Hello, A yearly allocation for the LCRC cluster has been requested with the following updated information: Submitter/PI: Kent E. Wardle Project Name: SepCFD Division: CSE Project title: CFD Simulation of Separations Processes Associated funding: - DOE-NE Fuel Cycle Research and Development (FCR&D) Program - DOE-NE Nuclear Energy Advanced Modeling and Simulation (NEAMS) Program - DOE NEUP Program Other Systems: - local 8-core workstation for meshing, postprocessing, and small-scale development - MCS Full-disclosure (SiCortex 5832) [FELBM work only] Science: This project uses the open-source CFD package OpenFOAM (www.opencfd.co.uk) which is based on the finite volume solution method of the Navier-Stokes equations on arbitrary polyhedral, unstructured meshes for computational fluid dynamics simulations. OpenFOAM is capable of simulation of a wide range of physics and the focus in this project is on simulation of turbulent, multi-phase (liquid-liquid and liquid-liquid-air) mixing and separation. The specific application of interest is for liquid-liquid extraction processes employed in recycling of used nuclear fuel. The focus is on advanced multiphase simulations of flow in annular centrifugal contactors---a key piece of equipment for these processes. In addition to this more ‘traditional’ CFD approach, a small portion of the project seeks to explore advanced CFD methods using a finite-element based unstructured, lattice Boltzmann simulation method. This is promising technique which can bridge the gap between mole cular and continuum scales and can potentially allow much more efficient computational scaling than traditional CFD techniques. Project description: This project seeks to apply the open-source CFD package OpenFOAM to simulation of the complex, multiphase flow in liquid-liquid separations devices used for processing nuclear fuel via solvent extraction with a particular focus on simulations of the flow in the annular centrifugal contactor (https://sites.google.com/site/kwardleweb/research/centrifugal-contactor). For such processes, complex physical phenomena including free surface flow and liquid-liquid droplet breakup and coalescence must be incorporated. Accurate simulation of such physics requires fine meshes to resolve phase boundaries as well as efficient computational performance to enable transient simulations spanning several seconds of flow time despite the very small time steps that are required (~10-6 s). The specific simulation techniques to be used for the OpenFOAM simulations are based on the Volume of Fluid (VOF) method with Large Eddy Simulation (LES) of turbulence. As part of FY11 w ork, a hybrid multiphase CFD method which couples multifluid (per-phase momentum equations) with VOF interface sharpening for selected phase interfaces was developed. With this new solver, a three phase system can be modeled enforcing a sharp interface for liquid–air interfaces and at the same time allow interphase dispersion and mixing for the liquid–liquid flow. The code package OpenFOAM uses a C++ object-oriented framework with dynamically linked libraries and high-level solvers to enable the user to write custom solvers and combine various physics models and equations enabling the evaluation of unique combinations of models such as are required for the development of advanced liquid-liquid mixing simulation capability as targeted in this work. Work during FY12 will focus on three-phase (liquid-liquid-air) simulations of the coupled mixing zone and separation zone of the annular centrifugal contactor using the newly developed coupled multifluid-VOF solver. The first part of the year will be spent on testing of the base solver after which several next-stage enhancements will be made. This includes the development of methods for switching between dispersed phase modeling and sharp interface modeling. The main goal for FY12 work will be the implementation of polydisperse droplet capturing capability including models for droplet breakup and coalescence leading to ultimate prediction of liquid—liquid interfacial area. This allocation supports nuclear fuel cycle simulations efforts under DOE-NE as well as potential collaborations with DOE-EM sites such as Savannah River where centrifugal contactors and other multiphase separators are in use. The simulations performed here will also support ongoing FY12 experimental work in CSE led by the PI aimed at obtaining detailed multiphase measurements in an engineering-scale annular centrifugal contactor using electrical tomographic techniques. It is hoped that the experimental effort can provide data for validation of the advanced multiphase CFD models developed as part of this work. Aside from the work outlined above, one additional component of this project seeks to explore 'advanced' CFD techniques which may be more inherently suited to massive-scale parallel computations. One such advanced method is based on a finite element, unstructured method for solution of the discrete Boltzmann Equation along with a free-energy based multiphase simulation technique. This method does not require a global pressure solve and depends primarily on nearest neighbor information allowing it to exhibit more ideal parallel scaling. While to-date many of these simulations have been done on other systems (e.g. full-disclosure), a small fraction of the project allocation may also be used for this developmental effort. This portion of the project has received independent funding from the DOE Nuclear Energy University Programs (NEUP) under a three-year collaboration beginning in FY12 with Dr. Taehun Lee of City College of New York. As this part of the work grows, it may be necessary to break this out into a separate allocation. This project has one PI (Wardle) and may have at most 1-2 additional participating researchers, collaborators, or students. Project URL: http://sites.google.com/site/kwardleweb/research/cfd Current FY Hours Used: undetermined amount New FY Requested allocation: 475000 Q1: 150000 Q2: 75000 Q3: 100000 Q4: 150000 Justification: The OpenFOAM code has been found to scale quite well on a number of different computational systems. The Volume of Fluid technique combined with the fine mesh resolution required for droplet and interface tracking necessitate simulation timesteps on the order of microseconds. As such, typical runs require about 100 wall hours for every 1 second of simulated flow time. On Fusion with more recent versions of the code with minor enhancements, we have observed that this can be reduced to ~50 hours/1 second of flow. A plot showing a comparison of the scaling (strong) of a characteristic OpenFOAM simulation (VOF interFoam solver) for a relatively small-size problem (530K tetrahedral cells) is posted here: www.mcs.anl.gov/~wardle/scaling_ann3d_530K_cosmeaFdSurveyorFusion.png . Note that the number of processors has been scaled to 2.0Ghz equivalents. OpenFOAM performs quite well on Fusion. During FY11 work, a number of runs using polyhedral meshes on the order of 3-5M cells were done on as many as 384 processors. Due to time step limitations of free-surface simulation methods, typical runs will only have meshes on the order of 1M cells (~6M points for polyhedral cells) and run on ~200 processors. As the newly developed solver is more computationally intensive (per-phase momentum equations, plus additional droplet balances in the future) the scaling will need to be explored again to ensure the optimum number of processors are used. Thank You, The LCRC Accounts System