Solving dynamic contact problems with local refinement in space and time
Introduction
The numerical simulation of dynamic contact problems plays an important role in many applications in mechanics, like the forming of sheet metal, crash tests and other examples of structures under impact, fracture dynamics, or tire rolling. Especially the latter application is a challenging task from the point of view of simulation, as the problem usually features a complex three-dimensional geometry, nonlinear elastic materials as well as dynamic effects. In addition, the contact zone is usually quite small compared to the size of the tire but needs to be resolved very accurately to get a good picture of the evolution of the contact pressures during rolling contact. More importantly, many other multiscale contact simulations require local refinement in the vicinity of the contact zone which is a priori unknown.
In order to be able to perform an accurate simulation of a car tire, there is a huge demand for a sound numerical scheme that combines a suitable multi-scale discretization in space and in time with an efficient solution algorithm for the frictional contact conditions. The algorithm has to be robust with respect to jumps in the material parameters as well as to provide an energy-consistent description of the dynamics. The aim of this work is to design such an algorithm by combining several well-established mathematical methods with some new approaches. All of them are described in the following.
The basis of the algorithm is provided by a decomposition of the original structure into several overlapping subdomains which have different mesh sizes. The global d-dimensional structure, d ∈ {2, 3}, is discretized with a relatively coarse mesh that does not resolve the details along the contact boundary, whereas at the contact area, an overlapping patch with a fine triangulation is introduced. A two-dimensional example of such a geometry is sketched in Fig. 1. In order to avoid an expensive volume coupling, the transfer between the subdomains is only performed at the inner (d − 1)-dimensional interface. Here, we employ the variationally consistent mortar method (see, e.g., [1], [2]) with dual Lagrange multipliers [3] to enforce the weak continuity of the traces.
The subject of domain decomposition methods is already well-established in the literature; we refer to [4], [5], [6] and the references therein for an overview of the topic. The construction of domain decomposition schemes which are robust with respect to the mesh size as well as to jumps in the material parameters has been the topic of several papers (e.g., [7], [8], [9]). In this work, we make use of the overlapping decomposition in order to obtain an iterative solution scheme whose convergence rate is bounded independently of the mesh size or the Lamé parameters in the subdomains.
The next important item is the treatment of the contact inequality constraints (see [10], [11], [12], [13], [14] and the references therein for an overview of the topic). These conditions are enforced in a variationally consistent way using dual Lagrange multipliers, allowing for the application of a primal–dual active set strategy [15], [16], [17]. This scheme can be interpreted as a semismooth Newton method [15] applied to a set of nonlinear equations describing the contact conditions [18], [19]. In combination with the iterative subdomain coupling, an inexact Newton scheme is obtained which still shows superlinear local convergence provided that appropriate stopping criteria for the inner iteration are satisfied [20], [21], [22].
The incorporation of inertia effects into the formulation makes the simulation of the nonlinear contact problem even more challenging. Standard time stepping algorithms like the trapezoidal rule generally lose their energy conservation property if applied to nonlinear problems. Possible remedies are presented in, e.g., [23], [24], [25] for nonlinear material laws and in [26], [27] for contact. But even with the energy-consistent formulation of [27], the computed results for the contact stresses show spurious oscillations in time. This is avoided employing a local modification of the mass matrix along the potential contact boundary [28], [29], [30].
The last feature of the algorithm is the possibility to use different time step sizes in the subdomains. Here, the main challenge lies in the construction of the interface constraints. A suboptimal choice can lead to numerical instability or to artificial dissipation at the interface, even if the time integrators in the subdomains are stable and energy-conserving [31], [32], [33]. In [34], [35], a time substepping scheme with linear interpolation of the velocity and the Lagrange multipliers has been proposed which is stable but dissipative at the interface. An improved energy-conserving scheme with linear interpolation of the multipliers has been analyzed in [36], [37]. However, both methods rely on the expensive exact solution of the resulting coupled system. In contrast, we present a time-discrete coupled system that uses different time step sizes in each subdomain, conserves the energy over a coarse time step and does not require the exact solution of the coupled problem.
At this point, one might ask whether the algorithm presented in this work possibly is just a combination of known numerical techniques. As a return, we recall that we aim for formulating an algorithm that is able to solve challenging dynamic contact problems with complex local geometries. This cannot be achieved without building upon the knowledge and experience gained in the above mentioned topics of numerical simulation, i.e., we combine domain decomposition with mortar coupling, contact modeling via semismooth Newton methods and energy-consistent time integration. However, the algorithm presented in this work also contains important new aspects. The main novelty is the consistent use of the overlapping domain decomposition approach for both the discretization and the solver, leading to a two-scale formulation both in space and in time which has been implemented with contact constraints and dynamic terms. This is complemented by an extensive theoretical and numerical investigation of the properties of the resulting method.
We now turn to the structure of the rest of this work. In Sections 2 Problem formulation for the linear setting, 3 Iterative coupling algorithm, 4 Numerical results for the linear setting, we present and investigate the proposed algorithm for a linear dynamic problem. Section 2 introduces the notation and the governing equations as well as the algebraic formulation of the resulting fully coupled system. In Section 3, we describe an efficient iterative solution scheme based on overlapping domain decomposition and analyse its convergence rate. Section 4 numerically confirms the theoretical results by means of several tests.
In Section 5, we extend the domain decomposition approach to nonlinear problems, leading to an inexact semismooth Newton method. A special focus is put on the approximation of the frictional contact conditions. Section 6 contains several numerical examples illustrating the efficiency and the robustness of the resulting iterative scheme.
The case of different time step sizes is analyzed in Section 7, where we present an energy-conserving formulation that can efficiently be solved by the iterative scheme used before. The numerical results in Section 8 illustrate that the scheme can considerably decrease the number of global systems to be solved and still provide a good local accuracy. Section 9 concludes the work with a short summary.
Section snippets
Problem formulation for the linear setting
This section contains the problem formulation as well as the basic notation for the rest of this work. In Section 2.1, the governing equations for the linear problem are stated in their strong and weak forms, whereas Section 2.2 introduces the spatial and temporal discretization, including some properties of trace spaces and mortar operators which will be used in the sequel. In Section 2.3, the algebraic Schur complement formulation is presented.
Since Sections 2.2 Discretization, 2.3 Schur
Iterative coupling algorithm
In this section, we derive and investigate an iterative solution scheme for the coupled system (14), where each step consists of the solution of a coarse problem on the full domain Ω and a subsequent solution of a local fine grid problem on ω. In Subsection 3.1, the algorithm is stated. The error propagation and the convergence rate of the iterative scheme are analyzed in Sections 3.2 Error propagation, 3.3 Condition number analysis, respectively, followed by a short discussion on how its
Geometry and parameters
For these first numerical tests, we consider the domain Ω = (0, 2) × (0, 1) which is split into the patch ω = (0.5, 1.5) × (0, 0.5) and the upper domain . Both subdomains are initially discretized by quadrilaterals of size H = h = 0.125; afterwards, we perform L additional refinements of the fine grid on ω. The resulting grid for L = 2 is depicted on the left side of Fig. 3. If not stated otherwise, we iteratively solve the equations of linear elasticity (1a), (1b), (1c) with l = 0 and do not apply damping,
Extension to nonlinear problems
Motivated by the challenge of simulating the large deformation dynamics of structures with contact, we have presented and analyzed a flexible iterative scheme that allows for a locally improved spatial resolution. In this section, we extend the considerations of Section 3 to the nonlinear case by incorporating nonlinear effects like material nonlinearity and contact with friction.
In Section 5.1, a general nonlinear setting is sketched, as well as an iterative solution scheme using the
Numerical tests for the nonlinear setting
In this section, we show some numerical results with two different three-dimensional geometries.
Local time subcycling
The domain decomposition approach introduced in the previous sections has been motivated by the fact that one is interested in a detailed resolution of the solution within the fine patch. Hence, in addition to the different spatial mesh sizes, it is desirable to use different time scales for the coarse and the fine problems, which is the topic of the current section. In Section 7.1, we present a mortar coupled solution with different time scales and analyse its asymptotical convergence order
Numerical results for time subcycling
In this section, we illustrate the performance of Algorithm 3 applied to two different two-dimensional geometries.
Conclusion
In this paper, we have developed an efficient iterative algorithm for the solution of frictional dynamic contact problems with locally refined geometries. For this, we have employed an overlapping domain decomposition with an independent fine mesh around the contact zone. For linear problems, we have shown that under some reasonable assumptions, the convergence rate of the corresponding algorithm is bounded independently of the material and discretization parameters. Further, we have
References (59)
- et al.
A primal-dual active set strategy for non–linear multibody contact problems
Comput. Methods Appl. Mech. Engrg.
(2005) - et al.
A mixed formulation for frictional contact problems prone to Newton like solution methods
Comput. Methods Appl. Mech. Engrg.
(1991) Exact energy and momentum conserving algorithms for general models in nonlinear elasticity
Comp. Methods Appl. Mech. Engrg.
(2000)- et al.
Energy-controlling time integration methods for nonlinear elastodynamics and low-velocity impact
Comput. Methods Appl. Mech. Engrg.
(2006) - et al.
Mass redistribution method for finite element contact problems in elastodynamics
Eur. J. Mech. A/Solids
(2008) A study of the stability of subcycling algorithms in structural dynamics
Comput. Methods Appl. Mech. Engrg.
(1998)- et al.
A numerical scheme to couple subdomains with different time-steps for predominantly linear transient analysis
Comput. Methods Appl. Mech. Engrg.
(2002) - et al.
Stable Lagrange multipliers for quadrilateral meshes of curved interfaces in 3D
Comput. Methods Appl. Mech. Engrg.
(2007) - et al.
Two-scale Dirichlet–Neumann preconditioners for elastic problems with boundary refinements
Comput. Methods Appl. Mech. Engrg.
(2007) Numerical methods for nonlinear three-dimensional elasticity
Nonlinear complementarity functions for plasticity problems with frictional contact
Comput. Methods Appl. Mech. Engrg.
The mortar finite element method with Lagrange multipliers
Numer. Math.
A new nonconforming approach to domain decomposition: The mortar element method
A mortar finite element method using dual spaces for the Lagrange multiplier
SIAM J. Numer. Anal.
Domain Decomposition Methods for Partial Differential Equations
Domain decomposition
Domain Decomposition Methods – Algorithms and Theory
Iterative methods for the solution of elliptic problems on regions partitioned into substructures
SIAM J. Numer. Anal.
Balancing Neumann–Neumann preconditioners for mixed approximations of heterogeneous problems in linear elasticity
Numer. Math.
FETI and Neumann–Neumann iterative substructuring methods: Connections and new results
Comm. Pure Appl. Math.
Unilateral Contact Problems: Variational Methods and Existence Theorems
Computational Contact and Impact Mechanics
Computational Contact Mechanics
Variationally consistent discretization schemes and numerical algorithms for contact problems
Acta Numer.
The primal–dual active set strategy as a semi-smooth Newton method
SIAM J. Optim.
A primal–dual active set algorithm for three-dimensional contact problems with Coulomb friction
SIAM J. Sci. Comput.
Formulation and comparison of algorithms for frictional contact problems
Int. J. Numer. Methods Engrg.
Inexact Newton methods
SIAM J. Numer. Anal.
Cited by (17)
Time domain boundary elements for elastodynamic contact
2023, Computer Methods in Applied Mechanics and EngineeringSpace–time adaptive finite elements for nonlocal parabolic variational inequalities
2019, Computer Methods in Applied Mechanics and EngineeringCitation Excerpt :Of particular current interest in numerical analysis have been dynamic contact problems for time-dependent equations, including adaptive mesh refinements [21,22], high-order [23] and Nitsche methods [24,25]. Their analysis is crucial for applications from tire dynamics [21,26] to blood flow in aortic valves [27]. This article considers the systematic error analysis of the four standard parabolic variational inequalities [17] associated to the fractional Laplacian as a nonlocal model operator: obstacle problems and friction, both in the domain and in the boundary.
Time domain boundary elements for dynamic contact problems
2018, Computer Methods in Applied Mechanics and EngineeringCitation Excerpt :While contact for time-dependent problems is of clear practical relevance, neither its analysis nor rigorous boundary element methods have been much explored. There is an extensive computational literature, including [4–9], but analytically even the existence of solutions to these free boundary problems is only known for flat contact area [10,11]. Some rigorous results have recently been obtained for Nitsche stabilized finite elements [12].
A coupling method for stochastic continuum models at different scales
2014, Probabilistic Engineering MechanicsCitation Excerpt :When this assumption is appropriate, the bulk of the model can be represented by a meso-scale model without degrading the evaluation of the quantity of interest, and the zone where the micro-scale is considered can be seen as a zoom in a global meso-scale model. This paper therefore stands in the line of classical adaptive mesh and local enrichment techniques [39–42], global–local iterative methods [43–45] and bridging methods [46–49], widely developed for deterministic applications. In particular, we construct our coupling technique in the Arlequin framework [46], which is a bridging method.
A multiscale method with patch for the solution of stochastic partial differential equations with localized uncertainties
2013, Computer Methods in Applied Mechanics and EngineeringCitation Excerpt :For non-conforming discretizations, the spectral content of A should depend on the meshes, and in particular on the ratio between coarse and fine mesh sizes. In the deterministic setting, bounds for the condition number of A could be obtained following the proof of Theorem 2 in [23]. These results could then be easily extended to the present stochastic framework when the same stochastic approximation spaces are used for the global and local solutions.