Pith. sign in

REVIEW 4 major objections 5 minor 56 references

Pressure Field Reconstruction with SIREN: A Mesh-Free Approach for Image Velocimetry in Complex Noisy Environments

T0 review · 4 major / 5 minor · reviewed 2026-08-10 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read This paper claims that a sinusoidal representation network can reconstruct pressure fields from image velocimetry data without a mesh, with architecture choice acting as a tunable noise filter, outperforming OS-MODI and GFI on…

desk verdict A useful, well-benchmarked application of SIREN to pressure reconstruction, with real promise for LPT, but the headline noise-filtering claim is demonstrated only on white noise and the paper needs code and held-out tuning before I'd rely on it. read the letter →

arxiv 2501.17987 v1 pith:FKE3AABO submitted 2025-01-29 cs.CV physics.flu-dyn

Renato F. Miotto , William R. Wolf , Fernando Zigunov This is my paper · ORCID
classification cs.CVphysics.flu-dyn
keywords pressurereconstructionSIRENimplicitneuralrepresentationparticleimagevelocimetrymesh-freemethodsgradientintegrationnoisefilteringunstructuredmeshes
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper proposes using a sinusoidal representation network (SIREN) to reconstruct pressure fields from image velocimetry data by integrating the pressure gradient directly, without building a mesh. The authors argue that because the network maps coordinates to pressure through sine activations, it can be initialized and sized so that its frequency bandlimit acts as a filter, removing noise from the measured velocity field while preserving the true pressure signal. On a randomly seeded unstructured mesh, the method reports a relative mean absolute error of 0.0031 versus 0.0213 for one-shot matrix omnidirectional integration and 0.0401 for Green's function integration, and it resolves spectral content roughly an order of magnitude beyond those methods. If correct, the approach turns architecture choice into a denoising knob and removes mesh quality as a source of error for particle tracking and complex geometries.

What carries the argument

The load-bearing object is the SIREN: a fully connected network with sine activation functions, trained with automatic differentiation so that $\nabla_x \Phi(x)$ approximates the measured pressure gradient from $\nabla p = -\rho Du/Dt + \mu \nabla^2 u$. Its frequency behavior follows a "frequency factoring" mechanism: the first layer acts as a Fourier feature map, and each sinusoidal neuron generates integer linear combinations of input frequencies with bounded amplitudes, so the network's spectral content is set at initialization by the input scaling $\omega_0 = c k_{\min}$ with $2 \lesssim c \lesssim 3$ and the hidden-layer scalings $\omega_i \approx 20$-$30$. This is what lets the same network act as an integrator and as a tunable bandpass filter.

What would settle it

Add correlated, boundary-dominated noise with spectral content inside the signal band to the JHTDB velocity field and compare SIREN RMAE with OS-MODI; if SIREN no longer beats the mesh-based solvers, the architecture-as-filter claim fails. Alternatively, compare SIREN's extrapolated wall pressure against the pressure transducers in the bluff-body experiment; the authors already warn that extrapolated values may be non-physical.

Watch

Extended reading notes

Core claim

The central claim is that a SIREN, trained so that the gradient of its output matches the pressure gradient implied by the momentum equation, is a mesh-free pressure integrator whose architecture controls the trade-off between resolution and noise sensitivity. The paper shows that the input frequency scale can be set from the smallest wavenumber of the domain, and that fewer hidden layers and nodes suppress high-frequency noise while deeper or wider networks resolve finer scales. In noise-free structured-mesh cases the method matches OS-MODI; under added velocity noise the shallow 1x64/20-30 network has the lowest error of all tested methods; on an ill-conditioned random mesh it is several times more accurate than OS-MODI and GFI. The authors present the bluff-body experiment as a qualitative demonstration only, and explicitly warn that extrapolated outer solutions should not be relied upon.

Load-bearing premise

The noise-robustness claim assumes that velocity-derived pressure-gradient noise lives at higher frequencies than the true pressure signal, so shrinking the network's bandlimit removes noise without removing signal.

Editorial extensions

If this is right

  • Users can integrate pressure from scattered particle-tracking or PIV points with no mesh generation, sidestepping ill-conditioned cells entirely.
  • Choosing a shallower or narrower SIREN trades high-wavenumber detail for noise rejection, giving the experimenter a single lever to tune denoising.
  • In the noiseless spectral test the SIREN resolves wavenumbers roughly ten times higher than OS-MODI and GFI, so small-scale pressure features become accessible.
  • Extrapolation outside the measured region is unreliable and may return non-physical values, so the method should be restricted to the data domain.
  • The method's training cost (about 16 minutes for the deep configuration on the benchmark) is moderate, though the authors do not claim a head-to-head optimized comparison.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Editorial extension: if real PIV noise is broad-band and concentrated at walls rather than spectrally separated from the signal, the architecture-as-filter advantage could shrink; a test with correlated noise would settle the question.
  • Editorial extension: the same frequency-scaling rule could apply to other inverse problems that integrate a measured gradient, such as force or heat-flux reconstruction from particle data.
  • Editorial extension: the $c$ range of 2-3 for $\omega_0$ is empirical, so a data-driven selection of the bandlimit from validation error would make the method more portable.
  • Editorial extension: volumetric Lagrangian particle tracking with voids is the natural next case, since the mesh-free property should matter most where connectivity is absent.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 5 minor

Summary. The paper proposes a mesh-free pressure reconstruction method based on SIREN, a sinusoidal implicit neural representation, and compares it against OS-MODI and GFI on a two-dimensional slice of the JHTD isotropic turbulence dataset with synthetic Gaussian velocity noise, on structured and unstructured meshes, and on a single experimental PIV wake dataset. The central claims are that SIREN is intrinsically mesh-free, that its architecture can be tuned to filter velocity-derived noise, and that it outperforms OS-MODI and GFI on ill-conditioned unstructured meshes. The authors introduce a frequency-scaling rule omega0 = c*kmin with c in [2,3] and show that shallower networks with smaller omega0 reduce noise sensitivity.

Significance. If validated, the proposed SIREN approach would offer a practically useful mesh-free alternative for pressure integration from PIV/LPT data, particularly in complex geometries where mesh generation is difficult. The benchmark protocol is a strength: all methods share the same source term and are compared directly against DNS ground truth. The spectral analysis in Fig. 4 is informative and suggests that SIREN can resolve substantially higher wavenumbers than the two mesh-based methods. However, the headline noise-robustness claim relies on an unverified spectral-separability premise and is demonstrated only with isotropic Gaussian noise on a single turbulence dataset, with no error bars or repeated realizations. The experimental case is qualitative and the authors explicitly caution that extrapolated SIREN values may be non-physical. The significance is therefore conditional pending stronger validation of the filtering mechanism.

major comments (4)
  1. [Section 3.1, Fig. 3] The RMAE curves in Fig. 3 are based on a single realization for each noise level, with no confidence intervals or repeated trials. The conclusion that SIREN 1x64/20-30 "emerges as the most effective" at 10% velocity error could be within run-to-run variability of the random noise draw. Without multiple noise realizations or an error analysis, the statistical significance of the difference between SIREN and OS-MODI is not established, and this is load-bearing for the abstract's claim of superior performance in noisy environments.
  2. [Sections 3.1 and 3.3] The noise-filtering claim rests on the premise that noise in the pressure-gradient source term occupies frequencies above the signal band, so that a SIREN with reduced architecture and omega0 = c*kmin filters noise without attenuating signal. The experiments add only isotropic Gaussian velocity noise, which is spectrally white. Real PIV/LPT error is spatially correlated (interrogation-window smoothing, peak locking, boundary bias) and often strongest near boundaries, so it may overlap the signal band. The single experimental case in Section 3.3 shows only qualitative agreement; the authors state that the SIREN outer solution "should not be relied upon" and "may yield non-physical results." This is insufficient to confirm the architecture-based filtering mechanism under realistic noise conditions.
  3. [Section 3.1, footnote 1] The bandlimit rule omega0 = c*kmin with c in [2,3] is calibrated on the same JHTD dataset used for the performance evaluation, and the paper reports only two subdomains of that dataset plus one experimental case. This is a test-set tuning concern: the architecture and scaling choices are selected on the same data used to demonstrate their benefit. A validation on an independent flow case (for example, the Taylor-Green vortex or a different turbulence database) would be needed to show that the rule and the reported noise robustness generalize beyond the tuned case.
  4. [Section 3.2] The unstructured-mesh comparison uses a single random seed for each mesh type (Perturbed and Random). The claimed advantage of SIREN on the random mesh (RMAE 0.0031 versus 0.0213 for OS-MODI and 0.0401 for GFI) might be sensitive to the particular point distribution drawn. Without multiple independent seeds or a perturbed-seed analysis, the robustness of this conclusion is not demonstrated. Since the mesh-free property is one of the two central contributions, this point needs a stronger statistical basis.
minor comments (5)
  1. [Equation (4)] The loss integral is written as ∫_Ω ||∇xΦ(x) - ∇p(x)|| dx, but Ω is a subset of R² and the integrand is a 2D vector; the notation should use dA or dΩ, and the norm without a subscript could be ambiguous.
  2. [Section 3.1] The phrase "results (not shown here for brevity)" for the epoch count is a missing support: either include the plot or state the saturation behavior in one sentence so the reader can assess the convergence criterion.
  3. [Figure 5] The caption describes "transparency levels" for the number of nodes, but the text refers to lighter, intermediate, and darker shades; this is confusing. The legend should be clarified to indicate whether opacity or color is used.
  4. [Section 3.1] The acronym RMAE is used without definition before first use; it should be spelled out as relative mean absolute error at its first occurrence.
  5. [Section 2.1] The initialization w0 ∼ U(−1/n, 1/n) uses n without explicitly defining it as the dimension of the input coordinates; this should be stated for clarity.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity: the pressure reconstruction is a genuine integration fit to externally supplied gradient data, and the self-citations are non-load-bearing.

full rationale

The central derivation is self-contained: SIREN is trained to minimize Eq. (4), a loss on the externally supplied pressure-gradient field ∇p(x) (derived from velocity data), not on the target pressure itself; the pressure field is then obtained by the network's implicit integration through automatic differentiation. This is a standard inverse-problem formulation, so the output is not equivalent to the input by definition. The architecture-based noise filtering is an emergent spectral property (supported by transfer-function measurements in Figs. 4-5), and although the ω0 = c kmin rule and per-section layer/node counts are selected with knowledge of the same JHTD benchmarks used for the headline comparisons, this is test-set hyperparameter selection rather than a fitted parameter renamed as a prediction; the comparison to OS-MODI and GFI is still an empirical performance claim, not a derivation forced by construction. The paper's self-citations (Zigunov & Charonko for the OS-MODI baseline, Zigunov et al. for material-acceleration processing and the experimental dataset) are methodological and data-source references, not load-bearing premises for the SIREN claim, and no uniqueness or existence theorem is imported from the authors' prior work. The disclosed limitation that extrapolated outer solutions 'should not be relied upon' is honest and does not create circularity. Thus the circularity burden is minimal; the main caveat is generalizability of the empirically tuned bandlimit rule, which is a robustness/correctness concern rather than a circularity concern.

Assumptions & free parameters 4 free parameters · 5 assumptions · 0 invented entities

The central claim rests on hand-set frequency scalings and architecture choices (the actual mechanism of the claimed denoising), on an imported theory of SIREN frequency composition (Novello et al.), on a synthetic Gaussian noise model of unproven realism, and on benchmark ground truth from DNS. No new physical entities are introduced. The free parameters are disclosed but are tuned on the same benchmarks used for evaluation.

free parameters (4)
  • Input frequency scaling omega0 = 20, 30, 5, 10, or 90 per case
    Set as omega0 = c*kmin with c between 2 and 3, found empirically in Section 3.1; it sets the network bandlimit and is the main lever behind the claimed noise filtering.
  • Hidden-layer frequency scaling omega_i>0 = 5 to 30 (5, 20, 30 used)
    Hand-tuned per case; higher values emphasize high frequencies and can destabilize training; part of the accuracy-versus-denoising trade documented in Figs. 3 and 5.
  • Network architecture (layers x nodes) = 1x64, 1x128, 1x256, 2x256, 3x256, 2x64
    Different architectures are selected in each section to produce each claimed property; this is a free choice made with knowledge of the benchmark ground truth.
  • Training schedule = ADAM, lr = 3e-5, 2000 epochs
    Fixed across cases; the epoch count was chosen 'to secure a robust approximation', with the authors noting fewer epochs also work.
assumptions (5)
  • domain assumption SIREN frequency-factoring theory: hidden neurons generate frequencies that are integer linear combinations of input frequencies with bounded amplitudes (Novello et al. [26]).
    Imported without re-derivation to justify the bandlimit-control mechanism in Section 3.1; the mechanism behind the claimed noise filtering depends on this theory.
  • domain assumption The 2D pressure reconstruction only needs the x and y components of the material acceleration.
    Section 3: 'only the x and y components of the material acceleration are fitted by the network'; the missing out-of-plane term is acknowledged as a source of common error for all methods.
  • domain assumption Synthetic noise model: Gaussian perturbation of the 3D velocity field, amplitude based on the maximum velocity in the slice, direction uniform over a sphere.
    Section 2.2; this noise model drives the headline noise-robustness comparisons, and its realism for real PIV/LPT data is not established.
  • standard math OS-MODI is equivalent to solving a pressure Poisson equation with Neumann boundary conditions via conjugate-gradient regularization (Pryce et al. [29, 30]).
    Appendix and Section 1; used to interpret baseline behavior; the equivalence is taken from the cited proofs.
  • domain assumption The Delaunay-triangulated meshes from the two point seeds are appropriate comparison targets, with the ground truth interpolated onto the same points.
    Section 3.2; aliasing from resampling 10,000 to 2,000 points is acknowledged and mitigated by comparing against the interpolated ground truth.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Pressure Field Reconstruction with SIREN: A Mesh-Free Approach for Image Velocimetry in Complex Noisy Environments." pith.science (2026). https://pith.science/paper/FKE3AABO

@misc{pith2026250117987,
  author       = {Pith},
  title        = {Pith review of: Pressure Field Reconstruction with SIREN: A Mesh-Free Approach for Image Velocimetry in Complex Noisy Environments},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/FKE3AABO}},
  note         = {Machine review of arXiv:2501.17987}
}
read the original abstract

This work presents a novel approach for pressure field reconstruction from image velocimetry data using SIREN (Sinusoidal Representation Network), emphasizing its effectiveness as an implicit neural representation in noisy environments and its mesh-free nature. While we briefly assess two recently proposed methods - one-shot matrix-omnidirectional integration (OS-MODI) and Green's function integral (GFI) - the primary focus is on the advantages of the SIREN approach. The OS-MODI technique performs well in noise-free conditions and with structured meshes but struggles when applied to unstructured meshes with high aspect ratio. Similarly, the GFI method encounters difficulties due to singularities inherent from the Newtonian kernel. In contrast, the proposed SIREN approach is a mesh-free method that directly reconstructs the pressure field, bypassing the need for an intrinsic grid connectivity and, hence, avoiding the challenges associated with ill-conditioned cells and unstructured meshes. This provides a distinct advantage over traditional mesh-based methods. Moreover, it is shown that changes in the architecture of the SIREN can be used to filter out inherent noise from velocimetry data. This work positions SIREN as a robust and versatile solution for pressure reconstruction, particularly in noisy environments characterized by the absence of mesh structure, opening new avenues for innovative applications in this field.

Figures

Figures reproduced from arXiv: 2501.17987 by the authors.

Figure 6
Figure 6. The aspect ratio, defined as the ratio between [PITH_FULL_IMAGE:figures/full_fig_p009_6.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

56 extracted references · 48 canonical work pages

  1. [1]

    Iliass Azijli, Andrea Sciacchitano, Daniele Ragni, Artur Palha, and Richard P. Dwight. A posteri- ori uncertainty quantification of piv-based pressure data. Experiments in Fluids, 57(5):72, Apr 2016. ISSN 1432-1114. doi: 10.1007/s00348-016-2159-z

    work page doi:10.1007/s00348-016-2159-z 2016

  2. [2]

    Seeing implicit neural rep- resentations as fourier series, 2021

    Nuri Benbarka, Timon H¨ ofer, Hamd ul-moqeet Riaz, and Andreas Zell. Seeing implicit neural rep- resentations as fourier series, 2021

    work page 2021

  3. [3]

    Steven J. Beresh. Time-resolved particle image ve- locimetry. Measurement Science and Technology, 32(10):102003, Jun 2021. ISSN 0957-0233. doi: 10.1088/1361-6501/ac08c5

    work page doi:10.1088/1361-6501/ac08c5 2021

  4. [4]

    Calicchia, Rajat Mittal, Jung-Hee Seo, and Rui Ni

    Michael A. Calicchia, Rajat Mittal, Jung-Hee Seo, and Rui Ni. Reconstructing the pressure field around swimming fish using a physics-informed neural network. Journal of Experimental Biology, 226(8):jeb244983, Apr 2023. ISSN 0022-0949. doi: 10.1242/jeb.244983

    work page doi:10.1242/jeb.244983 2023

  5. [5]

    Carter and Bharathram Ganap- athisubramani

    Douglas W. Carter and Bharathram Ganap- athisubramani. Data-driven determination of low- frequency dipole noise mechanisms in stalled air- foils. Experiments in Fluids, 64(2):41, Feb 2023. ISSN 1432-1114. doi: 10.1007/s00348-023-03577-z

    work page doi:10.1007/s00348-023-03577-z 2023

  6. [6]

    Charonko, Cameron V

    John J. Charonko, Cameron V. King, Barton L. Smith, and Pavlos P. Vlachos. Assessment of pressure field calculations from particle image ve- locimetry measurements. Measurement Science and Technology, 21(10):105401, Aug 2010. ISSN 0957-

    work page 2010

  7. [7]

    Dabiri, Sanjeeb Bose, Brad J

    John O. Dabiri, Sanjeeb Bose, Brad J. Gemmell, Sean P. Colin, and John H. Costello. An algorithm to estimate unsteady and quasi-steady pressure fields from velocity field measurements. Journal of Experimental Biology, 217(3):331–336, Feb 2014. ISSN 0022-0949. doi: 10.1242/jeb.092767

    work page doi:10.1242/jeb.092767 2014

  8. [8]

    de Kat and B

    R. de Kat and B. W. van Oudheusden. Instan- taneous planar pressure determination from piv in turbulent flow. Experiments in Fluids, 52(5):1089– 1106, May 2012. ISSN 1432-1114. doi: 10.1007/ s00348-011-1237-5

    work page 2012

Show all 56 references
  1. [9]

    Whitehead, and Zhao Pan

    Matthew Faiella, Corwin Grant Jeon Macmillan, Jared P. Whitehead, and Zhao Pan. Error propa- gation dynamics of velocimetry-based pressure field calculations (2): on the error profile. Measurement Science and Technology, 32(8):084005, May 2021. ISSN 0957-0233. doi: 10.1088/136...

    2021 doi

  2. [10]

    Comparative assessment for pressure field reconstruction based on physics-informed neural network

    Di Fan, Yang Xu, Hongping Wang, and Jinjun Wang. Comparative assessment for pressure field reconstruction based on physics-informed neural network. Physics of Fluids, 35(7):077116, Jul 2023. ISSN 1070-6631. doi: 10.1063/5.0157753

    2023 doi

  3. [11]

    A modified green’s function approach to particle image velocimetry pressure estimates with an application to microenergy harvesters

    Oleg Goushcha, Yiannis Andreopoulos, and Peter Ganatos. A modified green’s function approach to particle image velocimetry pressure estimates with an application to microenergy harvesters. Archive of Applied Mechanics, 93(3):1217–1239, Mar 2023. ISSN 1432-0681. doi: 10.1007/s0...

    2023 doi

  4. [12]

    Pedersen

    Atle Jensen and Geir K. Pedersen. Optimization of acceleration measurements using piv. Measurement Science and Technology, 15(11):2275, Oct 2004. ISSN 0957-0233. doi: 10.1088/0957-0233/15/11/ 013

    2004 doi

  5. [13]

    Gomit, T

    Young Jin Jeon, G. Gomit, T. Earl, L. Chatel- lier, and L. David. Sequential least-square recon- struction of instantaneous pressure field around a body from tr-piv. Experiments in Fluids, 59(2): 27, Jan 2018. ISSN 1432-1114. doi: 10.1007/ s00348-018-2489-0

    2018

  6. [14]

    Weakened acute type condition for tetra- hedral triangulations and the discrete maximum principle

    Sergey Korotov, Michal Kˇ r ´ ıˇ zek, and Pekka Neit- taanm¨ aki. Weakened acute type condition for tetra- hedral triangulations and the discrete maximum principle. Mathematics of Computation, 70:107– 119, 2001

    2001

  7. [15]

    Numba: a llvm-based python jit compiler

    Siu Kwan Lam, Antoine Pitrou, and Stanley Seib- ert. Numba: a llvm-based python jit compiler. In Proceedings of the Second Workshop on the LL VMCompiler Infrastructure in HPC, LL VM ’15, New York, NY, USA, 2015. Association for Com- puting Machinery. ISBN 9781450340052. doi: 1...

    2015

  8. [16]

    A public tur- bulence database cluster and applications to study 14 Renato F

    Yi Li, Eric Perlman, Minping Wan, Yunke Yang, Charles Meneveau, Randal Burns, Shiyi Chen, Alexander Szalay, and Gregory Eyink. A public tur- bulence database cluster and applications to study 14 Renato F. Miotto et al. lagrangian evolution of velocity increments in tur- bulenc...

  9. [17]

    Instantaneous pres- sure and material acceleration measurements using a four-exposure piv system

    Xiaofeng Liu and Joseph Katz. Instantaneous pres- sure and material acceleration measurements using a four-exposure piv system. Experiments in Fluids, 41(2):227–240, Aug 2006. ISSN 1432-1114. doi: 10.1007/s00348-006-0152-7

    2006 doi

  10. [18]

    Error propagation from the piv-based pressure gradient to the integrated pressure by the omnidirectional integration method

    Xiaofeng Liu and Jose Roberto Moreto. Error propagation from the piv-based pressure gradient to the integrated pressure by the omnidirectional integration method. Measurement Science and Technology, 31(5):055301, Feb 2020. ISSN 0957-

    2020

  11. [19]

    Moreto, and Seth Siddle- Mitchell

    Xiaofeng Liu, Jose R. Moreto, and Seth Siddle- Mitchell. Instantaneous Pressure Reconstruction from Measured Pressure Gradient using Rotating Parallel Ray Method, pages 1–8. AIAA SciTech Forum. American Institute of Aeronautics and As- tronautics, Jan 2016. doi: 10.2514/6.2016-1049. 0

    2016 doi

  12. [20]

    Joaquim R.R.A. Martins. Aerodynamic de- sign optimization: Challenges and perspectives. Computers & Fluids, 239:105391, 2022. ISSN 0045-

    2022

  13. [21]

    doi: 10.1088/1361-6501/ab6c28

    doi

  14. [22]

    A delaunay based numer- ical method for three dimensions: generation, for- mulation, and partition

    Gary L Miller, Dafna Talmor, Shang-Hua Teng, and Noel Walkington. A delaunay based numer- ical method for three dimensions: generation, for- mulation, and partition. In Proceedings of the twenty-seventhannual ACM symposium on Theory of computing, pages 683–692, 1995

    1995

  15. [23]

    N. J. Neeteson, S. Bhattacharya, D. E. Ri- val, D. Michaelis, D. Schanz, and A. Schr¨ oder. Pressure-field extraction from lagrangian flow mea- surements: first experiences with 4d-ptv data. Experiments in Fluids, 57(6):102, Jun 2016. ISSN 1432-1114. doi: 10.1007/s00348-016-2170-4

    2016 doi

  16. [24]

    Neeteson and David E

    Nathan J. Neeteson and David E. Rival. Pressure- field extraction on unstructured flow data using a voronoi tessellation-based networking algorithm: a proof-of-principle study. Experiments in Fluids, 56 (2):44, Feb 2015. ISSN 1432-1114. doi: 10.1007/ s00348-015-1911-0

    2015

  17. [25]

    Opti- mization of planar piv-based pressure estimates in laminar and turbulent wakes

    Jeffrey McClure and Serhiy Yarusevych. Opti- mization of planar piv-based pressure estimates in laminar and turbulent wakes. Experiments in Fluids, 58(5):62, Apr 2017. ISSN 1432-1114. doi: 10.1007/s00348-017-2337-7

    2017 doi

  18. [26]

    Taming the frequency factory of sinusoidal net- works, 2024

    Tiago Novello, Diana Aldana, and Luiz Velho. Taming the frequency factory of sinusoidal net- works, 2024

    2024

  19. [27]

    Zhao Pan, Jared Whitehead, Scott Thomson, and Tadd Truscott. Error propagation dynamics of piv-based pressure field calculations: How well does the pressure poisson solver perform inher- ently? Measurement Science and Technology, 27 (8):084012, Jul 2016. ISSN 0957-0233. doi: 1...

    2016

  20. [28]

    Percin, M

    M. Percin, M. Vanierschot, and B. W. van Oud- heusden. Analysis of the pressure fields in a swirling annular jet flow. Experiments in Fluids, 58(12): 166, Nov 2017. ISSN 1432-1114. doi: 10.1007/ s00348-017-2446-3

    2017

  21. [29]

    Whitehead, Geordie Richards, Barton L

    Mingyuan Nie, Jared P. Whitehead, Geordie Richards, Barton L. Smith, and Zhao Pan. Error propagation dynamics of piv-based pressure field calculation (3): what is the minimum resolvable pressure in a reconstructed field? Experiments in Fluids, 63(11):168, Oct 2022. ISSN 1432-1...

    2022 doi

  22. [30]

    Re- visit liu and katz (2006) and zigunov and charonko (2024b), part (i): on the equivalence of the omnidi- rectional integration and the pressure poisson equa- tion, 2024

    Connor Pryce, Lanyu Li, and Zhao Pan. Re- visit liu and katz (2006) and zigunov and charonko (2024b), part (i): on the equivalence of the omnidi- rectional integration and the pressure poisson equa- tion, 2024

    2006

  23. [31]

    On the spectral bias of neural networks

    Nasim Rahaman, Aristide Baratin, Devansh Arpit, Felix Draxler, Min Lin, Fred Hamprecht, Yoshua Bengio, and Aaron Courville. On the spectral bias of neural networks. In Kamalika Chaudhuri and Ruslan Salakhutdinov, editors, Proceedings of the 36th International Conference on Mac...

  24. [32]

    Wiswall, Alexander G

    Nazmus Sakib, Zhao Pan, James T. Wiswall, Alexander G. Mychkovsky, and Barton L. Smith. Experimental investigation on the effect of bound- ary conditions and spatial resolution on pressure from piv for oscillating internal flow. Experiments in Fluids, 65(3):28, Feb 2024. ISSN ...

    2024 doi

  25. [33]

    Pryce and Z

    C. Pryce and Z. Pan. On the equivalency of the omni-directional integration and the poisson equation for pressure field reconstruction based on image velocimetry data. In 1000 Islands Fluid Dynamics Meeting, 2024

    2024

  26. [34]

    Optimal pres- sure reconstruction based on planar particle im- age velocimetry and sparse sensor measurements

    Roshan Shanmughan, Pierre-Yves Passaggia, Nico- las Mazellier, and Azeddine Kourta. Optimal pres- sure reconstruction based on planar particle im- age velocimetry and sparse sensor measurements. Experiments in Fluids, 61(11):229, Oct 2020. ISSN 1432-1114. doi: 10.1007/s00348-0...

    2020 doi

  27. [35]

    Vincent Sitzmann, Julien N. P. Martel, Alexan- der W. Bergman, David B. Lindell, and Gordon Wetzstein. Implicit neural representations with pe- riodic activation functions, 2020

    2020

  28. [36]

    Pietro Sperotto, Sandra Pieraccini, and Miguel A. Mendez. A meshless method to compute pres- sure fields from image velocimetry. Measurement Science and Technology, 33(9):094005, Jun 2022. ISSN 0957-0233. doi: 10.1088/1361-6501/ac70a9

    2022 doi

  29. [37]

    Savelli, A-J

    N. Savelli, A-J. Buchner, and J. Westerweel. Force reconstruction from piv for an accelerating plate. In 15th International Symposium on Particle Image Velocimetry – ISPIV 2023, 2023

    2023

  30. [38]

    J. W. Van der Kindere, A. Laskari, B. Ganap- athisubramani, and R. de Kat. Pressure from 2D snapshot PIV. Experiments in Fluids, 60 (2):32, 2019. ISSN 1432-1114. doi: 10.1007/ s00348-019-2678-5

    2019

  31. [39]

    P. L. van Gent, D. Michaelis, B. W. van Oud- heusden, P.-´E. Weiss, R. de Kat, A. Laskari, Y. J. Jeon, L. David, D. Schanz, F. Huhn, S. Gese- mann, M. Novara, C. McPhaden, N. J. Neeteson, D. E. Rival, J. F. G. Schneiders, and F. F. J. Schrijer. Comparative assessment of pressu...

    2017

  32. [40]

    Lagrangian and eulerian pressure field evaluation of rod-airfoil flow from time-resolved tomographic piv

    Daniele Violato, Peter Moore, and Fulvio Scarano. Lagrangian and eulerian pressure field evaluation of rod-airfoil flow from time-resolved tomographic piv. Experiments in Fluids, 50(4):1057–1070, Apr 2011. ISSN 1432-1114. doi: 10.1007/s00348-010-1011-0

    2011 doi

  33. [41]

    Srinivasan, Ben Mildenhall, Sara Fridovich-Keil, Nithin Raghavan, Utkarsh Singhal, Ravi Ramamoorthi, Jonathan T

    Matthew Tancik, Pratul P. Srinivasan, Ben Mildenhall, Sara Fridovich-Keil, Nithin Raghavan, Utkarsh Singhal, Ravi Ramamoorthi, Jonathan T. Barron, and Ren Ng. Fourier features let net- works learn high frequency functions in low dimen- sional domains. In Advances in Neural Inf...

    2020

  34. [42]

    W. R. Wolf and S. K. Lele. Wideband fast multipole boundary element method: Application to acoustic scattering from aerodynamic bodies. International Journal for Numerical Methods in Fluids, 67(12):2108–2129, 2011. doi: https://doi. org/10.1002/fld.2486

    2011 doi

  35. [43]

    Pressure reconstruction from the measured pressure gradient using gaussian process regression, 2023

    Zejian You, Qi Wang, and Xiaofeng Liu. Pressure reconstruction from the measured pressure gradient using gaussian process regression, 2023

    2023

  36. [44]

    Jiacheng Zhang, Sayantan Bhattacharya, and Pav- los P. Vlachos. Uncertainty of piv/ptv based eu- lerian pressure estimation using velocity uncer- tainty. Measurement Science and Technology, 33 (6):065303, Mar 2022. ISSN 0957-0233. doi: 10. 1088/1361-6501/ac56bf

    2022

  37. [45]

    Charonko

    Fernando Zigunov and John J. Charonko. A fast, matrix-based method to perform omnidirectional pressure integration. Measurement Science and Technology, 35(6):065302, Mar 2024. ISSN 0957-

    2024

  38. [46]

    Green’s function inte- gral method for pressure reconstruction from mea- sured pressure gradient and the interpretation of omnidirectional integration

    Qi Wang and Xiaofeng Liu. Green’s function inte- gral method for pressure reconstruction from mea- sured pressure gradient and the interpretation of omnidirectional integration. Physics of Fluids, 35 (7), jul 2023. ISSN 1089-7666. doi: 10.1063/5. 0156777. URL http://dx.doi.org...

    2023 doi

  39. [47]

    Reynolds number and slant angle effects on the flow over a slanted cylinder afterbody

    Fernando Zigunov, Prabu Sellappan, and Farrukh Alvi. Reynolds number and slant angle effects on the flow over a slanted cylinder afterbody. Journal of Fluid Mechanics, 893:A11, 2020. ISSN 0022-

    2020

  40. [48]

    face-crossing

    Fernando Zigunov, Prabu Sellappan, and Far- rukh Alvi. A bluff body flow control experiment with distributed actuation and genetic algorithm- based optimization. Experiments in Fluids, 63 (1):23, Jan 2022. ISSN 1432-1114. doi: 10.1007/ s00348-021-03356-8. Appendix In what foll...

    2022

  41. [51]

    doi: 10.1088/1361-6501/ad2da5

    doi

  42. [52]

    One- shot omnidirectional pressure integration through matrix inversion

    Fernando Zigunov and John J Charonko. One- shot omnidirectional pressure integration through matrix inversion. Measurement Science and Technology, 35(12):125301, sep 2024. doi: 10.1088/ 1361-6501/ad7486

    2024

  43. [56]

    highlighted, the pressure reconstruction is more sensitive to the error located near a Neumann bound- ary and/or far away from a Dirichlet boundary. It is still not clear how the regularization reconciles with data er- rors, but this interplay might be related to the disparate...

  44. [233]

    doi: 10.1088/0957-0233/21/10/105401

    doi

  45. [1114]

    doi: 10.1007/s00348-017-2324-z

    doi

  46. [1120]

    doi: 10.1017/jfm.2020.214. A11

    2020 doi

  47. [2008]

    doi: 10.1080/14685240802376389

    doi

  48. [7930]

    2022.105391

    doi: https://doi.org/10.1016/j.compfluid. 2022.105391

    2022

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.