Multi-material swept face remapping on polyhedral meshes
Introduction
Hydrodynamic simulations of multi-material flows are important in many areas of research. For example, they are crucial for the design of Inertial Confinement Fusion (ICF) targets. Outer shells of these targets consist of several layers of different materials with different physical properties. In many cases, it is practical to assume that these materials are immiscible, meaning that each material occupies its own spatial region which is a subset of the entire computational domain. These regions are naturally moving along with the flow which makes it suitable for a Lagrangian framework, where an underlying, interface-aligned computational grid evolves with the flow. However, vorticial components of the flow can result in an unacceptable tangling of the grid.
In Arbitrary Lagrangian-Eulerian (ALE) methods [1], mesh untangling and smoothing called rezoning is employed to restore the quality of the computational grid to acceptable levels. Rezoning requires simulation fields to be accurately and conservatively transferred from the original Lagrangian grid to the new rezoned grid. A complication is that some cells of the rezoned mesh now overlap cells containing different materials from the Lagrangian mesh, creating new multi-material cells, even if the original Lagrangian mesh cells were all single material. Rather than subdivide the rezoned mesh along material boundaries which could potentially lead to poor quality cells again, one can choose to represent affected cells as multi-material or mixed cells and only track the volume fractions3 in those cells for carrying forward the Lagrangian simulation. However, when it is time to perform another rezone/remap step, one must contend with the scenario where a cell of the rezoned mesh overlaps one or more multi-material cells. Simply transferring quantities into this rezoned cell in proportion to volume fractions in the overlapped Lagrangian cells is inaccurate and diffusive resulting in smearing of material boundaries. A more accurate approach is to reconstruct the pure material sub-cells of a multi-material cell using a material interface reconstruction algorithm [3], [2], [4] and computing the overlap of the rezoned cell with these sub-cells rather than the full cell. Compared to single-material methods, multi-material remapping requires conservation of material volumes in addition to the standard requirements of mass, momentum, total energy, etc. conservation. All aspects of ALE methods are discussed in the seminal article [5], e.g.
In this paper, we focus only on a particular aspect of an ALE method: conservative interpolation or remapping of a single scalar quantity, such as the material density, in a multi-material flow simulation. More specifically, we assume that this quantity is adequately defined by its mean density and the volume which it occupies inside a polytopal computational cell. This is a common representation in both cell-centered [6] and staggered [7] ALE methods. Accurate multi-material remapping methods [8] exist and are based on the computation of exact intersections between reconstructed material polytopes on a Lagrangian mesh with the cells of the rezoned mesh. However, geometric calculation of these intersections is computationally expensive and this cost can easily dominate the whole ALE method. Therefore, many codes [9], [10], [11], [12] rely on simplified algorithms usually formulated using fluxes through faces of neighboring cells, and are aptly named flux-based, swept-face, or just swept remap. The flux formulation requires the displacement between the meshes to be limited. In particular for the presented method, the volume defined by a face displacement between the two meshes has to be smaller that the volume of a particular cell-face pyramid (3) in the contributing multi-material cell on the Lagrangian mesh, see (31) for details.
A second order, single material swept remap using piece-wise linear density reconstruction was introduced in [13]. Compared to a second-order intersection remap, the single material swept remap generally shows comparable errors while it can be more sensitive to some pathological mesh displacements (like an hour-glass distortion) and to the mesh orientation with respect to the underlying density field [14]. Possible combination of a swept remap with an intersection remap is investigated in [15]. For multi-material problems, hybrid algorithms [16], [17], [18] combine an intersection remap around material interfaces with a swept remap in pure material regions. The history of advection-like remapping methods for ALE simulations is summarized in [9]. One approximate advection based multi-material method [10] relies on additional corrections to ensure that computed material volumes remain positive and that they sum up to the bulk volume. Another method [19] is based on a simple assumption about a layered material layout oriented (either parallel or perpendicular to a face) in order to determine which materials should be fluxed through the face. Advection multi-material remap can be also implemented [11] in an operator split fashion dimension-by-dimension.
A multi-material swept-face remap using a Piece-wise Linear Interface Calculation (PLIC) [3] is used in the LANL code FLAG [20], [21]. Here we present the algorithm in 2D/3D and compare it to an accurate intersection remap. In 1D, swept-face remap is identical to intersection remap. Both remapping methods are implemented in an open source library called Portage [22].
This article is organized as follows: mesh notation, field description (for single and multi-material cases), and field reconstruction methods are provided in Section 2. In Sections 3 and 4, we provide a detailed description of the intersection and swept face remapping methods. In Section 5 we provide a brief overview of the various software libraries implementing the methods described and used for the numerical experiments. Finally, in Section 6, we present results for various comparisons between these two methods followed by discussions and conclusions in Section 7.
Section snippets
Mesh notation
We assume a computational domain , and two sets of meshes dividing the domain into star-convex4 polytopes on a source mesh and the same number of convex polytopes on a target mesh , so that For the remainder of this paper, entities without tildes refer to the source and entities with tildes refer to the target. In addition, we assume the
Intersection remap scheme
The mesh intersection based remap scheme involves finding the exact amount of each donor cell that contributes to a target cell, and is considered the most accurate remapping method. It does not impose any requirements on the source and target mesh cell types, and is applicable to general polyhedral meshes. However, in order to be consistent with the description of swept-face remap, we assume that the source and target meshes have the same number of nodes, and the same connectivity. The
Swept-face remap scheme
For intersection based remapping, we need to compute the exact intersections between polytopes from source and target meshes. This process can become computationally costly, particularly for 3D meshes, where the cells may not be convex, necessitating their decomposition into tetrahedral elements (per cell) which are used to compute the intersections. For the cases, like in ALE methods, where the target mesh is a slightly moved source mesh, one can reformulate the problem of remapping using
Software libraries
Portage The algorithms described above are implemented in an open source C++ library called Portage [22]. Portage is a scalable, extensible remap library supporting state-of-the-art remap schemes for polyhedral meshes and particles in 2D and 3D up to second-order accuracy. The algorithms implemented within Portage ensure critical properties such as local/global conservation and bounds preservation for mesh remap as well as supporting single and multi-material field remapping. The library
Numerical results
We present 4 test cases in this section. As a sanity check, we reproduce a cyclic remapping test case on quadrilateral cells from [17]. Then, we modify the original problem and introduce a per-material error measure in order to provide a deep analysis of a multi-material swept remap comparing it to intersection based remap. Next, we demonstrate the capability of Portage by remapping between two polyhedral meshes. Finally, we compare swept remapping to intersection remapping for a simple
Discussion and conclusions
We have presented a swept region multi-material remap. Compared to an accurate intersection remap, the main advantage of a swept remap is that it does not require computation of polytope/polytope intersections. However, the need of geometrical operations such as an intersection of a polytope with a half-space inside interface reconstruction itself makes the advantage less significant in regions with multi-material cells.
We have numerically demonstrated that a swept remap is second-order
CRediT authorship contribution statement
Jan Velechovsky: Formal analysis, Investigation, Software, Validation, Visualization, Writing – original draft. Evgeny Kikinzon: Software, Validation. Navamita Ray: Software, Writing – review & editing. Hoby Rakotoarivelo: Software. Angela Herring: Funding acquisition, Project administration, Supervision. Mack Kenamond: Conceptualization, Resources, Software, Writing – review & editing. Konstantin Lipnikov: Investigation, Software, Writing – review & editing. Mikhail Shashkov: Methodology,
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.
Acknowledgement
This work is supported by the U.S. Department of Energy for Los Alamos National Laboratory under contract 89233218CNA000001. We thank ASC NGC Ristra and Portage for support. We thank anonymous reviewers for their constructive comments. LA-UR-22-20333.
References (39)
- et al.
Volume of fluid (VOF) method for the dynamics of free boundaries
J. Comput. Phys.
(1981) - et al.
Reconstruction of multi-material interfaces from moment data
J. Comput. Phys.
(2008) - et al.
Establishing mesh topology in multi-material cells: enabling technology for robust and accurate multi-material simulations
Comput. Fluids
(2018) - et al.
Arbitrary Lagrangian–Eulerian methods for modeling high-speed compressible multimaterial flows
J. Comput. Phys.
(2016) - et al.
The construction of compatible hydrodynamics algorithms utilizing conservation of total energy
J. Comput. Phys.
(1998) - et al.
Conservative multi-material remap for staggered multi-material arbitrary Lagrangian–Eulerian methods
J. Comput. Phys.
(2014) Computational methods in Lagrangian and Eulerian hydrocodes
Comput. Methods Appl. Mech. Eng.
(1992)- et al.
3d cell-centered hydrodynamics with subscale closure model and multi-material remap
Comput. Fluids
(2020) - et al.
Second-order sign-preserving conservative interpolation (remapping) on general grids
J. Comput. Phys.
(2003) - et al.
Hybrid remap for multi-material ALE
Comput. Fluids
(2011)
Two-step hybrid conservative remapping for multimaterial arbitrary Lagrangian–Eulerian methods
J. Comput. Phys.
One-step hybrid remapping algorithm for multi-material arbitrary Lagrangian–Eulerian methods
J. Comput. Phys.
An arbitrary Lagrangian-Eulerian computing method for all flow speeds
J. Comput. Phys.
A high-order finite volume method for systems of conservation laws—multi-dimensional optimal order detection (mood)
J. Comput. Phys.
Fully multidimensional flux-corrected transport algorithms for fluids
J. Comput. Phys.
An efficient linearity-and-bound-preserving remapping method
J. Comput. Phys.
Conservative remapping of material-dependent fields between possibly misaligned material regions
J. Comput. Phys.
Optimization-based synchronized flux-corrected conservative interpolation (remapping) of mass and momentum for arbitrary Lagrangian–Eulerian methods
J. Comput. Phys.
An efficient linearity and bound preserving conservative interpolation (remapping) on polyhedral meshes
Comput. Fluids
Cited by (5)
Cell-centered indirect Arbitrary Lagrangian-Eulerian numerical strategy for solving 3D gas dynamics equations
2024, Journal of Computational PhysicsA localized subdomain smoothing MMALE particle method for efficient modeling FSI problems
2024, Computer Methods in Applied Mechanics and EngineeringRemapping Between Meshes with Isoparametric Cells: a Case Study
2023, Communications on Applied Mathematics and Computation