Show simple item record

dc.rights.licenseRestricted to current Rensselaer faculty, staff and students. Access inquiries may be directed to the Rensselaer Libraries.
dc.contributorSahni, Onkar
dc.contributorOberai, Assad
dc.contributorShephard, M. S. (Mark S.)
dc.contributorTejada-Martínez, Andrés E.
dc.contributor.authorTran, Steven
dc.date.accessioned2021-11-03T08:42:45Z
dc.date.available2021-11-03T08:42:45Z
dc.date.created2017-01-13T09:41:33Z
dc.date.issued2016-12
dc.identifier.urihttps://hdl.handle.net/20.500.13015/1828
dc.descriptionDecember 2016
dc.descriptionSchool of Engineering
dc.description.abstractWe apply the current LES methodology on a variety of problems. Specifically, we consider four cases: (i) a turbulent channel flow, (ii) flow over a straight circular cylinder, (iii) flow over a wavy circular cylinder, and (iv) flow over a straight square cylinder. For these cases, LES predictions are compared against the experimental and DNS datasets (as applicable). For the turbulent channel flow, we compare LES predictions based on the RBVMS and combined models and find them to be of similar quality. For this case we also investigate the influence of mesh topology and orientation as well as mesh refinement on LES predictions. We observe that a directional bias in the local patch of elements around mesh nodes must be avoided since it results in a detrimental effect on LES predictions while mesh refinement improves the predictions, as expected. For the straight circular cylinder case, a mesh refinement study is performed and LES predictions are compared between the RBVMS model and the combined model. Overall the combined model outperforms the RBVMS model and is employed for the rest of the cases. For example, the combined model accurately predicts important quantities, such as fluctuating velocity and lift force, while the RBVMS model shows a disagreement. For both the wavy circular cylinder and straight square cylinder cases, LES predictions (based on the combined model) show a good overall agreement with the experimental and DNS datasets (as applicable).
dc.description.abstractThe focus of this thesis is the formulation and development of a turbulence-resolving simulation tool based on finite elements. There are many challenges to the development of a reliable numerical approach for predicting the behavior of complex turbulent flows. In such flows, the difficulties emanate from two aspects: (i) complicated geometry and solution features and (ii) a balance between the computational cost and the amount of resolution and modeling that is required to perform accurate predictions. The finite element (FE) method has proven to be a valuable numerical tool to handle complicated features. Specifically, the ease with which the FE method can be applied on unstructured meshes makes it a powerful option for investigating many complex problems of interest. Furthermore, the adaptive meshing, high-order, and parallelization capabilities of the FE method can significantly increase the computational efficiency. In terms of turbulence resolution, direct numerical simulation (DNS) is the most reliable option but it requires a proper resolution of all the relevant turbulent scales (with no modeling). This makes DNS prohibitively expensive from a computational viewpoint for many problems of interest. An attractive alternative is large eddy simulation (LES) which resolves the flow structures on the order of the grid size and models the effects of the scales that are too small to be resolved by the grid (i.e., subgrid scales). %Therefore, LES involves subgrid-scale (SGS) modeling.
dc.description.abstractThis thesis develops an LES methodology that entails the following features: (i) a combined subgrid-scale model in the context of stabilized finite element methods, (ii) a local dynamic procedure based on the variational Germano identity (VGI), (iii) an averaging scheme that makes the local dynamic procedure robust for complex inhomogeneous turbulent flows, and (iv) extension to problems with moving and deforming objects.
dc.description.abstractIn particular, the combined SGS model uses the residual-based variational multiscale (RBVMS) approach along with the Smagorinsky eddy-viscosity model. The RBVMS model is used to represent the cross-stress terms while the eddy-viscosity model is used for the Reynolds stresses. The unknown parameters in the eddy-viscosity model are computed using a dynamic procedure. This dynamic procedure is based on a novel localized formulation of the variational Germano identity that is applicable to inhomogeneous turbulent flows. The resulting local dynamic procedure is made robust by employing a Lagrangian averaging scheme, which is equivalent to averaging along local pathtubes, and maintains the applicability of the current LES methodology to inhomogeneous turbulent flows. Finally, we extend the combined subgrid-scale model, local VGI and Lagrangian averaging scheme on deforming unstructured meshes based on an arbitrary Lagrangian-Eulerian (ALE) description.
dc.language.isoENG
dc.publisherRensselaer Polytechnic Institute, Troy, NY
dc.relation.ispartofRensselaer Theses and Dissertations Online Collection
dc.subjectAeronautical engineering
dc.titleDynamic subgrid-scale modeling for finite element based simulation of complex turbulent flows
dc.typeElectronic thesis
dc.typeThesis
dc.digitool.pid177813
dc.digitool.pid177814
dc.digitool.pid177815
dc.rights.holderThis electronic version is a licensed copy owned by Rensselaer Polytechnic Institute, Troy, NY. Copyright of original work retained by author.
dc.description.degreePhD
dc.relation.departmentDept. of Mechanical, Aerospace, and Nuclear Engineering


Files in this item

Thumbnail

This item appears in the following Collection(s)

Show simple item record