Research PaperRegularization of continuum damage mechanics models for 3-D brittle materials using implicit gradient enhancement
Introduction
It is widely acknowledged in the literature that local continuum damage models lead to localization issues [32], [23]. This means that the predictions of the finite element simulations change significantly when changing the size and orientation of the mesh used for discretization. The growth of damage tends to localize in the smallest band that can be captured by the spatial discretization resulting in damage localization at the smallest length scale set by the length of a single element in the finite element mesh. Consequently, the solutions for a case with a progressively damaged zone converge to a solution with a localization zone of zero volume when successive mesh refinements are applied. In the end, the numerical prediction fails to converge to a physically meaningful result. Therefore, the global response of the system shows a strong dependence on the spatial discretization. The causes and effects of localization have been discussed in detail in Borst et al. [4].
The introduction of non-locality in the constitutive model is one of the popular methods to regularize the localization of damage in order to eliminate mesh dependence. In a non-local damage formulation, the state of stress (or strain) at a point does not only depend on the state of strain and its history at that particular point but also depends on the state of the media in a finite neighbourhood of that point. There are two families of non-local formulations, namely, on the one hand, integral formulations using a kernel to regularize the stress or strain [1], [25], [2] and on the other hand, gradient-based formulations [8], [5], [27], [42], [21]. Non-local approaches introduce a length scale to the mathematical representation of the physical system and thus prevent the localization problem. The concept of non-locality for elastic deformations was introduced by Kröner [20] and later extended to hardening plasticity by Mühlhaus and Alfantis [27]. A breakage diffusion model for strength softening rock is studied by Mühlhaus and Gross [28]. Lasry et al. [22] introduced the concept of a non-local constitutive model as a localization limiter for a strain-softening material. These localization limiters [3], [19] force the damage to grow in a zone with a finite width that is independent of the spatial finite element discretization.
There are two forms of non-local gradient formulation namely explicit and implicit, where the non-local equivalent strain is related to the local equivalent strain through partial differential equations [44]. Explicit gradient enhancement is be considered as weakly non-local only, as it fails to regularize the stress (or strain) under some circumstances [55]. On the other hand, implicit gradient enhancement has the advantage of being strongly non-local. It can be shown that they are largely equivalent to the integral type [43], see also in Section 2.
The idea of implicit gradient enhancement is to introduce a second differential equation in addition to the quasi-static equilibrium equation [43]. The equation takes the form of a Helmholtz equation to calculate the non-local field as the counterpart of a local equivalent strain or another local internal variable. The damage evolution law now depends on the regularized equivalent strain rather than the local version. The regularized equivalent strain is obtained by solving the Helmholtz equation. Consequently, the momentum and Helmholtz equations form a non-linear, coupled system of partial differential equations (PDEs). We use the split-operator method [13], [15] and small time steps to solve the two equations sequentially by alternating between them while increasing loading. Section 2.4 details the equations solved using the split-operator approach. The split-operator algorithm is a robust scheme commonly used in implementations of gradient damage [26] and phase-field models [29]. The method is also used in 2-D nonlinear elastic brittle damage by Pires-Dominuges et al. [46] and for 3-D non-local microplane damage models by Zreid et al. [55].
In this paper we extend this implementation to the general 3-D case without any assumption on damage pattern. Our implementation support massive parallel computation which is crucial for the solution of problems with very large finite element meshes required to resolve the targeted spatial length scales of the damage zone. The implementation is applied to the case of a 3-D rectangular rock specimens with a pre-existing surface flaw under uniaxial compression. Local damage models have been investigated to simulate this experimental setup using unstructured meshes in 3-D where results have extensively been compared with a rich set of experimental observations [31]. In order to be able to represent the flaw the finite element method with an unstructured mesh is applied. The rock heterogeneity is modeled through random sampling at a fixed grid resolution which is then interpolated to the simulation meshes with varying mesh sizes. The allows an investigation on mesh dependence for the damage model results without changing material properties for meshes with different element sizes. To our knowledge, this paper presents the first non-local damage model applied to this experimental set-up.
The paper is organized as follows: In Section 2 we present the ingredients of the non-local model including the governing equations and the damage model based on the implicit gradient approach and the provision of heterogeneous material properties. Section 2.4 introduces the operator splitting and its implementation. Results of simulation of the damage propagation for three different application cases are presented in Section 3. The effect of localization length scale is discussed in Section 4. In Section 5 the mesh sensitivity analysis for all three application cases is discussed. An evaluation of the results is given in Section 6.
Section snippets
Methodology
We consider a rock specimen which is incrementally loaded until failure. Fig. 1 shows the geometry of the three application cases investigated later in this paper, see Section 3. In each loading step the new displacement vector is calculated and the resulting damage to the rock based on the new strain tensor is determined. Material properties are updated according to the damage model before the next loading step is applied. This iterative process has been implemented using
Damage pattern of local and non-local model
We apply the non-local damage code to three specimens with different shapes: a cubic specimen with a weak zone in Section 3.1, a homogeneous L-shaped specimen in Section 3.2 and a rectangular specimen with a pre-existing weak flaw in Section 3.3. Materials parameters; in particular Young’s modulus; are chosen according to the values published in the relevant experimental studies [24], [50]. For the three cases studied in this paper namely a cubic, L-Shaped and rectangular (with flaw) shaped
Effect of localization length scale
In the non-local model, the fracture width depends on the value of the non-local parameter c which is proportional to the square of localization length l as defined in (19). So the localization length is crucial for controlling the fracture width as the damage diffuses at the scale of the localization length. It also places limits on the maximum mesh element size that can be used so that the fracture width is resolved accurately. We make the assumption that we need at least two elements inside
Mesh sensitivity analysis
The aim of this section is to analyze the effect of the mesh size on the results for the non-local damage model and to show that the non-local model is, in fact, mesh independent. For comparison we also plot the damage plots for the local damage model as well, to show the problem of mesh dependency arising in local models. To control the computational cost and execution time for the simulations a relatively large localization length of mm for the non-local model has been chosen. The
Conclusions
Using the implicit gradient damage formulation a local damage model is modified with the objective to eliminate mesh dependence. The idea is to introduce regularization of the equivalent strain to eliminate mesh dependence and strain localization for the strain-softening damage model. This non-local version is constructed by solving an additional PDE – the Helmholtz equation. In contrast to the local damage model based on a dual damage initiation condition as investigated previously by Mondal
CRediT authorship contribution statement
S. Mondal: Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Project administration, Validation, Visualization, Writing - original draft. L.M. Olsen-Kettle: Conceptualization, Funding acquisition, Software, Investigation, Supervision, Writing - review & editing. L. Gross: Conceptualization, Resources, Software, Investigation, Supervision, Writing - review & editing.
Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Acknowledgements
This research is supported by the Australian Research Council Discovery Early Career Researcher Award DE140101398. The first author is grateful for PhD scholarship support, the tuition fee award and research assistance from this award, The University of Queensland and the Swinburne University of Technology, Australia respectively. The authors also gratefully acknowledge the software and high performance computing (HPC) support from The University of Queensland. This work uses infrastructure
References (55)
- et al.
Nonlocal microplane model for fracture, damage, and size effect in structures
J Eng Mech
(1990) - et al.
Nonlocal continuum damage, localization instability and convergence
J Appl Mech Trans ASME
(1988) - et al.
A study of localization limiters for strain-softening in statics and dynamics
Comput Struct
(1989) - et al.
Fundamental issues in finite element analyses of localization of deformation
Eng Comput
(1993) - et al.
On gradient-enhanced damage and plasticity models for failure in quasi-brittle and frictional materials
Comput Mech
(1995) - et al.
Microplane model m7 for plain concrete. ii: Calibration and verification
J Eng Mech
(2013) - et al.
Strong discontinuity approaches: An algorithm for robust performance and comparative assessment of accuracy
Int J Solids Struct
(2016) Computational modelling of gradient-enhanced damage in quasi-brittle materials
Mech Cohesive-frict Mater
(1999)- et al.
Electrical Resistivity Tomography using a finite element based BFGS algorithm with algebraic multigrid preconditioning
Geophys J Int
(2017) - et al.
Comparison of nonlocal approaches in continuum damage mechanics
Comput Struct
(1995)
Shape and size effects on the compressive strength of high-strength concrete
Cem Concr Res
Nonlocal implicit gradient-enhanced elasto-plasticity for the modelling of softening behaviour
Int J Plast
Experimental analysis and computational modelling of damage and fracture
Strongly non-local gradient-enhanced finite strain elastoplasticity
Int J Numer Meth Eng
Gmsh: A 3-d finite element mesh generator with built-in pre- and post-processing facilities
Int J Numer Meth Eng
Inversion of potential field data using the finite element method on parallel computers
Comput Geosci
Interface modeling in incompressible media using level sets in escript
Phys Earth Planet Inter
Nonlocal models for damage and fracture: Comparison of approaches
Int J Solids Struct
Elasticity theory of materials with long range cohesive forces
Int J Solids Struct
An anisotropic gradient damage model for quasi-brittle materials
Comput Methods Appl Mech Eng
Localization limiters in transient problems
Int J Solids Struct
Mesh dependence and stress singularity in local approach to creep-crack growth analysis
Trans Jpn Soc Mech Eng Ser A
Uniaxial strength and failure in sandstone containing a pre-existing 3-d surface flaw
Int J Fract
A symmetric over-nonlocal microplane model m4 for fracture in concrete
Int J Solids Struct
An integrated continuous–discontinuous approach towards damage engineering in sheet metal forming processes
Eng Fract Mech
A variational principle for gradient plasticity
Int J Solids Struct
Cited by (15)
Simulation of the dynamic cracking of brittle materials using a nonlocal damage model with an effective strain rate effect
2024, Computer Methods in Applied Mechanics and EngineeringA 2D/3D implicit gradient-enhanced nonlocal meso-scale damage model for deformation and fracturing of brittle materials
2023, Engineering Analysis with Boundary ElementsDamage delocalisation in skeletal muscle-inspired hierarchical armoured structures for impact protection
2022, Composite StructuresCitation Excerpt :Section 2.1.1 outlines the governing equations, impact boundary conditions and finite element method we employ to simulate impact loading conditions. In Section 2.1.2 we describe the damage model we employ following the approach of Mondal et al. [15–17] to model the three different materials present (alumina, Dyneema® and bone). Section 3 shows the results of our numerical simulations and how the structural hierarchy introduced can increase the protective capability of the ceramic layer.
Sensitivity of the damage response and fracture path to material heterogeneity present in a sandstone specimen containing a pre-existing 3-D surface flaw under uniaxial loading
2020, Computers and GeotechnicsCitation Excerpt :As a study case, we use a sandstone specimen with a single pre-existing flaw under uniaxial compression. This case has been investigated for various geometrical configurations in both experimental (Lu et al., 2015; Yang and Jing, 2011; Lee and Jeon, 2011) and numerical studies (Zhang et al., 2018; Pakzad et al., 2018; Mondal, 2019; Mondal et al., 2019, 2020) and hence sets a well understood but not oversimplified application case. We will explore to what extent the rock damage and strength are influenced by realizations of a randomized initial stiffness distribution of the rock and if and how the corresponding variation in damage is sensitive to geometrical flaw parameters such as depth, angle and width.