Analysis of high-Reynolds-number lid-driven cavity flow using enhanced dynamic mode decomposition

Wang, Ping, Bai, Huisong, Peng, Yong ORCID: https://orcid.org/0000-0003-3377-6131, Zhou, Jianguo ORCID: https://orcid.org/0000-0002-4262-1898, Xu, Guangyao and Peng, Yuji (2025) Analysis of high-Reynolds-number lid-driven cavity flow using enhanced dynamic mode decomposition. Physics of Fluids, 37 (7). 075195. ISSN 1070-6631

Preview

Accepted Version Available under License Creative Commons Attribution. Download (2MB) | Preview

Abstract

To address the challenge of modal characterization of complex turbulent structures in high Reynolds number cavity flow, this study integrates the time integration contribution-dynamic mode decomposition (TIC-DMD) and sparsity-promoting DMD (SPDMD) as multi-scale analysis methods. Utilizing particle image velocimetry experimental data (Re = 5 × 105 and Re = 2 × 106), it comprehensively analyzes the dynamic characteristics and modal reconstruction performance of high Reynolds number cavity flow. The findings show that the TIC-DMD effectively extracts the dominant vortex structures through a time-domain energy integration mechanism. At Re = 5 × 105, it achieves 61.02% reduction in reconstruction error compared to SPDMD when using a high modal number (N = 246), significantly enhancing its ability to capture multi-scale turbulence. In addition, the SPDMD suppresses noise interference through sparsity constraints, achieving a reconstruction error of 0.0593 with a low modal number (N = 7), a 75.79% improvement over the standard DMD. Both methods' first-order modes consistently and stably reconstruct the dominant vortex structures of the flow field, while the standard DMD suffers from mode fragmentation due to noise sensitivity. Further analysis reveals that SPDMD excels at low modal numbers, whereas TIC-DMD offers superior stability and accuracy in flow field reconstruction as the modal number increases, particularly for high Reynolds number flows. The modal analysis framework developed in this study introduces a novel paradigm for modeling complex flows. The framework proposes to integrate experimental data with the large eddy simulation benchmark database, thereby advancing engineering applications in high Reynolds number flow control.