Advection of sampling grids for efficient computation of trajectory-based quantities
Alessandro Rigazzi · Repository for Publications and Research Data (ETH Zurich) · 2008
Lagrangian methods in numerical flow visualization, such as the finite-time Lyapunov exponent, often require the integration of a very high number of trajectories. Usually one trajectory needs to be computed for each node of the sampling grid. This constitutes a very high cost especially for transient vector fields because there the trajectories need to get recomputed for each time step. This thesis addresses the problem in two ways. As a first improvement to the known methods we restrict sampling grids to areas of interest, where phenomena which are to be visualized appear, disregarding the rest of the domain. The second improvement is done by exploiting temporal coherence of trajectory-based quantities by reusing part of the trajectories. This is achieved by advecting the nodes of the sampling grid. As a case study, height ridges of finite-time Lyapunov exponent are extracted on synthetic Computational Fluid Dynamics datasets provided by partners of the ETHZ. Performances of the method are tested against those of a classic method for computation of such ridges.