Student Posters Repository

Flow in porous media occurs in many natural and engineered systems such as groundwater, contaminant transport and remediation, oil and gas exploration, CO₂ sequestration, packed beds, absorbent hygiene products, and textiles. The study of fundamental flow and particle transport processes at the pore scale (in the order of micrometers) is essential to understanding how small scale mechanisms affect larger, field-scale processes. Pore-scale imaging and modeling is one of the techniques used to investigate these fundamental mechanisms. We use x-ray computed microtomography (XMT) to acquire three-dimensional images of porous media. The XMT images are processed and subsequently meshed to allow simulations of fluid flow and particle transport. My research is focused on the development mesh improvement algorithms for tetrahedral meshes, fluid flow simulation using an image-based finite element method code, and algorithm development and simulation of nanoparticle transport.

Aerogel is a kind of open-cell nanoporous materials with high porosity, low density, high surface area, and low thermal conductivity, which exhibits great application potential for thermal insulation. The heat transfer within aerogel mainly includes thermal conduction form solid skeleton, gaseous heat transfer and thermal radiation. Among the path of heat transfer, the heat transfer through gas generally plays an important part, especially for the structure with high porosity. It is known that the gaseous thermal conductivity is influenced significantly by the atmosphere factors (pressure, temperature etc.). There will exist gas permeation induced by the pressure gradient within the material when the atmosphere pressure changes. The gas permeation will result in additional energy migration, which influences the thermal insulation performance of the materials.  In this study, we aim to reconstruct the complex and interconnected nanostructure of silica aerogel by three-dimensional (3D) diffusion-limited cluster-cluster aggregation (DLCA) method. Then, we will use a modified direct simulation Monte Carlo (DSMC) code which is suitable for gas flow in porous medium based on Bird’s to study the gaseous thermal conductivity and permeability in nanoporous material with Fourier’s law and Darcy’s law, respectively.

Aortic coarctation (CoA) causes pressure drop across the narrowing. Complications after repair, such as recoarctation, aneurysm formation and systemic hypertension are encountered in more than one third of adult patients throughout their lives, and need to be closely monitored. Cardiovascular MRI (CMR) is widely used for follow-up, but mainly provides anatomical information. Obtaining hemodynamic information is more challenging but very valuable to detect complications. The gold standard to assess CoA severity is the pressure drop across the narrowing measured with catheterization. Unfortunately, catheterization is an invasive procedure with a 1 in a 1000 risk of patient death or serious injury. Intervention is recommended if the pressure gradient exceeds 20 mmHg. Computational fluid dynamics (CFD) is a promising non-invasive method to calculate the pressure drop through the CoA, based on CMR images. The aim of this study was to evaluate the performance of a MR-based CFD approach using catheterization as the reference standard.

We adaptively combine and couple different plasma models (fully kinetic, hybrid, two-fluid, and in future also MHD models) and their corresponding codes in a hierarchic way. This requires careful adjustment of the models at the coupling boundaries and well-defined criteria for switching the model. We solve the kinetic equations with a semi-Lagrangian scheme and PFC (positive flux-conservative) reconstruction. We solve the fluid equations with a Kurganov-Levy scheme and CWENO reconstruction.  Additionally, we use FDTD for determining the electromagnetic fields. The kinetic equations live in 6-dimensional phase-space but can be parallelized massively. We solve them on GPUs for efficiency reasons.

In recent years, high-order finite-element schemes have received a significant attention in the computational fluid dynamics (CFD) community due to the several advantages they can offer over the more established 2nd order finite-volume counterparts. However, the resulting discretized equations that must be solved for steady-state and time-implicit turbulent flow problems have proven to be much stiffer and more difficult to converge efficiently and robustly. The objective of the present study is to investigate and develop robust, efficient, and scalable multilevel solution strategies and preconditioning techniques for stabilized finite-element flow solvers. The proposed solution strategy is essentially a spectral multigrid approach in which the solution on the mesh with lowest polynomial degree (p=1) is solved using a preconditioned Newton-Krylov method. We have developed an implicit line preconditioner which can be properly distrusted among processing elements without affecting the convergence behavior of the linear system and non-linear path. The Implicit lines are extracted from a stiffness matrix based on strong connections. To improve the robustness of the implicit-line relaxation, a double CFL strategy, with a lower CFL number in the preconditioner matrix, has been developed.

See my website:

https://behzadahrabi.wixsite.com/simulation

Flow in a channel with grooves parallel to the flow direction has been studied by means of spectrally accurate Direct Numerical Simulations, with the primary goal of establishing channel geometries that enhance achievable laminar mixing at possibly low drag increase. Stability with respect to small disturbances is investigated using the direct numerical simulations (DNS) of the Linearized Navier-Stokes (LNS) equations, performed by means of spectral element fluid flow solver. In these simulations, the long-term asymptotic response to initial perturbation is determined, and the least attenuated (or most amplified) form of the perturbation (the unstable mode) is identified and tracked over a range of geometric and flow parameters. Critical conditions for the onset of instabilities at a range of parameters are also obtained. The groove shape having the largest mixing capability is identified by analyzing the structure of the disturbance. Finally, nonlinear saturation of the unstable modes and the resulting secondary flows are examined by means of a direct numerical solution of the full nonlinear Navier-Stokes system. Our results indicate that secondary flows, at saturation, can maintain the drag reducing properties for a limited range of Reynolds numbers. Phase space trajectories obtained by probing the flow solution at selected points. Based on Phase trajectories, the condition leading to the chaos obtained.

In hazard assessment for geophysical mass flows, we seek to construct accurate and reliable maps that show regions with high risk of hazard. Therefore, acceptably accurate numerical simulation of complex geophysical mass flows is of crucial importance. Modeling mechanical behavior of such flows or the flow rheology as well as taking care of solving the highly nonlinear system of hyperbolic PDEs which includes shock and discontinuity present major difficulties of this work. In this contribution, we present TITAN2D, the parallel toolkit for geophysical mass flow simulations which use adaptive mesh refinement in conjunction with a shock-capturing solver. Recent implementation of our software offers multiple well-known choices for flow rheology which allows seamless access to all these options for the same site in the same computational workflow.

Traditional methods of candidate gene identification do not work for certain traits, including those that present as non-survival in an individual's offspring, for certain groups of organisms, including humans, or for traits controlled by multiple genes. A new method, Evolutionary Rate Covariation (ERC), has been developed in mammals, drosophila, and yeast that compares the evolutionary rates of genes when mapped to the same species tree. ERC can identify individual genes with unique evolutionary rates on individual branches of a phylogenetic tree. Plants have much higher rates of gene duplication than animals and fungi, making ERC more challenging to calculate. I have developed a bioinformatic pipeline for calculating ERC in plants, and I present my preliminary results in this poster, which show that the pipeline works, as well as my plans for using this pipeline to investigate one meiotic variation, polyploidy, in plants. Polyploidy is a main cause of miscarriage in humans and has greatly influenced evolution within all eukaryotic kingdoms.

My research is focused on investigating and improving the scalability of task-based programming models for future Exascale machines. Tasking/task-based approach refers to designing programs in terms of “tasks”, logically discrete sections of work to be done. This approach is a promising way to program Exascale applications: they are divided into a myriad of small tasks, so the system is overprovisioned with them (the number of tasks is much larger than the number of available cores). The task scheduler selects tasks and dynamically assigns them to idle threads, which can execute them. Details of task scheduling to hardware are hidden from programmers; the task-scheduling runtime is responsible for efficient execution of task-based programs. Thus, developers just focus on implementing their algorithms in terms of tasks, what decreases programming effort and gives portability across different hardware. Within this research, I have already found an optimal task granularity for task-based applications depending on what scheduler is deployed to arrange tasks to processing units. Also, I have come up with a promising approach for designing future Exascale programming models. Currently, I am addressing data locality issues as it is the main limitation for having a good scalability for task-based applications running on thousands cores.