# Finite Deformation Plasticity for Composite Structures ... jfish/ Finite Deformation Plasticity for...

date post

23-May-2018Category

## Documents

view

214download

0

Embed Size (px)

### Transcript of Finite Deformation Plasticity for Composite Structures ... jfish/ Finite Deformation Plasticity for...

Finite Deformation Plasticity for Composite Structures: Computational Models and Adaptive Strategies

Jacob Fish and Kamlun ShekDepartments of Civil, Mechanical and Aerospace Engineering

Rensselaer Polytechnic InstituteTroy, NY 12180

Abstract

We develop computational models and adaptive modeling strategies for obtaining anapproximate solution to a boundary value problem describing the finite deformation plas-ticity of heterogeneous structures. A nearly optimal mathematical model consists of anaveraging scheme based on approximating eigenstrains and elastic concentration factors ineach micro phase by a constant in the portion of the macro-domain where modeling errorsare small, whereas elsewhere, a more detailed mathematical model based on a piecewiseconstant approximation of eigenstrains and elastic concentration factors is utilized. Themethodology is developed within the framework of statistically homogeneous compos-ite material and local periodicity assumptions.

1.0 Introduction

In this manuscript, we develop a theory and methodology for obtaining an approximatesolution to a boundary value problem describing the finite deformation plasticity of heter-ogeneous structures. The theory is developed within the framework of statistically homo-geneous composite material and local periodicity assumptions. For readers interested intheoretical and computational issues dealing with various aspects of nonperiodic heteroge-neous media we refer to [7][9][28][37].

The challenge of solving structural problems with accurate resolution of microstructuralfields undergoing inelastic deformation is enormous. This subject has been an active areaof research in the computational mechanics community for more than two decades.Numerous studies have dealt with the utilization of the finite element method [12][13][18][21][22][24][30][34], the boundary element method [11], the Voronoi cell method[10], the spectral method [1], the transformation field analysis [5], and the Fourier seriesexpansion technique [26] for solving PDEs arising from the homogenization of nonlinearcomposites. The primary goals of these studies were twofold: (i) develop macroscopicconstitutive equations that would enable solution of an auxiliary problem with nonlinearhomogenized (smooth) coefficients, and (ii) establish bounds for overall nonlinear proper-ties [2][29][32][33][34][35].

1

Attempts at solving large scale nonlinear structural systems with accurate resolution ofmicrostructural fields are very rare [10][12][26] and successes were reported for smallproblems and/or special cases. This is because for linear problems a unit cell or a represen-tative volume problem has to be solved only once, whereas for nonlinear history depen-dent systems, it has to be solved at every increment and for each macroscopic (Gauss)point. Furthermore, history data has to be updated at a number of integration points equalto the product of the number of Gauss points in the macro and micro (unit cell) domains.

To illustrate the computational complexity involved we consider an elasto-plastic analysisof the composite flap problem [8] with fibrous microstructure as shown in Figures 1 and 2.The structural problem is discretized with 788 tetrahedral elements (993 degrees of free-dom), whereas fibrous microstructure is discretized with 98 elements in the fiber domainand 253 elements in the matrix domain, totaling 330 degrees of freedom. The CPU timeon SPARC 10/51 workstation for this problem was over 7 hours, as opposed to 10 secondsif von Mises metal plasticity was used instead, which means that 99.9% of CPU time isspent on stress updates.

With the exception of [6][12][19] most of the research activities focused on small defor-mation inelastic response of microconstituents and their interfaces. This is partially justi-fied due to high stiffness and relatively low ductility of fibrous composite materials.However, when hardening is low and the stress measures are comparable to the inelastictangent modulus, or in the case of thin structures undergoing large rotations, large defor-mation formulation is required.

One of the objectives of the present manuscript is to extend the recent formulation of themathematical homogenization theory with eigenstrains developed by the authors in [8] toaccount for finite deformation and thermal effects. In addition, adaptive strategy is devisedto ensure reliability and efficiency of computations. In Section 2 we derive a closed formexpression relating arbitrary transformation fields to mechanical fields in the phases. InSections 3 and 4 we employ an additive decomposition of the rate of deformation intoelastic rate of deformation, governed by hypoelasticty and inelastic rate of deformation.Section 3 focuses on the 2-point approximation scheme (for two phase materials), whereeach point represents an average response within a phase. The local response within eachphase is then recovered by means of post-processing. In Section 4 we describe the n-pointscheme model, where n denotes the number of elements in the microstructure. Section 5 isdevoted to modeling error estimation and adaptive strategy. We develop an adaptive 2/n-point model, where the 2-point scheme is used in regions where modeling errors are small,whereas elsewhere the n-point scheme is employed. Numerical experiments conducted inSection 6 investigate the 2-point, the n-point, and the adaptive 2/n-point schemes in thecontext of finite deformation plasticity.

2

2.0 Mathematical Homogenization with Eigenstrains for SmallDeformations

In this section we generalize the classical mathematical homogenization theory [3][4] forheterogeneous media to account for eigenstrains. We regard all inelastic strains, phasetransformation and temperature effects as eigenstrains in an otherwise elastic body. Wewill derive closed form expressions relating arbitrary eigenstrains to mechanical fields in amulti-phase composite medium. In this section attention is restricted to small deforma-tions.

The microstructure of a composite material is assumed to be locally periodic (Y-periodic)with a period represented by a unit cell domain or a Representative Volume Element(RVE), denoted by , as shown in Figure 3. Let be a macroscopic coordinate vector in

macro domain and be a microscopic position vector in . For any Y-periodic

function , we have in which vector is the basic period of the

microstructure and is a 3 by 3 diagonal matrix with integer components. Adopting the

classical nomenclature, any Y-periodic function can be represented as

(1)

where superscript denotes a Y-periodic function . The indirect macroscopic spatial

derivatives of can be calculated by the chain rule as

(2)

(3)

where the comma followed by a subscript variable or denotes a partial derivative

with respect to the subscript variable (i.e. and ). A semi-colon

followed by a subscript variable denotes a partial derivative with respect to the remain-

ing x components (2), but a full derivative with respect to , and vice versa when a semi-

colon is followed by subscript variable (3). Summation convention for repeated right

hand side subscripts is employed, except for subscripts x and y.

We assume that micro-constituents possess homogeneous properties and satisfy equilib-rium, constitutive, kinematics and compatibility equations as well as jump conditions atthe interface between the micro-phases. The corresponding boundary value problem isgoverned by the following equations:

(4)

(5)

x y x

f f x y,( ) f x y ky+,( )= yk

f

f x( ) f x y x( ),( )

ff

f,xi x( ) f x; i x y,( ) f,xi x y,( )

1--- f,yi x y,( )+=

f y; i x y,( ) f,yi x y,( ) f,xi x y,( )+ f x; i x y,( )= =

xi yif,xi f xi f,yi f yi

xiyi

yi

ij x; j bi+ 0 in =

ij Lijkl kl kl( ) in =

3

(6)

(7)

(8)

(9)

where , and are components of stress, strain and rotation tensors; and

are components of elastic stiffness and eigenstrain tensors, respectively; is a body force

assumed to be independent of ; denotes the components of the displacement vector;

the subscript pairs with regular and square parenthesizes denote the symmetric and anti-symmetric gradients defined as

(10)

denotes the macroscopic domain of interest with boundary ; and are boundary

portions where displacements and tractions are prescribed, respectively, such that

and ; denotes the normal vector on . We assume that the

interface between the phases is perfectly bonded, i.e. and at the

interface, , where is the normal vector to and is a jump operator.

In the following, displacements and eigenstrains are

approximated in terms of double scale asymptotic expansions on :

(11)

(12)

Strain and rotation expansions on can be obtained by substituting (11) into (6) and(7) with consideration of the indirect differentiation rule (2)

(13)

(14)

where strain and rotation components for various orders of are given as

ij u i x; j( ) in =

ij u i x; j[ ] in =

ui ui on u=

ij nj ti on t=

ij ij ij Lijkl ij

bi

y ui

u i x; j( ) 1

2--- ui x; j

uj x; i+( ) , u i x; j[ ]

12--- ui x; j

uj x; i( )

u tui ti

u t = u t= ni

ij nj[ ] 0= ui[ ] 0=

int ni int [ ]

ui x( ) ui x y,( )= ij x( ) ij x y,( )=

ui x y,( ) ui0 x y,( ) ui1 x y,( ) + +

ij x y,( ) ij0 x y,( ) ij1 x y,( ) + +

ij x y,( )1---ij1 x y,( ) ij0 x y,( ) ij1 x y,( ) + + +

ij x y,( )1---ij1 x y,( ) ij0 x y,( ) ij1 x y,( ) + + +

4

(15)

(16)

Recommended

*View more*