RACER Pre-proposal Proposal
|Name:||Dr. Yves C Dubief|
|Institution/Firm:||University of Vermont|
School Of Engineering
33 Colchester Ave
Burlington, VT 05405
|Project Title:||Multiscale modeling of sediment transport|
The objective is to develop the framework for the modeling of sediment transport phenomena caused by the action of water flow on granular materials found in river bed and ocean floors. The formation of specific morphological patterns on the surface of sediment beds has profound impacts on the formation of river beds or on the safety of bridge pillars and other man-made submerged structures. The mechanisms of interactions between flow and sediment transport are extremely complex and, as such, remain a great challenge of physics, geology and engineering. In particular, the dynamics of the flow near the surface of a sediment bed, which plays a dominant role in the morphology of the surface, is the most difficult to characterize. Experimentally, measurement techniques, typically of optical nature, are limited near the bed surface owing to unsteady modifications of the morphology and reflections caused by large variations in altitude of the surface. Numerical simulations are typically limited by the constraints of solving accurately for a turbulent flow and the transport of sediment particles. Turbulence alone requires large computational resources and high-fidelity algorithms that are not accessible to all research group. The simulation of sediment particles interacting through collision, adhesion or lubrication, has its own limitations, that relates to efficient algorithms to describe particle-particle interactions.
It is proposed to create a synergistic effort between University of New Hampshire and University of Vermont leveraging the expertise of UNH in sediment transport and erosion, and the expertise of UVM in high-fidelity numerical simulations of complex flow/surface interactions. The development of a multiscale algorithm for sediment erosion will be coordinated via a data server hosted at UVM, which will enable UNH researchers to actively participate in the development and analysis of large, massively parallel simulations of sediment erosion performed at UVM. We will use a massively parallel multiphysics flow solver, which can not only simulate turbulent flows in time and space but can also handle some of the largest number of particles in a flow simulated so far (around 100 millions). The proposed algorithm will be a modification of an erosion algorithm developed for ablation processes by the UVM PI. The UNH PI has an extensive database of experimental measurements in the formation of morphological patterns in sediment beds which will serve as benchmarks for the proposed simulations. The broader impact is (a) the design of predictive algorithms and theories for the formation of sand ripples and scour in river beds and coastal regions and (b) the establishment of strong collaboration between the two institutions leading to the submission of funding requests to federal agencies (NSF, DoD, NASA).
|Collaborative Project Details:||UNH PI: Dr. Diane L Foster is an Associate Professor of Mechanical Engineering. She has conducted funded research projects for agencies and research organizations including NSF, Sea Grant, ONR, USGS, and the National Ocean Partnership Program. She and her graduate students research explores the fluid-sediment interaction in coastal and nearshore environments. They have obtained observations in both the field (North Carolina, Florida, and Oregon, and Ohio) and large laboratories (Oregon, TUDelft). Prof. Foster serves as an ocean advisor for the Center of Science and Industry (COSI), and as a Co-PI for the Coastal Sediment Transport Model (CSTM).
UVM PI expertise: Dr. Yves Dubief is an Assistant Professor of the School of Engineering. He has conducted research projects for agencies including NSF, NASA, NIH and DoD. He and his students develop multiscale approaches to study multiphysics problems, such as, influence of polymer additives on turbulence, ablation phenomena during atmospheric reentry, blood coagulation and lubrication in articular joints. The group has developed strong expertise in continuum (computational fluid dynamics) and particle-based (molecular dynamics) numerical methods. Dr. Dubief is a fellow of the Vermont Advanced Computing Center (VACC). Dr. Dubief is also part of a multi-university collaborative effort to develop a multiphysics and multiscale flow solver for complex flow systems.
The project is built on the concept of collaborative numerical and experimental efforts. Both methods have their own set of limitations, which may be overcome if simulations and experiments contribute to one another. In the present case, we will use existing experimental data to set the boundary conditions of our simulations and to assess the quality of our simulations. Using principles of uncertainty quantifications, the most sensitive approximation built in our numerical model will be identified and investigated. Upon satisfactory agreement between simulation and experiment, the numerical data will be used to gain insight of the spatio-temporal structure of sediment transport in regions inaccessible to measurement techniques.
The collaboration will consist of regular visits and from the design of database web interface hosted at UVM. Part of the budget will be used to modify an existing workstation to become a database repository and analysis center. A web interface will be developed with the help of VACC programing experts to facilitate the access and analysis of the generated database by UNH researchers. An example of such interface may be found at http://turbulence.pha.jhu.edu/ . The objective is to develop an interface that could be easily adapted to any set of large data that may not be easily transferred via ethernet. This interface is anticipated to become a strong asset of future collaborations between VACC users and their collaborators outside of UVM.
|Project Overview - include goals, general expertise *:||The specific objective of this RACER proposal is to (I) demonstrate the potential for collaboration of UVM and UNH on sediment transport, (II) develop an algorithm that surpasses existing simulation software in sediment transport and (III) design a web interface for efficient access of large datasets produced from numerical simulations.
I) The goal is to study the scour formation around a vertical cylinder, an experiment recently carried out by Dr. Foster. The simulation will be initially performed using a continuum algorithm developed by Fringer at Stanford for sediment transport. Fringer's algorithm can easily be implemented in UVM's existing algorithm for the simulation of low-temperature ablation presented in 2010 at the Center for Turbulence Research Summer Program. UVM' ablation algorithms is based on level-set methods to define the flow/solid interface. This first simulation will focus on the identification of the sources of uncertainty on the continuum algorithms.
II) Leveraging the unique capabilities of our code in discrete element modeling (DEM, i.e. particle transport with particle-particle interactions), we will develop a hybrid method based on discrete particles and level-set methods. Using the table top experiment of Munro et al. (Phys. Fluids, 2009), we will investigate a two-layer approach of the surface of sediment bed. The first layer consists of particles, packed according to their particle-particle interactions. This first layer rests on a level-set defined surface whose morphology adjust to the constraints applied to the first layer. The two-layer approach is envisioned to reduce the computational cost by modeling the sediment bed as a continuum in regions where no deformation occurs. However, the first layer will enable a very accurate description of the physics of flow particles colliding with the bed and bed particles being extracted into the flow due to hydrodynamic forces. Typical DEM simulations found in recent publications have of the order of 10,000 particles. Our current code has performed the simulation of 30 million particles in a fluidized bed.
|||) A web interface will be developed to probe data generated by UVM's simulation. The web interface will enable external users to perform combinations of mathematical and statistical operations. It is envisioned that external users will upload a C or fortran subroutines which will be compiled by the post-processing solver on the UVM server. One analysis is done, the web interface will transfer the results back to the user. We will contact the bioinformatic group working with NECC to inquire about possible collaborations on this topic.