Pith. sign in

REVIEW 4 major objections 7 minor 40 references

Sparse displacement sensors plus a regularized ensemble Kalman filter recover both the displacement field and the hidden phase-field crack in stochastic brittle-fracture models.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

A phase-field-regularized EnKF updates both displacement and crack phase-field states from sparse noisy displacement data, recovering crack location and residual strength better than the open-loop ensemble.

T0 review reviewed 2026-07-15 challenge →

load-bearing objection Full-state EnKF for phase-field fracture is a genuine step past parametric crack filters; the regularization fix works for localization but can bias residual stiffness, and all evidence is still synthetic. the 4 major comments →

arxiv 2603.09728 v1 pith:I4XL2LDN submitted 2026-03-10 cs.CE

A Regularized Ensemble Kalman Filter for Stochastic Phase Field Models of Brittle Fracture

Lucas Hermann , Ralf J\"anicke , Knut Andreas Meyer , Ulrich R\"omer This is my paper
classification cs.CE
keywords phase-field fractureensemble Kalman filterdata assimilationbrittle fractureBayesian state estimationmicromorphic modelcrack path uncertainty
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

The reading

Uncertain initial damage makes crack paths and remaining structural strength non-unique in phase-field models of brittle fracture. This paper shows that an ensemble Kalman filter can assimilate sparse, noisy displacement measurements into the high-dimensional model state so that both displacements and the phase-field are corrected online. Because a plain Kalman shift produces unphysical fields (negative phase-field values, oscillations), the authors add a short staggered re-solve with an inflated length scale that acts as a proximal correction toward model-consistent states. One- and two-dimensional numerical examples demonstrate that the posterior ensemble collapses onto a synthetic ground-truth crack path and residual load capacity even though damage is never observed directly. The result matters for structural monitoring: when material defects leave the crack path uncertain, sensor data can still recover the damage field and tighten estimates of remaining strength.

Core claim

With only sparse noisy displacement observations, a regularized ensemble Kalman filter can assimilate the high-dimensional state (displacements and phase-field) of a stochastic micromorphic phase-field fracture model so that the posterior ensemble matches a synthetic ground-truth crack path and residual load capacity reasonably well, whereas the unassimilated ensemble does not.

What carries the argument

Phase-field-based regularization of the EnKF analysis: after the Kalman shift on each ensemble member, a few staggered residual solves with an inflated length scale L > ℓ and temporarily lifted irreversibility project the state back onto a model-admissible manifold while damping spurious oscillations.

Load-bearing premise

That a few staggered re-solves with a larger crack-width parameter and temporarily ignoring irreversibility restore physical states without systematically changing how fast the crack later grows.

What would settle it

On the 2-D single-edge-notch shear benchmark, if after assimilation the ensemble-mean peak reaction force systematically lies outside the reported posterior spread of the ground-truth peak, or if the inferred phase-field maximum drifts away from the true crack path when sensor density is increased or noise is reduced, the central claim fails.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

If this is right

  • Sparse displacement sensors alone can recover the full phase-field crack without measuring damage directly.
  • Residual structural-strength estimates tighten after each assimilation step, supporting better remaining-life decisions.
  • Non-unique crack paths caused by uncertain initial damage become identifiable online as data arrive.
  • The same filter-plus-regularization pattern can be applied whenever only kinematics are observed in a history-dependent continuum damage model.

Where Pith is reading between the lines

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

  • A strongly constrained variational smoother (single-step 4D-Var) could replace the ad-hoc proximal steps if the extra model solves become affordable.
  • The method is a natural candidate for online Digital Image Correlation data once model-error kernels and localization are re-tuned for experimental noise.
  • Because the phase-field is inferred only through correlation with displacement, the same idea may transfer to other dual-field continuum models (e.g., poroelasticity or plasticity) where one field is hard to observe.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 7 minor

Summary. The paper proposes a Bayesian data-assimilation procedure for stochastic micromorphic phase-field models of brittle fracture. A prior ensemble is generated from random initial damage, propagated with a monolithic FEM solver, and updated with sparse noisy displacement observations via an ensemble Kalman filter (EnKF). Because unconstrained Kalman shifts produce unphysical states (negative phase field, oscillatory displacements), the authors introduce a post-analysis regularization: staggered residual re-solves that temporarily lift irreversibility (φ′=0), inflate the length scale (L>ℓ), then restore the original ℓ. 1D tension-rod and 2D single-edge-notched shear (SENS) synthetic studies show that the regularized posterior ensemble localizes the crack path and tightens reaction-force and peak-force distributions toward a held-out ground truth, whereas the unassimilated ensemble does not. The contribution is framed as state inference (displacements and phase field) rather than parameter inversion.

Significance. If the method works as claimed, it offers a practical route to fuse sensor data (e.g. DIC-type displacements) with high-dimensional phase-field fracture simulations without parametrizing crack geometry a priori—an advantage over existing EKF/EnKF work on XFEM or few-parameter crack descriptions. The combination of EnKF state update with an explicit phase-field proximal correction is novel in this application area, and the manuscript is transparent about limitations (non-Gaussianity, cost, need for regularization). Planned code release and clear Algorithms 1–2 support reproducibility. The central scientific value is therefore real, but it rests on the claim that regularization restores model-consistent states without systematically distorting residual strength—an assumption that the present synthetic evidence only partially substantiates.

major comments (4)
  1. §3.3.3, Algorithm 2 and Fig. 6b–c: The load-bearing proximal correction uses L=4ℓ (Table 4), φ′=0, and identity-weighted residual solves. The text itself states that the wider phase field increases accumulated damage, lowers remaining stiffness, and can accelerate subsequent propagation. This bias can partially explain the collapse of reaction-force ensembles toward the reference in Figs. 14 and 17. Without an ablation that isolates (i) EnKF shift alone, (ii) regularization alone, and (iii) EnKF+regularization for several L/n_stagger values, it is not clear how much of the reported residual-strength match is genuine assimilation versus L-driven damage inflation. A quantitative sensitivity study on residual force error and crack-position error versus L and n_stagger is needed to support the central claim.
  2. §5, eqs. (35)–(36): The proximal interpretation replaces the ensemble covariance C_d by the identity for memory reasons. That choice severs the link between the Kalman analysis covariance and the projection weights, so the procedure is no longer a true proximal map of the EnKF objective. The manuscript should either restore a diagonal/localized approximation of C_d or clearly reframe the step as a heuristic staggered projection rather than a proximal correction, and discuss the effect on the Bayesian interpretation of the posterior ensemble.
  3. §§3.3–4 and Figs. 12–19: All validation is synthetic, with ground truth generated from the same micromorphic AT2 model (finer mesh, fixed initial damage outside the prior). While inverse crime is partially avoided, success metrics remain within one model family. The claim that the method recovers crack path and residual capacity “reasonably well” from displacements alone would be substantially stronger with at least one misspecified-physics or real DIC-style experiment, or—if that is out of scope—with explicit quantitative error tables (e.g. L2 phase-field error, crack-tip location error, peak-force bias) for prior vs. posterior across ensemble members, not only qualitative figures and histograms.
  4. §3.3 and the near-Gaussian discussion: EnKF theory is invoked for a problem with strong nonlinearity and history-dependent irreversibility (nucleation, abrupt loss of stiffness). The paper notes that accuracy is “less obvious” outside the linear-Gaussian setting but provides no diagnostic (e.g. ensemble collapse indicators, non-Gaussianity of phase-field marginals, or comparison to a particle-filter baseline on the 1D problem). A short diagnostic subsection quantifying when the Gaussian update fails (e.g. pre- vs. post-nucleation) would make the applicability bounds of the method clearer.
minor comments (7)
  1. Notation: a_n,k vs. a_n,i vs. a^F/A/R is dense; a short notation table early in §1.1 would help.
  2. Fig. 6 is split across pages with subcaptions a/b/c; combining into a single multi-panel figure with consistent axis scales would improve readability.
  3. Table 1 and Table 2: α is written as “β G_c/ℓ” without defining β; state the numerical value used.
  4. §3.2: Matérn hyperparameter learning is described but the optimized (ν,σ,l) values used in the 1D/2D examples are not reported; please list them.
  5. Related work: brief comparison to recent sequential data assimilation for continuum damage or phase-field fatigue (beyond the parametric EKF/XFEM citations) would better position the contribution.
  6. Typos/style: “F orecast” heading spacing; occasional missing spaces after commas in math mode; “statFEM” acronym introduced without expansion on first use in the main text.
  7. Appendix A: localization length l_loc=0.45 and inflation r=1.05 are given for 2D only; state the 1D choices and whether results are sensitive to them.

Circularity Check

0 steps flagged

No circular derivation: posterior match is empirical validation against held-out synthetic ground truth, not a quantity forced by construction or self-citation.

full rationale

The paper's load-bearing claim is that a regularized EnKF, given sparse noisy displacements, updates the high-dimensional state (u, d/φ) so the posterior ensemble tracks a synthetic ground-truth crack path and residual strength better than the unassimilated prior. That claim is tested, not defined: ground truth uses a fixed initial damage outside the prior ensemble and a finer mesh (explicit inverse-crime avoidance), with additive observation noise; success is visual/histogram comparison of posterior vs that held-out truth (Figs. 8–19, reaction-force PDFs). The EnKF analysis is the standard Kalman shift (Eqs. 26–27) applied to a Monte-Carlo forecast; the subsequent staggered re-solves with L>ℓ and φ′=0 (Algorithm 2, §5 proximal form) are a proposed post-processing correction whose side-effects on stiffness the authors themselves document—they are not algebraically identical to the reported crack location or load capacity. Hyperparameters (Matérn data-model, inflation r, localization, L=4ℓ) are chosen/tuned but do not force the crack path by construction. Self-citations ([16] micromorphic model with co-author Jänicke; [5] prior stochastic phase-field with co-author Römer) supply the forward model and UQ context; they do not import a uniqueness theorem or ansatz that makes the assimilation result tautological. No fitted input is renamed as a prediction of the same quantity. The derivation chain is methodological application plus numerical demonstration, self-contained against an external synthetic benchmark.

Axiom & Free-Parameter Ledger

7 free parameters · 5 axioms · 1 invented entities

The central claim rests on standard phase-field fracture and EnKF machinery plus several hand-chosen filter/regularization knobs and the modeling choice that a short unconstrained staggered re-solve restores physical states. No new physical particles or forces are invented; the ‘regularized EnKF’ is a procedural construct. Synthetic Gaussian observation and ensemble-Gaussian approximations are domain assumptions that limit transfer to real sensors and highly multimodal crack ensembles.

free parameters (7)
  • Regularization length scale L (relative to ℓ) = 4ℓ
    Chosen L=4ℓ in 2D (Table 4); controls smoothing of EnKF artifacts and accumulated damage; not derived from first principles.
  • Staggered regularization iteration count niter / n_stagger = 4 (2D); 1 or 4 (1D demos)
    Hand-set (niter=4 in Table 4; 1D figures compare 1 vs 4); changes inferred phase-field magnitude and stiffness.
  • Covariance inflation factor r = 1.05
    Appendix A; r=1.05 in Table 3; standard EnKF fix for finite ensembles, chosen not learned from external data.
  • Localization length l_loc = 0.45
    Table 3 l_loc=0.45; squared-exponential localization to kill spurious correlations; problem-specific choice.
  • Data-model Matérn hyperparameters w=(ν,σ,l) and sensor noise σ_e = σ_e=4e-4; others optimized per setup
    Calibrated by maximizing data likelihood / set (σ_e=4e-4 in Table 3); affect Kalman gain weight of data vs prior.
  • Micromorphic penalty scale α=β G_c/ℓ and initial-damage prior parameters = α=β G_c/ℓ; Beta(8,8) pore locations in 2D
    Model and prior setup (Tables 1–2; Beta/Normal/Uniform nuclei); shape the forecast ensemble that EnKF conditions.
  • Ensemble size n_ens and sensor count n_sens = 100 / 100 / 20
    n_ens=100, n_sens=100, n_obs=20 (Table 3); finite-sample approximation quality and data weight depend on these choices.
axioms (5)
  • domain assumption Micromorphic AT2 phase-field brittle fracture with volumetric-deviatoric split and local irreversibility/bounds on φ is an adequate continuum model of the crack process.
    §2 energy (4) and local solve (10); entire state vector is defined in this model.
  • domain assumption Finite-ensemble EnKF with inflation and localization adequately approximates Bayesian filtering for this nonlinear, history-dependent fracture system.
    §3.3 cites near-Gaussian theory [31,32,23]; authors note true filter convergence is only guaranteed in linear-Gaussian settings.
  • domain assumption Displacement observations are linearly related to the FE state via H plus independent Gaussian model mismatch and sensor noise (statFEM-style data model).
    §3.2 eqs. (17)–(20); Kalman shift closed form depends on this Gaussian structure.
  • ad hoc to paper A few staggered residual solves with φ′=0 and temporarily larger L act as a proximal map that restores model-admissible states without needing a full constrained 3D/4D-Var solve.
    §3.3.3 Algorithms 1–2 and §5 eqs. (36); central fix for unphysical EnKF analyses.
  • standard math Standard multivariate Gaussian conditioning / Kalman algebra and FEM residual minimization are valid numerical tools.
    Used throughout §§2–3 for Newton solves and EnKF updates.
invented entities (1)
  • Phase-field proximal regularization of EnKF analysis (large-L then original-ℓ staggered re-solve with lifted irreversibility) no independent evidence
    purpose: Map data-driven Kalman-shifted states back toward the phase-field admissible set and suppress oscillations so subsequent load steps converge.
    Introduced in §3.3.3 and interpreted as proximal correction in §5; not an independent physical object, but a paper-specific algorithmic entity the central claim depends on.

reviewed 2026-07-15 · how reviews work

0 comments
Cite this review

Pith. "Pith review of A Regularized Ensemble Kalman Filter for Stochastic Phase Field Models of Brittle Fracture." pith.science (2026). https://pith.science/paper/I4XL2LDN

@misc{pith2026260309728,
  author       = {Pith},
  title        = {Pith review of: A Regularized Ensemble Kalman Filter for Stochastic Phase Field Models of Brittle Fracture},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/I4XL2LDN}},
  note         = {Machine review of arXiv:2603.09728}
}
Share X Bluesky LinkedIn Reddit HN
read the original abstract

The phase-field approach to brittle fracture provides a continuum framework for modeling crack initiation and propagation without explicit representation of discrete crack surfaces, provided the spatial discretization is fine enough to resolve the regularization length scale. However, uncertain local material parameters due to material defects can strongly influence simulation results, such as crack paths and remaining structural strength. At the same time, the ability to continuously monitor structures using sensors allows complementing modeling predictions with, e.g., displacement measurements. In this contribution, we connect these two complementary sources of information and present a Bayesian inference procedure that allows updating the current model state with incoming sensor data. We construct a Bayesian prior for the model state (both displacements and phase-field) and employ an ensemble Kalman filter (EnKF) to perform the update. In the EnKF, the update is computed by performing a Kalman shift on each ensemble member. Since the standard EnKF may produce assimilated states that violate common modeling assumptions, we present a phase field-based regularization technique as a proximal step correction toward model-consistent updates. 1D and 2D numerical examples demonstrate the performance and accuracy of the proposed method and show that the updated state matches the ground truth reasonably well. Unlike traditional Bayesian inversion techniques, which have already been applied to brittle fracture, we infer not the model parameters but the model state, i.e., the displacement field and the phase-field. Although only displacements are observed, the strong correlation between both fields also allows inference of the posterior phase-field.

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 · 15 canonical work pages · 1 internal anchor

  1. [1]

    G. A. Francfort, J. J. Marigo, Revisiting brittle fracture as an energy minimization problem, Journal of the Mechanics and Physics of Solids 46 (8) (1998) 1319–1342.doi:10.1016/S0022-5096(98)00034-9

    doi:10.1016/s0022-5096(98)00034-9 1998

  2. [2]

    Miehe, F

    C. Miehe, F. Welschinger, M. Hofacker, Thermodynamically consistent phase-field models of fracture: Variational principles and multi-field FE implementations, International Journal for Numerical Methods in Engineering 83 (10) (2010) 1273–1311.doi:10.1002/nme.2861

    doi:10.1002/nme.2861 2010

  3. [3]

    Ambati, T

    M. Ambati, T. Gerasimov, L. De Lorenzis, Phase-field modeling of ductile fracture, Computational Mechanics 55 (5) (2015) 1017–1040.doi:10.1007/s00466-015-1151-4

    doi:10.1007/s00466-015-1151-4 2015

  4. [4]

    Kalina, T

    M. Kalina, T. Schneider, J. Brummund, M. K¨ astner, Overview of phase-field models for fatigue frac- ture in a unified framework, Engineering Fracture Mechanics 288 (2023) 109318.doi:10.1016/j. engfracmech.2023.109318

    doi:10.1016/j 2023

  5. [5]

    Gerasimov, U

    T. Gerasimov, U. R¨ omer, J. Vondˇ rejc, H. G. Matthies, L. De Lorenzis, Stochastic phase-field modeling of brittle fracture: Computing multiple crack patterns and their probabilities, Computer Methods in Applied Mechanics and Engineering 372 (2020) 113353.doi:10.1016/j.cma.2020.113353

    work page doi:10.1016/j.cma.2020.113353 2020

  6. [6]

    Nagaraja, U

    S. Nagaraja, U. R¨ omer, H. G. Matthies, L. De Lorenzis, Deterministic and stochastic phase-field model- ing of anisotropic brittle fracture, Computer Methods in Applied Mechanics and Engineering 408 (2023) 115960.doi:10.1016/j.cma.2023.115960

    work page doi:10.1016/j.cma.2023.115960 2023

  7. [7]

    TC. Chu, WF. Ranson, M. A. Sutton, Applications of digital-image-correlation techniques to experi- mental mechanics, Experimental mechanics 25 (1985) 232–244

    1985

  8. [8]

    Girolami, E

    M. Girolami, E. Febrianto, G. Yin, F. Cirak, The statistical finite element method (statFEM) for coherent synthesis of observation data and model predictions, Computer Methods in Applied Mechanics and Engineering 375 (2021) 113533.doi:10.1016/j.cma.2020.113533

    work page doi:10.1016/j.cma.2020.113533 2021

  9. [9]

    Duffin, E

    C. Duffin, E. Cripps, T. Stemler, M. Girolami, Statistical finite elements for misspecified models, Proceedings of the National Academy of Sciences 118 (2) (2021) e2015006118.doi:10.1073/pnas. 2015006118

    doi:10.1073/pnas 2021

  10. [10]

    T. Wu, B. Rosi´ c, L. De Lorenzis, H. G. Matthies, Parameter identification for phase-field modeling of fracture: A Bayesian approach with sampling-free update, Computational Mechanics 67 (2) (2021) 435–453.doi:10.1007/s00466-020-01942-x

    doi:10.1007/s00466-020-01942-x 2021

  11. [11]

    Pulikkathodi, E

    A. Pulikkathodi, E. Lacazedieu, L. Chamoin, Real-time inverse crack tracking in uncertain microstruc- tures using PGD-based model reduction and extended Kalman filtering, Computational Mechanics 71 (2) (2023) 311–332.doi:10.1007/s00466-022-02240-4

    work page doi:10.1007/s00466-022-02240-4 2023

  12. [12]

    F. A. Rochinha, A. Peirce, Monitoring hydraulic fractures: State estimation using an extended Kalman filter, Inverse Problems 26 (2) (2010) 025009.doi:10.1088/0266-5611/26/2/025009

    work page doi:10.1088/0266-5611/26/2/025009 2010

  13. [13]

    Kuncham, S

    E. Kuncham, S. Sen, P. Kumar, H. Pathak, An online model-based fatigue life prediction approach using extended Kalman filter, Theoretical and Applied Fracture Mechanics 117 (2022) 103143.doi: 10.1016/j.tafmec.2021.103143. 23

    work page doi:10.1016/j.tafmec.2021.103143 2022

  14. [14]

    J. Ping, D. Zhang, History matching of fracture distributions by ensemble Kalman filter combined with vector based level set parameterization, Journal of Petroleum Science and Engineering 108 (2013) 288– 303.doi:10.1016/j.petrol.2013.04.018

    work page doi:10.1016/j.petrol.2013.04.018 2013

  15. [15]

    Cervera, G

    M. Cervera, G. B. Barbat, M. Chiumenti, J.-Y. Wu, A Comparative Review of XFEM, Mixed FEM and Phase-Field Models for Quasi-brittle Cracking, Archives of Computational Methods in Engineering 29 (2) (2022) 1009–1083.doi:10.1007/s11831-021-09604-8

    doi:10.1007/s11831-021-09604-8 2022

  16. [16]

    Bharali, F

    R. Bharali, F. Larsson, R. J¨ anicke, A micromorphic phase-field model for brittle and quasi-brittle frac- ture, Computational Mechanics 73 (3) (2024) 579–598.doi:10.1007/s00466-023-02380-1

    work page doi:10.1007/s00466-023-02380-1 2024

  17. [17]

    De Lorenzis, T

    L. De Lorenzis, T. Gerasimov, Numerical Implementation of Phase-Field Models of Brittle Fracture, in: L. De Lorenzis, A. D¨ uster (Eds.), Modeling in Engineering Using Innovative Numerical Methods for Solids and Fluids, Springer International Publishing, Cham, 2020, pp. 75–101

    2020

  18. [18]

    Khimin, L

    D. Khimin, L. Kolditz, V. Kosin, K. Mang, T. Wick, Modeling, Discretization, Optimization, and Simulation of Phase-Field Fracture Problems (Nov. 2023).doi:10.15488/15172

    work page doi:10.15488/15172 2023

  19. [19]

    Heister, M

    T. Heister, M. F. Wheeler, T. Wick, A primal-dual active set method and predictor-corrector mesh adap- tivity for computing fracture propagation using a phase-field approach, Computer Methods in Applied Mechanics and Engineering 290 (2015) 466–495.doi:10.1016/j.cma.2015.03.009

    work page doi:10.1016/j.cma.2015.03.009 2015

  20. [20]

    Kristensen, Phase field methods for fracture - With a special focus on hydrogen assisted fracture, Ph.D

    P. Kristensen, Phase field methods for fracture - With a special focus on hydrogen assisted fracture, Ph.D. thesis, Technical University of Denmark (2022)

    2022

  21. [21]

    Evensen, Data Assimilation: The Ensemble Kalman Filter, Springer, Berlin, Heidelberg, 2009.doi: 10.1007/978-3-642-03711-5

    G. Evensen, Data Assimilation: The Ensemble Kalman Filter, Springer, Berlin, Heidelberg, 2009.doi: 10.1007/978-3-642-03711-5

    doi:10.1007/978-3-642-03711-5 2009

  22. [22]

    C. E. Rasmussen, C. K. I. Williams, Gaussian Processes for Machine Learning, Adaptive Computation and Machine Learning, MIT Press, Cambridge, Mass, 2006

    2006

  23. [23]

    Sanz-Alonso, A

    D. Sanz-Alonso, A. M. Stuart, A. Taeb, Inverse Problems and Data Assimilation (Feb. 2023).arXiv: 1810.06191,doi:10.48550/arXiv.1810.06191

    Pith/arXiv arXiv doi:10.48550/arxiv.1810.06191 2023

  24. [24]

    Evensen, The Ensemble Kalman Filter: Theoretical formulation and practical implementation, Ocean Dynamics 53 (4) (2003) 343–367.doi:10.1007/s10236-003-0036-9

    G. Evensen, The Ensemble Kalman Filter: Theoretical formulation and practical implementation, Ocean Dynamics 53 (4) (2003) 343–367.doi:10.1007/s10236-003-0036-9

    doi:10.1007/s10236-003-0036-9 2003

  25. [25]

    Gordon, D

    N. Gordon, D. Salmond, A. Smith, Novel approach to nonlinear/non-Gaussian Bayesian state estimation, IEE Proceedings F Radar and Signal Processing 140 (2) (1993) 107.doi:10.1049/ip-f-2.1993.0015

    doi:10.1049/ip-f-2.1993.0015 1993

  26. [26]

    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

    doi:10.1111/1467-9868.00294 2001

  27. [27]

    J. P. Kaipio, E. Somersalo, Statistical and Computational Inverse Problems, Vol. 160 of Applied Math- ematical Sciences, Springer, New York, NY, 2005.doi:10.1007/b138659

    doi:10.1007/b138659 2005

  28. [28]

    Doucet, S

    A. Doucet, S. Godsill, C. Andrieu, On sequential Monte Carlo sampling methods for Bayesian filtering, Statistics and Computing 10 (3) (2000) 197–208.doi:10.1023/a:1008935410038

    doi:10.1023/a:1008935410038 2000

  29. [29]

    Giles, M

    D. Giles, M. M. Graham, M. Giordano, T. Koskela, A. Beskos, S. Guillas, ParticleDA.jl v.1.0: A real- time data assimilation software platform, Geoscientific Model Development 17 (6) (2024) 2427–2445. doi:10.5194/gmd-2023-38

    work page doi:10.5194/gmd-2023-38 2024

  30. [30]

    Bengtsson, P

    T. Bengtsson, P. Bickel, B. Li, Curse-of-dimensionality revisited: Collapse of the particle filter in very large scale systems, in: Probability and Statistics: Essays in Honor of David A. Freedman, Vol. 2, Institute of Mathematical Statistics, 2008, pp. 316–335.doi:10.1214/193940307000000518

    doi:10.1214/193940307000000518 2008

  31. [31]

    Calvello, J

    E. Calvello, J. A. Carrillo, F. Hoffmann, P. Monmarch´ e, A. M. Stuart, U. Vaes, Statistical accuracy of the ensemble Kalman filter in the near-linear setting (Mar. 2025).arXiv:2503.16154,doi:10.48550/ arXiv.2503.16154

    Pith/arXiv arXiv 2025

  32. [32]

    Accuracy of the Ensemble Kalman Filter in the Near-Linear Setting

    E. Calvello, P. Monmarch´ e, A. M. Stuart, U. Vaes, Accuracy of the Ensemble Kalman Filter in the Near-Linear Setting (Feb. 2025).arXiv:2409.09800,doi:10.48550/arXiv.2409.09800. 24

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2409.09800 2025

  33. [33]

    Kirkesæther Brun, T

    M. Kirkesæther Brun, T. Wick, I. Berre, J. M. Nordbotten, F. A. Radu, An iterative staggered scheme for phase field brittle fracture propagation with stabilizing parameters, Computer Methods in Applied Mechanics and Engineering 361 (2020) 112752.doi:10.1016/j.cma.2019.112752

    work page doi:10.1016/j.cma.2019.112752 2020

  34. [34]

    Hofacker, C

    M. Hofacker, C. Miehe, Continuum phase field modeling of dynamic fracture: Variational principles and staggered FE implementation, International Journal of Fracture 178 (1-2) (2012) 113–129.doi: 10.1007/s10704-012-9753-8

    work page doi:10.1007/s10704-012-9753-8 2012

  35. [35]

    Schillings, A

    C. Schillings, A. M. Stuart, Analysis of the Ensemble Kalman Filter for Inverse Problems, SIAM Journal on Numerical Analysis 55 (3) (2017) 1264–1290.doi:10.1137/16M105959X

    doi:10.1137/16m105959x 2017

  36. [36]

    T. M. Hamill, J. S. Whitaker, Accounting for the Error due to Unresolved Scales in Ensemble Data Assimilation: A Comparison of Different Approaches, Monthly Weather Review 133 (11) (2005) 3132– 3147.doi:10.1175/MWR3020.1

    work page doi:10.1175/mwr3020.1 2005

  37. [37]

    P. L. Houtekamer, H. L. Mitchell, A Sequential Ensemble Kalman Filter for Atmospheric Data Assimi- lation, Monthly Weather Review 129 (1) (2001) 123–137

    2001

  38. [38]

    Gaspari, S

    G. Gaspari, S. E. Cohn, Construction of correlation functions in two and three dimensions, Quarterly Journal of the Royal Meteorological Society 125 (554) (1999) 723–757.doi:10.1002/qj.49712555417

    doi:10.1002/qj.49712555417 1999

  39. [39]

    Bezanson, A

    J. Bezanson, A. Edelman, S. Karpinski, V. B. Shah, Julia: A Fresh Approach to Numerical Computing, SIAM Review 59 (1) (2017) 65–98.doi:10.1137/141000671

    doi:10.1137/141000671 2017

  40. [40]

    Carlsson, F

    K. Carlsson, F. Ekre, Ferrite. jl contributors, Ferrite.jl, Zenodo (Jan. 2026).doi:10.5281/ZENODO. 13862652. 25

    doi:10.5281/zenodo 2026

This paper was first reviewed by grok-4.5 on July 15, 2026.