test(domains): add mixed u–p kernel regression

This commit is contained in:
Jukka Aho
2026-05-09 18:38:29 +03:00
parent 4df002e673
commit e899ebed6a
@@ -0,0 +1,142 @@
# This file is a part of JuliaFEM.
# License is MIT: see https://github.com/JuliaFEM/JuliaFEM.jl/blob/master/LICENSE.md
"""
`MixedUPKernel`: displacementpressure (P0) on the DOF-based assembler.
1. `K_uu` matches `ContinuumKernel` on the same mesh when `inv_bulk = 0`
(`Hex8` / `Lagrange{1}` brick and reference `Tet10` / `Lagrange{2}`).
2. `global_field_ranges` / `saddle_point_blocks` recover the `u` block and couplings.
3. Global `K` is numerically symmetric where tested.
4. `assemble!` allocates 0 bytes after warmup (same contract as other kernels).
"""
using Test
using JuliaFEM
using JuliaFEM: MixedUPKernel, ContinuumKernel, ContinuumFormulation, FullThreeD
using JuliaFEM: DOFBasedCOOAssembler, DOFBasedCOOCache, assemble!, extract_system
using JuliaFEM: create_unit_cube_mesh, create_elements!, @DOFSet, DOF, Displacement, Vertex
using JuliaFEM: global_field_ranges, saddle_point_blocks
using LinearAlgebra
using SparseArrays
function _hex8_unit_cube_single_element()
return create_unit_cube_mesh(Hex8; nx = 1, ny = 1, nz = 1)
end
@testset "MixedUPKernel vs ContinuumKernel (K_uu block)" begin
mesh = _hex8_unit_cube_single_element()
mat = LinearElastic(E = 210e9, ν = 0.3)
S_u = @DOFSet{u::DOF{Displacement{3}, Vertex}}
elem_u, h_u = create_elements!(mesh, Element{Hex8, Lagrange{1}, S_u})
ku = ContinuumKernel(ContinuumFormulation{FullThreeD}(), mat)
asm = DOFBasedCOOAssembler()
cache_u = DOFBasedCOOCache(elem_u, h_u, mesh, ku)
assemble!(cache_u, asm, ku, mesh)
K_u, _ = extract_system(cache_u)
S_up = @DOFSet{u::DOF{Displacement{3}, Vertex}, p::DOF{Float64, Cell}}
elem_up, h_up = create_elements!(mesh, Element{Hex8, Lagrange{1}, S_up})
kup = MixedUPKernel(ContinuumFormulation{FullThreeD}(), mat; inv_bulk = 0.0)
cache_up = DOFBasedCOOCache(elem_up, h_up, mesh, kup)
assemble!(cache_up, asm, kup, mesh)
K_up, _ = extract_system(cache_up)
n_u = 3 * length(mesh.nodes)
@test size(K_up, 1) == n_u + 1
Kuu = K_up[1:n_u, 1:n_u]
rel = norm(Kuu - K_u) / max(norm(K_u), 1.0)
@test rel < 1e-12
ru, rp = global_field_ranges(h_up)
@test ru == 1:n_u
@test rp == (n_u + 1):(n_u + 1)
blk = saddle_point_blocks(K_up, ru, rp)
@test blk.A Kuu
@test norm(blk.B) > 0.0
# Symmetric bilinear form for this kernel; general saddle-point Jacobians need not satisfy this.
@test blk.Bt transpose(blk.B)
# Pressurepressure row/column: single dof, zero compressibility
@test abs(K_up[n_u + 1, n_u + 1]) < 1e-20
end
@testset "MixedUPKernel global symmetry + K_pp sign" begin
mesh = _hex8_unit_cube_single_element()
mat = LinearElastic(E = 210e9, ν = 0.3)
κ = mat.E / (3 * (1 - 2 * mat.ν)) # isotropic bulk modulus
inv_bulk = 1.0 / κ
S_up = @DOFSet{u::DOF{Displacement{3}, Vertex}, p::DOF{Float64, Cell}}
elem_up, h_up = create_elements!(mesh, Element{Hex8, Lagrange{1}, S_up})
kup = MixedUPKernel(ContinuumFormulation{FullThreeD}(), mat; inv_bulk = inv_bulk)
asm = DOFBasedCOOAssembler()
cache_up = DOFBasedCOOCache(elem_up, h_up, mesh, kup)
assemble!(cache_up, asm, kup, mesh)
K, _ = extract_system(cache_up)
R = K - transpose(K)
@test norm(R) <= 1e-8 * max(1.0, norm(K))
n_u = 3 * length(mesh.nodes)
kpp = K[n_u + 1, n_u + 1]
@test kpp < 0.0
end
@testset "MixedUPKernel Tet10 quadratic — K_uu vs ContinuumKernel" begin
nodes = Vec{3, Float64}[reference_coordinates(Tet10())...]
conn = ntuple(i -> UInt32(i), 10)
mesh = Mesh{Tet10}(nodes, [conn])
mat = LinearElastic(E = 210e9, ν = 0.3)
S_u = @DOFSet{u::DOF{Displacement{3}, Vertex}}
elem_u, h_u = create_elements!(mesh, Element{Tet10, Lagrange{2}, S_u})
ku = ContinuumKernel(ContinuumFormulation{FullThreeD}(), mat)
asm = DOFBasedCOOAssembler()
cache_u = DOFBasedCOOCache(elem_u, h_u, mesh, ku)
assemble!(cache_u, asm, ku, mesh)
K_u, _ = extract_system(cache_u)
S_up = @DOFSet{u::DOF{Displacement{3}, Vertex}, p::DOF{Float64, Cell}}
elem_up, h_up = create_elements!(mesh, Element{Tet10, Lagrange{2}, S_up})
kup = MixedUPKernel(ContinuumFormulation{FullThreeD}(), mat; inv_bulk = 0.0)
cache_up = DOFBasedCOOCache(elem_up, h_up, mesh, kup)
assemble!(cache_up, asm, kup, mesh)
K_up, _ = extract_system(cache_up)
ru, rp = global_field_ranges(h_up)
n_u = 3 * length(mesh.nodes)
@test length(ru) == n_u
@test length(rp) == 1
blk = saddle_point_blocks(K_up, ru, rp)
@test size(K_u, 1) == n_u
rel = norm(Matrix(blk.A) - Matrix(K_u)) / max(norm(K_u), 1.0)
@test rel < 1e-11
R = K_up - transpose(K_up)
@test norm(R) <= 1e-8 * max(1.0, norm(K_up))
end
@testset "MixedUPKernel assemble! zero allocations" begin
mesh = _hex8_unit_cube_single_element()
mat = LinearElastic(E = 210e9, ν = 0.3)
S_up = @DOFSet{u::DOF{Displacement{3}, Vertex}, p::DOF{Float64, Cell}}
elem_up, h_up = create_elements!(mesh, Element{Hex8, Lagrange{1}, S_up})
kup = MixedUPKernel(ContinuumFormulation{FullThreeD}(), mat; inv_bulk = 1e-11)
asm = DOFBasedCOOAssembler()
cache_up = DOFBasedCOOCache(elem_up, h_up, mesh, kup)
for _ in 1:3
assemble!(cache_up, asm, kup, mesh)
end
GC.gc()
@test (@allocated assemble!(cache_up, asm, kup, mesh)) == 0
end
@testset "MixedUPKernel API guards" begin
mat = LinearElastic(E = 210e9, ν = 0.3)
k = MixedUPKernel(ContinuumFormulation{FullThreeD}(), mat)
@test dofs_per_node(k) == 4
@test_throws ErrorException get_field(k)
end