test(domains): add Stokes mixed kernel regression

This commit is contained in:
Jukka Aho
2026-05-09 18:38:29 +03:00
parent fc298ee014
commit 6e255a83f5
@@ -0,0 +1,97 @@
# This file is a part of JuliaFEM.
# License is MIT: see https://github.com/JuliaFEM/JuliaFEM.jl/blob/master/LICENSE.md
"""
`StokesMixedKernel`: Newtonian Stokes up on the DOF-based assembler.
1. `K_up`, `K_pu`, `K_pp` match `MixedUPKernel` on the same mesh (geometry-only blocks).
2. `K_uu` scales linearly with viscosity `μ`.
3. Global `K` is numerically symmetric; `assemble!` is allocation-free after warmup.
"""
using Test
using JuliaFEM
using JuliaFEM: StokesMixedKernel, MixedUPKernel, ContinuumFormulation, FullThreeD, LinearElastic
using JuliaFEM: DOFBasedCOOAssembler, DOFBasedCOOCache, assemble!, extract_system
using JuliaFEM: create_unit_cube_mesh, create_elements!, @DOFSet, DOF, Displacement, Vertex
using LinearAlgebra
function _hex8_unit_cube_single_element()
return create_unit_cube_mesh(Hex8; nx = 1, ny = 1, nz = 1)
end
@testset "StokesMixedKernel vs MixedUPKernel (pressure blocks)" begin
mesh = _hex8_unit_cube_single_element()
mat = LinearElastic(E = 210e9, ν = 0.3)
inv_bulk = 1.0 / (210e9 / 3)
S = @DOFSet{u::DOF{Displacement{3}, Vertex}, p::DOF{Float64, Cell}}
elem, handler = create_elements!(mesh, Element{Hex8, Lagrange{1}, S})
asm = DOFBasedCOOAssembler()
k_stokes = StokesMixedKernel(ContinuumFormulation{FullThreeD}(); μ = 1.0e-3, inv_bulk = inv_bulk)
c_s = DOFBasedCOOCache(elem, handler, mesh, k_stokes)
assemble!(c_s, asm, k_stokes, mesh)
K_s, _ = extract_system(c_s)
k_up = MixedUPKernel(ContinuumFormulation{FullThreeD}(), mat; inv_bulk = inv_bulk)
c_u = DOFBasedCOOCache(elem, handler, mesh, k_up)
assemble!(c_u, asm, k_up, mesh)
K_u, _ = extract_system(c_u)
n_u = 3 * length(mesh.nodes)
ru = 1:n_u
rp = (n_u + 1):(n_u + 1)
@test norm(K_s[ru, rp] - K_u[ru, rp]) < 1e-12 * max(1.0, norm(K_u[ru, rp]))
@test norm(K_s[rp, ru] - K_u[rp, ru]) < 1e-12 * max(1.0, norm(K_u[rp, ru]))
@test abs(K_s[rp[1], rp[1]] - K_u[rp[1], rp[1]]) < 1e-12 * max(1.0, abs(K_u[rp[1], rp[1]]))
end
@testset "StokesMixedKernel K_uu scales with μ" begin
mesh = _hex8_unit_cube_single_element()
S = @DOFSet{u::DOF{Displacement{3}, Vertex}, p::DOF{Float64, Cell}}
elem, handler = create_elements!(mesh, Element{Hex8, Lagrange{1}, S})
asm = DOFBasedCOOAssembler()
k1 = StokesMixedKernel(ContinuumFormulation{FullThreeD}(); μ = 1.0, inv_bulk = 0.0)
c1 = DOFBasedCOOCache(elem, handler, mesh, k1)
assemble!(c1, asm, k1, mesh)
K1, _ = extract_system(c1)
k2 = StokesMixedKernel(ContinuumFormulation{FullThreeD}(); μ = 2.0, inv_bulk = 0.0)
c2 = DOFBasedCOOCache(elem, handler, mesh, k2)
assemble!(c2, asm, k2, mesh)
K2, _ = extract_system(c2)
n_u = 3 * length(mesh.nodes)
ru = 1:n_u
rel = norm(K2[ru, ru] - 2 * K1[ru, ru]) / max(norm(K1[ru, ru]), 1.0)
@test rel < 1e-12
end
@testset "StokesMixedKernel symmetry + zero allocations" begin
mesh = _hex8_unit_cube_single_element()
S = @DOFSet{u::DOF{Displacement{3}, Vertex}, p::DOF{Float64, Cell}}
elem, handler = create_elements!(mesh, Element{Hex8, Lagrange{1}, S})
k = StokesMixedKernel(ContinuumFormulation{FullThreeD}(); μ = 1.0e-2, inv_bulk = 1e-6)
asm = DOFBasedCOOAssembler()
cache = DOFBasedCOOCache(elem, handler, mesh, k)
assemble!(cache, asm, k, mesh)
K, _ = extract_system(cache)
R = K - transpose(K)
@test norm(R) <= 1e-8 * max(1.0, norm(K))
for _ in 1:3
assemble!(cache, asm, k, mesh)
end
GC.gc()
@test (@allocated assemble!(cache, asm, k, mesh)) == 0
end
@testset "StokesMixedKernel API guards" begin
k = StokesMixedKernel(ContinuumFormulation{FullThreeD}(); μ = 1.0)
@test dofs_per_node(k) == 4
@test_throws ErrorException get_field(k)
@test_throws ArgumentError StokesMixedKernel(ContinuumFormulation{FullThreeD}(); μ = 0.0)
end