Vincent Lemiale

Modelling of equal channel angular pressing using a mesh-free method

Severe plastic deformation (SPD) processes are widely recognised as efficient techniques to produce bulk ultrafine-grained materials. As a complement to experiments, computational modelling is extensively used to understand the deformation mechanisms of grain refinement induced by large strain loading conditions. Although considerable research has been undertaken in the modelling of SPD processes, most of the studies have been accomplished using mesh-based methods, such as the finite element method (FEM). Mesh-based methods have inherent difficulties in modelling high-deformation processes because of the distortions in the mesh and the resultant inaccuracies and instabilities. As an alternative, a mesh-free method called smoothed particle hydrodynamics (SPH) is used. The effectiveness of this technique is highlighted for modelling of one of the most popular SPD techniques, equal channel angular pressing. A benchmark between SPH and FE calculation is performed. Furthermore, a number of simulations under different processing conditions are compared to existing literature data. A satisfactory agreement is found, which indicates that SPD processes can be approached by mesh-free methods, such as SPH.

A scenario-based risk framework for determining consequences of different failure modes of earth dams

Failure modes for earth dams are extensively reviewed and analysed using a three-pronged approach including a literature review, physical observations of a representative earth dam site and finite element structural analysis of the dam wall. Several failure scenarios are used for predicting consequences in terms of downstream inundation and damage. The fluid flow component is performed using the mesh-free smoothed particle hydrodynamics method. For a representative earthen dam, piping and landslip are identified as key failure modes based on a combination of finite element analysis, theory and physical observations. Inundation behaviour is very different for the two failure modes. The landslip failure is the most critical one for the dam studied with flood water breaking the river bank and affecting surrounding property and farmland. For the piping failures, water flow from the initial pipes formed for significant periods before they collapse, but the flow rates are small compared with that of the much larger landslip mode. After failure, fragments of the collapsing wall block the breach and can considerably restrict the flood discharge. In some cases, the water pressure is able to push the obstructing material downstream and some minor flooding occurs, but in others cases the breach can remain blocked with little flooding occurring. A prototype risk framework is developed using the small database of the pre-computed flooding scenarios and key variables that affect inundation such as water level in the reservoir. This can be used to estimate inundation maps for as yet non-computed scenarios through interpolation and superposition techniques. The implementation of the risk framework is demonstrated by the estimation of inundation maps for two in-between non-computed reservoir levels. Inundation due to multiple breaches is also estimated by superposition of three single-breach scenarios. Results are compared against the simulated multiple breach. A preliminary implementation of this risk framework into a geographic information system is also described.

Multiscale particle-in-cell modelling for advanced high strength steels

Advanced High Strength Steels (AHSS) offer outstanding characteristics for efficient and economic use of steel. The unique features of AHSS are direct result of careful heat treatment that creates martensite in the steel microstructure. Martensite and carbon content in the microstructure greatly affects the mechanical properties of AHSS, underlining more importance on microstructural discontinuities and their multiphase characteristics. In this paper, we present the Multiscale Particle-In-Cell (MPIC) method for microstructural modelling of AHSS. A specific particle method [1] usually used in fluid mechanics is adapted and implemented in a parallel multiscale framework. This multiscale method is based on homogenisation theories; with Particle-In-Cell (PIC) method in both micro and macroscale, and offers several advantages in comparison to finite element (FE) based formulation. Application of this method to a benchmark uniaxial tension test is presented and compared with conventional FE solutions.

Smoothed Particle Hydrodynamics Applied to the Modelling of Landslides

Landslides are among the most devastating natural hazards because they often initiate rapidly and mobilize very large volumes of material. While the mechanics of landslides is relatively well understood it is still extremely difficult to anticipate any particular event and estimate its potential consequences. Mesh-free methods are ideally suited to handle large deformations associated with slope failure but they often assume the mechanism of failure a priori. In this work we apply Smoothed Particle Hydrodynamics to simulate all phases of a landslide within one single numerical platform. A Drücker-Prager model is used to determine the onset of failure. The post-failure behaviour is accommodated naturally by the mesh-free nature of the method. The relevance of the method to the modelling of landslides is demonstrated on several examples of slope failure.

Aiming for Modeling-Assisted Tailored Designs for Additive Manufacturing

It is well recognized that there are gaps in knowledge on the strongly intertwined process–microstructure–property–performance relationships inherent in the metallic additive manufacturing processes. Computational modeling can assist with filling in some of these gaps by increasing in-depth understanding of these relationships and highlighting cause-and-effect. Additionally, it can capture the knowledge of materials scientists and engineers and apply established physics-based rules to simulate the processes and thus predict the final outcomes. Modeling can also help optimize processes. Some even predict that future generations of additive manufacturing machines will employ ‘model-assisted feed forward algorithms’ that would leapfrog feedback control methods. In the current article the authors describe the several computational efforts sponsored by CSIRO’s ‘Lab 22—Australia’s Centre for Additive Innovation’ aimed at modeling-assisted tailored design. The models in development, e.g. microstructure prediction (both fundamental and empirical), powder bed raking, and residual stress predictions, are described in some detail, and representative results are presented.