Nonlinear Elasticity

Note

The full explanation for the underlying FEM theory in this example can be found in the Hyperelasticity tutorial of the Ferrite.jl documentation.

Implementation

The following code is based on the Hyperelasticity tutorial from the Ferrite.jl documentation, with some comments removed for brevity. There are two main modifications:

  1. Second-order Lagrange shape functions are used for field approximation: ip = Lagrange{RefTriangle,2}()^2.
  2. Four quadrature points are used to accommodate the second-order shape functions: qr = QuadratureRule{RefTriangle}(4).
using Ferrite, Tensors, TimerOutputs, IterativeSolvers

using FerriteMultigrid

TimerOutputs.enable_debug_timings(AlgebraicMultigrid)
TimerOutputs.enable_debug_timings(FerriteMultigrid)

struct NeoHooke
    μ::Float64
    λ::Float64
end

function Ψ(C, mp::NeoHooke)
    μ = mp.μ
    λ = mp.λ
    Ic = tr(C)
    J = sqrt(det(C))
    return μ / 2 * (Ic - 3 - 2 * log(J)) + λ / 2 * (J - 1)^2
end

function constitutive_driver(C, mp::NeoHooke)
    # Compute all derivatives in one function call
    ∂²Ψ∂C², ∂Ψ∂C = Tensors.hessian(y -> Ψ(y, mp), C, :all)
    S = 2.0 * ∂Ψ∂C
    ∂S∂C = 2.0 * ∂²Ψ∂C²
    return S, ∂S∂C
end;

function assemble_element!(ke, ge, cell, cv, fv, mp, ue, ΓN)
    # Reinitialize cell values, and reset output arrays
    reinit!(cv, cell)
    fill!(ke, 0.0)
    fill!(ge, 0.0)

    b = Vec{3}((0.0, -0.5, 0.0)) # Body force
    tn = 0.1 # Traction (to be scaled with surface normal)
    ndofs = getnbasefunctions(cv)

    for qp in 1:getnquadpoints(cv)
        dΩ = getdetJdV(cv, qp)
        # Compute deformation gradient F and right Cauchy-Green tensor C
        ∇u = function_gradient(cv, qp, ue)
        F = one(∇u) + ∇u
        C = tdot(F) # F' ⋅ F
        # Compute stress and tangent
        S, ∂S∂C = constitutive_driver(C, mp)
        P = F ⋅ S
        I = one(S)
        ∂P∂F = otimesu(I, S) + 2 * F ⋅ ∂S∂C ⊡ otimesu(F', I)

        # Loop over test functions
        for i in 1:ndofs
            # Test function and gradient
            δui = shape_value(cv, qp, i)
            ∇δui = shape_gradient(cv, qp, i)
            # Add contribution to the residual from this test function
            ge[i] += (∇δui ⊡ P - δui ⋅ b) * dΩ

            ∇δui∂P∂F = ∇δui ⊡ ∂P∂F # Hoisted computation
            for j in 1:ndofs
                ∇δuj = shape_gradient(cv, qp, j)
                # Add contribution to the tangent
                ke[i, j] += (∇δui∂P∂F ⊡ ∇δuj) * dΩ
            end
        end
    end

    # Surface integral for the traction
    for facet in 1:nfacets(cell)
        if (cellid(cell), facet) in ΓN
            reinit!(fv, cell, facet)
            for q_point in 1:getnquadpoints(fv)
                t = tn * getnormal(fv, q_point)
                dΓ = getdetJdV(fv, q_point)
                for i in 1:ndofs
                    δui = shape_value(fv, q_point, i)
                    ge[i] -= (δui ⋅ t) * dΓ
                end
            end
        end
    end
    return
end;

function assemble_global!(K, g, dh, cv, fv, mp, u, ΓN)
    n = ndofs_per_cell(dh)
    ke = zeros(n, n)
    ge = zeros(n)

    # start_assemble resets K and g
    assembler = start_assemble(K, g)

    # Loop over all cells in the grid
    for cell in CellIterator(dh)
        global_dofs = celldofs(cell)
        ue = u[global_dofs] # element dofs
        assemble_element!(ke, ge, cell, cv, fv, mp, ue, ΓN)
        assemble!(assembler, global_dofs, ke, ge)
    end
    return
end;

Near Null Space (NNS)

In multigrid methods for problems with vector-valued unknowns, such as elasticity, the near null space represents the low energy mode or the smooth error that needs to be captured in the coarser grid when using SA-AMG (Smoothed Aggregation Algebraic Multigrid), more on the topic can be found in Schroder [1].

For 3D linear elasticity problems, the rigid body modes are:

  1. Translation in the x-direction,
  2. Translation in the y-direction,
  3. Translation in the z-direction,
  4. Rotation about the x-axis (i.e., $x_1$): each point (x, y, z) is mapped to (0, -z, y).
  5. Rotation about the y-axis (i.e., $x_2$): each point (x, y, z) is mapped to (z, 0, -x).
  6. Rotation about the z-axis (i.e., $x_3$): each point (x, y, z) is mapped to (-y, x, 0).

The function create_nns constructs the NNS matrix B ∈ ℝ^{n × 6}, where n is the number of degrees of freedom (DOFs) for the case of p = 1 (i.e., linear interpolation), because B is only relevant for AMG.

function create_nns(dh, fieldname = first(dh.field_names))
    @assert length(dh.field_names) == 1 "Only a single field is supported for now."

    coords_flat = zeros(ndofs(dh))
    apply_analytical!(coords_flat, dh, fieldname, x -> x)
    coords = reshape(coords_flat, (length(coords_flat) ÷ 3, 3))

    grid = dh.grid
    B = zeros(Float64, ndofs(dh), 6)
    B[1:3:end, 1] .= 1 # x - translation
    B[2:3:end, 2] .= 1 # y - translation
    B[3:3:end, 3] .= 1 # z - translation

    # rotations
    x = coords[:, 1]
    y = coords[:, 2]
    z = coords[:, 3]
    # Around x
    B[2:3:end, 4] .= -z
    B[3:3:end, 4] .= y
    # Around y
    B[1:3:end, 5] .= z
    B[3:3:end, 5] .= -x
    # Around z
    B[1:3:end, 6] .= -y
    B[2:3:end, 6] .= x

    return B
end

function _solve(N = 10)
    reset_timer!()

    # Generate a grid
    L = 1.0
    left = zero(Vec{3})
    right = L * ones(Vec{3})
    grid = generate_grid(Tetrahedron, (N, N, N), left, right)

    # Material parameters
    E = 10.0
    ν = 0.3
    μ = E / (2(1 + ν))
    λ = (E * ν) / ((1 + ν) * (1 - 2ν))
    mp = NeoHooke(μ, λ)

    # Finite element base
    ip = Lagrange{RefTetrahedron, 2}()^3
    qr = QuadratureRule{RefTetrahedron}(4)
    qr_facet = FacetQuadratureRule{RefTetrahedron}(3)
    cv = CellValues(qr, ip)
    fv = FacetValues(qr_facet, ip)

    # DofHandler
    dh = DofHandler(grid)
    add!(dh, :u, ip) # Add a displacement field
    close!(dh)

    function rotation(X, t)
        θ = pi / 3 # 60°
        x, y, z = X
        return t * Vec{3}(
            (
                0.0,
                L / 2 - y + (y - L / 2) * cos(θ) - (z - L / 2) * sin(θ),
                L / 2 - z + (y - L / 2) * sin(θ) + (z - L / 2) * cos(θ),
            )
        )
    end

    ch = ConstraintHandler(dh)
    # Add a homogeneous boundary condition on the "clamped" edge
    dbc = Dirichlet(:u, getfacetset(grid, "right"), (x, t) -> [0.0, 0.0, 0.0], [1, 2, 3])
    add!(ch, dbc)
    dbc = Dirichlet(:u, getfacetset(grid, "left"), (x, t) -> rotation(x, t), [1, 2, 3])
    add!(ch, dbc)
    close!(ch)
    t = 0.5
    Ferrite.update!(ch, t)

    # Neumann part of the boundary
    ΓN = union(
        getfacetset(grid, "top"),
        getfacetset(grid, "bottom"),
        getfacetset(grid, "front"),
        getfacetset(grid, "back"),
    )

    # Pre-allocation of vectors for the solution and Newton increments
    _ndofs = ndofs(dh)
    un = zeros(_ndofs) # previous solution vector
    u = zeros(_ndofs)
    Δu = zeros(_ndofs)
    ΔΔu = zeros(_ndofs)
    apply!(un, ch)

    # Create sparse matrix and residual vector
    K = allocate_matrix(dh)
    g = zeros(_ndofs)

    # FIXME this needs better integration
    dh_coarse = DofHandler(grid)
    add!(dh_coarse, :u, Lagrange{RefTetrahedron, 1}()^3) # Add a displacement field
    close!(dh_coarse)
    B = create_nns(dh_coarse)
    config_gal = pmultigrid_config(coarse_strategy = Galerkin())
    fe_space = FESpace(dh, cv, ch)
    pcoarse_solver = SmoothedAggregationCoarseSolver(; B)
    builder = PMultigridPreconBuilder(fe_space, config_gal; pcoarse_solver)

    # Perform Newton iterations
    newton_itr = -1
    NEWTON_TOL = 1.0e-8
    NEWTON_MAXITER = 30

    @info ndofs(dh)

    while true
        newton_itr += 1
        # Construct the current guess
        u .= un .+ Δu
        # Compute residual and tangent for current guess
        assemble_global!(K, g, dh, cv, fv, mp, u, ΓN)
        # Apply boundary conditions
        apply_zero!(K, g, ch)
        # Compute the residual norm and compare with tolerance
        normg = norm(g)
        if normg < NEWTON_TOL
            break
        elseif newton_itr > NEWTON_MAXITER
            error("Reached maximum Newton iterations, aborting")
        end

        # Compute increment using conjugate gradients
        fill!(ΔΔu, 0.0)
        @timeit "Setup preconditioner" Pl = builder(K)[1]
        @timeit "Galerkin CG" IterativeSolvers.cg!(ΔΔu, K, g; Pl, maxiter = 100, verbose=false)
        fill!(ΔΔu, 0.0)
        @timeit "Galerkin GMRES" IterativeSolvers.gmres!(ΔΔu, K, g; Pl, maxiter = 100, verbose=false)
        fill!(ΔΔu, 0.0)
        @timeit "CG" IterativeSolvers.cg!(ΔΔu, K, g; maxiter = 1000, verbose=false)
        fill!(ΔΔu, 0.0)
        @timeit "GMRES" IterativeSolvers.gmres!(ΔΔu, K, g; maxiter = 1000, verbose=false)

        apply_zero!(ΔΔu, ch)
        Δu .-= ΔΔu
        break
    end

    # Save the solution
    @timeit "export" begin
        VTKGridFile("hyperelasticity", dh) do vtk
            write_solution(vtk, dh, u)
        end
    end

    print_timer(title = "Analysis with $(getncells(grid)) elements", linechars = :ascii)
    return u
end

u = _solve();
[ Info: 27783
---------------------------------------------------------------------------------------
     Analysis with 6000 elements              Time                    Allocations
                                     -----------------------   ------------------------
          Tot / % measured:               4.97s /  86.4%            441MiB /  78.8%

Section                      ncalls     time    %tot     avg     alloc    %tot      avg
---------------------------------------------------------------------------------------
GMRES                             1    1.78s   41.5%   1.78s   4.68MiB    1.3%  4.68MiB
CG                                1    893ms   20.8%   893ms    652KiB    0.2%   652KiB
Setup preconditioner              1    858ms   20.0%   858ms    319MiB   91.7%   319MiB
  pmultigrid hierarchy            1    858ms   20.0%   858ms    319MiB   91.7%   319MiB
    extend_hierarchy!             1    817ms   19.0%   817ms    238MiB   68.4%   238MiB
      build prolongator           1    796ms   18.5%   796ms    150MiB   43.2%   150MiB
        assembly                  1    796ms   18.5%   796ms    150MiB   43.1%   150MiB
        row normalization         1   82.7μs    0.0%  82.7μs     0.00B    0.0%    0.00B
      RAP                         1   21.5ms    0.5%  21.5ms   87.7MiB   25.2%  87.7MiB
      build restriction           1    410ns    0.0%   410ns     48.0B    0.0%    48.0B
    coarse solver setup           1   39.4ms    0.9%  39.4ms   78.1MiB   22.5%  78.1MiB
      extend_hierarchy!           3   39.4ms    0.9%  13.1ms   77.9MiB   22.4%  26.0MiB
        RAP                       3   12.1ms    0.3%  4.04ms   33.7MiB    9.7%  11.2MiB
        restriction setup         3   11.8ms    0.3%  3.93ms   38.1MiB   10.9%  12.7MiB
        improve candidates        3   10.3ms    0.2%  3.43ms     0.00B    0.0%    0.00B
        fit candidates            3   4.12ms    0.1%  1.37ms   2.79MiB    0.8%   952KiB
        strength                  3    950μs    0.0%   317μs   2.99MiB    0.9%  1.00MiB
        aggregation               3   80.4μs    0.0%  26.8μs    144KiB    0.0%  47.9KiB
      coarse solver setup         1   48.2μs    0.0%  48.2μs   6.57KiB    0.0%  6.57KiB
      prologue                    1   13.2μs    0.0%  13.2μs    219KiB    0.1%   219KiB
      ml setup                    1   4.68μs    0.0%  4.68μs      224B    0.0%     224B
    coarsen order                 1   1.11ms    0.0%  1.11ms   2.49MiB    0.7%  2.49MiB
Galerkin GMRES                    1    384ms    8.9%   384ms   13.4MiB    3.9%  13.4MiB
  Postsmoother                   19    146ms    3.4%  7.70ms     0.00B    0.0%    0.00B
  Presmoother                    19    146ms    3.4%  7.68ms     0.00B    0.0%    0.00B
  Residual eval                  19   28.4ms    0.7%  1.49ms     0.00B    0.0%    0.00B
  Coarse solve                   19   23.5ms    0.5%  1.24ms    684KiB    0.2%  36.0KiB
    Presmoother                  57   8.94ms    0.2%   157μs     0.00B    0.0%    0.00B
    Postsmoother                 57   8.20ms    0.2%   144μs     0.00B    0.0%    0.00B
    Residual eval                57   2.92ms    0.1%  51.3μs     0.00B    0.0%    0.00B
    Restriction                  57   1.57ms    0.0%  27.6μs     0.00B    0.0%    0.00B
    Prolongation                 57   1.31ms    0.0%  23.1μs     0.00B    0.0%    0.00B
    Coarse solve                 19   29.7μs    0.0%  1.56μs     0.00B    0.0%    0.00B
  Prolongation                   19   1.07ms    0.0%  56.4μs     0.00B    0.0%    0.00B
  Restriction                    19    894μs    0.0%  47.1μs     0.00B    0.0%    0.00B
Galerkin CG                       1    362ms    8.4%   362ms   5.34MiB    1.5%  5.34MiB
  Postsmoother                   19    141ms    3.3%  7.40ms     0.00B    0.0%    0.00B
  Presmoother                    19    136ms    3.2%  7.15ms     0.00B    0.0%    0.00B
  Residual eval                  19   27.8ms    0.6%  1.46ms     0.00B    0.0%    0.00B
  Coarse solve                   19   23.2ms    0.5%  1.22ms    684KiB    0.2%  36.0KiB
    Presmoother                  57   8.78ms    0.2%   154μs     0.00B    0.0%    0.00B
    Postsmoother                 57   8.16ms    0.2%   143μs     0.00B    0.0%    0.00B
    Residual eval                57   2.85ms    0.1%  50.0μs     0.00B    0.0%    0.00B
    Restriction                  57   1.56ms    0.0%  27.4μs     0.00B    0.0%    0.00B
    Prolongation                 57   1.27ms    0.0%  22.2μs     0.00B    0.0%    0.00B
    Coarse solve                 19   32.5μs    0.0%  1.71μs     0.00B    0.0%    0.00B
  Prolongation                   19    993μs    0.0%  52.2μs     0.00B    0.0%    0.00B
  Restriction                    19    853μs    0.0%  44.9μs     0.00B    0.0%    0.00B
export                            1   14.3ms    0.3%  14.3ms   4.88MiB    1.4%  4.88MiB
---------------------------------------------------------------------------------------

This page was generated using Literate.jl.