Efficient Parallel-in-Time Solution of Time-Periodic Problems Using a Multi-Harmonic Coarse Grid Correction

Iryna Kulchytska-Ruchka; Sebastian Sch\"ops

REVIEW 3 major objections 4 minor 34 references

Efficient Parallel-in-Time Solution of Time-Periodic Problems Using a Multi-Harmonic Coarse Grid Correction

T0 review · 3 major / 4 minor · reviewed 2026-08-14 · deepseek-v4-flash

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

Pith's one-line read This paper claims that a simplified Newton iteration with a specially chosen initial approximation turns the periodic Parareal coarse correction into a block-diagonal frequency-domain problem, enabling parallel solution of each harmonic…

desk verdict A genuine algorithmic trick for making nonlinear periodic Parareal FFT-diagonalizable, with a convergence proof that covers less ground than the experiments suggest. read the letter →

arxiv 1908.05245 v2 pith:RRDM355M submitted 2019-08-13 math.NA cs.NA

Iryna Kulchytska-Ruchka , Sebastian Schöps This is my paper · ORCID

classification math.NAcs.NA MSC 34A3434A3634A3765L2078M10

keywords time-periodicproblemsparallel-in-timePararealalgorithmmulti-harmonicmethodfrequencydomainsolutionsimplifiedNewtoniterationeddycurrentproblemfastFouriertransform

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper claims that nonlinear time-periodic problems can be solved much faster in parallel by combining the periodic Parareal method with a multi-harmonic coarse-grid correction. The key idea is to linearize the nonlinear PP-PC system with a simplified Newton iteration whose initial guess is chosen as a fixed vector plus the known fine-coarse discrepancy at each time window. This special choice makes the Jacobian constant, block-cyclic, and diagonalizable by the discrete Fourier transform, so the correction splits into independent frequency components. The claim is supported by convergence analysis for a scalar model problem and by eddy-current experiments showing speedups of up to 271x for the 2D cable model and 176x for the 3D transformer model.

What carries the argument

The central object is the simplified Newton iteration (5.5) together with the initial approximation (5.8). Choosing $U^{(k+1,0)} = (Z + b_n)$ in every time window makes the Jacobian $G_d$ in (5.10) block-cyclic with identical diagonal blocks $Q_d(Z)$; the DFT diagonalizes that matrix into blocks $[\hat G_d]_{jj} = Q_d(Z) - C e^{-i\Delta T \omega_j}$, so each frequency component is solved independently and in parallel. The fast Fourier transform keeps the cost of the transforms low, so the coarse-grid correction adds little overhead.

What would settle it

Solve the scalar model problem (5.27) with a nonlinearity satisfying Assumption 5.4 but with $L_2/c_1$ large (e.g., a steep monotone magnetization curve), start with $Z=0$, and compute $h_0$; if $h_0 > 1/2$ and the simplified Newton iteration fails to converge in the ball predicted by Theorem 5.3, the claimed sufficient condition is not delivering the guarantee in practice.

Watch

Extended reading notes

Core claim

For a nonlinear time-periodic system $Mu' + K(u)u = j$, the paper's proposal is to solve the periodic Parareal coarse correction with a simplified Newton iteration whose initial guess is $(Z + b_n)$ in every time window. That choice freezes the Jacobian at a constant block-cyclic matrix $G_d$ with equal diagonal blocks, so the DFT converts the correction step into $N$ independent $d\times d$ solves, one per Fourier frequency. The paper argues that for nonlinearities with strong monotonicity and a Lipschitz derivative condition, the affine-covariant Lipschitz constant $\delta_0 = L_2/c_1$ makes the simplified Newton iteration convergent whenever the initial guess satisfies $h_0 \le 1/2$; the numerical study confirms this for a scalar model. For the nonlinear coaxial cable and a three-dimensional transformer, the same algorithm reduces wall-clock time by factors up to 271.5 and 176 relative to sequential time stepping, and in the nonlinear cable it beats the fine-grid multi-harmonic solver for all tested core counts above five.

Load-bearing premise

The method rests on choosing a constant vector $Z$ so that the simplified Newton initial guess $U^{(0)} = Z + b_n$ satisfies $h_0 = \delta_0 \|U^{(1)} - U^{(0)}\| \le 1/2$; the paper proves the ingredients of $\delta_0$ but verifies this closeness condition numerically only for a one-dimensional model problem, using $Z=0$ without a proof in the eddy-current and transformer experiments.

Editorial extensions

If this is right

The coarse-level bottleneck of periodic Parareal disappears: the block-cyclic correction system becomes $N$ independent frequency-domain solves, one per harmonic.
For linear time-periodic problems, frequency-domain treatment of the fine system scales near optimally, reducing effective linear solves by a factor of about 2100 at $N=50$ in the cable test.
For the nonlinear cable, the new PP-PC MH solver cuts wall-clock time by factors between 15 (at $N=5$) and 271.5 (at $N=50$) versus sequential time stepping.
In the three-dimensional transformer model, the same solver gives about a 176-fold wall-clock speedup at $N=20$ cores.
Increasing the number of subintervals improves the coarse solver, so fewer Parareal iterations are needed and the fine solve remains the dominant cost.

Reading between the lines

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

The same trick of freezing the Jacobian with a specially chosen starting point could be applied to other diagonalization-based time-parallel methods for nonlinear periodic problems; any nonlinearity satisfying strong monotonicity is a candidate.
A constructive rule for choosing $Z$ so that $h_0 \le 1/2$ holds a priori would remove the main practical limitation; the paper only verifies this numerically for one scalar problem and uses $Z=0$ in the engineering examples.
The reported speedups are tied to the specific benchmarks and core counts; the DFT overhead in PP-PC MH grows with $N$, so at very large core counts the sequential fine solves will dominate and the speedup will plateau.
The convergence bound $\delta_0 = L_2/c_1$ suggests that materials with steep, strongly monotone magnetization curves are the natural stress test for the method.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Desk Editor's Note

read the letter

Let me give you a direct take. The paper's core trick is real: in periodic Parareal with a periodic coarse problem (PP-PC), if you start the simplified Newton iteration at U^(0)=Z+b^(k) — a constant plus the fine-coarse discrepancy — the Jacobian becomes block-cyclic with identical diagonal blocks Qd(Z), so you can diagonalize the whole coarse correction with an FFT and solve for each frequency independently. That is new relative to the prior linear multi-harmonic work and to the fixed-point MH iteration, and it does what it claims: the 2D nonlinear cable benchmark shows substantial speedups over sequential time stepping, and the method beats the Jacobi-type PP-PC iteration.

The 1D convergence analysis is competent: the Lipschitz bound for the nonlinearity, the constant δ0=L2/c1, and the verification of h0≤0.5 for the scalar example are all sound. The paper also honestly flags where the theory stops.

The soft spot is exactly where the theory stops. The theorem guarantees convergence only if the initial guess satisfies (5.25), and for the PP-PC version that initial guess includes the fine-coarse correction b^(k). The paper verifies h0 only for the scalar TP-MH case where b^(k) is absent; there is no construction of Z that makes h0≤1/2 for the PP-PC or vector problems. The 2D and 3D runs use Z=0 and convergence is observed, not guaranteed. That is a real limitation, but the paper says so after Theorem 5.6.

One more thing the stress-test note gets right: the 3D transformer experiment does not use the simplified Newton Jacobian Qd(Z). It uses the more general constant-matrix fixed-point iteration (5.15) with an H built from a fixed reluctivity. So the 176x speedup is evidence for the constant-matrix MH framework, not for the specific simplified Newton algorithm analyzed in Section 5. That is worth fixing in revision—either rerun 3D with Qd(Z) or clearly present (7.4) as a separate variant.

Minor: no code or data are provided, so the effective-solve and wall-clock numbers are hard to reproduce independently. Not fatal, but a good manuscript would include them.

Overall: this is a solid, honest paper with a genuine algorithmic contribution and real experimental support, but the convergence guarantee is narrower than the experiments suggest. It deserves peer review; I would accept it conditionally, asking the authors to either close the h0 gap for the PP-PC initial guess or explicitly scope the claim, and to clarify the 3D linearization. I would cite it if I worked on parallel-in-time methods, and it is worth a reading group discussion.

Referee Report

3 major / 4 minor

Summary. This paper introduces a parallel-in-time algorithm for nonlinear time-periodic problems. It combines the periodic Parareal method with a simplified Newton iteration whose initial guess is chosen as U^{(k+1,0)} = (Z + b_n), making the Jacobian block-cyclic with constant diagonal block Q_d(Z). The system is diagonalized via DFT, allowing independent frequency solves. Convergence is analyzed for a scalar model problem and verified numerically. Experiments on 2D and 3D eddy current problems show speedups up to 271x and 176x over sequential time stepping.

Significance. The main contributions are the constant-Jacobian construction (Eqs. 5.8-5.10) and its FFT diagonalization (Eqs. 5.12-5.13). The derivation is correct and the method has potentially wide applicability in electrical engineering. The numerical experiments are extensive and include both effective linear-system counts and wall-clock times. However, the convergence theory is limited to a scalar model and the 3D demonstration uses a different fixed-point linearization. Therefore the significance is moderate and conditional on closing the theory-experiment gap.

major comments (3)

[§5.4.1–5.4.2] The convergence guarantee for the PP-PC variant is not established. Theorem 5.6 provides only the Lipschitz constant delta_0 = L_2/c_1, but Theorem 5.3 also requires h_0 = delta_0 ||U^(1)-U^(0)|| <= 1/2 from Eq. (5.25). The numerical verification in §5.4.2 concerns the TP-MH setting with initial guess u^(0) = [z,...,z] and contains no b^(k) terms; it does not verify h_0 for the PP-PC initial guess (5.32), where U^(1)-U^(0) depends on the fine-coarse discrepancy b^(k). In the 2D nonlinear experiment, Eq. (6.8) sets Z=0 and no h_0 check is reported. Since no construction of Z or bound on ||U^(1)-U^(0)|| is given, Theorem 5.3 does not guarantee convergence of Algorithm 5.1 for the main nonlinear examples.
[Remark 5.7 and §6.3] The analysis is explicitly limited to a scalar model problem, and the extension to d-dimensional systems is only a pointer to operator theory in Remark 5.7. The 2D eddy current example uses the same algorithm with Z=0, so in the absence of a d-dimensional analogue of Theorem 5.6, or at least a numerical check of h_0, convergence of the vector PP-PC MH method is an observed property rather than a proven one. The sentence after Theorem 5.6 acknowledges that a close initial guess of the required form may not exist, but this is exactly the load-bearing condition on which the convergence theorem rests.
[§7, Eq. (7.4)] The 3D transformer experiment does not use the simplified Newton method (5.5)-(5.9) with Jacobian Q_d(Z) from Eq. (5.13) that is the subject of the convergence analysis. Instead it uses the fixed-point iteration (5.15) with the constant matrix H built from a fixed reluctance Kbar = nu_0 * 10^{-3} in Eq. (7.4). Therefore the 176x speedup and the 3D demonstration support a different constant-matrix iteration, not the central simplified Newton claim. The paper should either analyze the fixed-point iteration used in Section 7 or explicitly restrict the 3D claims and reframe them as evidence for the broader constant-matrix correction idea.

minor comments (4)

[§2] In the paragraph after Eq. (2.9), the sentence 'Another advantage ... is that the is fully non-intrusive' contains a typo; it should read 'it is fully non-intrusive.'
[§5.3] After Eq. (5.20), 'd-dimensinal vector space' should be 'd-dimensional vector space.'
[Fig. 4] The caption of Figure 4 contains a stray '3' after 'cross-section of a coaxial cable'; it should be removed.
[Eq. (4.3)] The displayed block-cyclic matrix in Eq. (4.3) omits the top-right -C block; showing the full cyclic shift would make the transition to the eigenvalue formula (4.7) easier to follow.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the nonlinear constant-Jacobian construction is derived by explicit algebraic identities, and the convergence constants are obtained from the stated nonlinearity assumptions, not fitted or self-citational.

full rationale

The central claim is a mathematical construction rather than a fitted prediction: choosing the simplified Newton initial guess as U^(k+1,0) = (Z + b^(k)_n) in (5.8) makes the Jacobian at that point equal to the constant block-cyclic matrix Gd in (5.10), and the DFT diagonalization in (5.12)-(5.13) then follows by the explicit formula (4.7). This is an exact derivation, not an input disguised as an output. The convergence analysis also avoids circularity: Theorem 5.3 is taken from external literature (Deuflhard / Ortega-Rheinboldt), and Theorem 5.6 derives the Lipschitz constant δ0 = L2/c1 from Assumption 5.4 on the model nonlinearity. The numerical values L2 = 0.6 and c1 = 1 in Section 5.4.2 are read off from the chosen piecewise polynomial κ, not calibrated to force convergence. The 2D and 3D experiments are genuine simulations compared against sequential time stepping and existing PP-IC/PP-PC methods, and the reported speedups are measured wall-clock results, not reproduced targets. Self-citations to the authors' prior work (e.g., [20] for the linear PP-PC MH solver and [3], [29] for earlier applications) provide background and the linear-case building block, but the nonlinear constant-Jacobian construction is derived in Sections 5.1-5.2 from explicit equations. The acknowledged difficulty around the initial-guess condition h0 ≤ 1/2 and the deferral of the vector-valued case in Remark 5.7 are legitimate limitations of the proof, not circular reasoning. The paper therefore exhibits no significant circularity.

Assumptions & free parameters 2 free parameters · 4 assumptions · 0 invented entities

The central claim relies on standard numerical analysis (circulant diagonalization, simplified Newton theory) and physically motivated monotonicity assumptions. No new physical entities or fitted constants are introduced; the only hand-picked quantities are the initial guess Z and the 3D linearization constant K-bar, which affect convergence speed but not the formal structure.

free parameters (2)

Z (initial guess constant in Eq. 5.8) = 0 for the 2D and 3D experiments; average of previous iterate also tested
Chosen by hand; the convergence condition h0 ≤ 1/2 (Eq. 5.25) depends on it. The paper notes the choice is problem-specific and refers to [8] for guidance.
K-bar (constant matrix in 3D fixed-point iteration, Eq. 7.4) = ν-bar = ν0 * 10^-3 H/m
Ad hoc choice used to define the linearization matrix H for the 3D transformer; no systematic selection rule is provided.

assumptions (4)

domain assumption Nonlinearity κ (or ν) satisfies Assumption 5.4 / 6.1: strong monotonicity, lower bound c1, C^1 with vanishing derivative, Lipschitz g(t)=κ'(t)t.
Used in Lemma 5.5 and Theorem 5.6 to obtain the Lipschitz constant δ0 = L2/c1; physically motivated by magnetization curves from [18].
domain assumption The matrix pencil (M, K(v)) is regular, and the spatially discretized eddy-current problem is an index-1 DAE.
Stated in Section 2 and used to justify implicit Euler and consistent initial values.
standard math The block-circulant matrix Gd is diagonalized by the DFT matrix F.
This is the classical diagonalization of a circulant matrix; the spectrum is given in Eq. (5.13).
standard math The simplified Newton convergence theorem (Deuflhard Theorem 5.3) applies to the model problem.
External theorem; the affine-covariant Lipschitz condition is proven in Theorem 5.6, while the h0 condition is verified numerically for the 1D model.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Efficient Parallel-in-Time Solution of Time-Periodic Problems Using a Multi-Harmonic Coarse Grid Correction." pith.science (2026). https://pith.science/paper/RRDM355M

@misc{pith2026190805245,
  author       = {Pith},
  title        = {Pith review of: Efficient Parallel-in-Time Solution of Time-Periodic Problems Using a Multi-Harmonic Coarse Grid Correction},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/RRDM355M}},
  note         = {Machine review of arXiv:1908.05245}
}

read the original abstract

This paper presents a highly-parallelizable parallel-in-time algorithm for efficient solution of nonlinear time-periodic problems. It is based on the time-periodic extension of the Parareal method, known to accelerate sequential computations via parallelization on the fine grid. The proposed approach reduces the complexity of the periodic Parareal solution by introducing a simplified Newton algorithm, which allows an additional parallelization on the coarse grid. In particular, at each Newton iteration a multi-harmonic correction is performed, which converts the block-cyclic periodic system in the time domain into a block-diagonal system in the frequency domain, thereby solving for each frequency component in parallel. The convergence analysis of the method is discussed for a one-dimensional model problem. The introduced algorithm and several existing solution approaches are compared via their application to the eddy current problem for both linear and nonlinear models of a coaxial cable. Performance of the considered methods is also illustrated for a three-dimensional transformer model.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

34 extracted references · 28 canonical work pages

[1]

Außerhofer, O

S. Außerhofer, O. B ´ır´o, and K. Preis , An eﬃcient harmonic balance method for nonlinear eddy-current problems, IEEE Trans. Magn., 43 (2007), pp. 1229–1232, https://doi.org/10.1109/TMAG.2006.890961

work page arXiv 2007
[2]

Bachinger, U

F. Bachinger, U. Langer, and J. Sch ¨oberl, Numerical analysis of nonlinear multiharmonic eddy current problems, Numer. Math., 100 (2005), pp. 593–616, https://doi.org/10.1007/s00211-005-0597-2

doi:10.1007/s00211-005-0597-2 2005
[3]

D. Bast, I. Kulchytska-Ruchka, S. Sch ¨ops, and O. Rain , Accelerated steady-state torque computation for induction machines using parallel-in-time algorithms , IEEE Trans. Magn., 56 (2020), pp. 1–9, https: //doi.org/10.1109/TMAG.2019.2945510

arXiv 2020
[4]

B. K. Bose, Power Electronics And Motor Drives , Academic Press, Burlington, 2006

work page 2006
[5]

B´ır´o and K

O. B´ır´o and K. Preis , An eﬃcient time domain method for nonlinear periodic eddy current problems , IEEE Trans. Magn., 42 (2006), pp. 695–698

work page 2006
[6]

J. R. Brauer, Simple equations for the magnetization and reluctivity curves of steel , IEEE Trans. Magn., 11 (1975), p. 81, https://doi.org/10.1109/TMAG.1975.1058555

work page arXiv 1975
[7]

S. C. Brenner and L. R. Scott , The mathematical theory of ﬁnite element methods , vol. 15 of Texts in applied mathematics, Springer, New York, 3. ed. ed., 2008

work page 2008
[8]

Deuflhard, Newton methods for nonlinear problems: aﬃne invariance and adaptive algorithms , Springer, Berlin, 2004

P. Deuflhard, Newton methods for nonlinear problems: aﬃne invariance and adaptive algorithms , Springer, Berlin, 2004

work page 2004

Show all 34 references

[9]

M. J. Gander and E. Hairer , Nonlinear Convergence Analysis for the Parareal Algorithm , Springer Berlin Heidelberg, Berlin, Heidelberg, 2008, pp. 45–56, https://doi.org/10.1007/978-3-540-75199-1 4

2008 doi
[10]

M. J. Gander and L. Halpern , Time Parallelization for Nonlinear Problems Based on Diagonalization , Springer International Publishing, Cham, 2017, pp. 163–170, https://doi.org/10.1007/978-3-319-52389-7 15

2017 doi
[11]

M. J. Gander, L. Halpern, J. Rannou, and J. Ryan, A direct time parallel solver by diagonalization for the wave equation, SIAM J. Sci. Comput., 41 (2019), pp. A220–A245, https://doi.org/10.1137/17M1148347

2019 doi
[12]

M. J. Gander, Y.-L. Jiang, B. Song, and H. Zhang , Analysis of two parareal algorithms for time-periodic problems, SIAM J. Sci. Comput., 35 (2013), pp. A2393–A2415, https://doi.org/10.1137/130909172

2013 doi
[13]

M. J. Gander, I. Kulchytska-Ruchka, I. Niyonzima, and S. Sch¨ops, A new parareal algorithm for problems with discontinuous sources , SIAM J. Sci. Comput., 41 (2019), pp. B375–B395, https://doi.org/10.1137/ 18M1175653

2019
[14]

M. J. Gander, J. Liu, S.-L. Wu, X. Yue, and T. Zhou , ParaDiag: Parallel-in-Time Algorithms Based on the Diagonalization Technique , 2020, https://arxiv.org/abs/2005.09158. ArXiv: 2005.09158

2020 arXiv
[15]

M. J. Gander and S. Vandewalle , On the superlinear and linear convergence of the parareal algorithm , in 28 KULCHYTSKA-RUCHKA AND SCH ¨OPS Domain decomposition methods in science and engineering XVI, vol. 55 of Lecture Notes in Computational Science and Engineering, Springer,...

2007
[16]

Gyselinck, P

J. Gyselinck, P. Dular, C. Geuzaine, and W. Legros , Harmonic-balance ﬁnite-element modeling of elec- tromagnetic devices: a novel approach , IEEE Trans. Magn., 38 (2002), pp. 521–524, https://doi.org/10. 1109/20.996137

2002
[17]

T. Hara, T. Naito, and J. Umoto , Time-periodic ﬁnite element method for nonlinear diﬀusion equation , IEEE Trans. Magn., 21 (1985), pp. 2261–2264

1985
[18]

Heise, Analysis of a fully discrete ﬁnite element method for a nonlinear magnetic ﬁeld problem , SIAM J

B. Heise, Analysis of a fully discrete ﬁnite element method for a nonlinear magnetic ﬁeld problem , SIAM J. Numer. Anal., 31 (1994), pp. 745–759, http://www.jstor.org/stable/2158028

1994
[19]

J. D. Jackson , Classical Electrodynamics, Wiley & Sons, New York, 3rd ed., 1998, https://doi.org/10.1017/ CBO9780511760396

1998
[20]

Kulchytska-Ruchka, H

I. Kulchytska-Ruchka, H. De Gersem, and S. Sch ¨ops, An eﬃcient steady-state analysis of the eddy cur- rent problem using a parallel-in-time algorithm , in The Tenth International Conference on Computational Electromagnetics (CEM 2019), Edinburgh, UK, June 2019, https://doi.or...

2019
[21]

Lamour, R

R. Lamour, R. M ¨arz, and C. Tischendorf , Diﬀerential-Algebraic Equations: A Projector Based Anal- ysis, Diﬀerential-Algebraic Equations Forum, Springer, Heidelberg, 2013, https://doi.org/10.1007/ 978-3-642-27555-5

2013
[22]

Lions, Y

J.-L. Lions, Y. Maday, and G. Turinici , A parareal in time discretization of PDEs , Comptes Rendus de l’Acad´ emie des Sciences – Series I – Mathematics, 332 (2001), pp. 661–668, https://doi.org/10.1016/ S0764-4442(00)01793-6

2001
[23]

Mohan, T

N. Mohan, T. M. Undeland, and W. P. Robbins , Power electronics: converters, applications and design , Wiley, 3 ed., 2003

2003
[24]

Nakhla and J

M. Nakhla and J. Vlach , A piecewise harmonic balance technique for determination of periodic response of nonlinear systems , IEEE Trans. Circ. Syst., 23 (1976), pp. 85–91, https://doi.org/10.1109/TCS.1976. 1084181

1976 doi
[25]

J. M. Ortega and W. C. Rheinboldt, Iterative Solution of Nonlinear Equations in Several Variables, Society for Industrial and Applied Mathematics, Philadelphia, PA, USA, 2 ed., 2000, https://doi.org/10.1137/1. 9780898719468

doi
[26]

C. Pechstein, Multigrid-Newton-methods for nonlinear-magnetostatic problems , master’s thesis, Universit¨ at Linz, Linz, Austria, 2004, https://www.numa.uni-linz.ac.at/Teaching/Diplom/Finished/pechstein

2004
[27]

Pechstein and B

C. Pechstein and B. J ¨uttler, Monotonicity-preserving interproximation of b-h-curves , J. Comput. Appl. Math., 196 (2006), pp. 45–57, https://doi.org/10.1016/j.cam.2005.08.021

2006 doi
[28]

Sch ¨ops, H

S. Sch ¨ops, H. De Gersem, and T. Weiland , Winding functions in transient magnetoquasistatic ﬁeld-circuit coupled simulations , COMPEL, 32 (2013), pp. 2063–2083, https://doi.org/10.1108/ COMPEL-01-2013-0004

2013
[29]

Sch ¨ops, I

S. Sch ¨ops, I. Niyonzima, and M. Clemens , Parallel-in-time simulation of eddy current problems using parareal, IEEE Trans. Magn., 54 (2018), pp. 1–4, https://doi.org/10.1109/TMAG.2017.2763090

2018
[30]

Stoer and R

J. Stoer and R. Bulirsch , Numerische Mathematik 2 , Springer, Berlin, 3 ed., 2005

2005
[31]

L. N. Trefethen , Finite Diﬀerence and Spectral Methods for Ordinary and Partial Diﬀerential Equations , Cornell University, 1996

1996
[32]

Weiland , Time domain electromagnetic ﬁeld computation with ﬁnite diﬀerence methods , Int

T. Weiland , Time domain electromagnetic ﬁeld computation with ﬁnite diﬀerence methods , Int. J. Nu- mer. Model. Electron. Network. Dev. Field, 9 (1996), pp. 295–319, https://doi.org/10.1002/(SICI) 1099-1204(199607)9:4⟨295::AID-JNM240⟩3.0.CO;2-8

1996 doi
[33]

S.-L. Wu, H. Zhang, and T. Zhou , Solving time-periodic fractional diﬀusion equations via diagonalization technique and multigrid , Numerical Linear Algebra with Applications, 25 (2018), p. e2178, https://doi. org/10.1002/nla.2178

2018 doi
[34]

Zeidler, Nonlinear Functional Analysis and its Applications II/A

E. Zeidler, Nonlinear Functional Analysis and its Applications II/A. Linear Monotone Operators , Springer- Verlag, New York Inc., 1990

1990

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.

[1] [1]

Außerhofer, O

S. Außerhofer, O. B ´ır´o, and K. Preis , An eﬃcient harmonic balance method for nonlinear eddy-current problems, IEEE Trans. Magn., 43 (2007), pp. 1229–1232, https://doi.org/10.1109/TMAG.2006.890961

work page arXiv 2007

[2] [2]

Bachinger, U

F. Bachinger, U. Langer, and J. Sch ¨oberl, Numerical analysis of nonlinear multiharmonic eddy current problems, Numer. Math., 100 (2005), pp. 593–616, https://doi.org/10.1007/s00211-005-0597-2

doi:10.1007/s00211-005-0597-2 2005

[3] [3]

D. Bast, I. Kulchytska-Ruchka, S. Sch ¨ops, and O. Rain , Accelerated steady-state torque computation for induction machines using parallel-in-time algorithms , IEEE Trans. Magn., 56 (2020), pp. 1–9, https: //doi.org/10.1109/TMAG.2019.2945510

arXiv 2020

[4] [4]

B. K. Bose, Power Electronics And Motor Drives , Academic Press, Burlington, 2006

work page 2006

[5] [5]

B´ır´o and K

O. B´ır´o and K. Preis , An eﬃcient time domain method for nonlinear periodic eddy current problems , IEEE Trans. Magn., 42 (2006), pp. 695–698

work page 2006

[6] [6]

J. R. Brauer, Simple equations for the magnetization and reluctivity curves of steel , IEEE Trans. Magn., 11 (1975), p. 81, https://doi.org/10.1109/TMAG.1975.1058555

work page arXiv 1975

[7] [7]

S. C. Brenner and L. R. Scott , The mathematical theory of ﬁnite element methods , vol. 15 of Texts in applied mathematics, Springer, New York, 3. ed. ed., 2008

work page 2008

[8] [8]

Deuflhard, Newton methods for nonlinear problems: aﬃne invariance and adaptive algorithms , Springer, Berlin, 2004

P. Deuflhard, Newton methods for nonlinear problems: aﬃne invariance and adaptive algorithms , Springer, Berlin, 2004

work page 2004