On the comparison of LES data-driven reduced order approaches for hydroacoustic analysis

In this work, Dynamic Mode Decomposition (DMD) and Proper Orthogonal (POD) methodologies are applied to hydroacoustic dataset computed using Large Eddy Simulation (LES) coupled with Ffowcs Williams Hawkings (FWH) analogy. First, a low-dimensional description of the flow fields is presented modal decomposition analysis. Sensitivity towards DMD POD bases truncation rank discussed, extensive provided demonstrate ability both algorithms reconstruct all spatial temporal frequencies necessary support accurate noise evaluation. Results show that while capable capture finer coherent structures in wake region for same amount employed modes, reconstructed exhibit smaller magnitudes global spatiotemporal errors compared counterparts. Second, separate set modes generated half snapshots into two data-driven reduced models respectively, based on mid cast Interpolation (PODI). regard, results confirm predictive character approaches sufficiently accurate, relative superiority PODI over ones. This infers that, discrepancies induced due interpolation relatively low by integration linear regression operations DMD, present setup. Finally, post processing analysis evaluation FWH acoustic signals utilizing fluid dynamic as input demonstrates efficient predicting noises.