FlowFit Data Assimilation and TrackFit Uncertainty Quantification Advances

Abstract

Recent advances to our FlowFit data assimilation method for reconstructing an incompressible flow field based on scattered velocity and acceleration data and the TrackFit uncertainty quantification & fitting methods for estimating accurate particle trajectories are summarized in this work. Both methods are cost function minimizations approaches. In order to estimate high quality particle trajectories from noisy data a Wiener-like filter approach has been adopted based on a simple physical model of particle motion and cubic splines. Estimating ideal fitting parameters such as the lowpass cutoff frequency via spectral analysis is a challenge in this scenario especially given particle trajectory data of possibly short tracks and the heavily non-uniform power distribution across all frequencies. Standard methods for spectral estimation would suffer from spectral leakage artefacts. To deal with these conditions a dedicated spectral estimation method has been developed that does not suffer from spectral leakage effects as much as standard approaches. It is based on an auto-regressive analysis combined with prefiltering (pre-whitening) of the signal we would like to analyze.