Source-linked AI summary
Parameter-robust discretization and preconditioning of Biot's consolidation model
Jeonghun J. Lee, Kent-Andre Mardal, Ragnar Winther
TL;DR
The paper addresses the challenge of solving Biot systems whose parameters vary widely and whose meshes are refined. It combines a stable three-field finite-element formulation with operator preconditioning in weighted Hilbert spaces. The resulting preconditioners are reported to be robust across the targeted parameter regimes and discretization choices, with numerical experiments confirming the theoretical results.
Problem
Biot systems involve independent parameters spanning difficult regimes, including small permeability, large elastic moduli, small time steps, and changing discretization resolution.
Method
The paper develops a stable three-field finite-element formulation and constructs block diagonal preconditioners through operator preconditioning, parameter-dependent norms, and equivalent discrete operators.
Results
The preconditioners are robust with respect to model-parameter variations, finite-element choices satisfying the stability condition, and discretization parameters, as confirmed by numerical experiments.
Takeaways & Limitations
The proposed framework covers regimes relevant to geophysics and computational biomechanics, including large shear and bulk moduli, small hydraulic conductivity, and small time steps.
Takeaways & Limitations
The analysis assumes practical parameter scalings such as s0 scaling like α^2/λ and α being close to 1, while very large λ requires additional inf-sup control beyond stabilization terms.
Abstract
from arXiv · showhide
Biot's consolidation model in poroelasticity has a number of applications in science, medicine, and engineering. The model depends on various parameters, and in practical applications these parameters ranges over several orders of magnitude. A current challenge is to design discretization techniques and solution algorithms that are well behaved with respect to these variations. The purpose of this paper is to study finite element discretizations of this model and construct block diagonal preconditioners for the discrete Biot systems. The approach taken here is to consider the stability of the problem in non-standard or weighted Hilbert spaces and employ the operator preconditioning approach. We derive preconditioners that are robust with respect to both the variations of the parameters and the mesh refinement. The parameters of interest are small time-step sizes, large bulk and shear moduli, and small hydraulic conductivity.
1. Introduction
Biot’s model couples deformation in a saturated porous medium with viscous fluid flow, creating numerical challenges when physical parameters vary widely. This paper develops stable finite element discretizations and preconditioners designed to remain robust under parameter changes and mesh refinement.
- Model: Biot’s consolidation model describes elastic deformation and viscous fluid flow in a porous medium saturated by fluid.The unknowns are the elastic displacement u and fluid pressure pF.
- Motivation: Practical applications span geoscience and medicine, with elastic and permeability parameters varying across several orders of magnitude.Reported examples include permeability from 10^-14 to 10^-16 m^2 in central nervous-system tissue and approximately 10^-9 to 10^-21 m^2 in geophysics.
- Motivation: Such parameter variation motivates numerical methods whose behavior remains robust across relevant model-parameter regimes.The paper highlights small hydraulic conductivity, small time steps, and large elastic moduli as target regimes.
- Problem: The paper focuses on preconditioners that remain well behaved under both parameter variation and finite-element mesh refinement.This matters because iterative-solver convergence depends heavily on suitable preconditioners for large discrete systems.
- Approach: The proposed approach develops a new three-field Biot formulation and a corresponding parameter-robust block diagonal preconditioner using operator preconditioning.The paper also discusses stability, finite-element discretization, and numerical examples supporting the construction.
2. Preliminaries
The paper develops parameter-robust operator preconditioning for finite element discretizations, using parameter-dependent Hilbert spaces to control stability, conditioning, and mesh effects. Simplified examples motivate weighted norms and block-diagonal preconditioners for difficult permeability and incompressibility limits.
- Operator preconditioning: Parameter-dependent Hilbert spaces provide a systematic route to preconditioners whose operator norms and condition numbers remain bounded independently of model parameters.The framework maps parameter-dependent coefficient operators between weighted spaces and uses the resulting isomorphism to define the preconditioner.
- Operator preconditioning: Uniformly stable finite element discretizations transfer continuous preconditioner structures to discrete systems, while multilevel or domain-decomposition methods replace expensive exact inverses.This replacement is needed for effective discrete solvers rather than only theoretically robust operators.
- Biot parameter regimes: The Biot challenge combines independent elastic and porous-flow parameters, including Lamé parameters, permeability, and the Biot–Willis constant, across difficult limiting regimes.The examples specifically examine permeability tending to zero and the elastic material approaching incompressibility.
- Permeability limit: Small permeability creates a singular limit, yet the proposed AMG-preconditioned discretization remains asymptotically stable as κ decreases to the Stokes case κ = 0.The reported iteration counts and condition numbers increase as κ decreases but remain bounded in the limit; the zero eigenvalue is excluded from condition-number calculations.
- Nearly incompressible elasticity: Large λ weakens the stabilizing pressure term, so standard norms lose stability and a λ-dependent norm is required to control both mean-zero and mean-value pressure components.The divergence controls only the mean-value-zero pressure part, while the λ-dependent stabilization controls the mean component.
- Nearly incompressible elasticity: For the nearly incompressible example, preconditioner B2 appears uniformly bounded in iterations and condition numbers with respect to λ and mesh refinement, whereas B1 condition numbers grow linearly with λ.B1 can remain efficient because only one eigenvalue tends to zero while the rest of the spectrum stays bounded, but it is less robust by condition number.
3. Parameter-robust stability of the continuous problems
The paper analyzes parameter-robust stability for Biot formulations and uses weighted norms to construct formulations whose operators remain uniformly bounded or invertible across λ, α, and κ. It shows why the solid-pressure formulation is difficult to precondition robustly and motivates total pressure as an alternative.
- Difficulties in typical formulations: Uniform boundedness of the three-field operator can be obtained with parameter-dependent norms, but its inverse is not uniformly controlled as λ becomes large.The stabilization terms λ^-1(pS, qS) and λ^-1(pF, qF) are insufficient to control both pressure L2 norms for large λ.
- Difficulties in typical formulations: The solid-pressure formulation is difficult to make parameter-robust because one divergence field must control two independent pressure components.The shared coupling through (div v, qS + qF) does not provide a parameter-independent inf-sup control for both pressures.
- Difficulties in typical formulations: Iterations increase substantially with λ, especially when κ is small, whereas larger κ moderates this growth in the examined preconditioned system.This supports the interpretation that the κ-dependent pressure-gradient term provides a partial stabilization for pF.
- A new three-field formulation: The total-pressure variable pT := −λ div u + αpF is introduced to circumvent the failed inf-sup condition of the solid-pressure formulation.This reformulation uses unknowns (u, pT, pF) and leads to a different three-field system.
- A new three-field formulation: The total-pressure formulation defines an operator with bounds uniform in λ, α, and κ, and its inverse is parameter-independent under the stated assumptions.The theorem supplies an inf-sup constant independent of λ, α, and κ, implying invertibility with a parameter-independent inverse norm.
4. Discretization and construction of preconditioners
The paper proposes finite element discretizations of a three-field Biot formulation and derives block-diagonal preconditioners using parameter-dependent norms. Under suitable discrete inf-sup conditions, the resulting operators are stable uniformly across model parameters and discretization changes.
- 4. Discretization and construction of preconditioners: Finite element discretizations of the three-field formulation admit parameter-robust preconditioners for the discrete Biot systems.The construction applies to finite element spaces satisfying the required stability conditions.
- 4. Discretization and construction of preconditioners: The discrete formulation assumes that the velocity–total-pressure pair is a stable Stokes pair with an h-independent inf-sup constant.This stability assumption is the discrete version of the relevant continuous inf-sup condition.
- 4. Discretization and construction of preconditioners: The resulting preconditioner is parameter-robust because it is based on the mapping properties and parameter-dependent norms of the operator.The framework is applied to the discrete analogue of the three-field system.
- 4. Discretization and construction of preconditioners: Under the stated inf-sup assumptions, the discrete operator satisfies a parameter-uniform stability bound for the admissible parameters.The same conclusion is stated for both general boundary conditions and the case Γd = ∂Ω.
- 4. Discretization and construction of preconditioners: The parameter-dependent norms motivate a block-diagonal preconditioner whose first and third blocks correspond to standard second-order elliptic operators.The middle block is less standard and requires a separate effective construction.
5. A preconditioner for the operator λ−1I + I0
This section constructs an efficient approximation for the inverse of λ^-1I + I0, a nonstandard block in the parameter-robust preconditioner. The construction uses mass-matrix decompositions, a rank-one representation, and the Sherman–Morrison–Woodbury formula.
- 5. A preconditioner for the operator λ^-1I + I0: When λ ≥ 1, λ^-1I + I0 is spectrally equivalent to λ^-1Im + I0, reducing the construction to the latter operator.The reduction permits an approximate inverse based on finite element mass matrices.
- 5. A preconditioner for the operator λ^-1I + I0: The matrix for λ^-1Im + I0 can be written as M + (λ^-1 − 1)mm^T, a rank-one modification of the mass matrix.The rank-one structure follows from the mean-value component of the finite element basis.
- 5. A preconditioner for the operator λ^-1I + I0: The Sherman–Morrison–Woodbury formula supplies an inverse for the rank-one modification using the mass matrix and a scalar correction.The paper applies the formula to I + (c^-1 − 1)mm^TM^-1.
- 5. A preconditioner for the operator λ^-1I + I0: For piecewise linear continuous finite elements, a Jacobi preconditioner for the mass matrix yields an explicit approximation for the constructed inverse.The Jacobi preconditioner is based on the inverse diagonal of the mass matrix.
- 5. A preconditioner for the operator λ^-1I + I0: Although the resulting matrices are generally dense, matrix-vector products exploit the outer-product structure and avoid explicitly forming dense matrices.The same structure also supports the implementation for discontinuous finite elements.
6. Numerical results
Numerical experiments compare the proposed preconditioners across boundary conditions, finite element spaces, mesh refinements, parameter changes, and variable hydraulic conductivity. The reported iteration counts are generally robust, including for the MINI element and high conductivity contrasts.
- 6. Numerical results: The experiments replace exact inverses in the first and third preconditioner blocks with Hypre algebraic multigrid operators.The second block uses either the Section 5 construction or a standard Jacobi preconditioner, depending on the preconditioner.
- 6. Numerical results: Four test cases vary boundary conditions, preconditioners, and finite element spaces, with the first three using Taylor–Hood elements and the fourth using MINI elements.The experiments use the unit square and compare preconditioners with structures (4.6) and (4.4).
- 6. Numerical results: Case 1 shows iteration results that are fairly robust with respect to parameter changes and mesh refinements.The tables report iterations and condition numbers under a relative-residual convergence criterion of 10^-6.
- 6. Numerical results: The MINI-element case has larger iteration counts than Taylor–Hood but remains fairly robust under parameter changes.The full numerical results also report that Case 3 can require about 30–45% more iterations than Case 2.
- 6. Numerical results: The variable-conductivity experiment remains fairly robust for mesh refinements and parameter changes, including high contrasts of κ.This experiment uses nonconstant κ in the geometry specified for Table 8.
7. Conclusion
The paper develops parameter-robust discretizations and preconditioners for Biot’s consolidation model. Its theoretical and numerical results cover broad parameter variations, stable finite element choices, and discretization changes.
- 7. Conclusion: The proposed preconditioners are robust with respect to model parameters, admissible finite element spaces, and discretization parameters.The construction uses mapping properties and parameter-dependent norms.
- 7. Conclusion: The covered parameter regimes include large shear and bulk elastic moduli, small hydraulic conductivity, and small time-steps.These ranges include applications in geophysics and computational biomechanics.
- 7. Conclusion: Numerical experiments confirm the theoretical robustness results.