Simplicity of mesh generation and robustness against mesh entanglement during large deformations are key attractive features of particle based methods. These features can be exploited in number of engineering problems where traditional techniques suffer due to aforementioned limitations. Numerical modelling of particulate composites is one of such ideal engineering applications where particle based methods can be effectively used due to their simplicity and robustness. Complicated geometrical configurations of particulate composites obtained from techniques such as scanning electron microscopy (SEM) can be easily converted to particle based mesh without loosing much information. This enables more accurate analysis of the chosen composite materials. Therefore, a smooth particle hydrodynamics (SPH) based numerical technique is developed here to investigate the mechanical properties and evolution of debonding process in particulate composites. To perform the numerical study, a Lagrangian corrected SPH (CSPH) method is presented together with an appropriate numerical model for treating material interface discontinuity within the particulate composites. The material interface discontinuity is enforced using an innovative method which combines penalty formulation with a bilinear interface cohesive model for SPH method. The proposed SPH methodology is used in a number of numerical examples involving composite materials and related interface problems. The effect of penalty value on the interface model and of the smoothing length of the SPH method are also analysed during these simulations. The results illustrate the effectiveness, robustness and potential of the developed methodology. It is concluded that the proposed numerical techniques can be easily and effectively applied to simulate multi-phase composites with various interface conditions and, can provide useful information regarding the inherent mechanism of damage evolution and fracture of particulate or fibre reinforced composites. (C) 2009 Elsevier B.V. All rights reserved.

Sphere packing is an attractive way to generate high quality mesh. Several algorithms have been proposed in this topic, however these algorithms are not sufficiently fast for large scale problems. The paper presents an efficient sphere packing algorithm which is much faster and appears to be the most practical among all sphere packing methods presented so far for mesh generation. The algorithm packs spheres inside a domain using advancing front method. High efficiency has resulted from a concept of 4R measure, which localizes all the computations involved in the whole sphere packing process.

In this paper, we investigate boundary recovery, the problem that has troubled researchers ever since Delaunay-based methods were applied to generate mesh. There are a number of algorithms for boundary recovery already and most of them depend heavily on adding extra nodes. In this paper, we make an effort to seek a method to recover boundaries without using extra nodes.It was noted that some previous algorithms imposed artificial boundary constraints on a meshing problem at the recovering stage; we first try to discard these artificial constraints and thus make things easier. Then a new method is proposed by which the boundaries can be recovered by means of two operations: (1) creating a segment in the mesh and (2) removing a segment from the mesh. Both operations are special cases of a general local transformation called small polyhedron reconnection operation. The method works well when coupled with the sphere-packing method proposed by the first author. If the mesh sizing function is suitable, a good configuration of nodes will be created accordingly by the sphere-packing method and the boundary can be recovered by the local transformation presented here without inserting extra nodes. Copyright (c) 2007 John Wiley & Sons, Ltd.

The analytical stress-strain relation with heterogeneous parameters is derived for the heterogeneous brittle materials under a uniaxial extensional load, in which the distributions of the elastic modulus and the failure strength are assumed to be statistically independent. This theoretical solution gives an approximate estimate of the equivalent stress-strain relations for 3-D heterogeneous materials. In one-dimensional cases it may provide comparatively accurate results. The theoretical solution can help us to explain how the heterogeneity influences the mechanical behaviors. Further, a numerical approach is developed to model the non-linear behavior of three-dimensional heterogeneous brittle materials. The lattice approach and statistical techniques are applied to simulate the initial heterogeneity of heterogeneous materials. The load increment in each loading stage is adaptively determined so that the better approximation of the failure process can be realized. When the maximum tensile principal strain exceeds the failure strain, the elements are considered to be broken, which can be carried out by replacing its Young's modulus with a very small value. A 3-D heterogeneous brittle material specimen is simulated during a full failure process. The numerical results are in good agreement with the analytical solutions and experimental data.

By using the lattice model combined with finite element methods and statistical techniques, a numerical approach is developed to establish mechanical models of three-dimensional heterogeneous brittle materials. A special numerical code is introduced, in which a lattice model and statistical approaches are used to simulate the initial heterogeneity of material properties. The size of displacement-load step is adaptively determined so that only few elements would fail in each load step. When the tensile principal strain in an element exceeds the ultimate strain of this element, the element is considered broken and its Young's modulus is set to be very low. Some important behaviors of heterogeneous brittle materials are indicated using this code. Load-displacement curves and figures of three-dimensional fracture patterns are also numerically obtained, which are similar to those observed in laboratory tests.