Spacetime-spectral analysis of flowfields
Vilas J. Shinde · 2025
Read Spacetime-spectral analysis of flowfields by Vilas J. Shinde (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2512-20728.txt.
Introduction
Vilas J Shinde builds a spectral microscope that watches frequencies change in time, at the data's own resolution. Fourier analysis gives a signal's frequency content but assumes periodicity and discards when each tone was active. Short-time transforms and wavelets trade off time detail against frequency detail. Flowfields, especially turbulent ones with bursty extreme events, need better.
The proposed spectral mode split produces spectral-space and spectral-time modes, each ranked by both frequency and energy. Together they form a tone map resolved at the flowfield's own grid size rather than a windowed blur. Numerical cases show spacetime-tone analysis of flowfields, and the same split serves reduced-order modelling and denoising. Engineers studying unsteady flows gain a sharper tool for events that standard modal analysis smears away.
Catalog record
| CREATOR | Vilas J. Shinde |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 56 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2512-20728
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Spacetime-spectral analysis of flowfields Vilas J. Shinde Abstract The classical Fourier analysis of a time signal, in the discrete sense, provides the frequency content of signal under the assumption of periodicity. Although the original signal can be exactly recovered using an inverse transform, the time dependence of the spectrum remains inaccessible. There exist various time-frequency analysis techniques, such as the short time fast Fourier transform and wavelets, but those are fundamentally limited in achieving high resolution in both the time and frequency domains concurrently. For spatiotemporal flowfields, the frequency based modal decompositions generally provide spatial modes with a temporal counterpart that evolves at a constant frequency. However, an accurate time-local spectral contribution and its variation over time are highly desired to better understand the intermittent/extreme events, for instance, in turbulent flowfields. To this end, this paper presents a spectral mode decomposition that yields spectral-space and spectral-time modes, where the later along with the associated spectral energies provide a spectrogram that is at the resolution of flowfields. The spectral modes get both the frequency and energy ranks. The numerical examples demonstrate the use of technique not only in spacetime-frequency analysis but also in reduced-order modeling and denoising applications. keywords Spatiotemporal fields, Time frequency analysis, Fourier transform, Modal decomposition techniques † † affiliation: Department of Aerospace Engineering, Mississippi State University, Mississippi State, MS 39762, USA 1 Introduction Spatiotemporal fields are omnipresent in all branches of sciences, and for that matter humanities. The study of physical phenomena typically involves a set of variables ( e.g. , velocity, pressure, concentration, etc. in fluid sciences) with a spatial distribution that evolves in time; above all, it includes understanding, analysis, modeling, and prediction of these spatiotemporal variable fields. In fluid mechanics, turbulence is a prime example that exhibits complex spatiotemporal flowfields. Needless to say, turbulence plays a key role in transport phenomena, environmental flows, plasma physics, to mention a few. A turbulent flow is characterized by the presence of a wide range of length/time scales, comprising complex interactions across the scales ( 5 ; 43 ) . Further, it manifests spatiotemporal coherence, energy bursts, and intermittent/extreme events whose identification and understanding are of critical importance from practical standpoint. To this end, the spectral mode decomposition (SMD) presented in this paper facilitates the identification of spectral spatial coherence and its non-sinusoidal temporal dynamics, providing a highly resolved spectrogram. Time-frequency analysis methods have been extensively used in signal processing, primarily to identify time-local spectral events. A more popular Fourier transform (FT) when applied to the entire time domain provides overall frequency contributions; however, the time-local presence of frequencies is not retrieved. The Fourier transform consists of a constant amplitude sine/cosine waves, thus the time varying spectral information, i.e spectrogram, remains inaccessible. The spectrogram is commonly obtained by employing a short time Fourier transform (STFT) repeatedly. Generally, STFT spectrogram provides only the spectral magnitudes, losing track of the phase information; the reconstruction of the original signal based on spectrogram is an intricate process. To account for the frequency variation over time, the sine/cosine waves are modulated to change over time, leading to wavelets. It employs a specific multi-frequency kernel, known as a mother wavelet, to identify the time-local spectral presence. The more common Morlet is a combination of the sine and Gaussian functions. Wavelet transforms are effective for non-stationary signals, and are particularly effective with non-harmonic functions. Among several limitations of time-frequency analyses ( 10 ; 21 ) , the limitation of trade-off between the time and frequency can be significant, as resolving the lower frequencies require longer time duration. In addition, the time-frequency analyses are commonly employed to local/sparse spatial information and seldom to extract a highly-resolved spatial coherence. With the availability of scale-resolved experimental and numerical simulations of turbulent flows, numerous modal decomposition techniques have emerged in order to extract the spatial/temporal coherence in seemingly chaotic fluid motions ( 30 ; 42 ) . These techniques differ in their definition of the coherence and ranking of modes, as well as specific physical characteristics of their dynamics. The more fundamental modal decomposition techniques include: Fourier decomposition, proper orthogonal decomposition (POD) ( 20 ; 41 ) , and dynamics mode decomposition (DMD) ( 32 ; 31 ) . A common factor among all modal decomposition techniques is that they decompose spatiotemporal flowfields into spatial coherent patterns and their associated temporal dynamics. Although essentially these techniques build on Fourier transform/eigen-/singular-value decompositions, these bring out distinct spatiotemporal dynamics that pertains to their construct. Fourier spatial modes represent the spectral weights at specific frequency and the corresponding time coefficients are monotones of the associated frequencies. By definition, the temporal coefficients form an orthogonal set of temporal modes, whereas the spatial Fourier modes need not be orthogonal. POD is another popular technique that distills most energetic spatial/temporal modes for a given rank which is based on the energy. It employs the eigenvalue decomposition, leading to bi-orthonormal sets of space and time modes. Importantly, the POD modes may (and generally) comprise multiple frequencies in their spatiotemporal dynamics. The variance associated with the modes, i.e. the modal energy, defines the mode rank. Lastly, the DMD procedure seeks for spatial modes at a specific frequency and associated growth/decay rate. The temporal modes of DMD exhibit a single frequency that is growing or decaying in time at a constant rate. Similar to Fourier modes, the time modes are orthogonal by definition, whereas the spatial DMD modes need not be orthogonal. For stationary flows, the DMD modes show sustained dynamics or zero growth rate, similar to the Fourier modes. Nonetheless, the construct of Fourier and DMD modes are characteristically different; the Fourier modes comprise a constant spectral spatial coherence, while the DMD modes exhibit spectral spatial coherence that grows/decays with time. The FT/POD/DMD and their variants have been extremely useful in decoding the spatiotemporal dynamics, size reduction, and constructing efficient predictive models of complex turbulent flows. The two classes of techniques, namely the time-frequency analysis and the modal decomposition, separately seek to obtain, respectively, a well-resolved time-varying spectral contribution, and a converged spatial coherent patterns with associated temporal dynamics. In this regard, a modal decomposition technique with time-varying spectral dynamics is highly desired in order to investigate the spatiotemporal flowfields in a unified manner. The frequency based modal decomposition methods, i.e. FT, DMD, and spectral variants of POD ( 19 ; 46 ) , provide spatial modes corresponding to a unique frequency and associated temporal coefficients that evolve at a constant frequency. In turbulent spatiotemporal flowfields, it is crucial to know the time-local spectral contribution and its variation over time, which need not be at a fixed frequency that is associated with the mode ( 40 ; 22 ) . In this quest, a spectral mode decomposition procedure is developed in this paper, where the time coefficients that are associated with the spectral spatial modes capture the exact time-local modal contribution and its variation over time. The spectral spatial modes are essentially the Fourier modes at a unique frequency and with a specific spectral energy, which provides another measure for the modal rank. The proposed SMD method for the spacetime-frequency analysis of flowfields is demonstrated considering the following three numerical examples with increasing level of complexity: 1. transitional