Professional Documents
Culture Documents
10.0 version January 2009 Large Displacement Finite Element Analysis PART 2
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
CONTENTS
CONTENTS
6.0 KINEMATIC CONSTRAINTS
6.1 RIGID BODY
6.1.1 RIGID BODY MASS 6.1.2 RIGID BODY INERTIA 6.1.3 RIGID BODY FORCE AND MOMENT COMPUTATION 6.1.4 TIME INTEGRATION 6.1.5 RIGID BODY BOUNDARY CONDITIONS
5
5
5 6 7 8 8
8
9 13
16
16 16 17 18
18 19
22
23
8.0 INTERFACES
8.1 INTRODUCTION
8.1.1 LAGRANGE MULTIPLIER METHOD 8.1.2 PENALTY METHOD
26
26
27 28
28
30
30 31
31 31 33 36 37
39
39 40 40 42 44 44
45
45 46 46 47 48 48
52
52 53 53 53 53 53
20-jan-2009
CONTENTS
54
54 55 55 57 58 59 59 59 60 61 62
64
64
8.10 TYPE 15 - ELLIPSOIDAL SURFACE TO SEGMENT CONTACT 8.11 TYPE 16- NODE TO CURVED SURFACE CONTACT 8.12 TYPE 17- GENERAL SURFACE TO SURFACE CONTACT 8.13 SOME COMMON PROBLEMS
8.13.1 INCORRECT NEAREST MASTER NODE FOUND 8.13.2 INCORRECT NEAREST MASTER SEGMENT FOUND - B1 8.13.3 INCORRECT NEAREST MASTER SEGMENT FOUND - B2 8.13.4 INCORRECT NEAREST MASTER SEGMENT FOUND - B3 8.13.5 INCORRECT IMPACT SIDE - C1 8.13.6 NO MASTER NODE IMPACT - D1
65 66 66 67
67 67 68 68 69 69
72
74 76
76 79 82 83 93 97
99
99 103 104 105 106 107 109 110 113 114
118
121 122 124 127 127 128
129
129 130 131 132
133 133
133 134
20-jan-2009
CONTENTS
134 134 135 135 136 137 138
139
140 140 140
141 141 142 143 143
145
145 147 148 149 151 151 151 152 153
154
155 155 155 155 156 156 156 156
11.0 STATIC
11.1 STATIC SOLUTION BY EXPLICIT TIME-INTEGRATION
11.1.1 ACCELERATION CONVERGENCE
161
161
165
165
166 166
173
173 174 174
177 183
183 188 191
20-jan-2009
KINEMATIC CONSTRAINTS
Chapter
KINEMATIC CONSTRAINTS
20-jan-2009
KINEMATIC CONSTRAINTS
m = mM + mI
I
EQ. 6.1.1.1
xG =
mM xM + mI xI m
EQ. 6.1.1.2
20-jan-2009
KINEMATIC CONSTRAINTS
yG =
mM y M + mI y I m mM z M + mI z I m
EQ. 6.1.1.3
zG =
where
EQ. 6.1.1.4
M i I xx = J xx + m M ( yM yG ) + (z M zG ) + I xx + m i ( yi yG ) + ( zi zG ) EQ. 6.1.2.1
2 2 2
2
M i I yy = J yy + m M ( xM xG ) + ( z M zG ) + I yy + m i ( xi xG ) + ( zi zG )
2 2 2
) EQ. 6.1.2.2 )
M i I zz = J zz + m M ( xM xG ) + ( yM yG ) + I zz + m i ( xi xG ) + ( yi yG ) EQ. 6.1.2.3
2 2 2
2
M i I xy = J xy + m M ( xM xG ) + ( yM yG ) + I xy mi ( xi xG ) + ( yi yG ) i
EQ. 6.1.2.4
M i I yz = J yz + m M ( yM yG ) + (z M zG ) + I yz mi ( yi yG ) + ( zi zG ) i
EQ. 6.1.2.5
M i I xz = J xz + m M ( xM xG ) + ( z M zG ) + I xz mi ( xi xG ) + (zi zG ) i
) EQ. 6.1.2.6
where I ij is the moment of rotational inertia in the ij direction. J ij is the master node added inertia.
20-jan-2009
KINEMATIC CONSTRAINTS
r r r F = FM + Fi
i
EQ. 6.1.3.1
r r r r r M = M M + M i + Si G F i
i i
EQ. 6.1.3.2
where
r F M is the force vector at the master node r F i is the force vector at the slave nodes r M M is the moment vector at the master node r M i is the moment vector at the slave nodes r G is the vector from slave node to the center of mass
Resolving these into orthogonal components, the linear and rotational acceleration may be computed as: Linear Acceleration
i =
Fi m
EQ. 6.1.3.3
Rotational Acceleration
I11 = M 1 (I 3 I 2 )23
EQ. 6.1.3.4
I 2 2 = M 2 (I1 I 3 )13
EQ. 6.1.3.5
I 3 3 = M 3 (I 2 I1 )12
where Ii are the principal moments of inertia of the rigid body
EQ. 6.1.3.6
i are the rotational accelerations in the principal inertia frame (reference frame) i is the rotational velocity in the principal inertia frame (reference frame)
Mi are the moments in the principal inertia frame (reference frame)
20-jan-2009
KINEMATIC CONSTRAINTS
t +
t r t r = t + (t )t 2 2
EQ. 6.1.4.1
t +
t r t r = t + (t )t 2 2
EQ. 6.1.4.2
where v is the linear velocity vector. Rotational velocities are computed in the local reference frame. The velocities of slave nodes are computed by:
i = M + Si Gx
r r
i = M
A tied interface (Type 2) can be used to connect a fine mesh of Lagrangian elements to a coarse mesh or two different kinds of meshes (for example spring to shell contacts).
20-jan-2009
KINEMATIC CONSTRAINTS
A master and a slave surface are defined in the interface input cards. The contact between the two surfaces is tied. No sliding or movement of the slave nodes is allowed on the master surface. There are no voids present either. It is recommended that the master surface has a coarser mesh. Accelerations and velocities of the master nodes are computed with forces and masses added from the slave nodes. Kinematic constraint is applied on all slave nodes. They remain at the same position on their master segments. Tied interfaces are useful in rivet modeling, where they are used to connect springs to a shell or solid mesh.
20-jan-2009
KINEMATIC CONSTRAINTS
The mass of the slave node is transferred to the master nodes using the position of the projection on the segment and linear interpolation functions:
i i mmaster = mmaster + mslave i (p)
EQ. 6.2.1.1
Figure 6.2.3 Transfer of slave node efforts to the master nodes (SPOTflag=0)
S: mslave M: mislave
h
Fn S
F Ft Fn * i(P) Ft * i(P)
M * i(P)
The inertia of the slave node is also transferred to the master nodes by taking into account the distance d between the slave node and the mater surface:
i i I master = I master + I slave + mslave d 2 i (p)
EQ. 6.2.1.2
The term
mslave d 2 may increase the total inertia of the model especially when the slave node is far from the
master surface. The stability conditions are written on the master nodes:
EQ. 6.2.1.3
20-jan-2009
10
KINEMATIC CONSTRAINTS
The dynamic equilibrium of each master node is then studied and the nodal accelerations are computed. Then the velocities at master nodes can be obtained and updated to compute the velocity of the projected point P by the following expressions:
translation VPtranslation = Vmaster i ( p) i i
EQ. 6.2.1.4
EQ. 6.2.1.4
EQ. 6.2.1.5
With this formulation, the added inertia may be very large especially when the slave node is far from the mean plan of the master element.
Forces and moments transfer from slave to master nodes is described in Figure 6.2.5. The force applied at the slave node S is redistributed uniformly to the master nodes. In this way, only translational mode is excited. The moment M + CS F is redistributed to the master nodes by four forces Fi such that:
r Fi A CM i
EQ. 6.2.1.6
20-jan-2009
11
KINEMATIC CONSTRAINTS
CM
i
Fi = M + CS F
where
Fn S P C Fn / 4 Ft / 4 F3 S M C M+F^CS F1
F Ft
F4
F2
In this formulation the mass of the slave node is equally distributed to the master nodes. In conformity with effort transmission as described in 6.2.1.2, the spherical inertia is computed with respect to the center of the master element C:
Slave IC = I Slave + m Slave .d 2
EQ. 6.2.1.7
where d is distance from the slave node to the center of element. In order to insure the stability condition without reduction in the time step, the inertia of the slave node is transferred to the master nodes by an equivalent nodal mass computed by:
20-jan-2009
12
KINEMATIC CONSTRAINTS
m =
Slave
+m I
Slave
X iYi 2 2 X i + Zi Yi Z i
X i Zi Yi Z i 2 2 X i + Yi
EQ. 6.2.1.8
For this reason the formulation causes an increase of mass which may become very important especially when the node is far from the mean surface of the master shell element.
The distance between each master node in the box and the slave node is computed. The master node giving the minimum distance (dmin) is retained. The segment is chosen with the selected node, (if the selected node belongs to 2 segments, one is selected at random).
20-jan-2009
13
KINEMATIC CONSTRAINTS
d2 < d1
m3 is chosen
m1
m2 d1
S d2 m3 m4
Slave node
To compute the minimum distance, dmin, we have two solutions: 1- The slave node is an internal node for the master segment, as shown in Figure 6.2.9. The slave node is projected orthogonally on the master segment to give a distance that may be compared with other distances. Select the minimum distance:
20-jan-2009
14
KINEMATIC CONSTRAINTS
slave node m2 d m1 m3 m4
The segment that provides the minimum distance is chosen for the following computation. 2- The slave node is a node external to the master segment, as shown in Figure 6.2.10. The distance selected is that between the slave node and the nearest master node. Figure 6.2.10 .Nearest master node
slave node d m3 m2 m1 m4
The segment is chosen using the selected node, (if the selected node belongs to 2 segments, one is chosen at random).
20-jan-2009
15
KINEMATIC CONSTRAINTS
20-jan-2009
16
KINEMATIC CONSTRAINTS
A moving wall is a master slave option. Master node defines the wall position at each time step and imposes velocity on impacted slave nodes. Impacted slave node forces are applied to the master node. The slave node forces are computed with momentum conservation. The mass of the slave nodes is not transmitted to the master node, assuming a large rigid wall mass compared to the impacted slave node mass.
If penetration occurs, a new velocity must be computed. This new velocity is computed using one of three possible situations. These are: 1. Sliding 2. Sliding with Friction 3. Tied For a node which is allowed to slide along the face of the rigid wall, the new velocity V is given by:
v r r rr V =V V n n
( )
EQ. 6.3.3.1
20-jan-2009
17
KINEMATIC CONSTRAINTS
A friction coefficient can be applied between a sliding node and the rigid wall. The friction models are developed in section 8.6.4. For a node that is defined as tied, once the slave node contacts the rigid wall, its velocity is the same as that of the wall. The node and the wall are tied. Hence:
v V = 0
EQ. 6.3.3.2
N r r v N r I = Fi t = mi Vi W i=I i=I
EQ. 6.3.4.1
r r dI F= dt
EQ. 6.3.4.2
20-jan-2009
18
KINEMATIC CONSTRAINTS
The velocity of the group of nodes rigidly linked together is computed using momentum conservation (EQ. 6.5.0.1.). However, no global moment equilibrium is respected.
Vi =
m v
i =i N
i i
m
i =i
EQ. 6.4.0.1
Angular velocity for the ith DOF with respect to the global or a skew reference frame is:
n i i I j j j =1 i = N I ij j =1
For non-coincident nodes, no rigid body motion is possible. A rigid link is equivalent to an infinitely stiff spring type 8.
EQ. 6.4.0.2
6.5 Section
A section is a cut in the structure where forces and moments will be computed and stored in output files. It is defined by: A cutting plane A reference point to compute forces A direction of the section. Figure 6.5.1 Definition of a section for an oriented solid
M F
In RADIOSS the cutting plane is defined by a group of elements and its orientation by a group of nodes as shown in Figure 6.5.2.
20-jan-2009
19
KINEMATIC CONSTRAINTS
Then, a point is defined for the center of rotation of the section and a reference frame is attached to this point to compute the internal efforts. Figure 6.5.3 Center of rotation and its associated frame for a section
The resultant of all forces applied to the elements and its application point are computed by:
F = fi M = mi + ON i f i
Figure 6.5.4 Resultant of force and moment for a node I w.r.t the rotation point O
20-jan-2009
20
Chapter
LINEAR STABILITY
LINEAR STABILITY
(a) Stable
(b) Unstable
(c) Neutral
It is clear that the state (b) represents an unstable case since a small change in the position of the ball results the rolling either to the right or to the left. It is worthwhile to mention here that stability and equilibrium notions are quite different. A system in static equilibrium may be in unstable state and a system in evolution is not necessary unstable. A good understanding of the stability of equilibrium can be obtained by studying the load-deflection curves. A typical behavior of a structure in buckling is given in Figure 7.0.2. The load-deflection curves are slightly different for systems with and without imperfection. In the first case, the structure is loaded until the bifurcation point B corresponding to the first critical load level. Then, two solutions are mathematically acceptable: response without buckling (BA), response after buckling (BC). In the case of structures with imperfection, no bifurcation point is observed. The behavior before buckling is not linear and the turning point D is the limit point in which the slope of the curve changes sign. If the behavior before buckling is linear or the nonlinearity before the limit point is not high, the linear stability technique can be used to determine the critical load. The method is based on the perturbation of the equilibrium state. As the perturbations are small, the linearized model is used. The method is detailed in the following section. Figure 7.0.2 Bifurcation and limit points in load-deflection curves for system with and without imperfections B: Bifurcation point, D: Limit point
PCr PS
Displacement
20-jan-2009
22
LINEAR STABILITY
= 0
=>
Static equilibrium
EQ.7.1.0.1
Applying a small perturbation to the equilibrium state, the variation of the total potential energy can be written as:
(t + t ) = (t ) + 2
Where
EQ.7.1.0.2
is the second variation of the potential energy. Then, the three cases can be distinguished: => => Stable (case a) The energy increases around the equilibrium state. Unstable (case b) The energy decreases around the equilibrium state. Neutral stability (case c) EQ.7.1.0.5 EQ.7.1.0.4 EQ.7.1.0.3
2 > 0
2 < 0
=> =>
2 = 0
=> =>
2 = 2 int + 2 ext = 0
EQ.7.1.0.6
Where, the indices int and ext denote the interval and external parts of the total potential energy. After the application of the application of finite element method, the stability equation in a discrete form can be written as:
e 2 = 2 int + 2 e ext = 0 n n
EQ.7.1.0.7
2 e ext =
e 2 int =
0 Se
X {f } dS
n
0 e
EQ.7.1.0.8
Ve0
EQ.7.1.0.9
LINEAR STABILITY
{X cr } = {X L }
{Scr } = {S L } {Fcr } = {FL }
If we admit that the loading does not depend on the deformation state, the hypothesis part, we have: EQ.7.1.0.10
2 ext = 0 is then true. ] for the nonlinear Using EQ. 2.4.2.6 and denoting [e] for the linear part of Green-Lagrange strain tensor and [
{E} = {e}+ { }
Putting this equation in EQ.7.1.0.9, we obtain:
EQ.7.1.0.11
2 e ext =
Or Where k
V0e
( e [C ]{e}+ ( [C ]{e}+
L
e [C ][ L ] + S { 2 E})) dVe0
EQ.7.1.0.12
2 e ext = X
EQ.7.1.0.13
[]
stiffness matrix and C the elastic matrix. The linear theory of stability allows estimating the critical loads and their associated modes by resolving an eigenvalue problem:
[ ]
[ ]
[ ]
EQ.7.10.14
Linear stability assumes the linearity of behavior before buckling. If a system is highly nonlinear in the neighbourhood of the initial state C , moderate perturbations may lead to unstable growth. In addition, if case of path-dependent materials the use of method is not conclusive from an engineering point of view. However, the method is simple and provides generally good estimations of limit points. The resolution procedure consists in two main steps. First, the linear solution for the equilibrium of the system under the application of the load {FL } is obtained. Then, EQ.7.10.14 is resolved to compute the first desired critical loads and modes. The methods to compute the eigen values are those explained in section 4.2.
20-jan-2009
24
Chapter
INTERFACES
20-jan-2009
25
INTERFACES
8.0 INTERFACES
8.1 Introduction
Interfaces solve the contact and impact conditions between two parts of a model. Contact-impact problems are among the most difficult nonlinear problems to solve as they introduce discontinuities in the velocity time histories. Prior to the contact, the normal velocities of the two bodies which come into contact are not equal, while after impact the normal velocities must be consistent with the impenetrability condition. In the same way, the tangential velocities along interfaces are discontinuous when stick-slip behavior occurs in friction models. These discontinuities in time complicate the integration of governing equations and influence performance of numerical methods. Central to the contact-impact problem is the condition of impenetrability. This condition states that bodies in contact cannot overlap or that their intersection remains empty. The difficulty with the impenetrability condition is that it cannot be expressed in terms of displacements as it is not possible to anticipate which parts of the bodies will come into contact. For this reason, it is convenient to express the impenetrability condition in a rate form at each cycle of the process. This condition can be written as:
A B N = vN vN 0
EQ. 8.1.0.1
interpenetration. EQ. 8.1.0.1 simply expresses that when two bodies are in contact, they must either remain in contact and N = 0 , or they must separate and N < 0 . On the other hand, the tractions must observe the balance of momentum across the contact interface. This requires that the sum of the tractions on the two bodies vanish:
A B tN + tN =0
EQ. 8.1.0.2
EQ. 8.1.0.3
EQ 8.1.0.1 and EQ. 8.1.0.2 can be combined in a single equation stating that t N N = 0 . This condition simply expresses that the contact forces do not create work. If the two bodies are in contact, the interpenetration rate vanishes. On the other hand, if the two bodies are separated N < 0 but the surface tractions vanish. As a result, the product of the surface tractions and the interpenetration rate disappear in all cases. If we notice that the impenetrability condition is expressed as an inequality constraint, the condition:
tN N = 0
EQ. 8.1.0.4
can also be seen as the Kuhn-Tucker condition associated with the optimization problem consisting in minimizing the total energy (EQ. 2.10.0.5) subject to the inequality constraint 8.1.0.1. In practice, the solution to a contact problem entails in three steps: First, it is necessary to find for each point those points in the opposite body which will possibly come into contact. This is the geometrical recognition phase. The second phase is to check whether or not the bodies are in contact and, if the bodies are in contact, if they are sticking or slipping. This step makes uses of the geometrical information computed in the first phase. The last step will be to compute a satisfactory state of contact.
20-jan-2009
26
INTERFACES
The geometrical recognition phase is dependent on the type of interface. This will be discussed below in parallel with the description of interfaces. On the other hand, structural problems with contact-impact conditions lead to constrained optimization problems in which the objective function to be minimized is the virtual power subject to the contact-impact conditions. There are conventionally two approaches to solving such mathematical programming problems: the Lagrange multiplier method the Penalty method. Both of two methods are used in RADIOSS.
df =
EQ. 8.1.1.1
dg =
EQ. 8.1.1.2
Note that the differentials are all independent, so any combination of them can be set equal to 0 and the remainder must still give zero. This requires that
f g x + x k k g 2 = 0 , ...,
dxk = 0
EQ. 8.1.1.4
is
g1 = 0 ,
f = 1g1 + 2g 2 + ...
EQ. 8.1.1.5
The Lagrange multiplier method can be applied to contact-impact problems. In this case, the multivariate function is the expression of total energy subjected to the contact conditions:
&, & &) f ( x1 , x2 ,..., xn ) ( x, x x &, & &) = 0 g (x1 , x2 ,..., xn ) Q(x, x x
&, & & are the global vectors of d.o.f. The application of Lagrange multiplier method to the previous where x, x x equations gives the weak form as:
20-jan-2009
27
INTERFACES
EQ. 8.1.1.8 EQ. 8.1.1.9
K L
LT x f = 0 0
EQ. 8.1.1.10
The Lagrange multipliers are physically interpreted as surface tractions. The equivalence of the modified virtual power principle with the momentum equation, the traction boundary conditions and the contact conditions (impenetrability and surface tractions) can be easily demonstrated [28]. It is emphasized that the above weak form is an inequality. In the discretized form, the Lagrange multiplier fields will be discretized and the restriction of the normal surface traction to be compressive will result from constraints on the trial set of Lagrange multipliers.
Q = ( N )d
c
EQ. 8.1.2.1
where
( N ) = 0 ( N ) > 0
if if
N =0 N <0
is an arbitrary parameter known as the penalty parameter. The penalty function is an arbitrary function of the interpenetration and its rate. It is emphasized that the weak form including the virtual power and the penalty term EQ. 8.1.2.1 is not an inequality form. The penalty function will be defined in the description of interfaces.
20-jan-2009
28
INTERFACES
Interface Descriptions
Type Description
1 2 3 5 6 7 8 9 10 11 12 14 15 Tied contact (boundary) between an ALE part and a Lagrangian part. Tied contact. Used to simulate impacts and contacts on shell and solid elements. Surfaces should be simply convex. Used to simulate impacts and contacts between a master surface and a list of slave nodes. Best suited for beam, truss and spring impacts on a surface. Used to simulate impacts and contacts between two rigid surfaces. A general interface that removes the limitations of types 3 and 5.
Drawbead contact for stamping applications
ALE lagrange with void opening and free surface. Tied after impact with or without rebound. Edge to edge or line to line impact. Connects 2 fluid meshes with free, tied or periodic options ALE or EULER or LAG/ ALE or EULER or LAG. Ellipsoidal surfaces to nodes contact. Ellipsoidal surfaces to segments contact.
Each of these interfaces was developed for a specific application field. However, this application field is not the only selection criteria. Some limitations of the different algorithms used in each interface can also determine the choice. The algorithm limitations concern mainly the search of the impacted segment. This search may be performed directly (type 7, 10, 11 interfaces), or via a search of the nearest node (type 3, 5, 6 interfaces). Apart from the limitation of the nearest node search, some limitations exist for the choice between the segments connected to the nearest node. These limitations are the same for type 3, 5 and 6 interfaces. Type 3, 5 and 6 interfaces also have some limitations due to the orientation of the normal segment. Type 7 interface was written to emulate type 3 and 5 interfaces without algorithm limitations. With this interface, each node can impact one or more segments, on both sides, on the edges or on the corners of the segments. The only limitation to this interface concerns high impact speed and/or small gap. For these situations the interface will continue to work properly, but the time step can decrease dramatically. Types 3, 5 and 7 interfaces are compatible with all RADIOSS kinematic options. Type 1 interface is special option used with solid elements to provide mesh transition. Type 1 interface is used to connect a Lagrangian and an A.L.E. mesh Type 2 interface is used to connect a fine and a coarse Lagrangian mesh. All other interface types (3, 4, 5, 6 and 7) are used to simulate impacts and contacts. A node may belong to several interfaces.
20-jan-2009
29
INTERFACES
For a brick element, a segment is one face of the brick. Figure 8.2.2 Brick Element Segment
For a two dimensional element, a segment is one side. Figure 8.2.3 2D Element Segment
20-jan-2009
30
INTERFACES
capacity to make each point of the surface impact on itself capacity to impact on both sides of a segment (internal and external) possibility for a point of the surface to be wedged between an upper and a lower part processing of very strong concavities (will complete folding) reversibility of the contact, thereby authorizing unfolding after folding or the simulation of airbag deployment.
20-jan-2009
31
INTERFACES
The contact is detected based on the criteria of distance between a group of nodes and a meshed surface. The distance between a node and the surface of a triangular or quadrangular segment is evaluated, locally. Symmetrized contact nodes to surface The symmetrization of the previous formulation makes it possible to model a contact between two surfaces, as the group of nodes of the first surface can impact the second group and vice-versa. Figure 8.4.2 Contact edges to edges
This formulation makes it possible to model contacts between wire framed structures or between edges of twodimensional structures. Contact is detected based on the criteria of distance between two segments. This formulation can also be used to describe in an approximate way the surface to surface contact.
20-jan-2009
32
INTERFACES
Several approaches can be used to detect the contact between two surfaces. If the two surfaces are quadrangular, the exact calculation of the contact can be complex and quite expensive. An approximate solution may be made by combining the two previous formulations. By evaluating all the contacts of nodes to surface, as well as the edge to edge contacts, the only approximation is the partial consideration of the segment curvature.
When using direct search, at each cycle the distance is calculated from each entity (node, segment, edge) to all others. The quadratic cost (N*N) of this algorithm makes it unusable in case of auto-contact.
20-jan-2009
33
INTERFACES
In a simulation in fast dynamics, geometrical modifications of the structure are not very important during one cycle of calculation. It is then possible to consider neighborhood search algorithms using the information of the previous cycle of computation. If for a node the nearest segment is known at the previous cycle, it is then possible to limit the search for this node to the segments topologically close to the previous one (the segments having at least one common node). Furthermore, if an algorithm based on the search of neighboring nodes is used, then the search may be limited to the nodes of the segments connected to the previous closest node.
It remains however necessary to do a complete search before the first cycle of calculation. We will also see that this algorithm presents significant restrictions that limit its use to sliding surfaces (it cannot be used for autoimpacting surfaces). The cost of this algorithm is linear (N), except at the first cycle of computation, during which it is quadratic (N*N). The combination of this algorithm with one of the two following is also possible.
5 4 3 2 1 0 3
2 4 5 8 9 13 18 15 17 19 23 20 24 25 27 10 6
11 14 16 21 26
12
22
In order to limit the memory space needed, we first count the number of nodes in each box, thereby making it possible to limit the filling of those boxes that are not empty. In the two-dimensional example shown above, nodes are arranged in the boxes as described in the following table. With this arrangement, the calculation of distances between nodes of the same box is not a problem. On the other hand, taking into account the nodes of neighboring boxes is not straightforward, especially in the horizontal direction (if the arrangement is first made vertically, as in this case). One solution is to consider three columns of boxes at a time. Another solution, more powerful but using more memory, would be for each box to contain the nodes already located there plus those belonging to neighboring boxes. This is shown in the third series of columns in the following table. Once this
20-jan-2009
34
INTERFACES
sorting has been performed, the last step is to calculate the distances between the different nodes of a box, followed by the distance between these nodes and those of the neighboring box. In box 0,3 for example, fifteen distances must be calculated. 11-14, 11-3, 11-8, 11-12, 11-15, 11-16, 11-17, 11-18, 14-3, 14-8, 14-12, 14-15, 14-16, 14-17, 14-18.
Box
Nodes
0,0 0,1 0,2 0,3 1,1 1,2 1,3 1,4 1,5 2,1 2,2 2,3 2,4 2,5 3,3 4,1 4,2 4,4 5,0
26 21 16 11 14 22 15 17 8 3 1 23 19 13 5 2 9 24 20 4 6 25 27 10
21 15 16 17 22 26 8 3 8 1 2 8 3 1 1 4 9 9 24 11 12 14 15 17 21 22 26 8 12 15 16 17 18
15 16 17 19 21 23 26 11 12 13 14 18 16 19 21 22 23 5 2 3 9 5 2 3 5 11 13 14 15 16 17 19 5 5 10 12 13 15 17 18 22 23 8 3 5 6 13 19 20 9 8 10 12 15 17 18 19 9 10 12 13 18 11 12 13 14 18 12 18 3
15 17 19 22
20 25 27 10 24 10
In this example we have considered a search based on nodal proximity. It is possible to adapt this algorithm in order to arrange segments or even edges in the boxes. It will be necessary however to reserve more memory, for a segment can overlap several boxes.
20-jan-2009
35
INTERFACES
In order to illustrate the 3D quick sort, let us consider a search for node-segment proximity. At each step, space is successively divided into two equal parts, according to directions X, Y or Z. The group of nodes is thus separated into two subsets. We thereby obtain a tree organization with two branches per node. After each division, a check is made to determine whether the first of the two boxes must be divided. If so, it will be divided similar to the previous one. If not, the next branch is then checked. This recursive algorithm is identical to the regular fast sorting one. The segments are also sorted using the spatial pivot. The result of the test can lead to three possibilities: the segment is on the left side of the pivot, on the right side of the pivot or astride the pivot. In the first two cases, the segments are treated similar to the nodes, but in the third situation, the segment is duplicated and placed on both sides. Among the different criteria that can be used to stop the division are the following situations: the box does not contain any nodes the box does not contain any segments the box contains sufficiently few elements that the calculation of distance of all the couples is more economical the dimension of the box is smaller than a threshold
20-jan-2009
36
INTERFACES
penalty methods do not ensure kinematic contact continuity, but they add springs at the contact spots.
The first advantage to this formulation is its natural integration in an explicit code. Each contact is treated like an element and integrates itself perfectly into the code architecture, even if the programming is vectorial and parallel. Contrary to the kinematic formulations, the penalty method ensures the conservation of momentum and kinetic energy during impact. The formulation used in RADIOSS is a penalty type formulation. The choice of the penalty factor is a major aspect of this method. In order to respect kinematic continuity, the penalty spring must be as rigid as possible. If the impedance of the interface becomes higher than those of the structures in contact, some numerical rebounds (high frequency) can occur. To ensure the stability of the integration diagram, without having additional constraints, this rigidity must be low. With a too low penalty, the penetration of the nodes becomes too strong and the geometrical continuity is no longer ensured. The compromise selected consists in using a stiffness of the same order of magnitude than the stiffness of the elements in contact. This stiffness is non linear and increases with the penetration, so that a node is not allowed to cross the surface. These choices provide a clear representation of physics, without numerical generation of noise, but require the contact stiffness in the calculation of the criteria of stability of the explicit scheme to be taken into account.
20-jan-2009
37
INTERFACES
r t
r ni
Ns Ni
N4
r si
(-1,0)
(0,0)
(1,0)
r s
N1
(0,-1) N2
The projected point is calculated using isoparametric coordinates for a quadrangular segment and barycentric coordinates for a triangular segment. In the case of any quadrangular segment, the exact calculation of these coordinates leads to a system of two quadratic equations that can be solved in an iterative way. The division of the quadrangular segment into four triangular segments makes it possible to work with a barycentric coordinate system and gives equations that can be solved in an analytical way. From the isoparametric coordinates (s,t) of the projected point (Ni), we have all the necessary information for the detection and the processing of contact. The relations needed for the determination of s and t are as follows: the vector NiNs is normal to the segment at the point Ni and the normal to the segment is given by the vectorial product of the vectors si and ti.
2 N 1 Ni = (1 s ) N 1 N 3 + (1 + s ) N 1 N 4 D = N 1 Ni N 1 N i N 1 N 2 / N 1 N 2 N 1 N 2 D / s = 0
20-jan-2009 38
INTERFACES
The force of penalty is calculated as in node-to-segment contact. It is applied to the nodes N1, N2, N3, N4 and therefore ensures the equilibrium of forces and moments.
N3 Ni N2
N4 N1
8.5.1 Limitations
The main limitations of type 3 interfaces are: The two surfaces should be simply convex. The surface normals must face each other. A node may not exist on the master and slave side of an interface simultaneously. Surfaces must consist of either shell or brick elements. It is recommended that the two surface meshes be regular with a good aspect ratio. The interface gap should be kept small, if not zero. There are some search problems associated with this interface.
20-jan-2009
39
INTERFACES
Because the computation algorithm is performed twice, accuracy is improved over a type 5 interface. However, the computational cost is increased. The first pass solution solves the penetration of the nodes on surface 1 with respect to segments on surface 2. The second pass solves surface 2 nodes with respect to surface 1 segments.
20-jan-2009
40
INTERFACES
K =s
K1 K 2 K1 + K 2
EQ. 8.5.3.1
where s = Stiffness Scaling Factor. Default is 0.2. K1 = Surface 1 Stiffness K2 = Surface 2 Stiffness K = Overall Interface Spring Stiffness The scale factor, s, may have to be increased if:
K1 << K 2 or K 2 << K1
The calculation of the spring stiffness for each surface is determined by the type of elements. For example:
K1 =
1 s 10 s K 2 implies K = K 2 or K = K1 10 11 11
K = 0.5sEt
where E = Modulus of Elasticity t = Shell Thickness The stiffness does not depend on the shell size.
EQ. 8.5.3.2
K = 0.5
where
sBA2 V
EQ. 8.5.3.3
20-jan-2009
41
INTERFACES
r K Ft = C1C0 10
where K = Interface Spring Stiffness
EQ. 8.5.4.1
20-jan-2009
42
INTERFACES
Slave node
Vt Kt
Fn = K s P
where K s = K 0
EQ. 8.5.4.2
Gap Gap P
K0 = Initial interface spring stiffness (as in type 5) The tangential force computation is given by:
Ft = K t D
where K t =
EQ. 8.5.4.3
Kn 10
If the friction force is greater than the limiting situation, Ft > Fn , the frictional force is reduced to equal the limit, Ft = Fn , and sliding will occur. If the friction is less than the limiting condition, Ft Fn , the force is unchanged and sticking will occur. Time integration of the frictional forces is performed by:
EQ. 8.5.4.4
Fn
Xt
20-jan-2009
43
INTERFACES
20-jan-2009
44
INTERFACES
Referring to Figure 8.5.9, the first situation shows the mesh deforming in a way that allows the normals to be facing each other. However, in the second case, the deformation moves two surfaces with normals all facing the same direction, where contact will not be detected. Large rotations can have a similar effect, as shown in Figure 8.5.10 Figure 8.5.10 Initial and Deformed Mesh (Before Impact)
Kinematic motion may reposition the mesh so that normals do not correspond. It is recommended that possible impact situations be understood before a simulation is attempted.
8.6.1 Limitations
The main limitations of type 5 interface are: The master segment normals must be oriented from the master surface towards the slave nodes. The master side segments must be connected to solid or shell elements. The same node is not allowed to exist on both the master and slave surfaces. Some search problems (refer to the problems A1, B1, B2, B3, C1, D1 in chapter) It is recommended that the master surface mesh be regular, with a good aspect ratio and that a small or zero gap be used to detect penetration.
20-jan-2009
45
INTERFACES
Surface 1 K1 K2 Surface 2
K =s
K1 K 2 K 1+ K 2
EQ. 8.6.3.1
20-jan-2009
46
INTERFACES
K1 << K 2 or K 2 << K1
The calculation of the spring stiffness for each surface is determined by the type of elements. For example:
K1 =
s 1 10 s K 2 implies K = K 2 or K = K1 10 11 11
= C1 e (C V ) p 2 + C3 e (C V ) p + C5 e (C V )
2 4 6
EQ. 8.6.4.1
Renard law:
= C1 + (C3 C1 )
V V 2 C5 C5
V C5 C 6 C5
2
if V 0, C5
] ]
EQ. 8.6.4.2
= C3 (C3 C4 )
V C5 3 2 C 6 C5
if V C5 , C6
= C2
1 1 + (V C6 ) C2 C4
if V C6
where, C1 = s , C2 = d
20-jan-2009
47
INTERFACES
4 F = Ft + (1 ) Ft 1
where the coefficient a depends on the Ifiltr flag value.
EQ. 8.6.4.3
20-jan-2009
48
INTERFACES
4. Determine the new closest master node, M2, to node N at time t1. The new master node must belong to one of the segments S0, S1, S2 or S3. 5. Determine the new closest master segment (S2, S4, S5 and S6). The Starter CPU Cost is calculated with the following equation: CPUstarter = a x Number of Slave Nodes x Number of Master Nodes The Engine CPU Cost is calculated with the following equation: CPUengine = b x Number of Slave Nodes The algorithm used in the engine is less expensive but it does not work in some special cases. In Figure 8.6.4, if node N is moving from N(t0) to N(t1) and then to N(t2), the closest master nodes are M0 and M1. When the final node movement to N(t2) is taken, the impact on segment S will not be detected since none of the nodes on this segment are considered as the closest master node. Figure 8.6.4 Undetected Impact
20-jan-2009
49
INTERFACES
)(
EQ. 8.6.6.1
where P1 = Projection of N on plane (M B1 B2) The A value for segment S2 is given by:
)(
EQ. 8.6.6.2
where P2 = Projection of N on plane (M B2 B3) The same procedure is carried out for all master segments that node M is connected to. The closest segment is the segment for which A is a minimum. In some special cases (curved surfaces), it is possible that: 1. All values of A are positive. 2. More than one value of A is negative.
20-jan-2009
50
INTERFACES
F = Kp
where K = Interface Spring Stiffness
EQ. 8.6.6.3
Reaction forces F1, F2, F3 and F4 are applied on each master node (as shown in Figure 8.6.8) in the opposite direction to the penetration force, such that:
F1 + F2 + F3 + F4 = F
EQ. 8.6.6.4
Forces Fi (i = 1, 2, 3, 4) are functions of the position of the contact point, C. They are evaluated by:
20-jan-2009
51
INTERFACES
Fi =
di
dj
j =1
4
EQ. 8.6.6.5
where d i =
Ni (Sc , tc )
N (S , t )
j =1 2 j c c
mi
or more simply:
Fi = N i (Sc , tc )F
where Ni = Standard linear quadralateral interpolation functions Sc and tc = Isoparametric coordinates contact point The penalty method is used to reduce the penetration. This provides: Accuracy Generality Efficiency Compatibility
EQ. 8.6.6.6
8.7.1 Limitations
Some of the main limitations for this interface type are: Surface 1 must be part of one and only one rigid body. Surface 2 must be part of one and only one rigid body. The interface stiffness (user defined function) can reduce the time step. Other limitations are the same as for type 3 interfaces. Figure 8.7.1 Surfaces 1 and 2 with Facing Normals
20-jan-2009
52
INTERFACES
t = 0.1
M K
EQ. 8.7.5.1
where M = min(M rigid body 1 , M rigid body 2 ) K = Tangent of user force function Time step
The function f refers to a function number given in input and has to be provided by user.
20-jan-2009
53
INTERFACES
))
EQ. 8.7.6.1
r r q N i is the contribution to node N i of vector qi distributed on the segment penetrated by node Qi.
8.8.1 Limitations
All limitations encountered with interface types 3, 4 and 5 are solved with this interface. It is a fast search algorithm without limitations. Figure 8.8.1 Slave and Master Node Impact
There are no search limitations with this interface concerning node to surface contacts. All possible contacts are found.
20-jan-2009
54
INTERFACES
There is no limitation on the use of large and small segments on the same interface. This is recommended to have a good aspect ratio elements or a regular mesh to obtain reasonable results; however, it is not an obligation. There is no limitation to the size of the spring stiffness factor. The spring stiffness is much greater than interfaces 3 and 5, with the default stiffness factor set to 1.0. This is to avoid node penetrations larger than the gap size, removing problems that were associated with the other interfaces.
F = f ( p ) + visc 2 KM
dp dt
EQ. 8.8.2.1
The stiffness is much larger than the other interfaces to accommodate high speed impacts with minimal crossing of surfaces. The consequence of this is that a stable time step is calculated to maintain solution stability.
r r Ft = C tVt
EQ. 8.8.3.1
In the second method an artificial stiffness KS is input. The change of tangent force Ft is obtained the following equation:
Ft = K S t
where
EQ. 8.8.3.2
Fn = K s P + C
dP dt
EQ. 8.8.3.3
where K s = K 0 Gap P
Gap
C = VIS s 2 K s M
K0 = Initial interface spring stiffness (as in type 5) VISs = Critical damping coefficient on interface stiffness (input)
20-jan-2009
55
INTERFACES
EQ. 8.8.3.4
20-jan-2009
56
INTERFACES
If the friction force is greater than the limiting situation, Ft > Fn , the frictional force is reduced to equal the limit, Ft = Fn , and sliding will occur. If the friction is less than the limiting condition, Ft Fn , the force is unchanged and sticking will occur. Note that the friction coefficient Darmstad and Renard models as described in section 8.6.4. Time integration of the frictional forces is performed by:
EQ. 8.8.3.5
Fn
Vt
Type 7 interface has a gap that covers both edges of the segments, as shown in Figure 8.8.5.
20-jan-2009
57
INTERFACES
dP > 0 by: dt
t min
Gap p = 0.5 dP dt
EQ. 8.8.5.1
t nod =
2M K
EQ. 8.8.5.2
K1
K0 Gap 0
Penetration Gap 1
Using a larger gap size, curves 1 and 2 keep the same initial stiffness, hence the initial time step remains the same. Since the impact slowing force is applied over a greater distance, the stiffness is not changed as much, but increases. Increasing the initial interface stiffness, although decreasing the time step initially, will increase the time step if penetration is large.
20-jan-2009
58
INTERFACES
For shell elements, the master gap is equal to one half of the shell thickness. The slave gap is equal to one half of the largest thickness of all connected shell elements. For solid elements, the master gap is zero. If the slave node is only connected to solid elements, the slave gap is zero. For beam or truss elements connected to the slave node, the slave gap is one half of the square root of the section area. If a slave node is connected to different elements (shell, brick, beam, truss) the largest gap value is used. The total gap is the sum of the slave and master gaps. The total gap cannot be smaller than a minimum gap (user input gap).
20-jan-2009
59
INTERFACES
The reaction force is not a linear relation like the previous interfaces. There is a viscous damping which acts on the rate of penetration. The force computation is given by:
F = Ks P + C
dP dt
EQ. 8.8.9.1
20-jan-2009
60
INTERFACES
where K s = K 0 Gap P
Gap
C = VIS s 2 K s M
K0 = Initial interface spring stiffness (as in type 5) VISs = Critical damping coefficient on interface stiffness (input) The instantaneous stiffness is given by:
Gap 2 Kt = K0 (Gap P )2
Figure 8.8.10 Force - Penetration Graph
EQ. 8.8.9.2
A critical viscous damping is required to be defined on the type 7 input card for both damping on the spring stiffness and for interface friction damping.
20-jan-2009
61
INTERFACES
Position 1 in Figure 8.8.11 shows the force acting radially from the edge of the segment. The size of the force depends on the amount of penetration. At position 2 the force is normal to the segment surface. In position 3 two segments intersect and their gaps overlap. The result is that each segment applies a force to the node, normal to the respective segment; this can double the force for the distance of gap overlap.
Deep penetrations (i.e. close to gap value) cannot be avoided in most car crash simulations. They occur in the following cases: initial penetrations of adjacent plates edge impacts: wrong side contacts full collapse of one structural region rigid body impact on another rigid body or on fixed nodes or on very stiff structures impact between heavy stiff structures high impact speed small gap
20-jan-2009
62
INTERFACES
F =k
gp (g p )
EQ. 8.8.11.1
With k = 0.5 STFAC*E* t The elastic contact energy is calculated with the formulation:
(g p ) CE = kg p gLn g
When node to element mid-plane distance is smaller then 10-10 gap, the node is deactivated. The maximum potential contact energy is: elastic contact energy CE = 23 k g Drastic time step dropping is mostly due to cases where node is forced into the gap region.
EQ. 8.8.11.2
20-jan-2009
63
INTERFACES
The interface is designed to allow penetration of slave nodes. However, the contact algorithm does not ensure that a node will not cross the ellipsoid when sliding; nodes may slide along the ellipsoid until they fully cross the ellipsoid, resulting in that the structure itself fully crossing the surface and contact force is no longer applied to it (as shown in figure 8.9.2, where perfect sliding is considered). Increase interface stiffness or friction to avoid this problem.
20-jan-2009
64
INTERFACES
Interface Type 14 : No contact is detected between heel and coarse mesh of floor
Interface stiffness is a non-linear increasing function of penetration, computed in order to avoid penetrations up to half the ellipsoid:
20-jan-2009
65
INTERFACES
Depth
Penetration
A Kinematic Time Step is computed so that the element does not cross the line Lt within one time step. A friction coefficient m is input. Interface takes into account sliding/rolling effects. Coulomb Friction condition is expressed as:
20-jan-2009
66
INTERFACES
20-jan-2009
67
INTERFACES
20-jan-2009
68
INTERFACES
20-jan-2009
69
INTERFACES
20-jan-2009
70
Chapter
MATERIALS
20-jan-2009
71
MATERIALS
Law number in RADIOSS (MID) (1) (42) (19) (58) (28) (68) (50) (32) (43) (57) (25) (15) (14) (12) (53) (2) (2) (48) (44) (36) (60) (13) (21) (10) (63) (64) (65) (27) (54) (24) (22) (23) (52) (34) (35) (38) (40) (62) (70) (33) (4) (6) (3) (41) (49) (0)
von Mises hardening with brittle damage von Mises hardening with ductile damage von Mises with visco-plastic flow
Hydrodynamic
Visco-plastic Strain rate and temperature dependence on yield stress Turbulent viscous flow Elasto-plastic hydrodynamic Hydrodynamic material Elasto-plastic hydrodynamic with thermal softening Void material
Hill Hill (tabulated) 3-Parameter Barlat Composite Shell Composite Shell Chang-Chang Composite Solid Tsai-Wu formula for solid Foam model Johnson-Cook Zerilli-Armstrong Zhao Cowper-Symonds Tabulated piecewise linear Tabulated quadratic in strain rate Rigid material Drucker-Prager for rock or concrete Druger-Prager for rock or concrete by polynominal Hnsel model Ugine and ALZ approach Elastomer Aluminum, glass, etc. Predit rivets Reinforced concrete Ductile damage for solids and shells Ductile damage for solids Ductile damage for porous materials, Gurson Boltzman Generalized Kelvin-Voigt Tabulated law Generalized Maxwell-Kelvin Hyper visco-elastic Tabulated law - hyper visco-elastic Closed cell, elasto-plastic foam Johnson-Cook Hydrodynamic viscous von Mises isotropic hardening with polynomial pressure Lee-Tarver material Steinberg-Guinan Fictitious
Void
20-jan-2009
73
MATERIALS
These material laws are used to model purely elastic materials, or materials that remain in the elastic range. The Hookes law requires only two values to be defined; the Young's or elastic modulus E, and Poisson's ratio, . The law represents a linear relation between stress and strain. The Ogdens law is applied to slightly compressible materials as rubber or elastomer foams undergoing large deformation with an elastic behavior [34]. The strain energy W is expressed in a general form as a function of W (1 , 2 , 3 ) :
W =
p
p 1 + 2 + 3 3 p
p p p
)
p p
and
where i , ith principal stretch ( i = 1 + i , i being the ith principal engineering strain) and constants. This law is very general due to the choice of coefficient
p material
p .
For a compressible material, the components of Cauchy stress tensor are given in its principal reference:
i =
i W J i 0
EQ. 9.1.0.2
1 = 2 = 3 =
Considering the general case of compressible material and defining new variables:
3 * i i = J 1
EQ. 9.1.0.3
EQ. 9.1.0.4
*
( , , , J ) = W ( , , ). We can write:
* 1 * 2 * 3 1 2 3
i =
* 3 i W i W * j W * J = * + J i J J 1 j = j i i
EQ. 9.1.0.5
As
1 *j 2 1 *j J 1 j = J and = J 3 for i=j and = J 3 for ij, EQ. 9.1.0.5 is simplified to: 3 i i i 3 i
i =
* * * 3 1 * W 1 W J W i * * J J i 3 j =1 j
EQ. 9.1.0.6
20-jan-2009
74
MATERIALS
p=
1 3 W * = j 3 j =1 J
EQ. 9.1.0.7
p = K f bulk ( J ) ( J 1)
where f bulk a user defined function related to the bulk modulus K :
EQ. 9.1.0.8
K =
2(1 + ) 3(1 2 )
p
EQ. 9.1.0.9
p
EQ. 9.1.0.10
being the ground shear modulus, and the Poisson's ratio. W* is obtained using EQ. 9.1.0.1 and EQ. 9.1.0.4:
* * W * 1 , * 2 , 3 = p
p * * * * * * 1 + 2 + 3 3 p Ln 1 23 p p
p p p
EQ. 9.1.0.11
* i
W * * = p (i p 1) * p i
EQ. 9.1.0.12
1 = 2 = 3 = 1
2 3
1 = 2 = 3 =
using
and
* = * = * = 1 , we have also W * = i = 0
1
W .
Note: For an incompressible material we have too small time steps in explicit codes.
0 .5
. However,
= 0.495
Mooney-Rivlin material law admits two basic assumptions: The rubber is incompressible and isotropic in unstrained state, The strain energy expression depends on the invariants of Cauchy tensor.
I1 = 1 + 2 + 3
2 2
I 2 = 1 2 + 2 3 + 3 1
2 2 2 2 2 2 2 2
EQ. 9.1.0.13
20-jan-2009
75
MATERIALS
EQ. 9.1.0.14
1 = 2 C10 2 = 2 C01 1 = 2 2 = 2
EQ. 9.1.0.15
The purpose of this section is to describe the mathematical models related to composite and orthotropic materials.
9.2.1 Fabric law for elastic orthotropic shells (laws 19 and 58)
Two elastic linear and a nonlinear models exist in RADIOSS.
9.2.1.1 Fabric linear law for elastic orthotropic shells (law 19)
A material is orthotropic if its behavior is symmetrical with respect to two orthogonal plans. The fabric law allows to model this kind of behavior. This law is only available for shell elements and can be used to model an airbag fabric. Many of the concepts for this law are the same as for law 14 which is appropriate for composite solids. If axes 1 and 2 represent the orthotropy directions, the constitutive matrix C is defined in terms of material properties:
1 E 11 12 E11 C 1 = 0 0 0
21
E22 1 E22 0 0 0
0 0 1 G12 0 0
0 0 0 1 G23 0
0 0 0 0 1 G31
EQ. 9.2.1.1
where the subscripts denote the orthotropy axes. As the matrix C is symmetric:
20-jan-2009
76
MATERIALS
12
E11
21
E22
EQ. 9.2.1.2
Therefore, six independent material properties are the input of the material: E 11 = Young's modulus in direction 1 E 22 = Young's modulus in direction 2
12 = Poisson's ratio
G 12 , G 23 , G 31 = Shear moduli for each direction The coordinates of a global vector V is used to define direction 1 of the local coordinate system of orthotropy. The angle is the angle between the local direction 1 (fiber direction) and the projection of the global vector
The shell normal defines the positive direction for . Since fabrics have different compression and tension behavior, an elastic modulus reduction factor, RE, is defined that changes the elastic properties of compression. The formulation for the fabric law has a 11 reduction if 11 < 0 as shown in Figure 9.2.2.
20-jan-2009
77
MATERIALS
9.2.1.2 Fabric nonlinear law for elastic anisotropic shells (law 58)
This law is used with RADIOSS standard shell elements and anisotropic layered property (type 16). The fiber directions (warp and weft) define the local axes of anisotropy. Material characteristics are determined independently in these axes. Fibers are nonlinear elastic and follow the equation:
= E B 2 , with
d >0 d
EQ. 9.2.1.3
The shear in fabric material is only supposed to be function of the angle between current fiber directions (axes of anisotropy):
= G0 tan( ) 0 = G tan( ) + G A 0
and
if if
EQ. 9.2.1.4
> T
GT 1 + tan 2 (T )
GA = (G0 G ) tan( T ) ,
where
G=
with
0 = G0 tan( 0 )
= 0. If
T is a shear lock angle, GT is a tangent shear modulus at T , and G0 is a shear modulus at G0 = 0, the default value is calculated to avoid shear modulus discontinuity at T : G0 = G.
Figure 9.2.4 Elastic Compression Modulus Reduction
weft
warp
0 is an initial angle between fibers defined in the shell property (type 16).
The warp and weft fiber are coupled in tension and uncoupled in compression. But there is no discontinuity between tension and compression. In compression only fiber bending generates global stresses. Figure 9.2.5 illustrates the mechanical behavior of the structure.
20-jan-2009
78
MATERIALS
A local micro model describes the material behavior (Figure 9.2.6). This model represents just of a warp fiber wave length and of the weft one. Each fiber is described as a non linear beam and the two fibers are connected with a contacting spring. These local non linear equations are solved with Newton iterations at membrane integration point. Figure 9.2.6 Local frame definition
Weft
20-jan-2009
79
MATERIALS
The Hooke matrix defining the relation between the stress and strain tensors is diagonal, as there is no Poisson's effect:
11 E11 0 22 33 0 = 12 0 23 0 0 31
0 E22 0 0 0 0
0 0 E33 0 0 0
0 0 0 G12 0 0
0 0 0 0 G23 0
11 22 33 0 12 0 23 G31 31 0 0 0 E11 2
EQ. 9.2.2.1
EQ. 9.2.2.2
Plasticity may be defined by a volumic strain or strain dependent yield curve (Figure 9.2.8). The input yield stress function is always positive. If the material undergoes plastic deformation, its behavior is always orthotropic, as all curves are independent to each other. Figure 9.2.8 Honeycomb typical constitutive curve
The failure plastic strain may be input for each direction. If the failure plastic strain is reached in one direction, the element is deleted. The material law may include strain rate effects (law 50) or not (law 28).
20-jan-2009
80
MATERIALS
9.2.2.2 Cosserat medium for nonlinear pseudo-plastic orthotropic solids (law 68)
Conventional continuum mechanics approaches cannot incorporate any material component length scale. However, a number of important length scales as grains, particles, fibers, and cellular structures must be taken into account in a realistic model of some kinds of materials. To this end, the study of a microstructure material having translational and rotational degrees-of-freedom is underlying. The idea of introducing couple stresses in the continuum modelling of solids is known as Cosserat theory which returns back to the works of brothers Cosserat in the beginning of 20th century [110]. A recent renewal of Cosserat mechanics is presented in several works of Forest et al [111], [112], [113], and [114]. A short summary of these publications is presented in this section. Cosserat effects can arise only if the material is subjected to non-homogeneous straining conditions. A Cosserat medium is a continuous collection of particles that behave like rigid bodies. It is assumed that the transfer of the interaction between two volume elements through surface element dS occurs not only by means of a traction and shear forces, but also by moment vector as shown in Figure 9.2.9.
32 22 12
21
2 1 L
31
11
11
31
Surface forces and couples are then represented by the generally non-symmetrical force-stress and couple-stress tensors ij and ij (units MPA and MPa-m):
ti = ij n j ; mi = ij n j
The force and couple stress tensors must satisfy the equilibrium of momentums:
EQ. 9.2.2.3
&&i ij , j + f i = u
&& ij , j ikl kl + ci = I i
where f i are the volume forces, ci volume couples, the signature of the perturbation (i,k,l). In the often used couple-stress, the Cosserat micro-rotation is constrained to follow the material rotation given by the skew-symmetric part of the deformation gradient: EQ. 9.2.2.4
i = ijk u j ,k
1 2
EQ. 9.2.2.5
20-jan-2009
81
MATERIALS
The associated torsion-curvature and couple stress tensors are then traceless. If a Timoshenko beam is regarded as a one-dimensional Cosserat medium, constraint EQ. 9.2.2.5 is then the counterpart of the Euler-Bernoulli conditions. The resolution of the previous boundary value problem requires constitutive relations linking the deformation and torsion-curvature tensors to the force- and couple-stresses. In the case of linear isotropic elasticity, we have:
Symm. Skew Symm. ij = ekk ij + 2 eij + 2c eij Symm. Skew Symm. ij = kkk ij + 2 ij + 2 ij
EQ. 9.2.2.6
where eij
Symm.
and eij
Skew Symm.
deformation tensor. Four additional elasticity moduli appear in addition to the classical Lam constants. Cosserat elastoplasticity theory is also well-established. Von Mises classical plasticity can be extended to micropolar continua in a straightforward manner. The yield criterion depends on both force- and couple-stresses:
f ( , ) =
EQ. 9.2.2.7
where s denotes the stress deviator and ai, bi are the material constants. Cosserat continuum theory can be applied to several classes of materials with microstructures as honeycombs, liquid crystals, rocks and granular media, cellular solids and dislocated crystals.
EQ. 9.2.3.1
where the coefficients F, G, H, L, M and N are the constants obtained by the material tests in different orientations. The stress components ij are expressed in the Cartesian reference parallel to the three planes of anisotropy. EQ.9.2.3.1 is equivalent to von Mises yield criteria if the material is isotropic. In a general case, the loading direction is not the orthotropic direction. In addition, we are concerned with the plane stress assumption for shell structures. In planar anisotropy, the anisotropy is characterized by different strengths in different directions in the plane of the sheet. The plane stress assumption will allow to simplify EQ. 9.2.3.1, and write the expression of equivalent stress eq as:
eq = A1 112 + A2 22 2 A3 11 22 + A12 12 2
The coefficients Ai are determined using Lankfords anisotropy parameter r :
EQ. 9.2.3.2
R=
H=
R 1+ R
1 A1 = H 1 + r 00
EQ. 9.2.3.3
20-jan-2009
82
MATERIALS
1 1 A12 = 2 H (r45 + 0.5) r + r 00 90
A3 = 2 H
where the Lankfords anisotropy parameters r are determined by performing a simple tension test at angle
d r =
d 33
EQ. 9.2.3.4
eq
and
&0 )m 0 = a ( 0 )n .(
The yield stress variation is shown in Figure 9.2.10. Figure 9.2.10 Yield stress variation
EQ. 9.2.3.6
The strain rates are defined at integration points. The maximum value is taken into account:
d x d y d d , ,2 ( xy ) = max dt dt dt dt
EQ. 9.2.3.7
In RADIOSS, it is also possible to introduce the yield stress variation by a user defined function (law 43). Then, several curves are defined to take into account the strain rate effect. It should be noted that as Hills law is an orthotropic law, it must be used for elements with orthotropy properties as Type 9 and Type 10 in RADIOSS.
20-jan-2009
83
MATERIALS
The first case can be modeled by an isotropic material where the composite property is defined in element property definition as explained in chapter 5. However, in the case of composite shell with orthotropic layers the definition of a convenient material law is needed. Two dedicated material laws for composite orthotropic shells exist in RADIOSS: Material law COMPSH (25) with orthotropic elasticity, two plasticity models and brittle tensile failure, Material law CHANG (15) with orthotropic elasticity, fully coupled plasticity and failure models.
These laws are described here. The description of elastic-plastic orthotropic composite laws for solids is presented in the next section.
t1 = Tensile failure strain in direction 1 m1 = Maximum strain in direction 1 t 2 = Tensile failure strain in direction 2 m2
= Maximum strain in direction 2
20-jan-2009
84
MATERIALS
(1-dmax) E
9.2.4.2 Delamination
The delamination equations are:
31 = G31 (1 d 3 ) 31 23 = G23 (1 d 3 ) 23
where d 3 is the delamination damage factor. The damage evolution law is linear with respect to the shear strain. Let
2 2 = 31 + 23
then:
for for
d3 = 0 d3 = 1
= t =m
EQ. 9.2.4.4
20-jan-2009
85
MATERIALS
EQ. 9.2.4.5
Where F1 =
1cy
1
c 1y
1t y
;
F2 =
1
c 2 y
1
t 2 y
F11 = F44 =
c 12 y
t 1y
F22 =
1
t 2 y c 2y
t 12 y
F12 =
F11 F22 2
is the reduction factor. The six other parameters are the yield stresses in tension and compression for the orthotropy directions which can be obtained uniaxial loading tests:
1yt
F ( ) < 1
: elastic state : plastic admissible state : plastically inadmissible stresses EQ. 9.2.4.6
F ( ) = 1
F ( ) > 1
20-jan-2009
86
MATERIALS
For F ( ) = 1 the cross-sections of Tsai-Wu function with the planes of stresses in orthotropic directions is shown in Figure 9.2.13. Figure 9.2.13 Cross-sections of Tsai-Wu yield surface for F ( ) = 1
2
t 2 y
t 12 y
1cy 1t y
c 12 y
c 2 y
12
If F ( ) > 1 , the stresses must be projected on the yield surface to satisfy the flow rule. F ( ) is compared to a maximum value F W p varying in function of the plastic work W p during work hardening phase:
p n p
( ) F ( ) = F (W ) = 1 + bW
EQ. 9.2.4.7
with: b = Hardening parameter n = Hardening exponent Therefore, the plasticity hardening is isotropic as illustrated in Figure 9.2.14.
20-jan-2009
87
MATERIALS
If one of the two conditions is satisfied, the material is ruptured. The evolution of yield surface during work hardening of the material is shown in Figure 9.2.15. Figure 9.2.15 Evolution of Tsai-Wu yield surface
The model will allow the simulation of the brittle failure by formation of cracks. The cracks can either be oriented parallel or perpendicular to the orthotropic reference frame (or fiber direction), as shown in Figure
20-jan-2009
88
MATERIALS
for a given element, 9.2.16. For plastic failure, if the plastic work WP is larger than the maximum value WP then the element is considered to be ruptured. However, for a multi-layer shell, several criteria may be considered to model a total failure. The failure may happen: If WP > WP If WP > WP If WP > WP If WP > WP If WP > WP If WP > WP If
max
max
for one layer, for all layers, or tensile failure in direction 1 for each layer, or tensile failure in direction 2 for each layer, or tensile failure in directions 1 and 2 for each layer, or tensile failure in direction 1 for all layer, or tensile failure in direction 2 for all layer, or tensile failure in direction 1 and 2 for each layer.
max max
max max
WP > WP
max
max
If WP > WP
The last two cases are the most physical behaviors; but the use of failure criteria depends, at first, to the analysts choice. In RADIOSS the flag IOFF defines the used failure criteria in the computation. Figure 9.2.16 Crack orientation
In practice, the use of brittle failure model allows to estimate correctly the physical behavior of a large rang of composites. But on the other hand, some numerical oscillations may be generated due to the high sensibility of the model. In this case, the introduction of an artificial material viscosity is recommended to stabilize results. In addition, in brittle failure model, only tension stresses are considered in cracking procedure. The ductile failure model allows plasticity to absorb energy during a large deformation phase. Therefore, the model is numerically more stable. This is represented by CRASURV model in RADIOSS. The model makes also possible to take into account the failure in tension, compression and shear directions as described in the following.
& F (W p ) = 1 + bW pn 1 + c ln &o
EQ. 9.2.4.8
20-jan-2009
89
MATERIALS
Wp & W
p
: plastic work, : maximum value of the plastic work rates of the strain tensor, : reference plastic work rate for static loading, : strain rate coefficient (equal to zero for static loading).
& W po
c
The last equation implies the growing of the Tsai-Wu yield surface when the dynamic effects are increasing. The effects of strain rate are illustrated in Figure 9.2.17. Figure 9.2.17 Strain rate effect in work hardening
F44 =
( )
12 y
EQ. 9.2.4.9
Another modification concerns the parameters Fij in EQ. 9.2.4.5 which are expressed now in function of plastic work and plastic work rate as in EQ. 9.2.4.8:
2 2 F (W p ) = 1 = F1 (W p ) 1 + F2 (W p ) 2 + F11 (W p ) 12 + F22 (W p ) 2 + F44 (W p ) 12 + 2 F12 (W p ) 1 2
& c 1c n 1 + c1c Ln 1 + b1c (W p ) 1cy = 10 &o & n c c c c + W p2 c ) 1 c Ln 2 2 y = 20 1 + b2 ( & o & t 1t n 1 + c1t Ln 1 + b1t (W p ) 1t y = 10 &o & n t t t t + W p2t ) 1 c Ln 2 2 y = 20 1 + b2 ( &o & 12 n 1 + c12 Ln ) 12 y = 120 1 + b12 (W p &o
where the five sets of coefficients b, n and c should be obtained by experience. The work hardening is shown in Figure 9.2.18.
EQ. 9.2.4.10
20-jan-2009
90
MATERIALS
The CRASURV model will allow the simulation of the ductile failure of orthotropic shells. The plastic and failure behaviors are different in tension and in compression. The stress softening may also be introduced in the model to take into account the residual Tsai-Wu stresses. The evolution of CRASURV criteria with hardening and softening works is illustrated in Figure 9.2.19.
20-jan-2009
91
MATERIALS
S1 = Longitudinal tensile strength S 2 = Transverse tensile strength S12 = Shear strength C1 = Longitudinal compressive strength C 2 = Transverse compressive strength = Shear scaling factor.
where 1 is the fiber direction. The failure criterion for fiber breakage is written as: Tensile fiber mode :
2
11 > 0
2
11 12 e2 + f = S 1. 0 S1 12
EQ. 9.2.4.11
20-jan-2009
92
MATERIALS
11 < 0
0 failed EQ. 9.2.4.12 < 0 elastic plastic
11 ec2 = C 1.0 1
For matrix cracking, the failure criterion is: Tensile matrix mode : 22 > 0
22 12 e = S + S 1.0 2 12
2 m
22 < 0
EQ. 9.2.4.14
If the damage parameter is equal to or greater than 1.0, the stresses are decreased by using an exponential function to avoid numerical instabilities. We use a relaxation technique by decreasing gradually the stress:
[ (t )] = f (t ) [ d (tr )]
With: f (t ) = exp
EQ. 9.2.4.15
t tr T
and
t tr where:
t = the time, tr = the start time of relaxation when the damage criteria are assumed,
[ d (tr )] = the stress components at the beginning of damage (for matrix cracking [ d (tr )] =
d 22 ) d 12
v v v
20-jan-2009
93
MATERIALS
For the case of unidirectional orthotropy (i.e. E33 = E22 and G31 = G12 ) the material law (53) in RADIOSS allows to simulate an orthotropic elastic-plastic behavior by using a modified Tsai-Wu criteria.
ij (t ) , and strain rate, dij, from global reference frame to fiber reference frame.
EQ. 9.2.5.1
ij (t + t ) = ij (t ) + Dijkl d kl t
3. Transform the lamina stress,
The elastic constitutive matrix C of the lamina relates the non-null components of the stress tensor to those of strain tensor:
{ } = [D]{ }
EQ. 9.2.5.2
The inverse relation is generally developed in term of the local material axes and nine independent elastic constants:
1 E 11 11 22 33 = 12 23 31
21
E22 1 E22
31 32
E33
0 0 0 1 G12
0 0 0 0 1 G23
E33 1 E33
Symm.
0 0 11 22 0 33 12 0 23 0 31 1 G31
EQ. 9.2.5.3
where Eij are the Youngs modulus, Gij shear modulus and ij Poissons ratios. due to the distortion.
ij
20-jan-2009
94
MATERIALS
EQ. 9.2.5.4
Fi =
c i
t i
Fii =
it
c i
F12 =
where
1 2
(F11F22 )
1 F23 = F22 2
is the yield stress in direction i. c and t denote respectively for compression and tension. f W p
( )
represents the yield envelope evolution during work hardening with respect to strain rate effects:
& f (W p ) = 1 + B W pn 1 + c. ln &0
EQ. 9.2.5.5
where W p is the plastic work, B the hardening parameter, n the hardening exponent and c strain rate coefficient.
EQ. 9.2.5.6
If the maximum value is reached the material is failed. In EQ. 9.2.5.5, the strain rate affects on the evolution of yield envelope. However, it is also possible to take into & effects on the maximum stress max as shown in Figure 9.2.22. account the strain rate
20-jan-2009
95
MATERIALS
max
max
= y 1 + c. ln
= y 1 + c. ln
& &0
0 1 + c. ln max = max
0 max = max
f max
= max y
EQ. 9.2.5.7
The orthotropic plasticity behavior is modeled by a modified Tsai-Wu criterion (EQ. 9.2.5.4) in which:
F12 =
45 c
( )
2 45 c y
EQ. 9.2.5.8
where y is yield stress in 45 unidirectional test. The yield stresses in direction 11, 22, 12, 13 and 45, are defined by independent curves obtained by unidirectional tests (Figure 9.2.23). The curves give the stress variation in function of a so-called strain v :
v = 1 exp(Trace[ ])
EQ. 9.2.5.9
20-jan-2009
96
MATERIALS
ij
Traction
Compression
v = 1 exp(Trace[ ])
F = a K1 + K 2 + a K1 K 2 + c 2 K 2 2( e )
m m m
EQ. 9.2.6.1
where
is the yield stress, a and c are anisotropic material constants, m Barlats exponent and K1 and K 2 are
defined by :
K1 =
xx h yy
2
2
EQ. 9.2.6.2
xx h yy 2 K2 = + p 2 ( xy ) 2
where h and p are additional anisotropic material constants. All anisotropic material constants, except for p which is obtained implicitly, are determined from Barlat width to thickness strain ratio R from:
EQ. 9.2.6.3
20-jan-2009
97
MATERIALS
can be calculated according to [100] by:
R =
2 m ( e )
F F + yy xx
EQ. 9.2.6.4
where
direction. Let
F F 45 + xx yy
1 R45 = 0
EQ. 9.2.6.5
It is worthwhile to note that Barlats law reduces to Hills law when using m=2.
20-jan-2009
98
MATERIALS
n 1 + c ln = a + b p
where:
m & 1 &0
EQ. 9.3.1.1
&
&0
m = Temperature exponent
melt is the melting temperature in Kelvin degrees. The adiabatic conditions are assumed for temperature
computation:
= i +
Where:
C (Volume )
Eint
EQ. 9.3.1.2
20-jan-2009
99
MATERIALS
d x d y d d , ,2 ( xy ) = max dt dt dt dt
EQ. 9.3.1.3
xy = xy
With global plastification method, we have:
1 2
EQ. 9.3.1.4
d dEi / dt = dt VM
where Ei is the internal energy. For solid elements, the maximum value of the strain rate components is used:
EQ. 9.3.1.5
EQ. 9.3.1.6
20-jan-2009
100
MATERIALS
EQ. 9.3.1.7
For thin-walled structures, the equivalent strain is computed by the following approach. If is the main component of strain tensor, the kinematic assumptions of thin-walled structures allows to decompose the inplane strain into membrane and flexural deformations:
= z + m
Then, the expression of internal energy can by written as:
t 2 t 2 t 2
EQ. 9.3.1.8
Ei = dz = E 2 dz = E (z + m ) dz
2 t 2 t 2 t 2
EQ. 9.3.1.10
Therefore:
1 2 2 Ei = E ( z + + 2 m z ) dz = E 2 z 3 + m z + m z 2 3 t t
2 2 2 m 2
t 2
EQ. 9.3.1.11
1 2 2 Ei = E 2t 3 + m t = E eq t 12
EQ. 9.3.1.12
eq =
1 22 2 t + m 12
EQ. 9.3.1.13
EQ. 9.3.1.14
&m = m
& =
Therefore:
& =
EQ. 9.3.1.17
20-jan-2009
101
MATERIALS
Referring to EQ. 9.3.1.13, it can be seen that an equivalent strain rate can be defined using a similar expression to the equivalent strain:
&eq = eq
EQ. 9.3.1.18
&eq =
1 &2 2 2 &m k t + 12
EQ. 9.3.1.19
For solid elements, the strain rate is computed using the maximum element stretch:
EQ. 9.3.1.20
The strain rate at integration point, i in /ANIM/TENS/EPSDOT/i (1<i<n) is calculated by the following expression:
&i = &m
Where:
1 (2i 1) &b 1t 2 n
EQ. 9.3.1.21
&m is the membrane strain rate /ANIM/TENS/EPSDOT/MEMB &b is the bending strain rate /ANIM/TENS/EPSDOT/BEND
The strain rate in upper and lower layers are computed by:
/ANIM/TENS/EPSDOT/UPPER /ANIM/TENS/EPSDOT/LOWER
EQ. 9.3.1.24
a = 2 dt Fcut dt = time interval Fcut = cutting frequency & f = filtered strain rate
20-jan-2009
102
MATERIALS
= C0 + C1 exp C3 + C4 ln
where:
& n + C5 p & 0
EQ. 9.3.2.1
p = Plastic Strain
= Temperature (computed as in Johnson Cook plasticity)
20-jan-2009
103
MATERIALS
C0 = Yield Stress
n = Hardening Exponent
& = Strain Rate, must be 1 s-1 converted into user's time unit &0 = Reference Strain Rate
Additional inputs are:
max
allows to define element rupture as in the former law. The theoretical aspects related to strain rate
The damage is neglected in the model. The work hardening model is similar to the Johnson Cook model (law 2) without temperature effect where the only difference is in the strain rate dependent formulation. The equation that describes the stress during plastic deformation is:
1 1 &p = (a + b )1 + c
n p
EQ. 9.3.3.1
where,
&
= Strain Rate
1/p = strain rate exponent The implanted model in RADIOSS allows the cyclic hardening with a combined isotropic-kinematic approach. The coefficient Chard varying between zero and unity is introduced to regulate the weight between isotropic and kinematic hardening models. In isotropic hardening model, the yield surface inflates without moving in the space of principle stresses. The evolution of the equivalent stress defines the size of the yield surface, as a function of the equivalent plastic strain. The model can be represented in one dimensional case as shown in Figure 9.3.4. When the loading direction is changed, the material is unloaded and the strain reduces. A new hardening starts when the absolute value of the stress reaches the last maximum value (Figure 9.3.4(a)).
20-jan-2009
104
MATERIALS
Figure 9.3.4 Isotropic and Kinematic hardening models for deformation decrease (a) Isotropic hardening (b) Prager-Ziegler kinematic hardening
yt
yt
This law is available for solids and shells. Refer to the RADIOSS Input Manual for more information about element/material compatibilities.
EQ. 9.3.4.1
where:
A = Yield stress B = hardening parameter n = hardening exponent C = relative strain rate coefficient D = strain rate plasticity factor m = Relative strain rate exponent E = strain rate coefficient k = strain rate exponent In the case of material without strain rate effect, the hardening curve given by EQ. 9.3.4.1 is identical to those of Johnson-Cook. However, Zhao law allows a better approximation of strain rate dependent materials by introducing a nonlinear dependency. As described for Johnson-Cook law, a strain rate filtering can be introduced to smooth the results. The plastic flow with isotropic or kinematic hardening can be modeled as described for in section 9.3.3. The material failure is happened when the plastic strain reaches a maximum value as in Johnson-Cook model. However, two tensile strain limits are defined to reduce stress when rupture starts:
n +1 = n
where:
t 2 1 t 2 t1
EQ. 9.3.4.2
t1
t2
20-jan-2009
105
MATERIALS
When
1 > t 2
9.3.5 Tabulated piecewise linear and quadratic elasto-plastic laws (laws 36 and 60)
The elastic-plastic behavior of isotropic material is modeled with user defined functions for work hardening curve. The elastic portion of the material stress-strain curve is modeled using the elastic modulus, E, and Poisson's ratio, . The hardening behavior of the material is defined in function of plastic strain for a given strain rate (Figure 9.3.5). An arbitrary number of material plasticity curves can be defined for different strain rates. For a given strain rate, a linear interpolation of stress for plastic strain change, can be used. This is the case of law 36 in RADIOSS. However, in law 60 a quadratic interpolation of the functions allows to better simulate the strain rate effects on the behavior of material as it is developed in law 60. For a given plastic strain, a linear interpolation of stress for strain rate change is used. Compared to Johnson-Cook model (law 2), there is no maximum value for the stress. The curves are extrapolated if the plastic deformation is larger than the maximum plastic strain. The hardening model may be isotropic, kinematic or a combination of the two models as described in section 9.3.3. The material failure model is the same as in Zhao law. For some kinds of steels the yield stress dependence to pressure has to be incorporated especially for massive structures. The yield stress variation is then given by:
y = y 0 ( p ) f ( p )
EQ. 9.3.5.1
where p is the pressure defined by EQ. 2.7.2.1. Drucker-Prager model described in section 9.3.6 gives a nonlinear function for f ( p ) . However, for steel type materials where the dependence to pressure is low, a simple linear function may be considered:
y = y 0 ( p ) C p( p )
EQ. 9.3.5.2
where C is user defined constant and p the computed pressure for a given deformed configuration. Figure 9.3.5 Piecewise linear stress-strain curves
p
The principal strain rate is used for the strain rate definition:
d x d y d 1 d x d y = + + dt dt dt 2 dt dt
For strain rate filtering, refer to section 9.3.1.2.
d xy + dt
EQ. 9.3.5.3
20-jan-2009
106
MATERIALS
F = J 2 A0 + A1P + A2 P 2
where:
EQ. 9.3.6.1
1 sij sij 2
stresses represents a revolutionary surface with an axis parallel to the trisecting of the space as shown in Figure 9.3.7. This representation is in contrast with the von Mises criteria where yield criterion has a cylindrical shape. Drucker-Prager criterion is a simple approach to model the materials with internal friction because of the symmetry of the revolution surface and the continuity in variation of normal to the yield surface. The pressure in the material is determined in function of volumetric strain for loading phase:
P = f ( )
for loading
d > 0
EQ. 9.3.6.2
where f is a user defined (law 21) or a cubic polynomial function (law 10). For unloading phase, if the volumetric strain has a negative value, a linear relation is defined as:
P = C1
<0
EQ. 9.3.6.3
For unloading with a positive volumetric strain, another linear function may be used:
P = B
>0
EQ. 9.3.6.4
In RADIOSS Drucker-Prager model is used in laws 21 and 2. None of these laws can reproduce the monodimensional behavior. In addition, no viscous effect is taken into account.
J 2 and P.
20-jan-2009
107
MATERIALS
3
2
20-jan-2009
108
MATERIALS
The elastic-plastic behavior of the material is defined by Johnson-Cook model. However, the stress-strain curve for the material incorporates a last part related to damage phase as shown in Figure 9.3.10. The damage parameters are:
t1 m1
n = eff (1 d )
where 0 < d < 1 is the damage factor. The strains and the stresses in each direction are given by:
EQ. 9.3.7.1
1 = 2 =
(1 d1 )E
2
E
v 2 E
EQ. 9.3.7.2
v 1 E
EQ. 9.3.7.3
12 = 1 =
(1 d1 )G
12
EQ. 9.3.7.4
E (1 d1 ) (1 + v 2 ) 1 (1 d1 )v 2
EQ. 9.3.7.5
20-jan-2009
109
MATERIALS
2 =
E ( 2 + (1 d1 )v1 ) 1 (1 d1 )v 2
EQ. 9.3.7.6
= t = m
; ;
d=0 d=1
A linear damage model is used to compute the damage factor in function of material strain.
d=
t m t m ( tp ) m t
EQ. 9.3.7.7
The stress-strain curve is then modified to take into account the damage by EQ. 9.3.7.1. Therefore:
=E
EQ. 9.3.7.8
m t t tp
EQ. 9.3.7.9
The mathematical approach described here can be applied to the modeling of rivets. Predit law in RADIOSS allows achievement of this end by a simple model where for the elastic-plastic behavior a Johnson-Cook model or a tabulated law (36) may be used.
20-jan-2009
110
MATERIALS
f ( m , J 2 , f c , f b , ) c = 0
where:
EQ. 9.3.8.1
m =
I1 = mean stress 3
3 3 J3 3/ 2 2 J2
A schematic representation of the failure surface in the principal stress space is given in Figure 9.3.12. The yield surface is derived from the failure envelope by introducing a scale factor k( m , ). The meridian planes are presented in Figure 9.3.13. The steel directions are defined identically to material law 14 by a type 6 property set. If no property set is given in the element input data, r ,s , are taken respectively as direction 1, 2, 3. For quad elements, direction 3 is taken as the direction. Steel data properties are: E =Young's modulus
= Yield strength
E t = Tangent modulus
20-jan-2009
111
MATERIALS
Uniaxial tension
Uniaxial compression
1
Biaxial tension
Biaxial compression
Failure Surface
20-jan-2009
112
MATERIALS
vonm
fc
1.
0.5
P fc
-0.5 -1.0 -1.5 -2.0
0.0
0 1
o o If If
< dam = 0
dam Edam = (1 )E
1 + (1 ) 2
This implies an isotropic damage with the same effects in tension and compression. The inputs of the model are the starting damage strain dam and the slope of the softening curve Et as shown in Figure 9.3.14. For brick elements the damage law can be only applied to the deviatoric part of stress tensor sij and
Gdam =
Edam . This is the case of law (22) in RADIOSS. However, if the application of damage law to 2(1 + dam )
stress tensor
ij
20-jan-2009
113
MATERIALS
& may The strain rate definition and filtering for these laws are as explained in section 9.3.1. The strain rate affect or not the maximum stress value max according to the user's choice as shown in Figure 9.3.15.
Figure 9.3.15 Strain rate dependency (a) Strain rate effect on
max
max
= y 1 + c. ln
max
= y 1 + c. ln
& & 0
0 max = max
9.3.10 Ductile damage model for porous materials (Gurson law 52)
The Gurson constitutive law [64] models progressive microrupture through void nucleation and growth. It is dedicated to high strain rate elasto-viscoplastic porous metals. A coupled damage mechanical model for strain rate dependent voided material is used. The material undergoes several phases in the damage process as described in Figure 9.3.16.
20-jan-2009
114
MATERIALS
The constitutive law takes into account the void growth, nucleation and coalescence under dynamic loading. The evolution of the damage is represented by the void volume fraction, defined by:
f =
Va Vm Va
EQ. 9.3.10.1
Where Va , Vm are respectively the elementary apparent volume of the material and the corresponding elementary volume of the matrix. The rate of increase of the void volume fraction is given by:
EQ. 9.3.10.2
& = (1 f ) Trace D p f g
[ ]
2
EQ. 9.3.10.3
Where Trace[Dp] is the trace of the macroscopic plastic strain rate tensor. The nucleation rate of voids is given by the following expression:
& = f n
fN S N 2
1 M N 2 SN
&M
EQ. 9.3.10.4
Where fN is the nucleated void volume fraction, SN is the Gaussian standard deviation, N is the nucleated effective plastic strain and M is the admissible plastic strain. The viscoplastic flow of the porous material is described by:
2 eq 3 m = + 2q1 f cosh evp 2 2 q2 M M 2 = eq + 2q f 1 + q f 1 3 evp 2 M
2 if m > 0 1 + q3 f
EQ. 9.3.10.5
if m 0
20-jan-2009
115
MATERIALS
eq
M is
is the
f = f f = f + fu f c ( f f ) c c f F fc
Where:
f c critical void volume fraction at coalescence, f F critical void volume fraction at ductile fracture, fu the corresponding value of the coalescence function f u =
1 * , f ( f F ) = fu . q1
The variation of the specific coalescence function is shown in Figure 9.3.17. Figure 9.3.17 Variation of specific coalescence function
f * = fN
Nucluation
f * = fF f * = fC
Coalescence
Growth
f*
&M =
: Dp (1 f ) M
EQ. 9.3.10.7
Where is the Cauchy stress tensor, M is the admissible plastic stress and Dp is the macroscopic plastic strain rate tensor which can be written in the case of the associated plasticity as:
& evp Dp =
EQ. 9.3.10.8
with evp the yield surface envelope. The viscoplastic multiplier is deduced from the consistency condition:
EQ. 9.3.10.9
20-jan-2009
116
MATERIALS
evp
evp (1 f ) : I + A1 A2
EQ. 9.3.10.10
where:
; A = f N e A2 = (1 f ) M 1 S N 2
evp
1 M N 2 SN
EQ. 9.3.10.11
20-jan-2009
117
MATERIALS
D(t ) =
(t ) 0
EQ. 9.4.0.1
The creep behavior is mainly composed of three phases: (i) primary creep with fast decrease in creep strain rate, (ii) secondary creep with slow decrease in creep strain rate and (iii) tertiary creep with fast increase in creep strain rate. The creep strain rate is the slope of creep strain to time curve. Another kind of loading concerns viscoelastic materials subjected to a constant tensile strain,
the stress, (t ) which is called stress relaxation, gradually decreases. The tensile relaxation modulus is then defined as:
0 . In this case,
E (t ) =
(t ) 0
EQ. 9.4.0.2
Because viscoelastic response is a combination of elastic and viscous responses, the creep compliance and the relaxation modulus are often modeled by combinations of springs and dashpots. A simple schematic model of viscoelastic material is given by the Maxwell model shown in Figure 9.4.1. The model is composed of an elastic spring with the stiffness E and a dashpot assigned a viscosity . It is assumed that the total strain is the sum of the elastic and viscous strains:
= e +v
Figure 9.4.1 Maxwell model
EQ. 9.4.0.3
The time derivation of the last expression gives the expression of the total strain rate:
EQ. 9.4.0.4
e =v =
EQ. 9.4.0.5
20-jan-2009
118
MATERIALS
= E e
&v =
then
&e & = E
Combining EQ. 9.4.0.4, EQ. 9.4.0.6 and EQ. 9.4.0.7, an ordinary differential equation for stress is obtained:
& & = E
where
or
& & = E
EQ. 9.4.0.8
is the relaxation time. A solution to the differential equation is given by the convolution integral:
t t d (t ' ) d (t ' ) dt ' = R(t t ') dt ' dt ' dt '
(t ) = E e[ (t t ') / ]
EQ. 9.4.0.9
where R(t) is the relaxation modulus. The last equation is valid for the special case of Maxwell one-dimensional model. It can be extended to the multi-axial case by:
(t ) = Cijkl (t t ')
t
EQ. 9.4.0.10
where Cijkl are the relaxation moduli. The Maxwell model represents reasonably the material relaxation. But it is only accurate for secondary creep as the viscous strains after unloading are not taken into account. Another simple schematic model for viscoelastic materials is given by Kelvin-Voigt solid. The model is represented by a simple spring-dashpot system working in parallel as shown in Figure 9.4.2. Figure 9.4.2 Kelvin-Voigt model
& = E +
EQ. 9.4.0.11
When = 0 (no dashpot), the system is a linearly elastic system. When E=0 (no spring), the material behavior is expressed by Newton's equation for viscous fluids. In above relation, we have considered a one-dimensional model. For multiaxial situations, the equations can be generalized and rewritten in tensor form. The Maxwell and Kelvin-Voigt models are appropriate for ideal stress relaxation and creep behaviors. They are not adequate for most of physical materials. A generalization of these laws can be obtained by adding other springs to the initial models as shown in Figures 9.4.3 and 9.4.4. The equations related to the generalized Maxwell model are given as:
20-jan-2009
119
MATERIALS
EQ. 9.4.0.12 EQ. 9.4.0.13
= i + 1 =
&=
i
Ei &1 1 + E1 1
EQ. 9.4.0.14
The mathematical relations, which hold the generalized Kelvin-Voigt model are:
=e +k = e + k e =
E
;
EQ. 9.4.0.15
k =
V
Et
&k =
The combination of these equations allows to obtain the expression of stress and strain rates:
& E
&k +
& k + Et k =
& & = E
(E + Et )
EEt
EQ. 9.4.0.18
20-jan-2009
120
MATERIALS
The models described above concern the viscoelastic materials. The plasticity can be introduced in the models by using a plastic spring. The plastic element is inactive when the stress is less than the yield value. The modified model is able to reproduce creep and plasticity behaviors. The viscoplasticity law (33) in RADIOSS will allow to implement very general constitutive laws useful for a large rang of applications as low density closed cells polyurethane foam, honeycomb, impactors and impact limiters. The behavior of viscoelastic materials can be generalized to three dimensions by separating the stress and strain tensors into deviatoric and pressure components:
sij = 2 (t )
t 0 t
eij
EQ. 9.4.0.19
kk (t ) = 3K (t )
0
kk d
EQ. 9.4.0.20
where sij and eij are the stress and strain deviators. shear and bulk relaxation moduli.
P = K kk + Pair
with:
EQ. 9.4.1.1
Pair =
and:
P0 ; 1+
V
V 1 + 0 V0
EQ. 9.4.1.2
kk = ln V 0
EQ. 9.4.1.3
, the porosity, P0 the initial air pressure, 0 the initial volumetric strain and K the
bulk modulus. For deviatoric behavior, the generalized Maxwell model is used. The shear relaxation moduli in EQ. 9.4.0.19 is then defined as:
(t ) = Gl + Gs e t
Gs = G0 Gl
as the inverse of relaxation time
where G0 is the short time shear modulus, Gl the long time shear modulus and is the decay constant, defined
s : s = s
Gs
EQ. 9.4.1.6
with
The coefficients s , Gs and Gl are defined for the generalized Maxwell model as shown in Figure 9.4.5.
20-jan-2009
121
MATERIALS
governs the transition from the initial modulus G0 to the final modulus Gl .
m =
I1 is the mean stress, the deviatoric stresses sij at steps n and n+1 are computed by the 3
expressions:
n n n sij = ij ij m n +1 n &ij dt sij = sij +s
= 1 for i = j else = 0
with:
G + Gt 2G.Gt &ij = 2Ge &ij s sij (t ) + eij 0 0 G + Gt G.Gt &ij = Ge &ii s sii (t ) + eii 0 0
( for i j )
( for i = j )
EQ. 9.4.2.3
EQ. 9.4.2.4
20-jan-2009
122
MATERIALS
EQ. 9.4.2.5
Gt =
Et 2(1 + vt )
EQ. 9.4.2.6
& + E2 ). In EQ. 9.4.2.5 the coefficients A and B are defined for Young's modulus updates ( E = E1
Figure 9.4.6 Generalized Kelvin-Voigt model for RADIOSS law 35
KK t K + Kt dP &kk C2 kk + C3 kk = C1 K dt 3 + 2 0 3 + 2 0
where:
EQ. 9.4.2.7
K= Kt =
EQ. 9.4.2.8
EQ. 9.4.2.9
1 P = kk 3
EQ. 9.4.2.10
kk = ln V 0
and
EQ. 9.4.2.11
0 are the Navier Stokes viscosity coefficients which can be compared to Lame constants in elasticity. 2 V + 0 is called the volumetric coefficient of viscosity. For incompressible model, kk = 0 and
3
and
0 =
. In EQ. 9.4.2.11, C1, C2 and C3 are Boolean multipliers used to define different responses. For
example, C1=1, C2= C3=0 refers to a linear bulk model. Similarly, C1=C2=C3=1 corresponds to a visco-elastic bulk model. For polyurethane foams with closed cells, the skeletal spherical stresses may be increased by:
Pair =
P0 1+
EQ. 9.4.2.12
20-jan-2009
123
MATERIALS
is the volumetric strain, , the porosity, P0 the initial air pressure. In RADIOSS, the pressure may
an "equivalent air pressure" vs . The user can define this function used for open cell foams or for closed cell by defining a model identical to material law FOAM_PLAS (33) (see following sections).
9.4.3 Tabulated strain rate dependent law for viscoelastic materials (law 38)
The law incorporated in RADIOSS can only be used with solid elements. It can be used to model: polymers, elastomers, foam seat cushions, dummy paddings, hyperfoams, hypoelastic materials.
In compression, the nominal stress-strain curves for different strain rates are defined by user (Figure 9.4.7). Up to 5 curves may be input. The curves represent nominal stresses versus engineering strains. Figure 9.4.7 Nominal stress-strain curves defined by user input functions
The first curve is considered to represent the static loading. All values of the strain rate lower than the assumed static curve are replaced by the strain rate of the static curve. It is reasonable to set the strain rate corresponding to the first curve equal to zero. For strain rates higher than the last curve, values of the last curve are used. For a &1 and higher & 2 strain rates are & , two values of function at for the two immediately lower given value of read. The related stress is then computed as:
a & &1 = 2 + ( 1 2 ) 1 & & 2 1 b
EQ. 9.4.3.1
Parameters a and b define the shape of the interpolation functions. If a = b = 1, then the interpolation is linear. The figure 9.4.8 shows the influence of a and b parameters.
20-jan-2009
124
MATERIALS
The coupling between the principal nominal stresses in tension is computed using anisotropic Poisson's Ratio:
))
EQ. 9.4.3.2
t is the maximum Poisson's ratio in tension, c being the maximum Poisson's ratio in compression, and R, the exponent for the Poisson's ratio computation (in compression, Poisson's ratio is always equal to c ). In compression, material behavior is given by nominal stress vs nominal strain curves as defined by the user for different strain rates. Up to 5 curves may be input. The algorithm of the formulation follows several steps: 1. Compute principal nominal strains and strain rates. 2. Find corresponding stress value from the curve network for each principal direction. 3. Compute principal Cauchy stress. 4. Compute global Cauchy stress. 5. Compute instantaneous modulus, viscosity and stable time step. Stress, strain and strain rates must be positive in compression. Unloading may be either defined with an unloading curve, or else computed using the "static" curve, corresponding to the lowest strain rate (Figures 9.4.9 and 10).
20-jan-2009
125
MATERIALS
It should be noted that for stability reasons, damping is applied to strain rates with a damping factor:
EQ. 9.4.3.3
The stress recovery may be applied to the model in order to ensure that the stress tensor is equal to zero, in an undeformed state. A hysteresis decay may be applied when loading, unloading or in both phases by:
= H min (1, (1 e (t ) ))
EQ. 9.4.3.4
Where, H is the hysteresis coefficient, and the relaxation parameter. Confined air content may be taken into account, either by using a user defined function, or using the following relation:
V 1 V 0 Pair = P0 V V 0
EQ. 9.4.3.5
20-jan-2009
126
MATERIALS
dp &kk = K dt
Shear behavior is computed using a shear factor as follows:
EQ. 9.4.4.1
G (t ) = G + Gi e it
1
Pair =
with:
P0 1+ V 1 + 0 V0
EQ. 9.4.5.1
EQ. 9.4.5.2
20-jan-2009
127
MATERIALS
being the volumetric strain, , the porosity, P0 the initial air pressure,
s
0 the
initial volumetric
s &ij (t + t ) = ijs (t ) + ij E
E + Et
EEt s (t ) ij + ij t
EQ. 9.4.5.3
, or
y = A + B (1 + C )
s ij (t ) = ij (t ) (Pair ij )
1 3
128
MATERIALS
EQ. 9.4.6.3
The hypothesis of decomposition applied to the elastic potential allows separating the density of total initial energy into two parts:
= sp + dev
EQ. 9.4.6.4
where indices sp and dev denote respectively spherical and deviatoric parts. Using Hills criteria to express the density of hyperelastic potential for crushable foams, it can be written as a function of principal stretches ( 1 , 2 , 3 ) and deformation gradient:
0 (1 , 2 , 3 ) =
i =1
2i i 1 i i + 2 i + 3 i 3 + 1 J 2 1 i i
EQ. 9.4.6.5
and
i =
(1 2 i )
with
The reader is invited to consult references [101] and [102] for more details.
RADIOSS provides a material database incorporated in the installation. Many parameters are already defined by default and give accurate results. Some of them are described in the following sections: energy and pressure equations are solved simultaneously.
& (1 m ) &0
p = C0 + C1 + C2 2 + C3 3 + (C4 + C5 )En
20-jan-2009
129
MATERIALS
0 where = 0 melt
Material parameters are the same as in law 3. The parameters are: C = Strain rate coefficient
melt = Melting temperature max = Maximum Temperature. For T > Tmax : m=1 is used.
p C p = specific heat per unit volume 0
For an explanation about strain rate filtering, refer to chapter 9.3.1.2.
&ij Sij = 2 ve
p = C0 + C1 + C2 2 + C3 3 + (C4 + C5 )En
where Sij = Deviatoric stress tensor
= Kinematic viscosity
v=
EQ. 9.5.2.3
C 4 = C5 = 1
En 0 =
P0
A perfect gas allows compressibility and expansion and contraction with a rise in temperature. However, for many situations, especially very slow subsonic flows, an incompressible gas gives accurate and reliable results with less computation.
20-jan-2009
130
MATERIALS
C0 = C2 = C3 =C 4 = C5 = E0 = 0 C1 = 0 c 2
where, c = speed of sound
Incompressibility is achieved via a penalty method. The sound speed is set to at least 10 times the maximum velocity. This classical assumption is not valid when fluid and structures are coupled. In this case, set the sound speed in the fluid so that the first eigen frequency is at least 10 times higher in the fluid than in the structure.
EQ. 9.5.3.1
p = C0 + C1 + C2 2 + C3 3 + (C4 + C5 )En
where
EQ. 9.5.3.2
= / 1
0
C0, C1, C2, C3, C4, C5 = Hydrodynamic constants En = Energy per unit of initial volume The pressure and energy values are obtained by solving two linear equations:
Et + t = Et
1 1 1 (t + t + t ) p p 2 t t + t
0
EQ. 9.5.3.3
pt + t = F ( )t + t + G ( )t + t p Et + t
where F ( ) = C0 + C1 + C2 + C3
2 3
EQ. 9.5.3.4
G ( ) = C 4 + C 5
Input requires Young's or the elastic modulus, E, and Poisson's ratio, . These quantities are used only for the deviatoric part. They do not need to be consistent with the bulk modulus, C1. The plasticity material parameters are: A = Yield Stress B = Hardening Modulus n = Hardening Exponent
20-jan-2009
131
MATERIALS
A pressure cut off, pmin, can be given to limit the pressure in tension. The pressure cut off must be less than zero. Figure 9.5.1 shows a typical curve of the hydrodynamic pressure. Figure 9.5.1 Hydrodynamic Pressure Relationship
The coefficients C4 and C5 are used to control temperature effects. The initial energy per unit volume, En0, is also required. Simulations involving temperature effects can increase computational time substantially since energy and pressure equations are solved simultaneously.
Em = Ecold + cvTm
EQ. 9.5.4.1
Where Ecold is cold compression energy and Tm melting temperature supposed to be constant. If the internal energy E is less than Em , the shear modulus and the yield strength are defined by the following equations:
1 Em E E Ec G = G0 1 + b1 pV 3 h 300 e c v 1 E Ec Em E 1 + b2 pV 3 h 300 y =0 c e v
Where b1 , b2 , h and
f fE
fE
EQ. 9.5.4.2
EQ. 9.5.4.3
= 0 1 + p 0
]n
EQ. 9.5.4.4
20-jan-2009
132
MATERIALS
Failure Model CHANG ENERGY HASHIN JOHNSON LAD_DAMA PUCK TBUTCHER WILKINS FLD SPALLING TENSSTRAIN WIERZBICKI XFEM
d =
f
& * 1 + D5T & 0
EQ. 9.7.1.1
with:
]
* = m VM
EQ. 9.7.1.2
and the parameters Di the material constants. Failure is assumed to occur when d=1.
20-jan-2009
133
MATERIALS
VM : c = m + VM
EQ. 9.7.2.1
In a similar approach, a failure criteria based on a cumulative equivalent plastic strain was proposed by Wilkins. Two weight functions are introduced to control the combination of damage due to the hydrostatic and deviatoric loading components. The failure is assumed when the cumulative reaches a critical value dc. The cumulative damage is obtained by:
d c = W1W2 d p W1W2 ip
0
EQ. 9.7.2.2
i =1
with: W1 = where:
1 1 + aP
, P=
s2 s 2 1 3 jj , W2 = (2 A) , A = Max s ,s , s1 s2 s3 3 j =1 3 1
and
d = ( r ) dt
t
EQ. 9.7.3.1
material constant, t is the time when solid cracks and d another material
20-jan-2009
134
MATERIALS
major (%)
Failure zone
min (%)
1/ m
EQ. 9.7.6.1
max = C1e C
2
min = C3e C
4
and
are
for solids:
= =
m VM m VM
27 J 3 2 VM 3 27 2 1 2 3
with:
for shells:
m : hydrostatic stress
20-jan-2009
135
MATERIALS
f f
=const
max
min
-1 0
f
+1
t1 a damage factor D is applied to reduce the stress as shown in Fig. 9.7.3. The element is ruptured D=1. In addition, the maximum strains t1 and t 2 may depend on the strain rate by defining a scale
20-jan-2009
136
MATERIALS
where,
r = 0 (1 - D)b
20-jan-2009
137
Chapter
MONITORED VOLUME
20-jan-2009
138
MONITORED VOLUME
There are five types of MONITORED VOLUME: AREA Type: volume and surface output (post processing option, no pressure) PRES Type: user function defining pressure versus relative volume GAS Type: adiabatic pressure volume relation. P (V Vinc ) = Cst
with
cp cv
P = rT = (c p cv )T = ( 1)cvT = ( 1)
E V
COMMU Type: Chambered, communicating, folded airbag (airbag with communications) Same basic equations: Typical use of monitored volume is for tire, fuel tank, airbag.
For tire use PRES or GAS type monitored volume. For fuel tank use PRES or GAS type monitored volume. For simple unfolded airbag use AIRBAG type monitored volume. For chambered airbag use two ore more COMMU type monitored volumes. For folded airbag use a set of COMMU type monitored volumes.
20-jan-2009
139
MONITORED VOLUME
P(V Vi ) = P0 (V0 Vi )
= 1 a critical damping (shell mass and volume stiffness) is used. The viscous pressure q is:
q=
1 Pm fabric dV A V dt
where m fabric = At is the mass of fabric and A its surface. The applied pressure is P - Pext + q. The specific inputs for this type are: - -
- Pext - Po - Vi
20-jan-2009
140
MONITORED VOLUME
If the deflation is considered (isenthalpic outflow computation), the initial mass of gas, must also be input. This monitored volume is typically used to model tire pressure or simple fuel tank. For tire model Vi is zero and for fuel tank Vi is the fuel volume.
EQ. 10.3.1.1
E is the internal energy P is the pressure V is the monitored volume Vi incompressible volume
P=
( 1)E
V Vi
EQ. 10.3.1.2
where is the gas constant. For air, = 1.4. The two equations above allow the current volume to be determined. The energy and pressure can then be found.
P(t-dt) E(t-dt)
~ V = V Vi
W =
P(t ) + P(t dt ) ~ ~ V (t ) V (t dt ) 2
EQ. 10.3.2.1
P=
we have:
( 1)E
~ V ~ E (t ) E (t dt ) ( 1)V ~ + ~ V 2 (t ) V (t dt )
EQ. 10.3.2.2
W =
EQ. 10.3.2.3
20-jan-2009
141
MONITORED VOLUME
~ ~ ~ V = V (t ) V (t dt ) / 2 E = E (t ) E (t dt ) ~ 1 + ( 1) V W = 2 E 1 E + ~ ~ ~ V V V 1+ ~ 1 ~ 2V 2V
EQ. 10.3.2.4
~ V E W ( 1)E ~ 1 + V 2E
Computing the energy from basic principles:
EQ. 10.3.2.5
~ V E H out E = ( 1)E ~ 1 + V 2E
EQ. 10.3.2.6
H out can be estimated from, u(t-dt), the velocity at vent hole; this estimation will be described hereafter.
The variation of internal energy E can be given by:
EQ. 10.3.2.7
E (t ) = E (t dt ) + E
E (t ) P(t ) = ( 1) ~ V (t )
This pressure is then applied to the monitored volume to get: 1. New accelerations 2. New velocities
10.3.3 Venting
Venting, or the expulsion of gas from the volume, is assumed to be isenthalpic. The flow is also assumed to be unshocked, coming from a large reservoir and through a small orifice with effective surface area, A. Conservation of enthalpy leads to velocity, u, at the vent hole. The Bernouilli equation is then written as: (monitored volume)
1 vent
Pext
u2 (vent hole) 2
EQ. 10.3.3.1
20-jan-2009
142
MONITORED VOLUME
vent
Pext
(vent hole)
EQ. 10.3.3.2
EQ. 10.3.3.3
& out m
1/
Avent u
EQ. 10.3.3.4
E Avent u ~ V
EQ. 10.3.3.5
The vent hole area or scale factor area, Avent, can be defined in two ways: - a constant area taking into account a discharge coefficient - a variable area equal to the area of a specified surface, multiplied by a discharge coefficient. - a variable area equal to the area of the deleted elements within a specified surface, multiplied by a discharge coefficient.
u2 <
And,
2 2 2 P c = +1 +1
EQ. 10.3.4.1
Pcrit
2 1 = P +1
= 0.
20-jan-2009
143
MONITORED VOLUME
Only if the sloshing effect is neglected, pressure in a partial filled fuel tank can be modeled with a type GAS monitored volume. Use following input: - = 1.4 -
= 0.01
20-jan-2009
144
MONITORED VOLUME
criteria. t Pdef prevents deflation due to initial peak of pressure corresponding to the jetting activation. Figure 10.4.1 Uniform Pressure Airbag
P,V,, , ,
H in
H out
Injection of energy and mass Bag mechanics (i.e. unfolding, expansion, membrane tension, impacts ...) Exhaust through vent holes
EQ. 10.4.1.1
20-jan-2009
145
MONITORED VOLUME
P=
( 1)E
V
EQ. 10.4.1.2
where is the gas constant. For air, = 1.4 The two equations above allow the current airbag volume to be determined. The energy and pressure can then be found. To know the current airbag volume, derive energy and thus pressure. Considering a gas such that the constant pressure and the constant volume heat capacities per mass unit (respectively, cp and cv) vary in temperature T. The following temperature dependency of the constant pressure heat capacity is assumed:
c p = a + bT + cT 2
where a, b and c are the constants depending to characteristics of the gas. The cp and cv satisfy the Mayer relation:
EQ. 10.4.1.3
c p (T) - c v (T) =
R M
EQ. 10.4.1.4
With R is the universal gas constant depending to the unit system (R=8.3144 Jmol-1K-1):
cv (T ) =
de (T ) dT
h(t ) = e(t ) +
EQ. 10.4.1.5
Where, e is the specific energy and h the specific enthalpy per mass unit of the gas at temperature T. We can then obtain:
T
e(T ) = ecold +
Tcold
cv (T )dT
and
h(T ) = ecold +
Tcold
c (T )dT + M T
v
EQ. 10.4.1.6
Where, the lower index cold refers to the reference temperature Tcold. Now, assuming an ideal mixture of gas:
PV = nRT
with n the total number of moles:
EQ. 10.4.1.7
n = n (i ) =
i i
m (i ) M (i )
EQ. 10.4.1.8
where m(i) is the mass of gas i, and M(i) is the molar weight of gas i. It follows:
P=
with n =
nRT V
EQ. 10.4.1.9
m (i ) . (i ) i M
20-jan-2009
146
MONITORED VOLUME
Using EQ. 10.4.1.5, the variation of total gas energy can be written as:
T (t ) T ( t t ) (i ) (i ) (i ) (i) (i ) (i ) (i ) (i ) E = (m + m in m out ) e cold + cv (T )dT m e cold + cv (T )dT i Tcold Tcold i
) e
T ( t t ) (i ) cold
(i ) c ( T ) dT v Tcold
EQ. 10.4.2.2 On the other hand, the basic energy equation EQ. 10.4.1.1 of the airbag and the expression of enthalpy in EQ. 10.4.1.5 gives:
T ( i ) in R (i ) (i ) (i ) (i ) E = m in e cold + (i ) T in + cv (T )dT M Tcold i T ( i ) out (i ) R (i ) (i ) m out e cold + (i ) T out + cv (i ) (T )dT W M Tcold i
EQ. 10.4.2.2
Where m(i)in and T(i)in are characteristics of the inflator and are considered as input to the problem. m out and T(i)out can be estimated from the velocity at vent hole u(t). W is the variation of the external work. This estimation will be described hereafter. It comes from EQ. 10.4.2.1 and EQ. 10.4.2.2:
(i)
(m
i
(i )
+ m (i ) in m( i ) out
) c
T (t )
(i )
v T ( t t )
(T )dT =
EQ. 10.4.2.3 The variation of the external work can be written as:
W =
(P(t ) + P(t t )) (V (t ) V (t t ))
2
EQ. 10.4.2.4
20-jan-2009
147
MONITORED VOLUME
(i ) (i ) (i ) m + m in m out i
) c
T (t )
(i )
v T ( t t )
m (i ) + m( i ) in m (i ) out V (t ) V (t t ) RT (t ) (T )dT + (i ) M 2V (t ) i
EQ. 10.4.2.6
T (t )
v T (t t )
(i )
(i ) (i ) (i ) (i ) m + m in m out cv (T t t )(T (t ) T (t t )) + i (i ) (i ) (i ) m + m in m out V (t ) V (t t ) RT (t ) (i ) 2V (t ) M i EQ. 10.4.2.7 (i) T in T ( i ) out R i (i ) ( ) i (i ) R (i ) ( ) = m in T in + cv (T )dT m out ( i ) Tout + cv (T )dT M (i ) i M i T ( t t ) T ( t t )
m(i ) V (t ) V (t t ) M ( i ) RT (t t ) 2V (t t ) i
Which allows to determine the actual temperature T(t). The actual pressure then computed from the equation of perfect gas (EQ. 10.4.1.8).
20-jan-2009
148
MONITORED VOLUME
The mass versus time curve can be derived from the pressure curve if Tin is known:
& = m
& VM P RTin
EQ. 10.4.3.1
Where, M is the molecular weight of the injected gas. R is the perfect gas constant : The average estimate for temperature of injection is:
cp R = c p cv ; = . cv M
Tin =
where:
P VM m R
EQ. 10.4.3.2
P is the total pressure variation during the experiment. m is the total injected mass, which can be derived from the mass of propellant in the pyrotechnic inflator and the chemical reaction; 40% is a typical value for the ratio of the produced mass of gas to the solid propellant mass.
1
P
1 vent
Pext
u2 2
(vent hole)
EQ. 10.4.4.1
vent
Pext
(vent hole)
EQ. 10.4.4.2
EQ. 10.4.4.3
20-jan-2009
149
MONITORED VOLUME
with
m
i
(i )
averages of heat capacities at constant pressure and constant volume. The mass flow rate is given by:
1/
Avent u
EQ. 10.4.4.4
1/
Avent u
E V
P = ext P
1/
EQ. 10.4.4.5
The total mass flow rate is given by dmout where Avent is the vent hole surface.
Avent u
EQ. 10.4.4.6
The vent hole area or scale factor area, Avent, can be defined in two ways:
a constant area taking into account a discharge coefficient a variable area equal to the area of a specified surface multiplied by a discharge coefficient.
10.4.4.1 Supersonic outlet flow Vent pressure Pvent is equal to external pressure Pext for unshocked flow. For shocked flow, Pvent is equal to critical pressure Pcrit and U is bounded to critical sound speed:
u2 <
And,
2 2 2 P c = +1 +1
EQ. 10.4.4.7
Pcrit
2 1 = P +1
Pvent = max(Pcrit , Pext) 10.4.4.2 Outgoing mass per gas The mass flow of gas i is dm
(i )
out
V (i ) =
It comes finally
dm (i ) out =
n (i ) dmout n (i )
i
EQ. 10.4.4.8
20-jan-2009
150
MONITORED VOLUME
10.4.5 Porosity
The isenthalpic model is also used for porosity. In this case, one can define the outgoing surface by the following equation:
AOUT = f ( A) g ( P) h(t )
where:
EQ. 10.4.5.1
: Scale factor
It is also possible to define closure of the porous surface in the case of contact. The second model integrated in RADIOSS is called Chemkin model in which the mass flow due to the porosity is computed by:
EQ. 10.4.5.2
u = f (P Pext )
EQ. 10.4.5.3
A2
- Initial mass energy density is defined from the above values. There is no need to define an injected mass at time zero.
EQ. 10.4.7.1
20-jan-2009
151
MONITORED VOLUME
r n1 being the normalized vector between the projection of the center of the element upon segment
(N1,N3) and the center of element as shown in the figure.
is the angle between the vector MN 2 and the vector n1 . the distance between the center of the element and its projection of a point upon segment (N1, N3).
The projection upon the segment (N1, N3) is defined as the projection of the point in direction MN 2 upon the line (N1, N3) if it lies inside the segment (N1, N3). If this is not the case, the projection of the point upon segment (N1, N3) is defined as the closest node N1 or N3. If N3 coincides to N1, the dihedral shape of the jet is reduced to a conical shape. Figure 10.4.3 Jetting effect schema
20-jan-2009
152
MONITORED VOLUME
With a tank test it is possible to measure temperature at injection point or in the middle of the tank. For pressure the two values are equal. Figure 10.4.5 Tank experiment schema
EQ. 10.4.9.1
dT dP P T V dt dt & m= ( 1)CV T 2
EQ. 10.4.9.2
20-jan-2009
153
MONITORED VOLUME
m=
PV ( 1)CV T T2 dP dt
EQ. 10.4.9.3
Tin =
dT dP T P dt dt
EQ. 10.4.9.4
Or,
dT Tin = + dt & m T m
And if T is constant: Tin = T/
EQ. 10.4.9.5
20-jan-2009
154
MONITORED VOLUME
Injection of energy and mass Bag mechanics (i.e. unfolding, expansion, membrane tension, impacts ...) Exhaust through vent holes
This option is used to simulate chambered airbags and may be used to unfold an airbag. Each COMMU type monitored volume works like an AIRBAG type monitored volume with possible vent communication with some other monitored volume of COMMU type. A chambered airbag is therefore modeled with two or more COMMU type monitored volumes. Each monitored volume can have an inflater and an atmospheric vent hole. Monitored volume 1 can communicate with monitored volume 2 with or without communication from 2 to 1. Communicating area, deflation pressure or time from 1 to 2 can be different from corresponding values from 2 to 1. It is thereby possible to model a valve communication.
10.5.4 Venting
Same equations as for AIRBAG type monitored volume are used. Refer to Chapter 10.4.4. The mass flow rate is given by:
1/
Avent u
EQ. 10.5.4.1
1/
Avent u
E V
EQ. 10.5.4.2
These mass and energy flux are removed from the current volume and added to the communicating volume at next cycle.
20-jan-2009
155
MONITORED VOLUME
20-jan-2009
156
MONITORED VOLUME
M in
10.5.8.2 Example 2
Volume 1 communicates with volume 2 and volume 2 with volumes 1 and 3, but there is no communication from 3 to 2. Two COMMU type monitored volume communications can have common nodes or common shell property sets but this is optional. To model a folded airbag, one COMMU type monitored volume is used for each folded part. The boundary between two folded parts is closed with a dummy property set (fictitious property). The pressure in each folded part will be different and the area of communication will increase during inflation. With this model, the volume with inflater will inflate first and before than folded parts (better than "jetting" model).
20-jan-2009
157
MONITORED VOLUME
M in
Pext
10.5.8.3 Example 3
Volume 1 and volume 2 with common property set.
20-jan-2009
158
MONITORED VOLUME
Communication area Vol. 1 to 2: prop. 4 Vol. 2 to 1: prop. 4 Vol. 1 to 3: prop. 5 Vol. 3 to 1: prop. 5
Property
Property
Property
Dummy property
Dummy property
20-jan-2009
159
MONITORED VOLUME
Chapter
STATIC
20-jan-2009
160
STATIC
11.0 STATIC
Explicit scheme is generally used for time integration in RADIOSS, in which velocities and displacements are obtained by direct integration of nodal accelerations as described in chapter 4. With this approach, the time step is often small due to stability condition. For the static solution of structural mechanical problems as the steady state is a part of the transient response for a temporal-step load, the use of explicit scheme is usually possible if the computation time remains reasonable. However, in static or slow dynamic computations as duration of the study is large, many cycles are necessary to carry out the simulation. To resolve static problems, an alternative to explicit method is the implicit time-integration scheme. In this method, a system of nonlinear equations is obtained and then resolved by Newton-Raphson method. It can be shown that the implicit scheme is always stable. That results in a large time step with the explicit method. However, as a global stiffness matrix should be assembled and inverted, the method is relatively high cost per loading step. The primary difference between the explicit and implicit methods is that an explicit algorithm obtains the next value from known previous values. An implicit method assumes a solution to a problem and solves the equations simultaneously. As the global equilibrium equation is generally nonlinear, an iterative numerical resolution is generally used. The implicit method might fail when:
The material law is highly nonlinear. Complicated material behavior is easier to accommodate using an explicit method. The number of elements is too large. Explicit method does not require large matrix inversion, the I/O is less important and the memory required is also less. Matrices must be re-evaluated at each time step and for most of the iterations. The problem includes several contacts. Contact algorithms are very efficient in explicit programs. The static analysis is a pre-loading case before a fully dynamic behavior phase. In this case, the coupling of two phases is very common. Explicit approaches furnish an alternative to the previous cases.
In RADIOSS V5 both of implicit and explicit methods are available to study the static behavior of systems. The choice a method depends on the nature of the problem and the engineers feeling. The explicit approach is especially attractive for problems with highly nonlinear geometric and material behavior as all quantities may be treated as vectors, resulting in low storage requirements. The number of cycles to achieve convergence may be quite large, but global efficiency is generally observed. The implicit method is introduced to study efficiently static applications such as spring back in sheet metal forming or gravity loading or other initial state computations before / after dynamic simulations.
The loading is applied at a rate sufficiently slow to minimize the dynamic effects. The final solution is obtained by smoothing the curves. In case of elasto-plastic problems, one must minimize dynamic overshooting because of the irreversibility of the plastic flow.
20-jan-2009
161
STATIC
This method was first introduced by Otter [19] and has been used in several hydrodynamic codes (e.g. HEMP [2]). A nodal damping Cv is added to the momentum equation:
dv + C = f ext f int dt Fc = CV
EQ. 11.1.1.1
Fk = KX = Fkt t / 2 + KVt
0 =
Fk M
1 = 0 +
Fc CV = 0 M M
Vt + t / 2 = Vt t / 2 + 1 t
1 V = Vt t / 2 + 1t 2
Vt + t / 2 = Vt t / 2 + 0 t
We have:
CV t M
EQ. 11.1.1.8
Vt + t / 2 = Vt t / 2 + 0 t
1 C Vt t / 2 + 1t t 2 M t
EQ. 11.1.1.9
1 C Ct Vt + t / 2 = 1 Vt t / 2 + 0 1t M M 2
EQ. 11.1.1.10
C 1 Ct 0 t Vt + t / 2 = 1 1 t Vt t / 2 + M M 0
Approximation
EQ. 11.1.1.11
, we obtain :
Vt + t / 2 = (1 2 )Vt t / 2 + (1 ) 0 t
=C
2M t C = 2M t
20-jan-2009
162
STATIC
C = AM
or A =
and
A=
2 t
EQ. 11.1.1.14
2 t
EQ. 11.1.1.15
EQ. 11.1.1.16
C=
2M
EQ. 11.1.1.17
Where is the relaxation coefficient whose recommended value is 1. is less than or equal to the highest period of the system. These are the input parameters used in /DYREL option. The explicit time integration scheme is changed to compute the new velocities. The explicit time integration given in chapter 4 gives (EQ. 4.1.2.7):
Vt + t / 2 = Vt t / 2 + t t
which is now written as:
EQ. 11.1.1.18
Vt + t / 2 = (1 2 )Vt t / 2 + (1 ) t t
where:
EQ. 11.1.1.19
EQ. 11.1.1.20
This empirical methodology consists in setting to zero the nodal velocities each time the Kinetic Energy reaches a maximum.
Rayleigh damping
[ ]
[ ]
[ ]
EQ. 11.1.1.21
where and are the pre-defined constants. In modal analysis, the use of a proportional damping matrix allows to reduce the global equilibrium equation to n-uncoupled equations by using an orthogonal transformation. If the global equilibrium equation is expressed as:
EQ. 11.1.1.22
20-jan-2009
163
STATIC
EQ. 11.1.1.23
+ 12 0 0 + 2 2 T With [ ] [C ][ ] = . . . . 0 .
Each uncoupled equation is written as:
2 t && + 2 & i i i i + i i = f i
. . . .
. . . . . . 2 . . + n 0 0
EQ. 11.1.1.24
With 2 ii = + i Where
This leads to a system of n equations with two unknown variables and . Regarding to the range of the dominant frequencies of system, two frequencies are chosen. Using the pair of the most significant frequencies, two equations with two unknown variables can be resolved to obtain values for and . For high frequencies the role of
is more significant. However, for lower frequencies plays an important role (Fig. 11.1.1).
Fig. 11.1.1 Rayleigh damping variation for natural frequencies
Nonlinear part
2 i
Quasi-linear part
i
2
The Rayleigh damping method applied to explicit time-integration method leads to the following equations:
t t M t + Cv t = Fext Fint
EQ. 11.1.1.27
20-jan-2009
164
STATIC
With
C = M + K
t t dt Fint = Fint + Kv dt 2 t dt 2
dt
EQ. 11.1.1.28
vt = v
+t
dt 2
t t t M 0 = Fext Fint
Neglecting
, in
Kv t evaluation we have:
Kv = Kv
t
t t dt t t dt Fint 0 0 Fint = M dt dt
EQ. 11.1.1.31
And finally:
t t t ) 0 = M 1 (Fext Fint
t dt 2
EQ. 11.1.1.32
=
t
0 v
dt dt 1+ 2
t 0
t dt 0
)
EQ. 11.1.1.33
t+
dt 2
=v
dt 2
+ t dt
EQ. 11.1.1.34
The three approaches available in RADIOSS are Dynamic Relaxation (/DYREL), Energy Discrete Relaxation (/KEREL) and Rayleigh damping (/DAMP). Refer to RADIOSS Example Manual for application examples.
20-jan-2009
165
STATIC
The response to different load systems can be obtained by superposition, Removing all loads returns the structure to the reference position.
Perfect linear elasticity for any deformation, Infinitesimal deformation, Infinite strength.
Despite of obvious physically unrealistic limitations, the linear model can be a good approximation of portions of nonlinear response. As the computational methods for linear problems are efficient and low cost, RADIOSS linear solvers can be used to find equilibrium of quasi-linear systems. The Preconditioned Conjugate Gradient method is the iterative linear solver available in RADIOSS. The algorithm is allowing to save a lot of memory for usual application of RADIOSS as a sparse storage method is used. This means that only the non-zero terms of the global stiffness matrix are saved. In addition, the symmetry property of both stiffness and preconditioning matrices is worthwhile to save memory. The performance of conjugate gradient method depends highly to the preconditioning method. Several options are available in RADIOSS using the card /IMPL/SOLV/1. The simplest method is a so-called Jacobi method in which only the diagonal terms are taken into account. This choice is allowing to save considerable memory space; however, the performance may be poor. The incomplete Choleski is one of the best known effective preconditioning methods. However, it can result negative pivots in some special cases even if the stiffness matrix is definite positive. This results a low convergence of PCG algorithm. The problem can be resolved by using a stabilization method as described in [103]. Finally, the Factored Approximate Inverse method may be the best choice which is used by default in RADIOSS.
which is a
n (X + X ) = 0 (X ) +
2 1 X + X T X + ... X X 2 2
EQ. 11.2.2.1
where the subscripts n and 0 denote respectively final and initial configurations. The term
can be identified X
as the out-of-balance forces or gradient F, of the total potential energy which is the difference between the internal force vector Fint and the external force vector Fext. The term
KT. The principle of minimum energy and the equilibrium of stable state give:
= n (X + X ) 0 (X ) = 0
20-jan-2009
166
STATIC
EQ. 11.2.2.3
X T K TX > 0
EQ. 11.2.2.4
EQ. 11.2.2.3 can handle the solution of the nonlinear problem when an incremental method is used. The solution methods are generally based on continuous incremental and corrective phases. The most important class of corrective methods concerns the Newton-Raphson method and its numerous variants as modified, modifieddelayed, damped, quasi and so forth. All of these Newton-like methods require access to the past solution. In the following section we study the conventional and modified Newton methods under general increment control.
EQ. 11.2.2.5
depending on only one loading parameter . If the loading depends to only one loading variable independent to the state of deflection, we have:
with Fext (X, ) = FC + Fext (X) . This equation represents a system of n algebraic nonlinear equations
EQ. 11.2.2.6
Several techniques are available to resolve EQ. 11.2.2.5. In some situations, the parameter is fixed, and the equations are resolved to determine n components of X in order to verify EQ. 11.2.2.5. In this case, the technique is called load control method. Another technique called displacement control consists in fixing a component of X and searching for and n-1 other components of displacement vector X. A generalization of displacement control technique is allowing to imply several components of displacement vector by using an Euclidian norm. The method is called arc-length control and intended to enable solution algorithms to pass limit points (i.e. maximum and minimum loads). The techniques making possible to obtain the load-deflection curve by finding point by point the solution are called piloting techniques. When the piloting technique is chosen for a given step, the associated solution is obtained by an iterative resolution of so-called Newton-Raphson methods. At iteration i, the residual vector Ri is:
R i = Fext (X i , i ),-Fint (X i )
A correction
EQ. 11.2.2.7
R R R i +1 = R i + X + X
Combining EQ.11.2.2.8 with EQ. 11.2.2.7, we obtain:
i i KT X - Fext = R i
EQ. 11.2.2.8
EQ. 11.2.2.9
as R
i +1
= 0 and:
X i +1 = X i + X
20-jan-2009
167
STATIC
EQ. 11.2.2.10
i +1 = i +
i i
The tangent matrix K T is obtained by assembling the elementary matrices k T . It corresponds to:
KT =
Fint Fext X X
EQ. 11.2.2.11
Using load control technique, the standard Newton-Raphson method resolves EQ. 11.2.2.9 to EQ. 11.2.2.11 by applying a known load increment as illustrated in Figure 11.2.1. The tangent matrix is updated and triangulized at each iteration. This insures a quadratic convergence to exact solution. Figure 11.2.1 Standard Newton-Raphson resolution in the case of load control technique
iterations
KT
i
Fn = n F
Fn = n F
Xn
X n
Figure 11.2.2 Modified Newton-Raphson resolution in the case of load control technique
iterations
KT Fn = n F
Fn = n F
Xn
X n
20-jan-2009
168
STATIC
However, it is possible to save computation time which depends on the size of the problem and on the degree of the nonlinearity of the problem. The method is called modified Newton-Raphson which is based on the conservation of the tangent matrix for all iterations (Figure 11.2.2). This method can also be combined with the acceleration techniques as line-search explained in the following section. The convergence criteria may be based on Euclidian norm of residual forces, residual displacements or energy where an allowable tolerance is defined.
X i +1 = X i + X
EQ. 11.2.2.12
where is obtained to minimize the total potential energy or to satisfy the principle of virtual works. The techniques to determine use often a Raleigh-Ritz procedure with only one unknown parameter. The principle of virtual work can be written in the general form:
W (X, X ) = X R (X ) = 0
Considering EQ. 11.2.2.12, we can write:
EQ. 11.2.2.13
X = X
and:
EQ. 11.2.2.14
W = X R (X i + X ) = 0
Then
for all
EQ. 11.2.2.15
X R (X i + X ) = 0
which leads to a three-order polynomial equation in
C1+C 2 +C 3 2 + C4 3 = 0
The coefficients C1, C2, C3 and C4 can be expressed in terms of displacements Xi and the increment of displacements X .
20-jan-2009
169
STATIC
Figure 11.2.3 Various load-deflection curves and step-by-step solution by arc-length method
Static load
Static load
Dynamic B
A B
C
Displacement
(a) Snap-through (b) Snap-backs
Displacement
Static load
Static load
A n-1 n n+1 B
PCr PS
Displacement
Displacement
(c) Arc-length method: intersection of the equilibrium branch with the circle about the last solution The tracing of equilibrium branches are quite difficult. In arc-length method, instead of incrementing the load parameter, a measure of the arc length in the displacement-load parameter space is incremented. This is accomplished by adding a controlling parameter to the equilibrium equations. The arc-length method was originally introduced by Riks [116] and Wempner [117]. Considering a function f implying several components of the displacement vector X, the arc-length method consists in determining in each step the Euclidian norm of the increase in X: (d) Buckling with or without imperfections
f = X i +1 X n {X i +1 X n } (S n ) = 0
2
EQ. 11.2.2.18
X i +1 X n {X i +1 X n } = (S n )
And
a ( ) + b( ) + c = 0
2
20-jan-2009
170
STATIC
With:
a = X F {X F } ;
{X F } = [KTn ]1{Fext }
b = 2 X F {Y } ; c = Y {Y } (S n )
2
{Y } = {X F }+ {X i X n }
In each of the Newton-Raphson iterations, EQ. 11.2.2.20 must be resolved to select a real root. If there is no root, S n should be reduced. The most closed root to the last solution is retained in the case of two real roots. Figure 11.2.3(c) illustrates the intersection of the equilibrium branch with the circle about the last solution.
20-jan-2009
171
Chapter
RADIOSS PARALLELIZATION
20-jan-2009
172
RADIOSS PARALLELIZATION
S (P ) =
(T
EQ. 12.1.0.3
where TPar and TSeq are the computation times respectively related to parallel and non-parallel parts. As
The limit value can be obtained when the process number tends to infinite:
S ( ) =
1 TSeq
EQ. 12.1.0.5
Table 12.1.1 gives the speedup in function of number of processors and the rate of parallelization in the program. It can be seen that if the rate of parallelization is less 95%, the computation acceleration will not be greater than 20 however the number of processors. This means that to obtain a good scalability of a code, at least 99% of the program must be parallel.
20-jan-2009
173
RADIOSS PARALLELIZATION
Table 12.1.1 Speed-up in function of process number and parallelization Process number
Seq / P
16
32
64
128
Memory
C P U 1
C P U 2
...
C P U
In this case, all processors can access to a common memory space. From programming point of view, each process called parallel task, reach to the entire memory space allocated by the program. It is necessary to manage properly the access to this shared memory by introducing barriers and locking mechanisms. The SMP model programming has the advantage to be managed easily. However, the performance of the method depends on the ratio between the memory access speed and the CPU speed. The parallelism approach used in RADIOSS SMP is a multi-task programming type. The tasks are explicitly managed by the programmer. The computation tasks are attributed to the processors by a dynamic procedure as they are available. This approach is especially adapted to the super-computers used as computation server where the load of a given processor varies with respect to others. The SMP version is developed for computers with shared memory architecture and cannot be used efficiently on the super-computers with distributed memory or cluster structures. The RADIOSS SMP version is constantly improved since the first release. However, the efficiency of SMP version has to be updated to take into account the evolution of computer architecture in memory management and CPU speed.
20-jan-2009
174
RADIOSS PARALLELIZATION
processors with a better efficiency. It makes possible to use up to 64 processors. In addition, all RADIOSS Crash options are available in this version including Arithmetic Parallel option. The principle of program is based on Single Program Multiple Data where the same program runs with different data. RADIOSS Starter carries out a domain decomposition. Then, RADIOSS Engine has just to send data to different processors in an initialization step. Thereafter, each program runs over each sub domain. It is necessary to communicate information between processors to manage data on the border of domains. This is carried out in RADIOSS using MPI (Message Passing Interface) library. Figure 12.3.1 illustrates the architecture of muti-processor computers with distributed memory. The RADIOSS SPMD version runs independently to architecture of memory as well on the computers with shared memory or distributed memory or a set of work stations in a network.
Network interconnect
C P U 1
C P U 2
...
C P U P
M1
M2
...
Mp
20-jan-2009
175
RADIOSS PARALLELIZATION
References
20-jan-2009
176
REFERENCES
13.0 REFERENCES
[1] Mendelson A. Plasticity: Theory and Application MacMillan Co., New York, 1968. [2] Wilkins M., Calculation of elastic plastic flow LLNL, University of California UCRL-7322, 1981. [3] Noh W., Numerical methods in hydrodynamics calculations LLNL, Univer. of Calif. UCRL-52112, 1976. [4] Zienkiewicz O.C., The Finite element method in engineering science Mc Graw Hill, 1971. [5] Hallquist J., Theoretical Manual for DYNA3D LLNL, University of California, 1983. [6] Hughes T., Taylor R. , Kanoknukulchai , A simple and efficient F.E. for plate element Int. Journal Numerical Methods in Engineering , 11 1529-1543, 1977. [7] Belytschko T., Hsieh B. , Non Linear Transient Finite Element Analysis with convected coordinates Int. Journal Num. Methods in Engineering, 7 255-271, 1973. [8] Donea J. et al, An Arbitary Lagrangian-Eulerian finite element method for transient dynamic fluid-structure interactions Computer methods in applied mechanics, 1982. [9] Von Neumann J. Richtmeyer R., A method for the numerical calculation of hydrodynamical shocks Journal of applied physics, 1950. [10] Kosloff D. Frazier G. Treatment of hourglass pattern in low order finite element code International journal for numerical and analytical methods in geomechanics, 1978. [11] Johnson G., Cook W. A constitutive model and data for metals subjected to large strains, high strain rates and high temperatures, 7th Symposium on Ballistics, The Hague, 1983. [12] Flanagan D., Belytschko T. A Uniform Strain Hexahedron and Quadrilateral with Orthogonal Hourglass Control Int. Journal Numerical Methods in Engineering 17 679-706, 1981. [13] Iliouchine A. Plasticity Edition Eyrolles Paris, 1956. [14] Halphen B., On the velocity field in thermoplasticity finished , Laboratoire de Mcanique des Solides, Ecole Polytechnique, International Journal of Solids and Structures, Vol.11, pp 947-960, 1975. [15] Zeng Q., Combescure A., "A New One-point Quadrature, General Nonlinear Quadrilateral Shell Element with Physical Stabilization" Int. J. Numer. Meth. Engng. 42, 1307-1338, 1998. [16] Belytschko T., Lin J.L. and Tsay C.S. Explicit algorithms for the nonlinear dynamics of shells Computer Methods in Applied Mechanics and Engineering, 42:225-251, 1984. [17] Belytschko T. and Leviathan I. Physical stabilization of the 4-node shell element with one-point quadrature, Computer Methods in Applied Mechanics and Engineering, 113:321-350, 1992. [18] Belytschko T. and Leviathan I. Projection schemes for one-point quadrature shell elements, Computer Methods in Applied Mechanics and Engineering, 115:277-286, 1993. [19] Otter J.R.H., Mechanical Engineering and Design 3, 1965. [20] Zienkiewicz O.C. Lohner R., Accelerated Relaxation or Direct Solution, Int. J. for Numerical methods in Engineering, 1985. [21] Wilson L.T. et al. Incompatible displacement models, page 43. Academi Press, New York, 1973. [22] Hughes T.J.R.and Liu W.K. Nonlinear finite element analysis of shells: Part I: Three-dimensional shells, Computer Methods in Applied Mechanics and Engineering, 26:331-362, 1981. [23] Belytschko T. and Tsay C.S. A stabilization procedure for the quadrilateral plate element with one-point quadrature, Computer Methods in Applied Mechanics and Engineering, 55:259-300, 1986. [24] BelytschkoT. ,Wong B.L. and Chiang H.Y. Advances in one-point quadrature shell elements, Computer Methods in Applied Mechanics and Engineering, 96:93-107, 1989. [25] Hughes T.J.R. and Tezduyar T.E.. Finite elements based upon Mindlin plate theory with particular reference to the four-node bilinear isoparametric element, J. of Applied Mechanics, 48:587-596, 1981. [26] MacNeal R.H. Derivation of element stiffness matrices by assumed strain distributions, Nuclear Engrg. Des., 70:3-12, 1982.
20-jan-2009
177
REFERENCES
[27] Belytschko T. and Bachrach W.E. Efficient implementation of quadrilaterals with high coarse-mesh accuracy, Computer Methods in Applied Mechanics and Engineering, 54:279-301, 1986. [28] Engelmann B.E. and Whirley R.G. A new elastoplastic shell element formulation for DYNA3D. Report ugrl-jc-104826, Lawrence Livermore National Laboratory, 1990. [29] Liu W.K., Law E.S., Lam D. and Belytschko T. Resultant-stress degenerated-shell element, Int. J. Numer. Methods Engrg, 19:405-419, 1983. [30] Zhu Y. and Zacharia T. A new one-point quadrature, quadrilateral shell element with drilling degree of freedom, Computer Methods in Applied Mechanics and Engineering, 136:165-203, 1996. [31] Belytschko T. and Bindeman L. Assumed strain stabilization of the 4-node quadrilateral with 1-point quadrature for nonlinear problems, Computer Methods in Applied Mechanics and Engineering, 88:311340, 1991. [32] Dvorkin E. and Bathe K.J. A continuum mechanics four-node shell element for 35 general nonlinear analysis, Engrg Comput, 1:77-88, 1984. [33] Stolarski H., Belytschko T. and Lee S.H. A review of shell finite elements and corotational theories, Computational Mechanics Advances, 2:125-212, 1995. [34] R.W Ogden, Non-Linear Elastic Deformations, Ellis Horwood, 1984. [35] T. Belytschko and T.J.R. Hughes, Computational Methods for Transient Analysis, North-Holland, 1983. [36] T. Belytschko, Wing Kam Liu, B. Moran, Finite Elements for Nonlinear Continua and Structures, John Wiley, 1999. [37] M. Geradin, Structural Dynamics , Masson, 1993. [38] S. Ahmad, B.M. Irons and O. C. Zienkiewicz. Analysis of thick and thin shell structures by curved finite elements, Computer Methods in Applied Mechanics and Engineering, 2:419-451, 1970. [39] D.J. Allman. A quadrilateral finite element including vertex rotations for plane elasticity problems, Int. J. Numer. Methods Engrg, 26:717-739, 1988. [40] J.L. Batoz, G.Dhatt., Modeling of Structures by finite element , volume 3, HERMES, 1992. [41] T.J.R. Hughes and F. Brezzi. On drilling degrees of freedom, Computer Methods in Applied Mechanics and Engineering, 72:105-121, 1989. [42] G.H. MacNay II. Numerical modeling of tube crush with experimental comparison, Proceedings of the 7th Int. Conference on Vehicle Structural Mechanics, Society of Automotive Engs., pages 123-134, 1988. [43] Antoine Letellier. Contribution to the modeling of impacts of birds on the blades of aircraft engines. PhD thesis, University d'Evry, 1996. [44] R.H. MacNeal and R.L. Harder. A proposed standard set of problems to test finite element accuracy, Finite Elements Anal. Des., 11:3-20, 1985. [45] L. Morino, J.W. Leech and E.A. Witmer. An improved numerical calculation technique for large elasticplastic transient deformations of thin shells: Part 2-evaluation and applications, J. of Applied Mechanics, 38:429-436, 1971. [46] J.C. Simo and T.J.R. Hughes. On variational foundations of assumed strain methods, J. of Applied Mechanics, 53:51-54, 1986. [47] R.L. Taylor. Finite element for general shell analysis. 1979. [48] P. Underwood. Dynamic relaxation, chapter 5, pages 245-265. Elsevier Science, 1983. [49] P. C. Hammet, O.P. Marlowe & A.H. Stroud. Numerical integration over simplexes and cones, Math. Tables Aids Comp, 10, 130-7, 1956. [50] J.H. Argyris, An excursion into the large rotations, Computer Methods in Applied Mechanics and Engineering, 32, 85-155, 1982. [51] M. A. Crisfield, A consistent corotational formulation for nonlinear three-dimensional beam element, Computer Methods in Applied Mechanics and Engineering, 81, 131-150, 1990.
20-jan-2009
178
REFERENCES
[52] J.C. Simo, A finite strain beam formulation, Part I: The three-dimensional dynamic problem, Computer Methods in Applied Mechanics and Engineering, 49, 55-70, 1985. [53] G. A. Wempner, F.E., finite rotations and small strains of flexible shells, IJSS, 5, 117-153, 1969. [54] K.J. Bathe, Finite Element Procedures, Prentice Hall, 1996. [55] T. J. R. Hugues, Analysis of Transient Algorithms with Particular Reference to Stability Behavior Computational Methods for Transient Analysis, eds. T. Belytschko and T. J. R Hugues, 67-155, North Holland, 1983. [56] Courant R., K.O. Friedrichs, H. Levy, ber die partiellen Differenzensleichungen der Mathematiscen Physik, Math. A nn., Vol. 100, 32, 1928. [57] Tsai S.W., Wu E.M., A general theory of strength for anisotropic materials, Journal of Composite Materials, 58-80, 1971. [58] CHANG, F.K. and CHANG, K.Y., A Progressive Damage Model For Laminated Composites Containing Stress Concentrations, Journal of Composite Materials, Vol 21, 834-855, 1987. [59] CHANG, F.K. and CHANG, K.Y., Post-Failure Analysis of Bolted Composites Joints in Tension or ShearOut Mode Failure, Journal of Composite Materials, Vol 21, 809-833, 1987. [60] Matzenmiller, A. and Schweizerhof, K., Crashworthiness Simulation of composite Structures-a first step with explicit time integration Non linear Computational Mechanics- State of the Art Ed. p. Wriggers and W. Wagner, 1991. [61] Han Zhao, "A Constitutive Model for Metals over a Large Range of Strain Rates.", Materials Science & Engineering, A230, 1997. [62] Han Zhao, Gerard Gary,"The Testing and Behavior Modelling of Sheet Metals at Strain Rates from 10.e-4 to 10e+4 s-1", Materials Science & Engineering" A207, 1996. [63] Drucker D., Prager W., "Soil mechanics and plastic analysis of limit design", Quart. Appl. Math., Vol. 10, 157-165, 1952. [64] Gurson A. L. "Continuum theory of ductile rupture by void nucleation and growth: Part I Yield criteria and flow rules for porous ductile media", Journal of Engineering Materials and Technology, Vol. 99, 2-15, 1977. [65] Ted Belytschko and Lee P. Bindeman "Assumed strain stabilization of the eight node hexahedral element", Computer Methods in Applied Mechanics and Engineering, vol.105, 225-260, 1993. [66] Ted Belytschko, Itai Leviathan Projection schemes for one-point quadrature shell elements, Computer Methods in Applied Mechanics and Engineering, Vol. 115, 227-286, 1994. [67] Lighthill, M.J., On Sound Generated Aerodynamically, Part I :General Theory, Proc. Roy.Soc.,A211,564587, 1952. [68] Lighthill M.J., On Sound Generated Aerodynamically, Part II: Turbulence as a source of sound, Proc.Roy. Soc., A222, 1-32, 1954. [69] Curle, N., The influence of solid boundaries upon aerodynamic sound, Proc.Roy. Soc.Lond., A23, 505514, 1955. [70] Ffowcs Williams, J.E. and Hawkings D.L., Sound Generation by Turbulence and Surfaces in Arbitrary Motion, Phil.Trans.Roy.Soc., A, Vol. 264, No. 1151, pp.321-344, 1969. [71] R. Siegert, Numerical simulation of aero-acoustic sound generated by a simplified side mirror model. SIA, 1999. [72] Brooks A.N. and Hughes T.J.R., Streamline Upwind /Petrov-Galerkin Formulations for Convection Dominated Flows with particular Emphasis on the Incompressible Navier-Stokes Equations, Computer Methods in Applied Mechanics and Engineering, Vol. 32, 1982. [73] M. Sakurai, M. Endo, F. Pri, "Development of the exhaust systems radiation noise technology" Proceedings of Insert ASMEFEDSM03, ASME 2003 Fluids Engineering Division Summer Meeting, Honolulu, Hawaii, USA, July 6-11, 2003. [74] Bayliss and E.Turkell, Outflow Boundary Conditions for Fluid Dynamics, SIAM J.Sci.Stat. Comput.Vol.3 (2), 1982.
20-jan-2009
179
REFERENCES
[75] S. Vergne, J.M. Auger, N. G'Styr, F. Pri, "Simulation of cavity Aero-Elastic Noise induced by an external turbulent flow perturbed by a small ruler", Workshop on LES for Acoustics October 2002, DLR Goettingen, Germany 2002. [76] D. Obrist, D. Nicolopoulos, A. Jacques, "Aero Acoustic Simulation of a side mirror compared with experimental results", The 2002 International Congress and Exposition on Noise Control Engineering, Dearborn, 2002. [77] Y. Fukui, F. Pri, "Methodological developments for vibrating pipes noise prediction", The 2002 International Congress and Exposition on Noise Control Engineering, Dearborn, 2002. [78] J.BONET.T-S.L.LOK, Variational and Momentum Preservation Aspects of Smooth Particle Hydrodynamic Formulations, Computer Methods in Applied Mechanics and Engineering, vol.180, pp.97115 (1999). [79] L.B. Lucy, A numerical approach to the testing of the fission hypothesis, Astro. J., Vol. 82, 1013, 1977. [80] R. A. Gingold, J.J. Monaghan, SPH: Theory and application to non-spherical stars, Mon. Not. R. Astron. Soc., Vol. 181, 375, 1977. [81] P.W. Randles, L.D.Libersky, Smoothed Particle Hydrodynamics: Some recent improvements and applications, Computer Methods Appl. Mech. Engrg. vol.139, pp.375-408, 1996. [82] D.S. Balsara, Von Neumann Stability Analysis of Smoothed Particle Hydrodynamics Suggestions for Optimal Algorithms, Journal of Computational Physics, vol.121, pp.357-372, 1995. [83] J.J. Monaghan, Smoothed Particle Hydrodynamics, Annu.Rev.Astron.Astro-phys, vol.30, pp. 543-574, 1992. [84] J.W. Swegle, D.L. Hicks, S.W.Attaway, Smoothed Particle Hydrodynamics: Stability Analysis, Journal of Computational Physics, vol.116, pp.123-134, 1995. [85] J. Bonet, T-S.L. Lok, Variational and Momentum Preservation Aspects of Smooth Particle Hydrodynamic Formulations, Computer Methods in Applied Mechanics and Engineering, vol.180, pp.97-115, 1999. [86] J. Bonet, S. Kulasegram, Correction and Stabilization of Smooth Particle Hydrodynamics Methods with Applications in Metal Forming Simulations, Int. J. Numer. Meth. Engng., vol.47, pp.1189-1214, 2000. [87] M.A. Crisfield, Non linear finite element analysis of solids and structures, J. Wiley, Vol. 2, 1997. [88] D. Nicolopoulos, F. Pri, A. Jacques, Direct numerical simulation of Aero-Acoustic phenomena, MCUBE, Internal Report, February 2004. [89] M.F. Ausserer, S.W. Lee, An eighteen node solid element for thin shell analysis, Int. J. Numer. Meth. Engrg., Vol. 26, pp. 1345 .1364,1988. [90] H.C. Park, C. Cho and S.W. Lee, An efficient assumed strain element model with six dof per node for geometrically nonlinear shells, Int. J. Numer. Meth. Engrg.,Vol. 38, pp 4101-4122,1995. [91] K.Y.Sze and A.Ghali, A hexahedral element for plates, shells and beam by selective scaling, Int. J. Numer. Meth. Engrg., Vol. 36, pp 1519-1540,1993. [92] P. Betch and E. Stein, An assumed strain approach avoiding artificial thickness straining for a nonlinear 4node shell element, Computer Methods in Applied Mechanics and Engineering, Vol.11, pp. 899-909,1997. [93] M. Bischoff and E. Ramm, Shear deformable shell elements for large strains and rotations, Int. J. Numer. Meth.Engrg., Vol. 40, pp. 445-452,1997. [94] R. Hauptmann and K. Schweizerhof, A systematic development of solid-shell element formulations for linear and nonlinear analysis employing only displacement degrees of freedom, Int. J. Numer. Meth. Engrg., Vol.42, pp. 49-69,1988. [95] J.C. Simo and M.S. Rifai, A class of mixed assumed strain methods and the method of incompatible modes, Int. J. Numer. Meth. Engrg., Vol. 9, pp. 1595-1638,1990. [96] J.L. Meriam, Dynamics, John Wiley & sons, Second edition, 1975. [97] D.C. Sorensen, Implicit application of polynomial filters in a k-step Arnoldi method, SIAM J. Matrix Anal. Appl., Vol. 13, pp. 357-385, 1992.
20-jan-2009
180
REFERENCES
[98] A.S. Argon, J.Im, R. Safoglu, Cavity formation from inclusions in ductile fracture, Metallurgical Transactions, Vol. 6A, pp. 825-837, 1975. [99] F.R. Tuler, B.M. Butcher, A criteria for time dependence of dynamic fracture, International Journal of Fracture Mechanics, Vol. 4, N4, 1968. [100] F. Barlat, J. Lian, Plastic behavior and stretchability of sheet metals, Part I: A yield function for orthoropic sheets under plane stress conditions, International Journal of Plasticity, Volume 5, pp. 51-66, 1989. [101] J.C. Simo, On a fully three-dimensional finite strain viscoelastic damage model: Formulation and Computational Aspects, Computer Methods in Applied Mechanics and Engineering, Vol. 60, pp. 153-173, 1987. [102] J.C. Simo, T.J.R Hughes, Computational inelasticity, Mechanics and Materials, Springer, 1998. [103] M.A. Ajiz, A. Jennings, A robust incomplete Choleski-conjugate gradient algorithm, Int. J. Method Eng., Vol. 20, pp. 949-966, 1984. [104] R. Flectcher, Practical methods of optimization, 2nd edition, Wiley, 1987. [105] Batoz J.L., Guo Y.Q., Shakourzadeh H., "Nonlinear Analysis of thin shells with elasto-plastic element DKT12", Revue Europenne des Elments Finis, Vol. 7, N 1-2-3, pp. 223-239. 1998. [106] F. Sabourin, M. Brunet, Analysis of plates and shells with a simplified three-node triangle element, Thin-walled Structures, Vol. 21, pp. 209-223, Elsevier, 1995. [107] V. Faucher, A. Combescure, A time and space mortar method for coupling linear modal subdomains and nonlinear subdomains in explicit structural dynamics, Computer Methods in Applied Mechanics and Engineering, Vol. 192, pp. 509-533, 2002. [108] A. Cardona, M. Gradin, A superelement formulation for mechanism analysis, Computer Methods in Applied Mechanics and Engineering, Vol. 100, pp. 1-29, 1992. [109] V. Faucher, A. Combescure, Local modal reduction in explicit dynamics with domain decomposition. Part 1: extension to subdomains undergoing finite rigid rotations, International Journal for Numerical Methods in Engineering, Vol. 60, pp. 2531-2560, 2004. [110] E. Cosserat, F. Cosserat, Deformable body theory, Hermann, Paris, 1909. [111] S. Forest, K. Sab, Cosserat overall modeling of heterogeneous materials, Mechanics Research Communications, Vol. 25, No. 4, pp. 449-454, 1998. [112] S. Forest, G. Cailletaud and R. Sievert, A Cosserat Theory for Elastoviscoplastic Single Crystals at Finite Deformation , Archives of Mechanics, vol. 49, pp. 705-736, 1997. [113] J. Besson, F. Bultel, S. Forest, Plasticity Cosserat media. Application to particular composites particulaires, 4th Symposium Calculation of Structures, CSMA/Teksea, Toulouse, pp. 759-764, 1999. [114] S. Forest, Cosserat Media , ed. by K.H.J. Buschow, R.W. Cahn, M.C. Flemings, B. Ilschner, E.J. Kramer and S. Mahajan, Encyclopedia of Materials, Science and Technology, Elsevier, 2001. [115] T. Wierzbicki, From crashworthiness to fracture; Ten years of research at MIT, International RADIOSS Users Conference, Nice, June 2006. [116] E. Riks, An incremental approach to the solution of snapping and buckling problems, Int. J. Solids & Structs, Vol. 15, 529-551, 1979. [117] G.A. Wempner, Discrete approximations related to nonlinear theories of solids, Int. J. Solids & Structs, Vol. 7, 1581-1599, 1971.
20-jan-2009
181
Appendices
20-jan-2009
182
APPENDICES
14.0 APPENDICES
Appendix A Conversion Tables & Constants
Length
US & UK >> Metric system 1 Inch (in) - US 1 Inch (in) - UK 1 Foot (ft) = (12.in) - US 1 Foot (ft) = (12.in) - UK 1 Yard (yd) = (3.ft) - US 1 Yard (yd) = (3.ft) - UK 1 Mile (mi) = (1760.yd) - US 1 Mile (mi) = (1760.yd) - UK 1 Nautical mile (UK) 25.40005 mm 25.39996 mm 0.3048006 m 0.3047995 m 0.9144018 m 0.9143984 m 1.609347 km 1.609341 km 1.853181 km Metric system >> US & UK 1 millimeter (mm) 1 millimeter (mm) 1 meter (m) 1 meter (m) 1 meter (m) 1 meter (m) 1 kilometer (km) 1 kilometer (km) 1 kilometer (km) 0.03937 in (US) 0.03937 in (UK) 3.28083 ft (US) 3.28083 ft (UK) 1.093611 yd (US) 1.093611 yd (UK) 0.6213699 mi (US) 0.6213724 mi (UK) 0.5396127 n.mi (UK)
Surface
US & UK >> Metric system 1 Acre - US 1 Acre - UK 1 Square inch (sq in) - US 1 Square inch (sq in) - UK 1 Square foot (sq ft) = 144 sq in - US 1 Square foot (sq ft) = 144 sq in - UK 1 Square yard (sq yd) = 9 sq.ft - US 1 Square yard (sq yd) = 9 sq.ft - UK 1 Square mile (sq mi) = 640 acres - US 1 Square mile (sq mi) = 640 acres - UK 0.4046873 ha 0.4046842 ha 6.451626 cm 6.451578 cm
2 2 2 2
Metric system >> US & UK 1 hectare (ha) 1 hectare (ha) 1 Square centimeter (cm 1 Square centimeter (cm 1 Square meter (m 1 Square meter (m 1 Square meter (m 1 Square meter (m
2) 2) 2) 2) 2) 2) 2) 2)
2.471044 acre (US) 2.4711 acre (UK) 0.1549997 sq. in (US) 0.1550 sq.in (UK) 10.76387 sq.ft (US) 10.7639 sq.ft (UK) 1.195985 sq.yd (US) 1.1960 sq.yd (UK) 0.3861006 sq.mi (US) 0.3861 sq.mi (UK)
2 2
Volume
US/UK >> Metric system Volume 1 Cubic inch (cu in) - US 1 Cubic inch (cu in) - UK 1 Cubic foot (cu ft) - US 1 Cubic foot (cu ft - (UK) 1 Cubic yard (cu yd) - US 1 Cubic yard (cu yd) - UK Measure of capacity 16.3871 cm3 16.38698 cm3 28.31702 dm3 28.31670 dm3 0.7645594 m 0.7645509 m
3 3
3) 3)
0.06102509 cu in (US) 0.0610241 cu in (UK) 0.03531544 cu ft (US) 0.0353148 cu ft (UK) 1.307943 cu yd (US) 1.307957 cu yd (UK)
1 Cubic decimeter (dm ) 1 Cubic decimeter (dm ) 1 Cubic meter (m 1 Cubic meter (m
3) 3) 3
20-jan-2009
183
APPENDICES
33.814 fl oz 35.195 fl oz 0.0283782 bu (US) 0.02749692 bu (UK) 0.2641779 gal (US) 0.2199754 gal (UK) 2.113423 liq.pt (US) 1.759803 pt (UK)
Mass
Notes : Attention should be paid to not confuse mass and weight. The mass (kg) is an intrinsic characteristic of the body and is measured in kilogram. The weight is a force which depends on terrestrial attraction and it is the equivalent of the mass of a body by the acceleration of gravity (9.80665 at the sea level) and is measured in Newton [ N ]. The specific mass is the quotient of the mass of a body by its volume. The unit of mass of S.I. = kilogram. Metric system >> US/UK 64.79892 mg 64.79892 mg 28.34953 g 28.34953 g 0.4535924 kg 0.4535924 kg 45.35924 kg 45.35924 kg 1.016047 t 1.016047 t 1 milligram (mg) 1 milligram (mg) 1 gram (g) 1 gram (g) 1 kilogram (kg) 1 kilogram (kg) 1kilogram (kg) 1 kilogram (kg) 1 ton 1 ton 0.01543236 gr (US) 0.01543236 gr (UK) 0.03527396 oz av. (US) 0.03527396 oz av. (UK) 2.204622 lb av. (US) 2.204622 lb av. (UK) 0.02204622 sh.cwt (US) 0.02204622 ctl (UK) 0.9842064 l.tn (US) 0.9842064 tn (UK)
US/UK >> Metric system 1Grain (gr) - US 1Grain (gr) - UK 1Ounce (oz) - US 1Ounce (oz) - UK 1Pound (Ib) = 16 oz - US 1Pound (Ib) = 16 oz - UK 1Short hundredweight(sh cwt)= 100 Ib - US 1Cental (UK) 1Long ton (l tn) = 2240 Ib - US 1Ton (UK)
20-jan-2009
184
APPENDICES
Temperature
Conversion Type Degrees Celsius (C) to degrees Kelvin (K) Degrees Celsius (C) to degrees Fahrenheit (F) Degrees Celsius (C) to degrees Rankine (R) Degrees Fahrenheit (F) to degrees Kelvin (K) Degrees Fahrenheit (F) to degrees Rankine (R) Degrees Kelvin (K) to degrees Rankine (R) Degrees Rankine (R) to degrees Celsius (C) Degrees Rankine (R) to degrees Kelvin (K) Conversion Formula = (C) + 273.15 = (K) = [(C) * 1.8] + 32 = (F) = [(C) * 1.8] + 491.67 = (R) = [(F) * 0.555556] + 255.37 = (K) = (F) + 459.67 = (R) = (K) * 1.8 = (R) = [(R) - 491.67] * 0.555556 = (C) = (R) * 0.555556 = (K)
Pressure
Symbol 1 at 1 atm 1 bar 1 cmHg (0 C) 1 cmH2O 1 dyn/cm2 1 ft H2O 1 hPa 1 inHg (0 C) 1 inHg (15.56 C) 1 inH2O (15.56 C) 1 inH2O (4 C) 1 kgf/cm2 1 kgf/dm2 1 kgf/m2 1 kgf/mm2 1 kPa 1 kip/in2, ksi, KSI 1 MPa 1 mH2O, mCE (15.56 C) 1 mH2O, mCE (4 C) 1 bar 1 Hg (0 C) 1 mbar 1 mmHg, torr, Torr (0 C) 1 mmH2O, mmCE (15.56 C) 1 mmH2O, mmCE (4 C) 1 N/m2 1 Pa, N/m2 1 lbf/ft2 1 psi, PSI, lbf/in2 1 atm 1 tonf/cm2 (metric) 1 tonf/m2 (metric) 1 torr SI equivalent = 98066.5 Pa = 101325 Pa = 100000 Pa = 1333.22 Pa = 98.0638 Pa = 0.1 Pa = 2988.98 Pa = 100 Pa = 3386.38 Pa = 3377.18 Pa = 248.845 Pa = 249.082 Pa = 98066.5 Pa = 980.665 Pa = 9.80665 Pa = 9806650 Pa = 1000 Pa = 6894760 Pa = 1000000 Pa = 9797.06 Pa = 9806.38 Pa = 0.1 Pa = 0.133322 Pa = 100 Pa = 133.322 Pa = 9.79706 Pa = 9.80638 Pa = 1 Pa = 1 Pa = 47.8803 Pa = 6894.76 Pa = 101325 Pa = 98066500 Pa = 9806.65 Pa = 133.322 Pa Designation of unit measure - atmosphere (metric) - atmosphere (standard) - bar - centimeter of mercury (0 C) - centimeter of water (4 C) - dyne per square centimeter - foot of water (4 C) - hectopascal - inch of mercury (0 C) - inch of mercury (15.56 C) - inch of water (15.56 C) - inch of water (4 C) - kilogram force per square centimeter - kilogram force per square decimeter - kilogram force per square meter - kilogram force per square millimeter - kilopascal - kilopound force per square inch - megapascal - meter of water (15.56 C) - meter of water (4 C) - microbar (barye, barrie) - micron of mercury (millitorr) - millibar - millimeter of mercury (0 C) - millimeter of water (15.56 C) - millimeter of water (4 C) - newton per square meter - pascal - pound force per square foot - pound force per square inch - standard atmosphere - ton force (metric) per square centimeter - ton force (metric) per square meter - torr
20-jan-2009
185
APPENDICES
Designation of unit measure - acre foot per month - barrel (US, liq.) per day - barrel (US, petrol) per day - cubic centimeter per second - cubic foot per minute - cubic foot per second - cubic inch per minute - cubic inch per second - cubic meter per hour - cubic meter per minute - cubic meter per second - cubic yard per minute - cumec (musec) - gallon (UK) per day - gallon (UK) per hour - gallon (UK) per minute - gallon (UK) per second - gallon (US, liq.) per day - gallon (US, liq.) per hour - gallon (US, liq.) per minute - gallon (US, liq.) per second - liter per hour - liter per minute - liter per second
Material Constants
Density (kg/m^3) Aluminum 2024-T3 Aluminum 6061-T6 Aluminum 7079-T6 Concrete Copper - pure Gold - pure Iron Lead - pure Magnesium AZ31B-H24 Magnesium HK31A-H24 Nickel - pure Platinum Silver - pure Steel AISI 304 Steel AISI C1020 Tantalum Titanium B 120VCA Tungsten 2770 2700 2740 2242 8900 19320 7830 11340 1770 1790 8900 21450 10500 8030 7850 16600 4850 19300 1.379E+10 4.483E+10 4.414E+10 2.207E+11 1.469E+11 7.241E+10 1.931E+11 2.034E+11 1.862E+11 1.021E+11 3.448E+11 0.3900 0.3700 0.2900 0.2900 0.3500 0.3000 0.2800 0.4200 0.3500 0.3500 52.74 26.10 25.20 12.96 9.0000 19.80 17.82 11.34 6.4800 9.3600 4.5000 1.172E+11 7.448E+10 0.4200 16.56 4.39 Elastic Modulus (Pa) 7.310E+10 7.310E+10 7.172E+10 Poisson's ratio 0.3300 0.3300 0.3300 Thermal expansion coefficient (microns/m C) 22.68 24.30 24.66 Heat capacity J/kg(C)
963.00 963.00 963.00 1000 385.00 130.00 440.00 130.00 1047 544.00 461.00 130.00 235.00 503.00 419.00 126.00 544.00 138.00
20-jan-2009
186
APPENDICES
Heat Capacity a [J/(mol K)] 26,092000 29,659000 24,997350 25,567590 20,786000 20,786030 20,786030 33,066178 30,092000 19,995630 26,884120 -36,220000 27,679880 b [J/(mol K )] 8,2188E-03 6,1373E-03 5,5187E-02 6,0961E-03 2,8259E-10 4,8506E-13 4,8506E-13 -1,1363E-02 6,8325E-03 4,9771E-02 1,8678E-02 4,8475E-01 5,1149E-02
2
MW [kg/mol] 0,02801 0,032 0,04401 0,02801 0,03995 0,02018 0,004 0,00202 0,01802 0,01703 0,03408 0,07811 0,04401
c [J/(mol K )] -1,9761E-06 -1,1865E-06 -3,3691E-05 4,0547E-06 -1,4642E-13 -1,5829E-16 -1,5929E-16 1,1433E-05 6,7934E-06 -1,5376E-05 3,4342E-06 -3,1570E-04 -3,0645E-05
d [J/(mol K )] 1,5927E-10 9,5780E-11 7,9484E-09 -2,6713E-09 1,0921E-17 1,5251E-20 1,5251E-20 -2,7729E-09 -2,5345E-09 1,9212E-09 -3,3787E-09 7,7620E-08 6,8479E-09
e [J K/mol] 4,4434E+04 2,1966E+05 1,3664E+05 1,3102E+05 -3,6614E02 3,1963E-05 3,1963E-05 1,5856E+05 8,2139E+04 1,8917E+05 1,3588E+05 0,0000E+00 1,5791E+05
Temp.range [K] 298-6000 298-6000 298-1200 298-1300 298-6000 298-6000 298-6000 298-1000 500-1700 298-1400 298-1400 298-1500 298-1400
N2O
Velocity m/s m/s m/s m/s Km/s Km/s Km/s mm/s mm/s mm/s m/s m/s m/s dam/s dam/s dam/s 10 m/s
4
Kg/m3 m Kg/l
Energy Kmg2/s2 J mJ KJ MJ KJ GJ
G 9.81 9.81 9.81 9.81 9.81e-6 9.81e-6 9.81e-6 9.81e+3 9.81e+3 9.81e+3 9.81e-3 9.81e-3 9.81e-3 9.81e-4 9.81e-4 9.81e-4 9.81e-10
Kg/l
Kg/l m Kg/l
Kg/l
Kg/l M Kg/l K Kg/l G Kg/l M Kg/l K Kg/l G Kg/l Kg/l K Kg/l M Kg/l Kg/l
nJ mJ J mJ KJ dJ hJ 10 J 10 J
5 5
N KN N MN daN
20-jan-2009
187
APPENDICES
1 ~ p = K1 + K 2 2 + K 3 3 1 + E (1 + ) 2
with
EQ. B1
1 1 = 1 V 0
initial volume and p the pressure. The equation can be rewritten in polynomial expression in and linear in energy. Let Us be the shock velocity, C the sound speed and S a material constant. The particle velocity u and shock velocity Us are related by the following equation:
U S = C + Su
EQ. B2
The values of C and S for some materials are given in Table B1. The hydrodynamic shock pressure determined by:
pH =
with x = 1
0C 2 x
(1 Sx) 2
EQ. B3
= 1V 0
EH =
1 1 1 1 pH = pH x 2 0 2 0
EQ. B4
The change in pressure with respect to the initial state can be expressed as:
pH p0 = (EH E0 ) p p0 = (E E0 )
This leads to:
EQ. B5 EQ. B6
p pH = (E EH )
or p = pH + E
EQ. B7
x x 1 pH = pH 1 2 + E 0 2 0
and
EQ. B8
where
x = 1 = 0 0
E=
written as:
20-jan-2009
188
APPENDICES
~ p = pH 1 + (1 + ) E 2
EQ. B9
The hydrodynamic shock pressure pH given by EQ. B3 can be expressed in linear polynomial terms:
p H = K1 + K 2 2 + K 3 3
with:
EQ. B10
K1 = 0 C 2 K 2 = 0C 2 (2 S 1) K 3 = 0C 2 (S 1)(3S 1)
which allows to obtain the Me Grneisen equation given by EQ. B1. The equation of state used in RADIOSS is written in a linear polynomial expression:
~ p = c1 + c2 2 + c3 3 + (c4 + c5 )E
EQ. B11
The parameters ci can be identified by using EQ. B10 and Me Grneisen equation. Under the assumption =cst, the following setting is obtained:
c1 = K1
c2 = K 2 c3 = K 3
K1 K2
EQ. B12
c 4 = c5 = 1
If
= 0 ax = 0 a
1+
EQ. B13
where a is a material characteristic parameter given for some materials in Table B1. The second part of EQ. B9 is expressed as:
(1 + ) E = ( 0 + ( 0 + a ) )E
which leads to:
EQ. B14
c1 = K1
c2 = K 2
K1 = K1 2 S 1 0 2 2
EQ. B15
20-jan-2009
189
APPENDICES
c3 = K 3 c4 = 0
0
2
K 2 + aK1
EQ. B16
c5 = 0 a
Material
V0
(cm /g)
3
0
1.16 1.63 2.18 1.3 2.0 2.0 1.97 2.84 1.0 1.3 1.7 1.1 1.5 1.5 1.5 2.3
Mg(1): Alloy AZ31B, 96%Mg, 3% Al, 1% Zn. Al(2): Alloy 2024, 93,5% Al, 4.5% Cu, 1.5% Mg. Stainless(3):Alloy 304, 72% Fe, 19% Cr, 9% Ni.
20-jan-2009
190
APPENDICES
{ } = [D ]{ }
1 = D111 + D12 2 + D13 3 1 = ( + 2 )1 + ( 2 + 3 ) 1 = (1 + 2 + 3 ) + 21 1 = K kk + 2e1
with
{ } = [D ]{ }
{ } = [C ]{ }
1 E ; [D ] = (1 + )(1 2 ) 1 E E 1 E [C ] = ; Symm.
0 0 0 1 2 2
Symm.
E E 1 E
0 0 0 2(1 + ) E
0 0 0 0 2(1 + ) E
0 0 1 2 0 2 1 2 2 0 0 0 0 0 2(1 + ) E
0 0 0
0 0 0
{ } = [H ]{ }
1 = H111 + H12 2
20-jan-2009
191
APPENDICES
{ } = [H ]{ }
[H ] = E
1 2
0 1 0 1 0 2 0 0 0 0
0 0 0 1 2 0
0 1 2
0 0 0
1 0 0 0
0 0 2(1 + ) 0 0
0 0 0 2(1 + ) 0
0 2(1 + )
0 0 0
{ } = [H ]{ }
1 E [H ] = (1 + )(1 2 ) 0
1 0
0 0 1 2 2
{ } = [C ]{ }
1 1 + [C ] = 1 E 0 0
0 0 2
E,G E
E,B E
G,
G,B
B,
2(1+v)G
9 BG 3B + G
3(1-2v)B
(3 + 2 )
+
G=
B=K
E 2(1 + v )
E 3(1 2v )
3EB 9B E
B
3(1 2v )B 2(1 + v )
B
3 + 2 3
EG 9G 3E E 2G 2G
2(1 + v )G 3(1 2v )
Ev
3B E 6B
2Gv 1 2v
3B 2G 6 B + 2G
3B 2G 3
3Bv
2( + )
(1 + v )(1 2v )
20-jan-2009
(E 2G )G
3G E
(3B E )3B
9B E
(1 + v )
192
APPENDICES
12 21
=
Three Young modulus in orthotropic directions 1, 2 and 3 : E1, E2, E3 Three shear modulus in plans 12, 13 and 23 : G12, G13, G23 Three Poisson ratios satisfying the relations:
1 12 21 13 31 23 32 12 23 31 21 13 32 > 0
1 E1 12 E1 13 E1 [C ] = 0 0 0
E1 E2 E1 1 12 21 > 0 ;
13 31
=
= E3 E2 E3 1 13 31 > 0
23 32
1 23 32 > 0
{ } = [C ]{ }
21 E2 1 E2 23 E2 0 0 0
31 E3 32 E 1 E3 0 0 0
0 0 0 1 G12 0 0
0 0 0 0 1 G13 0
0 0 0 0 1 G23
20-jan-2009
193
APPENDICES
{ } = [C ]{ }
12 E1 1 E2 E2 0 0 0
12 E1 E2 1 E2 0 0 0
0 0 0 1 G12 0 0
0 0 0 0 1 G12 0
0 0 0 0 2(1 + ) E2
EA L [k ] =
0 12 EI 3 L3 (1 + 2 )
0 0 12 EI 2 L3 (1 + 2 )
0 0 0 GJ L
0 0 L K 33 2 0
0 L K 22 2 0 0 0
K11 0 0 0 0 0 K11
0 K 22 0 0 0 K 26 0 K 22
0 0 K 33 0 K 35 0 0 0 K 33
0 0 0 K 44 0 0 0 0 0 K 44
0 0 K 35 0 2 3 K 55 4 + 3 0 0 0 K 35 0 K 55
(4 + 3 )EI 2 L(1 + 3 )
(4 + 2 )EI 3 L(1 + 2 )
Symm.
K 26 0 0 0 2 2 K 66 4 + 2 0 K 26 0 0 0 K 66 0
with :
2 = 3 =
I=
bh 3 12
20-jan-2009
194