The VT-Rnet connection has been used to transfer large data sets that were generated while running numerical simulations of fluid flow and particles for a range of engineering and scientific applications. A speed-up factor of 1.5x was observed for file transfer rates between local workstations and data storage facilities at Virginia Tech’s Advanced Research Computing. With file sizes upward of 1 TB, such scale-ups can result in large savings in file transfer times, and accelerate setting up and debugging practices, as well as post-processing analyses.

Representative research studies are listed here, and the associated mesh and file sizes are included, to get a general picture of the magnitude of the problems.


Prediction of the behavior of fluid flow past particles of varying shapes has been generating interest from many researchers due to its diverse industrial applications. Examples are combustion of pulverized coal, fibre suspension flow in paper forming, pneumatic conveying of granular materials, pollutant transport in the atmosphere and transport of sediment grains in river. Among the early researches, most studies performed their research on spheres due to the simplicity of the geometry. Nevertheless, a sphere is usually not the best approximation of random shapes in nature. Our focus is on a low aspect-ratio cylindrical shaped particles with different packing conditions in the flow field.

In this study, the particle count varies from 286 to 859 in a domain composed by 128 million computational cells, accounting for solid fractions between 10% and 30%. The surface of each particle is discretized using triangular elements, with the total number of triangles ranging from 3.3 million to 10 million for different solid fractions. 30 different cases are simulated considering 3 solid fractions, 5 Reynolds number and 2 different particle assembly at each solid fraction and Reynolds number. This results in a total file size of 12.9 TB.


A collaboration between the College of Science and College of Engineering is aimed at analyzing and improving heat transfer for a complex fuel-rod geometry. Fluid flow and heat transfer simulations of airflow around the existing design are carried out. An unsteady inflow demands a time-accurate simulation, forcing the need for high temporal resolution. Design of an insulator encompassing the fuel-rod further complicates these calculations, with pore dimensions ranging from the micron-scale to a few centimeters. The need to resolve high temperature gradients and the wide range of physical dimensions, as well as the unsteady inflow, results in each simulations taking nearly a month to complete a multiple flow-through cycles.

System level approximations of the fuel-rod and the surrounding insulator are made by setting appropriate porosity, permeability and material properties. A Gaussian profile for the inflow velocity is used to model the unsteady airflow, and a heat sink mimics energy loss due to evaporation. These are laminar flow simulations that capture the dynamics of airflow around the complex outlines of the fuel-rod, and subsequent expansion of the flow into a larger diameter section. In addition to the airflow, conduction through the surround insulator and the fuel-rod are also included in these conjugate heat transfer calculations. Preliminary and optimized designs for the same conditions are converged at, keeping in mind prescribed parameters that include the maximum peripheral wall temperature and the volume flow-rate of the gas.

Immersed boundary simulations of such configurations require background fluids meshes with approximately 80 million cells, and fine features on the surface of the fuel-rod necessitate nearly 12 million triangular elements. For the insulator, an additional 47.5 million triangles are required. Such cell counts generate solution files of the order of multiple gigabytes, and with temporary varying solution, files of the order of multiple terabytes are generated, requiring high-speed connectivity for post-processing and visualization.


Khosronejad et al., 2012. Melville and Coleman, 2000.

Complex interactions between horse-shoe and wake vortices with particulate beds at the bottom of rivers can compromise the stability of piers immersed in the river beds. Studying these phenomena is crucial in extending the lifespan of mechanical structures in river and ocean beds.

This study involved the flow of water over a particulate bed around a square pier at a Reynolds number of 30,000 (based on pier side length). A discrete element method (DEM) was used to compute the particle bed dynamics, with each particle of size 0.85mm diameter. A representative particle model was used to reduce the total number of particles from upward of 100 million to approximately 7.2 million. To further reduce the cost of simulation, a background mesh consisting of 24 million cells was setup to represent a periodic domain for the large eddy simulations (LES).

Grids for flow over particle bed

Two separate simulations were conducted, where interaction between particles in the bed was represented using the soft-sphere and hard-sphere models, and conclusions were drawn based on the time-varying interaction between the particles, the fluid and the rectangular pier.

Solutions to particle bed simulations

Considering the unsteady nature of the problem, time-accurate solutions were generated to evaluate each model, and data sets corresponding to numerous instances in time had to be created, each resulting in file sizes upward of 5GB. The total size of all the files generated for each simulation was upward of 1TB even for this scaled down version of the problem. Future planned simulations of realistic particle sizes and bed dimensions are expected to yield even large file sizes, and would require data sets at higher frequencies, to accurately determine the effectiveness of each collision model.



The complex and articulated skin-like wing surface of bats allows them to be highly maneuverable fliers. Understanding the associated aerodynamics can be insightful in designing wing shapes for aircraft systems, and also for setting up actuator system for autonomous micro-air vehicles (MAVs).

Grid for computing flow around bat wing

This study used the kinematics of a bat (Hipposideros) to simulate air flow around the wing surface at a Reynolds number of 12,000. The bat’s wing was represented using 15 thousand triangular elements in an immersed boundary method (IBM) framework as a thin surface. The background fluid features were captured using a Cartesian mesh of approximately 33 million cells. Each solution file for this setup was approximately 8GB in size.

Unsteady flow around bat wing

Due to the highly unsteady nature of the flow, monitoring the time-accurate variation of the fluid coherent structures was necessary. 200 time-frames were identified, which resulted in a total file size of 1.6TB. That is for a single simulation of a single trajectory of bat flight. Future planned simulations of more complex kinematics demand larger numbers of time-frames to resolve wing articulations, potentially generating file sizes upward of tens of terabytes. This work is funded by the NSF under grant CBET-1510797.

Selected Publications

  1. Hosseinzadegan, H. and Tafti, D.K., “A 3D predictive model of thrombus growth in stenosed vessels with dynamic geometries,” J. of Medical and Biological Engineering, accepted July 2018.
  2. He, L. and Tafti, D.K., “Variation of drag, lift and torque in an assembly of ellipsoidal particles,” J. Powder Technology, accepted June 2018.
  3. Elghannay, H. and Tafti D. K., “Revised Partial Coupling for in Fluid-Particulate Systems,” J. of Computational Multiphase Flows, accepted July 2018.
  4. Hosseinzadegan, H. and Tafti, D.K., “Mini Review: Mechanisms of Platelet Activation, Adhesion and Aggregation,” Thromb Haemost Res (2017);1(2):1008.
  5. Elghannay, H. and Tafti D. K., “Sensitivity of Numerical Parameters on DEM Predictions of Sediment Transport” Particulate Science and Technology, (2017), http://dx.doi.org/10.1080/02726351.2017.1352638
  6. Hegde, M. Meenakshisundaram, Chartain, N., Sekhar S., Tafti D.K., Williams, C. and Long, T., “3D Printing All-Aromatic Polyimides using Mask-Projection Stereolithography: Processing the Nonprocessable, Advanced Materials 2017, 1701240. DOI:10.1002/adma.201701240
  7. He, L. and Tafti, D.K. “Heat Transfer in an Assembly of Ellipsoidal Particles at Low to Moderate Reynolds Numbers,” Int. J. of Heat Mass Transfer 114, Nov. 2017, pp. 324-336. DOI: https://doi.org/10.1016/j.ijheatmasstransfer.2017.06.068
  8. Hosseinzadegan, H. and Tafti, D.K., “Review: Modeling Thrombus Formation and Growth,” Biotechnology and Bioengineering, 2017. DOI: 10.1002/bit.26343.
  9. Yu, K., Elghannay, H.A. and Tafti D.K., “An Impulse Based Model for Spherical Particle Collisions with Sliding and Rolling,” J. Powder Technology 319, September 2017, pp. 102-116 DOI: https://doi.org/10.1016/j.powtec.2017.06.049
  10. Hosseinzadegan, H and Tafti, D. K., Prediction of Thrombus Growth: Effect of Stenosis and Reynolds Number, Cardiovascular Engineering and Technology, Vol. 8, No. 2, June 2017, pp. 164-181. DOI:10.1007/s13239-017-0304-3.
  11. He, L. and Tafti D. K., “Evaluation of Drag Correlations Using Particle Resolved Simulations of Spheres and Ellipsoid in Assembly,” J. Powder Technology 313, 15 May 2017, pp. 332-343.

Submitted for Publication

  1. Elghannay, H. and Tafti D. K. and Yu, K, “Evaluation of physics based hard-sphere model with the soft sphere model for dense fluid-particle flow systems,” Int. J. Multiphase Flow, submitted August 2018.
  2. Fan, X., Sekhar, S. Windes, P, Tafti, D.K. “Canonical Description of Wing Kinematics and Dynamics for a Straight Flying Insectivorous Bat (Hipposideros Pratti), Physics of Fluids, submitted June 2018.
  3. He L, and Tafti D.K. “A Supervised Machine Learning Approach for Predicting Variable Drag Forces on Spherical Particles in Suspension,” J. Powder Technology, submitted June 2018.
  4. Yu, K. and Tafti, D. K. “Size and temperature dependent collision and deposition model for micron-sized sand particles”, J. Turbomachinery, submitted May 2018.
  5. Windes, P., Fan, X., Bender, M.,Tafti, D.K., Müller, R. “A computational investigation of lift generation and power expenditure of Pratt’s Roundleaf bat (Hipposideros pratti) in forward flight,” PLOS ONE, submitted April 2018.
  6. Elghannay, H. and Tafti D. K., “Alternate Tangential Impact Treatments for the Soft-Sphere Collision Model,” Particulate Science and Technology, submitted March 2018.

M.S. Thesis

  1. Cody Dowd: A Study Of Centrifugal Buoyancy And Particulate Deposition In A Two Pass Ribbed Duct For The Internal Cooling Passages Of A Turbine Blade, 2016.
  2. Peter Windes (Co-advisor: B. Behkam), Computational Modeling Of Intracapillary Bacteria Transport In Tumor Microvasculature, 2016.
  3. Adam Norman, A Fundamental Study Of Advance Ratio, Solidity, Turbine Radius, And Blade Profile On The Performance Characteristics Of Vertical Axis Turbines (VATs), 2016.
  4. Steven Paul, A Computational Framework for Fluid-Thermal Coupling Of Particle Deposits, 2018.
  5. Xiaozhou Fan, Canonical Decomposition Of Wing Kinematics For A Straight Flying Insectivorous Bat (Hipposideros Pratti), 2017.

Ph.D. Thesis

  1. Husam El-Ghanney, Methods Development and Validation for Large Scale Simulations of Dense Particulate Flow Systems in CFD-DEM Framework, 2018.
  2. Long He, Study of Fluid Forces and Heat Transfer on Non-spherical Particles in Assembly Using Particle Resolved Simulation, 2017.
  3. Hamid Hosseinzadegan, A physio-chemical predictive model of dynamic thrombus formation and growth in stenosed vessels, 2017.