Abstract:
Damage models are developed within the continuum damage mechanics framework
which allows the description of material degeneration with general constitutive
equations. The difficulty in the description of damage behaviour increases with
increasing complexity of the material behaviour. This is especially true when it comes
to composite materials which have an orthotropic material behaviour. The conventional
description of damage, i.e. the local continuum damage mechanics description, leads
to strain-softening behaviour which is characterised by a decline in stress with
simultaneously increasing strain.
Due to strain-softening the tangent stiffness becomes negative which forces the wave
speed to become imaginary in dynamic problems. Consequently the partial differential
equations governing the dynamic problem change from hyperbolic to elliptic and,
therefore, the initial boundary value problem no longer has a unique solution. Due to
this the physical meaning becomes unrealistic. Strain-softening is limited to an infinitely
small area in which waves are not able to propagate in a process called wave trapping.
A displacement discontinuity in an area of width zero (localisation zone) develops. The
strain becomes infinite in this zone and is accompanied with a zero stress. Areas
outside the softening zone are not able to interact with the strain-softening domain. As
a consequence the strain-softening domain acts similar to a free boundary at which
waves reflect. The implementation of local continua with strain-softening behaviour in
finite element codes leads to additional numerical problems. Strain-softening behaviour
manifests itself in the smallest area possible which is a single point in analytical
considerations. This area is defined by the element discretisation in finite element
codes. Therefore, strain-softening leads to a pronounced mesh sensitivity of results in
addition to mathematical and physical issues.
This work aims to find a solution which removes problems associated to strain-
softening. Its aim is to represent material behaviour due to damage realistically and
enable numerical results to convergence to a unique solution.
The strain-softening problem is the focus of this work. It was investigated using a 1D
wave propagation problem described by Bažant and Belytschko [1]. This simple
experiment allows for an easy comparison of analytical and numerical results and
therefore gives an insight into the problems connected to strain-softening.
Furthermore, regularisation methods, specifically nonlocal and viscous methods, were
investigated. Regularisation methods add additional terms to constitutive equations
which keep the initial boundary value problem well-posed and enable a unique solution
independent of the element discretisation. It was found that these methods are indeed
capable of regulating the softening problem; however, they add additional difficulties in
the description of material behaviour.
A new approach to the strain-softening issues, unique at this point of time, was
developed in this work which implements damage as an equivalent damage force. This
approach is able to keep the initial boundary value problem stable and converge to a
unique solution without adding additional terms in the constitutive equations, such as
regularisation methods. This new approach to strain-softening was implemented for an
isotropic material with scalar damage variable in DYNA3D successfully. Numerical
results converged to a unique solution and were physically reasonable. The concept of
an equivalent damage force was further developed to orthotropic material behaviour.
This made an advanced representation, using an 8th rank damage tensor, necessary.
The 8th rank damage tensor is able to represent anisotropic damage and it is also the
most general damage representation possible.