Elsevier

Journal of Computational Physics

Volume 273, 15 September 2014, Pages 671-705
Journal of Computational Physics

On multi-time-step monolithic coupling algorithms for elastodynamics

https://doi.org/10.1016/j.jcp.2014.05.034Get rights and content

Abstract

We present a way of constructing multi-time-step monolithic coupling methods for elastodynamics. The governing equations for constrained multiple subdomains are written in dual Schur form and the continuity of velocities is enforced at system time levels. The resulting equations will be in the form of differential–algebraic equations. To crystallize the ideas we shall employ Newmark family of time-stepping schemes. The proposed method can handle multiple subdomains, and allows different time-steps as well as different time stepping schemes from the Newmark family in different subdomains. We shall use the energy method to assess the numerical stability, and quantify the influence of perturbations under the proposed coupling method. Two different notions of energy preservation are introduced and employed to assess the performance of the proposed method. Several numerical examples are presented to illustrate the accuracy and stability properties of the proposed method. We shall also compare the proposed multi-time-step coupling method with some other methods available in the literature.

Section snippets

Introduction and motivation

Coupled problems (such as fluid–structure interaction, structure–structure interaction and thermal–structure interaction) have been the subject of intense research in recent years in both computational mechanics and applied mathematics. The report compiled by the Blue Ribbon Panel on Simulation-Based Engineering Science emphasizes that the ability to solve coupled problems will be vital to accelerate the advances in engineering and science through simulation [1]. Developing stable and accurate

Newmark family of time-stepping schemes

Consider a system of second-order ordinary differential equations of the following form:Mu¨(t)+Ku(t)=f(t)t(0,T] where t denotes time, T denotes the time interval of interest, M is a symmetric positive definite matrix, K is a symmetric positive semidefinite matrix, and a superposed dot denotes derivative with respect to the time. The above system of equations can arise from a semi-discrete finite element discretization of the governing equations in linear elastodynamics [25]. In this case, M is

Governing equations for multiple subdomains

We now write governing equations for multiple subdomains. We will also outline various ways to write subdomain interface conditions, and discuss their pros and cons. To this end, let us divide the domain Ω into S non-overlapping subdomains, which will be denoted by Ω1,,ΩS. That is,Ω=i=1SΩiandΩiΩj=forij We shall assume that the meshes in the subdomains are conforming along the subdomain interface, as shown in Fig. 1. There are several ways to enforce the continuity along the interface, and

Proposed multi-time-step coupling method

The aim of this paper is to solve Eqs. (15a), (15b) numerically by allowing each subdomain to have its own time-step and its own time integrator from the Newmark family of time stepping schemes. We first introduce notation that will help in presenting the proposed multi-time-step coupling method in a concise manner.

Stability analysis using the energy method

We shall employ the energy method to show the stability of the proposed multi-time-step coupling method. The energy method is a popular strategy employed in Mathematical Analysis to derive estimates and to perform stability analysis. The method is widely employed in the theory of partial differential equations [39], and numerical analysis [25], [40]. The basic idea behind the energy method is to choose an appropriate norm (which is referred to as the energy norm) and show that the solution is

Split degree-of-freedom lumped parameter system

Consider a split agree of freedom whose motion can be described by the following system of ordinary differential/algebraic equations:mAu¨A(t)+kAuA(t)=fA(t)+λ(t)mBu¨B(t)+kBuB(t)=fB(t)λ(t)u˙A(t)u˙B(t)=0 The following parameters are used: mA=0.1, mB=0.005, and the stiffness of springs are kA=2.5 and kB=50. The subdomain time-steps are taken as ΔtA=0.02 and ΔtB=0.005. The system time-step is taken as Δt=0.02. The values of the external forces are taken to be zero, that is fA=0 and fB=0. The

On energy conserving vs. energy preserving coupling

In this section we address the energy preserving and energy conserving properties of the proposed multi-time-step coupling method. Two different notions of energy preserving will be considered. In particular, the following questions will be answered:

  • (a)

    Does the coupling method add or extract energy from the system of subdomains in comparison with the case of no coupling?

  • (b)

    Do the interface forces perform net work?

  • (c)

    Under what conditions does the coupling method conserve the total energy of the system

On the performance of backward difference formulae and implicit Runge–Kutta schemes

In the numerical analysis literature, backward difference formulae (BDF) and implicit Runge–Kutta (IRK) schemes have been the schemes of choice for solving DAEs [18], [41]. The following quote by Petzold has been a popular catch-phrase for promoting BDF schemes: “BDF is so beautiful that it is hard to imagine something else could be better” [41, p. 481]. This statement may be true for first-order DAEs that arise from modeling of physical systems involving dissipation. But these two families of

Representative numerical results

Using several canonical problems, we illustrate that the proposed multi-time-step coupling method possesses the following desirable properties:

  • (I)

    All subdomains can subcycle simultaneously. That is, Δti<Δti=1,,S.

  • (II)

    The method can handle multiple subdomains.

  • (III)

    Drift in displacements along the subdomain interface is not significant.

  • (IV)

    Under fixed subdomain time-steps, the accuracy of numerical solutions can be improved by decreasing the system time-step.

  • (V)

    For a fixed system time-step, accuracy of the

Concluding remarks

We have developed a multi-time-step coupling method that can handle multiple subdomains with different time-steps in different subdomains. The coupling method can couple implicit and explicit time-stepping schemes under the Newmark family even with disparate time-steps of more than two orders of magnitude in different subdomains. A systematic study on the energy preservation and energy properties of the proposed coupling method is presented, and the corresponding sufficient conditions are also

Acknowledgements

The authors acknowledge the support of the National Science Foundation under Grant No. CMMI 1068181. The opinions expressed in this paper are those of the authors and do not necessarily reflect that of the sponsor(s).

References (47)

  • T. Belytschko et al.

    Explicit integration of structural problems

    Finite Elem. Nonlinear Mech.

    (1977)
  • T. Belytschko et al.

    Stability of explicit–implicit mesh partitions in time integration

    Int. J. Numer. Methods Eng.

    (1978)
  • T.J.R. Hughes

    Generalization of selective integration procedures to anisotropic and nonlinear media

    Int. J. Numer. Methods Eng.

    (1980)
  • T.J.R. Hughes et al.

    Implicit–explicit finite elements in transient analysis: stability theory

    J. Appl. Mech.

    (1978)
  • P. Smolinski et al.

    Multi-time-step integration using nodal partitioning

    Int. J. Numer. Methods Eng.

    (1988)
  • W.J.T. Daniel

    The subcycled Newmark algorithm

    Comput. Mech.

    (1997)
  • P.B. Nakshatrala et al.

    A time-staggered partitioned coupling algorithm for transient heat conduction

    Int. J. Numer. Methods Eng.

    (2009)
  • A. Akkasale

    Stability of coupling algorithms

    (2011)
  • L. Petzold

    Differential/algebraic equations are not ODEs

    SIAM J. Sci. Stat. Comput.

    (1982)
  • U.M. Ascher et al.

    Computer Methods for Ordinary Differential Equations and Differential–Algebraic Equations

    (1998)
  • L. He et al.

    Convergence analysis of a parallel interfield method for heterogeneous simulations with dynamics substructuring

    Int. J. Numer. Methods Eng.

    (2008)
  • A. Bonelli et al.

    Novel generalized-α methods for interfield parallel integration of heterogeneous structural dynamics systems

    J. Comput. Appl. Math.

    (2010)
  • A. Bonelli et al.

    Novel partitioned time integration methods for DAE systems based on L-stable linearly implicit algorithms

    Int. J. Numer. Methods Eng.

    (2011)
  • Cited by (23)

    • High-efficient decoupling method for coupling systems with multiple subdomains and time steps

      2022, Mechanical Systems and Signal Processing
      Citation Excerpt :

      The main advantages of MMTS are: by using FETI method, a complex or large-scale structure can be divided into different subdomains with non-overlapping elements [12,13], and each subdomain is solved separately. Moreover, by imposing velocity continuity conditions on the interfaces shared nodes, Gravouil and Combescure (GC method) proposed [31] and improved [32–38] the MMTS methods to couple arbitrary Newmark scheme. However, energy conservation can only be retained for the case with a unique time step for all subdomains [32,39].

    • Convergence and stability analysis of heterogeneous time step coupling schemes for parabolic problems

      2017, Applied Numerical Mathematics
      Citation Excerpt :

      This implies that time integration algorithms can be used to solve each subdomain (partition) separately and then couple the solutions together. Much of this research has addressed hyperbolic problems [7,10,11,15,16,20,31,32,37]. On the other hand, subcycling methods for parabolic problems are not as mature as those associated with second-order transient problems (structural dynamics).

    • A monolithic multi-time-step computational framework for first-order transient systems with disparate scales

      2015, Computer Methods in Applied Mechanics and Engineering
      Citation Excerpt :

      The proposed monolithic multi-time-step coupling framework for first-order transient systems enjoys several attractive features, which will be illustrated in the subsequent sections by both theoretical analysis and numerical results. In the remainder of this paper, we shall closely follow the notation introduced for multi-time-step coupling in [8]. We shall consider transient advection–diffusion–reaction equation as the continuous model problem.

    View all citing articles on Scopus
    1

    Saeid Karimi, Graduate Student.

    View full text