feat(poroelastic): add Biot coupled u–p volume kernel

New BiotPoroelasticKernel with README for steady/transient coupled displacement and pore pressure.
This commit is contained in:
Jukka Aho
2026-05-11 02:23:06 +03:00
parent eac97e0b19
commit d646aa9d9e
2 changed files with 387 additions and 0 deletions
+24
View File
@@ -0,0 +1,24 @@
# SPDX-FileCopyrightText: 2015-2026 Jukka Aho
# SPDX-License-Identifier: MIT
# `src/domains/poroelastic/`
Steady quasi-static **Biot** coupling: solid displacement `u` with vertex
pore pressure `p`, Darcy flow in the pressure block, Biot coefficient `α`, and
optional fluid **storage** `storage_S` on pressure (L² mass for `assemble_M!`).
See `kernel.jl` for the weak form and `PorePressure` in `src/fields/api.jl`
for the scalar field tag.
Monolithic time stepping (e.g. backward Euler on `(K + M/dt)`) can use the
same `DOFBasedCOOCache` with `assemble!` + `assemble_M!`. The mass operator
includes `M_pp` (`storage_S`), `M_pu` and `M_up` (`α` times the divergence
pairings aligned with the steady `K_pu` / `K_up` blocks), and optional
consistent `M_uu` from solid density `ρ` (same as `ContinuumKernel` mass). Solid
acceleration dynamics beyond this `M_uu` slot are not part of
this kernel.
Tests: `test/domains/poroelastic/` (Terzaghi-style single-drainage column with
uniform IC vs Fourier when `α = 0` in `test_biot_terzaghi_uniform_ic.jl`, and
coupled `α > 0` refinement checks in `test_biot_terzaghi_coupled_refinement.jl`).
The thermo-poroelastic analogue for uncoupled **thermal** diffusion on a column
is `test/domains/thermo_poroelastic/test_thm_column_thermal_decay.jl`.
+363
View File
@@ -0,0 +1,363 @@
# SPDX-FileCopyrightText: 2015-2026 Jukka Aho
# SPDX-License-Identifier: MIT
#=
Linear quasi-steady **Biot poroelasticity** (small strain) on a single mesh:
vertex displacement `u` (same layout as [`ContinuumKernel`](@ref)) plus a
scalar **pore pressure** `p` at vertices (same low-order layout as
[`ThermoElasticKernel`](@ref) uses for temperature).
Weak form (steady stiffness blocks; no body loads shown):
: ε(u) : ε(v) dΩ α p · div(v) dΩ = rhs_u(v)
α q · div(u) dΩ + k p · q dΩ = rhs_p(q)
with `` and small-strain `ε(u)` from [`compute_stress`](@ref) on
`mech_material`, isotropic Darcy conductivity tensor `k` from
[`HydraulicConductivity`](@ref) (`k = K I` in 3D), and **Biot coefficient**
`α 0` (often denoted `b` in geotechnical notation).
The off-diagonal sign pattern matches [`ThermoElasticKernel`](@ref) with
`β α`: both coupling blocks use the same `-α (N_i/x_α) N_j` /
`-α N_i (N_j/x_β)` structure so the assembled `K` stays symmetric.
For transient **storage + volumetric strain rate** in the fluid mass balance,
`assemble_M!` / `apply_M!` add:
M_pp[i,j] = storage_S · N_i N_j dΩ,
M_pu[i,j] = α · N_i^p (N_j^u/x_{c_j}) dΩ,
M_up[i,j] = α · (N_i^u/x_{c_i}) N_j^p dΩ,
with `storage_S 0` and `α 0`. The `M_pu` / `M_up` blocks are Galerkin
partners of the steady `K_pu` / `K_up` couplings (assembled `M_pu = -K_pu`,
`M_up = -K_up` for the same mesh and quadrature). Optional solid density
`ρ 0` adds a consistent **`M_uu`** on displacements (same microkernel as
[`ContinuumKernel`](@ref) with `density = ρ`; `ρ = 0` skips). Rows with
`storage_S = 0` and `α = 0` still allow a transient-only mechanical mass when
`ρ > 0`.
`operator_is_posdef` is `false`: even though this symmetric operator can
be elliptic in some parameter regimes, downstream solvers should not
assume SPD (compare mixed `u``p` solids).
# DOF template
```julia
S = @DOFSet{u::DOF{Displacement{3}, Vertex}, p::DOF{PorePressure, Vertex}}
kernel = BiotPoroelasticKernel(
ContinuumFormulation{ThreeDimensional}(),
LinearElastic(E = 3.0e10, ν = 0.25),
HydraulicConductivity(K = 1.0e-12),
0.8,
0.0,
0.0,
)
```
=#
using Tensors
using ..JuliaFEM: AbstractKernel, AbstractFormulation
using ..JuliaFEM: ContinuumFormulation, AbstractContinuumTheory
using ..JuliaFEM: AbstractMaterial, LinearElastic, HydraulicConductivity
using ..JuliaFEM: hydraulic_conductivity_tensor
using ..JuliaFEM: ThermoElasticQPBuffer
using ..JuliaFEM: AssemblyMaterialWorkspace
using ..JuliaFEM: compute_stiffness_value
import ..JuliaFEM: qpoint_buffer_eltype, update_qpoint_buffer!, evaluate_entry,
evaluate_mass_entry,
reference_fields, get_field, dofs_per_node
using ..JuliaFEM: DOFLayoutEntry, field_idx, entity_local, component
"""
BiotPoroelasticKernel{Theory, MatM, MatF}
Coupled poroelastic kernel: `u` (3 DOFs / node) + pore pressure `p` (1 DOF /
node). See the file-level comment for the steady weak form.
# Fields
- `formulation::ContinuumFormulation{Theory}`
- `mech_material::MatM` — mechanical solid (`LinearElastic`, …)
- `flow_material::MatF` — [`HydraulicConductivity`](@ref) (Darcy mobility `K`)
- `α::Float64` — BiotWillis coefficient (set to `0.0` to recover a block-
diagonal elasticity + Darcy potential split for regression tests)
- `storage_S::Float64` — fluid storage coefficient on pressure (`≥ 0`);
`0` skips the `M_pp` block; `M_pu` and `M_up` still appear when `α > 0`
- `density::Float64` — solid mass density `ρ` (`≥ 0`) on `u` for `M_uu`; `0`
skips the displacement mass block (quasi-static default)
# Example
```julia
S = @DOFSet{u::DOF{Displacement{3}, Vertex}, p::DOF{PorePressure, Vertex}}
kernel = BiotPoroelasticKernel(
ContinuumFormulation{ThreeDimensional}(),
LinearElastic(E = 3.0e10, ν = 0.25),
HydraulicConductivity(K = 1.0e-12),
0.8,
0.0,
0.0,
)
```
"""
struct BiotPoroelasticKernel{Theory<:AbstractContinuumTheory,
MatM<:AbstractMaterial,
MatF<:HydraulicConductivity} <: AbstractKernel
formulation::ContinuumFormulation{Theory}
mech_material::MatM
flow_material::MatF
α::Float64
storage_S::Float64
density::Float64
function BiotPoroelasticKernel(
formulation::ContinuumFormulation{Theory},
mech_material::MatM,
flow_material::MatF,
α::Float64,
storage_S::Float64,
density::Float64,
) where {Theory<:AbstractContinuumTheory,
MatM<:AbstractMaterial,
MatF<:HydraulicConductivity}
α 0.0 || throw(ArgumentError("Biot coefficient α must be ≥ 0, got α = $α"))
storage_S 0.0 ||
throw(ArgumentError("storage_S must be ≥ 0, got storage_S = $storage_S"))
density 0.0 ||
throw(ArgumentError("density must be ≥ 0, got density = $density"))
return new{Theory, MatM, MatF}(
formulation,
mech_material,
flow_material,
α,
storage_S,
density,
)
end
end
function BiotPoroelasticKernel(
formulation::ContinuumFormulation{Theory},
mech_material::MatM,
flow_material::MatF,
α::Float64,
storage_S::Float64,
) where {Theory<:AbstractContinuumTheory,
MatM<:AbstractMaterial,
MatF<:HydraulicConductivity}
return BiotPoroelasticKernel(formulation, mech_material, flow_material, α, storage_S, 0.0)
end
function BiotPoroelasticKernel(
formulation::ContinuumFormulation{Theory},
mech_material::MatM,
flow_material::MatF,
α::Float64,
) where {Theory<:AbstractContinuumTheory,
MatM<:AbstractMaterial,
MatF<:HydraulicConductivity}
return BiotPoroelasticKernel(formulation, mech_material, flow_material, α, 0.0, 0.0)
end
"""Convenience constructor with `α = 0` and `storage_S = 0` (block-diagonal steady split)."""
function BiotPoroelasticKernel(
formulation::ContinuumFormulation{Theory},
mech_material::MatM,
flow_material::MatF,
) where {Theory<:AbstractContinuumTheory,
MatM<:AbstractMaterial,
MatF<:HydraulicConductivity}
return BiotPoroelasticKernel(formulation, mech_material, flow_material, 0.0, 0.0, 0.0)
end
@inline operator_is_posdef(::BiotPoroelasticKernel) = false
@inline dofs_per_node(::BiotPoroelasticKernel) = 4
function get_field(::K) where {K<:BiotPoroelasticKernel}
error("$(K) is multi-field — use `local_dof_layout(E)` and " *
"`elem.dof_indices` instead of `get_field(kernel)`.")
end
# Reuse the thermo-elastic per-IP buffer layout (elasticity tangent + rank-2 tensor).
@inline qpoint_buffer_eltype(::BiotPoroelasticKernel) = ThermoElasticQPBuffer
@inline function reference_fields(kernel::BiotPoroelasticKernel)
ε_ref = zero(SymmetricTensor{2,3,Float64,6})
σ_ref, 𝔻_ref, _ = compute_stress(kernel.mech_material, ε_ref, NamedTuple(), 0.0)
q_ref = zero(Vec{3,Float64})
k_ref = hydraulic_conductivity_tensor(kernel.flow_material)
return ((σ = σ_ref, 𝔻 = 𝔻_ref, q = q_ref, k = k_ref), NamedTuple())
end
@inline function update_qpoint_buffer!(
buffer::AbstractVector{ThermoElasticQPBuffer},
workspace::AssemblyMaterialWorkspace{FieldType, StateType},
::BiotPoroelasticKernel,
) where {FieldType, StateType}
fields = getfield(workspace, 1)
@inbounds for q in eachindex(buffer)
f = fields[q]
buffer[q] = ThermoElasticQPBuffer(f.𝔻, f.k)
end
return nothing
end
"""
evaluate_entry(kernel::BiotPoroelasticKernel, geometry_cache,
qp_vec, layout_i, layout_j, elem_id::Int) -> Float64
| (field_i, field_j) | block | contribution |
| ------------------ | ----- | ------------ |
| (1, 1) | K_uu | `Σ_q B_iα : : B_jβ · detJ_w` |
| (2, 2) | K_pp | `Σ_q ∇N_i · k · ∇N_j · detJ_w` |
| (1, 2) | K_up | `−α · Σ_q (∂N_i/∂x_α) · N_j · detJ_w` |
| (2, 1) | K_pu | `−α · Σ_q N_i · (∂N_j/∂x_β) · detJ_w` |
"""
@inline function evaluate_entry(
kernel::BiotPoroelasticKernel,
geometry_cache,
qp_vec::AbstractVector{ThermoElasticQPBuffer},
layout_i::DOFLayoutEntry,
layout_j::DOFLayoutEntry,
::Int,
)
fi = field_idx(layout_i)
fj = field_idx(layout_j)
node_i = entity_local(layout_i)
node_j = entity_local(layout_j)
comp_i = component(layout_i)
comp_j = component(layout_j)
n_ips = length(geometry_cache.detJ_w)
K_ij = 0.0
if fi == 1 && fj == 1
@inbounds for q in 1:n_ips
∇N_i = geometry_cache.∇N_data[q, node_i]
∇N_j = geometry_cache.∇N_data[q, node_j]
detJw = geometry_cache.detJ_w[q]
C = Tensor{4,3}(qp_vec[q].C)
K_ij += compute_stiffness_value(∇N_i, ∇N_j, C, comp_i, comp_j) * detJw
end
elseif fi == 2 && fj == 2
@inbounds for q in 1:n_ips
∇N_i = geometry_cache.∇N_data[q, node_i]
∇N_j = geometry_cache.∇N_data[q, node_j]
detJw = geometry_cache.detJ_w[q]
k_q = qp_vec[q].k
K_ij += (∇N_i k_q ∇N_j) * detJw
end
elseif fi == 1 && fj == 2
α = kernel.α
@inbounds for q in 1:n_ips
∇N_i = geometry_cache.∇N_data[q, node_i]
N_j = geometry_cache.N_data[q, node_j]
detJw = geometry_cache.detJ_w[q]
K_ij += -α * ∇N_i[comp_i] * N_j * detJw
end
else # fi == 2 && fj == 1
α = kernel.α
@inbounds for q in 1:n_ips
N_i = geometry_cache.N_data[q, node_i]
∇N_j = geometry_cache.∇N_data[q, node_j]
detJw = geometry_cache.detJ_w[q]
K_ij += -α * N_i * ∇N_j[comp_j] * detJw
end
end
return K_ij
end
"""
evaluate_mass_entry(::BiotPoroelasticKernel, ...)
| (field_i, field_j) | block | contribution |
| ------------------ | ----- | ------------ |
| (2, 2) | M_pp | `storage_S · Σ_q N_i N_j detJ_w` |
| (2, 1) | M_pu | `α · Σ_q N_i (∂N_j/∂x_{c_j}) detJ_w` |
| (1, 2) | M_up | `α · Σ_q (∂N_i/∂x_{c_i}) N_j detJ_w` |
| (1, 1) | M_uu | `ρ · Σ_q N_i N_j detJ_w` on matching displacement components (`ρ = 0` skips) |
| other | — | `0` |
`M_pu` and `M_up` match minus the steady `K_pu` and `K_up` microkernels for the
same `α` and layout. `M_uu` matches [`ContinuumKernel`](@ref) mass for the same
`ρ` and vertex layout.
"""
@inline function evaluate_mass_entry(
kernel::BiotPoroelasticKernel,
geometry_cache,
qp_buffer,
layout_i::DOFLayoutEntry,
layout_j::DOFLayoutEntry,
)
fi = field_idx(layout_i)
fj = field_idx(layout_j)
n_ips = length(geometry_cache.detJ_w)
if fi == 1 && fj == 1
ρ = kernel.density
ρ == 0.0 && return 0.0
component(layout_i) == component(layout_j) || return 0.0
node_i = entity_local(layout_i)
node_j = entity_local(layout_j)
M_ij = 0.0
@inbounds for q in 1:n_ips
N_i = geometry_cache.N_data[q, node_i]
N_j = geometry_cache.N_data[q, node_j]
detJw = geometry_cache.detJ_w[q]
M_ij += ρ * N_i * N_j * detJw
end
return M_ij
end
if fi == 2 && fj == 2
S = kernel.storage_S
S == 0.0 && return 0.0
node_i = entity_local(layout_i)
node_j = entity_local(layout_j)
M_ij = 0.0
@inbounds for q in 1:n_ips
N_i = geometry_cache.N_data[q, node_i]
N_j = geometry_cache.N_data[q, node_j]
detJw = geometry_cache.detJ_w[q]
M_ij += S * N_i * N_j * detJw
end
return M_ij
elseif fi == 2 && fj == 1
α = kernel.α
α == 0.0 && return 0.0
node_i = entity_local(layout_i)
node_j = entity_local(layout_j)
comp_j = component(layout_j)
M_ij = 0.0
@inbounds for q in 1:n_ips
N_i = geometry_cache.N_data[q, node_i]
∇N_j = geometry_cache.∇N_data[q, node_j]
detJw = geometry_cache.detJ_w[q]
M_ij += α * N_i * ∇N_j[comp_j] * detJw
end
return M_ij
elseif fi == 1 && fj == 2
α = kernel.α
α == 0.0 && return 0.0
node_i = entity_local(layout_i)
node_j = entity_local(layout_j)
comp_i = component(layout_i)
M_ij = 0.0
@inbounds for q in 1:n_ips
∇N_i = geometry_cache.∇N_data[q, node_i]
N_j = geometry_cache.N_data[q, node_j]
detJw = geometry_cache.detJ_w[q]
M_ij += α * ∇N_i[comp_i] * N_j * detJw
end
return M_ij
else
return 0.0
end
end