This action might not be possible to undo. Are you sure you want to continue?

by homogenization

H. Tollenaere & D. Caillerie*

Laboratoire Sols Solides Structures, BP 53X, 38041 Grenoble, France

(Received 1 December 1995; accepted 1 November 1997)

We present a method of continuous modeling of quasi-repetitive lattice structures

called discrete homogenization and derived from the periodic continuous medium

homogenization. The structure studied here is a two-dimensional quasi-periodic lattice

truss with pin-jointed nodes in the case of static deformation. A one-dimensional

simple illustrative example is given. ᭧1998 Elsevier Science Limited and Civil-

Comp Limited. All rights reserved.

1 INTRODUCTION

Taking advantage of the smallness of each elememtary

beam compared with the size of the whole structure, differ-

ent attempts have been made to model repetitive trusses

composed of a large number of beams by continuous equa-

tions.

1–9

So, a crane jib may be modeled as a beam, a roof or

a space antenna as a plate or a shell. The aim of this model-

ing is to simplify the analysis or the optimization of the

structure.

The method developed in this work is an adaptation of the

two-scale asymptotic expansions used in homogenization of

periodic media;

10,11

it differs partly from the other methods

proposed in the literature and does not suffer some of the

defects of them. Though related to the studies of refs

12,13

,

our model is different in the sense that we consider the

beams as one-dimensional structures. The method had

been developed for beam-like structures

14–17

and is now

settled for beam-like, plate-like or three-dimensional repe-

titive trusses of very general design, even for structures with

mechanism and for nearly repetitive structures (here it is

developed for two-dimensional nearly repetitive structures).

The same method has been applied to structures in large

displacement framework,

18

thanks to its characteristic fea-

tures (asymptotic expansion of displacements and tensions),

the method could be extended to such non-linear problems

as plasticity.

The structures under consideration are very large pin-

jointed trusses, so the unknowns are the displacements of

nodes and the tensions in the beams; the loads are applied on

the nodes. The structures are assumed to be large, which

means that they consist of a large number of elementary

beams, the lengths of which are small compared with the

size of the structures; the ratio of beam lengths over struc-

ture size is the small parameter e of the problem. The

method of discrete homogenization consists in assuming

asymptotic series expansions in powers of e for the node

displacements and the tensions in beams. The balance equa-

tions of nodes and force displacement relations of beams are

then developed by carrying these series expansions in them

and by using Taylor’s expansion of ﬁnite differences. Lasty,

an identiﬁcation of the terms of same power of e yields

partial differential equations satisﬁed by the ﬁrst terms of

the series expansions of the displacements and tensions.

Following the type of the structure under consideration

these partial differential equations are those of three- or

two-dimensional elasticity (the case of this study) or those

of a beam or a plate, so they deﬁne the continuous model of

the truss.

The continuous model is valid for the study of either

static deformations or vibrations; only the case of static

deformations is presented here. The calculations have

been completed for a quite general truss and the results

give general expressions of the elastic moduli of the equiva-

lent continuous medium. The reckoning of these moduli

needs only the solving the self-balance equations of one

elementary cell, which can be performed analytically for

simple cells and can be easily programmed on a computer

for more complicated cells.

A two-dimensional example would be too long to be

developed here and only a one-dimensional one is pre-

sented. It is however illustrative, indeed it shows that the

Advances in Engineering Software Vol. 29, No. 7–9, pp. 699–705, 1998

᭧ 1998 Elsevier Science Ltd and Civil-Comp Ltd

Printed in Great Britain. All rights reserved

0965-9978/98/$19.00 +0.00 PII: S 0 9 6 5 - 9 9 7 8 ( 9 8 ) 0 0 0 3 4 - 9

ADES 376

699

*Author to whom all correspondence should be addressed.

method enables the expected well-known result to be found,

and it is sufﬁcient to point out the differences of results

between this method and one used in litterature.

The Einstein convention of repeated index is used

throughout this paper.

2 NUMBERING OF NODES AND BEAMS

The truss under consideration is made of beams only linked

by their ends, so the geometry of the truss is completely

deﬁned by the positions of the nodes and the connectivity

between them, that is which of them are linked by a beam.

The lattice truss is assumed to be quasi-periodic, which

means that it is the repetition of nearly identical cells. We

use a system of numbering of nodes and beams that reﬂects

the regularity of the structure and leads to the curvilinear

coordinates of the equivalent continuum.

The truss is two-dimensional, so the cells composing it

are numbered by a system of two integers (v

1

, v

2

). The nodes

are then numbered by a triplet ˜ n ¼(n, v

1

, v

2

), which means

that the node ˜ n is the node n in the cell (v

1

, v

2

). The same

system is used for beams which are numbered

˜

b ¼(b, v

1

, v

2

).

Each beam links two nodes and is oriented, which means

that one the two nodes is the origin of the oriented beam,

denoted O(

˜

b). The other node is the end and is denoted E(

˜

b).

O(

˜

b) and E(

˜

b) are then mappings from the set of beams

˜

B

into the set of nodes

˜

N and O

¹1

( ˜ n) (respectively E

¹1

( ˜ n))

denotes the set of beams the origin (respectively the end) of

which is ˜ n.

In ˜ n ¼(n, v

1

, v

2

) and

˜

b ¼(b, v

1

, v

2

), the indices n and b

number the nodes and beams of a reference cell the sets of

which are denoted N

R

and B

R

.

The beams link nodes of the same cell or of neighboring

cells, as the orientation of a beam is arbitrary, it is chosen so

that the origin O(

˜

b) of the beam

˜

b ¼(b, v

1

, v

2

) is in the cell

(v

1

, v

2

), that is, a node numbered (n, v

1

, v

2

). The end E(

˜

b) of

the beam

˜

b ¼(b, v

1

, v

2

) is in the cell (v

1

, v

2

) or in a neigh-

boring one, anyway in a cell numbered (v

1

þ d

1

, v

2

þ d

2

)

with d

i

e{-1, 0, 1}, and E(

˜

b) is a node numbered (m, v

1

þ d

1

,

v

2

þ d

2

).

As the truss is repetitive, n, m, d

1

, d

2

depend only on b,

they are denoted respectively O

R

(b), E

R

(b), d

1b

, d

2b

.

3 GEOMETRY OF A QUASI-PERIODIC TRUSS

The lattices under consideration are structures with a large

number of cells, which is equivalent to saying that

e ¼1=(N

c

)

1=2

is a small parameter (N

c

being the number of

cells in the structure).

We use an asymptotic expansion method which assumes

that e tends to zero. Actually, that means that we consider a

sequence of structures with more and more cells which are

smaller and smaller. The sequence of structures has to be

completely deﬁned, that is to say that for each value of e it is

necessary to know where the nodes of the structure are

located in the space. Moreover, we intend to consider

trusses which are not exactly repetitive but quasi-periodic

which means that their geometry can vary slightly.

Then, we assume that the positions of the nodes are given

by an expansion of the form:

᭙˜ n ¼(n, v

1

, v

2

), r( ˜ n) ¼r

0

(l

1e

, l

2e

) þer

n1

(l

1e

, l

2e

)

þe

2

r

n2

(l

1e

, l

2e

) þ

…

ð1Þ

where l

ie

¼ ev

i

and where r

0

(l

1e

, l

2e

), r

n1

(l

1e

, l

2e

),

…

are

given functions of (l

1

,l

2

).

At ﬁrst order in e, all the nodes of a cell have approxi-

mately the same location given by r

0

(l

1

, l

2

), their positions

with respect to the point r

0

(l

1

, l

2

) are of the order of e and

given by er

n1

(l

1e

, l

2e

) þ

…

.

In this way, an exactly periodic lattice of periods eL

1

and

eL

2

is deﬁned by:

r( ˜ n) ¼l

1e

L

1

þl

2e

L

2

þer

n

which ﬁts eqn (1) with r

0

¼l

1

L

1

þl

2

L

2

and with r

n1

¼r

n

independent of (l

1

, l

2

).

Fig. 1 displays an example of a quasi-periodic structure

which is described by:

r((1, v

1

, v

2

)) ¼R(l

1e

) cos l

2e

p

2

_ _

i þsin l

2e

p

2

_ _

j

_ _

r((2, v

1

, v

2

)) ¼R(l

1e

)

_

cos l

2e

p

2

þe

p

5

_ _

i

þsin l

2e

p

2

þe

p

5

_ _

j

_

with R(l

1

) ¼R

i

þl

1

(R

e

¹R

i

), l

1e

¼ ev

1

and l

2e

¼ ev

2

.

Fig. 1. Example of a quasi-periodic structure in the cell (4,4).

700 H. Tollenaere, D. Caillerie

For this example, the r( ˜ n) can be developed in:

r((1, v

1

, v

2

)) ¼R(l

1e

) cos l

2e

p

2

_ _

i þsin l

2e

p

2

_ _

j

_ _

r((2, v

1

, v

2

)) ¼R(l

1e

) cos l

2e

p

2

_ _

i þsin l

2e

p

2

_ _

j

_ _

þe

p

5

R(l

1e

) ¹sin l

2e

p

2

_ _

i þcos l

2e

p

2

_ _

j

_ _

þe

2

[

…

]

Eqn (1) enables the geometrical characteristics of the

beams to be expanded. The length 1

˜

b

¼kr(E(

˜

b)) ¹

r(O(

˜

b))k and the unitary vector of the beam

e

˜

b

¼[r(E(

˜

b)) ¹r(O(

˜

b))]=1

˜

b

can be expanded in:

1

˜

b

¼eL

b0

(l

1e

, l

2e

) þe

2

L

b1

(l

1e

, l

2e

) þ

…

e

˜

b

¼e

b0

(l

1e

, l

2e

) þee

b1

(l

1e

, l

2e

) þ

…

The expressions of the L

bk

and e

bk

can be found in terms of

the ﬁelds r

nk

(l

1

, l

2

) by using Taylor expansions:

e

b0

¼

1

L

b0

r

E

R

(b)

¹r

O

R

(b)

þ

ץr

0

ץl

i

d

ib

_ _

(2)

It is assumed that, after deformation, the geometry of

the structure remains quasi-periodic, which means that

the positions of the nodes are given by expansions of

the form eqn (1). The difference between the positions

of a node before and after deformation is its displace-

ment u( ˜ n), which is then given by an expansion of the

form:

u( ˜ n) ¼u

0

(l

1e

, l

2e

) þeu

n1

(l

1e

, l

2e

)

þe

2

u

n2

(l

1e

, l

2e

) þ

…

, l

ie

¼ev

i

ð3Þ

Eqn (3) shows that at ﬁrst order in e, the displacements of

the nodes of the structures are approximately given by the

displacement ﬁeld u

0

(l

1

, l

2

). Consequently, the assump-

tion that the structure is quasi-periodic (as described by

eqn (1)) and remains quasi-periodic after deformation

yields that the displacement of nodes becomes, as e

decreases, a function of a domain.

Very often in literature

4–6

a displacement ﬁeld is deﬁned

on the whole space domain occupied by the structure. This

ﬁeld is an extension of the displacements of the nodes,

recalling the interpolations of ﬁnite elements. This kind of

extension would probably be necessary if a convergence

proof was considered; it is not for the use of the asymptotic

expansions developed here.

The following Sections 4–8 are devoted to the derivation

of the equations governing the ﬁeld u

0

(l

1

, l

2

) and giving

the equivalent continuous model.

4 EXPANSION OF BEAM TENSIONS

Consistently with the assumption about geometry, we

assume that the Young modulus E

˜

b

and the section A

˜

b

of

a beam are given by series expansion in power of e. This

implies that the stiffness k

˜

b

¼(E

˜

b

A

˜

b

)=1

˜

b

of the beam can be

expanded in:

k

˜

b

¼

1

e

k

b0

(l

1e

, l

2e

) þk

b1

(l

1e

, l

2e

) þek

b2

(l

1e

, l

2e

) þ

…

(4)

In the tension displacement relations for beams:

N

˜

b

¼k

˜

b

[u(E(

˜

b)) ¹u(O(

˜

b))]·e

˜

b

(5)

the differences such as u

0

(l

1e

þed

1

, l

2e

þed

2

) ¹

u

0

(l

1e

, l

2e

) can be expanded using a Taylor expansion

(ed

1

is small). Consequently the beam tensions N

˜

b

can be

expanded in:

N

˜

b

¼N

bO

(l

1e

, l

2e

) þeN

b1

(l

1e

, l

2e

) þe

2

N

b2

(l

1e

, l

2e

) þ

…

(6)

and by identiﬁcation of terms of same power in e, the

expansion of the tension displacement relation yields:

N

b0

¼k

b0

u

E

R

(b)1

¹u

O

R

(b)1

þ

ץu

0

ץl

i

d

ib

_ _

·e

b0

(7)

N

b1

¼

…

,

…

5 EXPANSION OF BALANCE EQUATIONS—

BALANCE EQUATION OF THE EQUIVALENT

CONTINUUM

To ﬁnd the equivalent continuum of the truss, it remains to

develop the balance equations of the nodes, which are:

᭙˜ n,

˜

beO

¹1

( ˜ n)

N

˜

b

e

˜

b

¹

˜

beE

¹1

( ˜ n)

N

˜

b

e

˜

b

þf

e= ˜ n

¼0 (8)

where f

e= ˜ n

is the external force applied on the node ˜ n. This

force is assumed to be:

f

e= ˜ n

¼ef

e=n

(l

1e

, l

2e

)

This shows that f

e= ˜ n

is assumed to be of the order of e, this

assumption will be proved consistent with the continuum

modeling and the order of magnitude of k

˜

b

, eqn (4).

The expansion of the balance eqn (8) is much easier when

the equations are written with a virtual power formulation

which reads:

᭙v( ˜ n),

˜

be

˜

B

N

˜

b

e

˜

b

·[v(O(

˜

b)) ¹v(E(

˜

b))] þ

˜ ne

˜

N

f

e= ˜ n

·v( ˜ n) ¼0

(9)

Sums such as

˜

be

˜

B

can be split in

v

1

, v

2

beB

R

and, in the

process of expansion, which implies that e is very small, the

sums e

2

v

1

, v

2 are approximated by integrals

_

q

dl

1

dl

2

.

We get then the virtual power formulation of the equili-

brium of the equivalent continuum by taking in eqn (9):

᭙˜ n ¼(n, v

1

, v

2

), v( ˜ n) ¼v

0

(l

1e

, l

2e

), l

ie

¼ev

i

where v

0

(l

1

, l

2

) is a smooth given virtual macroscopic

velocity ﬁeld.

Continuous modeling of lattice structures by homogenization 701

We get then:

᭙v

0

, ¹

q

S

i0

·

ץv

0

ץl

i

dl

1

dl

2

þ

q

f·v

0

dl

1

dl

2

¼0 (10)

with:

S

i0

¼

beB

R

N

b0

d

ib

e

b0

(11)

and

f ¼

neN

R

f

e=n

(12)

The previous process explains why the external forces f

e= ˜ n

are assumed to be of the order of e.

The vectors S

i0

describe the internal efforts of the equi-

valent continuum in the curvilinear coordinates system (l

1

,

l

2

). They are related to the stress tensor j

O

in Section 9.

Eqn (10) is the virtual power formulation of the balance

equation of the continuum equivalent to the structure.

6 SELF-EQUILIBRIUM OF THE REFERENCE

CELL

To establish the constitutive relation of the equivalent con-

tinuum we need to use the self balance equations of the cell

which is obtained from eqn (9) by taking:

᭙˜ n ¼(n, v

1

, v

2

), v( ˜ n) ¼ev(l

1e

, l

2e

)v

n

(l

1e

, l

2e

), l

ie

¼ev

i

Using the same process as in Section 5, we get:

᭙v, v

n

q

beB

R

N

bO

e

b0

·[v

E

R

(b)

¹v

O

R

(b)

]

_ _

v dl

1

dl

2

¼0 (13)

which yields:

᭙v

n

,

beB

R

N

b0

e

b0

·[v

E

R

(b)

¹v

O

R

(b)

] ¼0 (14)

This relation is the virtual formulation of the self-

equilibrium of the cell.

7 CONSTITUTIVE RELATIONS OF THE

EQUIVALENT CONTINUUM

In order to get the description of the continuum equivalent

to the structure, its balance eqn (10) has to be completed by

a constitutive relation linking S

i0

to ץu

0

=ץl

i

. The constitu-

tive relation is in fact given by eqns (7), (14) and (11).

Indeeed, ץu

0

=ץl

i

being considered as data, eqns (7) and

(14) are equations the unknowns of which are N

b0

and

u

n1

. Solving these equations gives N

b0

as a function of

ץu

0

=ץl

i

, then eqn (11) yields the wanted constitutive

relation.

As the eqn (14) is discrete, the procedure described above

can be carried out in two steps which makes the solving of

eqns (7) and (14) easier. First a base J

bk

of solutions of eqn

(14) is determined, then this base is orthogonolized in such a

way to eliminate the u

n1

and to express N

b0

in terms of

ץu

0

=ץl

i

.

Let J

bk

, the index k numbering the elements of the base,

denote a base of solutions of eqn (14). The tensions N

b0

which are also solutions of eqn (14) are then a linear com-

bination of the J

bk

s:

N

b0

¼

k

a

k

J

bk

Carrying this relation in the tension displacement eqn (7)

yields:

k

a

k

J

bk

¼k

b0

u

E

R

(b)

¹u

O

R

(b)

þ

ץu

0

ץl

i

d

ib

_ _

·e

b0

Orthogonalizing the J

bk

s by:

beB

R

J

bk

J

bh

k

b0

¼d

kh

enables the u

n1

to be eliminated and the a

k

to be

calculated:

a

k

¼f

ik

·

ץu

0

ץl

i

where

f

ih

¼

beB

R

J

bh

d

ib

e

b0

Then, we can reckon N

b0

and S

i0

in terms of the ץu

0

=ץl

i

:

N

b0

¼

k

J

bk

f

ik

_ _

·

ץu

0

ץl

i

(15)

S

i0

¼

k

f

ik

f

jk

·

ץu

0

ץl

i

_ _

This last relation is the constitutive relation of the equiva-

lent continuum, it reads:

S

i0

¼

k

f

ik

ᮏf

jk

_ _

@

ץu

0

ץl

i

(16)

where f

ik

ᮏf

jk

is the linear operator called tensorial or

diadic product of the two vectors f

ik

and f

jk

.

A@y denotes the image of the vector y by the operator A.

When A is the diadic product a ᮏb, A@y, is such that:

a ᮏb@y ¼(b·y)a

8 CONTINUOUS MODEL OF THE TRUSS IN

CURVILINEAR COORDINATES

The continuous model of the truss in curvilinear coordinates

is given then by the constitutive eqn (16) and the balance

702 H. Tollenaere, D. Caillerie

eqn (10) which yields:

ץS

i0

ץl

i

þf ¼0 (17)

Some boundary conditions have to be added to this prob-

lem. They can be deduced by expanding the boundary con-

ditions of the truss.

The well-posedness of boundary value problems for the

equivalent continuum comes down to the positive deﬁnite-

ness of the following quadratic form:

1

2

k

f

ik

ᮏf

jk

_ _

@

ץu

0

ץl

j

_ _

·

ץu

0

ץl

i

It can be studied and deduced from general assumptions on

the truss.

Remark: carrying u

0

, solution of eqns (16) and (17) com-

plete with boundary conditions, into eqn (15) gives the ﬁrst

term N

b0

of the expansion, eqn (6) of N

b

, that is an approx-

imation of the tensions in the beams of the truss.

9 TWO-DIMENSIONAL ELASTIC MODEL IN

SPACE COORDINATES

The equivalent continuum constitutive relation, eqn (16), is

linear, which could be expected since the considered truss is

linear. Moreover the continuous model, eqns (16) and (17),

should be a two-dimensional elastic model. This does not

appear obviously in eqns (16) and (17) written in curvilinear

coordinates. The best way to prove that the continuous

model is actually a two-dimensional elastic model is to

write the balance and constitutive equations in usual space

coordinates.

The cells of the lattice truss are located by the two inte-

gers (v

1

, v

2

) or in an equivalent manner by the two real

numbers (l

1e

, l

2e

) with l

ie

¼ev

i

, which take discrete values.

In the homogenization process, the discrete variables

become the continuous variables (l

1

, l

2

), which means

that (l

1

, l

2

) locate the material points of the equivalent

continuum. From the eqn (1), we see that the position of

the material point located by (l

1

, l

2

) is r

0

(l

1

, l

2

), then the

continuum model in space coordinates is given by the

change of variables:

(l

1

, l

2

) →r

0

(l

1

, l

2

): (18)

The easiest way to ﬁnd the relation between the vectors S

i0

and the stress tensor j

0

in the continuum is to make the

change of variables, eqn (18), in the virtual power formula-

tion eqn (9).

Let L

x

v be the gradient tensor of the virtual velocity ﬁeld

v. The change of variables, eqn (18), yields:

ץv

ץl

i

¼L

x

v@

ץr

0

ץl

i

(19)

The change of variables, eqn (18), in eqn (10) yields:

᭙v, ¹

q

S

i0

ᮏ

ץr

0

ץl

i

_ _

: L

x

v

1

g

dx

1

dx

2

þ

q

f·v

1

g

dx

1

dx

2

¼0

where g is the determinant of the matrix of the components

of the two vectors ץr

0

=ץl

1

and ץr

0

=ץl

2

and where : denotes

the scalar (contracted) product of two tensors.

The stress tensor j

0

is then:

j

0

¼

1

g

S

i0

ᮏ

ץr

0

ץl

i

(20)

which veriﬁes the balance equation:

div

x

j

0

þ

1

g

f ¼0 (21)

To derive the constitutive relation between j

0

and the strain

tensor e(u

0

), it is useful to obtain the expression of j

0

in

terms of N

b0

, L

b0

and e

b0

. This expression can be deduced

from eqns (20) and (11), but it is easiest to derive it from a

suitable choice of virtual velocities in the virtual power

formulation of the self-equilibrium of the cell, eqn (14).

Let E be any tensor and let v

n

¼E@r

n

, then eqn (2)

yields:

v

E

R

(b)

¹v

O

R

(b)

¼E@ L

b0

e

b0

¹

ץr

0

ץl

i

d

ib

_ _

and eqn (14) becomes:

beB

R

N

b0

e

b0

ᮏ L

b0

e

b0

¹

ץr

0

ץl

i

d

ib

_ _

_ _

: E ¼0

as E is any tensor, the previous relation added to eqns (11)

and (20) yield:

j

0

¼

1

g

beB

R

N

b0

L

b0

e

b0

ᮏe

b0

(22)

This shows that, as expected for an elastic model, the stress

tensor is symmetrical.

To get the constitutive relation in space coordinates, we

then have to carry the expression of ץu

0

=ץl

i

(see eqn (19)) in

eqn (15), itself carried in eqn (22). The result is then:

j

0

¼

1

g

k

z

k

[z

k

: e(u

0

)] (23)

where z

k

¼(ץr

0

=ץl

i

) ᮏf

ik

and e(u

0

) is the symmetrical

part of L

x

u

0

.

The continuous model of the truss in space coordinates is

then given by the balance eqn (21) and the constitutive

relation, eqn (23). This model is clearly a two-dimensional

elastic one, its elastic energy density being:

1

2g

k

[z

k

: e(u

0

)]

2

Continuous modeling of lattice structures by homogenization 703

10 A SIMPLE ONE-DIMENSIONAL EXAMPLE

We study a one-dimensional straight structure built by the

periodic repetition of two beams of stiffnesses k

c

/e and k

d

/e

and lengths l

c

¼ eL

c

and l

d

¼ eL

d

submitted to external

forces f

e= ˜ n

¼ef

e=n

e parallel to e, the unit vector parallel to

the structure, so the problem is entirely one-dimensional.

Only one index v is needed to number the cells. There are

two nodes 1 and 2 and two beams c and d in the reference

cell.

The beam c is oriented from node 1 to node 2, the beam d is

oriented from node 2 to node 1, then:

O

R

(c) ¼1, E

R

(c) þ2, d

c

¼0

O

R

(d) ¼2, E

R

(d) ¼1, d

d

¼0

(there is only one d per beam).

The problem is one-dimensional, so there is only one

coordinate l and only one vector S

0

¼S

0

e proportional to

e. The balance equation of the equivalent continuum reads

in this example:

ץS

0

ץl

þf ¼0

The self equilibrium of the reference cell reads:

N

c0

¹N

d0

¼0

N

d0

¼N

c0

¼0

Then there is only one J and only one f which are:

J¼

¸

¸

¸

¸

J

c

J

d

¼

k

c

k

d

k

c

þk

d

_ _

1=2¸

¸

¸

¸

1

1

f¼(d

c

J

c

þd

d

J

d

)e ¼

k

c

k

d

k

c

þk

d

_ _

1=2

e

u

0

¼u

0

e is parallel to e, it depends on l and the constitutive

relation is the well-known relation:

S

0

¼

k

c

k

d

k

c

þk

d

ץu

0

ץl

The structure is exactly periodic then:

r

0

(l) ¼l(L

c

þL

d

)e

Setting x ¼ l(L

c

þ L

d

) gives the constitutive relation in

space variable x:

j

0

¼(L

c

þL

d

)

k

c

k

d

k

c

þk

d

ץu

0

ץx

In the following, the previous result is compared to that of

the ‘static condensation’ method.

The method used in

4,5

consists in looking for the dis-

placement u( ˜ n) of any node ˜ n as the value of a function U(x)

at the location x( ˜ n) of the node. The differences u( ˜ m) ¹u( ˜ n)

in the tension displacement relations of the beams are

expanded using Taylor expansions to the second order, an

origin being chosen in the cell.

Choosing the node 2 as origin yields for example:

u((1, v þ1)) ¹u((2, v)) ¼1

d

UЈ þ

(1

d

)

2

2

UЉ

u((1, v)) ¹u((2, v)) ¼ ¹1

c

UЈ þ

(1

c

)

2

2

UЉ

The contribution of the beams c and d of the cell v to the

elastic energy of the structure is:

E

v

¼

1

2

_

k

c

e

¹1

c

UЈ þ

(1

c

)

2

2

UЉ

_ _

2

þ

k

d

e

1

d

UЈ þ

(1

d

)

2

2

UЉ

_ _

2

_

The method, called ‘static condensation’, used in

4,5

con-

sists then of setting ץE

v

=ץU" equal to zero, this enables to

express UЉ and E

v

in term of UЈ:

E

v

¼

1

2

k

c

k

d

(l

c

)

2

(l

d

)

2

(l

c

þl

d

)

e[k

c

(l

c

)

4

þk

d

(l

d

)

4

]

(UЈ)

2

(l

c

þl

d

)

The length (l

c

þ l

d

) of the cell is in E

v

the element of

integration dx in the passage of the energy of the discrete

structure to that of the equivalent continuous bar. Thus the

elastic stiffness of the equivalent continuous bar, given by

this method, is:

k

c

k

d

(L

c

)

2

(L

d

)

2

(L

c

þL

d

)

k

c

(L

c

)

4

þk

d

(L

d

)

4

REFERENCES

1. Abrate, S., Continuum modeling of lattice structures. Shock

Vibr. Dig., 1985, 18(1), 16–21.

2. Abrate, S., Continuum modeling of lattice structures for

dynamic analysis. Shock Vibr. Dig., 1988, 21(10), 3–8.

3. Abrate, S., Continuum modeling of lattice structures 3. Shock

Vibr. Dig., 1991, 23(3), 16–21.

4. Noor, A. K. and Andersen, C. M., Analysis of beam-like

lattice trusses. Comput. Meth. Appl. Mech. Engng, 1979,

20, 53–70.

5. Noor, A. K. and Nemeth, M., Micropolar beam models for

lattice gridswith rigid joints. Comput. Meth. Appl. Mech.

Engng, 1980, 21, 249–263.

704 H. Tollenaere, D. Caillerie

6. Noor, A. K., Continuum modeling for repetitive lattice struc-

tures. Appl. Mech. Rev., 1988, 41(7), 285–296.

7. Renton, J. D., Behavior of Howe, Pratt and Warren trusses. J.

Struct. Div. ASCE, 1969, 95, 193–212.

8. Renton, J. D., General properties of space grids. Int. J. Mech.

Sci., 1970, 12, 801–810.

9. Renton, J. D., The beamlike behavior of space trusses. AIAA

J., 1984, 22(2), 273–280.

10. Bensoussan, A., Lions, J. L. & Papanicolaou, G., Asymptotic

Analysis for Periodic Structures. North Holland, Amsterdam,

1978.

11. Sanchez-Palencia, E., Non Homogeneous Media and Vibra-

tion Theory. Springer, Berlin, 1980.

12. Cioranescu, D. and Saint Jean Paulin, J., Proble`mes de

Neumann et de Dirichlet dans des structures re´ticule´es de

faible e´paisseur, Vol. 303, Se´rie I, No. 1, pp. 7–10. CRAS,

Paris, 1986.

13. Cioranescu, D. and Saint Jean Paulin, J., Reinforced and honey-

comb structures. J. Math. Pures Appl., 1986, 65, 403–422.

14. Caillerie, D., Trompette, P. and Verna, P., Homogenenization

of Periodic Trusses, Congre`s IASS, Madrid, 1989.

15. Tollenaere, H., Mode`les bidimensionnels de tisse´s—

Homoge´ne´isation des treillis en vibrations libres. The`se de

l’Institut National Polytechnique de Grenoble (France), 1994.

16. Verna, P. and Caillerie, D., Homoge´ne´isation d’un treillis

plan hyperstatique inte´rieurement. Congre`s StruCoMe,

Paris, 1990.

17. Verna, P., Mode´lisation continue des structures discre`tes par

homoge´ne´isation: cas des treillis. The`se de l’Institut National

Polytechnique de Grenoble (France), 1991.

18. Moreau, G. and Caillerie, D., Continuum modelling of lattice

structures in large displacement framework. Proceedings of

Civil Comp ’95, Cambridge, B. H. V. Topping Ed., Civil-

Comp Press, Edinburgh, U.K., 1995.

Continuous modeling of lattice structures by homogenization 705

all the nodes of a cell have approximately the same location given by r0 (l1 . which means that it is the repetition of nearly identical cells. rn1 (l1e . v2 ). v1 . . The nodes ˜ are then numbered by a triplet n ¼ (n. v 2). method enables the expected well-known result to be found. and E(b) is a node numbered (m. anyway in a cell numbered (v 1 þ d 1. l2 ) are of the order of e and given by ern1 (l1e .4). Actually. v 2). they are denoted respectively O R(b). that is. v2 ). v1 . The lattice truss is assumed to be quasi-periodic. The same system is used for beams which are numbered ˜ b ¼ (b. n. ˜ ˜ denoted O(b). v2 ) is in the cell ˜ (v 1. l 1e ¼ ev 1 and l 2e ¼ ev 2. we assume that the positions of the nodes are given by an expansion of the form: ˜ ˜ n ¼ (n. v1 . it is chosen so ˜ ˜ that the origin O(b) of the beam b ¼ (b. and it is sufﬁcient to point out the differences of results between this method and one used in litterature. 2 NUMBERING OF NODES AND BEAMS The truss under consideration is made of beams only linked by their ends. At ﬁrst order in e. The end E(b) of ˜ the beam b ¼ (b. v 2 þ d 2) ˜ with d ie{-1. l2e ). ˜ ˜ ˜ O(b) and E(b) are then mappings from the set of beams B ˜ ˜ ˜ into the set of nodes N and O ¹ 1 (n) (respectively E ¹ 1 (n)) denotes the set of beams the origin (respectively the end) of ˜ which is n. l2 ). which means that one the two nodes is the origin of the oriented beam. ˜ ˜ In n ¼ (n. Moreover. so the geometry of the truss is completely deﬁned by the positions of the nodes and the connectivity between them. v1 . We use an asymptotic expansion method which assumes that e tends to zero. which is equivalent to saying that e ¼ 1=(Nc )1=2 is a small parameter (N c being the number of cells in the structure). Then. The other node is the end and is denoted E(b). 0. v1 .l 2). 1. The sequence of structures has to be Fig. v2 )) ¼ R(l1e ) cos l2e þ e i 2 5 p p i 2e þe j þ sin l 2 5 with R(l1 ) ¼ Ri þ l1 (Re ¹ Ri ). m. 1}. an exactly periodic lattice of periods eL1 and eL2 is deﬁned by: ˜ r(n) ¼ l1e L1 þ l2e L2 þ ern which ﬁts eqn (1) with r0 ¼ l1 L1 þ l2 L2 and with rn1 ¼ rn independent of (l 1. l2e ) þ …. We use a system of numbering of nodes and beams that reﬂects the regularity of the structure and leads to the curvilinear coordinates of the equivalent continuum. which means ˜ that the node n is the node n in the cell (v 1. The truss is two-dimensional. l2e ) þ ern1 (l1e . v 2). v 2) or in a neighboring one. d 1b. v 1 þ d 1. Each beam links two nodes and is oriented. E R(b). Caillerie completely deﬁned. the indices n and b number the nodes and beams of a reference cell the sets of which are denoted N R and B R. Tollenaere. as the orientation of a beam is arbitrary. v1 . v2 ) is in the cell (v 1. In this way. their positions with respect to the point r0 (l1 . a node numbered (n. 3 GEOMETRY OF A QUASI-PERIODIC TRUSS The lattices under consideration are structures with a large number of cells. r(n) ¼ r0 (l1e . l2e ) þ e2 rn2 (l1e . that is to say that for each value of e it is necessary to know where the nodes of the structure are located in the space. As the truss is repetitive. v 2). v2 )) ¼ R(l1e ) cos l2e i þ sin l2e j 2 2 h p p r((2. v1 . Fig. v2 ). v1 . d 2b. that is which of them are linked by a beam. we intend to consider trusses which are not exactly repetitive but quasi-periodic which means that their geometry can vary slightly. 2 2 v þ d ). Example of a quasi-periodic structure in the cell (4. v1 . that means that we consider a sequence of structures with more and more cells which are smaller and smaller. v 1. l 2). D.700 H. 1 displays an example of a quasi-periodic structure which is described by: h p p i r((1. The beams link nodes of the same cell or of neighboring cells. d 2 depend only on b. l2e ). … are given functions of (l 1. l2e ) þ … ð1Þ where l ie ¼ ev i and where r0 (l1e . so the cells composing it are numbered by a system of two integers (v 1. The Einstein convention of repeated index is used throughout this paper. v2 ) and b ¼ (b. v2 ). d 1.

4 EXPANSION OF BEAM TENSIONS Consistently with the assumption about geometry. the expansion of the tension displacement relation yields: ! u0 ib b0 b0 b0 ER (b)1 OR (b)1 ¹u þ i d ·e (7) N ¼k u l N b1 ¼ …. l2 ) and giving the equivalent continuous model. v1 . it remains to develop the balance equations of the nodes. l2 ) by using Taylor expansions: ! 1 ER (b) r0 ib b0 OR (b) ¹r þ id (2) e ¼ b0 r L l It is assumed that. l2e ) þ e2 N b2 (l1e . the sums e2 v1 . v2 are approximated by integrals q dl1 dl2 . The length 1b ¼ kr(E(b)) ¹ ˜ r(O(b))k and the unitary vector of the beam ˜ ˜ ˜ ˜ e b ¼ [r(E(b)) ¹ r(O(b))]=1b can be expanded in: 1b ¼ eLb0 (l1e . v2 ). This ˜ ˜ ˜ ˜ implies that the stiffness kb ¼ (Eb Ab )=1b of the beam can be . the r(n) can be developed in: h p p i r((1. which are: ˜ n. recalling the interpolations of ﬁnite elements. l2e ) þ eN b1 (l1e .Continuous modeling of lattice structures by homogenization ˜ For this example. v2 )) ¼ R(l1e ) cos l2e i þ sin l2e j 2 2 h p p i p 1e 2e 2e i þ cos l j þ e2 […] þ e R(l ) ¹ sin l 5 2 2 Eqn (1) enables the geometrical characteristics of the ˜ ˜ beams to be expanded. l2e ) þ eeb1 (l1e . the geometry of the structure remains quasi-periodic. in the ˜ ˜ process of expansion. l2e ) þ eun1 (l1e . The difference between the positions of a node before and after deformation is its displace˜ ment u(n). we ˜ ˜ assume that the Young modulus Eb and the section Ab of a beam are given by series expansion in power of e. which implies that e is very small. ˜ ˜ where f e=n is the external force applied on the node n. l2e ) ˜ This shows that f e=n is assumed to be of the order of e. it is not for the use of the asymptotic expansions developed here. eqn (4). which means that the positions of the nodes are given by expansions of the form eqn (1). as e decreases. l þ ed ) ¹ u0 (l1e . l2e ) þ … (6) ˜ eb ¼ eb0 (l1e . l2e ) þ ekb2 (l1e . l2e ) can be expanded using a Taylor expansion ˜ (ed 1 is small). l2e ) þ kb1 (l1e . … 5 EXPANSION OF BALANCE EQUATIONS— BALANCE EQUATION OF THE EQUIVALENT CONTINUUM To ﬁnd the equivalent continuum of the truss. after deformation. l2e ) þ …. l2e ) þ e2 un2 (l1e . l2e ) þ … ˜ 701 expanded in: 1 ˜ kb ¼ kb0 (l1e . lie ¼ evi where v0 (l1 . This force is assumed to be: ˜ f e=n ¼ ef e=n (l1e . Consequently the beam tensions N b can be expanded in: N b ¼ N bO (l1e . the displacements of the nodes of the structures are approximately given by the displacement ﬁeld u0 (l1 . Very often in literature 4–6 a displacement ﬁeld is deﬁned on the whole space domain occupied by the structure. the assumption that the structure is quasi-periodic (as described by eqn (1)) and remains quasi-periodic after deformation yields that the displacement of nodes becomes. this assumption will be proved consistent with the continuum ˜ modeling and the order of magnitude of kb . v1 . The expansion of the balance eqn (8) is much easier when the equations are written with a virtual power formulation which reads: ˜ v(n). v1 . v2 )) ¼ R(l1e ) cos l2e i þ sin l2e j 2 2 h p p i r((2. The following Sections 4–8 are devoted to the derivation of the equations governing the ﬁeld u0 (l1 . l2e ) þ e2 Lb1 (l1e . lie ¼ evi ð3Þ and by identiﬁcation of terms of same power in e. l2e ) þ … ˜ The expressions of the L bk and ebk can be found in terms of the ﬁelds rnk (l1 . l2e ) þ … e (4) In the tension displacement relations for beams: ˜ ˜ ˜ ˜ ˜ N b ¼ kb [u(E(b)) ¹ u(O(b))]·eb 0 1e 1 2e 2 (5) the differences such as u (l þ ed . ˜ ˜ beB ˜ ˜ ˜ ˜ N b eb ·[v(O(b)) ¹ v(E(b))] þ ˜ ˜ neN ˜ ˜ f e=n ·v(n) ¼ 0 (9) Sums such as beB can be split in v1 . v(n) ¼ v0 (l1e . l2e ). We get then the virtual power formulation of the equilibrium of the equivalent continuum by taking in eqn (9): ˜ ˜ n ¼ (n. a function of a domain. ˜ ˜ beO ¹ 1 (n) N b eb ¹ ˜ ˜ beE ¹ 1 (n) ˜ ˜ ˜ N b eb þ f e=n ¼ 0 ˜ ˜ (8) Eqn (3) shows that at ﬁrst order in e. This ﬁeld is an extension of the displacements of the nodes. Consequently. v2 beBR and. l2 ). l2 ) is a smooth given virtual macroscopic velocity ﬁeld. This kind of extension would probably be necessary if a convergence proof was considered. which is then given by an expansion of the form: ˜ u(n) ¼ u0 (l1e .

Eqn (10) is the virtual power formulation of the balance equation of the continuum equivalent to the structure. denote a base of solutions of eqn (14). A@y denotes the image of the vector y by the operator A. 7 CONSTITUTIVE RELATIONS OF THE EQUIVALENT CONTINUUM In order to get the description of the continuum equivalent to the structure. u0 = li being considered as data. The constitutive relation is in fact given by eqns (7). it reads: 2 3 u0 i0 ik jk f f @ i (16) S ¼ l k fjk is the linear operator called tensorial or where fik diadic product of the two vectors fik and fjk . A@y. l 2).702 We get then: v0 . v(n) ¼ ev(l1e . (14) and (11). eqns (7) and (14) are equations the unknowns of which are N b0 and un1 . vn 4 q enables the un1 to be eliminated and the a k to be calculated: ak ¼ fik · where fih ¼ beBR u0 li Jbh dib eb0 5 N e ·[v bO b0 ER (b) ¹v OR (b) ] v dl1 dl2 ¼ 0 (13) beBR Then. When A is the diadic product a b. the procedure described above can be carried out in two steps which makes the solving of This last relation is the constitutive relation of the equivalent continuum. then eqn (11) yields the wanted constitutive relation. Indeeed. its balance eqn (10) has to be completed by a constitutive relation linking Si0 to u0 = li . v2 ). Caillerie eqns (7) and (14) easier. we can reckon N b0 and Si0 in terms of the u0 = li : 4 5 u0 b0 bk ik J f · i (15) N ¼ l k Si0 ¼ k which yields: vn . v1 . is such that: a b@y ¼ (b·y)a 8 CONTINUOUS MODEL OF THE TRUSS IN CURVILINEAR COORDINATES The continuous model of the truss in curvilinear coordinates is given then by the constitutive eqn (16) and the balance . beBR N b0 eb0 ·[vER (b) ¹ vOR (b) ] ¼ 0 (14) u0 fik fjk · i l This relation is the virtual formulation of the selfequilibrium of the cell. Let J bk. They are related to the stress tensor j O in Section 9. The tensions N b0 which are also solutions of eqn (14) are then a linear combination of the J bks: N b0 ¼ k Si0 · v0 1 2 dl dl þ li q f·v0 dl1 dl2 ¼ 0 (10) N b0 dib eb0 (11) and f¼ neNR ˜ The previous process explains why the external forces f e=n are assumed to be of the order of e. Solving these equations gives N b0 as a function of u0 = li . then this base is orthogonolized in such a way to eliminate the un1 and to express N b0 in terms of u0 = li . the index k numbering the elements of the base. The vectors Si0 describe the internal efforts of the equivalent continuum in the curvilinear coordinates system (l 1. lie ¼ evi Using the same process as in Section 5. D. we get: v. l2e ). l2e )vn (l1e . f e=n (12) ak Jbk Carrying this relation in the tension displacement eqn (7) yields: ! u0 ak Jbk ¼ kb0 uER (b) ¹ uOR (b) þ i dib ·eb0 l k Orthogonalizing the J bks by: Jbk Jbh ¼ dkh kb0 beBR 6 SELF-EQUILIBRIUM OF THE REFERENCE CELL To establish the constitutive relation of the equivalent continuum we need to use the self balance equations of the cell which is obtained from eqn (9) by taking: ˜ ˜ n ¼ (n. ¹ with: Si0 ¼ beBR q H. Tollenaere. First a base J bk of solutions of eqn (14) is determined. As the eqn (14) is discrete.

it is useful to obtain the expression of j 0 in terms of N b0. eqn (18). Let E be any tensor and let vn ¼ E@rn . 1 f·v dx1 dx2 ¼ 0 g where g is the determinant of the matrix of the components of the two vectors r0 = l1 and r0 = l2 and where : denotes the scalar (contracted) product of two tensors. This model is clearly a two-dimensional elastic one. To get the constitutive relation in space coordinates. L b0 and eb0 . This expression can be deduced from eqns (20) and (11). Remark: carrying u0 . This does not appear obviously in eqns (16) and (17) written in curvilinear coordinates. l2 ): (18) To derive the constitutive relation between j 0 and the strain tensor e(u0 ). we see that the position of the material point located by (l 1. the previous relation added to eqns (11) and (20) yield: j0 ¼ 1 N b0 Lb0 eb0 g beBR eb0 (22) This shows that. eqn (23). Moreover the continuous model. as expected for an elastic model. which could be expected since the considered truss is linear. The result is then: j0 ¼ 1 g z k [z k : e(u0 )] k (23) The easiest way to ﬁnd the relation between the vectors Si0 and the stress tensor j 0 in the continuum is to make the change of variables. eqn (18). l2 ) → r0 (l1 . the stress tensor is symmetrical. The well-posedness of boundary value problems for the equivalent continuum comes down to the positive deﬁniteness of the following quadratic form: 42 3 5 1 u0 u0 ik jk f f @ j · i 2 l l k It can be studied and deduced from general assumptions on the truss. into eqn (15) gives the ﬁrst term N b0 of the expansion. is linear. but it is easiest to derive it from a suitable choice of virtual velocities in the virtual power formulation of the self-equilibrium of the cell. which take discrete values. In the homogenization process. then the continuum model in space coordinates is given by the change of variables: (l1 . its elastic energy density being: 1 2g [z k : e(u0 )]2 k . itself carried in eqn (22). l 2) is r0 (l1 . that is an approximation of the tensions in the beams of the truss. l 2e) with l ie ¼ ev i. l 2). eqn (18). eqn (16). ¹ S : Lx v dx1 dx2 i q g l þ q Some boundary conditions have to be added to this problem. The stress tensor j 0 is then: 1 j0 ¼ Si0 g r0 li (20) which veriﬁes the balance equation: 1 divx j0 þ f ¼ 0 g (21) 9 TWO-DIMENSIONAL ELASTIC MODEL IN SPACE COORDINATES The equivalent continuum constitutive relation. the discrete variables become the continuous variables (l 1.Continuous modeling of lattice structures by homogenization eqn (10) which yields: S þf ¼0 li i0 703 (17) The change of variables. should be a two-dimensional elastic model. we then have to carry the expression of u0 = li (see eqn (19)) in eqn (15). in the virtual power formulation eqn (9). The change of variables. eqn (6) of N b. eqn (14). solution of eqns (16) and (17) complete with boundary conditions. which means that (l 1. l 2) locate the material points of the equivalent continuum. They can be deduced by expanding the boundary conditions of the truss. then eqn (2) yields: r0 ib ER (b) OR (b) b0 b0 ¹v ¼ E@ L e ¹ i d v l and eqn (14) becomes: 4 5 r0 ib b0 b0 b0 b0 N e L e ¹ id : E¼0 l beBR as E is any tensor. From the eqn (1). Let Lx v be the gradient tensor of the virtual velocity ﬁeld v. The cells of the lattice truss are located by the two integers (v 1. v 2) or in an equivalent manner by the two real numbers (l 1e. l2 ). in eqn (10) yields: r0 1 i0 v. The best way to prove that the continuous model is actually a two-dimensional elastic model is to write the balance and constitutive equations in usual space coordinates. yields: v r0 i ¼ Lx v@ li l (19) where z k ¼ ( r0 = li ) fik and e(u0 ) is the symmetrical part of Lx u0 . eqns (16) and (17). The continuous model of the truss in space coordinates is then given by the balance eqn (21) and the constitutive relation.

In the following. D. Tollenaere. dd ¼ 0 (there is only one d per beam). u0 ¼ u0 e is parallel to e. Thus the elastic stiffness of the equivalent continuous bar. dc ¼ 0 OR (d) ¼ 2. Meth. Shock Vibr. S. 5. S. 23(3). the previous result is compared to that of the ‘static condensation’ method. The method used in 4. ER (d) ¼ 1. 2. Engng. v)) ¼ ¹ 1c U þ (1d )2 U 2 (1c )2 U 2 The beam c is oriented from node 1 to node 2. M. used in 4. v)) ¹ u((2. Meth. S. Mech. is: kc kd (Lc )2 (Ld )2 (Lc þ Ld ) kc (Lc )4 þ kd (Ld )4 REFERENCES 1. Choosing the node 2 as origin yields for example: u((1.. Continuum modeling of lattice structures. 53–70. Shock Vibr. K. 20. 16–21. Only one index v is needed to number the cells. the unit vector parallel to the structure. A. M.704 H. so the problem is entirely one-dimensional. v þ 1)) ¹ u((2. Appl. 1980.. 18(1). C. ER (c) þ 2. given by this method. Abrate. 1991. the beam d is oriented from node 2 to node 1. Analysis of beam-like lattice trusses... There are two nodes 1 and 2 and two beams c and d in the reference cell. K. A.5 consists then of setting Ev = U" equal to zero.. Micropolar beam models for lattice gridswith rigid joints. 4. and Andersen. 3. and Nemeth.. Dig. called ‘static condensation’. 1988. Comput. Engng.. Dig. 16–21. then: OR (c) ¼ 1. Noor. Continuum modeling of lattice structures 3. 3–8. Mech. Abrate. Caillerie Setting x ¼ l(L c þ L d) gives the constitutive relation in space variable x: j0 ¼ (Lc þ Ld ) k c k d u0 kc þ kd x 10 A SIMPLE ONE-DIMENSIONAL EXAMPLE We study a one-dimensional straight structure built by the periodic repetition of two beams of stiffnesses k c/e and k d/e and lengths l c ¼ eLc and l d ¼ eL d submitted to external ˜ forces f e=n ¼ ef e=n e parallel to e. The problem is one-dimensional. Comput. Appl. Shock Vibr. Noor. 21. 21(10). Abrate. Dig. this enables to express U and E v in term of U : Ev ¼ 1 kc kd (lc )2 (ld )2 (lc þ ld ) (U )2 (lc þ ld ) 2 e[kc (lc )4 þ kd (ld )4 ] The length (l c þ l d) of the cell is in E v the element of integration dx in the passage of the energy of the discrete structure to that of the equivalent continuous bar. v)) ¼ 1d U þ u((1. The balance equation of the equivalent continuum reads in this example: S þf ¼0 l The self equilibrium of the reference cell reads: N c0 ¹ N d0 ¼ 0 N d0 ¼ N c0 ¼ 0 Then there is only one J and only one f which are: c 2 c d 31=2 1 J kk J¼ ¼ d c þ kd k J 1 2 f ¼ (dc Jc þ dd Jd )e ¼ kc kd kc þ kd 31=2 e 0 The contribution of the beams c and d of the cell v to the elastic energy of the structure is: 4 2 1 kc (1c )2 v ¹ 1c U þ U E ¼ 2 2 e 2 32 5 kd d (1d )2 1 U þ U þ e 2 The method. an origin being chosen in the cell. The differences u(m) ¹ u(n) in the tension displacement relations of the beams are expanded using Taylor expansions to the second order. Continuum modeling of lattice structures for dynamic analysis.5 consists in looking for the dis˜ ˜ placement u(n) of any node n as the value of a function U(x) ˜ ˜ ˜ at the location x(n) of the node. 1985. it depends on l and the constitutive relation is the well-known relation: S0 ¼ kc k c k d u0 þ kd l The structure is exactly periodic then: r0 (l) ¼ l(Lc þ Ld )e . so there is only one coordinate l and only one vector S0 ¼ S0 e proportional to e.. 1979. 249–263.

1969. Paris. D. AIAA J. U. J. ` ´ 15. 22(2). 1991... Sci.. Struct. and Saint Jean Paulin. A. Cioranescu. Madrid. The beamlike behavior of space trusses. pp. J. D. Proceedings of Civil Comp ’95. J. K. 285–296. and Caillerie. Continuum modelling of lattice structures in large displacement framework. P. D. 7.. Pures Appl. P. L.K.. G.. Problemes de ´ ´ Neumann et de Dirichlet dans des structures reticulees de ´ ´ faible epaisseur.. Congres IASS. H... CivilComp Press. Continuum modeling for repetitive lattice structures. 95. 1986. 9. Non Homogeneous Media and Vibration Theory. 705 13. Paris. A. 8. 1978.. Noor. Modelisation continue des structures discretes par ´ ´ ` homogeneisation: cas des treillis. 65. H. Cambridge. J. and Saint Jean Paulin. 1984. Berlin. Mech. 10. CRAS. Pratt and Warren trusses.Continuous modeling of lattice structures by homogenization 6. 18.. 1988. ´ ` 17. Renton. 7–10. Caillerie.. Congres StruCoMe. 1970.. ASCE. Verna. Moreau. Rev. These de l’Institut National Polytechnique de Grenoble (France). Int. North Holland. Div. J. & Papanicolaou. Behavior of Howe. Vol. and Caillerie. 1986. Asymptotic Analysis for Periodic Structures. 801–810. 11. Edinburgh. Renton. Math. Cioranescu. P. J. D. B. 1990. 303. 41(7). 14. Renton. 273–280.. Springer. Trompette. No. V. Topping Ed. G. Amsterdam. J. 12. Mech. 403–422. D. 1980.. D.. Bensoussan. Sanchez-Palencia. Lions. Modeles bidimensionnels de tisses— ´ ´ ` Homogeneisation des treillis en vibrations libres. Homogeneisation d’un treillis ` ´ plan hyperstatique interieurement. These de l’Institut National Polytechnique de Grenoble (France). ` 12. D.. 1994. ´ ´ 16. . 1.. 1995. 193–212. Appl. Verna. 1989. Serie I. P.. E. General properties of space grids. Tollenaere. J. Reinforced and honeycomb structures. Homogenenization ` of Periodic Trusses. D... and Verna. J.

- Comparison of Current Test Requirements
- Added Mass
- ger3695e
- ger3569g
- 569
- GAS FOIL BEARING ANALYSIS AND THE EFFECT OF BUMP FOIL THICKNESS ON ITS PERFORMANCE CHARACTERISTICS USING A NON-LINEAR MATRIX EQUATION SOLVER
- GCRC-201-11111
- ds40pg
- Ecss e Hb 32 20 Part8a
- MIL-S-62549
- Basics Balancing 101
- 322_eu
- Generalized Displacements and the Accuracy of Classical Plate Theory

Sign up to vote on this title

UsefulNot useful- Economics of Construction - BOW
- truss
- Analysis of Steel Rooof Truss
- Roof Truss Design Procedure
- Download
- Design Steel Truss
- 15A11 Superstructures II
- Mutlu Seçer, Volume12-3,4
- Vector Mechanics - Workout Problems
- Rr410103 Structural Engineering III Design Drawing
- Lecture 3
- Report of Truss
- A380 Hangar
- A Ledger of Mark Contest
- Bridge
- chp-12
- BLDGTC5-RSW # FG-01
- Win_Tensor_FAQ.pdf
- Chapter+6+-+connection
- ch6
- ENGN.2060-012_Assignment_02_Solution.pdf
- Oct12
- wi301
- 20 Buckling Steel Print
- Ty 2rihwou3gs
- Mat Chapter 42
- Hassler Whitney - Tensor Products of Abelian Groups
- Morphology Development in Kenics Static Mixers
- Bridge Building Lab Report
- Staad Examples
- Continuous Modeling of Lattice Structures by Homogenization