Theory and numerics of geometrically non-linear gradient plasticity
Introduction
The objectives of the present work are the theory and in particular the appropriate numerics of a thermodynamically consistent geometrically non-linear gradient plasticity formulation. In particular we emphasize the physical motivation of higher gradients in terms of geometrically necessary dislocation densities. To this end we develop a general continuum mechanical framework of a dislocation based gradient model within multiplicative plasticity. In a simplified model of a single crystal with only one active slip system we later adopt the internal hardening variable as a substitute for the geometrically necessary dislocation density. We then derive and apply the appropriate numerics for the proposed prototype model.
Recently much attention has been focused on the physical motivation of the incorporation of higher gradients in plasticity. In addition to the meanwhile well-established regularizing effect w.r.t. discretization dependencies in standard numerical computations of localized plastic deformations in softening materials, higher gradients have been motivated by dislocations in crystal plasticity, see e.g. [1] and for an alternative framework [2]. In particular, higher gradients can be physically justified for single and polycrystalline material like metals, if one considers the dislocation density and incompatibility, respectively, see e.g. [3]. An alternative recent approach by Gurtin [4] deals as well with geometrically necessary dislocations, which are incorporated in a gradient theory of single crystal plasticity.
Several experimental evidences for microstructural interaction w.r.t. plastic deformation phenomena can be found in the literature, e.g. the Hall–Petch effect [5], [6], dislocation related hardening effects, see [7] or grain size effects in the works by Ashby [8]. In particular special dedication to various profound aspects of dislocations w.r.t. crystallographic characteristics can be found in an article by Amelinckx [9]. Furthermore torsion tests on copper wires were investigated by Fleck et al. [10], thereby scale effects motivate a particular strain gradient plasticity formulation as advocated by Fleck and Hutchinson [11].
Whenever the influence of material substructure cannot be neglected multifield theories enter the stage, see e.g. [12] and for a recent contribution [13]. Thereby one has to distinguish on the one hand theories that introduce extra observable fields resulting in couple stresses satisfying additional balance equations, e.g. the micropolar theory of the Cosserat brothers [14]. On the other hand, the introduction of internal variables and their gradients reflects the microstructural response. For a comprehensive study of the thermodynamics with internal variables we refer to Maugin [15]. Moreover in order to include nonlocal effects, apart from the here advocated particular gradient theory nonlocal integral models are a possibility, see e.g. [16].
For gradient continua, a variety of numerical strategies, different from the one proposed in this work, were investigated e.g. by Sluys et al. [17], Pamin [18], de Borst and Pamin [19], Peerlings et al. [20], [21], Mikkelsen [22], Steinmann [23], Comi [24], Chambon et al. [25], Svedberg and Runesson [26], Svedberg [27] and Nedjar [28].
For the local continuum description an early attempt to set up a mixed finite element formulation was provided by Pinsky [29], whereby a plastic strain-like variable is discretized in addition to the displacement field. An alternative proposal based on a complementary mixed finite element formulation is due to Simo et al. [30], wherein the flow rule is enforced in a weak sense at the element level. Likewise, a two-field finite element formulation for elasticity coupled to damage was proposed by Florez-Lopez et al. [31].
Our own recent attempts to the numerics of gradient plasticity at small strains are documented in [32].
Section snippets
Dislocation based gradient model
In the sequel we briefly reiterate some essentials of the continuum theory of dislocations within the setting of multiplicative elastoplasticity. Thereby, the main goal is the physical motivation for the incorporation of higher gradients of the plastic deformation gradient in the free energy and consequently in the yield condition.
Strong and weak form of the coupled problem
To set the stage for the following implementation for the simplified gradient model within a finite element environment we first summarize the pertinent set of equations for the solution of the coupled problem for the primary variables and κ in strong and weak form, see Table 1.
Firstly, neglecting inertia, the equilibrium subproblem is given by the material balance of linear momentum and the corresponding Neumann boundary conditions , which are tested by a virtual
Discretization in time and space of the coupled problem
The above set of equations has to be discretized in time, whereby we apply without loss of generality the implicit Euler backward method. Here one should note that in general the plastic incompressibility constraint , with SL(3) the unimodular group, is violated, for remedies see e.g. [44]. Then the time integration of the primary variables and κ renders a discretized time update for the values and κn+1. Finally the algorithmic set of equations has to be discretized in space. To
Active set search
The initially unknown decomposition of the discretization node point set into active and inactive subsets at time step tn+1 is determined iteratively by an active set search. Thereby, the strategy is borrowed from convex non-linear programming.
To this end, we first initialize the active working set at the start of the iteration by those nodes which currently violate the constraint RΦK⩽0
Then the trial iterate is computed from a global Newton–Raphson step which is
Monolithic iterative solution
An efficient algorithm for the solution of the coupled problem stated in the above sections is offered by a monolithic iterative strategy. Here, the discrete algorithmic balance of linear momentum together with the discrete algorithmic Kuhn–Tucker conditions are solved simultaneously. Then, for a given active working set a typical Newton–Raphson step reads as followswhereby the linearized residua are expressed by the corresponding iteration matrices, which
Numerical examples
With the above algorithm at hand we now investigate the performance of our gradient formulation. Firstly, a 1-D model problem of a bar under uniaxial tension as frequently used in literature is analyzed. Secondly, a 2-D model problem of a square panel under uniaxial tension is considered. Thereby, localization is triggered by imperfections and the influence of different discretization densities is investigated in comparison to the local formulation.
Conclusions
We have derived a thermodynamically consistent formulation and the corresponding discretized algorithmic format of a geometrically non-linear gradient elastoplastic model. Based on the relation between the dislocation density tensor and the plastic deformation gradient we firstly adopted a physically motivated gradient extension of the classical approach towards multiplicative elastoplasticity. To this end we essentially incorporated the dislocation density into the free Helmholtz energy.
References (45)
On the role of gradients in the localization of deformation and fracture
Int. J. Eng. Sci.
(1992)Views on multiplicative elastoplasticity and the continuum theory of dislocations
Int. J. Eng. Sci.
(1996)- et al.
On the continuum formulation of higher gradient plasticity for single and polycrystals
J. Mech. Phys. Solids
(2000)Erratum
(2001) The deformation and aging of mild steel
Proc. Phys. Soc. B
(1951)The deformation of plastically non-homogeneous materials
Phil. Mag.
(1970)- et al.
Strain gradient plasticity
Adv. Appl. Mech.
(1997) Continua with Microstructure
(1989)- et al.
Sur la theorie des corps deformables
(1909) - J. Pamin, Gradient-dependent plasticity in numerical simulation of localization phenomena, Ph.D. dissertation, Delft...
Formulation and computation of geometrically non-linear gradient damage
Int. J. Numer. Meth. Eng.
(1999)