91制片厂

Elizabeth Bradley, Department of Applied Mathematics, University of Colorado Boulder

Dynamics of data assimilation

Numerical solvers cannot track unmodelled effects like noise, and this becomes a particularly serious issue in complex nonlinear systems like fluid flows. One way to address this problem is to use observations of the system being modelled---e.g., velocity field data from a fluid flow---to periodically correct the simulation. Changing the state variables of a running solver, however, can set off all sorts of unfriendly dynamics, so this "data assimilation" process is not trivial. Moreover, gathering and processing data can be costly, both computationally and financially, so one should apply corrections only when and where they are needed---e.g., where the flow gradients are high. In this talk, I will discuss some of these issues in the context of a point-vortex model of an instrumental laboratory fluid flow: a planar air jet.