A class of structurally complete approximate Riemann solvers for trans- and supercritical flows with large gradients
Introduction
Flux computations are essential to the finite-volume method in Computational Fluid Dynamics (CFD). In Godunov schemes, used for compressible flows, the numerical fluxes at the intercell boundaries are evaluated by solving either an approximate or an exact Riemann problem. Many approximate Riemann solvers have been developed for practical use in CFD [1], and are typically non-iterative to prioritize computational speed. Such solvers also tend to assume a perfect gas to relate the thermodynamic variables. From these flux estimates it is possible to construct efficient numerical schemes for simulating flow problems of interest to scientists and engineers [2]. With the increasing interest in systems operating at thermodynamic conditions that depart from the ideal assumptions—often characterized by highly non-linear thermodynamic coupling and computationally expensive evaluation of fluid properties [3]—the accuracy and efficiency of the Riemann solvers become increasingly relevant.
For nearly four decades, there has been a continual pursuit to improve the representation of the wave structure within Riemann solvers. This is because improvements to the design of the Riemann solver can result in increased accuracy at minimal additional expense as compared to leveraging more involved discretization schemes. In some cases, it is impossible to resolve certain flow phenomena if the Riemann solver omits the necessary waves [4]. For instance, the HLLC solver [5] restored the missing contact discontinuity in the HLL flux [6], leading to improved resolution of material interfaces and sharp physical features in contexts such as supersonic and shallow water flows. Also, the HLLE [7] and HLLEM [8] solvers addressed issues with wavespeeds to ensure positively conservative results, particularly under vacuum conditions. To date, the highest fidelity approximate solvers consist of three-wave models [9], [10], [5], [11], [12], named as such because they model the three spatially distinct waves that are observed in the solution to the one-dimensional Riemann problem.
A major limitation of most approximate solvers is that rarefactions in the solution to the Riemann problem are simplified to discontinuous jumps—the spatially varying nature of rarefactions is lost despite their presence in exact solutions as well as in the underlying physical problem. This occurs due to the consideration of piecewise constant states between the wave fronts, resulting from the linearization of the governing equations. At a fully subsonic or supersonic state (Fig. 1.a), the rarefaction does not enclose the cell interface and the omission of an exact expansion wave is benign; the intercell flux is determined by other regions in the solution. However, if a rarefaction is present in a transonic scenario (Fig. 1.b), and it is approximated as a discontinuous jump, one can prove that the resulting weak solution violates the entropy condition. In the context of this work, we denote a transonic scenario when the head of the rarefaction is subsonic while the tail is supersonic. Because transonic scenarios are more prevalent at higher pressures and temperatures, the issue of entropy violations is especially relevant to the study of trans- and supercritical flows. Harten and Hyman [13], Osher [14], and Quirk [15] were among the first to explore entropy issues and fluxes due to transonic rarefactions in Riemann solvers. They showed that entropy violations are consistent with the mathematical definition of hyperbolic conservation laws but are thermodynamically inconsistent for the purposes of simulating real-world flows. Entropy-violating solutions frequently contain nonphysical phenomena such as expansion waves that suddenly decay into a shock front, also called expansion or rarefaction shocks.
Various entropy fixes have been developed over the years [13], [16], and they are generally modeled after Harten and Hyman's [13] approach of introducing a new intermediate state to approximate the lost rarefaction wave. The new intermediate state is often treated as a constant; alternatively, it can be linearly or polynomially interpolated between known states. This has the effect of introducing additional diffusivity in the flux terms to mitigate any expansion shocks. It has also been shown, in the case of perfect gases, that it is possible to calculate the flux analytically [17], [2], [1]. Even so, a simple and analytically correct means of restoring the expansion wave for arbitrary Riemann solvers—and especially under non-ideal thermodynamics—has not yet been demonstrated.
Recent contributions in the area of entropy stability have instead focused on extending existing entropy concepts to new applications. Few fundamental improvements have been made to the design of the Riemann solver itself. For example, studies have investigated entropy violations in boundary conditions [18], higher dimensions [19], [20], multicomponent flows [21], or hybridized Riemann solvers that switch or average between different flux estimates [22], [23]. The ideal gas assumption is usually made, and any entropy fixes follow the classical implementation or with minor optimizations. Other works have also examined entropy stability in the context of discontinuous Galerkin schemes [24], [25], [26], magnetohydrodynamics [27], Lagrangian gas dynamics [28], relativistic hydrodynamics [29], and nonclassical dense gases where rarefaction shocks are physically admissible [30], [31], [32], [33], [34]. The case of a single-species gas with arbitrary state equation obeying the Euler equations has thus far been overlooked. Studying this particular problem would facilitate the analysis of nonphysicalities attributable only to the Riemann solver.
We propose a class of structurally complete approximate Riemann solvers (StARS) that use recent derivations by Wang and Hickey [35] to analytically restore the expansion wave in pre-existing three-wave solvers. By structurally complete and approximate, it is meant that StARS provides explicit non-iterative means to compute: 1) wave speeds associated with the method of characteristics, i.e. normal shocks, contact discontinuities, rarefaction heads, and rarefaction tails; and 2) primitive and conservative variables as well as fluxes in each region between these waves—in particular, expansion waves are analytically reconstructed and not approximated as a constant or interpolated state. The result is a class of efficient approximate solvers that offer improved accuracy, the benefits of which are especially apparent under transonic flux conditions. Most importantly, this property of structural completeness is valid for both ideal and non-ideal thermodynamics. In this study, the Roe solver [36], whose entropy properties have been widely studied, is compared to a structurally complete version of the Roe solver (dubbed Roe-StARS) across compressible flow test cases where entropy violations arise. We also perform a comprehensive scaling analysis of flow conditions that give rise to transonic rarefactions, yielding a clear conceptual understanding of the thermodynamic and flow conditions in which such rarefactions occur. It is shown that transonic fluxes are particularly prevalent in trans- and supercritical flows with large thermophysical gradients.
The remaining paper is organized in the following manner. Section 2 describes the governing equations, thermodynamics, and numerical methods relevant to general aspects of Riemann solvers as well as the specific choice of solver and numerical scheme adopted in this study. Section 3 describes the general approach in restoring the expansion wave to an arbitrary three-wave solver, so that transonic fluxes are correctly accounted for. Section 4 conducts a scaling analysis of the flow conditions under which transonic fluxes occur and the errors that can arise if they are omitted. Finally, section 5 provides numerical results for a transcritical shock tube, shock tube with periodic bounds resulting in interfering shocks and rarefactions, a so-called gradient Riemann problem, and a two-dimensional Riemann problem. Provided at the end are a brief conclusion and appendices containing supporting derivations.
Section snippets
Governing equations
The flow is assumed to be inviscid, isentropic, and one-dimensional. Thus, the time-dependent Euler equations, in conservative form, are considered: where ρ is density, u is the velocity component in the x-direction, p is pressure, M is the molar mass of the fluid, e is specific internal energy on a molar basis, and t is time. The specific enthalpy, on a molar basis, may be expressed as where is the molar
Detecting the presence of a rarefaction at the cell interface
Fig. 1 depicts the rarefaction-contact-shock solution configuration that is often shown in textbooks and papers, although rarefactions and/or shocks can occur on the left, right, or both flanks of the star region (outlined in blue). The speed of various characteristic waves is denoted as S followed by the appropriate subscript. Pressure and velocity are uniform throughout the star region, and there is always a contact discontinuity wave located within the star region. Of interest are the
Scaling analysis
Here we perform a scaling analysis on the errors when transonic rarefactions are omitted from the Riemann solver. For demonstrative purposes, the Roe solver [36] and the Roe-StARS version are analyzed with nitrogen gas as the working fluid. The approach may be trivially extended to other three-wave solvers and media of interest.
Numerical results
Numerical results are compared for four test cases involving shocks and rarefactions with nitrogen gas: 1) a transcritical shock tube, 2) a shock tube with periodic boundaries and interfering waves, 3) a novel interpretation of the Riemann problem as the limiting case of general flow gradients; 4) a two-dimensional Riemann problem. The results of the third test case are analyzed together with the scaling analysis. The ensuing discussion focuses on the adverse effects of entropy violations and
Conclusion
A general and simple approach was described for developing structurally complete approximate Riemann solvers, by analytically restoring the expansion wave in pre-existing three-wave solvers. The accuracy improvement at each call of a StARS solver is on the order of a few percent, and thus their effect over millions of cells and time steps across the entire computational domain can be significant particularly for trans- and supercritical flows with large flow gradients. The restoration of the
CRediT authorship contribution statement
Jeremy C.H. Wang: Conceptualization, Formal analysis, Methodology, Software, Writing – original draft. Jean-Pierre Hickey: Conceptualization, Supervision, Writing – review & editing.
Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Acknowledgements
We thank the Natural Sciences and Engineering Research Council of Canada (NSERC) and University of Waterloo Graduate Fellowships for funding this research.
References (66)
- et al.
On Godunov-type methods near low densities
J. Comput. Phys.
(1991) An HLLC Riemann solver for magneto-hydrodynamics
J. Comput. Phys.
(2005)- et al.
Self-adjusting grid methods for one-dimensional hyperbolic conservation laws
J. Comput. Phys.
(1983) - et al.
Cell-vertex entropy-stable finite volume methods for the system of Euler equations on unstructured grids
Comput. Math. Appl.
(2021) - et al.
A simple parameter-free entropy correction for approximate Riemann solvers
C. R. Mecanique
(2010) - et al.
Hybrid entropy stable HLL-type Riemann solvers for hyperbolic conservation laws
J. Comput. Phys.
(2017) Approximate Riemann solvers, parameter vectors, and difference schemes
J. Comput. Phys.
(1981)- et al.
Efficient solution algorithms for the Riemann problem for real gases
J. Comput. Phys.
(1985) - et al.
Weighted essentially non-oscillatory schemes
J. Comput. Phys.
(1994) - et al.
Non-intrusive reduced order modeling of unsteady flows using artificial neural networks with application to a combustion problem
J. Comput. Phys.
(2019)
An entropy-stable hybrid scheme for simulations of transcritical real-fluid flows
J. Comput. Phys.
Computations of compressible multifluids
J. Comput. Phys.
Front tracking and two-dimensional Riemann problems
Adv. Appl. Math.
Two-dimensional Riemann problem for a hyperbolic system of nonlinear conservation laws I. Four j-cases
J. Differ. Equ.
Riemann Solvers and Numerical Methods for Fluid Dynamics
Finite-Volume Methods for Hyperbolic Problems
Deep-learning accelerated calculation of real-fluid properties in numerical simulation of complex flowfields
J. Comput. Phys.
The HLLC Riemann solver
Shock Waves
Restoration of the contact surface in the HLL-Riemann solver
Shock Waves
On upstream differencing and Godunov-type schemes for hyperbolic conservation laws
SIAM Rev.
On Godunov-type methods for gas dynamics
SIAM J. Numer. Anal.
A linearised Riemann solver for the time–dependent Euler equations of gas dynamics
Proc. R. Soc. Lond.
Riemann problems and the waf method for solving two–dimensional shallow water equations
Philos. Trans. Royal Soc. A
An HLLC Riemann solver for relativistic flows – I. Hydrodynamics
Mon. Not. R. Astron. Soc.
Riemann solvers, the entropy condition, and difference approximations
SIAM J. Numer. Anal.
A contribution to the great Riemann solver debate
Int. J. Numer. Methods Fluids
A non-parameterized entropy correction for Roe's approximate Riemann solver
Numer. Math.
Sonic flux formulae
SIAM J. Sci. Stat. Comput.
Entropy stable boundary conditions for the Euler equations
J. Comput. Phys.
Shock-stable Roe scheme combining entropy fix and rotated Riemann solver
AIAA J.
Formulation of entropy-stable schemes for the multicomponent compressible Euler equations
Comput. Methods Appl. Mech. Eng.
On the entropy conserving/stable implicit DG linearization of the Euler equations in entropy variables
Comput. Fluids
Entropy stable, robust and high-order DGSEM for the compressible multicomponent Euler equations
J. Comput. Phys.
Cited by (3)
FluxNet: A physics-informed learning-based Riemann solver for transcritical flows with non-ideal thermodynamics
2023, Computer Methods in Applied Mechanics and Engineering