"""Adaptive Self-consistent Clustering Analysis (ASCA).
This module includes the implementation of the Adaptive Self-consistent
Clustering Analysis (ASCA), a clustering-based reduced-order model proposed by
Ferreira et. al (2022) [#]_. Under infinitesimal strains, the Self-consistent
Clustering Analysis (SCA) proposed by Liu et. al (2016) [#]_ is recovered in
the absence of clustering adaptivity.
.. [#] Ferreira, B.P., Andrade Pires, F.M. and Bessa, M.A. (2022).
*Adaptivity for clustering-based reduced-order modeling of
localized history-dependent phenomena.* Comp Methods Appl M, 393
(see `here <https://www.sciencedirect.com/science/article/pii/
S0045782522000895?via%3Dihub>`_)
.. [#] Liu, Z., Bessa, M., and Liu, W.K. (2016).
*Self-consistent clustering analysis: An efficient multi-scale scheme
for inelastic heterogeneous materials.* Comp Methods Appl M, 396:319-341
(see `here <https://www.sciencedirect.com/science/article/pii/
S0045782516301499>`_)
The finite strain extension compatible with multiplicative kinematics of both
SCA and ASCA methods is also available as proposed by Ferreira (2022) [#]_
(see Sections 4.7-4.9). However, the development of an accurate self-consistent
scheme compatible with such as a formulation is still under investigation (see
Section 4.9).
.. [#] Ferreira, B.P. (2022). *Towards Data-driven Multi-scale
Optimization of Thermoplastic Blends: Microstructural
Generation, Constitutive Development and Clustering-based
Reduced-Order Modeling.* PhD Thesis, University of Porto
(see `here <https://repositorio-aberto.up.pt/handle/10216/
146900?locale=en>`_)
Besides the main class that implements the aforementioned methods, this
module also includes a class associated with the reference (fictitious)
homogeneous material, which arises in the formulation of the SCA and ASCA,
and an interface to implement any self-consistent scheme, required to determine
the properties of such a reference material.
Classes
-------
ASCA
Adaptive Self-Consistent Clustering Analysis (ASCA).
ReferenceMaterialOptimizer(ABC)
Elastic reference material properties optimizer interface.
InfinitesimalRegressionSCS(ReferenceMaterialOptimizer)
Infinitesimal strains format regression-based self-consistent scheme.
"""
#
# Modules
# =============================================================================
# Standard
from abc import ABC, abstractmethod
import time
import copy
# Third-party
import numpy as np
import numpy.matlib
import scipy.linalg
# Local
import ioput.info as info
import tensor.matrixoperations as mop
import tensor.tensoroperations as top
from clustering.citoperations import assemble_cit
from online.loading.macloadincrem import LoadingPath, IncrementRewinder, \
RewindManager
from clustering.adaptivity.crve_adaptivity import AdaptivityManager, \
ClusteringAdaptivityOutput
from ioput.incoutputfiles.homresoutput import HomResOutput
from ioput.incoutputfiles.efftanoutput import EffTanOutput
from ioput.incoutputfiles.refmatoutput import RefMatOutput
from ioput.miscoutputfiles.vtkoutput import VTKOutput
from ioput.miscoutputfiles.voxelsoutput import VoxelsOutput
#
# Authorship & Credits
# =============================================================================
__author__ = 'Bernardo Ferreira (bernardo_ferreira@brown.edu)'
__credits__ = ['Bernardo Ferreira', ]
__status__ = 'Stable'
# =============================================================================
#
# =============================================================================
[docs]class ASCA:
"""Adaptive Self-Consistent Clustering Analysis (ASCA).
The detailed formulation of this method can be found in
Ferreira et. al (2022) [#]_ and also in Ferreira (2022) [#]_.
.. [#] Ferreira, B.P., Andrade Pires, F.M. and Bessa, M.A. (2022).
*Adaptivity for clustering-based reduced-order modeling of
localized history-dependent phenomena.* Comp Methods Appl M, 393
(see `here <https://www.sciencedirect.com/science/article/pii/
S0045782522000895?via%3Dihub>`_)
.. [#] Ferreira, B.P. (2022). *Towards Data-driven Multi-scale
Optimization of Thermoplastic Blends: Microstructural
Generation, Constitutive Development and Clustering-based
Reduced-Order Modeling.* PhD Thesis, University of Porto
(see `here <https://repositorio-aberto.up.pt/handle/10216/
146900?locale=en>`_)
Attributes
----------
_n_dim : int
Problem number of spatial dimensions.
_comp_order_sym : list[str]
Strain/Stress components symmetric order.
_comp_order_nsym : list[str]
Strain/Stress components nonsymmetric order.
_global_strain_old_mf : numpy.ndarray (1d)
Last converged global vector of clusters strains stored in matricial
form.
_farfield_strain_old_mf : numpy.ndarray (1d), default=None
Last converged far-field strain tensor (matricial form).
_total_time : float
Total time (s) associated with online-stage.
_effective_time : float
Total time (s) associated with the solution of the equilibrium problem.
_post_process_time : float
Total time (s) associated with post-processing operations.
Methods
-------
get_time_profile(self)
Get time profile of online-stage.
solve_equilibrium_problem(self, crve, material_state, mac_load, \
mac_load_presctype, mac_load_increm, \
output_dir, problem_name='problem', \
clust_adapt_freq=None, \
is_solution_rewinding=False, \
rewind_state_criterion=None, \
rewinding_criterion=None, max_n_rewinds=1, \
is_clust_adapt_output=False, \
is_ref_material_output=False, \
is_vtk_output=False, \
vtk_data=None, is_voxels_output=False)
Solve clustering-based reduced-order equilibrium problem.
_init_global_strain_mf(self, crve, material_state, mode='last_converged')
Set clusters strains initial iterative guess.
_init_global_inc_strain_mf(self, n_total_clusters, mode='last_converged')
Set clusters incremental strains initial iterative guess.
_init_farfield_strain_mf(self, mode='last_converged')
Set far-field strain initial iterative guess.
_init_inc_farfield_strain_mf(self, mode='last_converged')
Set incremental far-field strain initial iterative guess.
_build_residual(self, crve, material_state, presc_strain_idxs, \
presc_stress_idxs, applied_mac_load_mf, ref_material, \
global_cit_mf, global_strain_mf, farfield_strain_mf=None, \
applied_mix_strain_mf=None, applied_mix_stress_mf=None)
Build Lippmann-Schwinger equilibrium residuals.
_build_jacobian(self, crve, material_state, presc_strain_idxs, \
presc_stress_idxs, global_cit_diff_tangent_mf)
Build Lippmann-Schwinger equilibrium Jacobian matrix.
_build_global_cit_diff_tangent_mf(self, crve, global_cit_mf, \
material_state, ref_material)
Build global cluster interaction - tangent modulus matrix.
_check_convergence(self, crve, material_state, presc_strain_idxs, \
presc_stress_idxs, applied_mac_load_mf, residual, \
applied_mix_strain_mf=None)
Check Lippmann-Schwinger equilibrium convergence.
_crve_effective_tangent_modulus(self, crve, material_state, \
global_cit_diff_tangent_mf, \
global_strain_mf=None, \
farfield_strain_mf=None)
CRVE tangent modulus and clusters strain concentration tensors.
_validate_csct(self, material_phases, phase_clusters, global_csct_mf, \
global_strain_mf, farfield_strain_mf)
Validate clusters strain concentration tensors computation.
_init_clusters_sct(self, material_phases, phase_clusters)
Initialize cluster strain concentration tensors.
_build_clusters_residuals(self, material_phases, phase_clusters, residual)
Build clusters equilibrium residuals dictionary.
_display_inc_data(mac_load_path)
Display loading increment data.
_display_scs_iter_data(ref_material, is_lock_prop_ref, mode='init', \
scs_iter_time=None)
Display reference material self-consistent scheme iteration data.
_display_nr_iter_data(mode='init', nr_iter=None, nr_iter_time=None, \
errors=[])
Display Newton-Raphson iteration data.
_set_output_files(self, output_dir, crve, problem_name='problem', \
is_clust_adapt_output=False, \
is_ref_material_output=None, \
is_vtk_output=False, vtk_data=None, \
is_voxels_output=None)
Create and initialize output files.
"""
[docs] def __init__(self, strain_formulation, problem_type,
self_consistent_scheme='regression', scs_parameters=None,
scs_max_n_iterations=20, scs_conv_tol=1e-4,
max_n_iterations=12, conv_tol=1e-6, max_subinc_level=5,
max_cinc_cuts=5, is_adapt_repeat_inc=True):
"""Constructor.
Parameters
----------
strain_formulation: {'infinitesimal', 'finite'}
Problem strain formulation.
problem_type : int
Problem type: 2D plane strain (1), 2D plane stress (2),
2D axisymmetric (3) and 3D (4).
self_consistent_scheme : {'regression',}, default='regression'
Self-consistent scheme to update the elastic reference material
properties.
scs_parameters : {dict, None}, default=None
Self-consistent scheme parameters (key, str; item,
{int, float, bool}).
scs_max_n_iterations : int, default=20
Self-consistent scheme maximum number of iterations.
scs_conv_tol : float, default=1e-4
Self-consistent scheme convergence tolerance.
max_n_iterations : int, default=12
Newton-Raphson maximum number of iterations.
conv_tol : float, default=1e-6
Newton-Raphson convergence tolerance.
max_subinc_level : int, default=5
Maximum level of loading subincrementation.
max_cinc_cuts : int, default=5
Maximum number of consecutive increment cuts.
is_adapt_repeat_inc : bool, default=False
True if loading increment is to be repeated after a clustering
adaptivity step, False otherwise.
"""
self._strain_formulation = strain_formulation
self._problem_type = problem_type
self._self_consistent_scheme = self_consistent_scheme
self._scs_parameters = scs_parameters
self._scs_max_n_iterations = scs_max_n_iterations
self._scs_conv_tol = scs_conv_tol
self._max_n_iterations = max_n_iterations
self._conv_tol = conv_tol
self._max_subinc_level = max_subinc_level
self._max_cinc_cuts = max_cinc_cuts
# Get problem type parameters
n_dim, comp_order_sym, comp_order_nsym = \
mop.get_problem_type_parameters(problem_type)
self._n_dim = n_dim
self._comp_order_sym = comp_order_sym
self._comp_order_nsym = comp_order_nsym
# Initialize last converged algorithmic variables
self._global_strain_old_mf = None
self._farfield_strain_old_mf = None
# Initialize times
self._total_time = 0.0
self._effective_time = 0.0
self._post_process_time = 0.0
# -------------------------------------------------------------------------
[docs] def get_time_profile(self):
"""Get time profile of online-stage.
Returns
-------
total_time : float
Total time (s) associated with online-stage.
effective_time : float
Total time (s) associated with the solution of the equilibrium
problem.
post_process_time : float
Total time (s) associated with post-processing operations.
"""
return self._total_time, self._effective_time, self._post_process_time
# -------------------------------------------------------------------------
[docs] def solve_equilibrium_problem(self, crve, material_state, mac_load,
mac_load_presctype, mac_load_increm,
output_dir, problem_name='problem',
clust_adapt_freq=None,
is_solution_rewinding=False,
rewind_state_criterion=None,
rewinding_criterion=None, max_n_rewinds=1,
is_clust_adapt_output=False,
is_ref_material_output=False,
is_vtk_output=False,
vtk_data=None, is_voxels_output=False):
"""Solve clustering-based reduced-order equilibrium problem.
The overall solution procedure of the Self-consistent Clustering
Analysis (SCA) under infinitesimal strains is summarized in Boxes C.2
(Newton-Raphson iterative scheme) and C.3 (self-consistent iterative
scheme) of Ferreira (2022) [#]_. The finite strain extension compatible
with multiplicative kinematics can be found in Box 4.2 (Newton-Raphson
iterative scheme) and the enrichement with clustering adaptivity
(Adaptive Self-consistent Clustering Analysis (ASCA)) is described in
Section 4.4.
.. [#] Ferreira, B.P. (2022). *Towards Data-driven Multi-scale
Optimization of Thermoplastic Blends: Microstructural
Generation, Constitutive Development and Clustering-based
Reduced-Order Modeling.* PhD Thesis, University of Porto
(see `here <https://repositorio-aberto.up.pt/handle/10216/
146900?locale=en>`_)
----
Parameters
----------
crve : CRVE
Cluster-Reduced Representative Volume Element.
material_state : MaterialState
CRVE material constitutive state.
mac_load : dict
For each loading nature type (key, {'strain', 'stress'}), stores
the loading constraints for each loading subpath in a
numpy.ndarray (2d), where the i-th row is associated with the i-th
strain/stress component and the j-th column is associated with the
j-th loading subpath.
mac_load_presctype : numpy.ndarray (2d)
Loading nature type ({'strain', 'stress'}) associated with each
loading constraint (numpy.ndarray of shape
(n_comps, n_load_subpaths)), where the i-th row is associated with
the i-th strain/stress component and the j-th column is associated
with the j-th loading subpath.
mac_load_increm : dict
For each loading subpath id (key, str), stores a numpy.ndarray of
shape (n_load_increments, 2) where each row is associated with a
prescribed loading increment, and the columns 0 and 1 contain the
corresponding incremental load factor and incremental time,
respectively.
output_dir : str
Absolute directory path of output files.
problem_name : str, default='problem'
Problem name.
clust_adapt_freq : dict, default=None
Clustering adaptivity frequency (relative to loading
incrementation) (item, int) associated with each adaptive
cluster-reduced material phase (key, str).
is_solution_rewinding : bool, default=False
Problem solution rewinding flag.
rewind_state_criterion : tuple, default=None
Rewind state storage criterion [0] and associated parameter [1].
rewinding_criterion : tuple, default=None
Rewinding criterion [0] and associated parameter [1].
max_n_rewinds : int, default=1
Maximum number of rewind operations.
is_clust_adapt_output : bool, default=False
Clustering adaptivity output flag.
is_ref_material_output : bool, default=False
Reference material output flag.
is_vtk_output : bool, default=False
VTK output flag.
vtk_data : dict, default=None
VTK output file parameters.
is_voxels_output : bool
Voxels output file flag.
"""
#
# Initializations
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Set online-stage initial time
init_time = time.time()
# Initialize online-stage post-processing time
self._post_process_time = 0.0
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Set default strain/stress components order according to problem
# strain formulation
if self._strain_formulation == 'infinitesimal':
comp_order = self._comp_order_sym
elif self._strain_formulation == 'finite':
comp_order = self._comp_order_nsym
else:
raise RuntimeError('Unknown problem strain formulation.')
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Initialize loading increment cut flag
is_inc_cut = False
# Initialize improved cluster incremental strains initial iterative
# guess flag
is_improved_init_guess = False
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Initialize global vector of clusters strain tensors and far-field
# strain tensor
if self._strain_formulation == 'infinitesimal':
# Set clusters infinitesimal strain tensors
global_strain_mf = np.zeros(
(crve.get_n_total_clusters()*len(self._comp_order_sym)))
# Set far-field strain tensor
farfield_strain_mf = np.zeros(len(self._comp_order_sym))
else:
# Set initialized deformation gradient (matricial form)
def_gradient_mf = np.array([1.0 if x[0] == x[1] else 0.0
for x in self._comp_order_nsym])
# Build clusters deformation gradients
global_strain_mf = np.tile(def_gradient_mf,
crve.get_n_total_clusters())
# Set far-field strain tensor
farfield_strain_mf = copy.deepcopy(def_gradient_mf)
# Initialize last converged algorithmic variables
self._global_strain_old_mf = copy.deepcopy(global_strain_mf)
self._farfield_strain_old_mf = copy.deepcopy(farfield_strain_mf)
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Initialize clustering adaptivity flag
is_crve_adaptivity = False
adaptivity_manager = None
if len(crve.get_adapt_material_phases()) > 0:
# Switch on clustering adaptivity flag
is_crve_adaptivity = True
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Set flag that controls if the macroscale loading increment where
# the clustering adaptivity is triggered is to be repeated
# considering the new clustering
is_adapt_repeat_inc = True
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Set clustering adaptivity frequency default
if clust_adapt_freq is None:
clust_adapt_freq = {mat_phase: 1 for mat_phase
in crve.get_adapt_material_phases()}
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Initialize clustering adaptivity manager
adaptivity_manager = AdaptivityManager(
self._strain_formulation, self._problem_type,
crve.get_adapt_material_phases(), crve.get_phase_clusters(),
crve.get_adaptivity_control_feature(),
crve.get_adapt_criterion_data(), clust_adapt_freq)
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Initialize increment rewinder manager
if is_solution_rewinding:
# Initialize increment rewinder manager
rewind_manager = RewindManager(
rewind_state_criterion=rewind_state_criterion,
rewinding_criterion=rewinding_criterion,
max_n_rewinds=max_n_rewinds)
# Initialize increment rewinder flag
is_inc_rewinder = False
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Initialize clusters state variables
material_state.init_clusters_state()
# Initialize clusters strain concentration tensors
clusters_sct_mf = self._init_clusters_sct(
material_state.get_material_phases(), crve.get_phase_clusters())
clusters_sct_old_mf = copy.deepcopy(clusters_sct_mf)
#
# Output files
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Set post-processing procedure initial time
procedure_init_time = time.time()
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Create and initialize output files
hres_output, efftan_output, ref_mat_output, voxels_output, \
adapt_output, vtk_output = self._set_output_files(
output_dir, crve, problem_name=problem_name,
is_clust_adapt_output=is_clust_adapt_output,
is_ref_material_output=is_ref_material_output,
is_vtk_output=is_vtk_output, vtk_data=vtk_data,
is_voxels_output=is_voxels_output)
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
if is_vtk_output:
# Write VTK file associated with the initial state
info.displayinfo('18')
vtk_output.write_vtk_file_time_step(
0, self._strain_formulation, self._problem_type, crve,
material_state, vtk_vars=vtk_data['vtk_vars'],
adaptivity_manager=adaptivity_manager)
# Get VTK output increment divider
vtk_inc_div = vtk_data['vtk_inc_div']
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Increment post-processing time
self._post_process_time += time.time() - procedure_init_time
#
# Elastic reference material
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Instantiate elastic reference material
ref_material = ElasticReferenceMaterial(self._strain_formulation,
self._problem_type,
self._self_consistent_scheme,
self._scs_conv_tol)
# Initialize initial value of elastic reference material properties
scs_init_properties = None
# Get initial value of elastic reference material properties
if (self._scs_parameters is not None) \
and {'E_init', 'v_init'}.issubset(self._scs_parameters.keys()):
# Get properties specifications
spec_1 = self._scs_parameters['E_init']
spec_2 = self._scs_parameters['v_init']
# Get properties
if (spec_1 == 'init_eff_tangent' and spec_2 == 'init_eff_tangent'):
scs_init_properties = \
crve.get_eff_isotropic_elastic_constants()
else:
scs_init_properties = {}
scs_init_properties['E'] = self._scs_parameters['E_init']
scs_init_properties['v'] = self._scs_parameters['v_init']
# Set initial value of elastic reference material properties
ref_material.init_material_properties(
material_state.get_material_phases(),
material_state.get_material_phases_properties(),
material_state.get_material_phases_vf(),
properties=scs_init_properties)
# Initialize reference material elastic properties locking flag
if self._self_consistent_scheme == 'none':
is_lock_prop_ref = True
else:
is_lock_prop_ref = False
#
# Loading path
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Instantiate macroscale loading path
mac_load_path = LoadingPath(
self._strain_formulation, self._problem_type, mac_load,
mac_load_presctype, mac_load_increm,
max_subinc_level=self._max_subinc_level,
max_cinc_cuts=self._max_cinc_cuts)
# Set initial homogenized state
mac_load_path.update_hom_state(material_state.get_hom_strain_mf(),
material_state.get_hom_stress_mf())
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Get increment counter
inc = mac_load_path.get_increm_state()['inc']
# Save macroscale loading increment (converged) state
if is_solution_rewinding and rewind_manager.is_rewind_available() \
and rewind_manager.is_save_rewind_state(inc):
# Set reference rewind time
rewind_manager.update_rewind_time(mode='init')
# Instantiate increment rewinder
inc_rewinder = IncrementRewinder(
rewind_inc=inc, phase_clusters=crve.get_phase_clusters())
# Save loading path state
inc_rewinder.save_loading_path(loading_path=mac_load_path)
# Save material constitutive state
inc_rewinder.save_material_state(material_state)
# Save elastic reference material
inc_rewinder.save_reference_material(ref_material)
# Save clusters strain concentration tensors
inc_rewinder.save_clusters_sct(clusters_sct_mf)
# Save algorithmic variables
inc_rewinder.save_asca_algorithmic_variables(global_strain_mf,
farfield_strain_mf)
# Set increment rewinder flag
is_inc_rewinder = True
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Setup first macroscale loading increment
applied_mac_load_mf, inc_mac_load_mf, n_presc_strain, \
presc_strain_idxs, n_presc_stress, presc_stress_idxs, \
is_last_inc = mac_load_path.new_load_increment()
# Get increment counter
inc = mac_load_path.get_increm_state()['inc']
# Display increment data
type(self)._display_inc_data(mac_load_path)
# Set increment initial time
inc_init_time = time.time()
# ---------------------------------------------------------------------
# Start incremental loading loop
while True:
#
# Clusters strain initial iterative guess
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Set clusters strain initial iterative guess
global_strain_mf = \
self._init_global_strain_mf(crve, material_state)
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Set far-field strain initial iterative guess
farfield_strain_mf = self._init_farfield_strain_mf()
#
# Self-consistent scheme iterative loop
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Initialize self-consistent scheme iteration counter
ref_material.init_scs_iter()
# Display reference material self-consistent scheme iteration data
type(self)._display_scs_iter_data(ref_material, is_lock_prop_ref,
mode='init')
# Set self-consistent scheme iteration initial time
scs_iter_init_time = time.time()
# -----------------------------------------------------------------
# Start self-consistent scheme iterative loop
while True:
#
# Global cluster interaction matrix assembly
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Update cluster interaction tensors with elastic reference
# material properties and assemble global cluster interaction
# matrix
global_cit_mf = assemble_cit(
self._strain_formulation, self._problem_type,
ref_material.get_material_properties(),
material_state.get_material_phases(),
crve.get_phase_n_clusters(), crve.get_phase_clusters(),
crve.get_cit_x_mf()[0], crve.get_cit_x_mf()[1],
crve.get_cit_x_mf()[2])
#
# Newton-Raphson iterative loop
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Initialize Newton-Raphson iteration counter
nr_iter = 0
# Display Newton-Raphson iteration header
type(self)._display_nr_iter_data(mode='init')
# Set Newton-Raphson iteration initial time
nr_iter_init_time = time.time()
# -------------------------------------------------------------
# Start Newton-Raphson iterative loop
while True:
#
# Cluster material state update and
# consistent tangent modulus
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Get clusters incremental strain in matricial form
clusters_inc_strain_mf = \
material_state.get_clusters_inc_strain_mf(
global_strain_mf)
# Perform clusters material state update and compute
# associated consistent tangent modulus
su_fail_state = material_state.update_clusters_state(
clusters_inc_strain_mf)
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Raise increment cut procedure if material cluster state
# update failed
if su_fail_state['is_su_fail']:
# Set increment cut flag
is_inc_cut = True
# Display increment cut
info.displayinfo('11', 'su_fail', su_fail_state)
# Leave Newton-Raphson equilibrium iterative loop
break
#
# Homogenized strain and stress tensors
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Update homogenized strain and stress tensors
material_state.update_state_homogenization()
#
# Global cluster interaction - tangent moduli matrix
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Compute global cluster interaction - tangent moduli
# matrix
global_cit_diff_tangent_mf = \
self._build_global_cit_diff_tangent_mf(
crve, global_cit_mf, material_state, ref_material)
#
# Lippmann-Schwinger equilibrium residuals
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Build Lippmann-Schwinger equilibrium residuals
residual = self._build_residual(
crve, material_state,
presc_strain_idxs, presc_stress_idxs,
applied_mac_load_mf, ref_material, global_cit_mf,
global_strain_mf,
inc_mac_load_mf=inc_mac_load_mf,
farfield_strain_mf=farfield_strain_mf,
farfield_strain_old_mf=self._farfield_strain_old_mf)
#
# Convergence evaluation
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Evaluate Lippmann-Schwinger equilibrium solution
# convergence
is_converged, conv_errors = self._check_convergence(
crve, material_state, presc_strain_idxs,
presc_stress_idxs, applied_mac_load_mf, residual)
# Display Newton-Raphson iteration header
type(self)._display_nr_iter_data(
mode='iter', nr_iter=nr_iter,
nr_iter_time=time.time()-nr_iter_init_time,
errors=conv_errors)
#
# Newton-Raphson iterative flow
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Control Newton-Raphson iteration loop flow
if is_converged:
# Leave Newton-Raphson iterative loop (converged
# solution)
break
elif nr_iter == self._max_n_iterations:
# Raise macroscale increment cut procedure
is_inc_cut = True
# Display increment cut
info.displayinfo('11', 'max_iter',
self._max_n_iterations)
# Leave Newton-Raphson equilibrium iterative loop
break
else:
# Increment iteration counter
nr_iter = nr_iter + 1
# Set Newton-Raphson iteration initial time
nr_iter_init_time = time.time()
#
# Lippmann-Schwinger equilibrium Jacobian matrix
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Build Lippmann-Schwinger equilibrium Jacobian matrix
jacobian = self._build_jacobian(crve, material_state,
presc_strain_idxs,
presc_stress_idxs,
global_cit_diff_tangent_mf)
#
# Lippmann-Schwinger equilibrium solution
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Solve Lippmann-Schwinger equilibrium system of linearized
# equilibrium equations
d_iter = numpy.linalg.solve(jacobian, -residual)
#
# Strains iterative update
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Update clusters incremental strain
global_strain_mf = global_strain_mf + \
d_iter[0:crve.get_n_total_clusters()*len(comp_order)]
# Update far-field strain
farfield_strain_mf = farfield_strain_mf + d_iter[
crve.get_n_total_clusters()*len(comp_order):]
#
# Self-consistent scheme
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# If raising a loading increment cut, leave self-consistent
# iterative loop
if is_inc_cut:
break
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Compute CRVE effective tangent modulus and clusters strain
# concentration tensors
eff_tangent_mf, clusters_sct_mf = \
self._crve_effective_tangent_modulus(
crve, material_state, global_cit_diff_tangent_mf)
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Set post-processing procedure initial time
procedure_init_time = time.time()
# Output reference material associated quantities (.refm file)
if is_ref_material_output:
ref_mat_output.write_file(
inc, ref_material,
material_state.get_hom_strain_mf(),
material_state.get_hom_stress_mf(),
farfield_strain_mf, applied_mac_load_mf['strain'],
eff_tangent_mf=eff_tangent_mf)
# Increment post-processing time
self._post_process_time += \
time.time() - procedure_init_time
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Solve self-consistent minimization problem
if is_lock_prop_ref:
# Skip update of reference material elastic properties
E_ref = ref_material.get_material_properties()['E']
v_ref = ref_material.get_material_properties()['v']
else:
# Compute reference material elastic properties through the
# solution of a self-consistent minimization problem
is_scs_admissible, E_ref, v_ref = \
ref_material.self_consistent_update(
material_state.get_hom_strain_mf(),
material_state.get_hom_strain_old_mf(),
material_state.get_hom_stress_mf(),
material_state.get_hom_stress_old_mf(),
eff_tangent_mf)
# If self-consistent scheme iterative solution is not
# admissible, either accept the current solution (first
# self-consistent scheme iteration) or perform one last
# self-consistent scheme iteration with the last converged
# increment reference material elastic properties
if not is_scs_admissible:
# Display reference material self-consistent scheme
# iteration footer
type(self)._display_scs_iter_data(
ref_material, is_lock_prop_ref, mode='end',
scs_iter_time=time.time() - scs_iter_init_time)
# Elastic reference material properties locking
if ref_material.get_scs_iter() == 0:
# Display locking of reference material elastic
# properties
info.displayinfo('14', 'locked_scs_solution')
# Leave self-consistent scheme iterative loop
# (accepted solution)
break
else:
# Display locking of reference material elastic
# properties
info.displayinfo('14', 'inadmissible_scs_solution')
# If inadmissible self-consistent scheme solution,
# reset reference material elastic properties to
# the last converged increment values
ref_material.reset_material_properties()
# Lock reference material elastic properties
is_lock_prop_ref = True
# Perform one last self-consistent scheme iteration
# with the last converged increment reference
# material elastic properties
type(self)._display_scs_iter_data(
ref_material, is_lock_prop_ref, mode='init')
# Proceed to last self-consistent scheme iteration
continue
#
# Convergence evaluation
# (self-consistent scheme)
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Check self-consistent scheme iterative solution convergence
is_scs_converged = \
ref_material.check_scs_convergence(E_ref, v_ref)
# Display reference material self-consistent scheme iteration
# footer
type(self)._display_scs_iter_data(
ref_material, is_lock_prop_ref, mode='end',
scs_iter_time=time.time()-scs_iter_init_time)
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Self-consistent scheme iterative flow
if is_scs_converged:
# Reset flag that locks reference material elastic
# properties
if self._self_consistent_scheme != 'none':
is_lock_prop_ref = False
# Leave self-consistent scheme iterative loop (converged
# solution)
break
elif ref_material.get_scs_iter() == self._scs_max_n_iterations:
# Display locking of reference material elastic properties
info.displayinfo('14', 'max_scs_iter',
self._scs_max_n_iterations)
# If the maximum number of self-consistent scheme
# iterations is reached without convergence, reset elastic
# reference material properties to last loading increment
# values
ref_material.reset_material_properties()
# Lock reference material elastic properties
is_lock_prop_ref = True
# Perform one last self-consistent scheme iteration with
# the last converged increment reference material elastic
# properties
type(self)._display_scs_iter_data(
ref_material, is_lock_prop_ref, mode='init')
else:
# Update reference material elastic properties
ref_material.update_material_properties(E_ref, v_ref)
# Increment self-consistent scheme iteration counter
ref_material.update_scs_iter()
# Display reference material self-consistent scheme
# iteration data
type(self)._display_scs_iter_data(
ref_material, is_lock_prop_ref, mode='init')
# Set self-consistent scheme iteration initial time
scs_iter_init_time = time.time()
#
# Loading increment cut
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
if is_inc_cut:
# Reset loading increment cut flag
is_inc_cut = False
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Reset elastic reference material properties to last loading
# increment values
ref_material.reset_material_properties()
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Perform loading increment cut and setup new loading increment
applied_mac_load_mf, inc_mac_load_mf, n_presc_strain, \
presc_strain_idxs, n_presc_stress, presc_stress_idxs, \
is_last_inc = mac_load_path.increment_cut(self._n_dim,
comp_order)
# Get increment counter
inc = mac_load_path.get_increm_state()['inc']
# Display increment data
type(self)._display_inc_data(mac_load_path)
# Set increment initial time
inc_init_time = time.time()
# Start new loading increment solution procedures
continue
#
# Clustering adaptivity
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# This section should only be executed if the loading increment
# where the clustering adaptivity condition is triggered is to be
# repeated considering the new clustering
if is_crve_adaptivity and is_adapt_repeat_inc \
and adaptivity_manager.check_inc_adaptive_steps(inc):
# Display increment data
if is_clust_adapt_output:
info.displayinfo('12', crve.get_adaptive_step() + 1)
# Build clusters equilibrium residuals
clusters_residuals_mf = self._build_clusters_residuals(
material_state.get_material_phases(),
crve.get_phase_clusters(), residual)
# Get clustering adaptivity trigger condition and target
# clusters
is_trigger, target_clusters, target_clusters_data = \
adaptivity_manager.get_target_clusters(
crve.get_phase_clusters(), crve.get_voxels_clusters(),
material_state.get_clusters_state(),
material_state.get_clusters_def_gradient_mf(),
material_state.get_clusters_def_gradient_old_mf(),
material_state.get_clusters_state_old(),
clusters_sct_mf, clusters_sct_old_mf,
clusters_residuals_mf, inc,
verbose=is_clust_adapt_output)
# Perform clustering adaptivity if adaptivity condition is
# triggered
if is_trigger:
# Display clustering adaptivity
info.displayinfo('16', 'repeat', inc)
# Set improved initial iterative guess for the clusters
# strain global vector (matricial form) after the
# clustering adaptivity
is_improved_init_guess = True
improved_init_guess = \
[is_improved_init_guess, global_strain_mf]
# Perform clustering adaptivity
adaptivity_manager.adaptive_refinement(
crve, material_state, target_clusters,
target_clusters_data, inc,
improved_init_guess=improved_init_guess,
verbose=is_clust_adapt_output)
# Get improved initial iterative guess for the clusters
# strain global vector (matricial form)
if is_improved_init_guess:
global_strain_mf = improved_init_guess[1]
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Get increment counter
inc = mac_load_path.get_increm_state()['inc']
# Display increment data
type(self)._display_inc_data(mac_load_path)
# Set increment initial time
inc_init_time = time.time()
# Start new loading increment solution
continue
#
# Solution rewinding
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Check rewind operations availability
if is_solution_rewinding and is_inc_rewinder \
and rewind_manager.is_rewind_available():
# Check analysis rewind criteria
is_rewind = rewind_manager.is_rewinding_criteria(
inc, material_state.get_material_phases(),
crve.get_phase_clusters(),
material_state.get_clusters_state())
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Rewind analysis if criteria are met
if is_rewind:
info.displayinfo('17', inc_rewinder.get_rewind_inc())
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Rewind loading path
mac_load_path = inc_rewinder.get_loading_path()
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Rewind material constitutive state
material_state = inc_rewinder.get_material_state(crve)
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Rewind elastic reference material
ref_material = inc_rewinder.get_reference_material()
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Rewind clusters strain concentration tensors
clusters_sct_old_mf = inc_rewinder.get_clusters_sct()
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Rewind algorithmic variables
self._global_strain_old_mf, \
self._farfield_strain_old_mf = \
inc_rewinder.get_asca_algorithmic_variables()
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Set post-processing procedure initial time
procedure_init_time = time.time()
# Rewind output files
inc_rewinder.rewind_output_files(
hres_output, efftan_output, ref_mat_output,
voxels_output, adapt_output, vtk_output)
# Increment post-processing time
self._post_process_time += \
time.time() - procedure_init_time
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Reset clustering adaptive steps
if is_crve_adaptivity:
adaptivity_manager.clear_inc_adaptive_steps(
inc_threshold=inc_rewinder.get_rewind_inc())
# Update total rewind time
rewind_manager.update_rewind_time(mode='update')
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Setup new loading increment
applied_mac_load_mf, inc_mac_load_mf, n_presc_strain, \
presc_strain_idxs, n_presc_stress, presc_stress_idxs, \
is_last_inc = mac_load_path.new_load_increment()
# Get increment counter
inc = mac_load_path.get_increm_state()['inc']
# Display increment data
type(self)._display_inc_data(mac_load_path)
# Set increment initial time
inc_init_time = time.time()
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Start new loading increment solution
continue
#
# Homogenized strain and stress tensors
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Get homogenized strain and stress tensors (matricial form)
hom_strain_mf = material_state.get_hom_strain_mf()
hom_stress_mf = material_state.get_hom_stress_mf()
# Build homogenized strain and stress tensors
hom_strain = mop.get_tensor_from_mf(hom_strain_mf, self._n_dim,
comp_order)
hom_stress = mop.get_tensor_from_mf(hom_stress_mf, self._n_dim,
comp_order)
# Get homogenized strain or stress tensor out-of-plane component
# (output only)
if self._problem_type == 1:
hom_stress_33 = material_state.get_oop_hom_comp()
#
# Increment homogenized results (.hres file)
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Initialize homogenized results dictionary
hom_results = dict()
# Build homogenized results dictionary
hom_results = {'hom_strain': hom_strain, 'hom_stress': hom_stress}
if self._problem_type == 1:
hom_results['hom_stress_33'] = hom_stress_33
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Set post-processing procedure initial time
procedure_init_time = time.time()
# Write increment homogenized results (.hres)
hres_output.write_file(
self._strain_formulation, self._problem_type, mac_load_path,
hom_results, time.time() - init_time - self._post_process_time)
# Write increment CRVE effective tangent modulus (.efftan)
efftan_output.write_file(
self._strain_formulation, self._problem_type, mac_load_path,
eff_tangent_mf)
# Increment post-processing time
self._post_process_time += time.time() - procedure_init_time
#
# Increment VTK file
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Write VTK file associated with the loading increment
if is_vtk_output and inc % vtk_inc_div == 0:
# Set post-processing procedure initial time
procedure_init_time = time.time()
# Write VTK file associated with the converged increment
info.displayinfo('18')
vtk_output.write_vtk_file_time_step(
inc, self._strain_formulation, self._problem_type, crve,
material_state, vtk_vars=vtk_data['vtk_vars'],
adaptivity_manager=adaptivity_manager)
# Increment post-processing time
self._post_process_time += time.time() - procedure_init_time
#
# Converged material state variables
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Update last converged material state variables
material_state.update_converged_state()
# Update last converged clusters strain concentration tensors
clusters_sct_old_mf = copy.deepcopy(clusters_sct_mf)
#
# Converged elastic reference material elastic properties
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Set reference material properties converged in the first loading
# increment
if inc == 1:
ref_material.set_material_properties_scs_init()
# Update converged elastic reference material elastic properties
ref_material.update_converged_material_properties()
#
# Converged algorithmic variables
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Update last converged global vector of clusters strain tensors
self._global_strain_old_mf = copy.deepcopy(global_strain_mf)
# Update last converged far-field strain tensor
self._farfield_strain_old_mf = copy.deepcopy(farfield_strain_mf)
#
# Clustering adaptivity output (.adapt file)
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Update clustering adaptivity output file with converged increment
# data
if is_crve_adaptivity:
# Set post-processing procedure initial time
procedure_init_time = time.time()
# Update clustering adaptivity output file
adapt_output.write_adapt_file(inc, adaptivity_manager, crve,
mode='increment')
# Increment post-processing time
self._post_process_time += time.time() - procedure_init_time
#
# Voxels cluster state based quantities output (.voxout file)
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Update clustering adaptivity output file with converged increment
# data
if is_voxels_output:
# Set post-processing procedure initial time
procedure_init_time = time.time()
# Update clustering adaptivity output file
voxels_output.write_voxels_output_file(
self._n_dim, comp_order, crve,
material_state.get_clusters_state(),
material_state.get_clusters_def_gradient_mf())
# Increment post-processing time
self._post_process_time += time.time() - procedure_init_time
#
# Incremental macroscale loading flow
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Update converged macroscale (homogenized) state
mac_load_path.update_hom_state(material_state.get_hom_strain_mf(),
material_state.get_hom_stress_mf())
# Display converged increment data
if self._problem_type == 1:
info.displayinfo(
'7', 'end', self._strain_formulation, self._problem_type,
hom_strain, hom_stress, time.time() - inc_init_time,
time.time() - init_time, hom_stress_33)
else:
info.displayinfo(
'7', 'end', self._strain_formulation, self._problem_type,
hom_strain, hom_stress, time.time() - inc_init_time,
time.time() - init_time)
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Save macroscale loading increment (converged) state
if is_solution_rewinding and rewind_manager.is_rewind_available() \
and rewind_manager.is_save_rewind_state(inc):
# Set reference rewind time
rewind_manager.update_rewind_time(mode='init')
# Instantiate increment rewinder
inc_rewinder = IncrementRewinder(
rewind_inc=inc, phase_clusters=crve.get_phase_clusters())
# Save loading path state
inc_rewinder.save_loading_path(loading_path=mac_load_path)
# Save material constitutive state
inc_rewinder.save_material_state(material_state)
# Save elastic reference material
inc_rewinder.save_reference_material(ref_material)
# Save clusters strain concentration tensors
inc_rewinder.save_clusters_sct(clusters_sct_mf)
# Save algorithmic variables
inc_rewinder.save_asca_algorithmic_variables(
global_strain_mf, farfield_strain_mf=farfield_strain_mf)
# Set increment rewinder flag
is_inc_rewinder = True
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Return if last loading increment, otherwise setup new loading
# increment
if is_last_inc:
# Set total time associated with online-stage
self._total_time = time.time() - init_time
# Set total time associated with the solution of the
# equilibrium problem
self._effective_time = self._total_time \
- self._post_process_time
# Output clustering adaptivity summary
if is_crve_adaptivity:
info.displayinfo('15', adaptivity_manager, crve,
self._effective_time)
# Finish solution of clustering-based reduced order equilibrium
# problem
return
else:
# Setup new loading increment
applied_mac_load_mf, inc_mac_load_mf, n_presc_strain, \
presc_strain_idxs, n_presc_stress, presc_stress_idxs, \
is_last_inc = mac_load_path.new_load_increment()
# Get increment counter
inc = mac_load_path.get_increm_state()['inc']
# Display increment data
type(self)._display_inc_data(mac_load_path)
# Set increment initial time
inc_init_time = time.time()
# -------------------------------------------------------------------------
[docs] def _init_global_strain_mf(self, crve, material_state,
mode='last_converged'):
"""Set clusters strains initial iterative guess.
Parameters
----------
crve : CRVE
Cluster-Reduced Representative Volume Element.
material_state : MaterialState
CRVE material constitutive state at rewind state.
mode : {'last_converged',}, default='last_converged'
Strategy to set clusters incremental strains initial iterative
guess.
Returns
-------
global_strain_mf : numpy.ndarray (1d)
Global vector of clusters strains stored in matricial form.
"""
# Set strain/stress components order according to problem strain
# formulation
if self._strain_formulation == 'infinitesimal':
comp_order = self._comp_order_sym
elif self._strain_formulation == 'finite':
comp_order = self._comp_order_nsym
else:
raise RuntimeError('Unknown problem strain formulation.')
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Get material phases
material_phases = material_state.get_material_phases()
# Get clusters associated with each material phase
phase_clusters = crve.get_phase_clusters()
# Get total number of clusters
n_total_clusters = crve.get_n_total_clusters()
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Set clusters strain initial iterative guess associated with the last
# converged solution
if mode == 'last_converged':
# Initialize initial iterative guess
global_strain_mf = np.zeros((n_total_clusters*len(comp_order)))
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Get last converged clusters state variables
clusters_state_old = material_state.get_clusters_state_old()
# Get last converged clusters deformation gradient
clusters_def_gradient_old_mf = \
material_state.get_clusters_def_gradient_old_mf()
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Initialize material cluster strain range indexes
i_init = 0
i_end = i_init + len(comp_order)
# Loop over material phases
for mat_phase in material_phases:
# Loop over material phase clusters
for cluster in phase_clusters[mat_phase]:
# Get last converged material cluster infinitesimal strain
# tensor (infinitesimal strains) or deformation gradient
# (finite strains)
if self._strain_formulation == 'infinitesimal':
strain_old_mf = \
clusters_state_old[str(cluster)]['strain_mf']
else:
strain_old_mf = \
clusters_def_gradient_old_mf[str(cluster)]
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Assemble to initial iterative guess
global_strain_mf[i_init:i_end] = strain_old_mf
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Update cluster strain range indexes
i_init = i_init + len(comp_order)
i_end = i_init + len(comp_order)
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
else:
raise RuntimeError('Unavailable strategy to set clusters strains '
'initial iterative guess.')
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Return
return global_strain_mf
# -------------------------------------------------------------------------
[docs] def _init_global_inc_strain_mf(self, n_total_clusters,
mode='last_converged'):
"""Set clusters incremental strains initial iterative guess.
Parameters
----------
n_total_clusters : int
Total number of clusters.
mode : {'last_converged',}, default='last_converged'
Strategy to set clusters incremental strains initial iterative
guess.
Returns
-------
global_inc_strain_mf : numpy.ndarray (1d)
Global vector of clusters incremental strains stored in matricial
form.
"""
if mode == 'last_converged':
# Set incremental initial iterative guess associated with the last
# converged solution
if self._strain_formulation == 'infinitesimal':
# Set clusters infinitesimal strain tensors
global_inc_strain_mf = \
np.zeros((n_total_clusters*len(self._comp_order_sym)))
else:
# Set initialized deformation gradient (matricial form)
def_gradient_mf = np.array([1.0 if x[0] == x[1] else 0.0
for x in self._comp_order_nsym])
# Build clusters deformation gradients
global_inc_strain_mf = np.tile(def_gradient_mf,
n_total_clusters)
else:
raise RuntimeError('Unavailable strategy to set clusters '
'incremental strains initial iterative guess.')
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Return
return global_inc_strain_mf
# -------------------------------------------------------------------------
[docs] def _init_farfield_strain_mf(self, mode='last_converged'):
"""Set far-field strain initial iterative guess.
Parameters
----------
mode : {'last_converged',}, default='last_converged'
Strategy to set incremental far-field strain initial iterative
guess.
Returns
-------
farfield_strain_mf : 1darray, default=None
Incremental far-field strain tensor (matricial form).
"""
if mode == 'last_converged':
# Set initial iterative guess associated with the last converged
# solution
farfield_strain_mf = copy.deepcopy(self._farfield_strain_old_mf)
else:
raise RuntimeError('Unavailable strategy to set far-field strain '
'initial iterative guess.')
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Return
return farfield_strain_mf
# -------------------------------------------------------------------------
[docs] def _init_inc_farfield_strain_mf(self, mode='last_converged'):
"""Set incremental far-field strain initial iterative guess.
Parameters
----------
mode : {'last_converged',}, default='last_converged'
Strategy to set incremental far-field strain initial iterative
guess.
Returns
-------
inc_farfield_strain_mf : numpy.ndarray (1d), default=None
Incremental far-field strain tensor (matricial form).
"""
if mode == 'last_converged':
# Set incremental initial iterative guess associated with the last
# converged solution
if self._strain_formulation == 'infinitesimal':
# Set far-field infinitesimal strain tensor
inc_farfield_strain_mf = np.zeros(len(self._comp_order_sym))
else:
# Set far-field deformation gradient
inc_farfield_strain_mf = \
np.array([1.0 if x[0] == x[1] else 0.0
for x in self._comp_order_nsym])
else:
raise RuntimeError('Unavailable strategy to set incremental '
'far-field strain initial iterative guess.')
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Return
return inc_farfield_strain_mf
# -------------------------------------------------------------------------
[docs] def _build_residual(self, crve, material_state, presc_strain_idxs,
presc_stress_idxs, applied_mac_load_mf, ref_material,
global_cit_mf, global_strain_mf, inc_mac_load_mf=None,
farfield_strain_mf=None, farfield_strain_old_mf=None):
"""Build Lippmann-Schwinger equilibrium residuals.
**Global residual function:**
.. math::
\\boldsymbol{R}_{n+1} =
\\begin{bmatrix}
\\boldsymbol{R}^{(I)}_{n+1} \\\\[5pt]
\\boldsymbol{R}^{(n_{\\text{c}}+1)}_{n+1}
\\end{bmatrix} =
\\begin{bmatrix} \\boldsymbol{0} \\\\[5pt]
\\boldsymbol{0} \\end{bmatrix} \\, , \\qquad
\\forall I = 1,2, \\, \\dots, \\, n_{\\text{c}} \\, ,
where :math:`\\boldsymbol{R}_{n+1}` is the global residual
function, :math:`\\boldsymbol{R}^{(I)}_{n+1}` is the :math:`I` th
material cluster equilibrium residual function,
:math:`\\boldsymbol{R}^{(n_{\\text{c}}+1)}_{n+1}` is the strain
and/or stress loading constraints residual function,
:math:`n_{c}` is the number of material clusters,
and :math:`n+1` denotes the current increment.
----
**Infinitesimal strains (incremental equilibrium formulation, \
incremental primary unknowns):**
*Equilibrium residuals*
.. math::
\\boldsymbol{R}^{(I)}_{n+1} =
\\Delta \\boldsymbol{\\varepsilon}_{\\mu,\\,n+1}^{(I)}
+ \\sum^{n_{\\text{c}}}_{J=1}
\\boldsymbol{\\mathsf{T}}^{(I)(J)} : \\left(
\\Delta \\hat{\\boldsymbol{\\sigma}}_{\\mu,\\,n+1}^{(J)}
- \\boldsymbol{\\mathsf{D}}^{e,\\, 0}:
\\Delta\\boldsymbol{\\varepsilon}_{\\mu,\\,n+1}^{(J)}
\\right)
- \\Delta \\boldsymbol{\\varepsilon}_{\\mu,\\,n+1}^{0}\\, ,
.. math::
\\forall I = 1, \\, \\dots, \\, n_{\\text{c}} \\, ,
where :math:`\\boldsymbol{R}^{(I)}_{n+1}` is the :math:`I` th
material cluster equilibrium residual function,
:math:`\\Delta \\boldsymbol{\\varepsilon}_{\\mu,\\,n+1}^{(I)}`
is the :math:`I` th material cluster incremental infinitesimal
strain tensor, :math:`\\boldsymbol{\\mathsf{T}}^{(I)(J)}` is
the cluster interaction tensor (fourth-order tensor) between
the :math:`I` th and :math:`J` th material clusters,
:math:`\\Delta \\boldsymbol{\\sigma}_{\\mu,\\,n+1}^{(J)}` is
the :math:`J` th material cluster incremental Cauchy stress
tensor (:math:`\\hat{(\\cdot)}` denotes the incremental nature
of the constitutive function),
:math:`\\boldsymbol{\\mathsf{D}}^{e,\\, 0}` is the elastic
tangent modulus of the reference homogeneous material,
:math:`\\Delta \\boldsymbol{\\varepsilon}_{\\mu,\\,n+1}^{0}` is
the incremental far-field infinitesimal strain tensor,
:math:`n_{c}` is the number of material clusters, and
:math:`n+1` denotes the current increment.
*Loading (homogenization-based) strain and/or stress constraints*
.. math::
\\boldsymbol{R}^{(n_{\\text{c}} + 1)}_{n+1} =
\\sum^{n_{\\text{c}}}_{I=1} f^{(I)}
\\Delta \\boldsymbol{\\varepsilon}_{\\mu,\\,n+1}^{(I)}
- \\Delta \\boldsymbol{\\varepsilon}_{n+1} \\, ,
where :math:`\\boldsymbol{R}^{(n_{\\text{c}}+1)}_{n+1}` is the
strain loading constraint residual function, :math:`f^{(I)}` is
the volume fraction of the :math:`I` th material cluster,
:math:`\\Delta \\boldsymbol{\\varepsilon}_{\\mu,\\,n+1}^{(I)}`
is the :math:`I` th material cluster incremental infinitesimal
strain tensor, :math:`\\Delta \\boldsymbol{\\varepsilon}_{n+1}`
is the macroscale incremental Cauchy stress tensor,
:math:`n_{c}` is the number of material clusters, and
:math:`n+1` denotes the current increment.
.. math::
\\boldsymbol{R}^{(n_{\\text{c}} + 1)}_{n+1} =
\\sum^{n_{\\text{c}}}_{I=1} f^{(I)}
\\Delta \\hat{\\boldsymbol{\\sigma}}_{\\mu,\\,n+1}^{(I)}
- \\Delta \\boldsymbol{\\sigma}_{n+1} \\, ,
where :math:`\\boldsymbol{R}^{(n_{\\text{c}}+1)}_{n+1}` is the
stress loading constraint residual function,
:math:`f^{(I)}` is the volume fraction of the :math:`I` th
material cluster,
:math:`\\Delta \\boldsymbol{\\sigma}_{\\mu,\\,n+1}^{(I)}` is
the :math:`I` th material cluster incremental Cauchy stress
tensor (:math:`\\hat{(\\cdot)}` denotes the incremental nature
of the constitutive function),
:math:`\\Delta \\boldsymbol{\\sigma}_{n+1}` is the macroscale
incremental Cauchy stress tensor, :math:`n_{c}` is the number
of material clusters, and :math:`n+1` denotes the current
increment.
----
**Infinitesimal strains (incremental equilibrium formulation, \
total primary unknowns):**
This formulation is mathematically equivalent to the incremental
formulation of the equilibrium problem. Based on the additive
nature of both infinitesimal strain and Cauchy stress tensors, this
functional format is suitable for a computational implementation
where total primary unknowns are adopted.
*Equilibrium residuals*
.. math::
\\begin{multline}
\\boldsymbol{R}^{(I)}_{n+1} =
\\boldsymbol{\\varepsilon}_{\\mu,\\,n+1}^{(I)}
+ \\sum^{n_{\\text{c}}}_{J=1}
\\boldsymbol{\\mathsf{T}}^{(I)(J)} : \\left(
\\hat{\\boldsymbol{\\sigma}}_{\\mu,\\,n+1}^{(J)}
- \\boldsymbol{\\mathsf{D}}^{e,\\, 0}:
\\boldsymbol{\\varepsilon}_{\\mu,\\,n+1}^{(J)} \\right)
- \\boldsymbol{\\varepsilon}_{\\mu,\\,n+1}^{0} \\\\
- \\left( \\boldsymbol{\\varepsilon}_{\\mu,\\,n}^{(I)}
+ \\sum^{n_{\\text{c}}}_{J=1}
\\boldsymbol{\\mathsf{T}}^{(I)(J)} : \\left(
\\boldsymbol{\\sigma}_{\\mu,\\,n}^{(J)}
- \\boldsymbol{\\mathsf{D}}^{e,\\, 0}:
\\boldsymbol{\\varepsilon}_{\\mu,\\,n}^{(J)} \\right)
- \\boldsymbol{\\varepsilon}_{\\mu,\\,n}^{0} \\right)
\\, ,
\\end{multline}
.. math::
\\forall I = 1, \\, \\dots, \\, n_{\\text{c}} \\, ,
where :math:`\\boldsymbol{R}^{(I)}_{n+1}` is the :math:`I` th
material cluster equilibrium residual function,
:math:`\\boldsymbol{\\varepsilon}_{\\mu}^{(I)}` is the
:math:`I` th material cluster infinitesimal strain tensor,
:math:`\\boldsymbol{\\mathsf{T}}^{(I)(J)}` is the cluster
interaction tensor (fourth-order tensor) between the
:math:`I` th and :math:`J` th material clusters,
:math:`\\boldsymbol{\\sigma}_{\\mu}^{(J)}` is the :math:`J` th
material cluster Cauchy stress tensor (:math:`\\hat{(\\cdot)}`
denotes the incremental nature of the constitutive function),
:math:`\\boldsymbol{\\mathsf{D}}^{e,\\, 0}` is the elastic
tangent modulus of the reference homogeneous material,
:math:`\\boldsymbol{\\varepsilon}_{\\mu}^{0}` is the far-field
infinitesimal strain tensor, :math:`n_{c}` is the number of
material clusters, :math:`n+1` denotes the current increment,
and :math:`n` denotes the last converged increment.
*Loading (homogenization-based) strain and/or stress constraints*
.. math::
\\boldsymbol{R}^{(n_{\\text{c}} + 1)}_{n+1} =
\\sum^{n_{\\text{c}}}_{I=1} f^{(I)}
\\boldsymbol{\\varepsilon}_{\\mu,\\,n+1}^{(I)}
- \\boldsymbol{\\varepsilon}_{n+1}
- \\left( \\sum^{n_{\\text{c}}}_{I=1} f^{(I)}
\\boldsymbol{\\varepsilon}_{\\mu,\\,n}^{(I)}
- \\boldsymbol{\\varepsilon}_{n} \\right) \\, ,
where :math:`\\boldsymbol{R}^{(n_{\\text{c}}+1)}_{n+1}` is the
strain loading constraint residual function, :math:`f^{(I)}` is
the volume fraction of the :math:`I` th material cluster,
:math:`\\boldsymbol{\\varepsilon}_{\\mu}^{(I)}` is the
:math:`I` th material cluster infinitesimal strain tensor,
:math:`\\boldsymbol{\\varepsilon}` is the macroscale
infinitesimal strain tensor, :math:`n_{c}` is the number of
material clusters, :math:`n+1` denotes the current increment,
and :math:`n` denotes the last converged increment.
.. math::
\\boldsymbol{R}^{(n_{\\text{c}} + 1)}_{n+1} =
\\sum^{n_{\\text{c}}}_{I=1} f^{(I)}
\\hat{\\boldsymbol{\\sigma}}_{\\mu,\\,n+1}^{(I)}
- \\boldsymbol{\\sigma}_{n+1} - \\left(
\\sum^{n_{\\text{c}}}_{I=1} f^{(I)}
\\boldsymbol{\\sigma}_{\\mu,\\,n}^{(I)}
- \\boldsymbol{\\sigma}_{n} \\right) \\, ,
where :math:`\\boldsymbol{R}^{(n_{\\text{c}}+1)}_{n+1}` is the
stress loading constraint residual function,
:math:`f^{(I)}` is the volume fraction of the :math:`I` th
material cluster, :math:`\\boldsymbol{\\sigma}_{\\mu}^{(I)}` is
the :math:`I` th material cluster Cauchy stress tensor
(:math:`\\hat{(\\cdot)}` denotes the incremental nature of the
constitutive function), :math:`\\boldsymbol{\\sigma}` is the
macroscale incremental Cauchy stress tensor, :math:`n_{c}` is
the number of material clusters, :math:`n+1` denotes the
current increment, and :math:`n` denotes the last converged
increment.
----
**Finite strains (total equilibrium formulation, \
total primary unknowns):**
*Equilibrium residuals*
.. math::
\\boldsymbol{R}^{(I)}_{n+1} =
\\boldsymbol{F}_{\\mu,\\,n+1}^{(I)}
+ \\sum^{n_{\\text{c}}}_{J=1}
\\boldsymbol{\\mathsf{T}}^{(I)(J)} : \\left(
\\hat{\\boldsymbol{P}}_{\\mu,\\,n+1}^{(J)}
- \\boldsymbol{\\mathsf{A}}^{e,\\, 0}:
\\boldsymbol{F}_{\\mu,\\,n+1}^{(J)} \\right)
- \\boldsymbol{F}_{\\mu,\\,n+1}^{0} \\, ,
.. math::
\\forall I = 1, \\, \\dots, \\, n_{\\text{c}} \\, ,
where :math:`\\boldsymbol{R}^{(I)}_{n+1}` is the :math:`I` th
material cluster equilibrium residual function,
:math:`\\boldsymbol{F}_{\\mu,\\,n+1}^{(I)}` is the :math:`I` th
material cluster deformation gradient,
:math:`\\boldsymbol{\\mathsf{T}}^{(I)(J)}` is the cluster
interaction tensor (fourth-order tensor) between the
:math:`I` th and :math:`J` th material clusters,
:math:`\\boldsymbol{P}_{\\mu,\\,n+1}^{(J)}` is the :math:`J` th
material cluster first Piola-Kirchhoff stress tensor
(:math:`\\hat{(\\cdot)}` denotes the incremental nature of the
constitutive function),
:math:`\\boldsymbol{\\mathsf{A}}^{e,\\, 0}` is the elastic
tangent modulus of the reference homogeneous material,
:math:`\\boldsymbol{F}_{\\mu,\\,n+1}^{0}` is the far-field
deformation gradient, :math:`n_{c}` is the number of material
clusters, and :math:`n+1` denotes the current increment.
*Loading (homogenization-based) strain and/or stress constraints*
.. math::
\\boldsymbol{R}^{(n_{\\text{c}} + 1)}_{n+1} =
\\sum^{n_{\\text{c}}}_{I=1} f^{(I)}
\\boldsymbol{F}_{\\mu,\\,n+1}^{(I)}
- \\boldsymbol{F}_{n+1} \\, ,
where :math:`\\boldsymbol{R}^{(n_{\\text{c}}+1)}_{n+1}` is the
strain loading constraint residual function,
:math:`f^{(I)}` is the volume fraction of the :math:`I` th
material cluster, :math:`\\boldsymbol{F}_{\\mu,\\,n+1}^{(I)}`
is the :math:`I` th material cluster deformation gradient,
:math:`\\boldsymbol{F}_{n+1}` is the macroscale deformation
gradient, :math:`n_{c}` is the number of material clusters, and
:math:`n+1` denotes the current increment.
.. math::
\\boldsymbol{R}^{(n_{\\text{c}} + 1)}_{n+1} =
\\sum^{n_{\\text{c}}}_{I=1} f^{(I)}
\\hat{\\boldsymbol{P}}_{\\mu,\\,n+1}^{(I)}
- \\boldsymbol{P}_{n+1}
where :math:`\\boldsymbol{R}^{(n_{\\text{c}}+1)}_{n+1}` is the
stress loading constraint residual function, :math:`f^{(I)}` is
the volume fraction of the :math:`I` th material cluster,
:math:`\\boldsymbol{P}_{\\mu, \\,n+1}^{(I)}` is the
:math:`I` th material cluster first Piola-Kirchhoff stress
tensor (:math:`\\hat{(\\cdot)}` denotes the incremental nature
of the constitutive function), :math:`\\boldsymbol{P}_{n+1}` is
the macroscale first Piola-Kirchhoff stress tensor,
:math:`n_{c}` is the number of material clusters, and
:math:`n+1` denotes the current increment.
----
Parameters
----------
crve : CRVE
Cluster-Reduced Representative Volume Element.
material_state : MaterialState
CRVE material constitutive state.
presc_strain_idxs : list[int]
Prescribed macroscale loading strain components indexes.
presc_stress_idxs : list[int]
Prescribed macroscale loading stress components indexes.
applied_mac_load_mf : dict
For each prescribed loading nature type
(key, {'strain', 'stress'}), stores the current applied macroscale
loading constraints in a numpy.ndarray of shape (n_comps,).
ref_material : ElasticReferenceMaterial
Elastic reference material.
global_cit_mf : numpy.ndarray (2d)
Global cluster interaction matrix. Assembly positions are assigned
according to the order of material_phases (1st) and phase_clusters
(2nd).
global_strain_mf : numpy.ndarray (1d)
Global vector of clusters strain tensors (matricial form).
inc_mac_load_mf : dict, default=None
For each loading nature type (key, {'strain', 'stress'}), stores
the incremental loading constraint matricial form in a
numpy.ndarray of shape (n_comps,).
farfield_strain_mf : numpy.ndarray (1d), default=None
Far-field strain tensor (matricial form).
farfield_strain_old_mf : numpy.ndarray (1d), default=None
Last converged far-field strain tensor (matricial form).
Returns
-------
residual : numpy.ndarray (1d)
Lippmann-Schwinger equilibrium residual vector.
"""
# Set strain/stress components order according to problem strain
# formulation
if self._strain_formulation == 'infinitesimal':
comp_order = self._comp_order_sym
elif self._strain_formulation == 'finite':
comp_order = self._comp_order_nsym
else:
raise RuntimeError('Unknown problem strain formulation.')
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Get material phases
material_phases = material_state.get_material_phases()
# Get clusters associated with each material phase
phase_clusters = crve.get_phase_clusters()
# Get total number of clusters
n_total_clusters = crve.get_n_total_clusters()
# Get clusters state variables
clusters_state = material_state.get_clusters_state()
# Get elastic reference material tangent modulus (matricial form)
ref_elastic_tangent_mf = ref_material.get_elastic_tangent_mf()
# Get homogenized strain and stress tensors (matricial form)
hom_strain_mf = material_state.get_hom_strain_mf()
hom_stress_mf = material_state.get_hom_stress_mf()
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Initialiatize and get last converged variables
if self._strain_formulation == 'infinitesimal':
# Get incremental homogenized strain and stress tensors
# (matricial form)
inc_hom_strain_mf = material_state.get_inc_hom_strain_mf()
inc_hom_stress_mf = material_state.get_inc_hom_stress_mf()
# Initialize last converged global vector of clusters strain
# tensors
global_strain_old_mf = np.zeros_like(global_strain_mf)
# Initialize last converged clusters polarization stress
global_pol_stress_old_mf = np.zeros_like(global_strain_mf)
# Get last converged clusters state variables
clusters_state_old = material_state.get_clusters_state_old()
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Initialize clusters polarization stress
global_pol_stress_mf = np.zeros_like(global_strain_mf)
# Initialize material cluster strain range indexes
i_init = 0
i_end = i_init + len(comp_order)
# Loop over material phases
for mat_phase in material_phases:
# Loop over material phase clusters
for cluster in phase_clusters[mat_phase]:
# Compute material cluster stress (matricial form)
stress_mf = clusters_state[str(cluster)]['stress_mf']
# Get material cluster strain (matricial form)
strain_mf = global_strain_mf[i_init:i_end]
# Add cluster polarization stress to global array
global_pol_stress_mf[i_init:i_end] = stress_mf - \
np.matmul(ref_elastic_tangent_mf, strain_mf)
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Compute last converged polarization stress
if self._strain_formulation == 'infinitesimal':
# Compute material cluster last converged stress
# (matricial form)
stress_old_mf = \
clusters_state_old[str(cluster)]['stress_mf']
# Get last converged material cluster strain
# (matricial form)
strain_old_mf = \
clusters_state_old[str(cluster)]['strain_mf']
global_strain_old_mf[i_init:i_end] = strain_old_mf
# Add last converged cluster polarization stress to global
# array
global_pol_stress_old_mf[i_init:i_end] = stress_old_mf \
- np.matmul(ref_elastic_tangent_mf, strain_old_mf)
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Update cluster strain range indexes
i_init = i_init + len(comp_order)
i_end = i_init + len(comp_order)
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Initialize residual vector
residual = np.zeros(n_total_clusters*len(comp_order)
+ len(comp_order))
# Compute clusters equilibrium residuals
residual[0:n_total_clusters*len(comp_order)] = \
np.subtract(
np.add(global_strain_mf,
np.matmul(global_cit_mf, global_pol_stress_mf)),
numpy.matlib.repmat(farfield_strain_mf, 1,
n_total_clusters))
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Add residual term for compatibility with incremental formulation
# under infinitesimal strains
if self._strain_formulation == 'infinitesimal':
residual[0:n_total_clusters*len(comp_order)] += np.reshape(
-np.subtract(
np.add(global_strain_old_mf,
np.matmul(global_cit_mf,
global_pol_stress_old_mf)),
numpy.matlib.repmat(farfield_strain_old_mf, 1,
n_total_clusters)),
(n_total_clusters*len(comp_order),))
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Compute homogenization constraints residuals
for i in range(len(comp_order)):
if i in presc_strain_idxs:
if self._strain_formulation == 'infinitesimal':
# Constraint compatible with incremental formulation
# under infinitesimal strains
residual[n_total_clusters*len(comp_order) + i] = \
inc_hom_strain_mf[i] - inc_mac_load_mf['strain'][i]
else:
residual[n_total_clusters*len(comp_order) + i] = \
hom_strain_mf[i] - applied_mac_load_mf['strain'][i]
else:
if self._strain_formulation == 'infinitesimal':
# Constraint compatible with incremental formulation
# under infinitesimal strains
residual[n_total_clusters*len(comp_order) + i] = \
inc_hom_stress_mf[i] - inc_mac_load_mf['stress'][i]
else:
residual[n_total_clusters*len(comp_order) + i] = \
hom_stress_mf[i] - applied_mac_load_mf['stress'][i]
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Return
return residual
# -------------------------------------------------------------------------
[docs] def _build_jacobian(self, crve, material_state, presc_strain_idxs,
presc_stress_idxs, global_cit_diff_tangent_mf):
"""Build Lippmann-Schwinger equilibrium Jacobian matrix.
*Infinitesimal strains:*
.. math::
\\boldsymbol{J}_{n+1} =
\\dfrac{\\partial \\boldsymbol{R}_{n+1}}{\\partial \\Delta
\\boldsymbol{\\varepsilon}_{\\mu,\\,n+1}} = \\begin{bmatrix}
\\dfrac{\\partial \\boldsymbol{R}^{(I)}_{n+1}}{\\partial
\\Delta \\boldsymbol{\\varepsilon}_{\\mu,\\,n+1}^{(K)}} &
\\dfrac{\\partial \\boldsymbol{R}^{(I)}_{n+1}}{\\partial
\\Delta \\boldsymbol{\\varepsilon}_{\\mu,\\,n+1}^{0}} \\\\[8pt]
\\dfrac{\\partial \\boldsymbol{R}^{(n_{\\text{c}} + 1)}_{n+1}
}{\\partial
\\Delta \\boldsymbol{\\varepsilon}_{\\mu,\\,n+1}^{(K)}} &
\\dfrac{\\partial
\\boldsymbol{R}^{(n_{\\text{c}} + 1)}_{n+1}}{\\partial
\\Delta \\boldsymbol{\\varepsilon}_{\\mu,\\,n+1}^{0}}
\\end{bmatrix} \\, , \\qquad \\forall I, \\, K=1, \\, \\dots,
\\, n_{\\text{c}} \\, ,
where :math:`\\boldsymbol{R}_{n+1}` is the global residual function
(assuming incremental equilibrium formulation and incremental
primary unknowns), :math:`\\boldsymbol{R}^{(I)}_{n+1}` is the
:math:`I` th material cluster equilibrium residual function,
:math:`\\boldsymbol{R}^{(n_{\\text{c}}+1)}_{n+1}` is the strain
and/or stress loading constraints residual function,
:math:`\\Delta \\boldsymbol{\\varepsilon}_{\\mu,\\,n+1}^{(K)}` is
the :math:`K` th material cluster incremental infinitesimal strain
tensor,
:math:`\\Delta \\boldsymbol{\\varepsilon}_{\\mu,\\,n+1}^{0}` is the
incremental far-field infinitesimal strain tensor, :math:`n_{c}` is
the number of material clusters, and :math:`n+1` denotes the
current increment.
The partial derivatives are defined as
.. math::
\\dfrac{\\partial \\boldsymbol{R}^{(I)}_{n+1}}{\\partial
\\Delta \\boldsymbol{\\varepsilon}_{\\mu, \\, n+1}^{(K)}} =
\\delta_{(I)(K)} \\boldsymbol{\\mathsf{I}} +
\\boldsymbol{\\mathsf{T}}^{(I)(K)} : \\left(
\\boldsymbol{\\mathsf{D}}^{(K)}_{n+1}
- \\boldsymbol{\\mathsf{D}}^{e,\\, 0} \\right) \\, ,
.. math::
\\dfrac{\\partial \\boldsymbol{R}^{(I)}_{n+1}}{\\partial
\\Delta \\boldsymbol{\\varepsilon}_{\\mu, \\, n+1}^{0}} =
- \\boldsymbol{\\mathsf{I}} \\, ,
.. math::
\\text{strain:} \\; \\dfrac{\\partial
\\boldsymbol{R}^{(n_{\\text{c}}+1)}_{n+1}}{\\partial \\Delta
\\boldsymbol{\\varepsilon}_{\\mu, \\, n+1}^{(K)}} = f^{(K)} \\,
\\boldsymbol{\\mathsf{I}} \\, , \\quad
\\text{stress:} \\; \\dfrac{\\partial
\\boldsymbol{R}^{(n_{\\text{c}}+1)}_{n+1}}{\\partial \\Delta
\\boldsymbol{\\varepsilon}_{\\mu, \\, n+1}^{(K)}} = f^{(K)} \\,
\\boldsymbol{\\mathsf{D}}^{(K)}_{n+1} \\, ,
.. math::
\\dfrac{\\partial \\boldsymbol{R}^{(n_{\\text{c}} + 1)}_{n+1}
}{\\partial
\\Delta \\boldsymbol{\\varepsilon}_{\\mu, \\, n+1}^{0}} =
\\mathbf{0} \\, ,
where :math:`\\delta_{(I)(K)}` is the Kronecker delta,
:math:`\\boldsymbol{\\mathsf{I}}` is the fourth-order identity
tensor, :math:`\\boldsymbol{\\mathsf{T}}^{(I)(K)}` is the cluster
interaction tensor (fourth-order tensor) between the :math:`I` th
and :math:`K` th material clusters,
:math:`\\boldsymbol{\\mathsf{D}}^{(K)}_{n+1}` is the consistent
tangent modulus of the :math:`K` th material cluster,
:math:`\\boldsymbol{\\mathsf{D}}^{e,\\, 0}` is the elastic tangent
modulus of the reference homogeneous material, :math:`f^{(K)}` is
the volume fraction of the :math:`K` th material cluster.
**Remark:** The Jacobian matrix is the same when the residual
functions are derived with respect to the total strains (assuming
incremental equilibrium formulation and total primary unknowns).
----
*Finite strains:*
.. math::
\\boldsymbol{J}_{n+1} =
\\dfrac{\\partial \\boldsymbol{R}_{n+1}}{\\partial
\\boldsymbol{F}_{\\mu,\\,n+1}} = \\begin{bmatrix}
\\dfrac{\\partial \\boldsymbol{R}^{(I)}_{n+1}}{\\partial
\\boldsymbol{F}_{\\mu,\\,n+1}^{(K)}} & \\dfrac{\\partial
\\boldsymbol{R}^{(I)}_{n+1}}{\\partial
\\boldsymbol{F}_{\\mu,\\,n+1}^{0}} \\\\[8pt]
\\dfrac{\\partial
\\boldsymbol{R}^{(n_{\\text{c}} + 1)}_{n+1}}{\\partial
\\boldsymbol{F}_{\\mu,\\,n+1}^{(K)}} & \\dfrac{\\partial
\\boldsymbol{R}^{(n_{\\text{c}} + 1)}_{n+1}}{\\partial
\\boldsymbol{F}_{\\mu,\\,n+1}^{0}}
\\end{bmatrix} \\, , \\qquad \\forall I, \\, K=1, \\, \\dots,
\\, n_{\\text{c}} \\, ,
where :math:`\\boldsymbol{R}_{n+1}` is the global residual
function, :math:`\\boldsymbol{R}^{(I)}_{n+1}` is the :math:`I` th
material cluster equilibrium residual function,
:math:`\\boldsymbol{R}^{(n_{\\text{c}}+1)}_{n+1}` is the strain
and/or stress loading constraints residual function,
:math:`\\boldsymbol{F}_{\\mu,\\,n+1}^{(K)}` is the :math:`K` th
material cluster deformation gradient,
:math:`\\boldsymbol{F}_{\\mu,\\,n+1}^{0}` is the far-field
deformation gradient, :math:`n_{c}` is the number of material
clusters, and :math:`n+1` denotes the current increment.
The partial derivatives are defined as
.. math::
\\dfrac{\\partial \\boldsymbol{R}^{(I)}_{n+1}}{\\partial
\\boldsymbol{F}_{\\mu, \\, n+1}^{(K)}} = \\delta_{(I)(K)}
\\boldsymbol{\\mathsf{I}} + \\boldsymbol{\\mathsf{T}}^{(I)(K)}
: \\left( \\boldsymbol{\\mathsf{A}}^{(K)}_{n+1}
- \\boldsymbol{\\mathsf{A}}^{e,\\, 0} \\right) \\, ,
.. math::
\\dfrac{\\partial \\boldsymbol{R}^{(I)}_{n+1}}{\\partial
\\boldsymbol{F}_{\\mu, \\, n+1}^{0}} =
- \\boldsymbol{\\mathsf{I}} \\, ,
.. math::
\\text{strain:} \\; \\dfrac{\\partial
\\boldsymbol{R}^{(n_{\\text{c}}+1)}_{n+1}}{\\partial
\\boldsymbol{F}_{\\mu, \\, n+1}^{(K)}} = f^{(K)} \\,
\\boldsymbol{\\mathsf{I}} \\, , \\quad
\\text{stress:} \\; \\dfrac{\\partial
\\boldsymbol{R}^{(n_{\\text{c}}+1)}_{n+1}}{\\partial
\\boldsymbol{F}_{\\mu, \\, n+1}^{(K)}} = f^{(K)} \\,
\\boldsymbol{\\mathsf{A}}^{(K)}_{n+1} \\, ,
.. math::
\\dfrac{\\partial \\boldsymbol{R}^{(n_{\\text{c}} + 1)}_{n+1}
}{\\partial \\boldsymbol{F}_{\\mu, \\, n+1}^{0}} = \\mathbf{0}
\\, ,
where :math:`\\delta_{(I)(K)}` is the Kronecker delta,
:math:`\\boldsymbol{\\mathsf{I}}` is the fourth-order identity
tensor, :math:`\\boldsymbol{\\mathsf{T}}^{(I)(K)}` is the cluster
interaction tensor (fourth-order tensor) between the :math:`I` th
and :math:`K` th material clusters,
:math:`\\boldsymbol{\\mathsf{A}}^{(K)}_{n+1}` is the material
consistent tangent modulus of the :math:`K` th material cluster,
:math:`\\boldsymbol{\\mathsf{A}}^{e,\\, 0}` is the elastic tangent
modulus of the reference homogeneous material, :math:`f^{(K)}` is
the volume fraction of the :math:`K` th material cluster.
----
Parameters
----------
crve : CRVE
Cluster-Reduced Representative Volume Element.
material_state : MaterialState
CRVE material constitutive state.
global_cit_diff_tangent_mf : numpy.ndarray (2d)
Global matrix similar to global cluster interaction matrix but
where each cluster interaction tensor is double contracted with the
difference between the associated material cluster consistent
tangent modulus and the elastic reference material tangent modulus.
Returns
-------
jacobian : numpy.ndarray (2d)
Lippmann-Schwinger equilibrium Jacobian matrix.
"""
# Set strain/stress components order according to problem strain
# formulation
if self._strain_formulation == 'infinitesimal':
comp_order = self._comp_order_sym
elif self._strain_formulation == 'finite':
comp_order = self._comp_order_nsym
else:
raise RuntimeError('Unknown problem strain formulation.')
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
_, foid, _, fosym, _, _, _ = top.get_id_operators(self._n_dim)
if self._strain_formulation == 'infinitesimal':
# Set fourth-order symmetric projection tensor (matricial form)
fosym_mf = mop.get_tensor_mf(fosym, self._n_dim, comp_order)
else:
# Set fourth-order identity tensor (matricial form)
foid_mf = mop.get_tensor_mf(foid, self._n_dim, comp_order)
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Get material phases
material_phases = material_state.get_material_phases()
# Get clusters associated with each material phase
phase_clusters = crve.get_phase_clusters()
# Get total number of clusters
n_total_clusters = crve.get_n_total_clusters()
# Get clusters volume fraction
clusters_vf = crve.get_clusters_vf()
# Get material consistent tangent modulus associated with each material
# cluster
clusters_tangent_mf = material_state.get_clusters_tangent_mf()
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Initialize Jacobian matrix
jacobian = np.zeros(2*(n_total_clusters*len(comp_order)
+ len(comp_order),))
# Compute Jacobian matrix component 11
i_init = 0
i_end = n_total_clusters*len(comp_order)
j_init = 0
j_end = n_total_clusters*len(comp_order)
if self._strain_formulation == 'infinitesimal':
jacobian[i_init:i_end, j_init:j_end] = \
scipy.linalg.block_diag(*(n_total_clusters*[fosym_mf, ])) \
+ global_cit_diff_tangent_mf
else:
jacobian[i_init:i_end, j_init:j_end] = \
scipy.linalg.block_diag(*(n_total_clusters*[foid_mf, ])) \
+ global_cit_diff_tangent_mf
# Compute Jacobian matrix component 12
i_init = 0
i_end = n_total_clusters*len(comp_order)
j_init = n_total_clusters*len(comp_order)
j_end = n_total_clusters*len(comp_order) + len(comp_order)
if self._strain_formulation == 'infinitesimal':
jacobian[i_init:i_end, j_init:j_end] = \
numpy.matlib.repmat(-1.0*fosym_mf, n_total_clusters, 1)
else:
jacobian[i_init:i_end, j_init:j_end] = \
numpy.matlib.repmat(-1.0*foid_mf, n_total_clusters, 1)
# Compute Jacobian matrix component 21
for k in range(len(comp_order)):
i = n_total_clusters*len(comp_order) + k
jclst = 0
for mat_phase in material_phases:
for cluster in phase_clusters[mat_phase]:
if k in presc_strain_idxs:
if self._strain_formulation == 'infinitesimal':
f_foid_mf = clusters_vf[str(cluster)]*fosym_mf
else:
f_foid_mf = clusters_vf[str(cluster)]*foid_mf
j_init = jclst*len(comp_order)
j_end = j_init + len(comp_order)
jacobian[i, j_init:j_end] = f_foid_mf[k, :]
else:
vf_tangent_mf = clusters_vf[str(cluster)]\
* clusters_tangent_mf[str(cluster)]
j_init = jclst*len(comp_order)
j_end = j_init + len(comp_order)
jacobian[i, j_init:j_end] = vf_tangent_mf[k, :]
# Increment column cluster index
jclst = jclst + 1
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Return
return jacobian
# -------------------------------------------------------------------------
[docs] def _build_global_cit_diff_tangent_mf(self, crve, global_cit_mf,
material_state, ref_material):
"""Build global cluster interaction - tangent modulus matrix.
*Infinitesimal strains:*
.. math::
\\boldsymbol{\\mathsf{T}}^{(I)(K)} : \\left(
\\boldsymbol{\\mathsf{D}}^{(K)}_{n+1}
- \\boldsymbol{\\mathsf{D}}^{e,\\, 0} \\right) \\, ,
\\qquad \\forall I, \\, K=1, \\, \\dots, \\, n_{\\text{c}} \\, ,
where :math:`\\boldsymbol{\\mathsf{T}}^{(I)(K)}` is the cluster
interaction tensor (fourth-order tensor) between the :math:`I` th and
:math:`K` th material clusters,
:math:`\\boldsymbol{\\mathsf{D}}^{(K)}_{n+1}` is the consistent tangent
modulus of the :math:`K` th material cluster,
:math:`\\boldsymbol{\\mathsf{D}}^{e,\\, 0}` is the elastic tangent
modulus of the reference homogeneous material, :math:`n_{c}` is the
number of material clusters. and :math:`n+1` denotes the current
increment.
----
*Finite strains:*
.. math::
\\boldsymbol{\\mathsf{T}}^{(I)(K)} : \\left(
\\boldsymbol{\\mathsf{A}}^{(K)}_{n+1}
- \\boldsymbol{\\mathsf{A}}^{e,\\, 0} \\right) \\, ,
\\qquad \\forall I, \\, K=1, \\, \\dots, \\, n_{\\text{c}} \\, ,
where :math:`\\boldsymbol{\\mathsf{T}}^{(I)(K)}` is the cluster
interaction tensor (fourth-order tensor) between the :math:`I` th and
:math:`K` th material clusters,
:math:`\\boldsymbol{\\mathsf{A}}^{(K)}_{n+1}` is the material
consistent tangent modulus of the :math:`K` th material cluster,
:math:`\\boldsymbol{\\mathsf{A}}^{e,\\, 0}` is the elastic tangent
modulus of the reference homogeneous material, :math:`n_{c}` is the
number of material clusters. and :math:`n+1` denotes the current
increment.
----
Parameters
----------
crve : CRVE
Cluster-Reduced Representative Volume Element.
global_cit_mf : numpy.ndarray (2d)
Global cluster interaction matrix. Assembly positions are assigned
according to the order of material_phases (1st) and phase_clusters
(2nd).
material_state : MaterialState
CRVE material constitutive state.
ref_material : ElasticReferenceMaterial
Elastic reference material.
Returns
-------
global_cit_diff_tangent_mf : numpy.ndarray (2d)
Global matrix similar to the global cluster interaction matrix but
where each cluster interaction tensor is double contracted with the
difference between the associated material cluster consistent
tangent modulus and the reference material elastic tangent modulus.
"""
# Get material consistent tangent modulus associated with each material
# cluster
clusters_tangent_mf = material_state.get_clusters_tangent_mf()
# Get elastic reference material tangent modulus
ref_elastic_tangent_mf = ref_material.get_elastic_tangent_mf()
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Build list which stores the difference between the material cluster
# consistent tangent modulus (matricial form) and the reference
# material elastic tangent modulus (matricial form)
diff_tangent_mf = list()
# Loop over material phases
for mat_phase in material_state.get_material_phases():
# Loop over material clusters
for cluster in crve.get_phase_clusters()[mat_phase]:
diff_tangent_mf.append(clusters_tangent_mf[str(cluster)]
- ref_elastic_tangent_mf)
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Build global matrix similar to the global cluster interaction matrix
# but where each cluster interaction tensor is double contracted with
# the difference between the associated material cluster consistent
# tangent modulus and the reference material elastic tangent modulus
global_cit_diff_tangent_mf = np.matmul(
global_cit_mf, scipy.linalg.block_diag(*diff_tangent_mf))
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Return
return global_cit_diff_tangent_mf
# -------------------------------------------------------------------------
[docs] def _check_convergence(self, crve, material_state, presc_strain_idxs,
presc_stress_idxs, applied_mac_load_mf, residual,
applied_mix_strain_mf=None):
"""Check Lippmann-Schwinger equilibrium convergence.
Parameters
----------
crve : CRVE
Cluster-Reduced Representative Volume Element.
material_state : MaterialState
CRVE material constitutive state.
presc_strain_idxs : list[int]
Prescribed macroscale loading strain components indexes.
presc_stress_idxs : list[int]
Prescribed macroscale loading stress components indexes.
applied_mac_load_mf : dict
For each prescribed loading nature type
(key, {'strain', 'stress'}), stores the current applied loading
constraints in a numpy.ndarray of shape (n_comps,).
residual : numpy.ndarray (1d)
Lippmann-Schwinger equilibrium residual vector.
applied_mix_strain_mf : numpy.ndarray (1d), default=None
Strain tensor (matricial form) that contains prescribed strain
components and (non-prescribed) homogenized strain components.
Returns
-------
is_converged : bool
True if Lippmann-Schwinger equilibrium iterative solution
converged, False otherwise.
errors : list[float]
List of errors associated with the Lippmann-Schwinger equilibrium
convergence evaluation.
"""
# Initialize convergence flag
is_converged = False
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Set strain/stress components order according to problem strain
# formulation
if self._strain_formulation == 'infinitesimal':
comp_order = self._comp_order_sym
elif self._strain_formulation == 'finite':
comp_order = self._comp_order_nsym
else:
raise RuntimeError('Unknown problem strain formulation.')
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Get total number of clusters
n_total_clusters = crve.get_n_total_clusters()
# Compute number of prescribed loading strain components
n_presc_strain = len(presc_strain_idxs)
# Compute number of prescribed loading stress components
n_presc_stress = len(presc_stress_idxs)
# Get homogenized strain and stress tensors (matricial form)
hom_strain_mf = material_state.get_hom_strain_mf()
hom_stress_mf = material_state.get_hom_stress_mf()
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Set strain and stress normalization factors
if n_presc_strain > 0 and not np.allclose(
applied_mac_load_mf['strain'][tuple([presc_strain_idxs])],
np.zeros(applied_mac_load_mf['strain'][tuple([
presc_strain_idxs])].shape), atol=1e-10):
strain_norm_factor = np.linalg.norm(
applied_mac_load_mf['strain'][tuple([presc_strain_idxs])])
elif not np.allclose(hom_strain_mf, np.zeros(hom_strain_mf.shape),
atol=1e-10):
strain_norm_factor = np.linalg.norm(hom_strain_mf)
else:
strain_norm_factor = 1.0
if n_presc_stress > 0 and not np.allclose(
applied_mac_load_mf['stress'][tuple([presc_stress_idxs])],
np.zeros(applied_mac_load_mf['stress'][tuple([
presc_stress_idxs])].shape), atol=1e-10):
stress_norm_factor = np.linalg.norm(
applied_mac_load_mf['stress'][tuple([presc_stress_idxs])])
elif not np.allclose(hom_stress_mf, np.zeros(hom_stress_mf.shape),
atol=1e-10):
stress_norm_factor = np.linalg.norm(hom_stress_mf)
else:
stress_norm_factor = 1.0
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Compute error associated with the clusters equilibrium residuals
error_1 = \
np.linalg.norm(residual[0:n_total_clusters*len(comp_order)]) \
/ strain_norm_factor
# Compute error associated with the homogenization constraints
# residuals
aux = residual[n_total_clusters*len(comp_order):]
if n_presc_strain > 0:
error_2 = \
np.linalg.norm(aux[presc_strain_idxs])/strain_norm_factor
if n_presc_stress > 0:
error_3 = \
np.linalg.norm(aux[presc_stress_idxs])/stress_norm_factor
# Criterion convergence flag is True if all residual errors
# converged according to the defined convergence tolerance
if n_presc_strain == 0:
error_2 = None
is_converged = (error_1 < self._conv_tol) \
and (error_3 < self._conv_tol)
elif n_presc_stress == 0:
error_3 = None
is_converged = (error_1 < self._conv_tol) \
and (error_2 < self._conv_tol)
else:
is_converged = (error_1 < self._conv_tol) \
and (error_2 < self._conv_tol) \
and (error_3 < self._conv_tol)
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Build convergence evaluation errors list
errors = [error_1, error_2, error_3]
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Return
return is_converged, errors
# -------------------------------------------------------------------------
[docs] def _crve_effective_tangent_modulus(self, crve, material_state,
global_cit_diff_tangent_mf,
global_strain_mf=None,
farfield_strain_mf=None):
"""CRVE tangent modulus and clusters strain concentration tensors.
*Infinitesimal strains:*
.. math::
\\overline{\\boldsymbol{\\mathsf{D}}}_{n+1} =
\\sum^{n_{\\text{c}}}_{I=1} f^{(I)}
\\boldsymbol{\\mathsf{D}}^{(I)}_{n+1} :
\\boldsymbol{\\mathsf{H}}^{(I)}_{n+1} \\, ,
.. math::
\\mathbf{H}^{(I)}_{n+1} = \\sum_{K=1}^{ n_{\\text{c}}}
\\left( \\mathbf{M}^{-1} \\right)_{(I)(K)} \\, ,
\\quad \\forall I = 1,2, \\, \\dots, \\, n_{\\text{c}} \\, ,
.. math::
\\mathbf{M} = \\begin{bmatrix}
\\dfrac{\\partial \\boldsymbol{R}^{(1)}_{n+1}}{\\partial
\\Delta \\boldsymbol{\\varepsilon}_{\\mu,\\,n+1}^{(1)}} & \\dots &
\\dfrac{\\partial \\boldsymbol{R}^{(1)}_{n+1}}{\\partial
\\Delta \\boldsymbol{\\varepsilon}_{\\mu,\\,n+1}^{(n_{\\text{c}})}}
\\\\[10pt] \\vdots & \\ddots & \\vdots \\\\[5pt]
\\dfrac{\\partial \\boldsymbol{R}^{(n_{\\mathrm{c}})}_{n+1}}{
\\partial \\Delta \\boldsymbol{\\varepsilon}_{\\mu,\\,n+1}^{(1)}} &
\\dots & \\dfrac{\\partial
\\boldsymbol{R}^{(n_{\\text{c}})}_{n+1}}{\\partial \\Delta
\\boldsymbol{\\varepsilon}_{\\mu,\\,n+1}^{(n_{\\text{c}})}}
\\end{bmatrix} \\, ,
where :math:`\\overline{\\boldsymbol{\\mathsf{D}}}_{n+1}` is the CRVE
homogenized consistent tangent modulus, :math:`f^{(I)}` is the volume
fraction of the :math:`I` th material cluster,
:math:`\\boldsymbol{\\mathsf{D}}^{(I)}_{n+1}` is the consistent tangent
modulus of the :math:`I` th material cluster,
:math:`\\boldsymbol{\\mathsf{H}}^{(I)}_{n+1}` is the :math:`I` th
material cluster strain concentration tensor,
:math:`\\boldsymbol{R}^{(I)}_{n+1}` is the :math:`I` th material
cluster equilibrium residual function,
:math:`\\Delta \\boldsymbol{\\varepsilon}_{\\mu,\\,n+1}^{(K)}` is the
:math:`K` th material cluster incremental infinitesimal strain tensor,
:math:`n_{c}` is the number of material clusters, and :math:`n+1`
denotes the current increment.
The residual derivatives are defined as
.. math::
\\dfrac{\\partial \\boldsymbol{R}^{(I)}_{n+1}}{\\partial
\\Delta \\boldsymbol{\\varepsilon}_{\\mu, \\, n+1}^{(K)}} =
\\delta_{(I)(K)} \\boldsymbol{\\mathsf{I}} +
\\boldsymbol{\\mathsf{T}}^{(I)(K)} : \\left(
\\boldsymbol{\\mathsf{D}}^{(K)}_{n+1} -
\\boldsymbol{\\mathsf{D}}^{e,\\, 0} \\right) \\, ,
.. math::
\\forall I, K = 1, \\, \\dots, \\, n_{\\text{c}} \\, .
**Remark:** The residual derivatives are the same when the residual
functions are derived with respect to the total strains.
----
*Finite strains:*
.. math::
\\overline{\\boldsymbol{\\mathsf{A}}}_{n+1} =
\\sum^{n_{\\text{c}}}_{I=1} f^{(I)}
\\boldsymbol{\\mathsf{A}}^{(I)}_{n+1} :
\\boldsymbol{\\mathsf{H}}^{(I)}_{n+1} \\, ,
.. math::
\\mathbf{H}^{(I)}_{n+1} = \\sum_{K=1}^{ n_{\\text{c}}} \\left(
\\mathbf{M}^{-1} \\right)_{(I)(K)} \\, , \\quad
\\forall I = 1,2, \\, \\dots, \\, n_{\\text{c}} \\, ,
.. math::
\\mathbf{M} = \\begin{bmatrix}
\\dfrac{\\partial \\boldsymbol{R}^{(1)}_{n+1}}{\\partial
\\boldsymbol{F}_{\\mu,\\,n+1}^{(1)}} & \\dots & \\dfrac{\\partial
\\boldsymbol{R}^{(1)}_{n+1}}{\\partial
\\boldsymbol{F}_{\\mu,\\,n+1}^{(n_{\\text{c}})}} \\\\[10pt]
\\vdots & \\ddots & \\vdots \\\\[5pt] \\dfrac{\\partial
\\boldsymbol{R}^{(n_{\\mathrm{c}})}_{n+1}}{\\partial
\\boldsymbol{F}_{\\mu,\\,n+1}^{(1)}} & \\dots &
\\dfrac{\\partial \\boldsymbol{R}^{(n_{\\text{c}})}_{n+1}}{
\\partial \\boldsymbol{F}_{\\mu,\\,n+1}^{(n_{\\text{c}})}}
\\end{bmatrix} \\, ,
where :math:`\\overline{\\boldsymbol{\\mathsf{A}}}_{n+1}` is the CRVE
homogenized material consistent tangent modulus, :math:`f^{(I)}` is the
volume fraction of the :math:`I` th material cluster,
:math:`\\boldsymbol{\\mathsf{A}}^{(I)}_{n+1}` is the material
consistent tangent modulus of the :math:`I` th material cluster,
:math:`\\boldsymbol{\\mathsf{H}}^{(I)}_{n+1}` is the :math:`I` th
material cluster strain concentration tensor,
:math:`\\boldsymbol{R}^{(I)}_{n+1}` is the :math:`I` th material
cluster equilibrium residual function,
:math:`\\boldsymbol{F}_{\\mu,\\,n+1}^{(K)}` is the :math:`K` th
material cluster deformation gradient, :math:`n_{c}` is the number of
material clusters, and :math:`n+1` denotes the current increment.
The residual derivatives are defined as
.. math::
\\dfrac{\\partial \\boldsymbol{R}^{(I)}_{n+1}}{\\partial
\\boldsymbol{F}_{\\mu, \\, n+1}^{(K)}} = \\delta_{(I)(K)}
\\boldsymbol{\\mathsf{I}} + \\boldsymbol{\\mathsf{T}}^{(I)(K)} :
\\left( \\boldsymbol{\\mathsf{A}}^{(K)}_{n+1} -
\\boldsymbol{\\mathsf{A}}^{e,\\, 0} \\right) \\, ,
.. math::
\\forall I, K = 1, \\, \\dots, \\, n_{\\text{c}} \\, .
----
Parameters
----------
crve : CRVE
Cluster-Reduced Representative Volume Element.
material_state : MaterialState
CRVE material constitutive state.
global_cit_diff_tangent_mf : numpy.ndarray (2d)
Global matrix similar to global cluster interaction matrix but
where each cluster interaction tensor is double contracted with the
difference between the associated material cluster consistent
tangent modulus and the elastic reference material tangent modulus.
global_strain_mf : numpy.ndarray (1d), default=None
Global vector of clusters strains stored in matricial form. Only
required for validation of cluster strain concentration tensors
computation.
farfield_strain_mf : numpy.ndarray (1d), default=None
Far-field strain tensor (matricial form). Only required for
validation of cluster strain concentration tensors computation.
Returns
-------
eff_tangent_mf : numpy.ndarray (2d)
CRVE effective material tangent modulus (matricial form).
clusters_sct_mf : dict
Fourth-order strain concentration tensor (matricial form)
(item, numpy.ndarray (2d)) associated with each material cluster
(key, str).
"""
# Set strain/stress components order according to problem strain
# formulation
if self._strain_formulation == 'infinitesimal':
comp_order = self._comp_order_sym
elif self._strain_formulation == 'finite':
comp_order = self._comp_order_nsym
else:
raise RuntimeError('Unknown problem strain formulation.')
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Set second-order identity tensor
_, foid, _, fosym, _, _, _ = top.get_id_operators(self._n_dim)
fosym_mf = mop.get_tensor_mf(fosym, self._n_dim, comp_order)
foid_mf = mop.get_tensor_mf(foid, self._n_dim, comp_order)
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Get material phases
material_phases = material_state.get_material_phases()
# Get clusters associated with each material phase
phase_clusters = crve.get_phase_clusters()
# Get total number of clusters
n_total_clusters = crve.get_n_total_clusters()
# Get clusters volume fraction
clusters_vf = crve.get_clusters_vf()
# Get material consistent tangent modulus associated with each material
# cluster
clusters_tangent_mf = material_state.get_clusters_tangent_mf()
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Compute equilibrium jacobian matrix (cluster strain concentration
# tensors system of linear equations coefficient matrix)
if self._strain_formulation == 'infinitesimal':
csct_matrix = \
scipy.linalg.block_diag(*(n_total_clusters*[fosym_mf, ])) \
+ global_cit_diff_tangent_mf
else:
csct_matrix = \
scipy.linalg.block_diag(*(n_total_clusters*[foid_mf, ])) \
+ global_cit_diff_tangent_mf
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Select clusters strain concentration tensors computation option:
#
# Option 1 - Solve linear system of equations
#
# Option 2 - Direct computation from inverse of equilibrium Jacobian
# matrix
#
option = 1
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
if option == 1:
# Compute cluster strain concentration tensors system of linear
# equations right-hand side
if self._strain_formulation == 'infinitesimal':
csct_rhs = numpy.matlib.repmat(fosym_mf, n_total_clusters, 1)
else:
csct_rhs = numpy.matlib.repmat(foid_mf, n_total_clusters, 1)
# Initialize system solution matrix (containing clusters strain
# concentration tensors)
global_csct_mf = np.zeros((n_total_clusters*len(comp_order),
len(comp_order)))
# Solve cluster strain concentration tensors system of linear
# equations
for i in range(len(comp_order)):
global_csct_mf[:, i] = numpy.linalg.solve(csct_matrix,
csct_rhs[:, i])
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
elif option == 2:
# Compute inverse of equilibrium jacobian matrix
csct_matrix_inv = numpy.linalg.inv(csct_matrix)
# Initialize system solution matrix (containing clusters strain
# concentration tensors)
global_csct_mf = np.zeros((n_total_clusters*len(comp_order),
len(comp_order)))
# Initialize cluster indexes
i_init = 0
i_end = i_init + len(comp_order)
j_init = 0
j_end = j_init + len(comp_order)
# Loop over material phases
for mat_phase_I in material_phases:
# Loop over material phase clusters
for cluster_I in phase_clusters[mat_phase_I]:
# Loop over material phases
for mat_phase_J in material_phases:
# Loop over material phase clusters
for cluster_J in phase_clusters[mat_phase_J]:
# Add cluster J contribution to cluster I strain
# concentration tensor
global_csct_mf[i_init:i_end, :] += \
csct_matrix_inv[i_init:i_end, j_init:j_end]
# Increment cluster index
j_init = j_init + len(comp_order)
j_end = j_init + len(comp_order)
# Increment cluster indexes
i_init = i_init + len(comp_order)
i_end = i_init + len(comp_order)
j_init = 0
j_end = j_init + len(comp_order)
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Validate cluster strain concentration tensors computation
is_csct_validation = False
if is_csct_validation:
self._validate_csct(material_phases, phase_clusters,
global_csct_mf, global_strain_mf,
farfield_strain_mf)
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Initialize effective tangent modulus
eff_tangent_mf = np.zeros((len(comp_order), len(comp_order)))
# Initialize clusters strain concentration tensors dictionary
clusters_sct_mf = {}
# Initialize cluster index
i_init = 0
i_end = i_init + len(comp_order)
# Loop over material phases
for mat_phase in material_phases:
# Loop over material phase clusters
for cluster in phase_clusters[mat_phase]:
# Get material cluster volume fraction
cluster_vf = clusters_vf[str(cluster)]
# Get material cluster consistent tangent (matricial form)
cluster_tangent_mf = clusters_tangent_mf[str(cluster)]
# Get material cluster strain concentration tensor
cluster_sct_mf = global_csct_mf[i_init:i_end, :]
# Store material cluster strain concentration tensor (matricial
# form)
clusters_sct_mf[str(cluster)] = cluster_sct_mf
# Add material cluster contribution to effective tangent
# modulus
eff_tangent_mf = eff_tangent_mf + \
cluster_vf*np.matmul(cluster_tangent_mf, cluster_sct_mf)
# Increment cluster index
i_init = i_init + len(comp_order)
i_end = i_init + len(comp_order)
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
return eff_tangent_mf, clusters_sct_mf
# -------------------------------------------------------------------------
[docs] def _validate_csct(self, material_phases, phase_clusters, global_csct_mf,
global_strain_mf, farfield_strain_mf):
"""Validate clusters strain concentration tensors computation.
This validation procedure requires the homogenized strain tensor
instead of the far-field strain tensor in the SCA formulation without
the far-field strain tensor.
----
Parameters
----------
material_phases : list[str]
RVE material phases labels (str).
phase_clusters : dict
Clusters labels (item, list[int]) associated with each material
phase (key, str).
global_csct_mf : numpy.ndarray (2d)
Global matrix of cluster strain concentration tensors (matricial
form).
global_strain_mf : numpy.ndarray (1d)
Global vector of clusters strains stored in matricial form.
farfield_strain_mf : numpy.ndarray (1d)
Far-field strain tensor (matricial form).
"""
# Set strain/stress components order according to problem strain
# formulation
if self._strain_formulation == 'infinitesimal':
comp_order = self._comp_order_sym
elif self._strain_formulation == 'finite':
comp_order = self._comp_order_nsym
else:
raise RuntimeError('Unknown problem strain formulation.')
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Initialize cluster index
i_init = 0
i_end = i_init + len(comp_order)
# Loop over material phases
for mat_phase in material_phases:
# Loop over material phase clusters
for cluster in phase_clusters[mat_phase]:
# Get material cluster strain concentration tensor
cluster_sct_mf = global_csct_mf[i_init:i_end, :]
# Compute cluster strain from strain concentration tensor
strain_mf = np.matmul(cluster_sct_mf, farfield_strain_mf)
# Compare cluster strain computed from strain concentration
# tensor with actual cluster strain. Raise error if equality
# comparison fails
if not np.allclose(strain_mf, global_strain_mf[i_init:i_end],
rtol=1e-05, atol=1e-08):
raise RuntimeError('Wrong computation of cluster strain '
'concentration tensor.')
# Increment cluster index
i_init = i_init + len(comp_order)
i_end = i_init + len(comp_order)
# -------------------------------------------------------------------------
[docs] def _init_clusters_sct(self, material_phases, phase_clusters):
"""Initialize cluster strain concentration tensors.
Parameters
----------
material_phases : list[str]
RVE material phases labels (str).
phase_clusters : dict
Clusters labels (item, list[int]) associated with each material
phase (key, str).
Returns
-------
clusters_sct_mf : dict
Fourth-order strain concentration tensor (matricial form)
(item, numpy.ndarray (2d)) associated with each material cluster
(key, str).
"""
# Set strain/stress components order according to problem strain
# formulation
if self._strain_formulation == 'infinitesimal':
comp_order = self._comp_order_sym
elif self._strain_formulation == 'finite':
comp_order = self._comp_order_nsym
else:
raise RuntimeError('Unknown problem strain formulation.')
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Initialize cluster strain concentration tensors dictionary
cluster_sct_mf = {}
# Loop over material phases
for mat_phase in material_phases:
# Loop over material phase clusters
for cluster in phase_clusters[mat_phase]:
# Initialize cluster strain concentration tensor (matricial
# form)
cluster_sct_mf[str(cluster)] = np.zeros((len(comp_order),
len(comp_order)))
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
return cluster_sct_mf
# -------------------------------------------------------------------------
[docs] def _build_clusters_residuals(self, material_phases, phase_clusters,
residual):
"""Build clusters equilibrium residuals dictionary.
This procedure is only carried out so that clusters equilibrium
residuals are conveniently stored to perform post-processing
operations.
----
Parameters
----------
material_phases : list[str]
RVE material phases labels (str).
phase_clusters : dict
Clusters labels (item, list[int]) associated with each material
phase (key, str).
residual : numpy.ndarray (1d)
Lippmann-Schwinger equilibrium residual vector.
Returns
-------
clusters_residuals_mf : dict
Equilibrium residual second-order tensor (matricial form)
(item, numpy.ndarray (1d)) associated with each material cluster
(key, str).
"""
# Set strain/stress components order according to problem strain
# formulation
if self._strain_formulation == 'infinitesimal':
comp_order = self._comp_order_sym
elif self._strain_formulation == 'finite':
comp_order = self._comp_order_nsym
else:
raise RuntimeError('Unknown problem strain formulation.')
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Initialize clusters equilibrium residuals dictionary
clusters_residuals_mf = {}
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Initialize cluster strain range indexes
i_init = 0
i_end = i_init + len(comp_order)
# Loop over material phases
for mat_phase in material_phases:
# Loop over material phase clusters
for cluster in phase_clusters[mat_phase]:
# Store cluster equilibrium residual
clusters_residuals_mf[str(cluster)] = residual[i_init:i_end]
# Update cluster strain range indexes
i_init = i_init + len(comp_order)
i_end = i_init + len(comp_order)
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Return
return clusters_residuals_mf
# -------------------------------------------------------------------------
[docs] @staticmethod
def _display_inc_data(mac_load_path):
"""Display loading increment data.
Parameters
----------
mac_load_path : LoadingPath
Macroscale loading path.
"""
# Get increment counter
inc = mac_load_path.get_increm_state()['inc']
# Get loading subpath data
sp_id, sp_inc, sp_total_lfact, sp_inc_lfact, sp_total_time, \
sp_inc_time, subinc_level = mac_load_path.get_subpath_state()
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Display loading increment data
info.displayinfo('7', 'init', inc, subinc_level, sp_id + 1,
sp_total_lfact, sp_total_time, sp_inc, sp_inc_lfact,
sp_inc_time)
# -------------------------------------------------------------------------
[docs] @staticmethod
def _display_scs_iter_data(ref_material, is_lock_prop_ref, mode='init',
scs_iter_time=None):
"""Display reference material self-consistent scheme iteration data.
Parameters
----------
ref_material : ElasticReferenceMaterial
Elastic reference material.
is_lock_prop_ref : bool
True if elastic reference material properties are locked, False
otherwise.
mode : {'init', 'end'}
Output mode: Self-consistent scheme iteration header (`init`) or
footer (`end`).
scs_iter_time : float
Total self-consistent scheme time (s).
"""
# Get elastic reference material properties
material_properties = ref_material.get_material_properties()
# Get self-consistent scheme iteration counter
scs_iter = ref_material.get_scs_iter()
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Display reference material self-consistent scheme iteration data
if mode == 'end':
# Display reference material self-consistent scheme iteration
# footer
info.displayinfo('8', mode, scs_iter_time)
else:
# Display reference material self-consistent scheme iteration
# header
if is_lock_prop_ref:
info.displayinfo('13', 'init', 0, material_properties['E'],
material_properties['v'])
else:
if scs_iter == 0:
info.displayinfo('8', 'init', scs_iter,
material_properties['E'],
material_properties['v'])
else:
# Get normalized iterative changes of elastic reference
# material properties associated with the last
# self-consistent iteration convergence evaluation
norm_dE = ref_material.get_norm_dE()
norm_dv = ref_material.get_norm_dv()
# Display reference material self-consistent scheme
# iteration header
info.displayinfo('8', 'init', scs_iter,
material_properties['E'], norm_dE,
material_properties['v'], norm_dv)
# -------------------------------------------------------------------------
[docs] @staticmethod
def _display_nr_iter_data(mode='init', nr_iter=None, nr_iter_time=None,
errors=[]):
"""Display Newton-Raphson iteration data.
Parameters
----------
mode : {'init', 'iter'}
Output mode: Newton-Raphson iteration header ('init') or solution
related metrics ('iter').
nr_iter : int
Newton-Raphson iteration counter.
nr_iter_time : float
Total Newton-Raphson iteration time (s).
errors : list[float]
List of errors associated with the Newton-Raphson convergence
evaluation.
"""
if mode == 'iter':
info.displayinfo('9', mode, nr_iter, nr_iter_time, *errors)
else:
info.displayinfo('9', mode)
# -------------------------------------------------------------------------
[docs] def _set_output_files(self, output_dir, crve, problem_name='problem',
is_clust_adapt_output=False,
is_ref_material_output=None, is_vtk_output=False,
vtk_data=None, is_voxels_output=None):
"""Create and initialize output files.
Parameters
----------
output_dir : str
Absolute directory path of output files.
crve : CRVE
Cluster-Reduced Representative Volume Element.
problem_name : str, default='problem'
Problem name.
is_clust_adapt_output : bool, default=False
Clustering adaptivity output flag.
is_ref_material_output : bool, default=False
Reference material output flag.
is_vtk_output : bool, default=False
VTK output flag.
vtk_data : dict, default=None
VTK output file parameters.
is_voxels_output : bool
Voxels output flag.
Returns
-------
hres_output : HomResOutput
Output associated with the homogenized results.
efftan_output : EffTanOutput
Output associated with the CRVE effective tangent modulus.
ref_mat_output : RefMatOutput
Output associated with the reference material.
voxels_output : VoxelsOutput
Output associated with voxels material-related quantities.
adapt_output : ClusteringAdaptivityOutput
Output associated with the clustering adaptivity procedures.
vtk_output : VTKOutput
Output associated with the VTK files.
"""
if output_dir[-1] != '/':
output_dir += '/'
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Set file where homogenized strain/stress results are stored
hres_file_path = output_dir + problem_name + '.hres'
# Instantiate homogenized results output
hres_output = HomResOutput(hres_file_path)
# Write homogenized results output file header
hres_output.init_file(self._strain_formulation)
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Set file where CRVE effective material tangent modulus is stored
efftan_file_path = output_dir + problem_name + '.efftan'
# Instantiate CRVE effective material tangent modulus output
efftan_output = EffTanOutput(efftan_file_path)
# Write homogenized results output file header
efftan_output.init_file(self._strain_formulation)
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
ref_mat_output = None
if is_ref_material_output:
# Set file where reference material data is stored
refm_file_path = output_dir + problem_name + '.refm'
# Instantiate reference material output
ref_mat_output = RefMatOutput(
refm_file_path, self._strain_formulation, self._problem_type,
self._self_consistent_scheme)
# Write reference material output file header
ref_mat_output.init_file()
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
voxels_output = None
if is_voxels_output:
# Set voxels material-related output file path
voxout_file_path = output_dir + problem_name + '.voxout'
# Instantiate voxels material-related output
voxels_output = VoxelsOutput(
voxout_file_path, self._strain_formulation, self._problem_type)
# Write voxels material-related output file header
voxels_output.init_voxels_output_file(crve)
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
vtk_output = None
if is_vtk_output:
# Set VTK output directories paths
pvd_dir = output_dir + 'post_processing/'
vtk_dir = pvd_dir + 'VTK/'
# Instantiante VTK output
vtk_output = \
VTKOutput(type='ImageData', version='1.0',
byte_order=vtk_data['vtk_byte_order'],
format=vtk_data['vtk_format'],
precision=vtk_data['vtk_precision'],
header_type='UInt64', base_name=problem_name,
vtk_dir=vtk_dir, pvd_dir=pvd_dir)
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
adapt_output = None
if is_clust_adapt_output:
# Set file where clustering adaptivity data is stored
adapt_file_path = output_dir + problem_name + '.adapt'
# Instantiate clustering adaptivity output
adapt_output = ClusteringAdaptivityOutput(
adapt_file_path, crve.get_adapt_material_phases())
# Write clustering adaptivity output file header
adapt_output.init_adapt_file(crve)
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Return
return hres_output, efftan_output, ref_mat_output, voxels_output, \
adapt_output, vtk_output
#
# Reference (fictitious) homogeneous material
# =============================================================================
[docs]class ElasticReferenceMaterial:
"""Elastic reference material.
Attributes
----------
_n_dim : int
Problem number of spatial dimensions.
_comp_order_sym : list[str]
Strain/Stress components symmetric order.
_comp_order_nsym : list[str]
Strain/Stress components nonsymmetric order.
_material_properties : dict
Elastic material properties (key, str) values
(item, {int, float, bool}).
_material_properties_old : dict
Last loading increment converged elastic material properties (key, str)
values (item, {int, float, bool}).
_material_properties_init : dict
Elastic material properties (key, str) values
(item, {int, float, bool}) initial guess.
_material_properties_scs_init : dict
Elastic material properties (key, str) values
(item, {int, float, bool}) converged in the first loading increment.
_elastic_tangent_mf : numpy.ndarray (2d)
Elastic tangent modulus in matricial form.
_elastic_compliance_matrix : numpy.ndarray (2d)
Elastic compliance in matrix form.
_scs_iter : int
Self-consistent scheme iteration counter.
_norm_dE : float
Normalized iterative change of Young modulus associated with the last
self-consistent iteration convergence evaluation.
_norm_dv : float
Normalized iterative change of Poisson ratio associated with the last
self-consistent iteration convergence evaluation.
Methods
-------
init_material_properties(self, material_phases, \
material_phases_properties, material_phases_vf, \
properties=None)
Set initial guess of elastic reference material properties.
update_material_properties(self, E, v)
Update elastic reference material properties.
update_converged_material_properties(self)
Update converged elastic reference material properties.
reset_material_properties(self)
Reset material properties to last loading increment values.
set_material_properties_scs_init(self)
Set material properties converged in the first loading increment.
get_material_properties(self)
Get elastic reference material properties.
get_elastic_tangent_mf(self)
Get elastic tangent modulus in matricial form.
get_elastic_compliance_matrix(self)
Get elastic compliance in matrix form.
init_scs_iter(self)
Initialize self-consistent scheme iteration counter.
update_scs_iter(self)
Update self-consistent scheme iteration counter.
get_scs_iter(self)
Get self-consistent scheme iteration counter.
get_norm_dE(self)
Get normalized iterative change of Young modulus.
get_norm_dv(self)
Get normalized iterative change of Poisson ratio.
self_consistent_update(self, strain_mf, strain_old_mf, stress_mf, \
stress_old_mf, eff_tangent_mf)
Compute reference elastic properties through self-consistent scheme.
_update_elastic_tangent(self)
Update reference material elastic tangent modulus and compliance.
_check_scs_solution(self, E, v)
Check admissibility of self-consistent scheme iterative solution.
check_scs_convergence(self, E, v)
Check self-consistent scheme iterative solution convergence.
get_available_scs(strain_formulation)
Get available self-consistent schemes.
lame_from_technical(E, v)
Get Lamé parameters from Young modulus and Poisson ratio.
technical_from_lame(lam, miu)
Get Young modulus and Poisson ratio from Lamé parameters.
"""
[docs] def __init__(self, strain_formulation, problem_type,
self_consistent_scheme, conv_tol=1e-4):
"""Constructor.
Parameters
----------
strain_formulation: {'infinitesimal', 'finite'}
Problem strain formulation.
problem_type : int
Problem type: 2D plane strain (1), 2D plane stress (2),
2D axisymmetric (3) and 3D (4).
self_consistent_scheme : {'regression',}, default='regression'
Self-consistent scheme to update the elastic reference material
properties.
conv_tol : float, default=1e-4
Self-consistent scheme convergence tolerance.
"""
self._strain_formulation = strain_formulation
self._problem_type = problem_type
self._self_consistent_scheme = self_consistent_scheme
self._conv_tol = conv_tol
self._material_properties = None
self._material_properties_old = None
self._material_properties_init = None
self._material_properties_scs_init = None
self._elastic_tangent_mf = None
self._scs_iter = 0
self._elastic_compliance_matrix = None
# Get problem type parameters
n_dim, comp_order_sym, comp_order_nsym = \
mop.get_problem_type_parameters(problem_type)
self._n_dim = n_dim
self._comp_order_sym = comp_order_sym
self._comp_order_nsym = comp_order_nsym
# -------------------------------------------------------------------------
[docs] def init_material_properties(self, material_phases,
material_phases_properties,
material_phases_vf, properties=None):
"""Set initial guess of elastic reference material properties.
Parameters
----------
material_phases : list[str]
RVE material phases labels (str).
material_phases_properties : dict
Constitutive model material properties (item, dict) associated with
each material phase (key, str).
material_phases_vf : dict
Volume fraction (item, float) associated with each material phase
(key, str).
properties : dict, default=None
Initial guess (item, float) of elastic reference material
properties (key, str). Expecting Young's modulus ('E') and
Poisson's coefficient ('v') for an isotropic elastic reference
material.
"""
if properties is None:
# If a initial guess of the elastic reference material properties
# is not provided, set them from the volume average of the actual
# material phases elastic properties
E = sum([material_phases_vf[phase]
* material_phases_properties[phase]['E']
for phase in material_phases])
v = sum([material_phases_vf[phase]
* material_phases_properties[phase]['v']
for phase in material_phases])
else:
# Set initial guess of elastic reference material properties
E = properties['E']
v = properties['v']
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Initialize elastic reference material properties
self._material_properties_init = {'E': E, 'v': v}
self._material_properties = \
copy.deepcopy(self._material_properties_init)
self._material_properties_old = \
copy.deepcopy(self._material_properties_init)
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Update elastic tangent modulus and elastic compliance matrix
self._update_elastic_tangent()
# -------------------------------------------------------------------------
[docs] def update_material_properties(self, E, v):
"""Update elastic reference material properties.
Parameters
----------
E : float
Young modulus of elastic reference material.
v : float
Poisson ratio of elastic reference material.
"""
# Update elastic reference material properties
self._material_properties['E'] = E
self._material_properties['v'] = v
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Update elastic tangent modulus and elastic compliance matrix
self._update_elastic_tangent()
# -------------------------------------------------------------------------
[docs] def update_converged_material_properties(self):
"""Update converged elastic reference material properties."""
self._material_properties_old = \
copy.deepcopy(self._material_properties)
# -------------------------------------------------------------------------
[docs] def reset_material_properties(self):
"""Reset material properties to last loading increment values."""
# Reset elastic reference material properties to last loading increment
# values
self._material_properties = \
copy.deepcopy(self._material_properties_old)
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Update elastic tangent modulus and elastic compliance matrix
self._update_elastic_tangent()
# -------------------------------------------------------------------------
[docs] def set_material_properties_scs_init(self):
"""Set material properties converged in the first loading increment."""
self._material_properties_scs_init = \
copy.deepcopy(self._material_properties)
# -------------------------------------------------------------------------
[docs] def get_material_properties(self):
"""Get elastic reference material properties.
Returns
-------
material_properties : dict
Elastic material properties (key, str) values
(item, {int, float, bool}).
"""
return copy.deepcopy(self._material_properties)
# -------------------------------------------------------------------------
[docs] def get_elastic_tangent_mf(self):
"""Get elastic tangent modulus in matricial form.
Returns
-------
elastic_tangent_mf : numpy.ndarray (2d)
Elastic tangent modulus in matricial form.
"""
return copy.deepcopy(self._elastic_tangent_mf)
# -------------------------------------------------------------------------
[docs] def get_elastic_compliance_matrix(self):
"""Get elastic compliance in matrix form.
Returns
-------
elastic_compliance_matrix : numpy.ndarray (2d)
Elastic compliance in matrix form.
"""
return copy.deepcopy(self._elastic_compliance_matrix)
# -------------------------------------------------------------------------
[docs] def init_scs_iter(self):
"""Initialize self-consistent scheme iteration counter."""
self._scs_iter = 0
# -------------------------------------------------------------------------
[docs] def update_scs_iter(self):
"""Update self-consistent scheme iteration counter."""
self._scs_iter += 1
# -------------------------------------------------------------------------
[docs] def get_scs_iter(self):
"""Get self-consistent scheme iteration counter.
Returns
-------
scs_iter : int
Self-consistent scheme iteration counter.
"""
return self._scs_iter
# -------------------------------------------------------------------------
[docs] def get_norm_dE(self):
"""Get normalized iterative change of Young modulus.
Returns
-------
norm_dE : float
Normalized iterative change of Young modulus associated with the
last self-consistent iteration convergence evaluation.
"""
return self._norm_dE
# -------------------------------------------------------------------------
[docs] def get_norm_dv(self):
"""Get normalized iterative change of Poisson ratio.
Returns
-------
norm_dv : float
Normalized iterative change of Poisson ratio associated with the
last self-consistent iteration convergence evaluation.
"""
return self._norm_dv
# -------------------------------------------------------------------------
[docs] def self_consistent_update(self, strain_mf, strain_old_mf, stress_mf,
stress_old_mf, eff_tangent_mf):
"""Compute reference elastic properties through self-consistent scheme.
Parameters
----------
strain_mf : numpy.ndarray (1d)
Homogenized strain (matricial form): infinitesimal strain tensor
(infinitesimal strains) or deformation gradient (finite strains)
strain_old_mf : numpy.ndarray (1d)
Last converged homogenized strain (matricial form): infinitesimal
strain tensor (infinitesimal strains) or deformation gradient
(finite strains)
stress_mf : numpy.ndarray (1d)
Homogenized stress (matricial form): Cauchy stress tensor
(infinitesimal strains) or first Piola-Kirchhoff stress tensor
(finite strains).
stress_old_mf : numpy.ndarray (1d)
Last converged homogenized stress (matricial form): Cauchy stress
tensor (infinitesimal strains) or first Piola-Kirchhoff stress
tensor (finite strains).
eff_tangent_mf : numpy.ndarray (2d)
CRVE effective material tangent modulus (matricial form).
Returns
-------
is_admissible : bool
True if self-consistent scheme iterative solution is admissible,
False otherwise.
E : float
Young modulus of elastic reference material.
v : float
Poisson ratio of elastic reference material.
"""
# Compute incremental homogenized strain and stress tensors according
# to problem strain formulation and self-consistent scheme
if self._strain_formulation == 'infinitesimal':
# Compute incremental homogenized infinitesimal strain tensor
inc_strain_mf = strain_mf - strain_old_mf
# Compute incremental homogenized Cauchy stress tensor
inc_stress_mf = stress_mf - stress_old_mf
else:
raise RuntimeError('A suitable self-consistent scheme has not '
'been developed under finite strains yet.')
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Compute elastic reference material properties based on the
# regression-based self-consistent scheme
if self._self_consistent_scheme in ('regression',):
# Initialize elastic reference material properties regression-based
# self-consistent scheme optimizer
if self._strain_formulation == 'infinitesimal':
# Initialize elastic reference material properties optimizer
ref_optimizer = InfinitesimalRegressionSCS(
self._strain_formulation, self._problem_type,
copy.deepcopy(self._material_properties_old),
inc_strain_mf, inc_stress_mf)
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Compute elastic reference material properties
E, v = ref_optimizer.compute_reference_properties()
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Check admissibility of self-consistent scheme solution
is_admissible = self._check_scs_solution(E, v)
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Return
return is_admissible, E, v
# -------------------------------------------------------------------------
[docs] def _update_elastic_tangent(self):
"""Update reference material elastic tangent modulus and compliance.
*Infinitesimal strains:*
.. math::
\\boldsymbol{\\mathsf{D}}^{e,\\,0} = \\lambda^{0} \\boldsymbol{I}
\\otimes \\boldsymbol{I} + 2 \\mu^{0} \\boldsymbol{\\mathsf{I}}_{s}
\\, ,
where :math:`\\boldsymbol{\\mathsf{D}}^{e,\\,0}` is the reference
material elastic tangent modulus, :math:`\\lambda^{0}` and
:math:`\\mu^{0}` are the elastic Lamé parameters,
:math:`\\boldsymbol{I}` is the second-order identity tensor, and
:math:`\\boldsymbol{\\mathsf{I}}_{s}` is the fourth-order symmetric
identity tensor.
.. math::
\\boldsymbol{\\mathsf{S}}^{e,\\,0} =
- \\dfrac{\\lambda^{0}}{2 \\mu^{0} (3\\lambda^{0} + 2\\mu^{0})}
\\boldsymbol{I} \\otimes \\boldsymbol{I} +
\\dfrac{1}{2 \\mu^{0}} \\boldsymbol{\\mathsf{I}}_{s} \\, ,
where :math:`\\boldsymbol{\\mathsf{S}}^{e,\\,0}` is the reference
material elastic compliance, :math:`\\lambda^{0}` and :math:`\\mu^{0}`
are the elastic Lamé parameters, :math:`\\boldsymbol{I}` is the
second-order identity tensor, and :math:`\\boldsymbol{\\mathsf{I}}_{s}`
is the fourth-order symmetric identity tensor.
----
*Finite strains:*
.. math::
\\boldsymbol{\\mathsf{A}}^{e,\\,0} = \\lambda^{0} \\boldsymbol{I}
\\otimes \\boldsymbol{I} + 2 \\mu^{0} \\boldsymbol{\\mathsf{I}}
\\, ,
where :math:`\\boldsymbol{\\mathsf{A}}^{e,\\,0}` is the reference
material hyperelastic tangent modulus, :math:`\\lambda^{0}` and
:math:`\\mu^{0}` are the elastic Lamé parameters,
:math:`\\boldsymbol{I}` is the second-order identity tensor, and
:math:`\\boldsymbol{\\mathsf{I}}` is the fourth-order identity tensor.
.. math::
\\boldsymbol{\\mathsf{S}}^{e,\\,0} =
- \\dfrac{\\lambda^{0}}{2 \\mu^{0} (3\\lambda^{0} + 2\\mu^{0})}
\\boldsymbol{I} \\otimes \\boldsymbol{I} +
\\dfrac{1}{2 \\mu^{0}} \\boldsymbol{\\mathsf{I}} \\, ,
where :math:`\\boldsymbol{\\mathsf{S}}^{e,\\,0}` is the reference
material hyperelastic compliance, :math:`\\lambda^{0}` and
:math:`\\mu^{0}` are the elastic Lamé parameters,
:math:`\\boldsymbol{I}` is the second-order identity tensor, and
:math:`\\boldsymbol{\\mathsf{I}}` is the fourth-order symmetric
identity tensor.
"""
# Get Young's Modulus and Poisson's ratio
E = self._material_properties['E']
v = self._material_properties['v']
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Compute Lamé parameters
lam, miu = type(self).lame_from_technical(E, v)
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Set required fourth-order tensors
_, foid, _, fosym, fodiagtrace, _, _ = \
top.get_id_operators(self._n_dim)
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Compute elastic tangent modulus and elastic compliance matrix
if self._strain_formulation == 'infinitesimal':
# Set symmetric strain/stress component order
comp_order = self._comp_order_sym
# Compute elastic tangent modulus and elastic compliance matrix
# according to problem type
if self._problem_type in [1, 4]:
# Compute elastic tangent modulus
elastic_tangent = lam*fodiagtrace + 2.0*miu*fosym
# Compute elastic compliance
elastic_compliance = -(lam/(2*miu*(3*lam + 2*miu))) \
* fodiagtrace + (1.0/(2.0*miu))*fosym
else:
# Set nonsymmetric strain/stress component order
comp_order = self._comp_order_nsym
# Compute elastic tangent modulus and elastic compliance matrix
# according to problem type
if self._problem_type in [1, 4]:
# Compute elastic tangent modulus (2D problem (plane strain),
# 3D problem)
elastic_tangent = lam*fodiagtrace + 2.0*miu*foid
# Compute elastic compliance
elastic_compliance = -(lam/(2*miu*(3*lam + 2*miu))) \
* fodiagtrace + (1.0/(2.0*miu))*foid
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Build elastic tangent modulus matricial form
elastic_tangent_mf = mop.get_tensor_mf(elastic_tangent, self._n_dim,
comp_order)
# Build elastic compliance matricial form
elastic_compliance_mf = mop.get_tensor_mf(elastic_compliance,
self._n_dim, comp_order)
# Build elastic compliance matrix (without matricial form associated
# coefficients)
elastic_compliance_matrix = np.zeros(elastic_compliance_mf.shape)
for j in range(len(comp_order)):
for i in range(len(comp_order)):
elastic_compliance_matrix[i, j] = \
(1.0/mop.kelvin_factor(i, comp_order)) \
* (1.0/mop.kelvin_factor(j, comp_order)) \
* elastic_compliance_mf[i, j]
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Update reference material elastic tangent modulus and compliance
# matrix
self._elastic_tangent_mf = elastic_tangent_mf
self._elastic_compliance_matrix = elastic_compliance_matrix
# -------------------------------------------------------------------------
[docs] def _check_scs_solution(self, E, v):
"""Check admissibility of self-consistent scheme iterative solution.
Parameters
----------
E : float
Young modulus of elastic reference material.
v : float
Poisson ratio of elastic reference material.
Returns
-------
is_admissible : bool
True if self-consistent scheme iterative solution is admissible,
False otherwise.
"""
# Set admissibility default value
is_admissible = False
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Evaluate admissibility conditions:
# Reference material Young modulus
if self._material_properties_init is None:
condition_1 = E > 0.0
else:
condition_1 = (E/self._material_properties_init['E']) >= 0.01
# Reference material Poisson ratio
condition_2 = v > 0.0 and (v/0.5) < 1.0
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Set admissibility of self-consistent scheme iterative solution
is_admissible = condition_1 and condition_2
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
return is_admissible
# -------------------------------------------------------------------------
[docs] def check_scs_convergence(self, E, v):
"""Check self-consistent scheme iterative solution convergence.
Parameters
----------
E : float
Young modulus of elastic reference material.
v : float
Poisson ratio of elastic reference material.
Returns
-------
is_converged : bool
True if self-consistent scheme iterative solution converged, False
otherwise.
"""
# Compute iterative variation of the reference material Young modulus
# and Poisson ratio
dE = E - self._material_properties['E']
dv = v - self._material_properties['v']
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Compute normalized iterative change of the reference material Young
# modulus and Poisson ratio
norm_dE = abs(dE/E)
norm_dv = abs(dv/v)
# Store normalized iterative change of reference material properties
self._norm_dE = norm_dE
self._norm_dv = norm_dv
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Check self-consistent scheme convergence
is_converged = (norm_dE < self._conv_tol) \
and (norm_dv < self._conv_tol)
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Return
return is_converged
# -------------------------------------------------------------------------
[docs] @staticmethod
def get_available_scs(strain_formulation):
"""Get available self-consistent schemes.
Parameters
----------
strain_formulation: {'infinitesimal', 'finite'}
Problem strain formulation.
Returns
-------
available_scs : tuple[str]
Available self-consistent schemes.
"""
if strain_formulation == 'infinitesimal':
available_scs = ('none', 'regression')
elif strain_formulation == 'finite':
available_scs = ('none',)
else:
raise RuntimeError('Unknown problem strain formulation.')
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
return available_scs
# -------------------------------------------------------------------------
[docs] @staticmethod
def lame_from_technical(E, v):
"""Get Lamé parameters from Young modulus and Poisson ratio.
Parameters
----------
E : float
Young modulus.
v : float
Poisson ratio.
Returns
-------
lam : float
Lamé parameter.
miu : float
Lamé parameter.
"""
# Compute Lamé parameters
lam = (E*v)/((1.0 + v)*(1.0 - 2.0*v))
miu = E/(2.0*(1.0 + v))
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Return
return lam, miu
# -------------------------------------------------------------------------
[docs] @staticmethod
def technical_from_lame(lam, miu):
"""Get Young modulus and Poisson ratio from Lamé parameters.
Parameters
----------
lam : float
Lamé parameter.
miu : float
Lamé parameter.
Returns
-------
E : float
Young modulus.
v : float
Poisson ratio.
"""
# Compute Young modulus and Poisson ratio
E = (miu*(3.0*lam + 2.0*miu))/(lam + miu)
v = lam/(2.0*(lam + miu))
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Return
return E, v
#
# Interface: Reference material properties optimizer
# =============================================================================
[docs]class ReferenceMaterialOptimizer(ABC):
"""Elastic reference material properties optimizer interface.
Attributes
----------
_n_dim : int
Problem number of spatial dimensions.
_comp_order_sym : list[str]
Strain/Stress components symmetric order.
_comp_order_nsym : list[str]
Strain/Stress components nonsymmetric order.
"""
[docs] @abstractmethod
def __init__(self, strain_formulation, problem_type):
"""Elastic reference material properties optimizer constructor.
Parameters
----------
strain_formulation: {'infinitesimal', 'finite'}
Problem strain formulation.
problem_type : int
Problem type: 2D plane strain (1), 2D plane stress (2),
2D axisymmetric (3) and 3D (4).
"""
pass
# -------------------------------------------------------------------------
[docs] @abstractmethod
def compute_reference_properties(self):
"""Compute elastic reference material properties.
Returns
-------
young : float
Young modulus of elastic reference material.
poiss : float
Poisson ratio of elastic reference material.
"""
pass
#
# Reference material properties optimizers
# =============================================================================
[docs]class InfinitesimalRegressionSCS(ReferenceMaterialOptimizer):
"""Infinitesimal strains format regression-based self-consistent scheme.
*Mimization problem:*
.. math::
\\left\\{ \\lambda^{0}_{n+1}, \\, \\mu^{0}_{n+1} \\right\\} =
\\underset{ \\left\\{ \\lambda ', \\, \\, \\mu ' \\right\\} }{
\\mathrm{argmin}} \\, || \\Delta \\boldsymbol{\\sigma}_{n+1} -
\\boldsymbol{\\mathsf{D}}^{e, \\, 0}_{n+1}(\\lambda ',\\mu ') :
\\Delta \\boldsymbol{\\varepsilon}_{n+1} ||^{2}
where :math:`\\lambda^{0}` and :math:`\\mu^{0}` are the elastic Lamé
parameters, :math:`\\Delta \\boldsymbol{\\sigma}` is the homogenized
incremental Cauchy stress tensor,
:math:`\\boldsymbol{\\mathsf{D}}^{e,\\,0}` is the reference material
elastic tangent modulus, :math:`\\Delta \\boldsymbol{\\varepsilon}` is
the homogenized incremental infinitesimal strain tensor, and
:math:`n+1` denotes the current increment.
----
*Solution:*
.. math::
\\begin{bmatrix}
\\lambda^{0}_{m+1} \\\\[10pt] \\mu^{0}_{m+1}
\\end{bmatrix}
=
\\begin{bmatrix}
\\text{tr} \\, \\left[ \\boldsymbol{I} \\right] \\,
\\text{tr} \\, \\left[ \\Delta
\\boldsymbol{\\varepsilon}_{m+1} \\right] & 2 \\,
\\text{tr} \\, \\left[ \\Delta
\\boldsymbol{\\varepsilon}_{m+1} \\right] \\\\[10pt]
\\text{tr} \\, \\left[ \\Delta
\\boldsymbol{\\varepsilon}_{m+1} \\right]^{2} & 2
\\Delta \\boldsymbol{\\varepsilon}_{m+1} :
\\Delta \\boldsymbol{\\varepsilon}_{m+1}
\\end{bmatrix}^{-1}
\\begin{bmatrix}
\\, \\text{tr} \\, \\left[ \\Delta
\\boldsymbol{\\sigma}_{m+1} \\right] \\\\[10pt]
\\Delta \\boldsymbol{\\sigma}_{m+1}: \\Delta
\\boldsymbol{\\varepsilon}_{m+1}
\\end{bmatrix} \\, .
Attributes
----------
_n_dim : int
Problem number of spatial dimensions.
_comp_order_sym : list[str]
Strain/Stress components symmetric order.
_comp_order_nsym : list[str]
Strain/Stress components nonsymmetric order.
"""
[docs] def __init__(self, strain_formulation, problem_type,
material_properties_old, inc_strain_mf, inc_stress_mf,
is_symmetrized=False):
"""Constructor.
Parameters
----------
strain_formulation: {'infinitesimal', 'finite'}
Problem strain formulation.
problem_type : int
Problem type: 2D plane strain (1), 2D plane stress (2),
2D axisymmetric (3) and 3D (4).
material_properties_old : dict
Last loading increment converged elastic reference material
properties (key, str) values (item, {int, float, bool}).
inc_strain_mf : numpy.ndarray (1d)
Incremental homogenized strain (matricial form).
inc_stress_mf : numpy.ndarray (1d)
Incremental homogenized stress (matricial form).
is_symmetrized : bool, default=False
True if a symmetric alternative stress-strain conjugate pair is
adopted in the finite strains regression-based self-consistent
scheme, False otherwise.
"""
self._strain_formulation = strain_formulation
self._problem_type = problem_type
self._material_properties_old = copy.deepcopy(material_properties_old)
self._inc_strain_mf = copy.deepcopy(inc_strain_mf)
self._inc_stress_mf = copy.deepcopy(inc_stress_mf)
self._is_symmetrized = is_symmetrized
# Get problem type parameters
n_dim, comp_order_sym, comp_order_nsym = \
mop.get_problem_type_parameters(problem_type)
self._n_dim = n_dim
self._comp_order_sym = comp_order_sym
self._comp_order_nsym = comp_order_nsym
# -------------------------------------------------------------------------
[docs] def compute_reference_properties(self):
"""Compute elastic reference material properties.
Returns
-------
E : float
Young modulus of elastic reference material.
v : float
Poisson ratio of elastic reference material.
"""
# Set strain/stress components order
if self._strain_formulation == 'infinitesimal':
comp_order = self._comp_order_sym
elif self._strain_formulation == 'finite':
if self._is_symmetrized:
comp_order = self._comp_order_sym
else:
comp_order = self._comp_order_nsym
else:
raise RuntimeError('Unknown problem strain formulation.')
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
if np.all([abs(self._inc_strain_mf[i]) < 1e-10
for i in range(self._inc_strain_mf.shape[0])]):
# Get last loading increment converged elastic reference material
# properties
E = self._material_properties_old['E']
v = self._material_properties_old['v']
# Return
return E, v
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Set second-order identity tensor
soid, _, _, _, _, _, _ = top.get_id_operators(self._n_dim)
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Initialize self-consistent scheme system of linear equations
# coefficient matrix and right-hand side
scs_matrix = np.zeros((2, 2))
scs_rhs = np.zeros(2)
# Get incremental strain and stress tensors
inc_strain = mop.get_tensor_from_mf(self._inc_strain_mf, self._n_dim,
comp_order)
inc_stress = mop.get_tensor_from_mf(self._inc_stress_mf, self._n_dim,
comp_order)
# Compute self-consistent scheme system of linear equations right-hand
# side
scs_rhs[0] = np.trace(inc_stress)
scs_rhs[1] = top.ddot22_1(inc_stress, inc_strain)
# Compute self-consistent scheme system of linear equations coefficient
# matrix
scs_matrix[0, 0] = np.trace(inc_strain)*np.trace(soid)
scs_matrix[0, 1] = 2.0*np.trace(inc_strain)
scs_matrix[1, 0] = np.trace(inc_strain)**2
scs_matrix[1, 1] = 2.0*top.ddot22_1(inc_strain, inc_strain)
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Limitation 1: Under isochoric loading conditions the first equation
# of the self-consistent scheme system of linear equations vanishes
# (derivative with respect to lambda). In this case, adopt the previous
# converged lambda and compute miu from the second equation of the
# self-consistent scheme system of linear equations
if (abs(np.trace(inc_strain))/np.linalg.norm(inc_strain)) < 1e-10 \
or np.linalg.solve(scs_matrix, scs_rhs)[0] < 0:
# Get previous converged reference material elastic properties
E_old = self._material_properties_old['E']
v_old = self._material_properties_old['v']
# Compute previous converged lambda
lam = (E_old*v_old)/((1.0 + v_old)*(1.0 - 2.0*v_old))
# Compute miu
miu = scs_rhs[1]/scs_matrix[1, 1]
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Limitation 2: Under hydrostatic loading conditions both equations of
# the self-consistent scheme system of linear equations become linearly
# dependent. In this case, assume that the ratio between lambda and miu
# is the same as in the previous converged values and solve the first
# equation of self-consistent scheme system of linear equations
elif np.all([abs(inc_strain[0, 0] - inc_strain[i, i])
/ np.linalg.norm(inc_strain) < 1e-10
for i in range(self._n_dim)]) and \
np.allclose(inc_strain, np.diag(np.diag(inc_strain)),
atol=1e-10):
# Get previous converged reference material elastic properties
E_old = self._material_properties_old['E']
v_old = self._material_properties_old['v']
# Compute previous converged reference material Lamé parameters
lam_old = (E_old*v_old)/((1.0 + v_old)*(1.0 - 2.0*v_old))
miu_old = E_old/(2.0*(1.0 + v_old))
# Compute reference material Lamé parameters
lam = (scs_rhs[0]/scs_matrix[0, 0])*(lam_old/(lam_old + miu_old))
miu = (scs_rhs[0]/scs_matrix[0, 0])*(miu_old/(lam_old + miu_old))
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Solve self-consistent scheme system of linear equations
else:
scs_solution = np.linalg.solve(scs_matrix, scs_rhs)
# Get reference material Lamé parameters
lam = scs_solution[0]
miu = scs_solution[1]
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Compute reference material Young modulus and Poisson ratio
E, v = ElasticReferenceMaterial.technical_from_lame(lam, miu)
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Return
return E, v