Predicting burst events in a forced 2D flow: A wavelet-based analysis

Anagha Madhusudanan,Rich R. Kerswell
2024-03-08
Abstract:Predicting and perhaps mitigating against rare, extreme events in fluid flows is an important challenge. Due to the time-localized nature of these events, Fourier-based methods prove inefficient in capturing them. Instead this paper uses wavelet-based methods to understand the underlying patterns in a forced flow over a 2-torus which has intermittent high-energy burst events interrupting an ambient low energy quiet flow. Three methods are examined to predict burst events: (1) a straightforward energy tracking approach which acts as the benchmark; (2) a wavelet proper orthogonal decomposition (WPOD) which uncovers the key flow patterns seen in the quiet regions and the bursting episodes; and (3) a wavelet resolvent analysis (WRA) that reveals the forcing structures that amplifies the underlying flow patterns. Both the wavelet-based approaches succeed in producing better (earlier) predictions than a simple energy criterion, with the WRA-based technique always bettering WPOD. However, both also produce false positives and the improvement of the WRA approach over WPOD is not as substantial as anticipated. We conjecture that this is because the mechanism for the bursts in the flow studied is found to be largely modal associated with the unstable eigenfunction of the Navier-Stokes operator linearized around the mean flow. The forcing approach should deliver much better improvements over the WPOD approach for generically non-modal bursting mechanisms where there is a lag between the imposed forcing and the final response pattern.
Fluid Dynamics
What problem does this paper attempt to address?