Linear Kelvin Wave Predictions in the z to 0 Limit

arXiv:2603.14945 · physics.flu-dyn · Submitted 2026-03-16 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: Physics. The day in general physics: instruments, methods and results across the field.

Kai: We're looking at the paper: "Linear Kelvin Wave Predictions in the z to 0 Limit".

Mira: Here's the gist of the paper. Linear wave theory captures the essential physics of free-surface flows at a fraction of the computational cost of nonlinear and viscous methods, making it attractive for design, realtime control, and surrogate modeling applications<ref:2603.14945#pg2>.

Kai: This paper develops a modified kernel for the Kelvin potential incorporating an elliptic spanwise line integration that naturally resolves this ill-posedness, yielding finite wave energy over the entire free surface<ref:2603.14945#pg2>.

Mira: The Problem with Traditional Methods Linear wave theory is attractive because it captures essential physics at a fraction of the computational cost of nonlinear and viscous flow models<ref:2603.14945#pg2>.

Kai: However, the Kelvin Green’s function for a translating point-source generates unbounded wave energy in the z → 0 limit, causing both numerical difficulties and physical inconsistencies<ref:2603.14945#pg2>.

Paper summary: Kai: So, to recap what we just covered, this paper looks at linear wave theory for free-surface flows and shows how they fixed the problem where point sources cause infinite energy near the surface by using an elliptic line integration <ref:2603.14945#pg2>.

Mira: They then show a specific source distribution, q of y sub p equals q zero divided by the square root of one minus y sub p squared over b squared for rectangular planforms, which regularizes this divergence and gives finite energy <ref:2603.14945#pg6>.

Lev: It’s interesting that they used a trial strength distribution based on the ship's geometry to fix the math, rather than just slapping some damping term in the boundary condition <ref:2603.14945#pg3>.

Kai: Right. And they followed that up with a fast numerical evaluator using contour deformation, getting speedups of one to fifty thousand over direct quadrature <ref:2603.14945#pg1>.

Mira: So the core claim is that this modified kernel gives well-behaved predictions right on the surface at depth zero, capturing Froude number trends and wave interference patterns that match what you expect for finitewidth ships <ref:2603.14945#pg10>.

Lev: What I think is important here is that they are showing that even though the point source singularity is known from Ursell in one thousand nine hundred sixty this specific approach provides a consistent spectral decay without needing arbitrary wavenumber filtering <ref:2603.14945#pg8>.

Kai: They also highlight that the destructive spanwise interference leads to a scaling of Aw equals b minus three for large b, which means the resistance coefficient Cw gradually decreases as the ship length increases <ref:2603.14945#pg11>.

Mira: It's a solid formulation because they avoided introducing non-physical Rayleigh damping parameters that complicate things further <ref:2603.14945#pg3>.

Lev: What I wonder is how this approach holds up when you start talking about more complex, non-linear effects, since this whole thing is built on the linear wave theory assumption <ref:2603.14945#pg2>.

Conclusion: Kai: So wrapping up the paper "Linear Kelvin Wave Predictions in the z to zero Limit," it seems they’ve successfully shown that you can handle these singular problems in linear wave theory using an elliptic source distribution <ref:2603.14945#pg8>.

Mira: They achieved a finite energy formulation by avoiding empirical damping, arbitrary filtering, and implicit mathematical formulations <ref:2603.14945#pg8>.

Lev: The speedup for the contour-deformation method is what makes it practical for real hardware applications, showing a ten thousand to fifty thousand times faster evaluation than direct quadrature <ref:2603.14945#pg1>.

Kai: And the implications are that you can get robust predictions on the free surface directly at depth zero, which is really useful for things like ship design and real-time control <ref:2603.14945#pg1>.

Mira: It means they have a consistent way to predict wave behavior based on linear theory, even when there are tricky mathematical issues with the point source singularity <ref:2603.14945#pg8>.

Lev: It’s about showing that the divergence isn't an intrinsic flaw in the theory itself, but how we set up our problem <ref:2603.14945#pg8>.

Kai: The paper demonstrates a new way to handle these problems without resorting to introducing non-physical damping parameters or arbitrary wavenumber filtering <ref:2603.14945#pg8>.

Gabriel D. Weymouth

Delft University of Technology

physics.flu-dyn

Submitted: 2026-03-16

Updated: 2026-11-04

License: http://creativecommons.org/licenses/by/4.0/

Importance score: 92/100

The gist: Linear wave theory captures the essential physics of free-surface flows at a fraction of the computational cost of nonlinear and viscous methods, making it attractive for design…

Key concepts

Kelvin Potential
This is the mathematical function used to describe the wave disturbance generated by a point source (like a ship) on a free surface. The traditional formulation of this potential causes problems near the surface, leading to infinite energy when analyzed mathematically.
Elliptic Spanwise Line Integration
This is a new mathematical technique introduced to modify the standard kernel. It involves integrating along an elliptic line in the spanwise direction. This modification effectively regularizes the ill-posed nature of the problem, ensuring that wave energy remains finite even as you approach the free surface ($z o 0$).

Terminology

Summary

Linear wave theory captures the essential physics of free-surface flows at a fraction of the computational cost of nonlinear and viscous methods, making it attractive for design, realtime control, and surrogate modeling applications<ref:2603. p. 0

Where the paper says it

Weymouth@tudelft. nl (Received xx; revised xx; accepted xx) Linear wave theory captures the essential physics of free-surface flows at a fraction of the computational cost of nonlinear and viscous methods, making it attractive for design, realtime control, and surrogate modeling applications.

14945#pg2>. This paper develops a modified kernel for the Kelvin potential incorporating an elliptic spanwise line integration that naturally resolves this ill-posedness, yielding finite wave energy over the entire free surface<ref:2603. p. 0
Where the paper says it

However, the Kelvin Green’s function for a translating point-source generates unbounded wave energy in the 𝑧 → 0 limit, causing both numerical difficulties and physical inconsistencies. This paper develops a modified kernel for the Kelvin potential incorporating an elliptic spanwise line integration that naturally resolves this ill-posedness, yielding finite wave energy over the entire free surface.

14945#pg2>.

The Problem with Traditional Methods

Linear wave theory is attractive because it captures essential physics at a fraction of the computational cost of nonlinear and viscous flow models<ref:2603.14945#pg2>. However, the Kelvin Green’s function for a translating point-source generates unbounded wave energy in the z → 0 limit, causing both numerical difficulties and physical inconsistencies<ref:2603. p. 0

Where the paper says it

nl (Received xx; revised xx; accepted xx) Linear wave theory captures the essential physics of free-surface flows at a fraction of the computational cost of nonlinear and viscous methods, making it attractive for design, realtime control, and surrogate modeling applications. However, the Kelvin Green’s function for a translating point-source generates unbounded wave energy in the 𝑧 → 0 limit, causing both numerical difficulties and physical inconsistencies.

14945#pg2>. This divergence is characterized by the wave elevation spectrum Sζ (k) peaking at k∗ ∼ 1/z with amplitude ∼ 1/z, diverging as z → 0 and making the downstream wake unresolvable<ref:2603. p. 0
Where the paper says it

The Kelvin Green’s function for a steady translating disturbance exactly satisfies the linear free-surface boundary condition and has a well-established integral representation (Peters 1949), but the potential is ill-posed as the source position approaches the freesurface. This singularity was first identified and analyzed by Ursell (1960), and as shown in Section 3, the wave elevation spectrum 𝑆𝜁 (𝑘) peaks at 𝑘 ∗ ∼ 1/|𝑧| with amplitude ∼ 1/|𝑧|, diverging as 𝑧 → 0 and making the downstream wake unresolvable.

14945#pg4>. Classic methods to avoid this singularity involve introducing a damping term in the free-surface boundary condition, but this complicates the Green’s function evaluation and introduces a non-physical Rayleigh damping parameter<ref:2603. p. 1
Where the paper says it

Weymouth (Havelock 1932; F¨urth et al. 2021), but this complicates the Green’s function evaluation and introduces a non-physical Rayleigh damping parameter.

14945#pg3>.

The Regularization Strategy

The authors establish a well-posed, finite-energy formulation of the linear free-surface problem in the z → 0 limit while retaining the explicit Green’s function framework<ref:2603.14945#pg3>. The first contribution is analytically developing the elliptic spanwise line kernel which naturally regularizes this divergence in Section 4<ref:2603. p. 1

Where the paper says it

This paper addresses this gap by establishing a well-posed, finite-energy formulation of the linear free-surface problem in the 𝑧 → 0 limit while retaining the explicit Green’s function framework. The first contribution of this paper, after introducing the Kelvin Green’s function and revisiting the divergence in the point-source spectrum in Section 3, is analytically developing the elliptic spanwise line kernel which naturally regularizes this divergence in Section 4.

14945#pg3>. For a rectangular planform, the potential on the contour reduces to an integral involving a physically motivated trial strength distribution q(yp)<ref:2603. p. 4
Where the paper says it

In contrast, the present work analytically evaluates the integral over a physically motivated trial strength distribution 𝑞, regularizing the integral and enabling universal finite energy predictions with an explicit Green’s function. For illustration, we focus on a rectangular planform of half-beam 𝑏 and length 𝐿.

14945#pg6>. This leads to the elliptic distribution q(yp) = q0 / √[1 - (y p/b)2]<ref:2603. p. 4
Where the paper says it

1), and positing a solution where the high wavenumber components of 𝑊 are controlled, the dominant operator acting with 𝑞 to enforce the velocity condition is the local logarithmic term: ∫ 𝑏 −𝑏 𝑞(𝑦 𝑝) log |𝑦 − 𝑦 𝑝 | d𝑦 𝑝 = const, |𝑦| < 𝑏. This is precisely the classical constant-downwash integral equation of finite-wing theory, whose solution on [−𝑏, 𝑏] is the elliptic distribution 𝑞(𝑦 𝑝) = 𝑞0 √︃ 1 − (𝑦 𝑝/𝑏) 2 (4.

14945#pg6>.

Numerical Evaluation and Speedup

The second contribution is the development of a fast and robust numerical evaluator for both kernels in Section 5 with an open-source Julia implementation<ref:2603.14945#pg3>. The partitioned contour-deformation approach is used to evaluate the wavelike kernels efficiently, achieving 104–105 speedup over direct quadrature while preserving the wake asymptotics<ref:2603. p. 0

Where the paper says it

This paper develops a modified kernel for the Kelvin potential incorporating an elliptic spanwise line integration that naturally resolves this ill-posedness, yielding finite wave energy over the entire free surface. We then present a fast evaluator for both point and line kernels using contour deformation adapted to the non-analytic Kelvin phase, achieving 104 -105 speedup over direct quadrature while preserving the wake asymptotics.

14945#pg9>. This method partitions the integral into non-oscillatory intervals and semi-infinite tails, using complex contour quadratures for the tails<ref:2603. p. 6
Where the paper says it

Partition boundaries and quadrature methods As in Gibbs et al. (2024), the wavelike integral is partitioned into two types of regions; (i) non-oscillatory intervals around stationary points, where standard quadrature is efficient, and (ii) the remaining semi-infinite tails, where complex contour quadratures converge exponentially.

14945#pg9>.

Results and Physical Insights

The resulting formulation produces well-behaved free-surface predictions directly on z = 0, capturing Froude number trends and wave interference patterns consistent with finitewidth ships<ref:2603.14945#pg10>. The theory predicts a strong decrease in the diverging wave elevation amplitude with increasing b for a given integrated line-source strength π/2 q0b<ref:2603. p. 8

Where the paper says it

Directly behind the planform the interaction of the stern corner waves generates a large rooster-tail in the near wake, leaving a rough but low amplitude wave-field in the far wake after the wave trains diverge. The theory predicts a strong decrease in the diverging wave elevation amplitude with increasing 𝑏 for a given integrated line-source strength 𝜋 2 𝑞0𝑏.

14945#pg11>. Furthermore, the destructive spanwise interference discussed above results in a scaling Aw ∼ b − 3 for large b, causing a gradual but monotonic decrease in the resistance coefficient Cw with increasing b<ref:2603. p. 9
Where the paper says it

The interference factor 1−cos(𝐿𝑘 𝑥) produces classical constructive and destructive interference between the bow and stern wave systems, resulting in oscillating resistance values with 𝐿. More uniquely, the destructive spanwise interference discussed above results in a scaling 𝐴𝑤 ∼ 𝑏 −3 for large 𝑏, causing a gradual but monotonic decrease in the resistance coefficient 𝐶𝑊 with increasing 𝑏.

14945#pg11>.

Conclusions

The singular behavior of the point-source Kelvin Green’s function as z → 0 is shown not to be intrinsic to linear wave theory. The elliptic source distribution yields a regularized kernel with finite energy and consistent spectral decay, without the introduction of empirical damping, arbitrary wavenumber filtering, or implicit formulations<ref:2603. p. 9

Where the paper says it

Conclusions The singular behavior of the point-source Kelvin Green’s function as 𝑧 → 0 is a fundamental inconsistency which produces unbounded wave energy even far downstream, but the present work shows that this pathology is not intrinsic to linear wave theory. An elliptic source distribution, motivated by the finite disturbance width, yields a regularized kernel with finite energy and consistent spectral decay, without the introduction of empirical damping, arbitrary wavenumber filtering, or implicit formulations.

14945#pg8>. The contour-deformation approach developed here is completely general, evaluating both point and line kernels efficiently (achieving 104–105 speedup over direct quadrature) while maintaining accuracy across the entire lower-half domain. p. 9
Where the paper says it

Accounting for local hull geometry while retaining the regularized kernel framework is a natural extension, at which point the predictions could be compared to experiments or high-fidelity nonlinear simulations. However, the contour-deformation approach developed here is completely general, evaluating both point and line kernels efficiently (achieving 104–105 speedup over direct quadrature) while maintaining accuracy across the entire lower-half domain.

References

Baar, J. J. M. & Price, W. p. 1

Where the paper says it

Even for a submerged source, efficient and robust evaluation of the Green’s function is challenging. Asymptotic series expansions have been developed (Baar & Price 1988b; Clarisse & Newman 1994), but require a patchwork of different series in different regions with no universally optimal series truncation criterion.

p. 1
Where the paper says it

Even for a submerged source, efficient and robust evaluation of the Green’s function is challenging. Asymptotic series expansions have been developed (Baar & Price 1988b; Clarisse & Newman 1994), but require a patchwork of different series in different regions with no universally optimal series truncation criterion.

G. 1988a Developments in the calculation of the wavemaking resistance of ships. p. 9
Where the paper says it

An elliptic source distribution, motivated by the finite disturbance width, yields a regularized kernel with finite energy and consistent spectral decay, without the introduction of empirical damping, arbitrary wavenumber filtering, or implicit formulations. The resulting formulation produces well-behaved free-surface predictions directly on 𝑧 = 0, capturing Froude number trends and wave interference patterns consistent with finitewidth ships.

Baar, J. J. M. & Price, W. G. 1988b Evaluation of the Wavelike Disturbance in the Kelvin Wave Source Potential<ref:2603. p. 1

Where the paper says it

After scaling all lengths by the Kelvin length ℓ = 𝒰2 /ℊ (where ℊ is the gravitational acceleration) and the velocity potential 𝜙 by 𝒰ℓ, the potential must satisfy the linear free-surface boundary condition on the undisturbed free-surface plane 𝜕𝑥 𝑥𝜙 + 𝜕𝑧𝜙 = 0 on 𝑧 = 0 as well as ∇ 2𝜙 = 0 in the lower half-space. The classic Kelvin Green’s function of a translating point source at 𝑥®𝑝 satisfies the boundary condition above by construction and is made up of near-field terms 𝑁 which act close to the disturbance and a wavelike term 𝑊 whose influence extends into the wake 𝐺 𝑥®𝑓 , 𝑥®𝑝 = 𝑁(®𝑥 𝑓 , 𝑥®𝑝) + 𝑊 (®𝑥 𝑓 − ®𝑥 𝑝 ′ ) 0 X0-

14945#pg3>.

Clarisse, J. p. 1

Where the paper says it

Even for a submerged source, efficient and robust evaluation of the Green’s function is challenging. Asymptotic series expansions have been developed (Baar & Price 1988b; Clarisse & Newman 1994), but require a patchwork of different series in different regions with no universally optimal series truncation criterion.

-M. & Newman, J. p. 1
Where the paper says it

Even for a submerged source, efficient and robust evaluation of the Green’s function is challenging. Asymptotic series expansions have been developed (Baar & Price 1988b; Clarisse & Newman 1994), but require a patchwork of different series in different regions with no universally optimal series truncation criterion.

N. 1994 Evaluation of the Wave-Resistance Green Function: Part 3— The Single Integral Near the Singular Axis<ref:2603. p. 1
Where the paper says it

Even for a submerged source, efficient and robust evaluation of the Green’s function is challenging. Asymptotic series expansions have been developed (Baar & Price 1988b; Clarisse & Newman 1994), but require a patchwork of different series in different regions with no universally optimal series truncation criterion.

14945#pg3>.

Cole, S. 1988 A simple example from flat-ship theory<ref:2603. p. 1

Where the paper says it

The waterline contour contributes directly to the potential and must be evaluated at 𝑧 = 0 (Baar & Price 1988a). Classic flat ship theories were developed to predict the pressure on high-speed, shallow-draft vessels (with 𝑧 ≈ 0) but these are high Froude number approximations, not exact solutions, and do not automatically regularize the singular wave energy at the body waterline (Tuck 1975; Cole 1988).

14945#pg3>.

Furth, M, Tan, M, Chen, Z & Arai, M ¨ 2021 A Dissipative Green’s Function Approach to Modeling Gravity Waves behind Submerged Bodies<ref:2603.14945#pg3>.

Gibbs, A., Hewett, D. P. & Huybrechs, D. 2024 Numerical evaluation of oscillatory integrals via automated steepest descent contour deformation<ref:2603. p. 1

Where the paper says it

Asymptotic series expansions have been developed (Baar & Price 1988b; Clarisse & Newman 1994), but require a patchwork of different series in different regions with no universally optimal series truncation criterion. Numerical steepest-descent is an established contour deformation approach for computing oscillatory integrals which has been applied to the Kelvin Green’s function (Iwashita & Ohkusu 1992), but these methods fail to converge near the Kelvin wedge boundary due to stationary point coalescence.

14945#pg9>.

Havelock, T. p. 1

Where the paper says it

Weymouth (Havelock 1932; F¨urth et al. 2021), but this complicates the Green’s function evaluation and introduces a non-physical Rayleigh damping parameter.

H. 1932 The theory of wave resistance. p. 5
Where the paper says it

Note that the wave energy remains finite for any 𝑏 > 0 since the point-source singularity is recovered only if the transition wavenumber 𝑘 𝑦 ∼ 1/𝑏 diverges. Wave Resistance The wave spectrum leads directly to the flat-ship wave resistance 𝒟𝑊 from the classic Havelock (1932) formula.

Iwashita, H. p. 1

Where the paper says it

Asymptotic series expansions have been developed (Baar & Price 1988b; Clarisse & Newman 1994), but require a patchwork of different series in different regions with no universally optimal series truncation criterion. Numerical steepest-descent is an established contour deformation approach for computing oscillatory integrals which has been applied to the Kelvin Green’s function (Iwashita & Ohkusu 1992), but these methods fail to converge near the Kelvin wedge boundary due to stationary point coalescence.

& Ohkusu, M. p. 1
Where the paper says it

Asymptotic series expansions have been developed (Baar & Price 1988b; Clarisse & Newman 1994), but require a patchwork of different series in different regions with no universally optimal series truncation criterion. Numerical steepest-descent is an established contour deformation approach for computing oscillatory integrals which has been applied to the Kelvin Green’s function (Iwashita & Ohkusu 1992), but these methods fail to converge near the Kelvin wedge boundary due to stationary point coalescence.

1992 The Green Function Method for Ship Motion at Forward Speed<ref:2603. p. 0
Where the paper says it

Key words: Kelvin waves, Green’s function, numerical integration 1. Introduction Linear potential theory for steady forward-speed wave problems captures the essential physics of ship-wave interaction at a fraction of the computational cost of nonlinear and viscous flow models.

14945#pg3>.

Newman, J. N. 1987 Evaluation of the Wave-Resistance Green Function: Part 1—The Double Integral<ref:2603. p. 1

Where the paper says it

Finally, Section 6 demonstrates the resulting wave fields and wave resistance of a rectangular flat-ship planform at 𝑧 = 0 across a range of Froude numbers. Kelvin Green’s Function for Linear Potential Flow We consider steady, incompressible and irrotational flow in the lower half-domain 𝑧 ≤ 0 with a reference frame translating in direction 𝑥 at speed 𝒰, Figure 1.

14945#pg3>.

Noblesse, F. p. 1

Where the paper says it

The singular energy is integrable and labeled as “spray drag” by these methods, but the singularity remains unresolved. The persistent difficulty of evaluating the potential at 𝑧 = 0 motivated the development of the Neumann-Michell theory which uses an implicit iterative scheme for the potential to avoid the evaluation entirely (Noblesse et al.

p. 1
Where the paper says it

The singular energy is integrable and labeled as “spray drag” by these methods, but the singularity remains unresolved. The persistent difficulty of evaluating the potential at 𝑧 = 0 motivated the development of the Neumann-Michell theory which uses an implicit iterative scheme for the potential to avoid the evaluation entirely (Noblesse et al.

1983 A Slender-Ship Theory of Wave Resistance<ref:2603. p. 9
Where the paper says it

More uniquely, the destructive spanwise interference discussed above results in a scaling 𝐴𝑤 ∼ 𝑏 −3 for large 𝑏, causing a gradual but monotonic decrease in the resistance coefficient 𝐶𝑊 with increasing 𝑏. This contrasts with classical thin-ship theory, which uses the beam-to-length ratio as a small parameter and predicts that wave resistance increases quadratically with beam, resulting in a constant 𝐶𝑊 regardless of 𝑏.

14945#pg3>.

Noblesse, F., Huang, F. & Yang, C. 2013 The Neumann–Michell theory of ship waves<ref:2603. p. 3

Where the paper says it

Approaches such as Neumann-Michell theory (Noblesse et al. 2013) avoid this difficulty by using integration by parts to reformulate the potential, but still require an explicit truncation of the wavelike spectrum, introducing a non-physical cutoff.

14945#pg8>.

Peters, A. S. 1949 A new treatment of the ship wave problem<ref:2603. p. 9

Where the paper says it

This contrasts with classical thin-ship theory, which uses the beam-to-length ratio as a small parameter and predicts that wave resistance increases quadratically with beam, resulting in a constant 𝐶𝑊 regardless of 𝑏. The new flat-ship theory captures the nonlinear variation in the potential with 𝑏, encoded in the Bessel function of the wavelike kernel, resulting in the drop in 𝐶𝑊 due to destructive diverging wave interference missing in thin-ship theory.

14945#pg3>.

Tuck, E. O. 1975 Low-Aspect-Ratio Flat-Ship Theory<ref:2603. p. 9

Where the paper says it

This contrasts with classical thin-ship theory, which uses the beam-to-length ratio as a small parameter and predicts that wave resistance increases quadratically with beam, resulting in a constant 𝐶𝑊 regardless of 𝑏. The new flat-ship theory captures the nonlinear variation in the potential with 𝑏, encoded in the Bessel function of the wavelike kernel, resulting in the drop in 𝐶𝑊 due to destructive diverging wave interference missing in thin-ship theory.

14945#pg3>.

Ursell, F 1960 On Kelvin’s ship-wave pattern<ref:2603. p. 0

Where the paper says it

The Kelvin Green’s function for a steady translating disturbance exactly satisfies the linear free-surface boundary condition and has a well-established integral representation (Peters 1949), but the potential is ill-posed as the source position approaches the freesurface. This singularity was first identified and analyzed by Ursell (1960), and as shown in Section 3, the wave elevation spectrum 𝑆𝜁 (𝑘) peaks at 𝑘 ∗ ∼ 1/|𝑧| with amplitude ∼ 1/|𝑧|, diverging as 𝑧 → 0 and making the downstream wake unresolvable.

14945#pg4>.

Weymouth, G. p. 0

Where the paper says it

Banner appropriate to article type will appear here in typeset article Linear Kelvin Wave Predictions in the 𝑧 → 0 Limit Gabriel D. Weymouth Ship Hydromechanics, Delft University of Technology, Delft, Netherlands Corresponding author: G.

D. 2026 KelvinFlatShip. p. 0
Where the paper says it

14945v3 [physics. flu-dyn] 30 Jun 2026

Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.

Improvements for AI systems

  1. Regularized Kernel Implementation: The AI system can now generate physically consistent wave predictions by incorporating an elliptic spanwise line integration that naturally resolves this ill-posedness of the Kelvin Green’s function, yielding finite wave energy over the entire free surface.

  2. Fast Evaluator Development: The system can utilize a fast evaluator for both point and line kernels using contour deformation adapted to the non-analytic Kelvin phase, achieving a 104-105 speedup over direct quadrature while preserving wake asymptotics.

  3. Robust Numerical Integration: The AI can perform accurate numerical integration of the wavelike kernel using a partitioned contour deformation approach, which is described as being efficient and universally robust for evaluations at the z = 0 limit, even when dealing with non-analytic phase functions.

  4. Flat-Ship Design Optimization: The system can predict wave fields and wave resistance for flat-ship plans across various Froude numbers, specifically capturing the nonlinear variation in the potential with b, which is encoded in the Bessel function of the wavelike kernel.

  5. Resistance Prediction Accuracy: The system can accurately calculate a well-posed wave resistance coefficient Cw that shows a monotonic decrease due to destructive diverging wave interference, contrasting with classical theory's prediction of quadratic increase.

Abstract

Linear wave theory captures the essential physics of free-surface flows at a fraction of the computational cost of nonlinear and viscous methods, making it attractive for design, real-time control, and surrogate modeling applications. However, the Kelvin Green's function for a translating point-source generates unbounded wave energy in the z to 0 limit, causing both numerical difficulties and physical inconsistencies. This paper develops a modified kernel for the Kelvin potential incorporating an elliptic spanwise line integration that naturally resolves this ill-posedness, yielding finite wave energy over the entire free surface. We then present a fast evaluator for both point and line kernels using contour deformation adapted to the non-analytic Kelvin phase, achieving 10 4 - 10 5 speedup over direct quadrature while preserving the wake asymptotics. Predictions on the most challenging z=0 limit demonstrate physically consistent wave patterns and wave resistance trends. An open-source Julia implementation is provided.

Related papers