar element under a full integration scheme, and is only
20% slower than the Belytschko and Tsay element. More recently, based on Belytschko and Leviathan's element,
Zhu and Zacharia [30] implemented the drilling rotation DOF in their one-point quadrature shell element; the
drilling rotation is independently interpolated by the Allman function [39] based on Hughes and Brezzi's [41]
mixed variational formulation.
The physical stabilization with assumed strain method seems to offer a rational way of developing a cost
effective shell element with a reduced integration scheme. The use of the assumed strains based on the mixed
variational principles, is powerful, not only in avoiding the locking problems (volumetric locking, membrane
shear locking, as in Belytschko and Bindeman [31]; transverse shear locking, as in Dvorkin and Bathe[32]), but
also in providing a new way to compute stiffness. However, as highlighted by Stolarski et al. [33], assumed
strain elements generally do not have rigorous foundations; there is almost no constraint for the independent
assumed strains interpolation. Therefore, a sound theoretical understanding and numerous tests are needed in
order to prove the legitimacy of the assumed strain elements.
The greatest uncertainty of the one-point quadrature shell elements with physical stabilization is with respect to
the nonlinear problems. All of these elements with physical stabilization mentioned above rely on the
assumptions that the spin and the material properties are constant within the element. The first assumption is
necessary to ensure the objectivity principle in geometrical nonlinear problems. The second was adapted in order
to extend the explicit evaluation of stabilized stiffness for elastic problems to the physical nonlinear problems.
We have found that the second assumption leads to a theoretical contradiction in the case of an elastoplastic
problem (a classic physical nonlinear problem), and results in poor behavior in case of certain crash
computations.
Zeng and Combescure [15] have proposed an improved 4-node shell element named QPPS with one-point
quadrature based on the physical stabilization which is valid for the whole range of its applications (see the
chapter 5.3.12). The formulation is based largely on that of Belytschko and Leviathan.
Based on the QPPS element, Zeng and Winkelmuller have developed a new improved element named QEPH
which is integrated in RADIOSS 44 version (see the chapter 5.3.13).
20-jan-2009 94
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
Plate theory assumes that one dimension (the thickness, z) of the structure is small compared to the other
dimensions. Hence, the 3D continuum theory is reduced to a 2D theory. Nodal unknowns are the velocities
( )
(v x′v y′v z ) of the mid plane and the nodal rotation rates ω x′ω y as a consequence of the suppressed z direction.
The thickness of elements can be kept constant, or allowed to be variable. This is user defined. The elements are
always in a state of plane stress, i.e. σ zz = 0 , or there is no stress acting perpendicular to the plane of the
element. A plane orthogonal to the mid-plane remains a plane, but not necessarily orthogonal as in Kirchhoff
∂vz ∂v z
theory, (where ε xz = ε yz = 0 ) leading to the rotations rates ωx = − and ω y = . In Mindlin plate
∂y ∂x
theory, the rotations are independent variables.
When the orthogonalized mode of the hourglass perturbation formulation is used, the characteristic length is
defined as:
L4 = 0.5
(L1 + L2 )
max (hm′ h f )
EQ. 5.3.3.5
where hm is the shell membrane hourglass coefficient and h f is the shell out of plane hourglass coefficient, as
mentioned in section 5.3.8.
( )
r r r
• Global Cartesian fixed system X = Xi + Yj + Zk
20-jan-2009 95
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
The vector normal to the plane of the element at the mid point is defined as:
x× y
n= EQ. 5.3.4.1
x× y
The vector defining the local direction is:
x y
t1 = t2 = EQ. 5.3.4.2
x y
Hence, the vector defining the local direction is found from the cross product of the two previous vectors:
t2 = n × t1 EQ. 5.3.4.3
Φ I ( x, y ) = a I + bI x + cI y + d I xy EQ. 5.3.5.2
It is also useful to write the shape functions in the Belytschko-Bachrach mix form [27]:
with
[ ( ) ( ) ]
Δ I = t I − t I x I bxI − t I y I byI ; t = (1,1,1,1)
20-jan-2009 96
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
The velocity of the element at the mid-plane reference point is found using the relations:
4
vx = ∑ Φ I vxI EQ. 5.3.5.4
I =1
4
v y = ∑ Φ I v yI EQ. 5.3.5.5
I =1
4
vz = ∑ Φ I v zI EQ. 5.3.5.6
I =1
where, vxI , v yI , vzI are the nodal velocities in the x,y,z directions.
4
ω y = ∑ Φ I ω yI EQ. 5.3.5.8
I =1
where ω xI and ω yI are the nodal rotational velocities about the x and y reference axes.
The velocity change with respect to the coordinate change is given by:
∂v x 4
∂Φ I
=∑ vxI EQ. 5.3.5.9
∂x I =1 ∂x
∂v x 4
∂Φ I
=∑ v xI EQ. 5.3.5.10
∂y I =1 ∂y
1 ⎛ ∂v ∂v y ⎞
e&xy = ⎜⎜ x + ⎟ EQ. 5.3.6.3
2 ⎝ ∂y ∂x ⎟⎠
1 ⎛ ∂v ∂v ⎞ 1 ⎛ ∂v ⎞
e&xz = ⎜ x + z ⎟ = ⎜ ω y + z ⎟ EQ. 5.3.6.4
2 ⎝ ∂z ∂x ⎠ 2 ⎝ ∂x ⎠
20-jan-2009 97
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
1 ⎛ ∂v y ∂v z ⎞ 1⎛ ∂v ⎞
e& yz = ⎜ + ⎟⎟ = ⎜⎜ − ω x + z ⎟⎟ EQ. 5.3.6.5
2 ⎜⎝ ∂z ∂y ⎠ 2⎝ ∂y ⎠
where e&ij is the membrane strain rate.
1 ⎛ ∂ω ∂ω ⎞
χ& xy = ⎜⎜ y − x ⎟⎟ EQ. 5.3.6.8
2 ⎝ ∂y ∂x ⎠
where χ& ij is the curvature rate.
This vector is computed from the velocity field vector {v}m and the shape function gradient {B}m :
where
⎡ ∂Φ1 ∂Φ 2 ∂Φ 3 ∂Φ 4 ⎤
⎢ 0 0 0 0 ⎥
⎢ ∂x ∂x ∂x ∂x ⎥
∂Φ1 ∂Φ 2 ∂Φ 3 ∂Φ 4 ⎥
[B]m =⎢ 0
⎢
0 0 0 EQ. 5.3.6.12
∂y ∂y ∂y ∂y ⎥
⎢ ∂Φ ∂Φ1 ∂Φ 2 ∂Φ 2 ∂Φ 3 ∂Φ 3 ∂Φ 4 ∂Φ 4 ⎥
⎢ 1 ⎥
⎢⎣ ∂y ∂x ∂y ∂x ∂y ∂x ∂y ∂x ⎦⎥
20-jan-2009 98
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
As with the membrane strain rate, the bending strain rate is computed from the velocity field vector. However,
the velocity field vector for the bending strain rate contains rotational velocities, as well as translations:
where
{v}b = {ω1y ,−ω1x , ω y2 ,−ωx2 , ω y3 ,−ω x3 , ω y4 ,−ω x4 , v1z , vz2 , vz3 , vz4 } EQ. 5.3.6.15
⎡ ∂Φ1 ∂Φ 2 ∂Φ 3 ∂Φ 4 ⎤
⎢ ∂x 0 0 0 0 0 0 0 0 ⎥
∂x ∂x ∂x
⎢ ∂Φ1 ∂Φ 2 ∂Φ 3 ∂Φ 4 ⎥
⎢ 0 0 0 0 0 0 0 0 ⎥
⎢ ∂y ∂y ∂y ∂y ⎥
⎢ ∂Φ 3 ∂Φ 3 ⎥
[B]b = ⎢ ∂Φ1 ∂Φ1 ∂Φ 2 ∂Φ 2 ∂Φ 4 ∂Φ 4
0 0 0 0 ⎥
⎢ ∂y ∂x ∂y ∂x ∂y ∂x ∂y ∂x ⎥
⎢Φ ∂Φ1 ∂Φ 2 ∂Φ 3 ∂Φ 4 ⎥
0 Φ2 0 Φ3 0 Φ4 0
⎢ 1 ∂x ∂x ∂x ∂x ⎥
⎢ ∂Φ1 ∂Φ 2 ∂Φ 3 ∂Φ 4 ⎥
⎢ 0 Φ1 0 Φ2 0 Φ3 0 Φ4 ⎥
⎣ ∂y ∂y ∂y ∂y ⎦
EQ. 5.3.6.16
20-jan-2009 99
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
z
4 3
y
b
x
1 2
m
a
Due to the lumped mass formulation used by RADIOSS, the lumped mass at a particular node is:
1
m= ρAt EQ. 5.3.6.17
4
The mass moments of inertia, with respect to local element reference frame, are calculated at node i by:
⎛ b2 + t 2 ⎞
I xx = m⎜⎜ ⎟⎟ EQ. 5.3.6.18
⎝ 12 ⎠
⎛ a2 + t 2 ⎞
I yy = m⎜⎜ ⎟⎟ EQ. 5.3.6.19
⎝ 12 ⎠
⎛ a2 + b2 ⎞
I zz = m⎜⎜ ⎟⎟ EQ. 5.3.6.20
⎝ 12 ⎠
ab
I xy = −m EQ. 5.3.6.21
16
⎛ A t2 ⎞
I xx = m⎜⎜ + ⎟⎟ EQ. 5.3.6.22
⎝ f 12 ⎠
I zz = I yy = I xx EQ. 5.3.6.23
I xy = 0 EQ. 5.3.6.24
where f is a regulator factor with default value f=12 for QBAT element and f=9 for other quadrilateral elements.
20-jan-2009 100
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
In elasticity it becomes:
int 0
with the constant part f being computed with one-point quadrature and the non constant part or hourglass
int hgr
part f being computed by perturbation stabilization (Ishell = 1, 2 ,3 ...) or by physical stabilization (Ishell =
22).
Along with the translational modes, the 4 node shell has 12 rotational modes: 4 out of plane rotation modes (1, 2,
3, 4), 2 deformation modes (5, 6), 2 rigid body or deformation modes (7, 8) and 4 hourglass modes (9, 10, 11,
12).
20-jan-2009 101
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
The hourglass velocity rate for the above vector is defined as:
∂qi
= ΓI viI = viI − vi 2 + vi 3 − vi 4 EQ. 5.3.8.2
∂t
The hourglass resisting forces at node I for in-plane modes are:
1 A ∂qi
f iIhgr = ρct hm ΓI EQ. 5.3.8.3
4 2 ∂t
For out of plane mode, the resisting forces are:
1 h ∂qi
f iIhgr = ρct 2 f ΓI EQ. 5.3.8.4
4 10 ∂t
where
i is the direction index,
I is the node index,
t is the element thickness,
c is the sound propagation speed,
A is the element area,
20-jan-2009 102
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
1 hr ∂r
miIhgr = ρcAt 2 i ΓI EQ. 5.3.8.9
50 2 ∂t
where h r is the shell rotation hourglass coefficient.
20-jan-2009 103
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
⎛ ∂v
(
viIHour = viI − ⎜ viI + iI ⋅ x j − x j
⎜ ∂x j
)⎞⎟⎟ EQ. 5.3.9.2
⎝ ⎠
Decomposition on the hourglass base vectors gives [12]:
∂qiα ⎛ ∂v ⎞
= ΓIα ⋅ viIHour = ⎜ viI − iI ⋅ x j ⎟ ⋅ ΓIα EQ. 5.3.9.3
∂t ⎜ ∂x j ⎟
⎝ ⎠
where
∂qiα
• are the hourglass modal velocities,
∂t
• ΓIα is the hourglass vectors, base.
∂vi ∂Φ j
Remembering that = ⋅ viJ and factorizing EQ. 5.1.5.5 gives:
∂x j ∂x j
∂qiα ⎛ ∂Φ J ⎞
= viI ⋅ ⎜ ΓIα − x j ΓJα ⎟ EQ. 5.3.9.4
∂t ⎜ ∂x j ⎟
⎝ ⎠
∂Φ J
γ Iα = ΓIα − x j ΓJα EQ. 5.3.9.5
∂x j
α
is the hourglass shape vector used in place of ΓI in EQ. 5.1.5.2.
20-jan-2009 104
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
20-jan-2009 105
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
⎡γ I φ , x 0 0⎤
[(B ) ]
m H
I
⎢
=⎢ 0 γ I φ , y 0⎥⎥ EQ. 5.3.9.12
⎣⎢γ I φ , y γ I φ , x 0⎥⎦
For bending:
⎡ 0 γ Iφ , x ⎤
[(B ) ]
b H
I
⎢
= ⎢− γ I φ , y 0 ⎥⎥ EQ. 5.3.9.13
⎢⎣ − γ I φ , x γ I φ , y ⎥⎦
It is shown in [16] that the non-constant part of the membrane strain rate does not vanish when a warped element
undergoes a rigid body rotation. Thus, a modified matrix [ B I ( ) m H
] is chosen using zγ = γ I z as a measure of
I
the warping:
⎡γ I φ, x 0 zγ bxI φ , x ⎤
[(B ) ]
m H
I
⎢
=⎢ 0 γ I φ, y zγ byI φ , y
⎥
⎥ EQ. 5.3.9.14
⎢γ I φ, y
⎣ γ I φ, x zγ (bxI φ , y +byIφ , x )⎥⎦
This matrix is different from the Belytschko-Leviathan [17] correction term added at rotational positions, which
couples translations to curvatures as follows:
⎡ 1 ⎤
⎢γ Iφ, x 0 0 0 −
4
zγ φ, x ⎥
[(B ) ]
m H
I
⎢
=⎢ 0
⎢
γ Iφ, y 0
1
4
zγ φ, y 0 ⎥
⎥
⎥
EQ. 5.3.9.15
⎢γ φ 1 1
γ I φ, x 0 zγ φ, x − zγ φ, y ⎥
⎢⎣ I , y 4 4 ⎥⎦
This will lead to “membrane locking” (the membrane strain will not vanish under a constant bending loading).
According to the general formulation, the coupling is presented in terms of bending and not in terms of
m
membrane, yet the normal translation components in ( BI ) do not vanish for a warped element due to the
tangent vectors ti (ξ ,η ) which differ from ti(0,0).
20-jan-2009 106
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
⎛ ε& ⎞
F1new = F1old + dm⎜ ε&11 + 22 ⎟ EQ. 5.3.9.17
⎝ 2 ⎠
⎛ ε& ⎞
F2 new = F2old + dm⎜ ε&22 + 11 ⎟ EQ. 5.3.9.18
⎝ 2 ⎠
ε&12
F3new = F3old + dm EQ. 5.3.9.19
3
Where: ρ 0 = density
AREA = area of the shell element surface
dt = time step
c = sound speed
In order to calibrate the dm value so that it represents the physical viscosity, one should obtain the same size for
all shell elements (Cf. AREA factor), then scale the physical viscosity value to the element size.
The stress and strain for a shell element can be written in vector notation. Each component is a stress or strain
feature of the element. The generalized strain ε can be written as:
where
eij is the membrane strain,
χ ij is the bending strain or curvature.
The generalized stress Σ can be written as:
t/2 t/2
where: N x = ∫ σ x dz
−t / 2
M x = − ∫ σ x zdz
−t / 2
t/2 t/2
Ny = ∫σ
−t / 2
y dz M y = − ∫ σ y zdz
−t / 2
t/2 t/2
N xy = ∫ σ xy dz
−t / 2
M xy = − ∫ σ xy zdz
−t / 2
t/2 t/2
N yz = ∫ σ yz dz
−t / 2
N xz = ∫σ
−t / 2
xz dz
20-jan-2009 107
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
{∑ el
}
(t + Δt ) = {∑(t )}+ L{ε&}Δt EQ. 5.3.10.3
⎡L 0⎤
L=⎢ m EQ. 5.3.10.5
⎣0 Lb ⎥⎦
⎡ Et − vEt ⎤
⎢1 − v 2 0 ⎥
1− v2
⎢ − vEt Et ⎥
Lm = ⎢ 0 ⎥ EQ. 5.3.10.6
⎢1 − v 1− v2
2
⎥
⎢ 0 Et ⎥
⎢⎣ 0
1 + v ⎥⎦
⎡ Et 3 − vEt 3 ⎤
0
⎢
(
⎢12 1 − v3
2
) (
12 1 − v 2 ) ⎥
⎥
− vEt Et 3
Lb = ⎢⎢ 0 ⎥⎥
⎢
(
12 1 − v 2 ) (
12 1 − v 2 ) ⎥
EQ. 5.3.10.7
⎢ Et 3 ⎥
0 0
⎢⎣ 12(1 + v ) ⎥⎦
E is the Young's or Elastic Modulus,
ν is Poisson's Ratio,
t is the shell thickness.
∂f &
and ε& p = λ EQ. 5.3.10.11
∂σ
f is the yield surface function for plasticity for associative hardening. The equivalent stress σ eq may be
expressed in form:
20-jan-2009 108
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
⎧σ xx ⎫ ⎡1 − 12 0⎤
⎪ ⎪ ⎢
with {σ } = ⎨σ yy ⎬ and [ A] = − 12
⎢ 1 0⎥⎥ for von Mises criteria. The normality law (EQ. 5.3.10.11) for
⎪σ ⎪ ⎢⎣ 0 0 3⎥⎦
⎩ xy ⎭
associated plasticity is written as:
{ }
Where σ represents stress components obtained by an elastic increment and [C] the elastic matrix in plane
*
stress. The equations EQ. 5.3.10.8 to 14 lead to obtain the nonlinear equation:
( )
f dε p = 0 EQ. 5.3.10.15
To determine the elastic-plastic state of a shell element, a number of steps have to be performed to check for
yielding and defining a plasticity relationship. Stress-strain and force-displacement curves for a particular ductile
material are shown in Figure 5.3.9.
Figure 5.3.9 - Material Curve
20-jan-2009 109
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
ε x = ex − zχ x EQ. 5.3.10.16
ε y = e y − zχ y EQ. 5.3.10.17
{ε } = {ε x , ε y , ε xy } EQ. 5.3.10.19
{σ } = {σ x , σ y , σ xy } EQ. 5.3.10.20
{σ (t + Δt )} = {σ (t )} + L{ε&}Δt
el
EQ. 5.3.10.21
⎡ E vE ⎤
⎢1 − v 2 0 ⎥
1 − v2
⎢ vE E ⎥
L=⎢ 0 ⎥
⎢1 − v 1 − v2
2
⎥
⎢ 0 E ⎥
0
⎢⎣ 1 + v ⎥⎦
The two shear stresses acting across the thickness of the element are calculated by:
E
σ yzel (t + Δt ) = σ yz (t ) + α e& yz Δt EQ. 5.3.10.22
1+ v
E
σ xzel (t + Δt ) = σ xz (t ) + α e&xz Δt
1+ v
where α is the shear factor. Default is Reissner's value of 5/6.
3. Von Mises yield criterion
The Von Mises yield criterion for shell elements is defined as:
σ vm
2
= σ x2 + σ y2 − σ xσ y + 3σ xy2 EQ. 5.3.10.23
For type 2 simple elastic-plastic material the yield stress is calculated using:
σ yield (t ) = a + bε p (t )
n
EQ. 5.3.10.24
This equation will vary according to the type of material being modeled.
4. Plasticity Check
The element's state of stress must be checked to see if it has yielded. These values are compared with the von
Mises and Yield stresses calculated in the previous step. If the von Mises stress is greater than the yield stress,
then the material will be said to be in the plastic range of the stress-strain curve.
20-jan-2009 110
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
(
dε zzp = − dε xxp + dε zzp ) EQ. 5.3.10.25
where the components of membrane strains dε xx , dε yy are computed by EQ. 5.3.10.13 as:
p p
⎧⎪dε xxp ⎪⎫ dε p ⎧σ ⎫
⎨ p ⎬= [A]2×2 ⎨ xx ⎬
⎪⎩dε yy ⎪⎭ σ y ⎩σ yy ⎭ EQ. 5.3.10.26
υ 1 − 2υ p
dε zz = − (dε + dε yy ) + dε zz EQ. 5.3.10.27
1 −υ 1−υ
xx
σ yield (t ) = a + bε p (t )
n
EQ. 5.3.10.28
σ vm − σ yield
ε& p Δt =
E
where ε& is the plastic strain rate.
p
20-jan-2009 111
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
Finally, the plastic stress is found by the method of radial return. In case of plane stress this method is
approximated because it cannot verify simultaneously the plane stress condition and the flow rule. The following
return gives a plane stress state:
σ yield el
σ ijpa = σ ij EQ. 5.3.10.30
σ vm
Iterative algorithm
Iplas=1
If flag 1 is used on card 1 of the shell property type definition, an incremental method is used. Step 5 is
performed using the incremental method described by Mendelson [1]. It has been extended to plane stress
situations. This method is more computationally expensive, but provides high accuracy on stress distribution,
especially when one is interested in residual stress or elastic return. This method is also recommended when
variable thickness is being used. After some calculations, the plastic stresses are defined as:
The value of Δε P must be computed to determine the state of plastic stress. This is done by an iterative
method. To calculate the value of Δε , the von Mises yield criterion for the case of plane stress is introduced:
P
σ x2 + σ y2 − σ xσ y + 3σ xy2 = σ yield
2
(t + Δt ) EQ. 5.3.10.35
and the values of σx, σy, σxy and σyield are replaced by their expression as a function of Δε P (EQS 5.3.8.31 to
5.3.8.34), with, for example:
σ yield (t + Δt ) = a + bε p (t + Δt )
n
EQ. 5.3.10.36
and :
ε p (t + Δt ) = ε p (t ) + Δε p EQ. 5.3.10.37
20-jan-2009 112
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
σ eq2 = A1σ xx2 + A2σ yy2 − A3σ xxσ yy + A12σ xy2 EQ. 5.3.10.38
⎡ A3 ⎤
⎢ A1 − 2 0 ⎥
with [ A] = ⎢ A2 0 ⎥
⎢ ⎥
⎢ sym A12 ⎥
⎣⎢ ⎦⎥
EQ. 5.3.10.14 is then written as:
⎡ ⎛ A3 ⎞ ⎛ A ⎞ ⎤
⎢1 + ⎜ A1 − 2 υ ⎟dR ⎜ A2υ − 3 ⎟dR
2⎠
0 ⎥
⎢ ⎝ ⎠ ⎝ ⎥
⎥ ; dR = dε
p
⎛ A3 ⎞ ⎛ A3 ⎞ E
[B] = ⎢ ⎜ A1υ − ⎟dR 1 + ⎜ A2 − υ ⎟dR
⎢ 0
⎥
⎝ 2⎠ ⎝ 2 ⎠ σ y 1−υ 2
⎢ 1−υ ⎥
⎢ 0 0 1+ A12 dR ⎥
⎢⎣ 2 ⎥⎦
Changing the stress variables to {σ }:
{σ } = [Q]{σ } EQ. 5.3.10.40
with:
⎡ A1 − A2 + C ⎤
⎢ 1 0⎥
⎛ A ⎞
⎢ − 2⎜ A2υ − 3 ⎟ ⎥
⎢ ⎝ 2 ⎠ ⎥
[Q] = ⎢⎢ A1 − A2 + C 1
⎥
0⎥ ; C = (1 − υ )( A
2
1 − A2 ) + [ A3 − ( A1 + A2 )υ ]
2 2
⎛ A ⎞
⎢ 2⎜ A1υ − 3 ⎟ ⎥
⎢ ⎝ 2 ⎠ ⎥
⎢ 0 0 1⎥
⎢ ⎥
⎣ ⎦
[ ]
The matrix B = [Q ][ B ][Q ]
−1
is diagonal:
⎡ ⎛ A υ C ⎞⎟ ⎤
⎢1 + dR⎜⎜ A2 − 3 − 0 0 ⎥
⎢ ⎝ 2 J Q ⎟⎠ ⎥
⎢ ⎛ ⎥
A υ C ⎞⎟
[B ] = ⎢⎢ 0 1 + dR⎜ A1 − 3 +
⎜ 2 J ⎟
0 ⎥ EQ. 5.3.10.41
⎥
⎝ Q ⎠
⎢ ⎥
⎢ ⎛ 1 − υ ⎞⎥
0 0 1 + dR. A12 ⎜ ⎟
⎢ ⎝ 2 ⎠⎥
⎣ ⎦
where J Q = 1 +
( A1 − A2 + C )2 is the Jacobien of [Q]. EQ. 5.3.10.40 is now written as:
⎛ A ⎞⎛ A ⎞
4⎜ A1υ − 3 ⎟⎜ A2υ − 3 ⎟
⎝ 2 ⎠⎝ 2 ⎠
20-jan-2009 113
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
This allows to give explicitly the expression of the yield surface EQ. 5.3.10.15:
( ) ( ) { } [ ]
f n +1 = σ eqn +1 − σ yn +1 = σ ∗ [B ] A [B ] σ ∗ − σ yn+1
2 2 t −t −1
{ } ( )2
=
A11 2 A22 2 A12 2
B12
B2 B3
A
σ xx + 2 σ yy + 2 σ xy + 12 σ xxσ yy − σ yn +1
B1B2
( )
2
EQ. 5.3.10.43
[ ]
with A 2×2 = [Q ]2×2 [A]2×2 [Q ]2×2 .
−t −1
f n +1
Δε np+1 = Δε np − EQ. 5.3.10.44
f n′+1
n
M x = t 2 ∑ wkM σ xkpa EQ. 5.3.11.2
k =1
Table 5.3.1 compares the coefficients used to the classical Newton quadrature in case of 3 integration points.
20-jan-2009 114
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
N 2 16 M 2
F= + 4 − σ yield
2
≤0 EQ. 5.3.11.4
t2 t
with
where
t/2 t/2
Nx = ∫σ Mx = ∫σ
pa pa
x dz x zdz
−t / 2 −t / 2
t/2 t/2
Ny = ∫ σ y dz My = ∫σ
pa pa
y zdz
−t / 2 −t / 2
t/2 t/2
N xy = ∫ σ xy dz M xy = ∫σ
pa pa
xy zdz
−t / 2 −t / 2
An extension of Iliouchine criterion for isotropic hardening is developed here. The yield surface can be
expressed as:
⎧ {N }⎫ ⎧ {N }⎫
t
f =⎨
{ M }⎬ [F ]⎨
{ M }⎬ − σ yβ
0
( )
2
=0 EQ. 5.3.11.7
⎩ ⎭ ⎩ ⎭
⎡1
⎢ h 2 [A]
1
[A]⎤⎥
⎛ h ⎞2
⎢ 2 3 ⎜⎜ sγh ⎟⎟ ⎥
⎢ ⎝ 6⎠ ⎥
with: [F ] = ⎢ ⎥ EQ. 5.3.11.8
1
⎢ sym 2
[ A] ⎥
⎢ ⎛ h2 ⎞ ⎥
⎢ ⎜⎜ γ ⎟⎟ ⎥
⎣ ⎝ 6 ⎠ ⎦
{N }t [A]{M }
and s = EQ. 5.3.11.9
{N }t [A]{M }
Where β and γ are scalar material characteristic constants, function of plastic deformation. They can be
identified by the material hardening law in pure traction and pure bending:
20-jan-2009 115
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
σ y (ε p )
In pure traction: f =
N2
h 2
(
− σ y0 β )2
=0→β =
σ y0
EQ. 5.3.11.10
In pure bending: f =
M2
(
− σ y0 β )2
=0→γ =
M
EQ. 5.3.11.11
(γh / 6)
2 2
σ y h2 / 6
3
σy /E+ ε p
If no hardening law in pure bending is used, γ is simply computed by γ = 2 varying between 1.0
σy / E +ε p
and 1.5.
The plasticity flow is written using the normality law:
⎧ {dε p }⎫ ∂f ⎧ {N }⎫
⎬ = dλ = 2dλ [F ]⎨
⎩{dΧ p }⎭
⎨
{M }⎬⎭
EQ. 5.3.11.12
⎧ {N }⎫ ⎩
∂⎨ ⎬
⎩{M }⎭
The equivalent plastic deformation ε P is proportional to the plastic work. Its expression is the same as in the
case of traction:
⎧{dε p }⎫ ⎧ {N }⎫
t
σ βdε = ⎨
0
y
{dχ
p
}⎬⎨
{ M }⎬ = 2dλ σ y β
0
( )2
EQ.5.3.11.13
⎩ p ⎭⎩ ⎭
This leads to:
H
dε p = 2dλσ y0 β and dβ = dε p EQ. 5.3.11.14
σ 0
y
where H is the plastic module. The derivative of function f in EQ. 5.3.11.7 is discontinuous when
{N }t {A}{M } =0. This can be treated when small steps are used by putting s=0 as explained in [87].
∂f
Then the derivative of f with respect to dε P ( ) is carried out. The derived equation is nonlinear in
∂dε P
internal efforts and is resolved by Newton-Raphson:
⎧ {N }⎫ ⎧ {N }⎫ ⎧ {N }⎫
⎨ ⎬ =⎨ ⎬ − 2dλ [D ][F ]⎨ ⎬ EQ. 5.3.11.15
⎩{M }⎭n +1 ⎩{M }⎭∗ ⎩{M }⎭n +1
⎧ {N }⎫ ⎧ {N }⎫ ⎧{dε }⎫
⎨ ⎬ =⎨ ⎬ − [D ]⎨ ⎬ EQ. 5.3.11.16
⎩{M }⎭* ⎩{M }⎭n ⎩{dχ }⎭
20-jan-2009 116
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
A - Notational conventions
• A bold letter denotes a vector or a tensor.
• An upper case index denotes a node number; a lower case index denotes a component of vector or
tensor.
• The Einstein convention applies only for the repeated index where one is subscript and another is
superscript, e.g.:
NI xI = ∑ NI xI .
• {} denotes a vector and [ ] denotes a matrix.
20-jan-2009 117
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
3
4
η Z
ζ
ξ Y
r X
n
1 2
a
Mid-plan
The geometry of the 4-node degenerated shell element, as shown in the Fig 5.3.11., is defined by its mid-surface
I
with coordinates denoted by Xp interpolated by the node coordinates X (I=1,4):
Xp(ξ ,η ) = N I X I EQ. 5.3.12.1
A generic point q within the shell is derived from point p on the mid-surface and its coordinate along the normal
(fiber):
Xq (ξ ,η , ζ ) = Xp + z ⋅ n with z = aζ
2 EQ. 5.3.12.3
with
(
v q (ξ ,η , ζ ) = v p + zβ = N I (ξ ,η ) v I + zβ I ) EQ. 5.3.12.5
β =ω×n
where β and ω are the rotational velocity vectors of the normal: β = β1 t 1 + β 2 t 2 = ω 2 t 1 − ω1 t 2
and (t1 , t 2 , n ) is base of the local coordinate system.
( )
v i = N I v iI + z − N Iω1I t I2i + N Iω 2I t 1iI ; i = 1,3
This velocity interpolation is expressed in the global system, but ω I must be defined first in the local nodal
coordinate system to ensure Mindlin's kinematic condition.
20-jan-2009 118
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
C - Strain-rate construction
The in-plane rate-of-deformation is interpolated by the usual displacement method.
( )
The rate-of-deformation tensor (or velocity-strains) D = L t + L t / 2 is defined by the velocity gradient tensor
T
L:
[ ] [
with Q = {t1}{t2 }{n} . ]
The Reissner-Mindlin conditions ε ζ = 0 and σ ζ = 0 requires that the strain and stress tensors are computed in
the local coordinate system (at each quadrature point).
After the linearization of Lt with respect to z, the in-plane rate-of-deformation terms are given by:
where the contravariant vectors g (α = 1,2 ) , dual to g α , satisfy the orthogonality condition: g ⋅ g β = δ β
α α α
2⎥
and [C1 ] = ⎢
t 2 ⋅ v p ,η ⎥⎦
EQ. 5.3.12.8
⎣ t1 ⋅ g t2 ⋅ g ⎦ ⎣t 2 ⋅ v p ,ξ
The curvature-translation coupling is presented in the bending terms for a warped element (the first two terms in
last Eq.)
The out-plane rate-of-deformation (transverse shear) is interpolated by the “assumed strain” method, which is
based on the Hu-Washizu variational principle.
If the out-plane rate-of-deformation is interpolated in the same manner for a full integration scheme, it will lead
to “shear locking”. It is known that the transverse shear strains energy cannot vanish when it is subjected to a
constant bending moment. Dvorkin-Bathe's [32] mix/collocation method has been proved very efficient in
overcoming this problem. This method consists in interpolating the transverse shear from the values of the
covariant components of the transverse shear strains at 4 mid-side points. i.e.:
[
γ ξ = (1 − η )γ ξ A1 + (1 + η )γ ξ A 2 / 2 ] EQ. 5.3.12.9
γη = [(1 − ξ )γ + (1 + ξ )γ η ]/ 2
B1 B2
η EQ. 5.3.12.10
Aα Bα
where γ ξ ,γη are the values of the covariant components at 4 mid-side points which vanish under a constant
bending moment (see Figure 5.3.12).
20-jan-2009 119
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
A2
γξ
4
A2
B1 γη 3
γη
γξ
B1 B2 B2
γη
A1
1 2
A1
γξ
The quadrature point is often chosen at (ξ = 0,η = 0) . The derivatives of the shape functions are:
N I ,ξ = (ξ I + hIη ) / 4 N I ,η = (η I + hI ξ ) / 4 EQ. 5.3.12.11
where hI = (1 − 1 1 − 1) .
This implies that all the terms computed at the quadrature point are the constant parts (with respect to (ξ ,η ) , and
the stabilizing terms (hourglass) are the non-constant parts.
The constant parts can be derived directly from the general formulations at the quadrature point without
difficulty. The difficulties in stabilization lie in correctly computing the internal force vector (or stiffness
matrices):
int
It would be ideal if the integration term f stab could be evaluated explicitly. But such is not the case, and the main
obstacles are the following:
For a non-coplanar element, the normal varies at each point so that it is difficult to write the non-constant part of
strains explicitly. For a physically nonlinear problem, the non-constant part of stress is not generally in an
explicit form. Thus, simplification becomes necessary.
The main modifications with respect to Q4γ24 shell element [40] are the following:
• Co-rotational coordinate system is used and the stresses are evaluated in 4 local systems at each Gauss
points..
20-jan-2009 120
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
A - Kinematic approximation
I I
The velocity interpolation using the nodal tangent vectors ( t1 , t 2 ) complicates the strain computation, especially
for transverse shear which is used mainly as a penalty function. To be consistent with the one-point quadrature
approach, the kinematic approximation is performed by:
[ (
vi = N I viI + z − δ i 2ω1I + δ i1ω2I )] EQ. 5.3.12.13
The projection consists in eliminating the nodal drilling rotations in order to reinforce Mindlin's kinematic
condition at the nodes. It has been pointed out [18] that without this projection, the element is too flexible and
cannot pass the Twisted Beam test.
This projection has a drawback of changing the invariance property to rigid body motion, i.e.: an warped element
being invariant to rigid body rotation will now strains under to rigid body rotation if the drill projection is
applied. To overcome this problem, a full projection proposed by Belytschko-Leviathan [66] which free either
drilling rotation and rigid motion should be used. This full projection is only used for QEPH element.
with: [ ( ) ( ) ]
Δ I = t I − t I x I bxI − t I y I b yI ; t I = (1,1,1,1)
A: area of element
The derivation of the shape functions is given by:
N I ,α = bαI + γ I φ,α (α = x, y )
where:
φ , x = ηξ , x + ξη , x = J −1η + J 12−1ξ
11
EQ. 5.3.12.16
φ , y = ηξ , y +ξη , y = J 21−1η + J 22−1ξ
The advantage of this shape function form is that a linear field expressed with Cartesian coordinates and a
bilinear field expressed with Natural coordinates are decomposed so that the constant part is directly formulated
with the Cartesian coordinates, and the non-constant part is to be approached separately.
20-jan-2009 121
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
with:
The development of the general formulations leads to the constant part, denoted by superscript 0, of the matrix
[BI]:
with
term of [(B ) ]
b 0
with ε = 2 x34 / y34 (f = (f
ij i + f j )/ 2 ) , which vanishes when the element is rectangular.
Thus, this term is not used in the program.
The constant part of the in-plane rate-of-deformation formulation without the H term is consistent with the result
of Belytschko's family shell element [24], [17], though this part has been obtained in a very different manner.
Letellier has given the same result in his thesis [43], and studies were also made of the quadratic terms with
respect to z.
Non-constant part
The main simplification for the non-constant part formulation, in order to overcome the difficulties described
above, is the following:
The element is considered to be flat.
In this case, the Jacobian matrix is written as:
⎡ x, ξ y, ξ 0 ⎤
⎡[J ] 0 ⎤
[J ] = [F0 ] = ⎢⎢ x,η 0 ⎥⎥ = ⎢
t
y,η ⎥ EQ. 5.3.12.19
0 aJ 0 / 2 J ⎦
⎢⎣ 0 0 aJ 0 / 2 J ⎥⎦ ⎣
and:
[( )( ) ( )( )]
J 0 = ξ I x I η I y I − η I x I ξ I y I / 16 = A / 4
20-jan-2009 122
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
⎡γ I φ , x 0 0⎤ ⎡ 0 γ Iφ, x ⎤
[(B ) ]
m H
I
⎢
=⎢ 0 γ Iφ, y
⎥
0⎥ ; [(B ) ]
b H
I
⎢
= ⎢− γ I φ , y 0 ⎥
⎥
EQ. 5.3.12.22
⎢γ I φ , y γ Iφ, x 0⎥⎦ ⎢− γ I φ , x γ I φ , y ⎥⎦
⎣ ⎣
It is shown in [16] that the non-constant part of membrane rate-strain does not vanish when a warped element
undergoes a rigid body rotation. Thus, a modified matrix BI [( )
m H
] is chosen:
⎡γ I φ , x 0 z γ b xI φ , x ⎤
[(B ) ]
m H
I
⎢
=⎢ 0 γ Iφ, y z γ b yI φ , y
⎥
⎥ EQ. 5.3.12.23
⎢γ I φ , y
⎣ γ Iφ, x z γ (b xI φ , y +b yI φ , x )⎥⎦
This matrix is different from the Belytschko-Leviathan correction term added at rotational positions, which
couples translations to curvatures as follows:
⎡ 1 ⎤
⎢γ I φ , x 0 0 0 −
4
zγ φ , x ⎥
[(B ) ]
m H
I
⎢
=⎢ 0
⎢
γ Iφ, y 0
1
4
zγ φ , y 0
⎥
⎥
⎥
EQ. 5.3.12.24
⎢γ φ , 1 1
γ Iφ, x 0 zγ φ , x − zγ φ , y ⎥
⎢⎣ I y 4 4 ⎥⎦
This will lead to 'membrane locking' (the membrane strain will not vanish under a constant bending loading).
According to the general formulation, the coupling is presented in the bending terms not in the membrane terms,
yet the normal translation components in (BIm)i3 do not vanish for a warped element due to the tangent vectors
ti (ξ ,η ) which differ from ti (0,0).
⎧ v zI ⎫
⎧ xz ⎫
D ⎪ I⎪
⎨ ⎬ = [BIc ]⎨ϖ x ⎬ EQ. 5.3.12.25
⎩ D yz ⎭ ⎪ϖ I ⎪
⎩ y⎭
[ ](ζ ,η ) = [BIcη ](η ) + [BIcξ ](ξ )
where: BIc
1 ⎡ 4η I y ξ I
I
− (ξ J + hJη I )η k y k y J (ξ J + hJη I )ηk y k x J ⎤
(B )
η 0
= ⎢ ⎥
Ic
16 A ⎣− 4η I x I ξ I (ξ J + hJη I )ηk x k y J − (ξ J + hJη I )η k x k x J ⎦
1 ⎡− 4ξ I y η I
I
(η J + hJ ξ I )ξ k y k y J − (η J + hJ ξ I )ξ k y k x J ⎤
(B )
ξ 0
= ⎢
(η J + hJ ξ I )ξ k x k x J ⎥⎦
EQ. 5.3.12.26
− (η J + hJ ξ I )ξ k x k y J
Ic
16 A ⎣ 4ξ I y Iη I
( )
and BIc
η H
( ) η ; (B )
= η I BIc
η 0
Ic
ξ H
( )ξ
= ξ I BIc
ξ 0
20-jan-2009 123
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
ξ, x (ξ ,η ) = ξ, x (0,0)η, x (ξ ,η ) = η, x (0,0)
which is true for a parallelogram element. Although this simplification is not necessary, it is justified by the fact
that the transverse shear terms serve mainly as a penalty function.
{f } = {( f ) }+ {( f ) }
I
int
I
int 0
I
int H
with {( f ) } the constant part being computed with one-point quadrature, and
int 0
I
It can be shown in the last equation that only the following scalar functions need to be integrated:
H xx = ∫ Aeφ, xφ, x dAe , H xy = ∫ Aeφ, xφ, y dAe , H yy = ∫ Aeφ, yφ, y dAe EQ. 5.3.12.27
⎧⎪q& b = γ Iϖ yI
bending : ⎨ xb EQ. 5.3.12.29
⎪⎩q& y = γ Iϖ x
I
[( ) ( )] [( ) (
⎧⎪q& s = 4hI vzI − ξ I y I η I + hI y I t I ϖ xI + ξ I x I η I + hI x I t I ϖ yI
shear : ⎨ xs
)]
[( ) ( )] [( )
⎪⎩q& y = 4hI vz − η I y ξ I + hI y t I ϖ x + η I x ξ I + hI x t I ϖ y
I I I I I I
( I
)] EQ. 5.3.12.30
20-jan-2009 124
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
(q )n +1 = (q )n + q&n + 1 Δt
2
Noting that if φ ,α is considered as constant over a time step, it is equivalent to the incremental stress
computation.
{f } = ∫ [B ] {σ }dv
I
int
ve
I
t e
The constitutive relation is written by either a tangent form: σ& = C t:D , or a projection form: σ = P (σ e ,...) ,
t
( )
where C σ ,... is the history-dependent tangent tensor ; {σ }= {σ }+ [C ]{D }Δt is the trial stress,
e
n +1 n n +1/ 2
and the function P consists of projecting the trial stress on the updated yield surface.
The decomposed form of the last Equation is written:
{f } = {( f ) }+ {( f )
I
int
I
int 0
I
int H
}
The constant part is computed by integration at each integration point through the thickness.
The stabilization part will be approached by relying on two hypotheses:
• Keep the same orthogonalities as in the elastic case, and
Use the unidimensional tangent modulus Et (ζ ) to evaluate the non-constant rate-stress, i.e.:
{σ& }H = (Et (ζ ) / E )[C ]{D}H EQ. 5.3.12.31
[ ]
where E is the Young's modulus and C is the matrix of elastic moduli.
where Et (ζ ) is obtained by the constant stress incremental computation along the thickness and Et (ζ ) = E in
the elastic zone, Et λm = Et / E; Et is the average value of Et (ζ ) and λb = min Et (ζ ) / E .
For the QPH shell, λm = 1
We have maintained the key orthogonalities without any significant deterioration in performance, although the
first two orthogonalities might have been slightly violated. In fact, it is simply due to these orthogonalities that a
one-point quadrature element dramatically reduces the computation cost; otherwise we return to the full
integration scheme.
Most of the physical stabilization elements have incorporated the following assumption:
The material response is constant within the element.
There are two alternatives to this assumption
20-jan-2009 125
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
Where α = m, b corresponds to the membrane and bending terms respectively. Note that the shear terms are
eliminated to avoid shear locking. The transverse shear terms can also be written as the same way:
⎧τ& xη ⎫⎪ ⎧⎪τ& xξ ⎫⎪
{τ&i }H = η ⎪⎨ ⎬+ξ⎨ ⎬ EQ. 5.3.12.36
⎪⎩τ& yη ⎪⎭ ⎪⎩τ& yξ ⎪⎭
We can now redefine 12 generalized hourglass stresses by integrating their rate ones, and the stress field can be
expressed by:
⎧σ xαη ⎫ ⎧σ xαξ ⎫
Membrane, bending : {σ } = {σ } + {σ } = {σ }
α
i
α 0
i
α H
i
α 0
i
⎪ α ⎪ ⎪ ⎪
+ η ⎨σ yη ⎬ + ξ ⎨σ αyξ ⎬
⎪ 0 ⎪ ⎪ 0 ⎪
⎩ ⎭ ⎩ ⎭
⎧τ& xη ⎫⎪ ⎧⎪τ& xξ ⎫⎪
Shear: {τ i } = {τ i }0 + {τ i }H = {τ i }0 + η ⎪⎨ ⎬+ξ⎨ ⎬
⎪⎩τ& yη ⎪⎭ ⎪⎩τ& yξ ⎪⎭
Even the redefinition for shear is not necessary as it is not included in the plastic yield criterion, but the same
stress calculation as the constant part with the updated Lagrangian formulation is always useful when large strain
is involved.
20-jan-2009 126
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
for any point in the solid element, where σ y is evaluated at the quadrature point.
As only one criteria is used for the non-constant part, two choices are possible:
• taking the value by some representative points, e.g. eight Gausse points
The second choice has been used in this element.
Elasto-plastic hourglass stress calculation:
The incremental hourglass stress is computed by:
• Elastic increment
(σ i )trH = (σ i )n + [C ]{ε&} Δt
H H
n +1
20-jan-2009 127
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
g1
t1 = EQ. 5.3.13.2
g1
Hence, the vector defining the local y direction is found from the cross product of the two previous vectors:
t 2 = n × t1 EQ. 5.3.13.3
φ1 = a1 + b1 x + c1 y EQ. 5.3.13.5
φ2 = a2 + b2 x + c2 y EQ. 5.3.13.6
φ3 = a3 + b3 x + c3 y EQ. 5.3.13.7
These shape functions are used to determine the velocity field in the element:
3
vx = ∑ φ I v xI EQ. 5.3.13.8
I =1
3
v y = ∑ φI v yI EQ. 5.3.13.9
I =1
3
vz = ∑ φI v zI EQ. 5.3.13.10
I =1
3
ω x = ∑ φI ω xI EQ. 5.3.13.11
I =1
3
ω y = ∑ φI ω yI EQ. 5.3.13.12
I =1
∂v x 3
∂φ
= ∑ I vxI EQ. 5.3.13.13
∂x I =1 ∂x
∂v x 3
∂φ
= ∑ I vxI EQ. 5.3.13.14
∂y I =1 ∂y
20-jan-2009 128
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
{v}b = {ω1y ,−ω1x , ω y2 ,−ω x&2ω y3 ,−ω x3 , v1z , vz2 , vz3 } EQ. 5.3.13.20
where:
⎡ ∂φ1 ∂φ2 ⎤
⎢ ∂x 0 0 0 0 0 0 0 ⎥
∂x
⎢ ∂φ1 ∂φ2 ∂φ3 ⎥
⎢ 0 0 0 0 0 0 ⎥
⎢ ∂y ∂y ∂y ⎥
[B] = ⎢ ∂φ1 ∂φ1 ∂φ2 ∂φ2 ∂φ3 ∂φ3
⎢ ⎥
0 0 0 ⎥ EQ. 5.3.13.22
⎢ ∂y ∂x ∂y ∂x ∂y ∂x ⎥
⎢φ ∂φ1 ∂φ2 ∂φ3 ⎥
0 φ2 0 φ3 0
⎢ 1 ∂x ∂x ∂x ⎥
⎢ ∂φ1 ∂φ2 ∂φ3 ⎥
⎢ 0 φ1 0 φ2 0 φ3 ⎥
⎣ ∂y ∂y ∂y ⎦
20-jan-2009 129
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
where:
ρ is the material density,
t is the shell thickness,
A is the reference plane surface area.
The mass moment of inertia about all axes is the same:
⎛ 2A t 2 ⎞
I xx = m⎜⎜ + ⎟⎟ EQ. 5.3.13.24
⎝ 6 12 ⎠
I zz = I yy = I xx EQ. 5.3.13.25
I xy = 0 EQ. 5.3.13.26
When nodal masses need to be calculated, the distribution is determined by the shape of the element as shown in
Figure 5.3.14.
Figure 5.3.14 - Mass distribution
α3
α1 α2
2
1
20-jan-2009 130
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
⎡c − s ⎤
⎢ s c ⎥ with c =cosθ and s = sinθ.
⎣ ⎦
⎡ε 11 ε 12 ⎤ ⎡ c s ⎤ ⎡ ε X ε XY ⎤ ⎡c − s ⎤
⎢ε ⎥=⎢ ⎥⋅⎢ ⋅
ε Y ⎥⎦ ⎢⎣ s c ⎥⎦
EQ. 5.3.14.1
⎣ 12 ε 22 ⎦ ⎣− s c ⎦ ⎣ε XY
⎡ε11 ε12 ⎤ ⎡ c s ⎤ ⎡c ⋅ ε X + s.ε XY − s ⋅ ε X + c.ε XY ⎤
⎢ε ⎥=⎢ ⎥⋅⎢ − s ⋅ ε XY + c.ε Y ⎥⎦
EQ. 5.3.14.2
⎣ 12 ε 22 ⎦ ⎣− s c ⎦ ⎣ c ⋅ ε XY + s.ε Y
⎡ε11 ε12 ⎤ ⎡ c 2 ⋅ ε X + 2cs.ε XY + s 2 .ε Y ( )
− cs ⋅ ε X + c 2 − s 2 .ε XY + cs.ε Y ⎤
⎥=⎢
( ) ⎥
⎢ε EQ. 5.3.14.3
⎣ 12 ε 22 ⎦ ⎣− cs ⋅ ε X + c − s .ε XY + cs.ε Y s 2 .ε X − 2cs.ε XY + c 2 .ε Y
2 2
⎦
The strain-stress relation in orthotropy directions is written as:
σ = σ 11 σ 22 σ 12 ; ε = ε 11 ε 22 ε 12 EQ. 5.3.14.6
⎡σ XX σ XY ⎤ ⎡c − s ⎤ ⎡σ 11 σ 12 ⎤ ⎡ c s ⎤
⎢σ = ⋅ ⋅ EQ. 5.3.14.7
⎣ XY σ YY ⎥⎦ ⎢⎣ s c ⎥⎦ ⎢⎣σ 12 σ 22 ⎥⎦ ⎢⎣− s c ⎥⎦
20-jan-2009 131
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
20-jan-2009 132
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
The shell normal defines the positive direction for φ . For elements with more than one layer, multiple φ angles
can be defined.
The fiber direction orientation may be updated by two different ways:
1. constant orientation in local corotational reference frame,constant orientation in local isoparametric
frame.The first formulation may lead to unstable models especially in the case of very thin shells (e.g.
airbag modelling). In Figure 5.3.17 the difference between the two formulations are illustrated for the case of
element traction and shearing.
Figure 5.3.17 - Fiber Direction Updating
Formulation 1
Shear
Formulation 2
Formulation 1 F α
F
Traction α
Formulation 2
20-jan-2009 133
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
20-jan-2009 134
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
5.3.14.7 Limitations
When modeling a composite material, there are two strategies that may be applied. The first, and simplest, is to
model the material in a laminate behavior. This involves using type 9 property shell elements. The second is to
model each ply of the laminate using one integration point. This requires either a type 10 or 11 element.
Modeling using the type 9 element allows global behavior to be modeled. Input is simple, with only the
reference vector as the extra information. A Tsai-Wu yield criterion and hardening law is easily obtained from
the manufacturer or a test of the whole material.
Using the type 10 or 11 element, one models each ply in detail, with one integration point per ply and tensile
failure is described in detail for each ply. However, the input requirements are complex, especially for the type
11 element.
Delamination is the separation of the various layers in a composite material. It can occur in situations of large
deformation and fatigue. This phenomenon cannot be modeled in detail using shell theory. A global criterion is
available in material law 25. Delamination can affect the material by reducing the bending stiffness and buckling
force.
hk hm
θk = and θm = EQ.5.3.15.1
2R 2R
However, for total rotation we have:
θk + θm = α k + α m EQ.5.3.15.2
θk =
(α k + α m )hk and θm =
(α k + α m )hm EQ.5.3.15.3
(hk + hm ) (hk + hm )
20-jan-2009 135
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
hm m
k
hk
i
O
θk θm
αm
αk
i, j
Consider the triangle element in Figure 5.3.19. The outward normal vectors at the three sides are defined and
denoted n1, n2 and n3. The normal component strain due to the bending around the element side is obtained
using plate assumption:
⎧ α1 ⎫
⎪α ⎪
⎧ε n1 ⎫ ⎡ (h1 +2 h5 ) 0 0 0 2
( h1 + h5 ) 0 ⎤⎪ 2 ⎪
⎪ ⎪ ⎢ ⎥ ⎪⎪α 3 ⎪⎪
⎨ε n 2 ⎬ = ⎢ 0 0 0 0
2 2
( h 2 + h6 ) ( h 2 + h6 ) ⎥ ⎨ ⎬ EQ.5.3.15.4
⎪γ ⎪ ⎢ 0 α
⎩ n3 ⎭ ⎣ 0 2
( h3 + h 4 )
2
( h3 + h 4 ) 0 0 ⎥⎦ ⎪ 4 ⎪
⎪α 5 ⎪
⎪ ⎪
⎩⎪α 6 ⎭⎪
The six mid-side rotations α i are related to the out-of-plane displacements of the six apex nodes as shown in
Figure 5.3.20 by the following relation:
20-jan-2009 136
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
where (h1 h2 h3 ) , (q1 h4 q2 ) , (r2 h5 r3 ) and (s3 h6 s1 ) are respectively the heights of the
triangles (1,2,3), (1,4,2), (2,5,3) and (3,6,1).
n1 2
n3
3
n2
1
The non-null components of strain tensor in the local element reference are related to the normal components of
strain by the following relation (see [40] and [106] for details):
⎡ ⎛ y3 − y 2 ⎞
2
⎛ y1 − y3 ⎞
2
⎛ y2 − y1 ⎞
2
⎤
⎢ ⎜⎜ ⎟⎟ ⎜⎜ ⎟⎟ ⎜⎜ ⎟⎟ ⎥
⎢ ⎝ l1 ⎠ ⎝ l2 ⎠ ⎝ l3 ⎠ ⎥
⎧ε xx ⎫ ⎢ 2 2 2 ⎥ ⎧ε n1 ⎫
⎪ ⎪ ⎢ ⎛ x2 − x3 ⎞ ⎛ x3 − x1 ⎞ ⎛ x1 − x2 ⎞ ⎥ ⎪⎨ε n 2 ⎪⎬ EQ.5.3.15.6
⎨ε yy ⎬ = ⎜⎜ ⎟⎟ ⎜⎜ ⎟⎟ ⎜⎜ ⎟⎟
⎪γ ⎪ ⎢ ⎝ l1 ⎠ ⎝ l2 ⎠ ⎝ l3 ⎠ ⎥⎪ ⎪
⎩ xy ⎭ ⎢ ⎛ y − y ⎞⎛ x − x ⎞ ⎛ y − y ⎞⎛ x − x ⎞ ⎛ y − y ⎞⎛ x − x ⎞⎥ ⎩γ n 3 ⎭
⎢2⎜ 3 2
⎟⎜ 2 3 ⎟ 2⎜ 1 3 ⎟⎜ 3 1 ⎟ 2⎜ 2 1 ⎟⎜ 1 2 ⎟⎥
⎢ ⎜⎝ l1 ⎟⎠⎜⎝ l1 ⎟⎠ ⎜⎝ l2 ⎟⎠⎜⎝ l2 ⎟⎠ ⎜⎝ l3 ⎟⎠⎜⎝ l3 ⎟⎠⎥
⎣ ⎦
20-jan-2009 137
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
5
2
ψ2
β2 γ2
4
ψ3
β3
3
ϕ3
β1 γ1
ϕ1 1
α k = −α m EQ.5.3.15.7
[cos β 2
h1
cos β 1
h2
− h13 0 0 0 ] EQ.5.3.15.8
The clamped condition is introduced by the symmetry in out-of-plane displacement i.e. wm=wk. This implies
α k = α m . The fourth row of the matrix in EQ.5.3.15.5 is then changed to:
[− cos β 2
h1
− cosh2β1 − h13 0 0 0 ] EQ.5.3.15.9
20-jan-2009 138
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
4
5
2
20-jan-2009 139
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
The derivation of solid-shell elements is more complicated than that of standard solid elements since they are
prone to the following problems:
• Shear and membrane locking with the hybrid strain formulation [89], [90], the hybrid stress [91], and
the Assumed and Enhanced Natural Strain formulations [92], [93], [94], [95].
• Trapezoidal locking caused by deviation of mid-plane from rectangular shape [8].
• Thickness locking due to Poisson’s ratio coupling of the in-plane and transverse normal stresses [89],
[90], [92], [94].
Solid shell elements in RADIOSS are the solid elements with a treatment of the normal stresses in the thickness
direction. This treatment consists of ensuring constant normal stresses in the thickness by a penalty method.
Advantage of this approach with respect to the plane-stress treatment is that it can simulate the normal
deformability and exhibits no discernible locking problems. The disadvantage is its possible small time step
since it is computed as solid element and the characteristic length is determined often using the thickness.
20-jan-2009 140
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
by Gauss-Lobatto schemes rise up to 9 integrations to enhance the quality of elasto-plastic behavior. The in-
plane integration may be done by a reduced 2x2 scheme or a fully integrated 3x3 points (Figure 5.4.3). A
reduced integration method is applied to the normal stress in order to avoid locking problems.
8 15 7
16
5 14 3
13 6
10
1 9 2
The distribution of mass is not homogenous over the nodes. The internal nodes receive three times more mass
than the corner nodes as shown in Figure 5.4.4.
s
r
t
s
t
t
(a) Two to nine integration points (b) In-plane 2x2 and 3x3 integration
through the thickness schemes
20-jan-2009 141
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
3/32 Mo
1/32 Mo
5.5.2 Stability
Determining the stability of truss elements is very simple. The characteristic length is defined as the length of the
element, i.e. the distance between N1 and N2 nodes.
L(t )
Δt ≤ EQ. 5.5.2.1
C
E
Where, L (t ) is the current truss length and C = is the sound speed.
ρ
20-jan-2009 142
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
5.5.4 Strain
The strain rate, as shown in Figure 5.5.2, is defined as:
( )
r r
∂v x ∂ V ⋅ ex
ε&x = = EQ. 5.5.4.1
∂x ∂x
A generalized force-strain graph can be seen in Figure 5.5.3. The force rate under elastic deformation is given
by:
where:
E is the Elastic Modulus,
A is the cross sectional area.
In the plastic region, the force rate is given by:
20-jan-2009 143
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
l
log
gap l0
In a general case, it is possible to introduce a gap distance in the truss definition. If gap is not null, the truss is
activated when the length of the element is equal to the initial length minus the gap value. This results a force-
strain curve shown in Figure 5.5.3(b).
20-jan-2009 144
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
20-jan-2009 145
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
The moments of inertia about the y and z axes are concerned with bending. They can be calculated using the
relationships:
The moment of inertia about the x axis concerns torsion. This is simply the summation of the previous two
moments of Ontario:
Ix = Iy + Iz EQ. 5.6.2.3
AL2
b=
max(I y , I z )
The force rate vector for an element is calculated using the relation:
⎡ F&x1 ⎤ EA ⎡+ 1 − 1⎤ ⎡υ x1 ⎤
⎢& ⎥= EQ. 5.6.4.1
⎣ Fx 2 ⎦ l ⎢⎣ − 1 + 1⎥⎦ ⎢⎣υ x 2 ⎥⎦
20-jan-2009 146
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
where:
E is the elastic modulus,
l is the beam element length,
With the force rate equation, the force vector is determined using explicit time integration:
5.6.4.2 Torsion
Torsional deformation occurs when the beam is loaded with a moment M about the X axis as shown in Figure
5.6.3.
Figure 5.6.3 Torsional Loading
⎡ M& x1 ⎤ GI x ⎡+ 1 − 1⎤ ⎡θ&x1 ⎤
⎢ & ⎥= ⎢ − 1 + 1⎥ ⎢ & ⎥ EQ. 5.6.4.3
⎣M x 2 ⎦ l ⎣ ⎦ ⎣θ x 2 ⎦
where:
G is the modulus of rigidity,
20-jan-2009 147
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
Two vector fields must be solved for forces and moments. The rate equations are:
⎡ v y1 ⎤
⎡ F& y1 ⎤ EI z ⎡+ 12 6l − 12 + 6l ⎤ ⎢⎢θ&z1 ⎥⎥
⎢& ⎥= 3
⎢⎣ Fy 2 ⎥⎦ l (1 + φ y ) ⎣ − 12 − 6l
⎢ EQ. 5.6.4.5
+ 12 − 6l ⎥⎦ ⎢v y 2 ⎥
⎢& ⎥
⎣θ z 2 ⎦
⎡ v y1 ⎤
⎡ M& z1 ⎤ EI z ⎡+ 6l (4 + φ )l
y
2
− 6l (2 − φ y )l ⎤ ⎢⎢θ&z1 ⎥⎥
2
⎢ & ⎥= 3 ⎢ ⎥
⎣ M z 2 ⎦ l (1 + φ y ) ⎢⎣+ 6l (2 − φ )l
y
2
− 6l (4 + φ y )l 2 ⎥⎦ ⎢v y 2 ⎥
EQ. 5.6.4.6
⎢& ⎥
⎣θ z 2 ⎦
144(1 + v )I z
where φy = ,
5 Al 2
ν is the Poisson's Ratio.
⎡ν z1 ⎤
⎢ ⎥
&
⎡ Fz1 ⎤ El y ⎡+ 12 6l − 12 + 6l ⎤ ⎢θ&y1 ⎥
⎢& ⎥ 3 = ⎢ ⎥ EQ. 5.6.4.9
⎣ Fz 2 ⎦ l (1 + Φ z ) ⎣ − 12 − 6l + 12 − 6l ⎦ ⎢ν z 2 ⎥
⎢& ⎥
⎣⎢θ y 2 ⎦⎥
20-jan-2009 148
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
⎡ν z1 ⎤
⎢ ⎥
⎡ M& y1 ⎤ El y ⎡+ 6l (4 + Φ z )l − 6l (2 − Φ z )l ⎤ ⎢θ&y1 ⎥
2 2
⎢ & ⎥= 3 ⎢ 2⎥
EQ. 5.6.4.10
⎢⎣ M y 2 ⎥⎦ l (1 + Φ z ) ⎣+ 6l (2 − Φ z )l − 6l (4 + Φ z )l ⎦ ⎢ν z 2 ⎥
2
⎢& ⎥
⎢⎣θ y 2 ⎥⎦
where:
144(1 + ν )I y
Φz = .
5 Al 2
Like bending about the Z axis, the factor Φ z introduces transverse shear.
• Elastic
• Elasto-plastic
An elastic beam can be defined with these four parameters. For accuracy and stability, the following limitations
should be respected:
12 I y I z = A4 EQ. 5.6.5.5
Ix = Iy + Iz EQ. 5.6.5.6
20-jan-2009 149
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
However, this model also gives good results for the circular or ellipsoidal cross-section. For tubular or H cross-
sections, plasticity will be approximated.
Recommendations:
⎛ ε& ⎞
σ y = (A + Bε pn )⎜⎜1 + C ln ⎟ EQ. 5.6.5.11
⎝ ε&0 ⎟⎠
The increment of plastic strain is:
ΔW plastic
Δε p = EQ. 5.6.5.12
σy
The equivalent strain rate is derived from the total energy rate:
ΔWtotal
ε& = EQ. 5.6.5.13
σ eq Δt
Yield stress:
Fx2 3 ⎛⎜ M x2 M y M z2 ⎞⎟
2
σ eq = + + + EQ. 5.6.5.14
A2 A ⎜⎝ I xx I yy I zz ⎟⎠
⎛ ⎛ ρSL ⎞⎛ L2 ⎞ ⎛ ρL ⎞ ⎛ ρL ⎞ ⎞
MAX ⎜⎜ ⎜ ⎟⎜⎜ ⎟⎟ + ⎜ ⎟ • MAX (I yy ; I zz ) ; ⎜ ⎟ • I xx ⎟⎟ EQ. 5.6.6.1
⎝ ⎝ 2 ⎠⎝ 12 ⎠ ⎝ 2 ⎠ ⎝ 2 ⎠ ⎠
20-jan-2009 150
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
No material data card is required for spring elements. However, the stiffness k and equivalent viscous damping
coefficient c are required. The mass m is required if there is any spring translation.
There are three other options defining the type of spring stiffness with the hardening flag:
• Linear Stiffness
• Nonlinear Stiffness
• Nonlinear Elasto-Plastic Stiffness
20-jan-2009 151
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
m
Δt = EQ. 5.7.1.1
k
For a dashpot only element:
m
Δt = EQ. 5.7.1.2
2c
For a parallel spring and dashpot element:
Δt =
( mk + c )− c
2
EQ. 5.7.1.3
k
The critical time step ensures that the stability of the explicit time integration is maintained, but it does not
ensure high accuracy of spring vibration behavior. Only two time steps are required during one vibration period
of a free spring to keep stability. However, if true sinusoidal reproduction is desired, the time step should be
reduced by a factor of at least 5.
If the spring is used to connect the two parts, the spring vibration period increases and the default spring time
step ensures stability and accuracy.
20-jan-2009 152
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
dl
F = k (l − l0 ) + c EQ. 5.7.2.1
dt
Figure 5.7.5 Linear Force-Displacement Curve
Δt =
( c + km )− c
2
EQ. 5.7.2.2
k
The stability criterion is the same as for the linear spring, but rather than being constant, the stiffness is
displacement dependent:
Δt =
(c 2
)
+ k ′m − c
EQ. 5.7.3.2
k′
20-jan-2009 153
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
where:
⎡ ∂ ⎤
k ′ = max ⎢ f (l − l0 )⎥ EQ. 5.7.3.3
⎣ ∂ (l − l0 ) ⎦
20-jan-2009 154
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
20-jan-2009 155
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
With δ= l-l o
20-jan-2009 156
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
where:
⎡ ∂ ⎤
C ′ = max ⎢ g (dl / dt )⎥ EQ. 5.7.8.3
⎣ ∂ (dl / dt ) ⎦
⎛ dl ⎞
F = f (l − l0 )g ⎜ ⎟ EQ. 5.7.9.1
⎝ dt ⎠
Graphs of this relationship for various values of g (dl / dt ) are shown in Figure 5.7.12.
Δt =
( C′ + k′M )− C′
2
EQ. 5.7.9.2
k′
20-jan-2009 157
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
where:
⎡ ∂ ⎤
K ′ = max ⎢ f (l − l0 )⎥ EQ. 5.7.9.3
⎣ ∂ (l − l0 ) ⎦
⎡ ∂ ⎤
C ′ = max ⎢ g (dl / dt )⎥ EQ. 5.7.9.4
⎣ ∂ (dl / dt ) ⎦
20-jan-2009 158
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
20-jan-2009 159
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
The same local x direction deformation, with a moving skew frame, can be seen in Figure 5.8.2.
Figure 5.8.2 Moving Skew Frame
In both cases, the forces are in equilibrium, but the moments are not. If the first two nodes defining the moving
skew system are the same nodes as the two spring element nodes, the behavior becomes exactly the same as that
of a type 4 spring element. In this case the momentum equilibrium is respected and local Y and Z deformations
are always equal to zero.
Fixed Nodes
If one of the two nodes is completely fixed, the momentum equilibrium problem disappears. For example, if
node 1 is fixed, the force computation at node 2 is not dependent on the location of node 1. The spring then
becomes a spring between node 1 and the laboratory, as shown in Figure 5.8.3.
Figure 5.8.3 Fixed node - Fixed skew frame
20-jan-2009 160
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
It is generally recommended that a general spring element (type 8) be used only if one node is fixed in all
directions or if the two nodes are coincident. If the two nodes are coincident, the translational stiffness’ have to
be large enough to ensure that the nodes remain near coincident during the simulation.
θ i = θ i 2 − θ i1 EQ. 5.8.11.2
where:
C is the equivalent viscous damping coefficient
20-jan-2009 161
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
⎛ ⎛ u& ⎞ ⎞
Fi = f i (ui )⎜⎜ A + B ln⎜⎜ i ⎟⎟ + g (u&i )⎟⎟ + Ci u&i EQ. 5.8.12.3
⎝ ⎝D⎠ ⎠
where:
5.8.13 Moments
Moments can be applied to a general spring element, as shown in Figure 5.8.5. For each DOF (i = x, y, z), the
moment is calculated by:
where:
C is the equivalent viscous damping coefficient
20-jan-2009 162
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
⎛ ⎛ θ& ⎞ ⎞
M i = f (θ i )⎜ A + B ln⎜⎜ i
⎜ ⎟ ( )⎟⎟ + C θ&
⎟ + g θ&i i i EQ. 5.8.13.3
⎝ ⎝D ⎠ ⎠
where:
Where :
Dyp
Dxn Dxp
Dyn
20-jan-2009 163
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
Δt =
( 2KM + C )− C2
EQ. 5.9.1.1
2K
This is the same as for type 4 spring elements, except that the stiffness is replaced with twice the stiffness to
ensure stability with high friction coefficients.
20-jan-2009 164
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
⎛ dδ dδ ⎞
F = K (δ1 + δ 2 ) + C ⎜ 1 + 2 ⎟ EQ. 5.9.8.2
⎝ dt dt ⎠
⎛ ⎛ dδ dδ ⎞ ⎛ βμ ⎞ ⎞
ΔF = max⎜⎜ K (δ1 + δ 2 ) + C ⎜ 1 + 2 ⎟, F tanh⎜ ⎟ ⎟⎟ EQ. 5.9.8.3
⎝ ⎝ dt dt ⎠ ⎝ 2 ⎠⎠
F1 = F + ΔF EQ. 5.9.8.4
20-jan-2009 165
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
F3 = F − ΔF EQ. 5.9.8.5
F2 = − F1 − F3 EQ. 5.9.8.6
where:
δ1 is the elongation of strand 1-2
δ2 is the elongation of strand 2-3
20-jan-2009 166
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
Where,
max(Ct) is the maximum translational damping
max(Cr) is the maximum rotational damping
20-jan-2009 167
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
If node 3 is not defined, the local skew frame that can be specified for the element is used to define the Z axis.
The X and Y axes are defined in the same manner as before.
r r r
z = x × yskew EQ. 5.10.10.4
20-jan-2009 168
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
If no skew frame and no third node is defined, the global Y axis is used to replace the Y skew axis. If the Y skew
axis is collinear with the local X axis, the local Y and Z axes are placed in a totally arbitrary position. The local
Y axis is defined at time zero, and is corrected at each cycle, taking into account the mean X axis rotation.
5.10.12 Tension
The tension component of the beam type spring element is independent of other forces. It is shown in Figure
5.10.4. The tension at each node is computed by:
20-jan-2009 169
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
where
5.10.13 Shear - XY
Shear in the Y direction along the face perpendicular to the X axis is a combination of forces and moments. This
can be seen in Figure 5.10.5.
Figure 5.10.5 XY Shear Forces and Moments
There are two mechanisms that can cause shear. The first is the beam double bending as shown in Figure 5.10.5.
The second is shear generated by node displacement, as shown in Figure 5.10.6, where node 2 is displaced.
Figure 5.10.6 Shear due to Node Displacement
20-jan-2009 170
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
M z 2 = M z1 EQ. 5.10.13.4
where:
⎛ θ + θ z1 ⎞
u y = u y 2 − u y1 − l ⎜ z 2 ⎟
⎝ 2 ⎠
f y (u y ) is the function defining the force-displacement relationship.
5.10.14 Shear - XZ
The XZ shear is orthogonal to the XY shear described in the previous section. The forces and moments causing
the shear can be seen in Figure 5.10.7.
M y 2 = M y1 EQ. 5.10.14.4
where
⎛ θ y 2 + θ y1 ⎞
u z = u z 2 − u z1 + l ⎜⎜ ⎟⎟
⎝ 2 ⎠
f z (u z ) is the function defining the force-displacement relationship.
5.10.15 Torsion
Torsional forces, shown in Figure 5.10.8, are calculated using the relations:
20-jan-2009 171
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
M x 2 = − M x1 EQ. 5.10.15.2
where:
M y 2 = − M y1 EQ. 5.10.16.2
where:
M z 2 = − M z1 EQ. 5.10.17.2
20-jan-2009 172
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
20-jan-2009 173
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
Nodes are numbered from 1 to n, and strands are numbered from 1 to n-1 (strand k goes from node Nk to node
Nk+1).
K C
Linear spring F = 0
δ + 0 δ&
L L
C &
Non linear spring F = f (ε ) ⋅ g (ε& ) + δ
L0
C &
or F = f (ε ) + δ if g function identifier is 0
L0
C &
or F = g (ε& ) + δ if f function identifier is 0
L0
L − L0
where, ε is engineering strain: ε =
L0
L0 is the reference length of element.
20-jan-2009 174
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
Assuming:
lk L
0
= 0 EQ. 5.11.2.2
lk L
where lk is the actual length of strand k.
K⎛ L ⎞
ΔFk (t ) = ΔFk (t − 1) + ⎜ δε k − δε ⎟⎟
0 ⎜
EQ. 5.11.2.3
l ⎝ lk ⎠
5.11.2.3 Friction
Friction is expressed at the nodes: if μ is the friction coefficient at node k, the pulley friction at node Nk is
expressed as:
⎛ βμ ⎞
ΔFk −1 − ΔFk ≤ (2 F + ΔFk −1 + ΔFk ) tanh⎜ ⎟ EQ. 5.11.2.4
⎝ 2 ⎠
⎛ βμ ⎞
When equation 5.11.2.4 is not satisfied, ΔFk −1 − ΔFk is reset to (2 F + ΔFk −1 + ΔFk ) tanh⎜ ⎟
⎝ 2 ⎠
All the ΔFk (k=1,n-1) are modified in order to satisfy all conditions upon ΔFk −1 − ΔFk (k=2,n-1), plus the
following condition on the force integral along the multistrand element:
∑ l (F + ΔF ) = LF
k =1, n −1
k k EQ. 5.11.2.5
This process could fail to satisfy equation 5.11.2.4 after the ΔFk ( k = 1, n − 1) modification, since no iteration is
made. However, in such a case one would expect the friction condition to be satisfied after a few time steps.
NOTE: Friction expressed upon strands (giving a friction coefficient μ along strand k) is related to pulley
friction by adding a friction coefficient μ/2 upon each nodes Nk and Nk+1 .
20-jan-2009 175
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
Ck2 + ρlk K k − Ck
Δt ≤ , ∀k EQ. 5.11.2.6
Kk
Mass of the multistrand
with K k = and (assuming 9.2.2.2) :
L0
⎛K F ⎞ ⎛ KL FL ⎞
K k = max⎜⎜ 0 , ⎟ = max⎜⎜ 0 , ⎟
0 ⎟
⎝ lk lk − l k ⎠ (
0 ⎟
⎝ lk L lk L − L ⎠ ) EQ. 5.11.2.7
⎛ dg ⎞
⎜ f (ε ) (ε& ) + C ⎟
dε& ⎠ = ⎛ f (ε ) dg (ε& ) + C ⎞ L
Ck = ⎝ ⎜ ⎟ 0 EQ. 5.11.2.8
lk0
⎝ dε ⎠ lk L
20-jan-2009 176
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
STIF0 is the spring stiffness before sensor activation. At sensor activation, the 2 input coefficients among D1,
STIF1, F1 and E1 determine the pretensioner characteristics. Let us recall the following relations between the 4
coefficients:
D1 ⋅ F1 F
E1 = , K1 = 1 EQ. 5.12.2.1
2 D1
STIF0 is also used as unloading stiffness before the end of the piston's slide, and as both loading and unloading
stiffness at the end of the piston's slide. STIF0 should be large enough to allow locking.
20-jan-2009 177
RADIOSS THEORY Version 10.0 ELEMENT LIBRARY
STIF0 is spring stiffness before sensor activation. Depending on the input, pretensioning force is defined as f(L-
L0), with either g(t-t0), or f(L-L0).g(t-t0), with L0 length of the spring at sensor activation time and at t0 sensor
activation time.
Similar use of STIF0 allows piston locking.
The force into the pretensioner spring is computed as: EQ. 5.12.2.2
if, Fp (t + dt ) ≥ 0
20-jan-2009 178