General-domain FC-based shock-dynamics solver I: Basic elements
Oscar P. Bruno, Daniel V. Leibovici · 2025
Read General-domain FC-based shock-dynamics solver I: Basic elements by Oscar P. Bruno, Daniel V. Leibovici (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2506-16076.txt.
Introduction
Oscar Bruno and Daniel Leibovici present the core of a shock solver that works on complex shapes in a 2025 paper. Shock waves in gas flows demand methods that stay sharp without ringing. Their scheme blends smooth spectral math with a neural net that spots shocks so fake thickness lands only there.
A patchwork of overlapping tiles lets the solver fit curved, complex domains under any edge rules. No hand-tuned knobs or fix-up limiters are needed, and the layout runs well in parallel. Tests on standard 2D flow cases show the method at work. A second part extends it to sharp-cornered obstacles.
Catalog record
| CREATOR | Oscar P. Bruno, Daniel V. Leibovici |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 98 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2506-16076
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
General-domain FC-based shock-dynamics solver I: Basic elements Oscar P. Bruno, Daniel V. Leibovici Abstract This contribution, Part I in a two-part article series, presents a general-domain version of the FC-SDNN (Fourier Continuation Shock-detecting Neural Network) spectral scheme for the numerical solution of nonlinear conservation laws, which is applicable under arbitrary boundary conditions and in general domains . Like the previous simple-domain contribution (Journal of Computational Physics X 15 , (2022)), the present approach relies on the use of the Fourier Continuation (FC) method for accurate spectral representation of non-periodic functions (without the limiting CFL time-step constraints inherent in other spectral schemes) in conjunction with smooth artificial viscosity assignments localized in regions detected by means of a Shock-Detecting Neural Network (SDNN). Relying on such techniques, the present Part I paper introduces a novel multi-patch/subpatch artificial viscosity-capable domain decomposition strategy for complex domains with smooth boundaries , and it illustrates the methodology by means of a variety of computational results produced by an associated parallel implementation of the resulting shock-capturing algorithm in a present-day computing cluster. The subsequent Part II contribution then extends the algorithm to enable treatment of obstacles with non-smooth boundaries, it considers questions concerning parallelization and accuracy, and it presents comparisons with physical theory and prior experimental and computational results. The resulting multi-patch FC-SDNN algorithm does not require use of problem-dependent algorithmic parameters or positivity-preserving limiters, and, on account of its use of an overlapping-patch discretization, it is geometrically flexible and efficiently parallelized. A variety of numerical tests for the 2D Euler equations are presented, including the simulation of supersonic and hypersonic flows and shocks past physical obstacles at high speeds, such as e.g. Mach 25 re-entry flow speeds. Keywords: Shocks, Artificial viscosity, Conservation laws, Multipatch domain decomposition, Overlapping patches, Fourier continuation, Non-periodic domain, Spectral method, Machine learning, Neural networks. 1 Introduction In a sequence of two papers (whose Part I is the present contribution) we present a general-domain FC-SDNN (Fourier Continuation Shock-detecting Neural Network) spectral scheme for the numerical solution of nonlinear conservation laws under arbitrary boundary conditions and in general domains . Like its simple-domain counterpart [ 7 ] , the approach proposed here enjoys limited spatial dispersion away from shocks and other discontinuities and does not suffer from the limiting CFL constraints inherent in other spectral schemes. The present Part I paper introduces a novel multi-patch/subpatch artificial viscosity-capable domain decomposition strategy for complex domains with smooth boundaries , and it illustrates the methodology by means of a variety of computational results produced by an associated parallel implementation of the resulting shock-capturing algorithm in a present-day computing cluster. The Part II contribution [ 8 ] , in turn, extends the algorithm to enable treatment of obstacles with non-smooth boundaries; it includes results validating the accuracy and energy conservation properties of the proposed multi-patch-based artificial viscosity method; and it introduces a parallel implementation of high parallel efficiency and scalability. The proposed approach relies on use of the FC-Gram Fourier Continuation method [ 1 , 2 , 9 , 31 ] for spectral representation of non-periodic functions in conjunction with new multi-patch versions of the Shock-Detecting Neural Network (SDNN) and smooth artificial viscosity methods introduced in [ 45 , 7 ] . The neural network approach, which was introduced in [ 45 ] , utilizes Fourier series to discretize the gas dynamics and related equations in two-dimensional periodic rectangular domains, eliminates Gibbs ringing at shocks (whose locations are determined by means of an artificial neural network) by assigning artificial viscosity to a small number of discrete points in the immediate vicinity of the shocks. The use of the classical Fourier spectral method in that contribution restricts the method’s applicability to periodic problems (thus precluding, in particular, applicability to geometries containing physical boundaries), and its highly localized viscosity assignments give rise to a degree of non-smoothness, resulting in certain types of unphysical oscillations manifested as serrated contour-lines in the flow fields. As demonstrated in [ 7 ] , the FC-SDNN approach leverages smooth viscosity assignments to deliver accurate, essentially oscillation-free solutions for non-periodic problems with arbitrary boundary conditions. Building on this foundation, the present two-part contribution extends the methodology to general geometries via multi-domain decompositions. Illustrative results include Euler simulations of re-entry-speed flows past obstacles—reaching, for instance, Mach 25—highlighted in Section 7.2 ; cf. Remark 6 . The computational solution of systems of conservation laws has been tackled by means of a variety of numerical methods. The brief account provided in [ 7 , Sec. 1] reviews some of the important methodologies developed in this field over the last few decades, including: Low-order finite volume and finite difference methods, possibly equipped with slope limiters [ 28 , 29 ] ; ENO [ 20 ] and WENO [ 30 , 23 ] schemes as well as a hybrid FC/WENO solver [ 47 ] ; Artificial viscosity-based methods in the context of finite difference discretizations [ 44 , 52 , 16 , 26 , 27 ] and Discontinuous Galerkin discretizations [ 37 ] ; The entropy-viscosity (EV) method [ 19 , 24 ] ; The spectral viscosity method [ 33 , 49 , 50 , 51 , 32 ] , that relies on convolution with a compactly supported kernel and which, while demonstrated numerically on the basis of 1D problems, may, like the EV method, potentially result in oversmearing; The C-method [ 43 , 40 , 41 ] , which augments the hyperbolic system with an additional equation used to determine a spatio-temporally smooth viscous term, and which, like the EV method, requires selection of several problem-dependent parameters and algorithmic variations; and, finally, a number of ML-based techniques [ 42 , 15 , 48 , 45 , 46 ] which enhance the performance of classical shock capturing schemes. The relative advantages provided by the proposed FC-SDNN, which were reviewed in [ 7 , Sec. 1] , are also present in the general-domain multi-patch algorithms introduced in the present Part I-II paper sequence—resulting, in particular, in accurate solutions with smooth contour lines for flows in general domains. As indicated above, the FC-SDNN method presented in this paper provides a general solver for shock-wave problems governed by conservation laws by incorporating 1) The shock-detection methodology mentioned above; in conjunction with, 2) The aforementioned rapidly convergent FC-Gram Fourier Continuation representations (which, eliminating the Gibbs phenomenon that would otherwise arise from lack of periodicity, provide rapidly convergent Fourier expansions of non-periodic functions); as well as, 3) Smooth artificial viscosity assignments [ 7 ] based on use of certain newly-designed smooth viscosity windows —which results in (a) Smooth flow profiles away from discontinuities, as well as (b) Sharp shock resolution, while (c) Enabling applicability to gas-dynamics problems including very strong shocks, without recourse to positivity-preserving limiters for the density and pressure fields. Unlike polynomial spectral methods such as the one based on use of Chebyshev representations, the proposed approach does not give rise to quadratically refined meshes near interval endpoints, and it utilizes viscosity assignments that decay proportionally to the spatial mesh size, resulting in, 4) CFL restrictions that decay linearly with the spatial meshsize. Finally, in view of its use of an overlapping-patch domain decomposition strategy [ 1 , 2 , 14 ] , 5) The