Scientific Computing Research Environment for the Mathematical Sciences (SCREMS)

点击翻译按钮获取中文摘要

ABSTRACTPI: Traian IliescuProposal: 0322852Scientific computing equipment is being requested for rapid solution of large eddy simulation (LES) models. In particular, the investigators study improved boundary conditions for complex engineering flows (one of the main hurdles in the development of LES). Current models suffer from failure to correctly capture subfilter-scale motion in the regions where boundary layer theory is not valid such as recirculation, separated flows, etc. The initial approach is to use approximate deconvolution methods for partial recovery of subfilter scale information. Additionally, the investigators intend to study the role LES may play in the development of reduced-order modeling for state estimation in feedback control. The ability for LES to compute large scale structures efficiently and accurately could be important for the real-time state estimation required for most flow control applications.The investigators intend to develop a scientific computing platform for addressing a number of modeling issues (such as appropriate boundary conditions and convolution kernels) for general three-dimensional turbulent flows. More mathematically sound models will lead to better understanding of flow phenomena such as density currents believed to have a significant influence on global ocean models. Better models are also important for atmospheric models and other geophysical flows. Furthermore, we intend to use this computational platform to address a number of practical optimization and control issues such as the optimal placement of control actuators and flow sensors, and the construction of fast (i.e. near real-time) reduced-order models for fluid systems.