0 Votes +0 Votes -

5 vues12 pagesoverview

Sep 22, 2016

© © All Rights Reserved

PDF, TXT ou lisez en ligne sur Scribd

overview

© All Rights Reserved

5 vues

overview

© All Rights Reserved

- Hidden Figures: The American Dream and the Untold Story of the Black Women Mathematicians Who Helped Win the Space Race
- Hidden Figures Young Readers' Edition
- The Law of Explosive Growth: Lesson 20 from The 21 Irrefutable Laws of Leadership
- The E-Myth Revisited: Why Most Small Businesses Don't Work and
- The Wright Brothers
- The Power of Discipline: 7 Ways it Can Change Your Life
- The Other Einstein: A Novel
- The Kiss Quotient: A Novel
- State of Fear
- State of Fear
- The 10X Rule: The Only Difference Between Success and Failure
- Being Wrong: Adventures in the Margin of Error
- Algorithms to Live By: The Computer Science of Human Decisions
- The Black Swan
- Prince Caspian
- The Art of Thinking Clearly
- A Mind for Numbers: How to Excel at Math and Science Even If You Flunked Algebra
- The Last Battle
- The 6th Extinction
- HBR's 10 Must Reads on Strategy (including featured article "What Is Strategy?" by Michael E. Porter)

Vous êtes sur la page 1sur 12

www.elsevier.com/locate/compstruct

composite plates: a sub-laminates nite element approach

Marco Gherlone *, Marco Di Sciuva

Department of Aerospace Engineering, Politecnico di Torino, Corso Duca degli Abruzzi, 24, 10129 Torino, Italy

Available online 18 September 2006

Abstract

In this work a plate mixed nite element based on a rened zig-zag plate model is formulated for the analysis of multilayered composite plates subjected to thermal and mechanical loads. The zig-zag plate model is characterised by: (i) through-the-thickness cubic inplane displacements and linear transverse displacement, (ii) transverse shear stresses continuity, (iii) satisfaction of traction equilibrium

conditions on top and bottom faces of the laminate thus allowing for the presence of tangential distributed loads, (iv) the transverse

normal deformability is taken into account by means of a r33 supposed constant along the thickness and by the use of the complete constitutive equations of orthotropic materials, (v) use, as degrees of freedom of the model, of displacements and transverse shear stresses of

external faces of the laminate. In particular, this last property may be extended to the associated plate nite element that, having its nodal

degrees of freedom on the external faces, allows a through-the-thickness discretisation in some sub-domains called sub-laminates. Some

numerical results are presented in order to show that the use of more sub-laminates improves the capability of the approach to capture

with high accuracy the though-the-thickness distributions of transverse displacements, strains and stresses also for very thick plates.

2006 Elsevier Ltd. All rights reserved.

Keywords: Multilayered plates; Thermo-mechanics; Transverse deformability; Sub-laminates; Plate mixed nite element

1. Introduction

Multilayered composite plates and shells are extensively

used in the eld of aeronautical, civil and marine engineering due to their high specic stiness and strength. The

increasing use of composite components as primary structural parts of more complex structures leads to the adoption of very thick panels and beams; high thick-to-side

length ratios exacerbate some of the peculiar aspects of

multilayered composite structures [15]:

transverse anisotropy: remarkable variations in the

mechanical properties from layer to layer are, for example, typical of the more and more successful sandwich

structures; transverse anisotropy leads to a more complex interaction among the dierent deformation modes

*

E-mail addresses: marco.gherlone@polito.it (M. Gherlone), marco.

disciuva@polito.it (M. Di Sciuva).

1

Tel.: +39 0115646808; fax: +39 0115646899.

0263-8223/$ - see front matter 2006 Elsevier Ltd. All rights reserved.

doi:10.1016/j.compstruct.2006.08.004

gives to the strain and stress states a more evident 3D

nature;

high transverse deformability: typical composite materials (bre reinforced plastics) exhibit shear moduli-toin-plane moduli ratios that are lower than those of

isotropic materials, thus, also for thin structures, the

transverse shear deformability may be remarkable; also

transverse normal deformability may be not negligible

when thick laminates are considered;

damage sensibility: composite structures may experience

damage as a consequence of operating conditions or

manufacturing operations; in any case, due to the discontinuous nature of a multilayered structure, the damage occurrence probability and the eects of damages on

the load carrying capabilities of composite structures are

issues of main importance;

load carrying capacity under environmental constraints:

bre reinforced plastics are particularly sensitive to the

action of temperature variations, moisture, radiations;

126

Nomenclature

a, b, h in-plane dimensions and thickness of the plate

L

common value of the sides for a square plate

N

number of orthotropic layers

n

number of sub-laminates

(k)

h

thickness of the kth layer

(k) (k) +

z , z thickness coordinates of the lower and upper

interfaces of the kth layer

V

volume of the plate

X

reference surface of the plate

S

lateral surface of the plate

C

boundary curve of the plate

xi (i = 1, 2, 3) Cartesian orthogonal coordinates

xa (a = 1, 2) in-plane orthogonal coordinates

x3

thickness coordinate

Vi

displacement eld

eij (ca3) strain tensor (transverse shear components)

rij (sa3) stress tensor (transverse shear components)

r

sa

rth term in the Taylor expansion of the smeared

contribution to the displacement eld

(k)

/a

intensity of the zig-zag contribution to the displacement eld in the kth layer

(k)

Sa

intensity of the in-plane displacement jump at

the interface between the kth and the (k + 1)th

layer

(k)

Tab sliding constants at the interface between the kth

and the (k + 1)th layer

Ei (i = 1, 2, 3) Young moduli of the material

Gij

shear moduli of the material

mij

Poisson coecients of the material

ai (i = 1, 2, 3) thermal expansion coecients of the material

q

density of the material

Qabvw in-plane reduced transformed stiness coecients

Qa3v3 transverse shear transformed stiness coecients

S33, Rab transverse normal and coupling terms of the

constitutive equations

xab,k33 thermo-mechanical terms of the constitutive

equations

H

temperature variation with respect to the reference value

enthalpy

strain energy

tangential loads distributed on bottom and top

faces

q0 ; qh transverse loads distributed on bottom and top

faces

ri i 1; 2; 3 distributed loads applied to the lateral

cylindrical surface

H iab ; Clab i 1; . . . ; 4 l 0; h through-the-thickness

cubic Hermitian shape functions

L1, L2 through-the-thickness linear Lagrange polynomial

S 0a ; S ha in-plane displacements of the bottom and top

faces

w0, wh transverse displacement of the bottom and top

faces

s0a3 ; sha3 transverse shear stresses of the bottom and top

faces

u0i ; uhi i 1; . . . ; 8 degrees of freedom of the bottom

and top faces for in-plane displacement V1

v0i ; vhi i 1; . . . ; 8 degrees of freedom of the bottom

and top faces for in-plane displacement V2

w0i ; w0;xi ; w0;yi ; whi ; wh;xi ; wh;yi i 1; . . . ; 4 degrees of freedom of the bottom and top faces for transverse

displacement V3 and bending rotations

s0xzi ; shxzi i 1; . . . ; 8 degrees of freedom of the bottom

and top faces for transverse shear stresses s13

s0yzi ; shyzi i 1; . . . ; 8 degrees of freedom of the bottom

and top faces for transverse shear stresses s23

Pi (i = 1, . . . , 8) parabolic 2D shape functions (serendipity)

Hi, Hxi, Hyi (i = 1, . . . , 4) cubic 2D shape functions (Hermite)

[M], [K] mass and stiness matrix of the plate nite element

{FM}, {FH} nodal forces (mechanical and thermal contributions) of the plate nite element

{q}

degrees of freedom of the plate nite element

{h}

nodal values of the temperature for the plate nite element

coecient of the matrix is greater than that of bres,

when temperature variations are present, high values

of transverse stresses may appear at the interfaces

between dierent layers thus causing the risk of delaminations [5].

eect (exacerbated for thick laminates), i.e. in-plane

displacements have 1st derivatives (with respect to the

thickness coordinate) that are discontinuous at the interfaces between layers with dierent mechanical

properties;

2. damage in composites may be of several types; the more

typical one is the delamination, i.e. a jump of in-plane

displacements at a particular interface where the adhesive lm is not working correctly (if opening of delami-

As a consequence, models for the analysis of multilayered composite structures should account for the following

capabilities:

H

/

p0a ; pha

discontinuous); in any case, damage modelling implies

a more complicated through-the-thickness distribution

of displacements eld;

3. transverse stresses must be through-the-thickness continuous and their values on top and bottom faces must

be equal to the applied distributed tangential and transverse loads;

4. thermal eects must be evaluated at least as prescribed

temperatures giving equivalent thermal loads.

There are essentially two approaches for modelling composite structures: (i) models that have only displacements as

primary unknowns and (ii) mixed models with displacements and stresses (mainly transverse ones) as unknowns

[6]. Moreover, depending on the assumed distribution along

the thickness for the primary unknowns, we may have [4]:

(a) equivalent single layer (ESL) theories (the a priori supposed behaviour for the unknowns is based on the whole

thickness of the laminate), (b) layerwise (LW) theories (the

distribution of the unknowns is assumed layer per layer).

Examples of ESL models based on displacements (i)(a)

are the classical lamination theories (CLPT [7], FSDT [8],

HSDT [9]). They are easy to implement and give good

results in term of global structure responses (displacements,

frequencies, critical loads) for thin plates and beams with a

reduced transverse anisotropy; they are, in a general case,

not able to assess points 13. LW mixed models (ii)(b)

are suitable to satisfy conditions 13; their main drawback

is the computational complexity, also due to the fact that

the number of unknowns depends on the number of layers.

From this point of view, LW theories based on displacements (i)(b) are interesting; for some of them the number

of unknowns increases with the number of layers [10,11],

for some others (zig-zag models by Di Sciuva [1215]) the

number of degrees of freedom does not change and is equal

to that of ESL classical theories. These latter theories are

able to assess point 1, point 2 (at least for not severe delaminations) and point 3 (only for transverse shear stresses

vanishing on top and bottom faces), thus they are a good

compromise between computational simplicity and accuracy. In this work we will present an improved version of

the classical zig-zag models able to increase the modelling

capabilities of the original ones.

The rst aspect for which these have to get better is transverse normal deformability. As is well known, the transverse

stresses in general and the normal one in particular, may be

calculated in three ways: using theories in which these stresses are primary unknowns (a priori evaluation) or adopting an a posteriori strategy, computing them by means of

the constitutive equations or integrating the indenite equilibrium equations. The last approach is, in most cases, the

more accurate but in some others is not able to recover the

correct stress on the external faces; the use of the constitutive

equations could be the necessary way if we are using a displacements-based theory as the zig-zag one is. The problem

with this approach is that results may be inaccurate when

127

strains from displacements and then stresses from strains

thus obtaining a discontinuous transverse normal stress. A

better way could be that of using mixed constitutive equations, i.e. the complete material constitutive equations but

with transverse normal stress as an input and the corresponding strain as an output; transverse normal stress is supposed to have a known through-the-thickness distribution

whose intensity is calculated in such a way that the two existing transverse normal strains (that obtained by the mixed

constitutive relations and that obtained dierentiating the

transverse displacement) are least squares compatible.

Aitharaju and Averill [18] proposes a constant r33 while

Tessler suggests [19] a cubic one. As a matter of fact, this

strategy is very similar to that adopted in a mixed theory in

which the transverse normal stress is an unknown; the dierence is that the compatibility condition, imposed a priori,

relates the stress unknown to the displacements ones so that

the theory is based only on the latter ones.

Moreover, for some applications in which tangential distributed loads act on the top and bottom faces of the laminate, transverse shear stresses should assume on them the

value of the external loads [20].

The Hermitian zig-zag plate model (HZZ) [21,22] is a

renement of the zig-zag plate models pioneered by Di Sciuva [1215], being able to assess also transverse normal

deformability (considering a through-the-thickness linear

transverse displacement and following Averill and Tesslers

approaches for calculating r33 supposed to be constant

along the thickness) and traction equilibrium conditions

on top and bottom laminate faces (in case of external tangential applied loads).

The sub-laminates methodology, proposed among

others by Averill and Yip [20], may be coupled with

the HZZ model thus obtaining further modelling

improvements:

for very thick laminates or, in any case, for sandwich

panels, transverse displacements may exhibit a non-linear distribution;

the constant value of r33 may give an idea of the average value of the stress but can be in some cases

inaccurate;

for some applications (piezoelectric layers bonded on

the external faces of the laminate) it can be important

to model structures with discretely varying thickness;

original zig-zag models and also HZZ are able to

model only sliding-type defects and the intensity of this

damage must not be excessive in order to avoid the socalled slip-locking phenomenon [23]; thus, there is the

need for avoiding incorrect results and to model also

other kinds of damage (for example, opening mode

damages).

The key idea of the sub-laminates strategy [20,21,23] is to

divide the whole laminate thickness in some sub-domains

and to adopt for each of them a particular distribution of

128

displacements and/or stresses. Thus we obtain an higherorder approximation without increasing the order of the

underlying theory but localising it in some reduced portions

of the thickness domain. In [21,23,24] the beam nite element based on the HZZ has been developed and applied

to the analysis of undamaged and damaged composite

beams subjected to mechanical loads, in [25,26] there is an

analysis of thermally loaded undamaged and damaged

beams and plates and in [27] the HZZ + sub-laminates

approach is assessed in several cases of undamaged and

damaged plates subjected to mechanical and thermal loads.

Icardi has also proposed a coupling between sub-laminates

strategy and zig-zag theories [28]; in this case the transverse

displacement is of the fourth-order and also the continuity

of the transverse normal stress may be respected.

The aim of this article is to present the formulation of the

plate nite element based on the HZZ model and on the sublaminate approach. The underlying theory is briey summarized and the case of thermal eects is also considered; then

the details of the nite element derivation are presented and

a discussion about the concept of sub-laminate is provided.

Some numerical results are nally shown in order to demonstrate the accuracy of the FEM model by means of comparisons with exact elasticity solutions; in particular, the eect

of increasing the number of sub-laminates is considered.

strain and stress vectors are dened as eij and rij, respectively; ca3 and sa3 are also used to indicate the two transverse shear components. The plate dimensions with

respect to the three axes are a, b and h; (k)h is the thickness

of the kth layer while (k)z+ and (k)z are the distances

between the upper and lower face of the kth layer and

the reference surface X (thus (1)z = 0 and (N)z+ = h).

2.2. Displacements, strains and stresses

The in-plane components of the displacement eld are

written as [22]

_

V a xb ; x3 ; t V a xb ; x3 ; t V a xb ; x3 ; t Ve a xb ; x3 ; t

where

Hermitian zig-zag model, we briey recall its basic hypotheses and properties. For the whole set of detailed equations

and derivations, see Di Sciuva and Gherlone [22].

V a xb ; x3 ; t

The Cartesian indicial notation is adopted with Latin

indices ranging from 1 to 3 and Greek indices ranging from

1 to 2. The Einstenian notation (summation over repeated

indices) is also used. An overbar will denote prescribed

quantities, for example external loads f . Dierentiation

will be represented with a comma, i.e. f,s = of/os. Superscript

(k) on the left of any quantity means aliation to the kth

layer; superscript (r) on the right of any quantity identies

the rth element in a series expansion. Integration over the

whole thickness coordinate will be indicated by h i.

Let us consider a multilayered composite plate built up

with N orthotropic layers with dierent physical and

mechanical properties (Fig. 1). V is the region occupied

by the plate, X the reference surface (coinciding with the

plate lower face), C its boundary curve and S is the lateral

cylindrical boundary surface of V. Moreover, Sp, Su and

Cp, Cu denote the two parts of S and C, where tractions

and displacements, respectively, are prescribed. A Cartesian orthogonal system of coordinates xj is used; xa denotes

the in-plane coordinates coinciding with the axes of the reference surface and x3 is the thickness coordinate. The corresponding displacements are indicated with Vj. The full

3

X

xr3 sr

a xb ; t

r0

the thickness coordinate, commonly used in the single-layer

or smeared laminate models;

k1

X

_

q

V a xb ; x3 ; t

/a xb ; tx3 q z

3

q1

is a term (here dened for the kth layer) with jumps in the

1st derivative with respect to x3; when enforcing the transverse shear stresses continuity, the (q)/a are determined;

k

X

q

Ve a xb ; x3 ; t

S a xb ; t

4

q1

enabling interfacial imperfections of axial type to be incorporated (no opening mode damages with transverse displacement discontinuities can be modelled). A linear slip

law is adopted

q

S a xx ; t q T ab xx q sb3 xx ; q z ; t

(q)

where Tab P 0 are the sliding constants of the qth interface ((q)Tab = 0 means perfectly bonded layers while

(q)

Tab ! 1 means completely debonded layers thus leading

sa3 (xx,(q) z+) = 0).

The transverse displacement is supposed to be throughthe-thickness linear

to

(q)

V 3 xb ; x3 ; t L1 x3 w0 xb ; t L2 x3 w1 xb ; t

polynomials.

For the strain eld, the small strains hypothesis is

adopted, thus giving

2eij V i;j V j;i

for i 6 j

e33 S 33 Rab eab S 33 r33 S 33 k33 H

The denitions of Qabvw, Rab, Qa3v3, S33, xab and k33 are

derived in Appendix A.

The virtual variation of the enthalpy H corresponds to

the virtual variation of the strain energy / due to the fact

that the temperature is imposed and known

dH d/ rab deab r33 de33 sa3 dca3

V a H 1ab S 0a H 2ab w0;b H 3ab S ha H 4ab wh;b C0ab s0b3 Chab shb3

V 3 L1 w0 L2 wh

orthotropic materials; a mixed form [22] of the stress-strain

relationship is used, with the thermal eect included (H is

the temperature variation with respect to a reference

value):

Using Eqs. (7) and (8) yields the following mixed form:

d/ Qabvw evw deab S 33 r33 dr33 Qa3v3 cv3 dca3 xab Hdeab

10

129

14

H iab

where

of the Hermite cubic polynomials and Clab have similar

properties [22]. The name Hermitian zig-zag plate

model has been chosen to remind these particular properties of the through-the-thickness shape functions.

2.4. Derivation of the transverse normal stress

We are going now to present some details of the derivation of r33 because in [22] the case without thermal eects

was considered and thermal eects are here for the rst

time taken into account. The transverse normal stress r33

is supposed constant through-the-thickness of the laminate

and its value is determined in such a way that the two possible transverse normal strains, namely e33 obtained by the

constitutive Eq. (8) and V3,3, are compatible [22]; Aitharaju and Averill [18,29] proposes a compatibility conditions

that derives from the Reissners mixed variational principle

hV 3;3 e33 i 0

15

i.e. to minimize the following integral quantity

hV 3;3 e33 2 i

There are essentially three conditions to be satised in

order to obtain the nal expression of the Hermitian zigzag plate model starting from Eqs. (1)(8).

16

hV 3;3 e33 S 33 i 0

17

11

written in a compact form

p1

Averills condition

hV 3;3 e33 pi 0

18

p S 33 Tesslers condition

Substituting Eq. (6) and the second of Eq. (8) into Eq. (18),

after some manipulations, we obtain

k

sb3 xx ; k z ; t k1 sb3 xx ; k1 z ; t; k 1; . . . ; N 1

sa3 xx ; 0; t

p0a xx ; t

sa3 xx ; h; t

pha xx ; t

p0a

12

33

pha

where

and

are the prescribed tangential loads distributed on bottom and top faces, respectively.

3. Use of the displacements and transverse shear stresses of

the external laminate faces as degrees of freedom of the

model

S 0a xx ; t V a xx ; 0; t

w0 xx ; t V 3 xx ; 0; t

s0a3 xx ; t 1 sa3 xx ; 0; t

S ha xx ; t V a xx ; h; t

wh xx ; t V 3 xx ; h; t

sha3 xx ; t N sa3 xx ; h; t

r33 p1cb S 0c;b p2cb w0;cb p3cb S hc;b p4cb wh;cb w0cb s0c3;b

13

19

where

hpS 33 Rab H iac i

;

hpS 33 i

hpS 33 Rab Cqac i

wqcb

;

hpS 33 i

hpi

D

hhpS 33 i

hpS 33 k33 Hi

rH

33

hpS 33 i

picb

i 1; 2; 3; 4

q 0; h

20

Eq. (19) shows that r33 is a function of the Hermitian degrees of freedom (13) and of temperature.

130

conditions

By means of the principle of virtual work, it is possible

to obtain the equations of motion and related boundary

conditions of the proposed theory

Z

hQabvw evw deab S 33 r33 dr33 Qa3v3 cv3 dca3 xab Hdeab i dX

X

Z

Z

hqV i dV i i dX p0a dS 0a

q0 dV 03

pha dS ha

X

h

q dV

h

3 dX

N Z

X

k1

k S

ri dV i dS p

21

tial loads applied on the bottom and top faces,

q0 and qh

the corresponding transversal loads and ri are the loads applied to the lateral cylindrical surface Sp (Fig. 1). We remind that the temperature distribution is known a priori

(Eq. (9)). See [22] for the details of the equations of motion

and boundary conditions for the case of mechanical loads.

In the present case the thermal resultant forces and moment should be added to the mechanical ones; we omit

their expressions for sake of brevity.

respectively, Fig. 2), the nite element approximation will

also be based on this separation. A subscript will denote

the number of the node to which the degree of freedom corresponds; a separate node numbering is used for the top

and bottom faces (Fig. 2).

3.1. FEM approximation: shape functions and degrees of

freedom

Let us discuss the mixed FEM approximation adopted

for the plate nite element; the generalised displacements

of the plate model are those dened in Eq. (13).

The 4 in-plane displacements S 01 ; S 02 ; S h1 and S h2 and the

4 transverse shear stresses s013 ; s023 ; sh13 and sh23 appear with

their rst derivatives in the strain energy expression thus

needing a C0 approximation. We adopt the parabolic 2D

shape functions used for the classical QUAD8 membrane

elements, thus obtaining, per each generalised displacements, 8 degrees of freedom, 4 placed at the corners and

4 placed at the centre of each side (black and white nodes

in Fig. 2, respectively). Considering, for example, the bottom face in-plane displacements and transverse shear stresses, we have the following FEM approximation (analogous

expressions hold for the top face)

Based on the theory briey summarized in Section 2, we

now formulate the relative plate nite element (Fig. 2). In

the context of nite element approximation a modied

notation will be used for sake of simplicity; u, v and w will

be the nodal displacements corresponding to the x1-, x2and x3-axis, respectively, while sxz and syz are the

transverse shear stresses. Considering that the degrees of

freedom of the model are located at the top and bottom

S 01 xa ; t

8

X

P i xa u0i t

i1

S 02 xa ; t

8

X

P i xa v0i t

i1

s013 xa ; t

8

X

22

P i xa s0xzi t

i1

s023 xa ; t

8

X

P i xa s0yzi t

i1

result, we have 32 FEM degrees of freedom per each face (4

per each node), for a total of 64 dofs (Fig. 2). There are

fundamentally two reasons for choosing the parabolic instead of the linear shape functions [23]. The rst reason is

that, also for simple cases of loads and boundary conditions, the distributions of in-plane displacements in x1

and x2 is not simply linear if evaluated at the top and bottom faces. The second reason is that, with an additive midnode per each side and per each face of the element, we can

describe interfacial damages (with the sub-laminate approach) with a high mesh density without using too many

elements.

A C1 approximation is required for the transverse displacements w0 and wh due to the presence of their second

derivatives in the strain energy expression; the corresponding rotational degrees of freedom w,x and w,y will appear in

the FEM approximation due to the adoption of the 2D

cubic Hermitian polynomials. As for the bottom face transverse displacement, we have

w0 xa ; t

4

X

fF H tg T fhtg

H i xa w0i t H xi xa w0;xi t

i1

H yi xa w0;yi t

23

where Hi, Hxi and Hyi are the Hermite cubic polynomials

and their derivatives. In this case we have 12 FEM degrees

of freedom per each face (3 per each corner node), for a total of 24 (Fig. 2). It is important to remind that the Hermite

polynomials, used as shape functions for the FEM approximation of transverse degrees of freedom along x1 and x2,

are not related with the Hermitian through-the-thickness

shape functions (Section 2.3).

The total number of FEM degrees of freedom per each

element is thus 64 + 24 = 88. The degrees of freedom vector will be denoted by {q(t)}.

3.2. Derivation of FEM equations of motion

Introducing the FEM approximations (Eqs. (22) and

(23) and similar ones for top face generalised displacements) into the expressions of displacements (14), strains

(7) and stresses (8) and (19) and these into the virtual work

principle expression (21), we nally obtain the equation of

motion of the plate nite element

Mf

qtg Kfqtg fF M tg fF H tg

24

{FM(t)} and {FH(t)} the mechanical and thermal equivalent

nodal forces, respectively. We omit the details of their derivation because they can be obtained via the classical FEM

procedure.

A few words have to be spent to describe the thermal

contribution to the load vector. This vector comes from

two terms in the virtual variation of the strain energy (21)

Z

hS 33 r33 dr33 xab Hdeab i dX

25

X

normal stress is limited to the term containing the temperature (Eqs. (19) and (20))

Z

hS 33 rH

33 dr33 xab Hdeab i dX

X

Z

hpS 33 k33 Hi

dr33 xab Hdeab dX

S 33

26

hpS 33 i

X

Without considering further details of the derivation, it can

be readily seen that {FH(t)} can be determined after integration, over the thickness and over the reference surface

of the element, of a function containing the temperature

variation H. In order to simplify this integration, we have

supposed to approximate the temperature function assuming that it is through-the-thickness linear and distributed in

terms of x1 and x2 as well as in-plane displacements and

transverse shear stresses (Eq. (22)). Thus, we have only to

know the 16 nodal values of temperature (8 nodes per each

element face) contained in the vector {h} and then it is possible to write

131

27

stacking sequence and mechanical properties but not on

the temperature.

3.3. The sub-laminates approach

It is important to spend a few words on the sub-laminates approach and on how it can be used to improve

the modelling capabilities of the plate nite element based

on the Hermitian zig-zag plate theory.

At rst, we focus our attention on the meaning of sublaminates; a sub-laminate is a group of adjacent layers

belonging to the whole laminate. We must not think to layers only in the classical physical sense, but also in a more

mathematical one; if needed, a physical layer may be

divided in two or more mathematical layers with the

same mechanical properties. As a consequence, a sub-laminate may correspond to the whole laminate thickness, to a

group of physical layers or to a group of mathematical ones.

The division of the whole thickness of a laminate in more

sub-laminates and the assumption that the Hermitian zigzag plate theory is valid in each of them, constitute an eective method to improve the theory performances without

the need for increasing the through-the-thickness polynomial order of the displacements. For example, the transverse displacement may be supposed piecewise linear

instead of simply linear and this may be obtained without

improving the underlying theory complexity but only placing its validity domain in more laminate sub-regions. From

the FEM point of view, the possibility to divide the thickness in some sub-domains appears to be a normal mesh procedure. Thus, when we are meshing a plate, we will not have

only to sub-divide the (x1, x2) domain, but also the thickness

dimension. As a nal consequence, the meshing procedure

is that of a 3D volume and the nite elements we are considering may be used as brick elements; and this is a simplication of the whole modelling work also from the

computational and programming point of view. Nevertheless, the dierence between a 3D nite element model and

the present 2D sub-laminates approach must be claried.

When using 3D nite elements for meshing a structure,

the shape functions used along the three axes are usually

the same and are part of the FEM approximation; the plate

nite element based on the Hermitian zig-zag model has traditional FEM shape functions along x1 and x2 axes while,

along x3, the shape functions are those of the underlying

theory, i.e. with no FEM approximation. These thickness

shape functions give to the model some higher-order capabilities (transverse shear stresses continuity, for example)

that cannot be fullled by traditional FEM shape functions.

Moreover, the use of more sub-laminates along the thickness does not improve at the same degree of accuracy the

evaluation of all quantities; as it will be shown below (Section 4), in-plane displacements, strains and stresses are

correctly evaluated also with one through-the-thickness

1

0.9

0.8

0.7

0.6

x3

with more sub-laminates. For this reason, there is the possibility to choose the number of elements along the thickness depending on the quantity one has to calculate.

When we assemble the nite elements, we must take care

to guarantee the required continuities of displacements and

transverse shear stresses along the thickness. Inside each

sub-laminate nite element the continuity is guaranteed

by the Hermitian theory itself (axial displacements are continuous except when an interlayer slip must be modelled and

transverse shear stresses are always continuous). When

passing from a sub-laminate to another, the continuities

may be imposed (when necessary, i.e. always for shear stresses and in the case of undamaged interfaces for displacements) using the traditional assembling procedure. The

importance of having placed the nodes and degrees of freedom of the theory and of the relative plate nite element on

the external top and bottom faces is thus evident. Moreover, the possibility to impose or not the continuity of some

quantities at the interface between adjacent sub-laminates

leads to a modelling capability that is wider than that

oered in each single sub-laminate by the Hermitian theory.

With a discontinuity in the transverse displacement it is possible to model fracture modes of the I type (opening modes).

The possibility to model discontinuities between sub-laminates in terms of axial displacements appears not to be an

improvement with respect to the Hermitian model, considering that also in each sub-laminate we have this modelling

capability; but it has already been shown [23] that, for

severe level of sliding type damages, the latter approach

may be erroneous while the sub-laminate one is successful.

0.5

0.4

0.3

0.2

HZZ1

HZZ2

0.1

Pagano

0

-1

V1(L,L/2,x3)

-4

x 10

1

HZZ1

0.9

HZZ2

HZZ4

0.8

Pagano

0.7

0.6

x3

132

0.5

0.4

0.3

0.2

4. Numerical results

0.1

assess the accuracy of the proposed model and of the developed plate mixed nite element.

We consider the case of a cross-ply square plate

(a = b = L), simply supported on the four edges and subjected to a transverse double-sine distributed load applied

on the top face

px

px

1

2

qh x1 ; x2 sin

sin

a

b

6.5

6.6

6.7

6.8

6.9

-4

x 10

1

0.9

0.8

0.7

0.6

0.5

0.4

0.3

0.2

HZZ1

HZZ2

0.1

Table 1

Material mechanical properties

V3(L/2,L/2,x3)

x3

listed in Table 1; the laminate is regular, the stacking sequence is (0/90/0/90) with total thickness 1 [mm]. The

side length-to-thickness ratio is 4. The problem has an exact solution [1] with which the solution obtained with the

present approach (Hermitian zig-zag model + sub-laminate

approach) will be compared: the problem has been solved

via the HZZ 2D nite element modelling only 1/4 of the

Pagano

0

E1 [MPa]

E2, E3 [MPa]

G23 [MPa]

m12, m13

m23

111,000

7900

3300

2000

0.33

0.49

10

11(L/2,L/2,x3)

15

x 10

-5

1

HZZ1

HZZ2

0.9

Pagano

0.8

0.7

0.6

x3

plate with 8 8 n elements (n being the number of sublaminates along the thickness). When 2 sub-laminates are

used, each of them consists of 2 layers. Figures showing

through-the-thickness distributions adopt the following

units: [mm] for the thickness coordinate and for displacements, [MPa] for stresses (strains are non-dimensional).

In Figs. 310 some through-the-thickness distributions

of displacements, strains and stresses are shown. Figs. 3,

5 and 8 demonstrate that also with one sub-laminate inplane quantities are well captured by the Hermitian model.

This is dierent from the case of transverse quantities (Figs.

4, 6, 7, 9 and 10) for which the use of a greater number of

thickness subdivisions is desiderable. Of particular interest

is that the procedure for calculating the transverse normal

stress r33 (constant along the thickness of each sub-laminate) leads to the determination of a sort of average value

per each sub-laminate (Fig. 9).

133

0.5

0.4

0.3

0.2

0.1

0

-10

-8

-6

-4

-2

0

11(L/2,L/2,x3)

1

HZZ1

0.9

1

HZZ1

HZZ2

0.8

HZZ2

0.9

HZZ4

HZZ4

Pagano

0.8

Pagano

0.7

0.7

0.6

x3

x3

0.6

0.5

0.5

0.4

0.4

0.3

0.3

0.2

0.2

0.1

0.1

-5

5

33(L/2,L/2,x3)

10

15

-5

x 10

0.1

0.2

0.3

0.4

0.5

0.6

33(L/2,L/2,x3)

0.7

0.8

0.9

Fig. 9. r33(L/2, L/2, x3) through-the-thickness distribution.

1

HZZ1

0.9

HZZ1

HZZ2

0.9

HZZ2

HZZ4

HZZ4

0.8

Pagano

0.7

0.7

0.6

0.6

0.5

x3

x3

0.8

0.5

0.4

0.4

0.3

0.3

0.2

0.2

0.1

0.1

0

-5.5

-5

-4.5

-4

-3.5

-3

-2.5

13(L,L/2,x3)

-2

-1.5

-1

-0.5

-4

x 10

Pagano

-1

-0.8

-0.6

13(L,L/2,x3)

-0.4

-0.2

134

5. Conclusions

The paper presents a FEM approach for the analysis of

multilayered composite plates subjected to mechanical

loads and temperature variations. The underlying theory

is a zig-zag one that guarantees the continuity of transverse

shear stresses (that can be also dierent from zero on top

and bottom faces due to the presence of tangential loads)

and takes into account transverse normal deformability.

The theory itself and the relative plate mixed nite element

have their degrees of freedom placed on the external faces;

the plate element may also be regarded as a 3D brick element for this characteristic but it maintains the same computational logic of a traditional 2D element. As a

consequence, it is possible to divide the whole laminate

thickness in some sub-domains each of them corresponding

to a nite element sub-laminate thus improving the accuracy of the whole approach. The sub-laminate methodology appears to be very eective in increasing the

modelling capabilities of any lamination theory, because:

1. the nal FEM model is more accurate than the underlying theory without the need for increasing the order of

expansion of displacements and stresses along the

thickness;

2. the number of sub-laminates may be chosen depending

on the needed degree of accuracy and on the quantity

(in-plane or transverse, displacement or stresses) one

has to analyse;

3. the division of the thickness in some sub-domains gives

a modelling freedom that is unknown to the classical

lamination theories and to their FEM implementations;

it is possible to simulate interlayer damages of dierent

kinds and to study plates with dierent thicknesses;

4. from a computational and programming point of view,

the sub-laminates approach may be simply faced as a

normal 3D FEM modelling.

The basic concepts and hypotheses of the theory and of

the related plate mixed nite element have been presented

and discussed. Some numerical results about a square laminate subjected to transverse distributed loads have been

shown to assess the accuracy of the FEM approach; the latter has proven to be eective in evaluating the static

response of thick plates to mechanical loads.

Further modelling capabilities that have to be tested are:

(i) application of tangential loads to the external faces of

the laminates, (ii) modelling of damaged plates (sliding

type delaminations), (iii) analysis of plates response to

temperature distributions.

Acknowledgements

(MIUR), that the authors gratefully acknowledge.

Appendix A. Constitutive equations

Let us suppose that any layer of the laminate is referred

to both the geometric coordinate system (x1, x2, x3) and

to its own material coordinate system x1 ; x2 ; x3 (axes

of orthotropy). It is supposed that x3 x3 . The superscript

* will indicate a quantity when referred to the material axes.

Moreover, we denote with superscript M the mechanical

component of the strain tensor and with H the thermal

one. For sake of simplicity, we will use a vector notation.

We know that [4]

8 9

>

> r11 >

>

>

>

>

>

> r22 >

>

>

>

>

>

=

< r >

33

M

fr g C fe g ()

>

s13 >

>

>

>

>

>

>

>

>

>

>

s

>

23 >

>

;

: >

s12

3 8 M 9

2

C 11 C 12 C 13

0

0

0

e11 >

>

>

>

>

7

6 C C C

M >

>

>

0

0

0 7 >

>

> e22 >

6 12

22

23

>

>

>

>

7

6

=

6 C 13 C 23 C 33

0

0

0 7 < eM

33

7

6

6

M

0

0 C 44

0

0 7

>

> c13 >

7 >

6 0

>

>

>

7 >

6

>

M >

4 0

0

0

0 C 55

0 5 >

c

>

> 23 >

>

;

: M >

0

0

0

0

0 C 66

c12

A:1

while, in geometric axes,

9

8

r11 >

>

>

>

>

>

>

>

> r22 >

>

>

>

>

>

=

<r >

33

M

frg C fe g ()

>

s13 >

>

>

>

>

>

>

>

>

>

>

s

>

>

23

>

>

;

:

s12

2

C 11 C 12 C 13

0

0

6C

0

0

6 12 C 22 C 23

6

6 C 13 C 23 C 33

0

0

6

6 0

0

0 C 44 C 45

6

6

4 0

0

0 C 45 C 55

C 16

C 36

A:2

where the relation between [C] and [C*] depends on h, the

angle between the x1 and x1 axes [4].

The mechanical strains are the dierence between the

total strains and the thermal ones; in material axes

of National Interest, titled Damage tolerance design of

stiened and sandwich composite panels for fuselages:

modelling, experimental testing and monitoring funded

C 26

3 8 M9

C 16

e11 >

>

>

>

>

>

7

>

>

>

C 26 7 >

eM

>

22 >

>

>

>

7 >

<

M

C 36 7

e33 =

7

M

0 7

>

> c13 >

7 >

>

>

>

7 >

>

M >

>

0 5 >

c

>

23 >

>

;

: M>

C 66

c12

A:3

where the thermal strains are related to temperature variation H = T T0 by means of the thermal expansion coecients {a*}

feH g fa gH

A:4

Qlm C lm

with

fa g a11

a22

a33

0 0

A:6

feH g fagH

A:7

and

fag b c2 a11 s2 a22 s2 a11 c2 a22 a33 0 0 2csa11 a22 cT

A:8

Substituting Eq. (A.6) into Eq. (A.2) we have that

frg Cfeg feH g Cfeg fkgH

A:9

where

fkg Cfag

A:10

k22

k33

k12 cT

A:11

moving the transverse normal stress r33 on the right side

and the transverse normal strain on the left one. In indicial

notation we thus obtain

rab Qabvw evw S 33 Rab r33 xab H

A:12

sa3 Qa3v3 cv3

where

2

Q1111

6Q

6 1211

Qabvw () 6

4 Q2111

2

Q11

6Q

6 16

6

4 Q16

Q2211

Q16

Q66

Q66

Q1112

Q1121

Q1212

Q1221

Q2112 Q2121

Q2212 Q2221

3

Q16 Q12

Q66 Q26 7

7

7

Q66 Q26 5

Q1122

A:14

References

with

fkg b k11

l; m 4; 5

i; j 1; 2; 6

A:5

0

Q1222 7

7

7

Q2122 5

Q2222

Q1313 Q1323

Q44 Q45

Qa3v3 ()

Q2313 Q2323

Q45 Q55

R11 R12

C 13 C 63

Rab

R21 R22

C 63 C 23

1

S 33

C 33

x11 x12

R11

k11 k12

xab

S 33 k33

x21 x22

k12 k22

R21

R12

R22

A:13

and

Qij C ij C i3 C j3 S 33 ;

135

and sandwich plates. J Compos Mater 1970;4:2034.

[2] Di Sciuva M, Gherlone M. The post-buckling behaviour of laminated

composite and sandwich shallow shells with damaged interfaces. In:

Proceedings of the IV international colloquium on computation of

shell and spatial structures, Chania (Crete), 2000, proceedings on CD.

[3] Di Sciuva M, Gherlone M, Librescu L. Implications of damaged

interfaces and of other non-classical eects on the load carrying

capacity of multilayered composite shallow shells. Int J Non-Linear

Mech 2002;37:85167.

[4] Reddy JN. Mechanics of laminated composite plates. New

York: CRC Press; 1997.

[5] Savoia M, Reddy JN. Three-dimensional thermal analysis of laminated composite plates. Int J Solids Struct 1995;32(5):593608.

[6] Reissner E. On a certain mixed variational theorem and a proposed

application. Int J Numer Meth Eng 1984;20:13668.

[7] Reissner E, Stavsky Y. Bending and stretching of certain types of

aelotropic elastic plates. J Appl Mech 1961;28:4028.

[8] Whitney JM, Pagano NJ. Shear deformation in heterogeneous

anisotropic plates. J Appl Mech 1970;37:10316.

[9] Reddy JN. A simple higher-order theory for laminated composite

plates. J Appl Mech 1984;51:74552.

[10] Reddy JN. A generalization of two-dimensional theories of laminated

composite plates. Commun Appl Numer Meth 1987;3(1):17380.

[11] Lu X, Liu D. An interlaminar shear stress continuity theory for both

thin and thick composite laminates. J Appl Mech 1992;59:5029.

[12] Di Sciuva M. Bending, vibration and buckling of simply supported

thick multilayered orthotropic plates: an evaluation of a new

displacement model. J Sound Vib 1986;105:42542.

[13] Di Sciuva M. An improved shear-deformation theory for moderately

thick multilayered anisotropic shells and plates. J Appl Mech

1987;54:58996.

[14] Di Sciuva M. Multilayered anisotropic plate models with continuous

interlaminar stresses. Compos Struct 1992;22(3):14967.

[15] Di Sciuva M. Geometrically non-linear theory of multilayered plates

with interlayer slips. AIAA J 1997;35(11):17539.

[16] Kant T, Swaminathan K. Estimation of transverse/interlaminar

stresses in laminated composites a selective review and survey of

current developments. Compos Struct 2000;49:6575.

[17] Kant T, Marur SR, Rao GS. Analytical solution to the dynamic

analysis of laminated beams using higher order rened theories.

Compos Struct 1998;40(1):19.

[18] Aitharaju VR, Averill RC. C0 zigzag kinematic displacement models

for the analysis of laminated composites. Mech Compos Mater Struct

1999;6:3156.

[19] Cook GM, Tessler A. A {3, 2}-order bending theory for laminated

composite and sandwich beams. Composites Part B 1998;29B:56576.

[20] Averill RC, Yip YC. Thick beam theory and nite element model

with zig-zag sublaminate approximations. AIAA J 1996;34(8):

162732.

[21] Di Sciuva M, Gherlone M. Finite element analysis of damaged

multilayered composite beams with transverse deformability. In:

Proceedings of the VI international conference on computational

structures technology, Prague (Czech Republic), 2002, proceedings on

CD, paper 129.

[22] Di Sciuva M, Gherlone M. A global/local third-order Hermitian

displacement eld with damaged interfaces and transverse extensibility: analytical formulation. Compos Struct 2003;4:41931.

[23] Di Sciuva M, Gherlone M. A global/local third-order Hermitian

displacement eld with damaged interfaces and transverse extensibility: FEM formulation. Compos Struct 2003;4:43344.

136

sandwich beams: a FEM approach based on the Hermitian zigzag model. In: Proceedings of the MDP-8 Cairo University

conference on mechanical design and production, Cairo (Egypt),

2004. p. 6518.

[25] Di Sciuva M, Gherlone M. Quasi-3D analysis of damaged multilayered composite laminates subjected to thermo-mechanical loads. In:

Proceedings of the V international congress on thermal stresses and

related topics, Blacksburg, Virginia (USA), 2003. p. TA-10-2-1TA10-2-4.

[26] Gherlone M, Di Sciuva M. A higher-order fem approach for the

analysis of damaged composite multilayered plates subjected to

thermal stresses, Wien (Austria), 2005. p. 2536.

[27] Gherlone M, Di Sciuva M. Thermo-mechanics of undamaged and

damaged multilayered composite plates: assessment of the FEM sublaminates approach. Compos Struct, inpress, doi:10.1016/j.compstruct.

2006.08.005.

[28] Icardi U. Higher-order zig-zag model for analysis of thick composite

beams with including transverse normal stress and sublaminates

approximation. Compos Part B: Eng 2001;32:34354.

[29] Aitharaju VR. A C0 zig-zag nite element for analysis of damaged

laminated and woven composites. PhD thesis, Department of

Materials Science and Mechanics, Michigan State University, 1997.

- 14ME770 – Finite Element AnalysisTransféré parsaravana pandian
- Strand7 Theoretical Manual Contents.pdfTransféré pardvegaucentral
- Finite Element MethodsTransféré parpraneeth4a
- Simulation Thermal Behaviour of Concrete by FEA a STAR Review (ICT Newsletter)Transféré pars80johnny
- tmpBBD0.tmpTransféré parFrontiers
- Finite Elelemt Methods (Update 1)Transféré parMazhar Ali
- FemTransféré parluffydmon
- 2 Marks Qiestions With AnswersTransféré parUbaid Bahussain
- 50494_prefTransféré parsmanoj354
- Smooth 1Transféré parLeong Wen Chun
- How to Find Errors in Finite-element ModelsTransféré parVenky Moka
- Surrey SE1M94 2001 PaperTransféré parphilipyap
- Finite Element Analysis of a Nose Gear During LandingTransféré parSezgin Bayrak
- Leteraure Review (Ben)Transféré pardurgam ramu
- 9789811055676-c2Transféré parPaul Turner
- Analysis of Reinforced Concrete (RC) Beams Using Nonlinear Finite ElementTransféré parSonia Faisal
- AE 8603 CMS SyllabusTransféré parRaajeshkrishna C.R
- [30]ENIEF2008_RieraTransféré parsacharya2011
- 6 Th Sem Syllabus AeroTransféré parJackson Jack
- Vulnerability Assessment of Buried Pipelines (1)Transféré pardiego
- An Axisymmetric Bending and Shear Stress Analysis of of Functionally Graded Circular Plate Based on Unconstrained Third Order Shear Deformation Theory via Differential Quadrature MethodTransféré paresatjournals
- KMSCFDWeb Assignment Module5Transféré parsadiksnm
- Advancing Ship Battle Damage Response by Integrating Distributed and Hull Systems Models.pptTransféré parFernando Raúl LADINO
- Finite Element Method Linear Rectangular Element for Solving Finite Nanowire Superlattice Quantum Dot StructuresTransféré partheijes
- fem termTransféré parAnkita Soni
- 6 Chapter 3Transféré parsju65
- Ch2 Variational PrincipleTransféré parabi
- Ch2_Variational_Principle.pdfTransféré parabi
- Concrete Construction Article PDF_ Welded Splices of Reinforcing Bars (1)Transféré parmrnaeem
- UMAT ExampleTransféré parAnonymous WAgzcf9

- BSNL Electronics Communication Engineering 20702Transféré parDeepak Sharma
- MATERIAL SCIENCE 79 IMPORTANT MCQ WWW.ALLEXAMREVIEW.COM.pdfTransféré parNandakumar Reddy
- Applied Heat and mass transferTransféré parabdullah1s
- JJ310 STRENGTH OF MATERIAL Chapter 4(a) Bending StressTransféré parAh Tiang
- 905 Uniformidad de DosisTransféré parmarcelo1315
- Thermodynamics PrinciplesTransféré parMuhammad Jafar
- NeverWet White Paper General 2013Transféré parKevin Ke
- 12-395-lTransféré parAndykaYayanSetiawan
- ElasticityTransféré parRisa M. Lim
- Menter Sst NasaTransféré partornapp
- Liquid membranes for gas vapor separations.pdfTransféré parwahyuthp43
- SoM Qus BankTransféré parReddy Babu
- Analytical Computation of Multiple Shrink Fitted ComponentsTransféré parParag Naik
- Solving Conformal Coating Problems in the Application ProcessTransféré parMael Azmi
- Counter Current ExchangeTransféré parZalina Razali
- All Unitized Reinfoced FemaleTransféré parwaweng22
- How to Prevent Weld FailureTransféré parKeith Rutherford
- Silva 2016Transféré parUmer Murtaza
- ndtTransféré parkoib789
- Mechanical Properties for Preliminary Design of Structures Made From PVC Coated FabricTransféré parFilthynotes
- AISC K FactorsTransféré parcklcon
- SESA6047 Exam 2013-2014_FINALTransféré parrrrrrrrrraaaaaaajjjj
- DryingTransféré parEiman Uzmi
- EHTC_10_Versailles_Feickert_Kirchhoff_final_präsentation_AutoFENA3D.pdfTransféré parAnonymous 1EQNc1
- Phy543-HW-combo (1).pdfTransféré parAbdul Jalil Mondal
- Thermal Shock Effect on the Glass Thermal Stress Response and Crack PrppagationTransféré parArmin Ber
- Dislocations and StrenghteningTransféré parAmber Williams
- Characterization Technique of NanofibersTransféré parInternational Journal for Scientific Research and Development - IJSRD
- Shell OpeningTransféré pard_maziero
- Mechanical Properties of MaterialsTransféré parcataice

## Bien plus que des documents.

Découvrez tout ce que Scribd a à offrir, dont les livres et les livres audio des principaux éditeurs.

Annulez à tout moment.