Conference Agenda

Concrete exhibits complex fracture behaviors, making accurate simulation and analysis essential for understanding material damage mechanisms. This study presents a numerical approach utilizing the novel Scaled Boundary Finite Element Method (SBFEM) to simulate the damage in concrete. The SBFEM is a semi-analytical method capable of using polyhedral elements with arbitrary number of nodes, edges and faces, which is highly complementary with efficient octree mesh generation algorithm. Therefore, octree meshes can be generated directly from the digital scan images of concrete samples, which captures material heterogeneity with high fidelity. Furthermore, the computational demands of damage simulation, especially in multi-scale problems, are addressed by leveraging modern High-Performance-Computing (HPC) techniques. The problem domain is partitioned into a number of parts with similar size and distributed to multiple computational units. The partition scheme is designed to minimize the data communication between parts, therefore maximizing the computational efficiency. By combining the SBFEM with HPC, this approach is particularly well-suited for simulating concrete’s fracture processes, enabling efficient modeling of damage initiation, propagation, and coalescence in complex geometries. Several numerical examples are presented in this work, demonstrating the accuracy and computational efficiency of the proposed approach. This study provides a robust tool for researchers and engineers to address the challenges of damage mechanics in concrete structures.

This paper presents three-dimensional finite element simulations of wooden structures using metal connections that employ an elastic-plastic damage model in order to simulate the nonlinear behaviour of wood. The wood constitutive law relies upon orthotropic material parameters, associated plasticity and continuum damage mechanics (CDM) to take into account the following properties of wood : anisotropy, brittle failure in tension, plasticity and ductile failure in compression. The model used in this paper is implemented as a user subroutine of the finite element software ZSoil. Experimental uniaxial compression tests on small wood samples are utilized to conduct a calibration of the material parameters. This calibration is performed by generating a Polynomial Chaos Expansion surrogate model of the true finite-element model, which is then used for a Bayesian inversion on the experimental data from tests on small wood samples. The constitutive model with the calibrated material parameters is then used to numerically reproduce experimental tests on wooden structures. The results demonstrate the model’s capability to reasonably approximate the nonlinear behaviour of wood along with its interaction with metal connections, opening up interesting prospects for engineers to better understand and optimise wooden structures.

Predictive modeling of crack initiation is of extreme interest to the engineering community. Within the context of elastic fracture mechanics, the Scaled Boundary Finite Element Method (SBFEM) has proven both efficient and accurate in estimating Stress Intensity Factors (SIFs). More specifically, the SBFEM allows for an analytical evaluation of the stress field as it approaches the crack tip while reducing the dimensionality of the numerical problem by one. In contrast to other methods, e.g., the FEM or the XFEM, no adjustments to the solution procedure are required, and SIFs can be conveniently extracted during post-processing. However, experimental observations on crack initiation are typically underlined by a significant statistical dispersion. This is mainly due to uncertainties arising from the geometry of the crack tip and the variability of the material properties due to e.g., inhomogeneities at the micro or meso-material scale. To this end, this study presents an efficient approach for estimating SIFs under uncertainty. A stochastic scaled boundary finite element method is developed, and the merits and bottlenecks of a non-intrusive implementation are investigated. Furthermore, a comparative study is performed vis-`a-vis the discretization method employed to generate sample domains, i.e., the Expansion Optimal Estimation (EOLE) and the Karhunen-Lo`eve Expansion (KLE).

Predictive modelling of fracture in materials is critical for understanding progressive failure at the material or structure scale. Traditional fracture modeling techniques often require laborious algorithms to track propagating cracks. Conversely, the phase field method has been established as an appealing alternative, mainly due to its favorable implementational features. Yet, criticism to the phase field method involves its ability to accurately resolve crack nucleation and its associated computational costs. In this work, we integrate adaptive quadtree meshing with the Virtual Element Method with the objective of significantly reducing the computational costs of phase field simulations. Quadtree meshing is a hierarchical grid-based technique used for adaptive mesh refinement in 2D simulations. Using the VEM, hanging nodes in the quadtree decomposition are naturally treated and a conforming mesh is always established in contrast to conventional methods. This optimizes computational resources by refining the mesh locally, improving accuracy in complex regions while maintaining coarser elements elsewhere. Different adaptivity criteria are explored and benchmarks pertaining to mode I and mode II brittle fracture are examined in terms of accuracy and efficiency when compared to the standard finite element method.