pith. sign in

arxiv: 2606.10335 · v1 · pith:WTXZ23BDnew · submitted 2026-06-09 · ⚛️ physics.comp-ph · cs.NA· math.NA· physics.flu-dyn

A Physics-Informed B-Spline Framework for Continuous Approximation of Flow Data

Junoh Jung , David Lenz , Emil Constantinescu , Tom Peterka This is my paper

Pith reviewed 2026-06-27 11:21 UTC · model grok-4.3

classification ⚛️ physics.comp-ph cs.NAmath.NAphysics.flu-dyn
keywords B-splinesphysics-informedflow approximationPDE residualscontinuous reconstructionNavier-Stokesfunctional approximationconvection-diffusion
0
0 comments X

The pith

Embedding PDE residuals into B-spline optimization produces continuous flow approximations that better respect governing physics.

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

The paper introduces a method to create continuous, differentiable approximations of discrete flow data using B-splines while enforcing physical constraints from the PDEs. It formulates an optimization problem for the spline control points that trades off matching the input data against minimizing the residuals of the governing equations and boundary conditions. This approach is tested on standard flow problems and shown to yield reconstructions with smaller PDE residuals and improved consistency with balance laws compared to data-only spline fits. For cases where the input data itself violates the physics, the method also achieves lower approximation errors than alternatives.

Core claim

By determining B-spline control points through minimization of a loss that includes both data fidelity and PDE residual terms, the resulting multivariate functional approximations preserve the exact differentiability and local support of splines while reducing unphysical residuals in the reconstructed fields for convection-diffusion, Burgers, and Navier-Stokes problems.

What carries the argument

Tensor-product B-splines whose control points are found by solving an optimization problem that balances data fidelity with PDE residuals, initial conditions, and boundary conditions, evaluated using analytical derivatives of the basis functions.

If this is right

  • Reconstructed fields exhibit reduced PDE residuals and better global balance-law consistency.
  • Lower approximation errors occur when the input data is physically inconsistent.
  • The method offers computational advantages over physics-informed neural networks for the tested equations.
  • Continuous fields remain compact and locally supported, enabling efficient downstream analysis and visualization.

Where Pith is reading between the lines

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

  • Such reconstructions could improve reliability of derived quantities like gradients or integrals computed from discrete simulation outputs.
  • The framework might apply to other types of physical data beyond fluid flows if the governing equations are known.
  • Integration with existing simulation codes could allow on-the-fly physics-informed post-processing without full field storage.

Load-bearing premise

The optimization balancing data fidelity with PDE residuals can be solved to produce B-spline coefficients that meaningfully reduce physical residuals in the tested problems.

What would settle it

Numerical experiments on the listed flow equations where the PI-MFA fields do not show lower PDE residuals or improved balance consistency relative to standard MFA.

Figures

Figures reproduced from arXiv: 2606.10335 by David Lenz, Emil Constantinescu, Junoh Jung, Tom Peterka.

Figure 1
Figure 1. Figure 1: Schematic of the proposed PI-MFA. Boxes and components shown in the gray-shaded background represent existing work, whereas those in the white background highlight the new contributions of this study. PDEs, boundary conditions (BCs), and initial conditions (ICs) directly into the approximation. Physics￾informed neural and operator-learning methods have shown that enforcing governing laws can improve robust… view at source ↗
Figure 2
Figure 2. Figure 2: One-dimensional convection-diffusion problem on 𝑥 ∈ [0, 1] with homogeneous Dirichlet bound￾ary conditions. Left: initial condition at 𝑡 = 0 (blue dashed) and terminal-time profiles at 𝑇 = 0.4 obtained from a non-conservative finite difference simulation (orange) and the analytical reference solution (black). Right: spatiotemporal fields of the analytical solution 𝜌ana, the numerical solution 𝜌num, and the… view at source ↗
Figure 3
Figure 3. Figure 3: Results of standard MFA data fitting, shown as a function of the number of control points. The number of control points in space and time (𝑁𝑥, 𝑁𝑡) is indicated at the top of each panel. 14 [PITH_FULL_IMAGE:figures/full_fig_p014_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: One-dimensional sweeps to choose the loss weights 𝜆 for the 1D convection–diffusion equation: (a) vary 𝜆𝑑𝑎𝑡 𝑎 while holding the other 𝜆 values fixed; (b) vary 𝜆 𝑝𝑑𝑒; (c) vary 𝜆𝐼𝐶; and (d) vary 𝜆𝐵𝐶, each with all other weights fixed. ℒtotal ℒpde ℒdata -3 10 -4 10 -5 10 -6 10 -2 10 -1 10 -5 10 -6 10 -4 10 -3 10 -2 10 -1 10 0 10 min min Pareto front Pareto-optimal ℒpde ℒphysics min ℒtotal 1.6 1.4 1.2 1.0 0.8 … view at source ↗
Figure 5
Figure 5. Figure 5: Results of the 4D parameter sweep over (𝜆data, 𝜆pde, 𝜆IC, 𝜆BC), shown in the (Ldata, Lpde) plane. The black curve denotes the Pareto front. Specific configurations are highlighted by distinct markers: the green diamond (min Lpde), the yellow star (min Ltotal), and the red triangle (min Lphysics). The magenta circle identifies the balanced Pareto-optimal configuration that minimizes the distance to the theo… view at source ↗
Figure 6
Figure 6. Figure 6: Convergence histories of the optimization losses for three representative weight configurations: min Lphysics, min Lpde, and the Pareto-optimal case. These configurations correspond to the specific 𝜆 selections identified in the parameter sweep ( [PITH_FULL_IMAGE:figures/full_fig_p018_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Continuous-field approximations (left column), strong-form PDE residuals (middle column), and actual relative errors with respect to the analytical solution (right column). Rows correspond to standard MFA (a–c), Reg-MFA (d–f), PINN (g–i), and PI-MFA under three representative weight selections: min Lpde (j–l), Pareto-optimal (m–o), and min Lphysics (p–r). 19 [PITH_FULL_IMAGE:figures/full_fig_p019_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Temporal evolution of the integrated mass-balance (flux) residual for the 1D convection-diffusion problem. The exact analytical solution conserves mass (zero residual). The non-conservative numerical data computed by the finite difference method (FDM) exhibits a severe negative drift, which the purely data￾driven MFA and Reg-MFA baselines inherently absorb. The PI-MFA configurations correct this physical d… view at source ↗
Figure 9
Figure 9. Figure 9: Snapshots of the 2D coupled viscous Burgers solution at 𝑡 = 0, 0.5, 1.0, 1.5, and 2.0. The left panels show the 𝑢 field, and the right panels show the 𝑣 field at the corresponding times. The FEM solution serves as the high-accuracy reference, while the input data is shown separately as physically inconsistent data. The absolute error with respect to the reference solution is reported in the rightmost colum… view at source ↗
Figure 10
Figure 10. Figure 10: Volumetric rendering of the standard MFA reconstructions for the 2D coupled Burgers velocity fields, 𝑢(𝑥, 𝑦, 𝑡) and 𝑣(𝑥, 𝑦, 𝑡), evaluated across four distinct control-point grid resolutions (𝑃𝑥, 𝑃𝑦, 𝑃𝑡). For each grid density, the pointwise absolute error relative to the simulation data and the corresponding strong￾form PDE residuals are displayed. 25 [PITH_FULL_IMAGE:figures/full_fig_p025_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: Convergence histories of the unweighted loss components during L-BFGS optimization for the four representative PI-MFA configurations applied to the 2D coupled Burgers problem. Each panel tracks the evolution of the total loss alongside the individual data-misfit, PDE-residual, initial-condition, and boundary-condition contributions for both the 𝑢 and 𝑣 velocity fields. 30 [PITH_FULL_IMAGE:figures/full_fi… view at source ↗
Figure 12
Figure 12. Figure 12: Volumetric comparison of the reconstructed 2D Burgers velocity fields (𝑢 and 𝑣) across standard MFA, regularized MFA (Reg-MFA), a PINN baseline, and four PI-MFA configurations. For each velocity component, the panels display the continuous approximation, the corresponding strong-form PDE residual, and the pointwise actual error relative to the high-fidelity FEM reference solution. The aggregated MSE for t… view at source ↗
Figure 13
Figure 13. Figure 13: Temporal evolution of the global integral-balance residuals, (a) 𝑟𝑢,𝑀 (𝑡) and (b) 𝑟𝑣,𝑀 (𝑡), for the two-dimensional coupled Burgers equations (see Eq. (4.18)). 33 [PITH_FULL_IMAGE:figures/full_fig_p033_13.png] view at source ↗
Figure 14
Figure 14. Figure 14: Snapshots of the horizontal velocity (𝑢), vertical velocity (𝑣), and kinematic pressure (𝑝) for the regularized 2D lid-driven cavity flow at 𝑡 = 0.01, 2.5, 5, 7.5, and 10. For each flow variable, the top row displays the exact high-fidelity FEM reference solution, the middle row shows the physically inconsistent low-fidelity observation data (degraded by spatial underresolution and numerical diffusion), a… view at source ↗
Figure 15
Figure 15. Figure 15: Standard, purely data-driven MFA reconstruction of the horizontal velocity (𝑢), vertical velocity (𝑣), and kinematic pressure (𝑝) space-time fields for the 2D lid-driven cavity problem. The volumetric plots depict the continuous approximations evaluated on the selected 20 × 20 × 50 control-point grid (top row) alongside their corresponding pointwise data-fitting errors relative to the corrupted observatio… view at source ↗
Figure 16
Figure 16. Figure 16: Convergence histories of the PI-MFA composite loss components for the 2D lid-driven cavity problem. All configurations are warm-started from the data-only MFA solution. Cases 1 and 2 rapidly stagnate as they overfit the coarse input data, leaving the physical residuals (dashed lines) elevated. Case 3 successfully drives the PDE residuals toward zero over a prolonged optimization by relaxing the data const… view at source ↗
Figure 17
Figure 17. Figure 17: Continuous space-time volumetric rendering (𝑥, 𝑦, 𝑡) of the 𝑢-velocity, 𝑣-velocity, and pressure (𝑝) fields for the lid-driven cavity problem. The high-fidelity FEM serves as the benchmark reference. The data-driven MFA accurately interpolates the noisy low-fidelity data but inherits its corrupted, smeared pressure field. The Reg-MFA and PINN baselines introduce excessive mathematical blurring and high￾fr… view at source ↗
Figure 18
Figure 18. Figure 18: Volumetric rendering of the strong-form PDE residual errors (𝑥-momentum, 𝑦-momentum, and divergence) over the space-time domain based on the metric in Eq. (4.1). 43 [PITH_FULL_IMAGE:figures/full_fig_p043_18.png] view at source ↗
Figure 19
Figure 19. Figure 19: One-dimensional convection-diffusion benchmarks Case 2 and Case 3 with homogeneous Dirichlet boundary conditions. Left: initial condition at 𝑡 = 0 (blue dashed) and solutions at 𝑇 = 0.4 from a non-conservative finite difference simulation (colored) and the analytical reference (black). Right: spatiotemporal fields of the analytical solution, the numerical solution, and the pointwise error 𝑒(𝑥, 𝑡) = 𝜌num(𝑥… view at source ↗
Figure 20
Figure 20. Figure 20: Results of standard MFA data fitting for Case 2, shown as a function of the number of control points. The parameter values used in the supplementary experiments are Case 2: 𝐴sin = 0.35, 𝐴b = 0.55, 𝑥0 = 0.25, 𝜎 = 0.06, Case 3: 𝐴1 = 0.7, 𝐴2 = −0.35, 𝑥1 = 0.28, 𝑥2 = 0.72, 𝑠1 = 0.05, 𝑠2 = 0.09 [PITH_FULL_IMAGE:figures/full_fig_p051_20.png] view at source ↗
Figure 21
Figure 21. Figure 21: Results of standard MFA data fitting for Case 3, shown as a function of the number of control points. 52 [PITH_FULL_IMAGE:figures/full_fig_p052_21.png] view at source ↗
read the original abstract

Continuous approximations of flow data are useful for downstream analysis, differentiation, and visualization, but purely data-driven reconstructions do not, in general, preserve the governing physics. This limitation becomes particularly important when input data are physically inconsistent, whether due to low-fidelity discretizations or unmodeled discrepancies. In such cases, reconstructed fields may exhibit inaccurate PDE residuals, violated balance laws, or unreliable derived quantities. To address this, we propose a physics-informed B-spline framework that embeds physical constraints directly into the reconstruction process. The method constructs compact, continuously differentiable representations of discrete fields using tensor-product B-splines and determines spline control points by solving an optimization problem balancing data fidelity with residuals of the governing PDEs, alongside initial and boundary conditions. Leveraging exact analytical derivatives of the B-spline basis enables efficient and accurate evaluation of physical residuals without storing full-resolution fields. We refer to this approach as physics-informed multivariate functional approximation (PI-MFA). Numerical studies on the 1D convection-diffusion, 2D coupled Burgers, and 2D incompressible Navier-Stokes equations show PI-MFA reduces PDE residuals and improves global balance-law consistency. Compared with standard and regularized MFA, PI-MFA produces more physically faithful reconstructions and, for physically inconsistent data, lower approximation errors, while offering computational advantages over tested physics-informed neural networks. Overall, PI-MFA preserves the compactness, local support, and exact differentiability of classical spline spaces while producing reliable continuous flow fields for scientific analysis and visualization.

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

1 major / 1 minor

Summary. The manuscript introduces physics-informed multivariate functional approximation (PI-MFA), a framework that represents discrete flow data via tensor-product B-splines and determines the control points through an optimization that balances data-fidelity terms against PDE residuals together with initial and boundary conditions. Exact analytic derivatives of the B-spline basis are used to evaluate the physics residuals efficiently. Numerical demonstrations are reported on the 1D convection-diffusion equation, the 2D coupled Burgers equations, and the 2D incompressible Navier-Stokes equations, with the claim that PI-MFA yields lower PDE residuals, improved global balance-law consistency, and smaller approximation errors on physically inconsistent data than standard or regularized MFA while remaining computationally cheaper than the tested physics-informed neural networks.

Significance. If the reported improvements are quantitatively confirmed, the method supplies a compact, locally supported, and exactly differentiable alternative to neural-network-based physics-informed reconstructions. The use of analytic B-spline derivatives for residual evaluation without storing full-resolution fields is a clear technical advantage for downstream analysis and visualization tasks.

major comments (1)
  1. [Abstract / Numerical studies] Abstract and Numerical studies section: the central claim that 'numerical studies ... show PI-MFA reduces PDE residuals and improves global balance-law consistency' and produces 'lower approximation errors' is not accompanied by any quantitative metrics, error norms, residual values, error bars, optimization tolerances, or data-exclusion criteria. Without these numbers the support for the load-bearing assertion that the physics-informed optimization yields meaningfully better reconstructions cannot be assessed.
minor comments (1)
  1. The balancing weights between the data and physics terms are free parameters; a brief sensitivity study or default selection strategy would clarify reproducibility.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for their thorough review and constructive feedback on our manuscript. We address the major comment below and will revise the manuscript to incorporate the suggested quantitative details.

read point-by-point responses

  1. Referee: [Abstract / Numerical studies] Abstract and Numerical studies section: the central claim that 'numerical studies ... show PI-MFA reduces PDE residuals and improves global balance-law consistency' and produces 'lower approximation errors' is not accompanied by any quantitative metrics, error norms, residual values, error bars, optimization tolerances, or data-exclusion criteria. Without these numbers the support for the load-bearing assertion that the physics-informed optimization yields meaningfully better reconstructions cannot be assessed.

    Authors: We agree that explicit quantitative metrics are needed to substantiate the claims regarding reduced PDE residuals, improved balance-law consistency, and lower approximation errors. In the revised version, we will augment the Numerical studies section with tables reporting L2 and L-infinity norms of PDE residuals, global mass/momentum balance errors, and data approximation errors for PI-MFA versus standard MFA, regularized MFA, and the tested PINNs on each benchmark problem. We will also document the optimization tolerances employed (e.g., convergence criteria for the control-point solver) and any data-exclusion criteria used in the experiments. These additions will allow direct quantitative assessment of the improvements. revision: yes

Circularity Check

0 steps flagged

No significant circularity; method is an explicit optimization construction

full rationale

The paper defines PI-MFA as the solution of an optimization problem whose objective explicitly combines a data-fidelity term with PDE-residual, initial-condition, and boundary-condition terms; the B-spline representation and its analytic derivatives are standard and independent of the target flow fields. Numerical studies are empirical demonstrations on standard test problems rather than predictions that reduce to the fitted inputs by construction. No self-citation is invoked as a load-bearing uniqueness theorem, no ansatz is smuggled via prior work, and no renaming of known results occurs. The derivation chain is therefore self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

Based solely on the abstract; the central claim rests on the existence of a solvable optimization that trades off data fidelity against PDE residuals and on the representational power of tensor-product B-splines for the flow fields. No explicit free parameters, axioms, or invented entities are detailed.

free parameters (1)
  • balancing weights between data and physics terms
    The optimization balances data fidelity with PDE residuals, implying weighting hyperparameters whose specific values are not provided.
axioms (1)
  • domain assumption Tensor-product B-splines possess sufficient approximation power to represent the solution fields of the tested PDEs while allowing exact derivative evaluation
    The method relies on this property to enable efficient residual computation without full-resolution storage.

pith-pipeline@v0.9.1-grok · 5815 in / 1403 out tokens · 35518 ms · 2026-06-27T11:21:54.316174+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

68 extracted references · 58 canonical work pages · 1 internal anchor

  1. [1]

    Piegl, W

    L. Piegl, W. Tiller, The NURBS Book, 2nd Edition, Springer, Berlin, Heidelberg, 1997.doi: 10.1007/978-3-642-59223-2

    work page doi:10.1007/978-3-642-59223-2 1997

  2. [2]

    de Boor, A Practical Guide to Splines, revised Edition, Vol

    C. de Boor, A Practical Guide to Splines, revised Edition, Vol. 27 of Applied Mathematical Sciences, Springer, New York, 2001.doi:10.1007/978-1-4612-6333-3

    work page doi:10.1007/978-1-4612-6333-3 2001

  3. [3]

    L. L. Schumaker, Spline Functions: Basic Theory, 3rd Edition, Cambridge Mathematical Library, Cambridge University Press, Cambridge, 2007.doi:10.1017/CBO9780511618994

    work page doi:10.1017/cbo9780511618994 2007

  4. [4]

    P. H. C. Eilers, B. D. Marx, Flexible smoothing with B-splines and penalties, Statistical Science 11 (2) (1996) 89–121.doi:10.1214/ss/1038425655

    work page doi:10.1214/ss/1038425655 1996

  5. [5]

    M. J. Johnson, Z. Shen, Y. Xu, Scattered data reconstruction by regularization in B-spline and associated wavelet spaces, Journal of Approximation Theory 159 (2) (2009) 197–223.doi:10.1016/j.jat. 2009.02.005

    work page doi:10.1016/j.jat 2009

  6. [6]

    Merchel, B

    S. Merchel, B. J¨ uttler, D. Mokriˇs, M. Pan, Fast formation of matrices for least-squares fitting by tensor- product spline surfaces, Computer-Aided Design 150 (2022) 103307.doi:10.1016/j.cad.2022. 103307

    work page doi:10.1016/j.cad.2022 2022

  7. [7]

    T. J. R. Hughes, J. A. Cottrell, Y. Bazilevs, Isogeometric analysis: CAD, finite elements, NURBS, exact geometry and mesh refinement, Computer Methods in Applied Mechanics and Engineering 194 (39–41) (2005) 4135–4195.doi:10.1016/j.cma.2004.10.008

    work page doi:10.1016/j.cma.2004.10.008 2005

  8. [8]

    Bazilevs, L

    Y. Bazilevs, L. Beir ˜ao da Veiga, J. A. Cottrell, T. J. R. Hughes, G. Sangalli, Isogeometric analysis: approximation, stability and error estimates forℎ-refined meshes, Mathematical Models and Methods in Applied Sciences 16 (7) (2006) 1031–1090.doi:10.1142/S0218202506001455

    work page doi:10.1142/s0218202506001455 2006

  9. [9]

    J. A. Cottrell, T. J. R. Hughes, A. Reali, Studies of refinement and continuity in isogeometric structural analysis, Computer Methods in Applied Mechanics and Engineering 196 (41–44) (2007) 4160–4183. doi:10.1016/j.cma.2007.04.007

    work page doi:10.1016/j.cma.2007.04.007 2007

  10. [10]

    Auricchio, L

    F. Auricchio, L. Beir ˜ao da Veiga, T. J. R. Hughes, A. Reali, G. Sangalli, Isogeometric collocation methods, Mathematical Models and Methods in Applied Sciences 20 (11) (2010) 2075–2107.doi: 10.1142/S0218202510004878

    work page doi:10.1142/s0218202510004878 2010

  11. [11]

    Auricchio, L

    F. Auricchio, L. Beir ˜ao da Veiga, T. J. R. Hughes, A. Reali, G. Sangalli, Isogeometric collocation for elastostatics and explicit dynamics, Computer Methods in Applied Mechanics and Engineering 249–252 (2012) 2–14.doi:10.1016/j.cma.2012.03.026

    work page doi:10.1016/j.cma.2012.03.026 2012

  12. [13]

    Montardini, G

    M. Montardini, G. Sangalli, L. Tamellini, Optimal-order isogeometric collocation at Galerkin super- convergent points, Computer Methods in Applied Mechanics and Engineering 316 (2017) 741–757. doi:10.1016/j.cma.2016.09.043

    work page doi:10.1016/j.cma.2016.09.043 2017

  13. [14]

    Peterka, Y

    T. Peterka, Y. S. G. Nashed, I. R. Grindeanu, V. S. Mahadevan, R. Yeh, X. Tricoche, Foundations of multivariate functional approximation for scientific data, in: 2018 IEEE 8th Symposium on Large Data Analysis and Visualization (LDA V), 2018, pp. 61–71.doi:10.1109/LDAV.2018.8739195. 55

    work page doi:10.1109/ldav.2018.8739195 2018

  14. [15]

    D. Lenz, R. Yeh, V. Mahadevan, I. Grindeanu, T. Peterka, Customizable adaptive regularization techniques for B-spline modeling, Journal of Computational Science 71 (2023) 102037.doi:10. 1016/j.jocs.2023.102037

    arXiv 2023

  15. [16]

    J. Sun, D. Lenz, H. Yu, T. Peterka, Scalable volume visualization for big scientific data modeled by functional approximation, IEEE Transactions on Visualization and Computer Graphics 30 (12) (2024) 8637–8651.doi:10.1109/TVCG.2024.3353594

    work page doi:10.1109/tvcg.2024.3353594 2024

  16. [17]

    J. Sun, D. Lenz, H. Yu, T. Peterka, MFA-DVR: Direct volume rendering of MFA models, Journal of Visualization 27 (2024) 109–126.doi:10.1007/s12650-023-00946-y

    work page doi:10.1007/s12650-023-00946-y 2024

  17. [18]

    Peterka, D

    T. Peterka, D. Lenz, I. Grindeanu, V. S. Mahadevan, Towards adaptive refinement for multivariate functional approximation of scientific data, in: 2023 IEEE 13th Symposium on Large Data Analysis and Visualization (LDA V), 2023, pp. 32–41.doi:10.1109/LDAV60332.2023.00011

    work page doi:10.1109/ldav60332.2023.00011 2023

  18. [19]

    Grindeanu, T

    I. Grindeanu, T. Peterka, V. S. Mahadevan, Y. S. G. Nashed, Scalable, high-order continuity across block boundaries of functional approximations computed in parallel, in: 2019 IEEE International Conference on Cluster Computing (CLUSTER), 2019, pp. 1–9.doi:10.1109/CLUSTER.2019.8891018

    work page doi:10.1109/cluster.2019.8891018 2019

  19. [20]

    V. S. Mahadevan, D. Lenz, I. Grindeanu, T. Peterka, Accelerating multivariate functional approximation computation with domain decomposition techniques, Journal of Computational Science 78 (2024) 102268.doi:10.1016/j.jocs.2024.102268

    work page doi:10.1016/j.jocs.2024.102268 2024

  20. [21]

    G. Ma, D. Lenz, T. Peterka, H. Guo, B. Wang, Critical point extraction from multivariate functional approximation, in: 2024 IEEE Topological Data Analysis and Visualization (TopoInVis), 2024, pp. 12–22.doi:10.1109/TopoInVis64104.2024.00006

    work page doi:10.1109/topoinvis64104.2024.00006 2024

  21. [22]

    J. Sun, D. Lenz, H. Yu, T. Peterka, Adaptive multi-resolution encoding for interactive large-scale volume visualization through functional approximation, in: 2024 IEEE 14th Symposium on Large Data Analysis and Visualization (LDA V), IEEE, 2024, pp. 33–42.doi:10.1109/LDAV64567.2024.00006

    work page doi:10.1109/ldav64567.2024.00006 2024

  22. [23]

    M. Guo, A. Manzoni, M. Amendt, P. Conti, J. S. Hesthaven, Multi-fidelity regression using artificial neural networks: Efficient approximation of parameter-dependent output quantities, Computer Methods in Applied Mechanics and Engineering 389 (2022) 114378.doi:10.1016/j.cma.2021.114378

    work page doi:10.1016/j.cma.2021.114378 2022

  23. [24]

    A. A. Howard, M. Perego, G. E. Karniadakis, P. Stinis, Multifidelity deep operator networks for data-driven and physics-informed problems, Journal of Computational Physics 493 (2023) 112462. doi:10.1016/j.jcp.2023.112462

    work page doi:10.1016/j.jcp.2023.112462 2023

  24. [25]

    Kiener, S

    A. Kiener, S. Langer, P. Bekemeyer, Data-driven correction of coarse grid CFD simulations, Computers & Fluids 264 (2023) 105971.doi:10.1016/j.compfluid.2023.105971

    work page doi:10.1016/j.compfluid.2023.105971 2023

  25. [26]

    S. Kang, E. M. Constantinescu, Learning subgrid-scale models with neural ordinary differential equa- tions, Computers & Fluids 261 (2023) 105919.doi:10.1016/j.compfluid.2023.105919

    work page doi:10.1016/j.compfluid.2023.105919 2023

  26. [27]

    Reviewofsummation-by-partsschemesforinitial-boundary- value problems

    P. Sousa, C. V. Rodrigues, A. Afonso, Enhancing CFD solver with machine learning techniques, Computer Methods in Applied Mechanics and Engineering 429 (2024) 117133.doi:10.1016/j. cma.2024.117133

    work page doi:10.1016/j 2024

  27. [28]

    S. Kang, E. M. Constantinescu, Enhancing low-order discontinuous Galerkin methods with neural ordinary differential equations for compressible Navier–Stokes equations (2025).arXiv:2310.18897, doi:10.48550/arXiv.2310.18897

    work page doi:10.48550/arxiv.2310.18897 2025

  28. [29]

    J. Jung, E. Constantinescu, Learning differentiable weak-form corrections to accelerate finite element simulations (2026).arXiv:2601.20019,doi:10.48550/arXiv.2601.20019

    work page doi:10.48550/arxiv.2601.20019 2026

  29. [30]

    S. Garg, S. Chakraborty, B. Hazra, Physics-integrated hybrid framework for model form error identifi- cation in nonlinear dynamical systems, Mechanical Systems and Signal Processing 173 (2022) 109039. doi:10.1016/j.ymssp.2022.109039. 56

    work page doi:10.1016/j.ymssp.2022.109039 2022

  30. [31]

    & Karniadakis, G

    M. Raissi, P. Perdikaris, G. E. Karniadakis, Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations, Journal of Computational Physics 378 (2019) 686–707.doi:10.1016/j.jcp.2018.10.045

    work page doi:10.1016/j.jcp.2018.10.045 2019

  31. [32]

    J. Yu, L. Lu, X. Meng, G. E. Karniadakis, Gradient-enhanced physics-informed neural networks for forward and inverse PDE problems, Computer Methods in Applied Mechanics and Engineering 393 (2022) 114823.doi:10.1016/j.cma.2022.114823

    work page doi:10.1016/j.cma.2022.114823 2022

  32. [33]

    D. Shu, Z. Li, A. Barati Farimani, A physics-informed diffusion model for high-fidelity flow field reconstruction, Journal of Computational Physics 478 (2023) 111972.doi:10.1016/j.jcp.2023. 111972

    work page doi:10.1016/j.jcp.2023 2023

  33. [34]

    L¨ utjens, B

    B. L¨ utjens, B. Leshchinskiy, C. Requena-Mesa, F. Chishtie, N. D´ıaz-Rodr´ıguez, O. Boulais, A. Pi˜na, D. Newman, A. Lavin, Y. Gal, C. Ra¨ıssi, Physics-informed GANs for coastal flood visualization (2020). arXiv:2010.08103,doi:10.48550/arXiv.2010.08103

    work page doi:10.48550/arxiv.2010.08103 2020

  34. [35]

    Chakravarty, S

    A. Chakravarty, S. Misra, C. S. Rai, Visualization of hydraulic fracture using physics-informed cluster- ing to process ultrasonic shear waves, International Journal of Rock Mechanics and Mining Sciences 137 (2021) 104568.doi:10.1016/j.ijrmms.2020.104568

    work page doi:10.1016/j.ijrmms.2020.104568 2021

  35. [36]

    Banerjee, K

    C. Banerjee, K. Nguyen, C. Fookes, G. E. Karniadakis, Physics-informed computer vision: A review and perspectives, ACM Computing Surveys 57 (1) (2024) 1–38.doi:10.1145/3689037

    work page doi:10.1145/3689037 2024

  36. [37]

    Ohashi, L

    N. Ohashi, L. K. Hwang, B. Kwon, Physics-informed neural networks for multi-field visualization with single-color laser induced fluorescence, AI Thermal Fluids 1 (2025) 100005.doi:10.1016/j.aitf. 2025.100005

    work page doi:10.1016/j.aitf 2025

  37. [38]

    M. G. Cox, The numerical evaluation of B-splines, IMA Journal of Applied Mathematics 10 (2) (1972) 134–149.doi:10.1093/imamat/10.2.134

    work page doi:10.1093/imamat/10.2.134 1972

  38. [39]

    de Boor, On calculating with B-splines, Journal of Approximation Theory 6 (1) (1972) 50–62

    C. de Boor, On calculating with B-splines, Journal of Approximation Theory 6 (1) (1972) 50–62. doi:10.1016/0021-9045(72)90080-9

    work page doi:10.1016/0021-9045(72)90080-9 1972

  39. [40]

    D. C. Liu, J. Nocedal, On the limited memory BFGS method for large scale optimization, Mathematical Programming 45 (1–3) (1989) 503–528.doi:10.1007/BF01589116

    work page doi:10.1007/bf01589116 1989

  40. [41]

    Qiu, LBFGS++ (LBFGSpp): A header-only C++ library for L-BFGS and L-BFGS-B algorithms, GitHub repository, version v0.4.0, commit c524a40 (2025)

    Y. Qiu, LBFGS++ (LBFGSpp): A header-only C++ library for L-BFGS and L-BFGS-B algorithms, GitHub repository, version v0.4.0, commit c524a40 (2025). URLhttps://github.com/yixuan/LBFGSpp

    2025

  41. [42]

    Crank, The Mathematics of Diffusion, 2nd Edition, Clarendon Press, Oxford, 1975

    J. Crank, The Mathematics of Diffusion, 2nd Edition, Clarendon Press, Oxford, 1975

    1975

  42. [43]

    Haberman, Applied Partial Differential Equations: With Fourier Series and Boundary Value Prob- lems, 5th Edition, Pearson, Boston, 2013

    R. Haberman, Applied Partial Differential Equations: With Fourier Series and Boundary Value Prob- lems, 5th Edition, Pearson, Boston, 2013

    2013

  43. [44]

    D. Lenz, O. Marin, V. S. Mahadevan, R. Yeh, T. Peterka, Fourier-informed knot placement schemes for B-spline approximation, Mathematics and Computers in Simulation 213 (2023) 374–393.doi: 10.1016/j.matcom.2023.05.017

    work page doi:10.1016/j.matcom.2023.05.017 2023

  44. [45]

    Pareto, Manual of Political Economy, A

    V. Pareto, Manual of Political Economy, A. M. Kelley, New York, 1971

    1971

  45. [46]

    R. T. Marler, J. S. Arora, Survey of multi-objective optimization methods for engineering, Structural and Multidisciplinary Optimization 26 (6) (2004) 369–395.doi:10.1007/s00158-003-0368-6

    work page doi:10.1007/s00158-003-0368-6 2004

  46. [47]

    R. T. Marler, J. S. Arora, The weighted sum method for multi-objective optimization: new in- sights, Structural and Multidisciplinary Optimization 41 (6) (2010) 853–862.doi:10.1007/ s00158-009-0460-7

    2010

  47. [48]

    F. M. Rohrhofer, S. Posch, C. G ¨oßnitzer, B. C. Geiger, Data vs. physics: The apparent Pareto front of physics-informed neural networks, IEEE Access 11 (2023) 86252–86261.arXiv:2105.00862, doi:10.1109/ACCESS.2023.3302892. 57

    work page doi:10.1109/access.2023.3302892 2023

  48. [49]

    Heldmann, S

    F. Heldmann, S. Berkhahn, M. Ehrhardt, K. Klamroth, PINN training using biobjective optimization: The trade-off between data loss and residual loss, Journal of Computational Physics 488 (2023) 112211. doi:10.1016/j.jcp.2023.112211

    work page doi:10.1016/j.jcp.2023.112211 2023

  49. [50]

    Bischof, M

    R. Bischof, M. A. Kraus, Multi-objective loss balancing for physics-informed deep learning, Computer Methods in Applied Mechanics and Engineering 439 (2025) 117914.doi:10.1016/j.cma.2025. 117914

    work page doi:10.1016/j.cma.2025 2025

  50. [51]

    J. D. Cole, On a quasi-linear parabolic equation occurring in aerodynamics, Quarterly of Applied Mathematics 9 (3) (1951) 225–236.doi:10.1090/qam/42889

    work page doi:10.1090/qam/42889 1951

  51. [52]

    R. J. LeVeque, Finite Volume Methods for Hyperbolic Problems, Cambridge University Press, Cam- bridge, 2002.doi:10.1017/CBO9780511791253

    work page doi:10.1017/cbo9780511791253 2002

  52. [53]

    Clark di Leoni, A

    P. Clark di Leoni, A. N. Hirata, P. D. Mininni, Dynamics of partially thermalized solutions of the Burgers equation, Physical Review Fluids 3 (2018) 014603.doi:10.1103/PhysRevFluids.3.014603

    work page doi:10.1103/physrevfluids.3.014603 2018

  53. [54]

    L. C. Evans, Partial Differential Equations, 2nd Edition, Vol. 19 of Graduate Studies in Mathematics, American Mathematical Society, Providence, RI, 2010

    2010

  54. [55]

    Borthwick, Introduction to Partial Differential Equations, Springer, Cham, 2017.doi:10.1007/ 978-3-319-48936-0

    D. Borthwick, Introduction to Partial Differential Equations, Springer, Cham, 2017.doi:10.1007/ 978-3-319-48936-0

    2017

  55. [56]

    M. C. Kennedy, A. O’Hagan, Bayesian calibration of computer models, Journal of the Royal Statistical Society: Series B (Statistical Methodology) 63 (3) (2001) 425–464.doi:10.1111/1467-9868. 00294

    work page doi:10.1111/1467-9868 2001

  56. [57]

    Hooker, Forcing function diagnostics for nonlinear dynamics, Biometrics 65 (3) (2009) 928–936

    G. Hooker, Forcing function diagnostics for nonlinear dynamics, Biometrics 65 (3) (2009) 928–936. doi:10.1111/j.1541-0420.2008.01172.x

    work page doi:10.1111/j.1541-0420.2008.01172.x 2009

  57. [58]

    R. E. Morrison, T. A. Oliver, R. D. Moser, Representing model inadequacy: A stochastic operator approach, SIAM/ASA Journal on Uncertainty Quantification 6 (2) (2018) 457–496.doi:10.1137/ 16M1106419

    2018

  58. [59]

    Masud, S

    A. Masud, S. Nashar, S. A. Goraya, Physics-constrained data-driven variational method for missing physics and model-form error, Computer Methods in Applied Mechanics and Engineering 417 (2023) 116295.doi:10.1016/j.cma.2023.116295

    work page doi:10.1016/j.cma.2023.116295 2023

  59. [60]

    D. A. Ham, P. H. J. Kelly, L. Mitchell, C. J. Cotter, R. C. Kirby, K. Sagiyama, N. Bouziani, S. Vorder- wuelbecke, T. J. Gregory, J. Betteridge, D. R. Shapero, R. W. Nixon-Hill, C. J. Ward, P. E. Farrell, P. D. Brubeck, I. Marsden, T. H. Gibson, M. Homolya, T. Sun, A. T. T. McRae, F. Luporini, A. Gregory, M. Lange, S. W. Funke, F. Rathgeber, G.-T. Bercea,...

    work page doi:10.25561/104839 2023

  60. [61]

    U. Ghia, K. N. Ghia, C. T. Shin, High-Re solutions for incompressible flow using the Navier–Stokes equations and a multigrid method, Journal of Computational Physics 48 (3) (1982) 387–411.doi: 10.1016/0021-9991(82)90058-4

    work page doi:10.1016/0021-9991(82)90058-4 1982

  61. [62]

    Shen, Hopf bifurcation of the unsteady regularized driven cavity flow, Journal of Computational Physics 95 (1) (1991) 228–245.doi:10.1016/0021-9991(91)90261-I

    J. Shen, Hopf bifurcation of the unsteady regularized driven cavity flow, Journal of Computational Physics 95 (1) (1991) 228–245.doi:10.1016/0021-9991(91)90261-I

    work page doi:10.1016/0021-9991(91)90261-i 1991

  62. [63]

    Kochkov, J

    D. Kochkov, J. A. Smith, A. Alieva, Q. Wang, M. P. Brenner, S. Hoyer, Machine learning-accelerated computational fluid dynamics, Proceedings of the National Academy of Sciences 118 (21) (2021) e2101784118.doi:10.1073/pnas.2101784118

    work page doi:10.1073/pnas.2101784118 2021

  63. [64]

    Pathak, M

    J. Pathak, M. Mustafa, K. Kashinath, E. Motheau, T. Kurth, M. Day, Using machine learning to augment coarse-grid computational fluid dynamics simulations (2020).arXiv:2010.00072,doi: 10.48550/arXiv.2010.00072. 58

    work page doi:10.48550/arxiv.2010.00072 2020

  64. [65]

    Z. Li, N. Kovachki, K. Azizzadenesheli, B. Liu, K. Bhattacharya, A. Stuart, A. Anandkumar, Fourier neural operator for parametric partial differential equations, in: International Conference on Learning Representations, 2021.arXiv:2010.08895,doi:10.48550/arXiv.2010.08895. URLhttps://openreview.net/forum?id=c8P9NQVtmnO

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2010.08895 2021

  65. [66]

    T. J. R. Hughes, The Finite Element Method: Linear Static and Dynamic Finite Element Analysis, Prentice-Hall, Englewood Cliffs, NJ, 1987

    1987

  66. [67]

    Ern, J.-L

    A. Ern, J.-L. Guermond, Theory and Practice of Finite Elements, Vol. 159 of Applied Mathematical Sciences, Springer, New York, 2004.doi:10.1007/978-1-4757-4355-5

    work page doi:10.1007/978-1-4757-4355-5 2004

  67. [68]

    Raissi, A

    M. Raissi, A. Yazdani, G. E. Karniadakis, Hidden fluid mechanics: Learning velocity and pressure fields from flow visualizations, Science 367 (6481) (2020) 1026–1030.doi:10.1126/science.aaw4741

    work page doi:10.1126/science.aaw4741 2020

  68. [69]

    J. A. Cottrell, T. J. R. Hughes, Y. Bazilevs, Isogeometric Analysis: Toward Integration of CAD and FEA, Wiley, Chichester, 2009.doi:10.1002/9780470749081. Government License (will be removed at publication): The submitted manuscript has been created by UChicago Argonne, LLC, Operator of Argonne National Laboratory (“Argonne”). Argonne, a U.S. Department o...

    work page doi:10.1002/9780470749081 2009