morphogen.fracture.phasefield¶
Brittle fracture of voxel images by the phase-field method: cracks that start, branch and merge in a periodic microstructure without being meshed or tracked.
A field d (0: sound, 1: broken) regularizes the cracks over the length l
(Francfort & Marigo 1998; Bourdin, Francfort & Marigo 2000): the energy of the cell is
E(u, d) = int g(d) psi(eps(u)) dV + int G_c gamma(d, grad d) dV, g(d) = (1 - d)^2 + k,
with the crack density gamma = d^2 / (2 l) + l |grad d|^2 / 2 (the model AT2: the
damage grows from the first load) or gamma = 3/8 (d / l + l |grad d|^2) (AT1:
elastic up to the stress sqrt(3 G_c E / (8 l)) in uniaxial tension), whose integral
tends to the area of the cracks as l vanishes: the work to break the cell is the
toughness G_c times that area, whatever the voxels, provided they resolve l
(linear elements over-estimate it by the factor 1 + h / (2 l), h their size: at
least two voxels per length).
The problem is solved by alternate minimization (Bourdin et al. 2000): at each step of a loading path, until the damage no longer moves,
the displacement at fixed damage: the periodic elastic problem of the voxels of the FFT solver (
morphogen.homogenization.fft), each voxel with its stiffness timesgof its damage (conjugate gradients preconditioned in Fourier space, the reference medium scaled by the degradation of the nodes; the macroscopic components prescribed as stresses are condensed);the damage at fixed displacement:
(G_c / l + 2 psi+) d - G_c l lap(d) = 2 psi+(AT2) withpsi+the tensile energy of each voxel, under the bound that the damage does not decrease (irreversibility, by a primal-dual active set: the obstacle problem ofAT1, whose damage is zero below its elastic limit and whose cracks have the finite width4 l); trilinear elements with a lumped reaction, conjugate gradients. Withirreversibility="history"the bound is replaced by the largestpsi+reached at each voxel (the history field of Miehe, Hofacker & Welschinger 2010): simpler, but it widens the cracks and over-estimates their energy.
The stress is degraded as a whole (sigma = g(d) C : eps: the hybrid formulation of
Ambati, Gerasimov & De Lorenzis 2015, linear at fixed damage) while the energy that
drives the damage is its tensile part only: all of it (split="none"), the spectral
split of Miehe et al. ("spectral": lambda <tr eps>+^2 / 2 + mu sum <eps_i>+^2, by
default) or the volumetric one of Amor et al. ("volumetric": K <tr eps>+^2 / 2 + mu
dev eps : dev eps), so that compression does not break. Each phase has its toughness;
the phases without one do not break (their stiffness is not degraded), the voids carry
neither stress nor cracks. Initial cracks are voxels given as broken: the damage of their
nodes is held at 1.
The results are those of a loading path (the macroscopic stress and strain in time),
with the mean damage, the regularized area of the cracks and the energies; the strength
of the cell (its largest stress) and its work of fracture are quantities for the
statistics of representative volumes (PhaseFieldResult.strength(),
PhaseFieldResult.work()).
In finite strains (rubbers, foams, soft tissues): morphogen.fracture.
phasefield_finite, the same damage problem with the energy of the deformation gradient
and the nonlinear FFT solver for the displacement.
- morphogen.fracture.phasefield.MODELS = ('AT2', 'AT1')¶
(c_w, the local term) of
gamma = (w(d) / l + l |grad d|^2) / c_w- Type:
Models of the crack density
- morphogen.fracture.phasefield.SPLITS = ('spectral', 'volumetric', 'none')¶
Splits of the energy that drives the damage
- class morphogen.fracture.phasefield.PhaseField(toughness: float | dict, length: float, model: str = 'AT2', split: str = 'spectral', residual: float = 1e-06, irreversibility: str = 'bound')[source]¶
Bases:
objectThe fracture of the phases of a sample:
toughness, the critical energy release rateG_cof all the phases (one value) or of each (a mapping phase ->G_c; the phases left out do not break), the regularizationlengthl(at least two voxels), themodel(MODELS), thesplitof the driving energy (SPLITS), theresidualstiffnesskof a broken voxel and theirreversibility:"bound"(the damage of each node never decreases) or"history"(the largest driving energy reached; see the module).- classmethod from_config(spec) PhaseField[source]¶
A mapping of the parameters of
PhaseField.
- class morphogen.fracture.phasefield.PhaseFieldResult(name: str, bc: str, path: LoadingPath, times: ndarray, strain: ndarray, stress: ndarray, temperature: ndarray, plastic: ndarray, phases: dict = <factory>, iterations: ndarray = <factory>, info: dict = <factory>, fields: dict = <factory>, support: object | None = None, kinematics: str = 'small', cauchy: ndarray | None = None, log_strain: ndarray | None = None, porosity: ndarray | None = None, series: dict = <factory>, scalars: dict = <factory>, orientation: ndarray | None = None, damage: ndarray | None = None, crack_surface: ndarray | None = None, elastic_energy: ndarray | None = None, fracture_energy: ndarray | None = None)[source]¶
Bases:
PathResultThe response of a cell that breaks along a loading path: a
PathResult(the macroscopicstrainandstressat thetimes; its variableplasticis the mean damage of the phases that break, as that of each phase inphases) with thedamage(the same), the regularized area of the cracks per unit volume (crack_surface), the elastic and fracture energies per unit volume, and the alternate minimizations of each step (iterations). Withfields: the damage, strain and stress of the voxels at chosen times.strength()is the strength of the cell andwork()the work done on it: with the stored energy released, more than the fracture energy when the cell breaks at once (the cracks of a brittle cell run faster than its loading).
- class morphogen.fracture.phasefield.PhaseFieldProblem(image: VoxelImage, materials: dict[int, Material], fracture, *, initial=None, grains=None, orientations=None, tol: float = 1e-07, staggered_tol: float = 0.001, max_staggered: int = 200)[source]¶
Bases:
objectThe brittle fracture of a periodic voxel image (see the module):
materialsof its phases (elastic; the crystals turned by the orientations of theirgrains),fracture(PhaseField, or a mapping of its parameters),initial: the voxels broken from the start (a boolean array of the shape of the image).tol: the relative residual of the conjugate gradients;staggered_tol: the largest change of the damage that ends the alternate minimization of a step, of at mostmax_staggerediterations.- fixed¶
those of the voxels broken from the start
- Type:
the nodes held at d = 1
- degradation(d: ndarray | None = None) ndarray[source]¶
gof each voxel for the nodal damaged(the current one by default): 1 in the phases that do not break.
- crack_surface(d: ndarray | None = None, weight: ndarray | None = None) float[source]¶
The regularized area of the cracks,
int gamma(d, grad d) dVover the phases that break; withweight(the toughness of the voxels), the fracture energy.
- run(path, *, name: str = 'path', fields=False, monitor=None) PhaseFieldResult[source]¶
Follow a loading path (
LoadingPath, or a mapping for it; its components prescribed as strains or as stresses, with a change of temperature):PhaseFieldResult.fields: true for the damage, strain and stress of the voxels at the end of each segment of the path,"steps"at the end of each step, or a list of times of its steps.monitor(time, records so far)after each step; a true value ends the path there.
- morphogen.fracture.phasefield.phase_field_fracture(image: VoxelImage, materials: dict[int, Material], fracture, path, *, name: str = 'path', fields=False, kinematics: str = 'small', **options) PhaseFieldResult[source]¶
The response of a periodic voxel image that breaks along a loading path: a
PhaseFieldProblem(itsoptions:initial,grains,orientations,tol,staggered_tol,max_staggered) run alongpath; withkinematics="finite", in finite strains (FinitePhaseFieldProblem: hyperelastic or elastic phases, nominal strains and stresses, its options).