Convergence study of multi-field singular value decomposition for turbulence fields
The convergence behavior of the multi-field singular value decomposition (MFSVD) is systematically examined considering different examples. MFSVD extends conventional SVD by simultaneously decomposing multiple fields, enabling efficient representation of nonlinear correlations. Tests using artificial multi-scale matrices demonstrate that the method reproduces power-law-like singular value spectra and decomposes turbulent structures into orthogonal spatial modes. The method is then applied to turbulence fields, i.e. drift-wave turbulence, obtained from numerical simulations. While single-field SVD requires several thousand modes to accurately reconstruct turbulent fields due to the presence of a wide range of interacting scales, MFSVD exhibits significantly faster convergence for nonlinear quantities. In particular, quadratic quantities such as particle flux are recovered with substantially fewer modes. This improved convergence is attributed to the ability of MFSVD to incorporate phase correlations among multiple fields. These results highlight the potential of MFSVD for reduced-order modeling of multi-field turbulence with lower computational cost.
This page was built for publication: Convergence study of multi-field singular value decomposition for turbulence fields
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q7030016)
