Constraint Penalization Method in the Lattice Boltzmann Method (LBM) for Fluid–Structure Interaction
Tristan Millet, Erwan Liberge · 2025
Read Constraint Penalization Method in the Lattice Boltzmann Method (LBM) for Fluid–Structure Interaction by Tristan Millet, Erwan Liberge (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2511-01629.txt.
Introduction
Tristan Millet and Erwan Liberge simulate fluids flowing around rigid bodies with the lattice Boltzmann method, a scheme that moves particle distributions across a grid. The hard part is forcing the fluid inside a solid region to move exactly with the body, without tracking the interface with extra machinery.
Their constraint penalization method extends the fictitious domain idea: a penalty term applied directly to the fluid velocity field enforces rigid-body motion, so no Lagrange multipliers or explicit interface forces are needed. The lattice Boltzmann algorithm keeps its locality and simplicity, and the fluid-solid coupling becomes implicit. Numerical experiments reproduce rigid-body motion and hydrodynamic interactions accurately at minimal extra cost, demonstrated on particle sedimentation from a simple case upward in complexity.
Catalog record
| CREATOR | Tristan Millet, Erwan Liberge |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 26 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2511-01629
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Constraint Penalization Method in the Lattice Boltzmann Method (LBM) for Fluid–Structure Interaction Tristan Millet, Erwan Liberge Abstract A constraint penalization method is introduced within the Lattice Boltzmann (LBM) framework to model fluid–structure interactions involving rigid bodies. The proposed approach extends the fictitious domain concept by enforcing the rigid-body motion through a penalization term directly applied to the fluid velocity field, eliminating the need for explicit Lagrange multipliers or interface force computation. This formulation preserves the locality and simplicity of the LBM algorithm while ensuring an implicit coupling between the fluid and solid regions. Numerical experiments demonstrate that the method accurately reproduces rigid-body motion and hydrodynamic interactions with minimal additional computational cost. The method is applied to particle sedimentation, starting with a simple example and progressing to increasingly complex cases. keywords Lattice Boltzmann Method (LBM), Fluid Structure Interaction (FSI), Constraint Penalization, Fictitious Domain, particle sedimentation 1 Introduction The Lattice Boltzmann Method (LBM) is a mesoscopic numerical approach for simulating fluid flows. Unlike traditional computational fluid dynamics (CFD) techniques—such as finite difference or finite element methods—which discretize the Navier–Stokes equations directly, the LBM models the fluid as a collection of particle populations evolving on a discrete lattice in space, time, and velocity. These populations obey simplified versions of the Boltzmann equation that account for collision and streaming processes. Through a Chapman–Enskog expansion [ 1 ] , the macroscopic Navier–Stokes equations can be recovered, while retaining the advantages of a simple, local, and highly parallelizable algorithmic structure. Due to these properties, the LBM has been successfully applied to particle-laden flows and fluid–structure interaction (FSI) problems [ 2 , 3 , 4 , 5 ] . Several strategies have been developed to model the coupling between fluids and solids in LBM, including the momentum exchange (ME) method [ 6 ] , the immersed boundary method (IBM) [ 4 ] , and volume penalization method [ 5 ] . All of these approaches share a common feature: the explicit computation of fluid–solid forces, which requires the transfer of information between the fluid solver and the solid. This coupling can significantly slow down simulations and may introduce numerical errors. In this work, we focus on adapting the fictitious domain method initially proposed by Glowinski and Patankar [ 7 , 8 , 9 ] . The central idea of this approach is to treat the solid region as a fluid and enforce rigid-body motion through Lagrange multipliers. T. Coupez et al. [ 10 , 11 ] revisited this concept and compared the enforcement of rigid-body constraints using either Lagrange multipliers or a penalization approach. Their findings indicate that penalization alone is generally sufficient, and that Lagrange multipliers only provide a noticeable increase in accuracy for very fine meshes. To our knowledge, this penalization strategy has so far been implemented exclusively within the finite element framework [ 12 ] . The present study aims to extend this methodology to the LBM framework, offering a potentially efficient and accurate alternative for simulating rigid solids in fluid flows without the explicit computation of interface forces. The first part presents the theoretical context, from the idea at the macroscopic level to its application within the framework of the LBM. Next, five applications are proposed. The first is a simple case of a floating particle in a fluid flow, followed by examples dealing with variations in the shape, number, and density of particles in sedimentation. 2 Theoretical background 2.1 Monolithic Formulation for Fluid–Solid Interaction In the monolithic approach to fluid–structure interaction, both the fluid and the solid are treated within a single computational domain, allowing the enforcement of solid constraints without explicitly computing interface forces. In the fluid domain \Omega_{f} , the incompressible Navier–Stokes equations read \begin{cases}\rho_{f}\left(\dfrac{\partial\mathbf{u}_{f}}{\partial t}+\mathbf{u}_{f}\cdot\nabla\mathbf{u}_{f}\right)=\mathbf{f}_{f}+\nabla\cdot\bm{\sigma}_{f},\\[2.84526pt] \nabla\cdot\mathbf{u}_{f}=0,\end{cases} (1) with the Cauchy stress tensor decomposed as \bm{\sigma}_{f}=-p\,\mathbf{I}+\bm{\tau},\qquad\bm{\tau}=2\mu_{f}\mathbf{D}[\mathbf{u}_{f}],\qquad\mathbf{D}[\mathbf{u}_{f}]=\frac{1}{2}(\nabla\mathbf{u}_{f}+\nabla\mathbf{u}_{f}^{T}), (2) where \bm{\tau} is the viscous stress tensor. In the solid domain \Omega_{s} , rigid-body constraints are enforced either via a Lagrange multiplier \bm{\lambda} or through penalization of the constraint \mathbf{D}[\mathbf{u}_{s}]=0 : \begin{cases}\rho_{s}\left(\dfrac{\partial\mathbf{u}_{s}}{\partial t}+\mathbf{u}_{s}\cdot\nabla\mathbf{u}_{s}\right)=\mathbf{f}_{s}+\nabla\cdot\bm{\sigma}_{s},\\[2.84526pt] \nabla\cdot\mathbf{u}_{s}=0,\end{cases} (3) with \bm{\sigma}_{s}=-p\,\mathbf{I}+2\mu_{s}\mathbf{D}[\mathbf{u}_{s}]+\mathbf{D}[\bm{\lambda}], (4) where \mu_{s} is a penalization viscosity enforcing near-rigid behavior in \Omega_{s} , and \bm{\lambda} represents the Lagrange multiplier enforcing rigid-body motion. Combining the fluid and solid, the governing equations over the full domain \Omega=\Omega_{f}\cup\Omega_{s} can be written as [ 10 ] [ 8 ] \rho(\mathbf{x})\left(\frac{\partial\mathbf{u}}{\partial t}+\mathbf{u}\cdot\nabla\mathbf{u}\right)=\rho(\mathbf{x})\mathbf{g}+\nabla\cdot\bm{\sigma}(\mathbf{x}), (5) with \displaystyle\bm{\sigma}(\mathbf{x}) \displaystyle=\bm{\sigma}_{s}\,\chi(\mathbf{x})+\bm{\sigma}_{f}\,[1-\chi(\mathbf{x})], (6) \displaystyle\mu(\mathbf{x}) \displaystyle=\mu_{s}\,\chi(\mathbf{x})+\mu_{f}\,[1-\chi(\mathbf{x})], (7) \displaystyle\rho(\mathbf{x}) \displaystyle=\rho_{s}\,\chi(\mathbf{x})+\rho_{f}\,[1-\chi(\mathbf{x})], (8) where \chi is the indicator function of the solid domain: \chi(\mathbf{x})=\begin{cases}1,&\mathbf{x}\in\Omega_{s},\\ 0,&\mathbf{x}\in\Omega_{f}.\end{cases} (9) This monolithic formulation allows the treatment of fluid and solid within a single system, avoiding the explicit evaluation of fluid–solid interaction forces. Finally, following the observations of Coupez [ 10 ] , which indicate that Lagrange multipliers only significantly improve accuracy for very fine meshes, we opt for a constraint penalization approach, thus avoiding the additional computational expense of Uzawa iterations loop. 2.2 Incompressible Lattice Boltzmann Method: Velocity-Based Approach (D2Q9) To solve the incompressible Navier–Stokes equations, we employ a Lattice Boltzmann Equation (LBE) formulation based on the velocity field [ 13 , 14 ] . Unlike conventional momentum-based schemes, which evolve the macroscopic momentum \rho\mathbf{u} , the velocity-based formulation directly updates the velocity \mathbf{u} and the hydrodynamic pressure p . This approach allows for a monolithic treatment of regions with different densities, enabling the consistent handling of multi-density flows, which is challenging in conventional LBM formulations. The evolution of the discrete distribution function f_{\alpha}(\mathbf{x},t) , associated with the discrete velocity \mathbf{c}_{\alpha} , is governed by f_{\alpha}(\mathbf{x}+\mathbf{c}_{\alpha}\Delta t,t+\Delta t)-f_{\alpha}(\mathbf{x},t)=-\frac{1}{\tau_{f}}\left[f_{\alpha}(\mathbf{x},t)-f_{\alpha}^{eq}(\mathbf{x},t)\right]+F_{\alpha}\,\Delta t+S_{\alpha}\,\Delta t, (10) where \tau_{f} is the dimensionless relaxation time, \Delta t is the time step, and F_{\alpha} represents a forcing term, and S_{\alpha} a source term. The difference between F_{\alpha} and S_{\alpha} will be discussed later. For the two-dimensional nine-velocity (D2Q9) model, the discrete velocity set is illustrated in Fig. 1 . Each discrete velocity \mathbf{c}_{\alpha} is associated with a weight coefficient \omega_{\alpha} , defined as \omega_{\alpha}=\begin{cases}4/9,&\alpha=0,\\ 1/9,&\alpha=1,3,5,7,\\ 1/36,&\alpha=2,4,6,8,\end{cases} (11) and the lattice speed is given by c=\Delta x/\Delta t . \Delta x \Delta x \mathbf{c}_{1} \mathbf{c}_{2} \mathbf{c}_{3} \mathbf{c}_{4} \mathbf{c}_{5} \mathbf{c}_{6} \mathbf{c}_{7} \mathbf{c}_{8} \mathbf{c}_{0} Figure 1 : Discrete velocity set for the two-dimensional nine-velocity (D2Q9) lattice used in the present LBM implementation. The equilibrium distribution function f_{\alpha}^{eq} satisfies the following moment constraints: \sum_{\alpha}f_{\alpha}^{eq}=0,\qquad\sum_{\alpha}f_{\alpha}^{eq}\mathbf{c}_{\alpha}=\mathbf{u},\qquad\bm{\Pi}^{eq}=\sum_{\alpha}f_{\alpha}^{eq}\mathbf{c}_{\alpha}\mathbf{c}_{\alpha}=\mathbf{u}\mathbf{u}+p\mathbf{I}, (12) where p=p_{h}/\rho denotes the normalized hydrodynamic pressure. The explicit expression of the equilibrium distribution function reads f_{\alpha}^{eq}=\begin{cases}-(1-\omega_{0})\dfrac{p}{c_{s}^{2}}-\omega_{0}\dfrac{\mathbf{u}\cdot\mathbf{u}}{2c_{s}^{2}},&\alpha=0,\\[5.69054pt] \omega_{\alpha}\left[\dfrac{p}{c_{s}^{2}}+\dfrac{\mathbf{c}_{\alpha}\cdot\mathbf{u}}{c_{s}^{2}}+\dfrac{(\mathbf{c}_{\alpha}\cdot\mathbf{u})^{2}}{2c_{s}^{4}}-\dfrac{\mathbf{u}\cdot\mathbf{u}}{2c_{s}^{2}}\right],&\alpha\neq 0,\end{cases} (13) where c_{s}=c/\sqrt{3} is the lattice sound speed. If a force density \mathbf{f} acts on the system (solid or fluid), the corresponding forcing term in the LBE ( 10 ) is given by [ 15 ] : F_{\alpha}=\left(1-\frac{1}{2\tau}\right)w_{\alpha}\left[\frac{\mathbf{c}_{\alpha}-\mathbf{u}}{c_{s}^{2}}+\frac{(\mathbf{c}_{\alpha}\cdot\mathbf{u})}{c_{s}^{4}}\,\mathbf{c}_{\alpha}\right]\cdot\frac{\mathbf{f}}{\rho}. (14) The macroscopic velocity and pressure are then recovered as [ 16 , 13 ] : \displaystyle\mathbf{u} \displaystyle=\sum_{\alpha}f_{\alpha}\mathbf{c}_{\alpha}+\frac{\Delta t}{2\rho}\mathbf{f}, (15a) \displaystyle p \displaystyle=\frac{1}{1-\omega_{0}}\left(\sum_{\alpha\neq 0}f_{\alpha}c_{s}^{2}-\frac{\omega_{0}}{2}|\mathbf{u}|^{2}\right). (15b) Through a Chapman–Enskog expansion [ 16 ] , the viscous stress tensor is given by \bm{\tau}=2\nu\rho\mathbf{D}[\mathbf{u}]=-\left(1-\frac{\Delta t}{2\tau}\right)\bm{\Pi}^{neq}, (16) where \nu=c_{s}^{2}(\tau-\frac{\Delta t}{2}) is the kinematic viscosity, and \bm{\Pi}^{neq} is the non-equilibrium momentum flux tensor defined as \bm{\Pi}^{neq}=\sum_{\alpha}(f_{\alpha}-f_{\alpha}^{eq}). To penalize the rigid body constraint , we introduce the following source term in Eq. ( 10 ): S_{\alpha}=\frac{\omega_{\alpha}}{2c_{s}^{4}}\,\mathbf{E}:\left(\mathbf{c}_{\alpha}\otimes\mathbf{c}_{\alpha}-c_{s}^{2}\,\mathbf{I}\right), (17) where \mathbf{E}=\alpha\bm{\Pi}^{neq} and \alpha is a real-valued parameter. A Chapman–Enskog expansion then shows that the viscous stress tensor is modified as \frac{\bm{\tau}}{\rho}=2\frac{\nu}{1-\alpha\tau}\mathbf{D}[\mathbf{u}]. (18) Hence, the source term introduces an effective viscosity on the solid