Improving Flow Analysis Using In Situ Lagrangian Techniques
Sudhanshu Sane
Committee: Hank Childs (chair), Boyana Norris, Michal Young
Directed Research Project(Jan 2018)
Keywords: Flow Visualization; Lagrangian Analysis; In Situ

We introduce a new approach, consisting of two complimentary techniques, to perform in situ Lagrangian flow analysis. Previous work had placed constraints on the placement and duration of basis flows, namely putting them in regular positions, and forcing all basis flows to terminate at the same time. With our work, we relax these constraints in order to achieve better accuracy. Our first technique guides the creation of basis flows. It differs from previous work in that we allow the duration of basis flows to vary. The varying duration allows for us to create long-lived basis flows, which improves accuracy by preventing error propagation when interpolating new particle trajectories during post hoc analysis. Our technique also differs from previous work in that we identify regions of greatest need, via Delaunay triangulation, and then place basis flow seeds in these regions. Our second technique is for interpolation of basis flows of arbitrary seed placement and durations: this technique is necessary since our first technique generates such data and since there is no previous work on interpolation of basis flows with arbitrary durations. We show that our complementary techniques produce improved accuracy and demonstrate the potential for reduced data storage. This means we can provide improved accuracy with same or less storage, or same accuracy with less storage.