A novel approach to rigid spheroid models in viscous flows using operator splitting methods

Benjamin K. Tapley, Elena Celledoni, Brynjulf Owren, Helge I. AnderssonNumer. Algorithms 2019 · DOI

Introduction

Simulating the dynamics of non-spherical particles in viscous flows is critical for understanding phenomena ranging from plankton transport in oceans to fiber orientation in industrial manufacturing. However, modeling a rigid spheroid involves solving a complex system of 13 coupled ordinary differential equations (ODEs) that track position, velocity, angular momentum, and orientation (typically via quaternions). Standard numerical integration techniques, such as Runge-Kutta or Adams-Bashforth methods, often struggle with these systems due to stability issues in "stiff" regimes (where fluid forces are dominant) and a failure to preserve the underlying geometric and physical properties of the motion.

Coordinate systems for a rigid spheroid in a viscous flow.

Methodology: The Strang Splitting Approach

To overcome these challenges, this research introduces a second-order symmetric operator splitting method, specifically a Strang splitting scheme. The core innovation lies in decoupling the particle's motion into two distinct sub-systems that can both be solved analytically:

  1. Sub-system 1 ($f_1$): Describes a free rigid body moving under gravity. This is a conservative, Hamiltonian system. The authors utilize exact solutions for axisymmetric rigid bodies to evolve this step.
  2. Sub-system 2 ($f_2$): Describes the dissipative effects of the fluid, including Stokes drag and torques.

By using the "variation of constants" formula, the authors derive analytical solutions for the momentum decay caused by fluid resistance. This analytical approach ensures that the only numerical error introduced is due to the splitting itself, rather than the approximation of individual steps. The complexity of these analytical solutions depends on the spheroid's aspect ratio $\lambda$, as shown in the table of coefficients below.

Analytical coefficients for the dissipative sub-system based on the spheroid aspect ratio λ.

Numerical Benchmarking

The method was evaluated using two complex fluid environments: the Ethier-Steinman Flow (a 3D analytic solution to the Navier-Stokes equations) and an Oscillating Shear Flow. Performance was compared against standard second-order Adams-Bashforth (AB2) and fourth-order Runge-Kutta (RK4) solvers.

Key Findings

  • Superior Accuracy and Efficiency: In systems where fluid forces act as a small perturbation, the splitting method (SP2) achieves an error proportional to $O(\epsilon h^2)$, making it significantly more accurate than AB2. For a fixed computational budget, SP2 produced solutions with over ten times less error than traditional methods.
  • Robust Stability: In "stiff" regimes where standard solvers like AB2 often diverge or "blow up" unless extremely small time-steps are used, the splitting method remained perfectly stable.
  • Order Reduction: The study identifies a theoretical limitation known as "order reduction." In extremely stiff conditions, the accuracy of the splitting method may drop from second-order to first-order. The authors successfully visualized this transition, providing a clear roadmap for when higher-resolution steps are required.

Conclusion

The proposed operator splitting method provides a robust, efficient, and physically consistent framework for simulating rigid spheroids. By leveraging exact solutions for the split sub-systems, the method preserves the qualitative dissipative behavior of the fluid-particle interaction better than general-purpose ODE solvers. Future work aims to extend this approach to triaxial ellipsoids and integrate it into large-scale computational fluid dynamics (CFD) simulations where fluid velocity data is provided on discrete grids.

Summary updated

BibTeX
@article{tapley2019novel,
  author = {Benjamin K. Tapley and Elena Celledoni and Brynjulf Owren and Helge I. Andersson},
  title = {{A novel approach to rigid spheroid models in viscous flows using operator splitting methods}},
  journal = {Numer. Algorithms},
  year = {2019},
  doi = {10.1007/s11075-019-00666-1},
}
Figure