Research on spectral estimation parameters for application of spectral proper orthogonal decomposition in train wake flows

Abstract

Coherent structures in surrounding flows around ground vehicles play an important role in characterizing their aerodynamic features. However, due to restrictions of traditional reduced-order models, extracting physically meaningful coherent structures from turbulent flows with massive separation still remains a challenging issue. The spectral proper orthogonal decomposition (SPOD), which extracts modes optimally representing the space-time flow statistics, enables the feasibility of further modeling and control of vehicle aerodynamic features. This study intends to investigate the influence of spectral estimation parameters on SPOD results, so as to serve as fundamentals for future works into this topic. The time-resolved pressure field obtained from a large-eddy simulation considering a generic high-speed train is used as the snapshot database. Spectral estimation parameters including block number, frequency resolution, and cutoff frequency are, respectively, discussed to quantify their impacts on both SPOD spectra and modes. The results reveal that, with the increasing of block number, higher reliability and accuracy of SPOD prediction can be achieved, with the block number of 20–30 that leads both requirements of efficiency and precision. The frequency axis with finer resolution reproduces more detailed spectral information, with the eigenvalue distribution and spatial distribution of mode energy under acceptable accuracy when dimensionless frequency resolution reaches 0.025. Moreover, the reducing of cutoff frequency results in increasing unresolved energy content, which will be distributed mostly near the corresponding cutoff frequency, and more minor scaled spatial structures in SPOD modes. The findings and approaches could also work as references for wider application field of SPOD approach.