Implicit Bonded Discrete Element Method with Manifold Optimization

Jia-Ming Lu, Geng-Chen Cao, Chenfeng Li, Shi-Min HuACM Trans. Graphics 2025 · DOI

TL;DR

This paper introduces a novel simulation method that combines an implicit integration scheme with the Bonded Discrete Element Method (BDEM) and manifold optimization. This approach significantly accelerates fracture simulations, achieving speedups of 2.1 to 9.8 times over explicit methods, while maintaining or improving accuracy and stability, making it ideal for complex dynamic fracture scenarios.

What Problem Does This Paper Address?

Simulating fracture in materials is a computationally intensive task, especially when aiming for realistic visual effects. Traditional methods like the Finite Element Method (FEM) and Material Point Method (MPM) struggle with capturing the intricate details of fracture propagation and can be computationally expensive. The Bonded Discrete Element Method (BDEM) offers a more physically grounded approach to fracture, but its explicit integration schemes require very small time steps, limiting its practical application for complex, dynamic scenarios. Furthermore, solving the stiff constraints and rotational dynamics inherent in BDEM, particularly with quaternions, presents a significant computational challenge. This paper aims to bridge this gap by developing an implicit BDEM that is both stable and efficient, capable of handling large time steps and complex deformations.

Methods

The core of this work lies in two main innovations:

  • Optimization-Based Implicit Integrator for BDEM: Instead of relying on explicit time integration, the authors reformulate the BDEM simulation as a constrained optimization problem. This is achieved by deriving a constant mass matrix formulation for spherical particles and using a discrete variational principle. This approach allows for larger time steps, leading to significant speedups.
  • Manifold Optimization: To efficiently solve the quaternion-constrained system (which represents rotations), the authors employ manifold optimization. This technique transforms the constrained optimization problem into an unconstrained one on a specific manifold (a hypersphere for quaternions). They present both first-order and second-order manifold optimization algorithms, with the latter leveraging the Hessian for faster convergence. A nullspace operator is also introduced to further reduce the number of unknowns.

The paper also details how standard fracture criteria are adapted for the implicit setting and discusses collision handling using a simple potential model.

Datasets

Not discussed in the paper.

Key Insights and Ideas

  • Implicit BDEM is Achievable and Beneficial: The paper demonstrates that it's possible to create a stable and efficient implicit integration scheme for BDEM, overcoming the limitations of explicit methods that require tiny time steps.
  • Manifold Optimization for Quaternions is Key: The use of manifold optimization is crucial for efficiently handling the rotational dynamics represented by quaternions. It allows for faster convergence and more stable solutions compared to traditional constrained optimization techniques like penalty or Lagrange multipliers.
  • Trade-off for Speed: The authors introduce an approximation in their integrator that reduces its order of accuracy but significantly boosts computational efficiency, highlighting a practical trade-off for real-time applications.
  • Scale Consistency: The proposed method exhibits superior scale consistency in fracture simulations compared to FEM and MPM, meaning fracture behavior remains predictable across different simulation scales. This is a significant advantage for artists and designers.
  • Material Parameters Drive Fracture: Unlike some continuum methods that require additional artificial rules for fracture, this BDEM approach relies solely on material parameters, leading to more physically grounded and authentic fragmentation.

Key Results

  • Significant Speedup: The proposed implicit BDEM with manifold optimization achieves computational speedups of 2.1 to 9.8 times over explicit BDEM methods across various scenarios (e.g., beam stretch, beam drape, chocolate falling, plate impact).
  • Improved Stability and Accuracy: The implicit nature and optimization-based approach lead to more stable simulations, allowing for larger time steps without sacrificing accuracy.
  • Realistic Fracture: The method captures fine details of fracture, including extreme deformations and material failure, as demonstrated in the ceramic plate and fabric tearing examples.
  • Competitive Performance: Compared to FEM and MPM, the method shows better scale consistency and more realistic collision effects in fragmentation.

Limitations and Future Work

  • Collision Handling Simplification: The paper acknowledges using a rudimentary collision potential for simplicity, which might limit the fidelity of collision effects. Future work will involve integrating more sophisticated collision models.
  • GPU Acceleration: While the BDEM system is amenable to parallelization, the current exposition focuses on a clear computational comparison without GPU acceleration. The authors plan to explore GPU acceleration for further performance gains.
  • Fracture Surface Reconstruction: The reconstruction and rendering of fracture surfaces play a role in the final visual outcome. The authors aim to explore fracture surface reconstruction methods tailored to BDEM.

Summary updated

BibTeX
@article{lu2025implicit,
  author = {Jia-Ming Lu and Geng-Chen Cao and Chenfeng Li and Shi-Min Hu},
  title = {{Implicit Bonded Discrete Element Method with Manifold Optimization}},
  journal = {ACM Trans. Graphics},
  year = {2025},
  doi = {10.1145/3711852},
}
Figure