An assessment of semi-discrete central schemes for hyperbolic conservation laws.

Abstract

High-resolution finite volume methods for solving systems of conservation laws have been widely embraced in research areas ranging from astrophysics to geophysics and aero-thermodynamics. These methods are typically at least second-order accurate in space and time, deliver non-oscillatory solutions in the presence of near discontinuities, e.g., shocks, and introduce minimal dispersive and diffusive effects. High-resolution methods promise to provide greatly enhanced solution methods for Sandia's mainstream shock hydrodynamics and compressible flow applications, and they admit the possibility of a generalized framework for treating multi-physics problems such as the coupled hydrodynamics, electro-magnetics and radiative transport found in zpinch physics. In this work, we describe initial efforts to develop a generalized "black-box" conservation law framework based on modern high-resolution methods and implemented in an object-oriented software framework. The framework is based on the solution of systems of general non-linear hyperbolic conservation laws using Godunov-type central schemes. In our initial 1 Keywords and Phrases: high-resolution, hyperbolic conservation laws, Godunov, semidiscrete, central schemes 2 Authors listed in alphabetic order.3 efforts, we have focused on central or central-upwind schemes that can be implemented with only a knowledge of the physical flux function and the minimal/maximal eigenvalues of the Jacobian of the flux functions, i.e., they do not rely on extensive Riemann decompositions. Initial experimentation with high-resolution central schemes suggests that contact discontinuities with the concomitant linearly degenerate eigenvalues of the flux Jacobian do not pose algorithmic difficulties. However, central schemes can produce significant smearing of contact discontinuities and excessive dissipation for rotational flows. Comparisons between "black-box" central schemes and the piecewiseparabolic method (PPM), which relies heavily on a Riemann decomposition, shows that roughly equivalent accuracy can be achieved for the same computational cost with both methods. However, PPM clearly outperforms the central schemes in terms of accuracy at a given grid resolution and the cost of additional complexity in the numerical flux functions. Overall we have observed that the finite volume schemes, implemented within a well-designed framework, are extremely efficient with (potentially) very low memory storage. Finally, we have found by computational experiment that second and third-order strong-stability preserving (SSP) time integration methods with the number of stages greater than the order provide a useful enhanced stability region. However, we observe that non-SSP and non-optimal SSP schemes with SSP factors less than one can still be very useful if used with time-steps below the standard CFL limit. The "well-designed" integration schemes that we have examined appear to perform well in all instances where the time step is maintained below the standard physical CFL limit. Acknowled...

Generalized fourier analyses of the advection–diffusion equation—Part II: two‐dimensional domains

“…The ultimate decision on a 'best' method for a given problem can be easily discerned by using the computational e ciency, i.e. CPU time, to achieve a given level of accuracy as a metric (see for example, Reference [4]). …”

On the practical importance of the SSP property for Runge-Kutta time integrators for some common Godunov-type schemes

“…It also has a truncation error half that of the optimal 2N method. We refer to this method as Optimal SSP33 ( In this section we summarize the spatial discretizations used for our numerical experiments, the numerical ux methods we employ, including two approximate central-upwind uxes and an exact Riemann ux, and the associated reconstruction methods [25]. We comment on two non-oscillatory reconstruction methods found in the literature and show by test and analysis that they are actually not third-order in general as previously claimed.…”

“…We acknowledge the important contributions of Mark A. Christon to the development and documentation of the conservation law code capability associated with this work [25]. The NEVADA code framework developed at Sandia National Laboratories provides the baseline software infrastructure.…”

Nonmonotone chemotactic invasion: High-resolution simulations, phase plane analysis and new benchmark problems

“…Typically HCL algorithms are tested with a suite of benchmark problems starting with single species linear advection [2,3,6,10]. Linear advection has the advantage of being conceptually simple, analytically tractable and clearly identifies susceptibility to numerical diffusion and artificial oscillation.…”

“…Although various other HCL benchmarks are used for algorithm development, such as traffic flow problems, acoustic dynamics and the Buckley-Leverett equation [2,3,6,10], there does not appear to be any particular problem more complex than the shock tube problem that has become a clear standard choice for benchmarking. We propose a new test case which has arisen in computational biology relevant to chemotactic cell invasion [15,20].…”