A Mixed Finite Element Method for Elasticity Problem

This paper describes a numerical solution for plane elasticity problem. It includes algorithms for discretization by mixed finite element methods. The discrete scheme allows the utilization of Brezzi Douglas Marini element (BDM1) for the stress tensor and piecewise constant elements for the displacement. The numerical results are compared with some previously published works or with others coming from commercial code like ABAQUS. keywords—Elasticity problem; Mixed Finite element method; BDM1 approximation; ABAQUS


INTRODUCTION
Mixed finite element methods for linear elasticity are based on approximations of a stress-displacement system derived from the Hellinger-Reissner variational principle [7], in which both displacements and stresses were approximated simultaneously.
The mathematical analysis and applications of mixed finite element methods have been widely developed since the seventies.A general analysis for this kind of methods was first developed by Brezzi [8].We also have to mention the papers by Babuska [14] and by Crouzeix and Raviart [15] which, although for particular problems, introduced some of the fundamental ideas for the analysis of mixed methods.
Many mixed finite element methods have been developed for plane elasticity, and generally speaking, they can be grouped into two categories: methods that enforce the symmetry of the stress weakly, and methods that enforce the symmetry exactly (strongly).In the former category, the stress tensor is not necessarily symmetric, but rather orthogonal to anti-symmetric tensors up to certain moments.Weakly imposed stress symmetry methods also introduce a new variable into the formulation that approximates the anti-symmetric part of the gradient of u; see for example [2][3].On other hand, exactly symmetric stress methods have been much more difficult to construct.The first class of inf_sup stable methods was the so called composite elements [4] [5].
Section II presents the model problem used in this paper.The discretization by mixed finite elements described is in section III.Numerical experiments carried out within the framework of this publication and their comparisons with other results are shown in section IV.

II. GOVERNING EQUATIONS
The equilibrium equations and boundary conditions are Where n is the unit outward normal.In the above, σ is the Cauchy stress, and f is the body force per unit volume.
The constitutive relation is given by Hooke's law: (5) Where C is the Hooke tensor, C is assumed here to have constant coefficients.Its inverse (compliance tensor) will be denoted by E. Hence (6) We consider small strains and displacements.The kinematics equations therefore consist of the straindisplacement relation (7) Where is the symmetric part of the gradient operator, and the boundary condition (8) www.ijacsa.thesai.org (10) (11) Then the standard weak formulation of the equilibrium equations is the following: Find and such that: One can see that ( 12)-( 13) practically coincide with the variational formulation of the Hellinger-Reissner principle.The use of this principle in the framework of finite elements can be traced Back to the pioneering work of Herrmann [9] and Hellan [10].The interest in using the stress field σ as an independent variable is questionable in as simple a case as the present one, but it is clear in more general and more complicated problems involving nonlinearities, plasticity, and so on.
Let the bilinear forms a and b, and the linear forms l and s such that: The underlying weak formulation ( 12)-( 13) may be restated as:

Let
, be a family of rectangulations of Ω.The edges of elements will be denoted (i=1, 2, 3 or i=1, 2, 3, 4) in the two-dimensional case.Let us deal first with the abstract framework (23)-( 24).Assume that we are www.ijacsa.thesai.orggiven two sequences and of subspaces E and Ψ, respectively.

We set . (27)
We have the following approximation theorem THEOREM 2. Assume that Then for every l 1 E′ and l 2 Ψ′, and for every h > 0, the discrete problem (30) (31) Has a unique solution.Moreover, there exists a constant such that (32) The dependence of on and can be easy traced [8].Clearly if (21) and ( 22) hold with constants and independent of h, then (32) holds with a constant independent of h.We define in general, for m integer ≥ 0, We are now ready for the error estimates.THEOREM 3. If (σ, u) is the solution of ( 12)-( 13) and (σ h , u h ) is the solution of (30)-(31), there exist a constant C > 0 such that: (37) Discretization of the mixed formulations, for linear elliptic operators, many examples of successful discretization of ( 12)-( 13) are known.The first ones were introduced by Raviart and Thomas in [11] and then reelaborated and extended to more general cases by Nedelec [12].Other families of possible discretization were introduced years later by Brezzi, Douglas, and Marini [1] [13].
To give a more precise definition of our mixed finite element approximation we shall need a few definitions.Let us define on an element K. P k : the space of polynomials of degree ≤ k.
We shall also need polynomial spaces on the edges of the elements . (38) In the two-dimensional, for the triangular elements we have

Restricting
to have a normal trace in yields a space larger than , but having essentially the same properties, that we denote The dimension of is thus dim = (42) For the triangular case we thus have the following inclusions between the spaces just defined (43) We consider the space obtained basically from the space of Brezzi-Douglas-Marini.The discrete scheme allows the utilization of BDM 1 for the stress tensor and piecewise constant elements for the displacement.
A mixed finite element approximation of ( 12)-( 13) is defined by Find and such that (49) .( 50) We obtain a system of linear equations (51)

Where
The matrix associated for the system (51) is symmetric indefinite.We use the iterative methods Minimum Residual Method (MINRES) for solving the symmetric system.

IV. NUMERICAL SIMULATIONS Example 1. Circular Void in a Finite Plate
Here a void of radius 0.3 is placed in the center of a plate of size 3 × 3 which is subjected to a unit stress in the y-direction.
The stress plot for σ yy s is in excellent agreement with the expected 3σ stress concentration at the edges of the hole.V. CONCLUSION We were interested in this work in the numeric solution for equilibrium equations.It includes algorithms for discretization by mixed finite element methods.The discrete scheme allows the utilization of BDM 1 for the stress tensor and piecewise constant elements for the displacement.Our results agree with ABAQUS.
Numerical results are presented to see the performance of the method, and seem to be interesting by comparing them with other recent results.

Fig. 1 .
Fig. 1.Body With Internal Boundary Subjected To Loads.We set

REMARK 1 .
THEOREM 1.Let E and Ψ be real Hilbert spaces, a bilinear form on E × E, and a bilinear form an E × Ψ. Set (20) And assume that: (21) (22) Then for every l 1 E′ and l 2 Ψ′ there exist a unique solution of the problem If problem (21)-(22) has a unique solution for every and then (20) holds and the bilinear form a(ξ 1 , ξ 2 ) restricted to K, is nonsingular (in the sense that it induces an isomorphism from K onto K').Clearly if one assumes that a (ξ 1 , ξ 2 ) is symmetric and positive semi definite, then (21) and (22) are necessary and sufficient for the existence and uniqueness of the solution of (23)-(24).REMARK 2. It is clear that if a (ξ 1 , ξ 2 ) is symmetric; the solution of (23)-(24) minimizes the functional (25) On the subspace of E, (26) And the formulation (23)-(24) corresponds to the introduction in (25)-(26) of the Lagrange multiplier .
derivatives being taken in the sense of distributions.On this space, we shall use the semi-norm www.ijacsa.thesai.org

Fig. 6 .
Fig. 6.Curve Of The Displacement Ux And Uy Along Hole In A Finite Plate

Fig. 10 .
Fig. 10.Curve Of Displacement Ux And Curve Of Displacement Uy Along Inclusion In A Finite Plate.