pith. sign in

arxiv: 2604.25955 · v1 · submitted 2026-04-26 · 🧮 math.NA · cs.NA· physics.flu-dyn

Mode-realigned pointwise interpolation (MRPWI) for efficient POD-Galerkin parametric reduced-order models

Lei Du , Shengqi Zhang This is my paper

Pith reviewed 2026-05-08 05:35 UTC · model grok-4.3

classification 🧮 math.NA cs.NAphysics.flu-dyn
keywords mode realignmentPOD-Galerkinparametric reduced-order modelspointwise interpolationGrassmann manifold interpolationproper orthogonal decompositioncylinder flow
0
0 comments X

The pith

A two-step sign and rotation alignment of POD modes lets pointwise interpolation build parametric reduced-order models that match Grassmann accuracy at far lower cost.

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

The paper introduces mode-realigned pointwise interpolation to construct POD-Galerkin parametric reduced-order models. It first aligns the signs of POD modes computed at different parameter values and then applies a rotation to synchronize them further, after which direct pointwise interpolation supplies the modes at new parameter values. On the cylinder-flow test case the resulting models reach accuracy comparable to Grassmann-manifold interpolation while running substantially faster and still remain close to direct numerical simulation. A reader would care because many engineering studies need repeated high-fidelity solutions across a range of parameters, and any reliable speed-up without loss of fidelity makes such studies more feasible.

Core claim

The central claim is that POD modes obtained at sampled parameter values can be synchronized by a two-step realignment consisting of sign alignment followed by rotation alignment. Once synchronized, the modes admit accurate pointwise interpolation to unseen parameters, producing POD-Galerkin parametric reduced-order models whose accuracy is comparable to models built with Grassmann manifold interpolation yet whose construction cost is markedly lower, as shown by the high-fidelity cylinder-flow results.

What carries the argument

The two-step mode realignment procedure (sign alignment followed by rotation alignment) inside MRPWI, which synchronizes POD modes across parameter values so that ordinary pointwise interpolation becomes reliable.

If this is right

  • PROMs built with MRPWI exhibit accuracy comparable to those built with GMI.
  • MRPWI construction cost is significantly lower than GMI construction cost.
  • The resulting models maintain high fidelity to direct numerical simulation on the cylinder-flow example.
  • The approach replaces manifold operations with simple algebraic steps while preserving the POD-Galerkin structure.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same realignment steps might be applied to other snapshot-based bases such as dynamic mode decomposition or reduced-basis methods.
  • Once the modes are synchronized, the resulting parametric ROMs could be embedded inside outer-loop tasks such as optimization or uncertainty quantification without repeated manifold projections.
  • If the alignment cost remains negligible for large numbers of modes, the method could scale to three-dimensional flow problems where GMI becomes prohibitive.

Load-bearing premise

POD modes at different parameter values differ from one another only by a sign flip and an orthogonal rotation that can be removed without discarding essential information or creating artifacts that degrade interpolation at unseen parameters.

What would settle it

A parameter value where the MRPWI-interpolated modes, after sign and rotation alignment, produce reduced-order solutions whose error relative to direct numerical simulation exceeds the error obtained with Grassmann manifold interpolation.

Figures

Figures reproduced from arXiv: 2604.25955 by Lei Du, Shengqi Zhang.

Figure 1
Figure 1. Figure 1: Flow over a cylinder: computational domain with cylinder diameter doubled for visualization. the inlet (𝑥 = −15𝐷), corresponding to a Dirichlet boundary condition. A natural boundary condition is enforced at the outlet (𝑥 = +35𝐷), given by [−𝑝𝑰 + 𝜈∇𝒖] ⋅ 𝒏 = 0, where 𝒏 is the unit normal vector. For the top (𝑦 = +15𝐷) and bottom (𝑦 = −15𝐷) boundaries, periodic boundary conditions are applied. On the cylinde… view at source ↗
Figure 2
Figure 2. Figure 2: RLEs with respect to number of modes 𝑁𝑟 for (a) velocity and (b) pressure. (2-point: 𝑅𝑒 ∈ {120, 140}; 4-point: 𝑅𝑒 ∈ {100, 120, 140, 160}). ROMs 2-point GMI 4-point GMI 2-point MRPWI 4-point MRPWI 10 20 30 40 "Re 0 0.5 1 1.5 2 7" u (%) (a) 10 20 30 40 "Re 0 2 4 6 8 10 7"p (%) (b) view at source ↗
Figure 3
Figure 3. Figure 3: RLEs with respect to system parameter interval Δ𝑅𝑒 for (a) velocity and (b) pressure. (2-point: 𝑅𝑒 ∈ {125,135}, {120,140}, {115,145}, {110,150}; 4-point: 𝑅𝑒 ∈ {115,125,135,145}, {100,120,140,160}, {85,115,145,175}, {70,110,150,190}). ROMs GMI MRPWI 2 3 4 5 Np 0.1 0.2 0.3 0.4 0.5 7" u (%) (a) 2 3 4 5 Np 0 0.5 1 1.5 2 2.5 7"p (%) (b) view at source ↗
Figure 4
Figure 4. Figure 4: RLEs with respect to number of neighbors 𝑁𝑝 for (a) velocity and (b) pressure. (𝑅𝑒 ∈ {120,140},{100,120,140}, {100,120,140,160}, {80,100,120,140,160}). : Preprint submitted to Elsevier Page 7 of 9 view at source ↗
read the original abstract

As a cornerstone of reduced-order modeling, the POD-Galerkin framework has garnered widespread attention and remains one of the most widely adopted approaches. Constructing POD-Galerkin PROMs involves integrating this framework with advanced interpolation techniques to obtain POD modes at target (unseen) parameters. While Grassmann manifold interpolation (GMI) serves as an accurate baseline, mode-realigned pointwise interpolation (MRPWI) is proposed to develop highly efficient PROMs that maintain comparable accuracy. Notably, the MRPWI employs a two-step mode realignment procedure, consisting of sign alignment and rotation alignment, to effectively synchronize the POD modes. Demonstration and evaluation of the constructed POD-Galerkin PROMs are conducted by examining flow over a cylinder. These models exhibit high fidelity in comparison to direct numerical simulation and standard POD-Galerkin ROMs. PROMs constructed via MRPWI achieve accuracy comparable to those using GMI, while providing significantly higher computational efficiency.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

3 major / 2 minor

Summary. The paper proposes mode-realigned pointwise interpolation (MRPWI) as an efficient alternative to Grassmann manifold interpolation (GMI) for constructing parametric POD-Galerkin reduced-order models (PROMs). It introduces a two-step mode realignment procedure (sign alignment followed by rotation alignment) to synchronize POD modes across parameter values, enabling direct pointwise interpolation. The approach is demonstrated on cylinder flow, where the resulting PROMs are reported to achieve high fidelity relative to DNS and standard POD-Galerkin ROMs, with accuracy comparable to GMI-based models but significantly higher computational efficiency.

Significance. If the central claims hold, MRPWI would offer a lightweight procedural alternative to manifold interpolation for parametric ROM construction, which could be valuable in applications such as parametric studies of fluid flows where repeated mode interpolation is required. The method's emphasis on avoiding manifold operations while preserving accuracy addresses a practical bottleneck in reduced-order modeling.

major comments (3)
  1. Abstract: the claims of 'accuracy comparable to those using GMI' and 'significantly higher computational efficiency' are presented without any quantitative error metrics (e.g., L2 or H1 norms), CPU timings, or tabulated comparisons between MRPWI, GMI, and full-order solutions, leaving the efficiency-accuracy tradeoff unsupported by data.
  2. The two-step realignment procedure (sign alignment + rotation alignment) implicitly assumes that all parametric variation in the POD basis is captured by sign changes and orthogonal transformations within a fixed subspace. No analysis or test is provided for cases of mode crossing, energy redistribution across singular values, or non-rotational distortions, which would violate this assumption and potentially degrade interpolation accuracy at unseen parameters.
  3. Cylinder-flow demonstration: the evaluation reports 'high fidelity' and 'comparable accuracy' but supplies no convergence data with respect to the number of retained modes, no error tables versus parameter values, and no explicit quantification of how the realignment step affects interpolation error relative to unaligned pointwise interpolation or GMI.
minor comments (2)
  1. Notation for the realignment operators (sign flip and rotation matrix) should be defined explicitly with equations rather than described only procedurally.
  2. The manuscript would benefit from a clear statement of the parameter range and sampling strategy used for the cylinder-flow training set, as well as the specific unseen parameter values used for testing.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive and detailed comments. We address each major comment point by point below, proposing specific revisions to strengthen the manuscript.

read point-by-point responses

  1. Referee: Abstract: the claims of 'accuracy comparable to those using GMI' and 'significantly higher computational efficiency' are presented without any quantitative error metrics (e.g., L2 or H1 norms), CPU timings, or tabulated comparisons between MRPWI, GMI, and full-order solutions, leaving the efficiency-accuracy tradeoff unsupported by data.

    Authors: We agree that the abstract would be strengthened by including specific quantitative support. In the revised manuscript we will add concise statements of key metrics (e.g., relative L2 errors and wall-clock timings) drawn from the cylinder-flow results already presented in Section 4. revision: yes

  2. Referee: The two-step realignment procedure (sign alignment + rotation alignment) implicitly assumes that all parametric variation in the POD basis is captured by sign changes and orthogonal transformations within a fixed subspace. No analysis or test is provided for cases of mode crossing, energy redistribution across singular values, or non-rotational distortions, which would violate this assumption and potentially degrade interpolation accuracy at unseen parameters.

    Authors: The referee correctly identifies the core modeling assumption. MRPWI is intended for regimes in which the POD subspace remains stable and parametric changes manifest primarily as sign flips and orthogonal transformations, as occurs in the cylinder example. We will insert a dedicated paragraph (new subsection 3.4) that explicitly states this assumption, discusses the conditions under which it may fail (mode crossing, singular-value reordering), and illustrates the expected behavior with a short synthetic test case. revision: yes

  3. Referee: Cylinder-flow demonstration: the evaluation reports 'high fidelity' and 'comparable accuracy' but supplies no convergence data with respect to the number of retained modes, no error tables versus parameter values, and no explicit quantification of how the realignment step affects interpolation error relative to unaligned pointwise interpolation or GMI.

    Authors: We acknowledge that the current presentation of results is insufficiently quantitative. In the revised Section 4 we will add: (i) a convergence study of PROM error versus number of retained modes, (ii) tabulated L2/H1 errors for each tested parameter value, and (iii) an ablation table that isolates the contribution of the sign-alignment and rotation-alignment steps relative to plain pointwise interpolation and to GMI. revision: yes

Circularity Check

0 steps flagged

No circularity: MRPWI is a direct procedural alternative to GMI

full rationale

The paper defines MRPWI via an explicit two-step realignment (sign alignment followed by rotation alignment) applied to POD modes computed at sampled parameters, then performs pointwise interpolation on the aligned modes. Accuracy and efficiency claims rest on numerical comparison against DNS and GMI on the cylinder-flow example, without any fitted parameter being relabeled as a prediction, without self-definitional loops, and without load-bearing self-citations that close the argument. The central derivation therefore remains independent of its own outputs.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claim depends on the domain assumption that POD modes at different parameters are alignable by sign and rotation without loss of fidelity for subsequent pointwise interpolation.

axioms (1)
  • domain assumption POD modes computed at different parameter values share a common structure that can be synchronized via sign alignment followed by rotation alignment.
    Invoked directly in the description of the two-step realignment procedure that enables pointwise interpolation.

pith-pipeline@v0.9.0 · 5469 in / 1195 out tokens · 62767 ms · 2026-05-08T05:35:13.921123+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

40 extracted references · 2 canonical work pages · 1 internal anchor

  1. [1]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION word.in bbl.in ":" * " " * FUNCTION f...

  2. [2]

    , year 2010

    author Amsallem, D. , year 2010 . title Interpolation on manifolds of CFD-based fluid and finite element-based structural reduced-order models for on-line aeroelastic predictions . publisher Stanford University

    2010

  3. [3]

    , author Farhat, C

    author Amsallem, D. , author Farhat, C. , year 2008 . title Interpolation method for adapting reduced-order models and application to aeroelasticity . journal AIAA journal volume 46 , pages 1803--1813

    2008

  4. [4]

    , author Zimmermann, R

    author Bendokat, T. , author Zimmermann, R. , author Absil, P.A. , year 2024 . title A grassmann manifold handbook: Basic geometry and computational aspects . journal Advances in Computational Mathematics volume 50 , pages 6

    2024

  5. [5]

    , author Grivet-Talocia, S

    author Benner, P. , author Grivet-Talocia, S. , author Quarteroni, A. , author Rozza, G. , author Schilders, W. , author Silveira, L.M. , year 2021 . title System-and data-driven methods and algorithms . publisher De Gruyter

    2021

  6. [6]

    , author Gugercin, S

    author Benner, P. , author Gugercin, S. , author Willcox, K. , year 2015 . title A survey of projection-based model reduction methods for parametric dynamical systems . journal SIAM review volume 57 , pages 483--531

    2015

  7. [7]

    , author Bruneau, C.H

    author Bergmann, M. , author Bruneau, C.H. , author Iollo, A. , year 2009 . title Enablers for robust pod models . journal Journal of Computational Physics volume 228 , pages 516--538

    2009

  8. [8]

    , author Kutz, J.N

    author Brunton, S.L. , author Kutz, J.N. , year 2022 . title Data-driven science and engineering: Machine learning, dynamical systems, and control . publisher Cambridge University Press

    2022

  9. [9]

    , author Noack, B.R

    author Brunton, S.L. , author Noack, B.R. , author Koumoutsakos, P. , year 2020 . title Machine learning for fluid mechanics . journal Annual review of fluid mechanics volume 52 , pages 477--508

    2020

  10. [10]

    , author Loiseau, J.C

    author Callaham, J.L. , author Loiseau, J.C. , author Brunton, S.L. , year 2023 . title Multiscale model reduction for incompressible flows . journal Journal of Fluid Mechanics volume 973 , pages A3

    2023

  11. [11]

    , author Wang, Q

    author Chen, W. , author Wang, Q. , author Hesthaven, J.S. , author Zhang, C. , year 2021 . title Physics-informed machine learning for reduced-order modeling of nonlinear problems . journal Journal of computational physics volume 446 , pages 110666

    2021

  12. [12]

    , author Fischer, P.F

    author Deville, M.O. , author Fischer, P.F. , author Mund, E.H. , year 2002 . title High-order methods for incompressible fluid flow . volume volume 9 . publisher Cambridge university press

    2002

  13. [13]

    , author Zhang, S

    author Du, L. , author Zhang, S. , author Yang, Y. , year 2025 . title Interpolation-based parametric reduced-order models via galerkin projection and dynamic mode decomposition . journal Physics of Fluids volume 37

    2025

  14. [14]

    , author Zhang, S

    author Du, L. , author Zhang, S. , author Zhang, R. , author Zhang, S. , year 2026 . title Interpolation-based parametric reduced-order models with dynamic mode decomposition . journal Journal of Computational Physics volume 544 , pages 114436

    2026

  15. [15]

    Deep Learning for Subspace Regression

    author Fanaskov, V. , author Trifonov, V. , author Rudikov, A. , author Muravleva, E. , author Oseledets, I. , year 2025 . title Deep learning for subspace regression . journal arXiv preprint arXiv:2509.23249

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  16. [16]

    , year 2002

    author Feeny, B. , year 2002 . title On proper orthogonal co-ordinates as indicators of modal activity . journal Journal of Sound and Vibration volume 255 , pages 805--817

    2002

  17. [17]

    , author Quaini, A

    author Hess, M.W. , author Quaini, A. , author Rozza, G. , year 2020 . title A comparison of reduced-order modeling approaches using artificial neural networks for pdes with bifurcating solutions . journal arXiv preprint arXiv:2010.07370

    work page arXiv 2020

  18. [18]

    , author Kutz, J.N

    author Hinze, M. , author Kutz, J.N. , author Mula, O. , author Urban, K. , year 2023 . title Model order reduction and applications: Cetraro, Italy 2021 . publisher Springer Nature

    2023

  19. [19]

    , year 2012

    author Holmes, P. , year 2012 . title Turbulence, coherent structures, dynamical systems and symmetry . publisher Cambridge university press

    2012

  20. [20]

    , author Feng, L

    author Kapadia, H. , author Feng, L. , author Benner, P. , year 2024 . title Active-learning-driven surrogate modeling for efficient simulation of parametric nonlinear systems . journal Computer Methods in Applied Mechanics and Engineering volume 419 , pages 116657

    2024

  21. [21]

    , author Brunton, S.L

    author Kutz, J.N. , author Brunton, S.L. , author Brunton, B.W. , author Proctor, J.L. , year 2016 . title Dynamic mode decomposition: data-driven modeling of complex systems . publisher SIAM

    2016

  22. [22]

    , author Lesoinne, M

    author Lieu, T. , author Lesoinne, M. , year 2004 . title Parameter adaptation of reduced order models for three-dimensional flutter analysis , in: booktitle 42nd AIAA Aerospace Sciences Meeting and Exhibit , p. pages 888

    2004

  23. [23]

    , author Cammi, A

    author Lorenzi, S. , author Cammi, A. , author Luzzi, L. , author Rozza, G. , year 2016 . title Pod-galerkin method for finite volume approximation of navier--stokes and rans equations . journal Computer Methods in Applied Mechanics and Engineering volume 311 , pages 151--179

    2016

  24. [24]

    , year 1970

    author Lumley, J.L. , year 1970 . title Stochastic tools in turbulence

    1970

  25. [25]

    , author Peplinski, A

    author Massaro, D. , author Peplinski, A. , author Schlatter, P. , year 2023 . title The flow around a stepped cylinder with turbulent wake and stable shear layer . journal Journal of Fluid Mechanics volume 977 , pages A3

    2023

  26. [26]

    , author Schlatter, P

    author Massaro, D. , author Schlatter, P. , year 2024 . title Global stability of the flow past a stepped cylinder . journal Journal of Fluid Mechanics volume 988 , pages A1

    2024

  27. [27]

    , author Balabane, M

    author Mendez, M.A. , author Balabane, M. , author Buchlin, J.M. , year 2019 . title Multi-scale proper orthogonal decomposition of complex fluid flows . journal Journal of Fluid Mechanics volume 870 , pages 988--1036

    2019

  28. [28]

    , author Ianiro, A

    author Mendez, M.A. , author Ianiro, A. , author Noack, B.R. , author Brunton, S.L. , year 2023 . title Data-driven fluid mechanics: combining first principles and machine learning . publisher Cambridge University Press

    2023

  29. [29]

    , author Papas, P

    author Noack, B.R. , author Papas, P. , author Monkewitz, P.A. , year 2005 . title The need for a pressure-term representation in empirical galerkin models of incompressible shear flows . journal Journal of Fluid Mechanics volume 523 , pages 339--365

    2005

  30. [30]

    , author Xiao, D

    author Pan, X. , author Xiao, D. , author Wang, L. , author Yang, X. , year 2025 . title A physics-data combined neural network-based finite volume parametric reduced order model . journal Physics of Fluids volume 37

    2025

  31. [31]

    , year 1984

    author Patera, A.T. , year 1984 . title A spectral element method for fluid dynamics: laminar flow in a channel expansion . journal Journal of computational Physics volume 54 , pages 468--488

    1984

  32. [32]

    , author Zhang, Y.J

    author Prakash, A. , author Zhang, Y.J. , year 2025 . title Nonintrusive projection-based reduced order modeling using stable learned differential operators . journal Computer Methods in Applied Mechanics and Engineering volume 442 , pages 117946

    2025

  33. [33]

    , year 2005

    author Rowley, C.W. , year 2005 . title Model reduction for fluids, using balanced proper orthogonal decomposition . journal International Journal of Bifurcation and Chaos volume 15 , pages 997--1013

    2005

  34. [34]

    , year 1987

    author Sirovich, L. , year 1987 . title Turbulence and the dynamics of coherent structures. i. coherent structures . journal Quarterly of applied mathematics volume 45 , pages 561--571

    1987

  35. [35]

    , author Kirby, M

    author Sirovich, L. , author Kirby, M. , year 1987 . title Low-dimensional procedure for the characterization of human faces . journal Journal of the Optical Society of America A volume 4 , pages 519--524

    1987

  36. [36]

    , author Brunton, S.L

    author Taira, K. , author Brunton, S.L. , author Dawson, S.T. , author Rowley, C.W. , author Colonius, T. , author McKeon, B.J. , author Schmidt, O.T. , author Gordeyev, S. , author Theofilis, V. , author Ukeiley, L.S. , year 2017 . title Modal analysis of fluid flows: An overview . journal AIAA journal volume 55 , pages 4013--4041

    2017

  37. [37]

    , author Hemati, M.S

    author Taira, K. , author Hemati, M.S. , author Brunton, S.L. , author Sun, Y. , author Duraisamy, K. , author Bagheri, S. , author Dawson, S.T. , author Yeh, C.A. , year 2020 . title Modal analysis of fluid flows: Applications and outlook . journal AIAA journal volume 58 , pages 998--1022

    2020

  38. [38]

    , author Schmidt, O.T

    author Towne, A. , author Schmidt, O.T. , author Colonius, T. , year 2018 . title Spectral proper orthogonal decomposition and its relationship to dynamic mode decomposition and resolvent analysis . journal Journal of Fluid Mechanics volume 847 , pages 821--867

    2018

  39. [39]

    , author Peraire, J

    author Willcox, K. , author Peraire, J. , year 2002 . title Balanced model reduction via the proper orthogonal decomposition . journal AIAA journal volume 40 , pages 2323--2330

    2002

  40. [40]

    , author Mak, S

    author Zhang, R. , author Mak, S. , author Dunson, D. , year 2022 . title Gaussian process subspace prediction for model reduction . journal SIAM Journal on Scientific Computing volume 44 , pages A1428--A1449

    2022