Center for Fluid Mechanics
And
The Fluids, Thermal and Chemical Processes Group
Of
The Division Of Engineering
Seminar Series
Department of Mechanical Sciences and Engineering University of Illinois, Urbana-Champaign Urbana, IL | |
Abstract: The talk will focus on our efforts using adjoint-based optimization to study the fluid mechanics of sound generation by free shear flows, with application to jet noise. An optimal control algorithm is shown to be able to reduce the noise from a model flow by over 10dB, which affords a unique opportunity to investigate the noise mechanisms that are suppressed by the control. Direct comparison of the noisy flow to its quieted counterpart shows that it is superficially unchanged.
However, projection onto the flow's empirical eigenfunctions (POD modes) indicates that the controlled flow has a more regular underlying wave-packet character. It is concluded that the control exploits the sensitivity of such wave-packet-like noise sources to their specific details in order to `find' small perturbations that make the flow acoustically inefficient.
Joint Stochastic/PDE Seminar
Princeton University | |
Abstract:
We consider complex-valued solutions of a three-dimensional
Navier-Stokes system without external forcing. We show that
there exists an open set in the space of 10-parameter families
of initial conditions such that for each family from this set
there are values of parameters for which the solution blows up
in finite time.
(Joint work with Prof. Y.G. Sinai)
Brown University -
Computer Science Department
Associate Professor of Radiology, Harvard Medical School | |
Abstract: Two applications of diffusion MRI will be discussed: the mapping of connectional neuroanatomy and the imaging of cortical architectonics.
Is diffusion MRI of neural connectivity a well- or ill-posed problem? To address this, we compared a simple deterministic tractography of ex vivo DSI with the recent isotope tracer studies of Schmahmann and Pandya in the macaque. In brief, the DSI accurately demonstrated principal cortical association pathways, both as to 3D trajectory and point-to-point connectivity (see below). Further, the DSI demonstrates a characteristic motif of cortical connection, namely, the division of specific zones' cortical connections into long and short association, commissural, striatal and projection pathways. Compared with these studies of the macaque, present human in vivo DSI tractography, while more accurate than comparable diffusion tensor tractography, remains highly oversimplified. These data indicate DSI tractography of connectional neuroanatomy can, under good conditions, compare favorably with gold standard tracer.
In DSI of the neocortex, routine analysis of directions of maximum diffusion shows pervasive radial orientations, parallel to cortical micro-columns, but only weak and intermittent intracortical structure. Analysis of the 2D planes of maximum diffusion, planar order, shows a quite different picture, revealing a strong horizontal anisotropy of the cortex. Further, coherent regions of this horizontal anisotropy show significant correlation with cytoarchiteconic regions. This suggests that planar structure may be a key to the noninvasive imaging of cortical structure.
Biography: Van Wedeen's research focuses on the development and application of MRI methods to investigate tissue motion and flow, and more recently tissue biomechanics in human subjects in the normal state and in disease.
Transatlantic Seminar
Scientific Computing Seminar
Please Note Change in Day, Time and Place for This Seminar Only |
Abstract: Through the advent of enhanced medical imaging, computational flow modelling can now be applied to anatomically realistic, in-vivo arterial geometries. Whilst this type of modelling can provide plentiful data on difficult to measure quantities such as wall shear stress distributions, our understanding of the fundamental fluid mechanics within these types of geometries remains somewhat limited. Furthermore if we hope to apply computational flow modelling to provide clinical useful data some form of data reduction will be necessary that ideally should be designed to capture the fundamental dynamics of the problem. It is therefore only through our understanding of the fundamental dynamics that this type of reduced modelling might be possible. These types of challenges have encouraged us to study the fluid mechanics under physiological flow conditions within idealised models. In this presentation we will discuss the generation and development of complex flow features generated under both steady and pulsatile flow within simplified geometries ranging from idealised anastomosis, to helical pipes and simplified stenoses configurations.
Scientific Computing Seminar
Massachusetts Institute of Technology, Cambridge, MA | |
Abstract: Discontinuous Galerkin methods have gained popularity in the simulation of aerodynamic flows because of their ability to discretize hyperbolic equations with high-order accuracy on complex geometries. However, they are rarely used for simulation of large scale problems because of their high cost, in particular for viscous problems, and the lack of efficient solvers. We present our Compact Discontinuous Galerkin (CDG) method for discretization of the viscous terms in the Navier-Stokes equations. The method is closely related to the Local Discontinuous Galerkin (LDG) method, but it uses a compact stencil, is computationally more efficient, and gives no additional fill-in with block-ILU(0) preconditioners. We solve the resulting linear systems using an efficient iterative method, based on GMRES preconditioned by a combination of block-ILU and coarse scale corrections. We obtain highest possible performance by using a block-based matrix format and optimized linear algebra libraries. To demonstrate the efficiency and accuracy of our methods we show examples of flows in complex geometries, high-Reynolds number flows, transonic flows, and aeroacoustics.
PDE Seminar
Department of Mathematics Colloquium
<--- 2006 Index