🔢 mathematics

A Hybridizable Discontinuous Galerkin Method for Wave Propagation in Elastic Beam Networks

This paper proposes and analyzes a hybridizable discontinuous Galerkin method combined with an energy-conservative implicit time discretization and a two-level overlapping additive Schwarz preconditioner to efficiently and accurately solve elastic wave propagation problems on networks, achieving optimal error estimates and uniform convergence while overcoming severe CFL restrictions and reducing the global system size to depend only on the number of network nodes.

Original authors: Moritz Hauck (Karlsruhe Institute of Technology), Joseph Holten (Karlsruhe Institute of Technology), Axel Målqvist (Chalmers University of Technology and University of Gothenburg), Andreas Rupp (Saarl

Published 2026-08-24

📖 6 min read🧠 Deep dive

CC BY 4.0

Original authors: Moritz Hauck (Karlsruhe Institute of Technology), Joseph Holten (Karlsruhe Institute of Technology), Axel Målqvist (Chalmers University of Technology and University of Gothenburg), Andreas Rupp (Saarland University), Lucia Swoboda (Chalmers University of Technology and University of Gothenburg)

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine a world built not of solid blocks, but of a vast, intricate web of thin, flexible threads. This is the reality of many materials we use every day, from the paper in a book to the cardboard in a shipping box. These materials are essentially networks of fibers, where countless tiny strands cross and bond to form a structure that is both light and surprisingly strong. When such a material is struck or shaken, waves of energy travel through this web, causing the fibers to vibrate, bend, and twist. Understanding how these waves move is crucial for engineers who design everything from medical stents to sound-dampening insulation. However, simulating this behavior on a computer is notoriously difficult. The fibers are so thin and numerous that trying to model every single microscopic detail in three dimensions would overwhelm even the most powerful supercomputers. Instead, scientists often treat the material as a simplified map of lines and junctions, a network where the physics happens along the lines and the connections happen at the points where they meet.

The challenge lies in the sheer variety of the fiber lengths within these networks. Some segments are microscopic, while others stretch much longer. This vast difference in scale creates a mathematical nightmare for standard computer simulations. If a researcher tries to calculate the movement of the waves step-by-step in time, the computer is forced to take incredibly tiny steps to keep the short fibers stable. This requirement, known as a time-step restriction, means that simulating even a fraction of a second of vibration could take years of computing time. To solve this, a team of researchers has developed a new mathematical approach that bypasses these bottlenecks, allowing them to simulate elastic waves moving through complex fiber networks with unprecedented speed and accuracy.

The researchers focused on a specific type of physics model called the Timoshenko beam theory. Unlike simpler models that treat fibers as perfectly rigid rods, this theory accounts for the fact that real fibers can shear and rotate as they bend, which is essential for accurately capturing the behavior of short, thick fibers found in materials like paper. To handle the complexity of the network, the team employed a technique known as the hybridizable discontinuous Galerkin method. In plain terms, this is a way of breaking the problem into small, manageable pieces along each fiber while keeping the connections between them flexible. The brilliance of their approach is that it allows the computer to solve the messy details of each fiber segment locally and independently. Once those local details are resolved, the computer can condense the entire problem down to a much smaller system that only involves the junction points where the fibers meet. This reduction is powerful because the size of the final calculation depends only on the number of junctions, not on how finely the fibers were divided or how complex the math used to describe them.

To ensure the simulation remains stable over time, the team combined this spatial method with a specific time-stepping strategy that conserves energy. In a physical system like a vibrating sheet of paper, energy is neither created nor destroyed; it simply shifts between the motion of the fibers and the tension within them. The researchers' method preserves this balance perfectly in the computer model. This is a critical feature because it prevents the simulation from drifting into nonsense or blowing up numerically, which often happens when trying to simulate waves over long periods. By using an implicit method, they avoided the severe time-step restrictions that plague older techniques, allowing them to take larger, more efficient steps through time without losing accuracy.

However, solving the resulting equations still presented a hurdle. The mathematical systems generated by these networks are often "ill-conditioned," meaning they are extremely sensitive to small errors and can be very slow to solve. Standard tools used to speed up these calculations often fail because they do not understand the unique, web-like structure of the material. To overcome this, the team introduced a specialized pre-solver, a mathematical tool that prepares the system for a quick solution. They treated the network as if it were a continuous object at a larger scale, overlaying a coarse grid on top of the fine network. This allowed them to use a two-level strategy: one level to handle the large-scale behavior of the web and another to fix the local details. This approach proved remarkably robust, keeping the number of calculation steps low regardless of how small the time steps were or how large the network became.

The team tested their method with a series of rigorous experiments. First, they ran simulations on a simple, cross-shaped network where they knew the exact answer in advance. The results matched the theoretical predictions perfectly, showing that the method converges at the fastest possible rate as the mesh is refined. They then moved to a more realistic scenario, simulating a piece of paper roughly eight millimeters square, containing nearly two million nodes and over three million fiber segments. Even with this massive complexity, the method held up, with the number of calculation steps remaining steady as they changed the size of the sub-regions or the time steps.

Finally, they tackled an industrially relevant problem: a twenty-millimeter square sheet of fiber network with over ten million nodes and nearly fifteen million edges. This simulation required tracking the movement of a wave generated by a tap at the center of the sheet. The system contained over sixty million unknowns and required significant computing power, yet the method solved it efficiently. The simulation ran for about five hours on a high-performance computer cluster, tracking the wave as it traveled nearly ten millimeters across the sheet, reflecting the anisotropic nature of the fiber orientation. The energy in the system remained conserved throughout the run, with only a tiny, expected drift caused by the limits of computer precision. The results confirmed that the method is not just a theoretical curiosity but a practical tool capable of handling the scale of real-world materials.

This work represents a significant step forward in the computational modeling of fiber-based materials. By combining a clever spatial reduction technique with a stable time-stepping scheme and a robust pre-solver, the researchers have provided a way to simulate wave propagation in complex networks that was previously out of reach. Their findings suggest that it is now possible to model the dynamic behavior of materials like paper and cardboard with a level of detail and efficiency that could eventually lead to better-designed products and a deeper understanding of how these ubiquitous materials respond to stress and impact. The method stands as a testament to how mathematical innovation can unlock the secrets of the complex, interconnected structures that surround us.

Technical Summary: A Hybridizable Discontinuous Galerkin Method for Wave Propagation in Elastic Beam Networks

Problem Statement
This paper addresses the numerical simulation of elastic wave propagation in fiber-based materials, such as paper and cardboard. These materials are modeled as spatial networks consisting of slender, one-dimensional components (edges) connected at junctions (nodes). The physical behavior of individual fibers is described by dynamic Timoshenko beam equations, which account for shear deformation and cross-sectional rotation, making them suitable for both slender and thick beams. The challenge lies in the geometric complexity of the network and the wide disparity in fiber segment lengths, which often renders standard explicit time-stepping schemes impractical due to severe Courant–Friedrichs–Lewy (CFL) restrictions. Furthermore, the resulting linear systems are typically ill-conditioned, posing difficulties for iterative solvers.

Methodology
The authors propose a Hybridizable Discontinuous Galerkin (HDG) method combined with an energy-conservative implicit time discretization.

Spatial Discretization (HDG):
- The method employs a hybrid dual mixed formulation where the primal variables (displacement and rotation) and dual variables (force and moment resultants) are approximated by piecewise polynomials of degree $p$ on each edge.
- A key feature of HDG is the introduction of hybrid unknowns defined only on the network nodes. This allows for the static condensation of the interior edge degrees of freedom.
- At each time step, the global system is reduced to a linear system defined solely on the network nodes. Crucially, the size of this condensed system depends only on the number of nodes and is independent of the polynomial degree $p$ used for the edge discretization.
- A stabilization parameter $\tau$ is introduced to penalize the jump between the primal trace and the hybrid unknown, ensuring stability and trace control.
Time Discretization:
- To avoid the CFL restrictions associated with explicit schemes (which would require time steps proportional to the smallest fiber segment), the authors employ an implicit time-stepping scheme, specifically the Crank–Nicolson method.
- This choice is motivated by the large variation in fiber lengths and ensures energy conservation for the semi-discrete and fully discrete systems in the absence of external forces and time-dependent boundary data.
Preconditioning:
- To solve the resulting ill-conditioned global condensed system efficiently, a two-level overlapping additive Schwarz preconditioner is introduced.
- This approach constructs an artificial coarse mesh over the domain containing the network. The preconditioner utilizes a coarse space based on continuous piecewise trilinear functions on this artificial mesh and local subspaces associated with the mesh elements.
- The method relies on the observation that the network behaves as a continuous object at sufficiently coarse scales, allowing the transfer of classical domain decomposition analysis techniques to the network setting.

Key Contributions and Theoretical Results

A Priori Error Analysis: The paper establishes optimal-order a priori error estimates for both spatial and temporal discretizations.
- For the spatial discretization, the error converges with order $O(h^{p+1-|s|})$ , where $h$ is the mesh size and $s$ relates to the scaling of the stabilization parameter $\tau$ . For the standard choice $\tau \sim 1$ , the optimal rate $O(h^{p+1})$ is achieved.
- For the temporal discretization using Crank–Nicolson, the method achieves second-order convergence $O(\Delta t^2)$ .
- The analysis combines projection-based techniques from HDG wave equation literature with arguments tailored to the Timoshenko beam network setting.
Energy Conservation: The formulation is proven to be energy-conservative. The semi-discrete and fully discrete schemes preserve the total energy of the system when external forces and time-dependent boundary data are absent.
Preconditioner Convergence: The authors prove the uniform convergence of the preconditioned conjugate gradient (PCG) method. This is achieved by establishing a spectral equivalence between the condensed HDG operator and a combination of weighted graph mass and Laplacian operators. Under assumptions of network homogeneity and connectivity on coarse scales, the condition number of the preconditioned system is shown to be bounded independently of the mesh size and time step.

Numerical Results
Numerical experiments validate the theoretical findings:

Convergence: Experiments on a coordinate cross network confirm the predicted optimal convergence rates in both space and time. The results also indicate superconvergence for the hybrid variable in specific norms.
Preconditioner Robustness: Tests on a realistic paper network model (approx. 1.7 million nodes) demonstrate that the iteration counts of the PCG method remain nearly uniform with respect to subdomain size and time step size, confirming the theoretical robustness.
Scalability: A large-scale simulation on a 20 mm $\times$ 20 mm fiber network (approx. 10.2 million nodes, 14.6 million edges) with 61 million unknowns in the trace system was successfully performed. The solver required an average of 44 iterations per time step and ran efficiently on 128 MPI ranks, demonstrating scalability to industrially relevant problem sizes.

Significance
The paper claims that the proposed method offers a robust and efficient framework for simulating wave propagation in complex fiber networks. By exploiting the zero-dimensional nature of network junctions, the HDG approach significantly reduces the global problem size at each time step, decoupling it from the polynomial degree. The use of an implicit time scheme overcomes the limitations of explicit methods in networks with highly varying segment lengths. Furthermore, the introduction of a specialized two-level preconditioner addresses the ill-conditioning inherent in these problems, enabling the solution of large-scale systems that were previously computationally prohibitive. The work provides a mathematically rigorous foundation for modeling the dynamics of fiber-based materials, with potential applications in understanding the mechanical behavior of paper and similar composites.

More like this