Meshless animation of fracturing solids

Mark Pauly, Richard Keiser, Bart Adams, Philip Dutré, Markus H. Gross, Leonidas J. GuibasACM Trans. Graphics 2005 · DOI

TL;DR

This paper introduces a novel meshless animation framework for simulating the fracture of elastic and plastic materials. By decoupling simulation nodes from explicit surface meshes and dynamically adapting shape functions, it achieves more realistic and controllable fracture patterns with greater efficiency than traditional methods.

What Problem Does This Paper Address?

Simulating the realistic fracture of solid objects in computer graphics is a notoriously difficult problem. Traditional methods, often based on Finite Element Methods (FEM), rely on maintaining a complex, conforming mesh that represents the object's geometry. When an object fractures, this mesh must be dynamically cut and re-meshed to accurately represent the new surfaces and topology. This process is computationally expensive, prone to stability issues, and can lead to undesirable artifacts like jagged fracture lines or "button-popping." Furthermore, manually specifying all the parameters for complex fracture events is often infeasible for animators. This paper aims to overcome these limitations by using a meshless approach that avoids the complexities of remeshing and offers more intuitive control over fracture behavior.

Methods

The core of this work lies in its meshless simulation framework. Instead of relying on a fixed mesh, the simulation uses a set of discrete simulation nodes scattered throughout the object's volume.

  • Meshless Discretization: The paper employs a point-based approach, similar to Smoothed Particle Hydrodynamics (SPH) or particle-based methods, to represent the object's volume. Shape functions are constructed using techniques like Moving Least Squares (MLS) approximation, which depend only on the spatial distribution of nodes, not a predefined connectivity.
  • Dynamic Shape Function Adaptation: A key innovation is how the method handles discontinuities introduced by cracks. It uses a transparency method to adapt shape functions. When a ray between two simulation nodes intersects a crack surface, the interaction between these nodes is modulated based on their distance to the crack front. This ensures smooth displacement fields across crack fronts without the need for remeshing.
  • Explicit Fracture Surface Modeling: Fracture surfaces are explicitly modeled using elliptical splats called surfels. These surfels are dynamically created and updated as cracks propagate. This allows for fine geometric detail, independent of the simulation node resolution.
  • Crack Propagation and Topology Control: Crack initiation is driven by stress thresholds. Propagation is controlled by stress tensor eigenvalues and material properties. The framework handles complex topological events like splitting, merging, and termination of crack fronts using a small set of operations.
  • Dynamic Re-sampling: To maintain simulation stability and accuracy, the density of simulation nodes is dynamically adjusted. Nodes are added in areas of high stress concentration (near crack fronts) or where strong deformations occur, ensuring adequate sampling resolution.
  • Snapping: A snapping technique is used to prevent the creation of overly small fragments, ensuring that any features are of a size comparable to the local node spacing.

Datasets

Not discussed in the paper. The paper focuses on a simulation framework and demonstrates its capabilities with various synthetic examples.

Key Insights and Ideas

  • Decoupling Simulation and Geometry: The most significant insight is that by using a meshless approach for simulation and an explicit, surfel-based representation for fracture surfaces, the simulation domain can be decoupled from the geometric representation. This dramatically simplifies handling topological changes and allows for higher geometric detail without proportionally increasing simulation complexity.
  • Dynamic Shape Functions for Cracks: The transparency method for adapting shape functions is crucial. It elegantly handles the discontinuities introduced by cracks by modulating node interactions, providing a smooth and stable simulation without explicit remeshing.
  • Surfels for Detailed Fracture Surfaces: Using surfels to represent fracture surfaces allows for the creation of highly detailed and visually rich crack patterns, independent of the underlying simulation grid. This is a major advantage for visual fidelity in animation.
  • Topological Primitives for Fracture: Identifying splitting, merging, initiation, and termination as fundamental topological events simplifies the management of complex fracture scenarios.

Key Results

The paper demonstrates the framework's ability to simulate a wide range of fracture behaviors, from brittle fracture of stiff objects to ductile deformation and fracture of highly plastic materials.

  • Efficiency: The meshless approach, combined with dynamic re-sampling and localized updates of transparency weights, leads to efficient simulations. The paper reports average simulation times of around 22 seconds per frame for complex brittle fracture scenarios.
  • Detail and Realism: The method produces highly detailed fracture surfaces and complex crack patterns, as evidenced by the visual examples. The decoupling of simulation resolution from surface detail allows for significantly higher surface detail than comparable FEM approaches.
  • Control: The framework offers explicit control mechanisms, such as crack painting and propagation history, allowing animators to guide fracture behavior for artistic purposes.

Limitations and Future Work

  • Small Fragments and Stability: The authors acknowledge that for very stiff objects that fracture into numerous small fragments, a sufficiently dense sampling of these fragments is required for stable shape function evaluation. This can inflate the number of simulation nodes and slow down computations. For such cases, approximating small fragments as rigid bodies is suggested.
  • Essential Boundary Conditions: Enforcing essential boundary conditions (e.g., fixed points) in meshless methods can be computationally more involved than in FEM. While the current system uses penalty forces for collision resolution, more advanced techniques like Lagrange multipliers or the Nitsche method are mentioned as potential future integrations, albeit with higher computational costs.
  • Self-Intersections: The penalty force approach for collision resolution does not guarantee that the simulation is entirely free of self-intersections, especially with highly deformable bodies.
  • Future Work: The authors suggest several promising directions:
  • Hierarchical Schemes: Extending the method to a hierarchical scheme for more efficient simulation of phenomena at different scales.
  • Data Structures: Improving data structures for better trade-offs between local caching and dynamic re-computation.
  • Advanced Constitutive Models: Incorporating more complex material models.
  • Thin Shells and Fluids: Extending the framework to handle thin shells and coupling with fluid simulations.

Summary updated

BibTeX
@article{pauly2005meshless,
  author = {Mark Pauly and Richard Keiser and Bart Adams and Philip Dutré and Markus H. Gross and Leonidas J. Guibas},
  title = {{Meshless animation of fracturing solids}},
  journal = {ACM Trans. Graphics},
  year = {2005},
  doi = {10.1145/1073204.1073296},
}
Figure