Elsevier

Computers & Fluids

Volume 117, 31 August 2015, Pages 247-261
Computers & Fluids

A space–time adaptive discontinuous Galerkin scheme

https://doi.org/10.1016/j.compfluid.2015.05.002Get rights and content

Highlights

  • High order discontinuous Galerkin scheme in a space time formulation.

  • Local time-stepping and space–time refinement.

  • Polymorphic grid cells with hanging nodes.

  • Local basis enrichment and adapted time approximation.

  • Implementation and control of multiple adaptation techniques.

Abstract

A discontinuous Galerkin scheme for unsteady fluid flows is described that allows a very high level of adaptive control in the space–time domain. The scheme is based on an explicit space–time predictor, which operates locally and takes the time evolution of the data within the grid cell into account. The predictor establishes a local space–time approximate solution in a whole space–time grid cell. This enables a time-consistent local time-stepping, by which the approximate solution is advanced in time in every grid cell with its own time step, only restricted by the local explicit stability condition. The coupling of the grid cells is solely accomplished by the corrector which is determined by the numerical fluxes. The considered discontinuous Galerkin scheme allows non-conforming meshes, together with p-adaptivity in 3 dimensions and h/p-adaptivity in 2 dimensions. Hence, we combine in this scheme all the flexibility that the discontinuous Galerkin approach provides. In this work, we investigate the combination of the local time-stepping with h- and p-adaptivity. Complex unsteady flow problems are presented to demonstrate the advantages of such an adaptive framework for simulations with strongly varying resolution requirements, e.g. shock waves, boundary layers or turbulence.

Introduction

An important advantage of the discontinuous Galerkin (DG) approach consists of its high flexibility with respect to the spatial approximation: The approximation of the solution is represented by a piecewise polynomial and the degree of the local polynomials determine the order of accuracy in space. Hence, increasing the order can be achieved through a simple enrichment of the local basis. This can also be handled locally in regions where we aim a higher resolution. Geometrical flexibility is also an important feature of the DG method since it can be formulated on general unstructured grids with non-conforming meshes to handle complex geometries. The DG method is thus an ideal candidate, for which general adaptive strategies as local grid refinement (h-refinement) or local choice of the order of accuracy (p-adaptivity) can be applied, see, e.g., [25]. Due to this distinguished versatile flexibility the DG method is particularly suited for multi-scale problems which require small regions of high resolution while other parts can be discretized uniformly.

Adaptive concepts for discontinuous Galerkin schemes have been successfully applied, e.g., by Burgess and Mavriplis [7], Kopera and Giraldo [26] or Hartmann and Houston [22], [23]. All these approaches based on DG schemes with implicit or semi-implicit time integration. As we are particularly interested in unsteady flow problems, the adaptivity of the time approximation is an additional issue and is a focal point in this paper. Especially for wave dominated problems like acoustics and propagating shock waves, the time accuracy has to be preserved. Regions with local refinement and unsteady phenomena need small time steps. Allowing large time steps for a transient problem would lead to numerical diffusion and to loss of information. For an explicit scheme, a global time-stepping approach can cause inefficiency, since the smallest time step forced by small sized grid cells dominates the whole simulation. To overcome this inefficiency, we employ an explicit time approximation with local time-stepping, which allows each grid cell to take the maximum locally stable time step. The local time-stepping enables a powerful adaptation framework, which also takes the coupling of spatial and temporal scales into account since any local h-refinement or p-enrichment in space inherently implies an adaptation of the time step as well. Together with the explicit local time-stepping approach, the artificial viscosity based shock capturing introduced by Persson and Peraire for an implicit DG scheme [35] now also becomes an attractive method to robustly resolve shock waves.

In the following we describe an explicit discontinuous Galerkin scheme that incorporates the following properties:

  • high order accurate spatial approximations on general unstructured grids with triangles, quadrilaterals, tetrahedra, hexahedra, pyramids, prisms,

  • arbitrary high order in space and time,

  • non-conforming meshes in 2 and 3 dimensions,

  • fully conservative,

  • shock capturing with artificial viscosity,

  • local time-stepping with the time step from the local stability constraint,

  • h/p-adaptivity in 2 dimensions and p-adaptivity in 3 dimensions.

Different building blocks of our DG approach have been constructed and already been published in a series of papers [30], [15], [16], [19]. The scope of this work is the investigation of the combination and interaction of explicit local time-stepping, artificial viscosity based shock capturing and h/p-adaptivity. To demonstrate the high potential of the adaptivity framework, simulation results of various complex unsteady flow problems are presented.

The paper is organized as follows. We first describe the main components of the space approximation within the semi-discrete approach in Section 2. The time consistent local time-stepping method and the determination of the time step are introduced in Section 3 to obtain the fully discrete formulation. The adaptivity framework (h- and p-adaptivity) and its implementation with respect to the local time-stepping scheme is discussed in Section 4. In Section 4 we also provide a brief overview of the artificial viscosity based shock capturing as part of the whole adaptation strategy. In Section 5, we focus on the simulation results and present different applications of the space–time-adaptive framework for two- and three-dimensional complex flows and also address the parallel performance and dynamic load balancing attributes of the implementation. Conclusions and final remarks are given in Section 6.

Section snippets

Approximation in space

For simplicity we restrict the derivations to a scalar advection–diffusion equation, which is formulated in conservation form asut+·fa(u)=·fd(u,u).Here, u=u(x,t) denotes the conserved solution variable, fa is the advection flux and fd the diffusion flux. The diffusion in the flux formulation may be rewritten as·fd(u,u)=·μ(u)u,where μ=μ(u) is the non-linear diffusion coefficient.

First, we consider the discretization in space. The spatial domain Ω is subdivided into

Time-consistent local time-stepping

The locality of the proposed explicit space–time discontinuous Galerkin scheme allows a completely new time-marching technique: Each grid cell may run with its own time step in a time-consistent manner. The local time-stepping algorithm is explained in the next subsection, followed by the determination of the maximum time step in each element.

Adaptivity in the space–time domain

The discontinuous trial space allows spatial adaptivity by changing the mesh size (h-adaptivity) or changing the local polynomial degree (p-adaptivity) in a straightforward way. The strategy to extend this to the space–time domain with respect to the local time-stepping is described in the following, whereas h-adaptivity is applied in two dimensions and p-adaptivity in two and three dimensions.

Numerical results

In this section, we first investigate the two-dimensional setup with shock capturing. The h/p-adaptivity with local time stepping is demonstrated on a transonic laminar flow around an airfoil. The combination of local time stepping and the capturing of strong moving shocks is presented for the standard double Mach reflection problem [45]. In three-dimensions, we combine the local time stepping with p-adaptivity to simulate the flow past a sphere. Finally, we present a simulation of an unsteady

Conclusions

We described a high order discontinuous Galerkin scheme for the numerical solution of unsteady convection–diffusion equations. The DG scheme is formulated in a space–time domain based on a predictor–corrector formulation and generalizes the solution in the small approach of the finite volume framework. The space–time discretization is kept explicit by starting with the predictor that determines approximate values at all intermediate time levels needed for high order accuracy in time. The

Acknowledgements

The research presented in this paper was supported in parts by the Deutsche Forschungsgemeinschaft (DFG) in the context of the Cluster of Excellence Simulation Technology (SimTech), the Research Training Group GRK 1095 ‘Aero-thermodynamic design of a scramjet propulsion system’ and the German Ministry of Research (BMBF) in the collaborative project Software for Scalable Parallel Computers.

References (46)

  • M.A. Kopera et al.

    Analysis of adaptive mesh refinement for IMEX discontinuous Galerkin solutions of the compressible Euler equations with application to atmospheric simulations

    J Comput Phys

    (2014)
  • L. Krivodonova et al.

    Shock detection and limiting with discontinuous Galerkin methods for hyperbolic conservation laws

    Appl Numer Math

    (2004)
  • J. Qiu et al.

    The discontinuous Galerkin method with Lax–Wendroff type time discretizations

    Comput Methods Appl Mech Eng

    (2005)
  • J. Qiu et al.

    Hermite WENO schemes and their application as limiters for Runge–Kutta discontinuous Galerkin method II: two dimensional case

    Comput Fluids

    (2005)
  • J.J.W. van der Vegt et al.

    Space-time discontinuous Galerkin finite element method with dynamic grid motion for inviscid compressible flows I. General formulation

    J Comput Phys

    (2002)
  • H. van der Ven et al.

    Space-time discontinuous Galerkin finite element method with dynamic grid motion for inviscid compressible flows II. Efficient flux quadrature

    Comput Methods Appl Mech Eng

    (2002)
  • B. van Leer

    Towards the ultimate conservative difference scheme V: a second order sequel to Godunov’s method

    J Comput Phys

    (1979)
  • P. Woodward et al.

    The numerical simulation of two-dimensional fluid flow with strong shocks

    J Comput Phys

    (1984)
  • J. Zhu et al.

    Runge–Kutta discontinuous Galerkin method using WENO limiters II: unstructured meshes

    J Comput Phys

    (2008)
  • Aluru S, Sevilgen FE. Parallel domain decomposition and load balancing using space-filling curves. In: Proceedings of...
  • D.N. Arnold et al.

    Unified analysis of discontinuous Galerkin methods for elliptic problems

    SIAM J Numer Anal

    (2002)
  • D.S. Balsara et al.

    A sub-cell based indicator for troubled zones in RKDG schemes and a novel class of hybrid RKDG+HWENO schemes

    J Comput Phys

    (2008)
  • N.K. Burgess et al.

    hp-adaptive discontinuous Galerkin methods for the Navier–Stokes equations

    AIAA J

    (2012)
  • Cited by (19)

    • On explicit discontinuous Galerkin methods for conservation laws

      2021, Computers and Fluids
      Citation Excerpt :

      In the rest of this paper, we focus on the moment schemes. Numerical tests as well as extensions for the STE-DG methods can be found in [32,10–12]. Preliminary 1D numerical solutions for the moment schemes are shown below.

    • A dynamically load-balanced parallel p-adaptive implicit high-order flux reconstruction method for under-resolved turbulence simulation

      2020, Journal of Computational Physics
      Citation Excerpt :

      In the literature, three major groups of adaptation methodologies can be found. The first one is the feature-based adaptation [38–40], the second one is the truncation-error-based or discretization-error-based adaptation [41–43], and the last one is the output-based or adjoint-based adaptation [44–47]. Feature-based adaptation methods are usually ad hoc and heavily rely on empirical parameters; however, their ease of implementation and reasonable robustness make them a good choice for adaptation.

    • Efficient simulation of one-dimensional two-phase flow with a high-order h-adaptive space-time Discontinuous Galerkin method

      2017, Computers and Fluids
      Citation Excerpt :

      Since all characteristics should exit an element face in the same direction, this would yield very flat elements (in time) if the eigenvalues of the system have a very large positive and negative component. Another technique proposed by Gassner et al. [14] involves a set of elements that are unstructured in space and extruded in time, where the time length of an element is variable. The flux contribution to an element is applied separately from the volume contribution after the volume contributions of all neighbouring elements have been computed.

    • A Sub-element Adaptive Shock Capturing Approach for Discontinuous Galerkin Methods

      2023, Communications on Applied Mathematics and Computation
    View all citing articles on Scopus
    View full text