diff --git a/CMakeLists.txt b/CMakeLists.txt index cb7813a1b..e36180b8d 100644 --- a/CMakeLists.txt +++ b/CMakeLists.txt @@ -1,228 +1,228 @@ #=============================================================================== # @file CMakeLists.txt # # @author Guillaume Anciaux # @author Nicolas Richart # # @date creation: Mon Jun 14 2010 # @date last modification: Fri Jan 22 2016 # # @brief main configuration file # # @section LICENSE # # Copyright (©) 2010-2012, 2014, 2015 EPFL (Ecole Polytechnique Fédérale de # Lausanne) Laboratory (LSMS - Laboratoire de Simulation en Mécanique des # Solides) # # Akantu is free software: you can redistribute it and/or modify it under the # terms of the GNU Lesser General Public License as published by the Free # Software Foundation, either version 3 of the License, or (at your option) any # later version. # # Akantu is distributed in the hope that it will be useful, but WITHOUT ANY # WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR # A PARTICULAR PURPOSE. See the GNU Lesser General Public License for more # details. # # You should have received a copy of the GNU Lesser General Public License # along with Akantu. If not, see . # # @section DESCRIPTION #------------------------------------------------------------------------------- # _ _ # | | | | # __ _| | ____ _ _ __ | |_ _ _ # / _` | |/ / _` | '_ \| __| | | | # | (_| | < (_| | | | | |_| |_| | # \__,_|_|\_\__,_|_| |_|\__|\__,_| # #=============================================================================== #=============================================================================== # CMake Project #=============================================================================== cmake_minimum_required(VERSION 3.5.1) # add this options before PROJECT keyword set(CMAKE_DISABLE_SOURCE_CHANGES ON) set(CMAKE_DISABLE_IN_SOURCE_BUILD ON) if(CMAKE_VERSION VERSION_GREATER 3.12) cmake_policy(SET CMP0074 NEW) endif() set(AKANTU_MAJOR_VERSION 4) set(AKANTU_MINOR_VERSION 0) set(AKANTU_PATCH_VERSION 0) -set(AKANTU_COPYRIGHT "2010-2020, EPFL (Ecole Polytechnique Fédérale de Lausanne) Laboratory (LSMS - Laboratoire de Simulation en Mécanique des Solides)") +set(AKANTU_COPYRIGHT "2010-2021, EPFL (Ecole Polytechnique Fédérale de Lausanne) Laboratory (LSMS - Laboratoire de Simulation en Mécanique des Solides)") set(AKANTU_MAINTAINER "Nicolas Richart") -set(AKANTU_HOMEPAGE_URL "htpps://akantu.ch") +set(AKANTU_HOMEPAGE_URL "https://akantu.ch") if(CMAKE_VERSION VERSION_GREATER 3.12) project(Akantu - HOMEPAGE_URL "htpps://akantu.ch") + HOMEPAGE_URL "https://akantu.ch") else() project(Akantu) endif() enable_language(CXX) #=============================================================================== # Misc. config for cmake #=============================================================================== set(AKANTU_CMAKE_DIR "${PROJECT_SOURCE_DIR}/cmake") set(CMAKE_MODULE_PATH ${CMAKE_MODULE_PATH} "${PROJECT_SOURCE_DIR}/cmake") set(CMAKE_MODULE_PATH ${CMAKE_MODULE_PATH} "${PROJECT_SOURCE_DIR}/cmake/Modules") set(BUILD_SHARED_LIBS ON CACHE BOOL "Build shared libraries.") set(CMAKE_EXPORT_COMPILE_COMMANDS ON CACHE BOOL "Enable/Disable output of compile commands during generation" FORCE) mark_as_advanced(BUILD_SHARED_LIBS) if(NOT AKANTU_TARGETS_EXPORT) set(AKANTU_TARGETS_EXPORT AkantuTargets) endif() include(CMakeVersionGenerator) include(CMakePackagesSystem) include(CMakeFlagsHandling) include(AkantuPackagesSystem) include(AkantuMacros) include(AkantuCleaning) #cmake_activate_debug_message() include(GNUInstallDirs) set(CMAKE_INSTALL_RPATH "${CMAKE_INSTALL_PREFIX}/lib") # add the automatically determined parts of the RPATH # which point to directories outside the build tree to the install RPATH set(CMAKE_INSTALL_RPATH_USE_LINK_PATH TRUE) #=============================================================================== # Version Number #=============================================================================== # AKANTU version number. An even minor number corresponds to releases. define_project_version() #=============================================================================== # Options #=============================================================================== # Debug set(CMAKE_CXX_FLAGS_RELEASE "-O3 -DNDEBUG -DAKANTU_NDEBUG" CACHE STRING "Flags used by the compiler during release builds" FORCE) #add_flags(cxx "-Wall -Wextra -pedantic -Werror") if (CMAKE_CXX_COMPILER_ID STREQUAL "GNU" OR CMAKE_CXX_COMPILER_ID STREQUAL "Clang") add_flags(cxx "-Wall -Wextra -pedantic") # -Weffc++ set(CMAKE_CXX_FLAGS_DEBUG "${CMAKE_CXX_FLAGS_DEBUG_INIT} -ggdb3" CACHE STRING "Flags used by the compiler during debug builds" FORCE) set(CMAKE_CXX_FLAGS_RELWITHDEBINFO "${CMAKE_CXX_FLAGS_RELWITHDEBINFO_INIT} -ggdb3" CACHE STRING "Flags used by the compiler during debug builds" FORCE) option (FORCE_COLORED_OUTPUT "Always produce ANSI-colored output (GNU/Clang only)." FALSE) mark_as_advanced(FORCE_COLORED_OUTPUT) if(FORCE_COLORED_OUTPUT) if (CMAKE_CXX_COMPILER_ID STREQUAL "Clang") add_flags(cxx "-fcolor-diagnostics") else() add_flags(cxx "-fdiagnostics-color=always") endif() endif() else() add_flags(cxx "-Wall") endif() option(AKANTU_EXAMPLES "Activate examples" OFF) option(AKANTU_TESTS "Activate tests" OFF) option(AKANTU_RUN_IN_DOCKER "Set the approriate flage tu run in docker" OFF) set(AKANTU_PREFERRED_PYTHON_VERSION 3 CACHE STRING "Preferred version for python related things") mark_as_advanced(AKANTU_PREFERRED_PYTHON_VERSION AKANTU_RUN_IN_DOCKER) include(AkantuExtraCompilationProfiles) #=============================================================================== # Dependencies #=============================================================================== declare_akantu_types() package_list_packages(${PROJECT_SOURCE_DIR}/packages EXTRA_PACKAGES_FOLDER ${PROJECT_SOURCE_DIR}/extra_packages NO_AUTO_COMPILE_FLAGS) ## meta option \todo better way to do it when multiple package give enable the ## same feature if(AKANTU_SCOTCH) set(AKANTU_PARTITIONER ON) else() set(AKANTU_PARTITIONER OFF) endif() if(AKANTU_MUMPS) set(AKANTU_SOLVER ON) else() set(AKANTU_SOLVER OFF) endif() #=============================================================================== # Akantu library #=============================================================================== add_subdirectory(src) #=============================================================================== # Documentation #=============================================================================== if(AKANTU_DOCUMENTATION OR AKANTU_DOCUMENTATION_MANUAL) add_subdirectory(doc) else() set(AKANTU_DOC_EXCLUDE_FILES "${PROJECT_SOURCE_DIR}/doc/manual" CACHE INTERNAL "") endif() #=============================================================================== # Python interface #=============================================================================== package_is_activated(python_interface _python_act) if(_python_act) if(IS_ABSOLUTE "${CMAKE_INSTALL_PREFIX}") set(AKANTU_PYTHON_INSTALL_PREFIX ${CMAKE_INSTALL_PREFIX}) else() set(AKANTU_PYTHON_INSTALL_PREFIX "${PROJECT_BINARY_DIR}/${CMAKE_INSTALL_PREFIX}") endif() add_subdirectory(python) endif() #=============================================================================== # Examples and tests #=============================================================================== include(AkantuTestsMacros) include(AkantuExampleMacros) if(AKANTU_TESTS) option(AKANTU_BUILD_ALL_TESTS "Build all tests" ON) find_package(GMSH REQUIRED) endif() # tests add_test_tree(test) if(AKANTU_EXAMPLES) if(AKANTU_TESTS) option(AKANTU_TEST_EXAMPLES "Run the examples" ON) endif() find_package(GMSH REQUIRED) add_subdirectory(examples) endif() #=============================================================================== # Install and Packaging #=============================================================================== include(AkantuInstall) option(AKANTU_DISABLE_CPACK "This option commands the generation of extra info for the \"make package\" target" ON) mark_as_advanced(AKANTU_DISABLE_CPACK) if(NOT AKANTU_DISABLE_CPACK) include(AkantuCPack) endif() diff --git a/doc/dev-doc/conf.py.in b/doc/dev-doc/conf.py.in index 5d3563252..9649b901a 100644 --- a/doc/dev-doc/conf.py.in +++ b/doc/dev-doc/conf.py.in @@ -1,242 +1,242 @@ # -*- coding: utf-8 -*- # # Configuration file for the Sphinx documentation builder. # # This file does only contain a selection of the most common options. For a # full list see the documentation: # http://www.sphinx-doc.org/en/master/config # -- Path setup -------------------------------------------------------------- # If extensions (or modules to document with autodoc) are in another directory, # add these directories to sys.path here. If the directory is relative to the # documentation root, use os.path.abspath to make it absolute, like shown here. # import os import shutil # import subprocess # import sys # sys.path.insert(0, os.path.abspath('.')) # -- Project information ----------------------------------------------------- project = '@PROJECT_NAME@' -copyright = '@AKANTU_COPYRING@' +copyright = '@AKANTU_COPYRIGHT@' author = '@AKANTU_MAINTAINER@' -version = '@AKANTU_VERSION_MAJOR@.@AKANTU_VERSION_MINOR@' +version = '@AKANTU_MAJOR_VERSION@.@AKANTU_MINOR_VERSION@' # The full version, including alpha/beta/rc tags release = '@AKANTU_VERSION@' # -- General configuration --------------------------------------------------- # If your documentation needs a minimal Sphinx version, state it here. # # needs_sphinx = '1.0' # Number figures numfig = True # Add any Sphinx extension module names here, as strings. They can be # extensions coming with Sphinx (named 'sphinx.ext.*') or your custom # ones. extensions = [ 'sphinx.ext.autodoc', 'sphinx.ext.intersphinx', 'sphinx.ext.coverage', 'sphinx.ext.mathjax', 'sphinx.ext.ifconfig', 'sphinx.ext.viewcode', 'sphinxcontrib.bibtex', 'breathe', ] read_the_docs_build = os.environ.get('READTHEDOCS', None) == 'True' # Add any paths that contain templates here, relative to this directory. templates_path = ['_templates'] # The suffix(es) of source filenames. # You can specify multiple suffix as a list of string: # # source_suffix = ['.rst', '.md'] source_suffix = '.rst' # The master toctree document. master_doc = 'index' # The language for content autogenerated by Sphinx. Refer to documentation # for a list of supported languages. # # This is also used if you do content translation via gettext catalogs. # Usually you set "language" from the command line for these cases. language = None # List of patterns, relative to source directory, that match files and # directories to ignore when looking for source files. # This pattern also affects html_static_path and html_extra_path. exclude_patterns = ["CMakeLists.txt"] # The name of the Pygments (syntax highlighting) style to use. pygments_style = 'sphinx' primary_domain = 'cpp' highlight_language = 'cpp' bibtex_bibfiles = ['manual/manual-bibliography.bib'] # -- Options for HTML output ------------------------------------------------- # The theme to use for HTML and HTML Help pages. See the documentation for # a list of builtin themes. # if read_the_docs_build: html_theme = 'default' else: html_theme = 'sphinx_rtd_theme' # Theme options are theme-specific and customize the look and feel of a theme # further. For a list of options available for each theme, see the # documentation. # # html_theme_options = {} # Add any paths that contain custom static files (such as style sheets) here, # relative to this directory. They are copied after the builtin static files, # so a file named "default.css" will overwrite the builtin "default.css". html_static_path = ['_static'] html_logo = '_static/logo_only_akantu.svg' # Custom sidebar templates, must be a dictionary that maps document names # to template names. # # The default sidebars (for documents that don't match any pattern) are # defined by theme itself. Builtin themes are using these templates by # default: ``['localtoc.html', 'relations.html', 'sourcelink.html', # 'searchbox.html']``. # # html_sidebars = {} html_sidebars = { '**': [ 'relations.html', # needs 'show_related': True theme option to display 'searchbox.html', ]} math_eqref_format = "Eq. {number}" # MathJax configuration mathjax_config = { 'extensions': [ "tex2jax.js", "siunitx.js" ], 'TeX': { 'Macros': { 'st': [r'\mathrm{#1}', 1], 'mat': [r'\mathbf{#1}', 1], 'half': [r'\frac{1}{2}', 0], }, 'extensions': ["AMSmath.js", "AMSsymbols.js", "sinuitx.js"], }, } # -- Options for HTMLHelp output --------------------------------------------- # Output file base name for HTML help builder. htmlhelp_basename = 'Akantudoc' # -- Options for LaTeX output ------------------------------------------------ latex_elements = { # The paper size ('letterpaper' or 'a4paper'). # # 'papersize': 'letterpaper', # The font size ('10pt', '11pt' or '12pt'). # # 'pointsize': '10pt', # Additional stuff for the LaTeX preamble. # 'preamble': r'''\usepackage{amsmath}''', # Latex figure (float) alignment # # 'figure_align': 'htbp', } # Grouping the document tree into LaTeX files. List of tuples # (source start file, target name, title, # author, documentclass [howto, manual, or own class]). latex_documents = [ (master_doc, 'Akantu.tex', 'Akantu Documentation', 'Nicolas Richart', 'manual'), ] # -- Options for manual page output ------------------------------------------ # One entry per manual page. List of tuples # (source start file, name, description, authors, manual section). man_pages = [ (master_doc, 'akantu', 'Akantu Documentation', [author], 1) ] # -- Options for Texinfo output ---------------------------------------------- # Grouping the document tree into Texinfo files. List of tuples # (source start file, target name, title, author, # dir menu entry, description, category) texinfo_documents = [ (master_doc, 'Akantu', 'Akantu Documentation', author, 'Akantu', 'One line description of project.', 'Miscellaneous'), ] # -- Options for Epub output ------------------------------------------------- # Bibliographic Dublin Core info. epub_title = project epub_author = author epub_publisher = author epub_copyright = copyright # The unique identifier of the text. This can be a ISBN number # or the project homepage. # epub_identifier = '' # A unique identification for the text. # # epub_uid = '' # A list of files that should not be packed into the epub file. epub_exclude_files = ['search.html'] # -- Extension configuration ------------------------------------------------- if read_the_docs_build: akantu_path = "." else: akantu_path = "@CMAKE_CURRENT_BINARY_DIR@" os.makedirs(os.path.join(akantu_path, '_static'), exist_ok=True) shutil.copyfile( os.path.join('@CMAKE_CURRENT_SOURCE_DIR@', html_logo), os.path.join(akantu_path, html_logo)) # print("akantu_path = '{}'".format(akantu_path)) breathe_projects = {"Akantu": os.path.join(akantu_path, "xml")} breathe_default_project = "Akantu" breathe_default_members = ('members', 'undoc-members') breathe_implementation_filename_extensions = ['.c', '.cc', '.cpp'] breathe_show_enumvalue_initializer = True # -- Options for intersphinx extension --------------------------------------- intersphinx_mapping = { 'numpy': ('https://docs.scipy.org/doc/numpy/', None), 'scipy': ('https://docs.scipy.org/doc/scipy/reference', None), } diff --git a/doc/dev-doc/manual/appendix/material-parameters.rst b/doc/dev-doc/manual/appendix/material-parameters.rst index 4d0d11dfb..4ab43b393 100644 --- a/doc/dev-doc/manual/appendix/material-parameters.rst +++ b/doc/dev-doc/manual/appendix/material-parameters.rst @@ -1,4 +1,129 @@ .. _app-material-parameters: Material Parameters =================== + +Linear elastic isotropic +------------------------ + +Keyword: :ref:`elastic ` + +Parameters: + +- ``rho``: (*Real*) Density +- ``E``: (*Real*) Young's modulus +- ``nu``: (*Real*) Poisson's ratio +- ``Plane_stress``: (*bool*) Plane stress simplification (only 2D problems) + + +Linear elastic anisotropic +-------------------------- + +Keyword: :ref:`elastic_anisotropic ` + +Parameters: + +- ``rho``: (*Real*) Density +- ``n1``: (*Vector*) Direction of main material axis +- ``n2``: (*Vector*) Direction of second material axis +- ``n3``: (*Vector*) Direction of third material axis +- ``C..``: (*Real*) Coefficient ij of material tensor C (all the 36 values in + Voigt notation can be entered) +- ``alpha``: (*Real*) Viscous propertion (default is 0) + + +Linear elastic anisotropic +------------------------ + +Keyword: :ref:`elastic_orthotropic ` + +Parameters: + +- ``rho``: (*Real*) Density +- ``n1``: (*Vector*) Direction of main material axis +- ``n2``: (*Vector*) Direction of second material axis (if applicable) +- ``n3``: (*Vector*) Direction of third material axis (if applicable) +- ``E1``: (*Real*) Young's modulus (n1) +- ``E2``: (*Real*) Young's modulus (n2) +- ``E3``: (*Real*) Young's modulus (n3) +- ``nu1``: (*Real*) Poisson's ratio (n1) +- ``nu2``: (*Real*) Poisson's ratio (n2) +- ``nu3``: (*Real*) Poisson's ratio (n3) +- ``G12``: (*Real*) Shear modulus (12) +- ``G13``: (*Real*) Shear modulus (13) +- ``G23``: (*Real*) Shear modulus (23) + + +Neohookean (finite strains) +--------------------------- + +Keyword: :ref:`neohookean ` + +Parameters: + +- ``rho``: (*Real*) Density +- ``E``: (*Real*) Young's modulus +- ``nu``: (*Real*) Poisson's ratio +- ``Plane_stress``: (*bool*) Plane stress simplification (only 2D problems) + + +Standard linear solid +--------------------- + +Keyword: :ref:`sls_deviatoric ` + +Parameters: + +- ``rho``: (*Real*) Density +- ``E``: (*Real*) Young's modulus +- ``nu``: (*Real*) Poisson's ratio +- ``Plane_stress``: (*bool*) Plane stress simplification (only 2D problems) +- ``Eta``: (*Real*) Viscosity +- ``Ev``: (*Real*) Stiffness of viscous element + + +Elasto-plastic linear isotropic hardening +----------------------------------------- + +Keyword: :ref:`plastic_linear_isotropic_hardening ` + +Parameters: + +- ``rho``: (*Real*) Density +- ``E``: (*Real*) Young's modulus +- ``nu``: (*Real*) Poisson's ratio +- ``h``: (*Real*) Hardening modulus +- ``sigma_y``: (*Real*) Yield stress + + +Marigo +------ + +Keyword: :ref:`marigo ` + +Parameters: + +- ``rho``: (*Real*) Density +- ``E``: (*Real*) Young's modulus +- ``nu``: (*Real*) Poisson's ratio +- ``Plane_stress``: (*bool*) Plane stress simplification (only 2D problems) +- ``Yd``: (*Random*) Hardening modulus +- ``Sd``: (*Real*) Damage energy + + +Mazars +------ + +Keyword: :ref:`mazars ` + +Parameters: + +- ``rho``: (*Real*) Density +- ``E``: (*Real*) Young's modulus +- ``nu``: (*Real*) Poisson's ratio +- ``At``: (*Real*) Traction post-peak asymptotic value +- ``Bt``: (*Real*) Traction decay shape +- ``Ac``: (*Real*) Compression post-peak asymptotic value +- ``Bc``: (*Real*) Compression decay shape +- ``K0``: (*Real*) Damage threshold +- ``beta``: (*Real*) Shear parameter diff --git a/doc/dev-doc/manual/constitutive-laws.rst b/doc/dev-doc/manual/constitutive-laws.rst index 1ae6c6c71..3416daffe 100644 --- a/doc/dev-doc/manual/constitutive-laws.rst +++ b/doc/dev-doc/manual/constitutive-laws.rst @@ -1,917 +1,925 @@ .. _sect-smm-cl: Constitutive Laws ----------------- In order to compute an element’s response to deformation, one needs to use an appropriate constitutive relationship. The constitutive law is used to compute the element’s stresses from the element’s strains. In the finite-element discretization, the constitutive formulation is applied to every quadrature point of each element. When the implicit formulation is used, the tangent matrix has to be computed. | The chosen materials for the simulation have to be specified in the mesh file or, as an alternative, they can be assigned using the at ``element_material`` vector. For every material assigned to the problem one has to specify the material characteristics (constitutive behavior and material properties) using the text input file (see :ref:`sect-io-material`). | In order to conveniently store values at each quadrature in a material point ``Akantu`` provides a special data structure, the at :cpp:class:`InternalField `. The internal fields are inheriting from the at :cpp:class:`ElementTypeMapArray `. Furthermore, it provides several functions for initialization, auto-resizing and auto removal of quadrature points. Sometimes it is also desired to generate random distributions of internal parameters. An example might be the critical stress at which the material fails. To generate such a field, in the text input file, a random quantity needs be added to the base value: All parameters are real numbers. For the uniform distribution, minimum and maximum values have to be specified. Random parameters are defined as a :math:`base` value to which we add a random number that follows the chosen distribution. The `Uniform `__ distribution is gives a random values between in :math:`[min, max)`. The `Weibull `__ distribution is characterized by the following cumulative distribution function: .. math:: F(x) = 1- e^{-\left({x/\lambda}\right)^m} which depends on :math:`m` and :math:`\lambda`, which are the shape parameter and the scale parameter. These random distributions are different each time the code is executed. In order to obtain always the same one, it possible to manually set the *seed* that is the number from which these pseudo-random distributions are created. This can be done by adding the following line to the input file *outside* the material parameters environments: .. code-block:: seed = 1.0 where the value 1.0 can be substituted with any number. Currently ``Akantu`` can reproduce always the same distribution when the seed is specified *only* in serial. The value of the *seed* can be also specified directly in the code (for instance in the main file) with the command: .. code-block:: RandGenerator::seed(1.0) The same command, with empty brackets, can be used to check the value of the *seed* used in the simulation. The following sections describe the constitutive models implemented in ``Akantu``. In Appendix `7 <#app:material-parameters>`__ a summary of the parameters for all materials of ``Akantu`` is provided. Elastic ``````` The elastic law is a commonly used constitutive relationship that can be used for a wide range of engineering materials (*e.g.*, metals, concrete, rock, wood, glass, rubber, etc.) provided that the strains remain small (*i.e.*, small deformation and stress lower than yield strength). The elastic laws are often expressed as :math:`\boldsymbol{\sigma} = \boldsymbol{C}:\boldsymbol{\varepsilon}` with where :math:`\boldsymbol{\sigma}` is the Cauchy stress tensor, :math:`\boldsymbol{\varepsilon}` represents the infinitesimal strain tensor and :math:`\boldsymbol{C}` is the elastic modulus tensor. .. _sect-smm-linear-elastic-isotropic: Linear isotropic '''''''''''''''' The linear isotropic elastic behavior is described by Hooke’s law, which states that the stress is linearly proportional to the applied strain (material behaves like an ideal spring), as illustrated in -Figure `[fig:smm:cl:elastic] <#fig:smm:cl:elastic>`__. + :numref:`fig:smm:cl:el`. + +.. figure:: figures/cl/stress_strain_el.svg + :alt: Elastic strain-stress curve + :name: fig:smm:cl:el + :width: 60.0% + + Stress-strain curve of elastic material and schematic representation of + Hooke's law, denoted as a spring. The equation that relates the strains to the displacements is: point) from the displacements as follows: .. math:: \label{eqn:smm:strain_inf} \boldsymbol{\varepsilon} = \frac{1}{2} \left[ \nabla_0 \boldsymbol{u}+\nabla_0 \boldsymbol{u}^T \right] where :math:`\boldsymbol{\varepsilon}` represents the infinitesimal strain tensor, :math:`\nabla_{0}\boldsymbol{u}` the displacement gradient tensor according to the initial configuration. The constitutive equation for isotropic homogeneous media can be expressed as: .. math:: \label{eqn:smm:material:constitutive_elastic} \boldsymbol{\sigma } =\lambda\mathrm{tr}(\boldsymbol{\varepsilon})\boldsymbol{I}+2 \mu\boldsymbol{\varepsilon} where :math:`\boldsymbol{\sigma}` is the Cauchy stress tensor (:math:`\lambda` and :math:`\mu` are the the first and second Lame’s coefficients). In Voigt notation this correspond to .. math:: \begin{aligned} \left[\begin{array}{c} \sigma_{11}\\ \sigma_{22}\\ \sigma_{33}\\ \sigma_{23}\\ \sigma_{13}\\ \sigma_{12}\\ \end{array}\right] &= \frac{E}{(1+\nu)(1-2\nu)}\left[ \begin{array}{cccccc} 1-\nu & \nu & \nu & 0 & 0 & 0\\ \nu & 1-\nu & \nu & 0 & 0 & 0\\ \nu & \nu & 1-\nu & 0 & 0 & 0\\ 0 & 0 & 0 & \frac{1-2\nu}{2} & 0 & 0 \\ 0 & 0 & 0 & 0 & \frac{1-2\nu}{2} & 0 \\ 0 & 0 & 0 & 0 & 0 & \frac{1-2\nu}{2} \\ \end{array}\right] \left[\begin{array}{c} \varepsilon_{11}\\ \varepsilon_{22}\\ \varepsilon_{33}\\ 2\varepsilon_{23}\\ 2\varepsilon_{13}\\ 2\varepsilon_{12}\\ \end{array}\right]\end{aligned} .. _sect-smm-linear-elastic-anisotropic: Linear anisotropic '''''''''''''''''' This formulation is not sufficient to represent all elastic material behavior. Some materials have characteristic orientation that have to be taken into account. To represent this anisotropy a more general stress-strain law has to be used. For this we define the elastic modulus tensor as follow: .. math:: \begin{aligned} \left[\begin{array}{c} \sigma_{11}\\ \sigma_{22}\\ \sigma_{33}\\ \sigma_{23}\\ \sigma_{13}\\ \sigma_{12}\\ \end{array}\right] &= \left[ \begin{array}{cccccc} c_{11} & c_{12} & c_{13} & c_{14} & c_{15} & c_{16}\\ c_{21} & c_{22} & c_{23} & c_{24} & c_{25} & c_{26}\\ c_{31} & c_{32} & c_{33} & c_{34} & c_{35} & c_{36}\\ c_{41} & c_{42} & c_{43} & c_{44} & c_{45} & c_{46}\\ c_{51} & c_{52} & c_{53} & c_{54} & c_{55} & c_{56}\\ c_{61} & c_{62} & c_{63} & c_{64} & c_{65} & c_{66}\\ \end{array}\right] \left[\begin{array}{c} \varepsilon_{11}\\ \varepsilon_{22}\\ \varepsilon_{33}\\ 2\varepsilon_{23}\\ 2\varepsilon_{13}\\ 2\varepsilon_{12}\\ \end{array}\right]\end{aligned} To simplify the writing of input files the :math:`\boldsymbol{C}` tensor is expressed in the material basis. And this basis as to be given too. This basis :math:`\Omega_{{\mathrm{mat}}} = \{\boldsymbol{n_1}, \boldsymbol{n_2}, \boldsymbol{n_3}\}` is used to define the rotation :math:`R_{ij} = \boldsymbol{n_j} . \boldsymbol{e_i}`. And :math:`\boldsymbol{C}` can be rotated in the global basis :math:`\Omega = \{\boldsymbol{e_1}, \boldsymbol{e_2}, \boldsymbol{e_3}\}` as follow: .. math:: \begin{aligned} \boldsymbol{C}_{\Omega} &= \boldsymbol{R}_1 \boldsymbol{C}_{\Omega_{{\mathrm{mat}}}} \boldsymbol{R}_2\\ \boldsymbol{R}_1 &= \left[ \begin{array}{cccccc} R_{11} R_{11} & R_{12} R_{12} & R_{13} R_{13} & R_{12} R_{13} & R_{11} R_{13} & R_{11} R_{12}\\ R_{21} R_{21} & R_{22} R_{22} & R_{23} R_{23} & R_{22} R_{23} & R_{21} R_{23} & R_{21} R_{22}\\ R_{31} R_{31} & R_{32} R_{32} & R_{33} R_{33} & R_{32} R_{33} & R_{31} R_{33} & R_{31} R_{32}\\ R_{21} R_{31} & R_{22} R_{32} & R_{23} R_{33} & R_{22} R_{33} & R_{21} R_{33} & R_{21} R_{32}\\ R_{11} R_{31} & R_{12} R_{32} & R_{13} R_{33} & R_{12} R_{33} & R_{11} R_{33} & R_{11} R_{32}\\ R_{11} R_{21} & R_{12} R_{22} & R_{13} R_{23} & R_{12} R_{23} & R_{11} R_{23} & R_{11} R_{22}\\ \end{array}\right]\\ \boldsymbol{R}_2 &= \left[ \begin{array}{cccccc} R_{11} R_{11} & R_{21} R_{21} & R_{31} R_{31} & R_{21} R_{31} & R_{11} R_{31} & R_{11} R_{21}\\ R_{12} R_{12} & R_{22} R_{22} & R_{32} R_{32} & R_{22} R_{32} & R_{12} R_{32} & R_{12} R_{22}\\ R_{13} R_{13} & R_{23} R_{23} & R_{33} R_{33} & R_{23} R_{33} & R_{13} R_{33} & R_{13} R_{23}\\ R_{12} R_{13} & R_{22} R_{23} & R_{32} R_{33} & R_{22} R_{33} & R_{12} R_{33} & R_{12} R_{23}\\ R_{11} R_{13} & R_{21} R_{23} & R_{31} R_{33} & R_{21} R_{33} & R_{11} R_{33} & R_{11} R_{23}\\ R_{11} R_{12} & R_{21} R_{22} & R_{31} R_{32} & R_{21} R_{32} & R_{11} R_{32} & R_{11} R_{22}\\ \end{array}\right]\\\end{aligned} .. _sect-smm-linear-elastic-orthotropic: Linear orthotropic '''''''''''''''''' A particular case of anisotropy is when the material basis is orthogonal in which case the elastic modulus tensor can be simplified and rewritten in terms of 9 independents material parameters. .. math:: \begin{aligned} \left[\begin{array}{c} \sigma_{11}\\ \sigma_{22}\\ \sigma_{33}\\ \sigma_{23}\\ \sigma_{13}\\ \sigma_{12}\\ \end{array}\right] &= \left[ \begin{array}{cccccc} c_{11} & c_{12} & c_{13} & 0 & 0 & 0 \\ & c_{22} & c_{23} & 0 & 0 & 0 \\ & & c_{33} & 0 & 0 & 0 \\ & & & c_{44} & 0 & 0 \\ & \multicolumn{2}{l}{\text{sym.}} & & c_{55} & 0 \\ & & & & & c_{66}\\ \end{array}\right] \left[\begin{array}{c} \varepsilon_{11}\\ \varepsilon_{22}\\ \varepsilon_{33}\\ 2\varepsilon_{23}\\ 2\varepsilon_{13}\\ 2\varepsilon_{12}\\ \end{array}\right]\end{aligned} .. math:: \begin{aligned} c_{11} &= E_1 (1 - \nu_{23}\nu_{32})\Gamma \qquad c_{22} = E_2 (1 - \nu_{13}\nu_{31})\Gamma \qquad c_{33} = E_3 (1 - \nu_{12}\nu_{21})\Gamma\\ c_{12} &= E_1 (\nu_{21} - \nu_{31}\nu_{23})\Gamma = E_2 (\nu_{12} - \nu_{32}\nu_{13})\Gamma\\ c_{13} &= E_1 (\nu_{31} - \nu_{21}\nu_{32})\Gamma = E_2 (\nu_{13} - \nu_{21}\nu_{23})\Gamma\\ c_{23} &= E_2 (\nu_{32} - \nu_{12}\nu_{31})\Gamma = E_3 (\nu_{23} - \nu_{21}\nu_{13})\Gamma\\ c_{44} &= \mu_{23} \qquad c_{55} = \mu_{13} \qquad c_{66} = \mu_{12} \\ \Gamma &= \frac{1}{1 - \nu_{12} \nu_{21} - \nu_{13} \nu_{31} - \nu_{32} \nu_{23} - 2 \nu_{21} \nu_{32} \nu_{13}}\end{aligned} The Poisson ratios follow the rule :math:`\nu_{ij} = \nu_{ji} E_i / E_j`. .. _sect-smm-cl-neohookean: Neo-Hookean ''''''''''' The hyperelastic Neo-Hookean constitutive law results from an extension of the linear elastic relationship (Hooke’s Law) for large deformation. Thus, the model predicts nonlinear stress-strain behavior for bodies undergoing large deformations. .. figure:: figures/cl/stress_strain_neo.svg :alt: Neo-hookean Stress-strain curve. :name: fig:smm:cl:neo_hookean :width: 40.0% Neo-hookean Stress-strain curve. As illustrated in :numref:`fig:smm:cl:neo_hookean`, the behavior is initially linear and the mechanical behavior is very close to the corresponding linear elastic material. This constitutive relationship, which accounts for compressibility, is a modified version of the one proposed by Ronald Rivlin :cite:`Belytschko:2000`. The strain energy stored in the material is given by: .. math:: \label{eqn:smm:constitutive:neohookean_potential} \Psi(\boldsymbol{C}) = \frac{1}{2}\lambda_0\left(\ln J\right)^2-\mu_0\ln J+\frac{1}{2} \mu_0\left(\mathrm{tr}(\boldsymbol{C})-3\right) where :math:`\lambda_0` and :math:`\mu_0` are, respectively, Lamé’s first parameter and the shear modulus at the initial configuration. :math:`J` is the jacobian of the deformation gradient (:math:`\boldsymbol{F}=\nabla_{\!\!\boldsymbol{X}}\boldsymbol{x}`): :math:`J=\text{det}(\boldsymbol{F})`. Finally :math:`\boldsymbol{C}` is the right Cauchy-Green deformation tensor. Since this kind of material is used for large deformation problems, a finite deformation framework should be used. Therefore, the Cauchy stress (:math:`\boldsymbol{\sigma}`) should be computed through the second Piola-Kirchhoff stress tensor :math:`\boldsymbol{S}`: .. math:: \boldsymbol{\sigma } = \frac{1}{J}\boldsymbol{F}\boldsymbol{S}\boldsymbol{F}^T Finally the second Piola-Kirchhoff stress tensor is given by: .. math:: \boldsymbol{S} = 2\frac{\partial\Psi}{\partial\boldsymbol{C}} = \lambda_0\ln J \boldsymbol{C}^{-1}+\mu_0\left(\boldsymbol{I}-\boldsymbol{C}^{-1}\right) The parameters to indicate in the material file are the same as those for the elastic case: ``E`` (Young’s modulus), ``nu`` (Poisson’s ratio). .. _sect-smm-cl-sls: Visco-Elasticity '''''''''''''''' Visco-elasticity is characterized by strain rate dependent behavior. Moreover, when such a material undergoes a deformation it dissipates energy. This dissipation results in a hysteresis loop in the stress-strain curve at every loading cycle (see :numref:`fig:smm:cl:visco-elastic:hyst`). In principle, it can be applied to many materials, since all materials exhibit a visco-elastic behavior if subjected to particular conditions (such as high temperatures). .. figure:: figures/cl/stress_strain_visco.svg :name: fig:smm:cl:visco-elastic:hyst :align: center :width: 40.0% Characteristic stress-strain behavior of a visco-elastic material with hysteresis loop .. figure:: figures/cl/visco_elastic_law.svg :name: fig:smm:cl:visco-elastic:model :align: center :width: 40.0% Schematic representation of the standard rheological linear solid visco-elastic model The standard rheological linear solid model (see Sections 10.2 and 10.3 of :cite:`simo92`) has been implemented in ``Akantu``. This model results from the combination of a spring mounted in parallel with a spring and a dashpot connected in series, as illustrated in :numref:`fig:smm:cl:visco-elastic:model`. The advantage of this model is that it allows to account for creep or stress relaxation. The equation that relates the stress to the strain is (in 1D): .. math:: \frac{d\varepsilon(t)}{dt} = \left ( E + E_V \right ) ^ {-1} \cdot \left [ \frac{d\sigma(t)}{dt} + \frac{E_V}{\eta}\sigma(t) - \frac{EE_V}{\eta}\varepsilon(t) \right ] where :math:`\eta` is the viscosity. The equilibrium condition is unique and is attained in the limit, as :math:`t \to \infty`. At this stage, the response is elastic and depends on the Young’s modulus :math:`E`. The mandatory parameters for the material file are the following: ``rho`` (density), ``E`` (Young’s modulus), ``nu`` (Poisson’s ratio), ``Plane_Stress`` (if set to zero plane strain, otherwise plane stress), ``eta`` (dashpot viscosity) and ``Ev`` (stiffness of the viscous element). Note that the current standard linear solid model is applied only on the deviatoric part of the strain tensor. The spheric part of the strain tensor affects the stress tensor like an linear elastic material. .. _sect-smm-cl-plastic: Plastic ``````` Small-Deformation Plasticity '''''''''''''''''''''''''''' The small-deformation plasticity is a simple plasticity material formulation which accounts for the additive decomposition of strain into elastic and plastic strain components. This formulation is applicable to infinitesimal deformation where the additive decomposition of the strain is a valid approximation. In this formulation, plastic strain is a shearing process where hydrostatic stress has no contribution to plasticity and consequently plasticity does not lead to volume change. :numref:`fig:smm:cl:Lin-strain-hard` shows the linear strain hardening elasto-plastic behavior according to the additive decomposition of strain into the elastic and plastic parts in infinitesimal deformation as .. math:: \boldsymbol{\varepsilon} &= \boldsymbol{\varepsilon}^e +\boldsymbol{\varepsilon}^p\\ \boldsymbol{\sigma} &= 2G(\boldsymbol{\varepsilon}^e) + \lambda \mathrm{tr}(\boldsymbol{\varepsilon}^e)\boldsymbol{I} .. figure:: figures/cl/isotropic_hardening_plasticity.svg :name: fig:smm:cl:Lin-strain-hard :align: center Stress-strain curve for the small-deformation plasticity with linear isotropic hardening. In this class, the von Mises yield criterion is used. In the von Mises yield criterion, the yield is independent of the hydrostatic stress. Other yielding criteria such as Tresca and Gurson can be easily implemented in this class as well. In the von Mises yield criterion, the hydrostatic stresses have no effect on the plasticity and consequently the yielding occurs when a critical elastic shear energy is achieved. .. math:: \label{eqn:smm:constitutive:von Mises} f = \sigma_{{\mathrm{eff}}} - \sigma_y = \left(\frac{3}{2} {\boldsymbol{\sigma}}^{{\mathrm{tr}}} : {\boldsymbol{\sigma}}^{{\mathrm{tr}}}\right)^\frac{1}{2}-\sigma_y (\boldsymbol{\varepsilon}^p) .. math:: \label{eqn:smm:constitutive:yielding} f < 0 \quad \textrm{Elastic deformation,} \qquad f = 0 \quad \textrm{Plastic deformation} where :math:`\sigma_y` is the yield strength of the material which can be function of plastic strain in case of hardening type of materials and :math:`{\boldsymbol{\sigma}}^{{\mathrm{tr}}}` is the deviatoric part of stress given by .. math:: \label{eqn:smm:constitutive:deviatoric stress} {\boldsymbol{\sigma}}^{{\mathrm{tr}}}=\boldsymbol{\sigma} - \frac{1}{3} \mathrm{tr}(\boldsymbol{\sigma}) \boldsymbol{I} After yielding :math:`(f = 0)`, the normality hypothesis of plasticity determines the direction of plastic flow which is normal to the tangent to the yielding surface at the load point. Then, the tensorial form of the plastic constitutive equation using the von Mises yielding criterion (see equation 4.34) may be written as .. math:: \label{eqn:smm:constitutive:plastic contitutive equation} \Delta {\boldsymbol{\varepsilon}}^p = \Delta p \frac {\partial{f}}{\partial{\boldsymbol{\sigma}}}=\frac{3}{2} \Delta p \frac{{\boldsymbol{\sigma}}^{{\mathrm{tr}}}}{\sigma_{{\mathrm{eff}}}} In these expressions, the direction of the plastic strain increment (or equivalently, plastic strain rate) is given by :math:`\frac{{\boldsymbol{\sigma}}^{{\mathrm{tr}}}}{\sigma_{{\mathrm{eff}}}}` while the magnitude is defined by the plastic multiplier :math:`\Delta p`. This can be obtained using the *consistency condition* which impose the requirement for the load point to remain on the yielding surface in the plastic regime. Here, we summarize the implementation procedures for the small-deformation plasticity with linear isotropic hardening: #. Compute the trial stress: .. math:: {\boldsymbol{\sigma}}^{{\mathrm{tr}}} = {\boldsymbol{\sigma}}_t + 2G\Delta \boldsymbol{\varepsilon} + \lambda \mathrm{tr}(\Delta \boldsymbol{\varepsilon})\boldsymbol{I} #. Check the Yielding criteria: .. math:: f = (\frac{3}{2} {\boldsymbol{\sigma}}^{{\mathrm{tr}}} : {\boldsymbol{\sigma}}^{{\mathrm{tr}}})^{1/2}-\sigma_y (\boldsymbol{\varepsilon}^p) #. Compute the Plastic multiplier: .. math:: \begin{aligned} d \Delta p &= \frac{\sigma^{tr}_{eff} - 3G \Delta P^{(k)}- \sigma_y^{(k)}}{3G + h}\\ \Delta p^{(k+1)} &= \Delta p^{(k)}+ d\Delta p\\ \sigma_y^{(k+1)} &= (\sigma_y)_t+ h\Delta p \end{aligned} #. Compute the plastic strain increment: .. math:: \Delta {\boldsymbol{\varepsilon}}^p = \frac{3}{2} \Delta p \frac{{\boldsymbol{\sigma}}^{{\mathrm{tr}}}}{\sigma_{{\mathrm{eff}}}} #. Compute the stress increment: .. math:: {\Delta \boldsymbol{\sigma}} = 2G(\Delta \boldsymbol{\varepsilon}-\Delta \boldsymbol{\varepsilon}^p) + \lambda \mathrm{tr}(\Delta \boldsymbol{\varepsilon}-\Delta \boldsymbol{\varepsilon}^p)\boldsymbol{I} #. Update the variables: .. math:: \begin{aligned} {\boldsymbol{\varepsilon^p}} &= {\boldsymbol{\varepsilon}}^p_t+{\Delta {\boldsymbol{\varepsilon}}^p}\\ {\boldsymbol{\sigma}} &= {\boldsymbol{\sigma}}_t+{\Delta \boldsymbol{\sigma}} \end{aligned} We use an implicit integration technique called *the radial return method* to obtain the plastic multiplier. This method has the advantage of being unconditionally stable, however, the accuracy remains dependent on the step size. The plastic parameters to indicate in the material file are: :math:`\sigma_y` (Yield stress) and ``h`` (Hardening modulus). In addition, the elastic parameters need to be defined as previously mentioned: ``E`` (Young’s modulus), ``nu`` (Poisson’s ratio). Damage `````` In the simplified case of a linear elastic and brittle material, isotropic damage can be represented by a scalar variable :math:`d`, which varies from :math:`0` to :math:`1` for no damage to fully broken material respectively. The stress-strain relationship then becomes: .. math:: \boldsymbol{\sigma} = (1-d)\, \boldsymbol{C}:\boldsymbol{\varepsilon} where :math:`\boldsymbol{\sigma}`, :math:`\boldsymbol{\varepsilon}` are the Cauchy stress and strain tensors, and :math:`\boldsymbol{C}` is the elastic stiffness tensor. This formulation relies on the definition of an evolution law for the damage variable. In ``Akantu``, many possibilities exist and they are listed below. .. _sect-smm-cl-damage-marigo: Marigo '''''' This damage evolution law is energy based as defined by Marigo :cite:`marigo81a`, :cite:`lemaitre96a`. It is an isotropic damage law. .. math:: \begin{aligned} Y &= \frac{1}{2}\boldsymbol{\varepsilon}:\boldsymbol{C}:\boldsymbol{\varepsilon}\\ F &= Y - Y_d - S d\\ d &= \left\{ \begin{array}{l l} \mathrm{min}\left(\frac{Y-Y_d}{S},\;1\right) & \mathrm{if}\; F > 0\\ \mathrm{unchanged} & \mathrm{otherwise} \end{array} \right.\end{aligned} In this formulation, :math:`Y` is the strain energy release rate, :math:`Y_d` the rupture criterion and :math:`S` the damage energy. The non-local version of this damage evolution law is constructed by averaging the energy :math:`Y`. .. _sect-smm-cl-damage-mazars: Mazars '''''' This law introduced by Mazars :cite:`mazars84a` is a behavioral model to represent damage evolution in concrete. This model does not rely on the computation of the tangent stiffness, the damage is directly evaluated from the strain. The governing variable in this damage law is the equivalent strain :math:`\varepsilon_{{\mathrm{eq}}} = \sqrt{<\boldsymbol{\varepsilon}>_+:<\boldsymbol{\varepsilon}>_+}`, with :math:`<.>_+` the positive part of the tensor. This part is defined in the principal coordinates (I, II, III) as :math:`\varepsilon_{{\mathrm{eq}}} = \sqrt{<\boldsymbol{\varepsilon_I}>_+^2 + <\boldsymbol{\varepsilon_{II}}>_+^2 + <\boldsymbol{\varepsilon_{III}}>_+^2}`. The damage is defined as: .. math:: \begin{aligned} D &= \alpha_t^\beta D_t + (1-\alpha_t)^\beta D_c\\ D_t &= 1 - \frac{\kappa_0 (1- A_t)}{\varepsilon_{{\mathrm{eq}}}} - A_t \exp^{-B_t(\varepsilon_{{\mathrm{eq}}}-\kappa_0)}\\ D_c &= 1 - \frac{\kappa_0 (1- A_c)}{\varepsilon_{{\mathrm{eq}}}} - A_c \exp^{-B_c(\varepsilon_{{\mathrm{eq}}}-\kappa_0)}\\ \alpha_t &= \frac{\sum_{i=1}^3<\varepsilon_i>_+\varepsilon_{{\mathrm{nd}}\;i}}{\varepsilon_{{\mathrm{eq}}}^2}\end{aligned} With :math:`\kappa_0` the damage threshold, :math:`A_t` and :math:`B_t` the damage parameter in traction, :math:`A_c` and :math:`B_c` the damage parameter in compression, :math:`\beta` is the shear parameter. :math:`\alpha_t` is the coupling parameter between traction and compression, the :math:`\varepsilon_i` are the eigenstrain and the :math:`\varepsilon_{{\mathrm{nd}}\;i}` are the eigenvalues of the strain if the material were undamaged. The coefficients :math:`A` and :math:`B` are the post-peak asymptotic value and the decay shape parameters. .. _sect:smm:CLNL: Non-Local Constitutive Laws ``````````````````````````` Continuum damage modeling of quasi-brittle materials undergo significant softening after the onset of damage. This fast growth of damage causes a loss of ellipticity of partial differential equations of equilibrium. Therefore, the numerical simulation results won't be objective anymore, because the dissipated energy will depend on mesh size used in the simulation. One way to avoid this effect is the use of non-local damage formulations. In this approach a local quantity such as the strain is replaced by its non-local average, where the size of the domain, over which the quantitiy is averaged, depends on the underlying material microstructure. ``Akantu`` provides non-local versions of many constitutive laws for damage. Examples are for instance the material :ref:`sect-smm-cl-damage-mazars` and the material :ref:`sect-smm-cl-damage-marigo`, that can be used in a non-local context. In order to use the corresponding non-local formulation the user has to define the non-local material he wishes to use in the text input file: -.. code-block:: +.. code-block:: none material constitutive_law_non_local [ name = material_name rho = $value$ ... ] where ``constitutive_law_non_local`` is the name of the non-local constitutive law, *e.g.* `marigo_non_local`. In addition to the material the non-local neighborhood, that should be used for the averaging process needs to be defined in the material file as well: -.. code-block:: +.. code-block:: none non_local neighborhood_name weight_function_type [ radius = $value$ ... weight_function weight_parameter [ damage_limit = $value$ ... ] ] for the non-local averaging, *e.g.* ``base_wf``, followed by the properties of the non-local neighborhood, such as the radius, and the weight function parameters. It is important to notice that the non-local neighborhood must have the same name as the material to which the neighborhood belongs! The following two sections list the non-local constitutive laws and different type of weight functions available in ``Akantu``. \subsection{Non-local constitutive laws} Let us consider a body having a volume :math:`V` and a boundary :math:`\Gamma`. The stress-strain relation for a non-local damage model can be described as follows: .. _eq:non-local-const: .. math:: \vec{\sigma} = (1-\bar{d}) \vec{D}:\epsilon with :math:`\vec{D}` the elastic moduli tensor, :math:`\sigma` the stress tensor, :math:`\epsilon` the strain tensor and :math:`\bar{d}` the non-local damage variable. Note that this stres-strain relationship is similar to the relationship defined in Damage model except :math:`\bar{d}`. The non-local damage model can be extended to the damage constitutive laws: :ref:`sect-smm-cl-damage-marigo` and :ref:`sect-smm-cl-damage-mazars`. The non-local damage variable :math:`\bar{d}` is defined as follows: .. _eq:non-local-const: .. math:: \bar{d}(\vec{x}) = \int_{V}W(\vec{x}, \vec{y}) d(\vec{y}) dV(\vec{y}) with :math:`W(\vec{x},\vec{y})` the weight function which averages local damage variables to describe the non-local interactions. A list of available weight functions and its functionalities in \akantu are explained in the next section. Non-local weight functions '''''''''''''''''''''''''' The available weight functions in ``Akantu`` are follows: - ``base_weight_function``: This weight function averages local damage variables by using a bell-shape function on spatial dimensions. - ``damaged_weight_function``: A linear-shape weight function is applied to average local damage variables. Its slope is determined by damage variables. For example, the damage variables for an element which is highly damaged are averaged over large spatial dimension (linear function including a small slope). - ``remove_damaged_weight_function``: This weight function averages damage values by using a bell-shape function as ``base_weight_function``, but excludes elements which are fully damaged. - ``remove_damaged_with_damage_rate_weight_function``: A bell-shape function is applied to average local damage variables for elements having small damage rates. - ``stress_based_weight_function``: Non local integral takes stress states, and use the states to construct weight function: an ellipsoid shape. Detailed explanations of this weight function are given in Giry et al. :cite:`giry13a`. .. _sec-cohesive-laws: Cohesive Constitutive laws `````````````````````````` .. _ssect-smm-cl-coh-snozzi: Linear Irreversible Law ''''''''''''''''''''''' .. figure:: figures/cl/linear_cohesive_law.svg :alt: Irreversible cohesive laws for explicit simulations. :name: fig:smm:coh:linear_cohesive_law :align: center :width: 60.0% Irreversible cohesive laws for explicit simulations. `Akantu` includes the Snozzi-Molinari :cite:`snozzi_cohesive_2013` linear irreversible cohesive law (see :numref:`fig:smm:coh:linear_cohesive_law`). It is an extension to the Camacho-Ortiz :cite:`camacho_computational_1996` cohesive law in order to make dissipated fracture energy path-dependent. The concept of free potential energy is dropped and a new independent parameter :math:`\kappa` is introduced: .. math:: \kappa = \frac{G_\mathrm{c, II}}{G_\mathrm{c, I}} where :math:`G_\mathrm{c, I}` and :math:`G_\mathrm{c, II}` are the necessary works of separation per unit area to open completely a cohesive zone under mode I and mode II, respectively. Their model yields to the following equation for cohesive tractions :math:`\vec{T}` in case of crack opening :math:`{\delta}`: .. math:: \vec{T} = \left( \frac{\beta^2}{\kappa} \Delta_\mathrm{t} \vec{t} + \Delta_\mathrm{n} \vec{n} \right) \frac{\sigma_\mathrm{c}}{\delta} \left( 1- \frac{\delta}{\delta_\mathrm{c}} \right) = \hat{\vec T}\, \frac{\sigma_\mathrm{c}}{\delta} \left( 1- \frac{\delta}{\delta_\mathrm{c}} \right) :label: eq-smm-coh-tractions where :math:`\sigma_\mathrm{c}` is the material strength along the fracture, :math:`\delta_\mathrm{c}` the critical effective displacement after which cohesive tractions are zero (complete decohesion), :math:`\Delta_\mathrm{t}` and :math:`\Delta_\mathrm{n}` are the tangential and normal components of the opening displacement vector :math:`\vec{\Delta}`, respectively. The parameter :math:`\beta` is a weight that indicates how big the tangential opening contribution is. The effective opening displacement is: .. math:: \delta = \sqrt{\frac{\beta^2}{\kappa^2} \Delta_\mathrm{t}^2 + \Delta_\mathrm{n}^2} In case of unloading or reloading :math:`\delta < \delta_\mathrm{max}`, tractions are calculated as: .. math:: \begin{eqnarray} T_\mathrm{n} &= \Delta_\mathrm{n}\, \frac{\sigma_\mathrm{c}}{\delta_\mathrm{max}} \left( 1- \frac{\delta_\mathrm{max}}{\delta_\mathrm{c}} \right) \\ T_\mathrm{t} &= \frac{\beta^2}{\kappa}\, \Delta_\mathrm{t}\, \frac{\sigma_\mathrm{c}}{\delta_\mathrm{max}} \left( 1- \frac{\delta_\mathrm{max}}{\delta_\mathrm{c}} \right) \end{eqnarray} so that they vary linearly between the origin and the maximum attained tractions. As shown in :numref:`fig:smm:coh:linear_cohesive_law`, in this law, the dissipated and reversible energies are: .. math:: \begin{eqnarray} E_\mathrm{diss} &= \frac{1}{2} \sigma_\mathrm{c}\, \delta_\mathrm{max}\\[1ex] E_\mathrm{rev} &= \frac{1}{2} T\, \delta \end{eqnarray} Moreover, a damage parameter :math:`D` can be defined as: .. math:: D = \min \left( \frac{\delta_\mathrm{max}}{\delta_\mathrm{c}},1 \right) which varies from 0 (undamaged condition) and 1 (fully damaged condition). This variable can only increase because damage is an irreversible process. A simple penalty contact model has been incorporated in the cohesive law so that normal tractions can be returned in case of compression: .. math:: T_\mathrm{n} = \alpha \Delta_\mathrm{n} \quad\text{if}\quad \Delta_\mathrm{n}\quad <\quad 0 where :math:`\alpha` is a stiffness parameter that defaults to zero. The relative contact energy is equivalent to reversible energy but in compression. The material name of the linear decreasing cohesive law is ``material_cohesive_linear`` and its parameters with their respective default values are: - ``sigma_c = 0`` - ``delta_c = 0`` - ``beta = 0`` - ``G_c = 0`` - ``kappa = 1`` - ``penalty = 0`` where ``G_c`` corresponds to :math:`G_\mathrm{c, I}`. A random number generator can be used to assign a random :math:`\sigma_\mathrm{c}` to each facet following a given distribution (see Section :ref:`sect-smm-cl`). Only one parameter between ``delta_c`` and ``G_c`` has to be specified. For random :math:`\sigma_\mathrm{c}` distributions, the chosen parameter of these two is kept fixed and the other one is varied. The bi-linear constitutive law works exactly the same way as the linear one, except for the additional parameter ``delta_0`` that by default is zero. Two examples for the extrinsic and intrinsic cohesive elements and also an example to assign different properties to inter-granular and trans-granular cohesive elements can be found in the folder ``examples/cohesive_element/``. .. _ssect:smm:cl:coh-friction: Linear Cohesive Law with Friction ''''''''''''''''''''''''''''''''' This law represents a variation of the linear irreversible cohesive of the previous section, which adds friction. The friction behavior is approximated with an elasto-plastic law, which relates the friction force to the relative sliding between the two faces of the cohesive element. The slope of the elastic branch is called ``penalty_for_friction``, and is defined by the user, together with the friction coefficient, as a material property. The friction contribution evolves with the damage of the cohesive law: it is null when the damage is zero, and it becomes maximum when the damage is equal to one. This is done by defining a current value of the friction coefficient (mu) that increases linearly with the damage, up to the value of the friction coefficient defined by the user. The yielding plateau of the friction law is given by the product of the current friction coefficient and the local compression stress acting in the cohesive element. Such an approach is equivalent to a node-to-node contact friction. Its accuracy is acceptable only for small displacements. The material name of the linear cohesive law with friction is ``material_cohesive_linear_friction``. Its additional parameters with respect to those of the linear cohesive law without friction, with the respective default values, are: - ``mu = 0`` - ``penalty_for_friction = 0`` .. _ssect:smm:cl:coh-fatigue: Linear Cohesive Law with Fatigue '''''''''''''''''''''''''''''''' This law represents a variation of the linear irreversible cohesive law of the previous section, that removes the hypothesis of elastic unloading-reloading cycles. With this law, some energy is dissipated also during unloading and reloading with hysteresis. The implementation follows the work of :cite:`nguyen2001`. During the unloading-reloading cycle, the traction increment is computed as .. math:: \dot{T} = \begin{cases} K^- \, \dot{\delta} & \text{if $\dot{\delta} < 0$} \\ K^+ \, \dot{\delta} & \text{if $\dot{\delta} > 0$} \\ \end{cases} where :math:`\dot{\delta}` and :math:`\dot{T}` are respectively the effective opening displacement and the cohesive traction increments with respect to time, while :math:`K^-` and :math:`K^+` are respectively the unloading and reloading incremental stiffness. The unloading path is linear and results in an unloading stiffness .. math:: K^- = \frac{T_\mathrm{max}}{\delta_\mathrm{max}} where :math:`T_\mathrm{max}` and :math:`\delta_\mathrm{max}` are the maximum cohesive traction and the effective opening displacement reached during the precedent loading phase. The unloading stiffness remains constant during the unloading phase. On the other hand the reloading stiffness increment :math:`\dot{K}^+` is calculated as .. math:: \dot{K}^+ = \begin{cases} - K^+ \, \dot{\delta} / \delta_\mathrm{f} & \text{if $\dot{\delta} > 0$} \\ \left( K^+ - K^- \right) \, \dot{\delta} / \delta_\mathrm{f} & \text{if $\dot{\delta}$ < $0$} \end{cases} where :math:`\delta_\mathrm{f}` is a material parameter (refer to :cite:`vocialta15` for more details). During unloading the stiffness :math:`K^+` tends to :math:`K^-`, while during reloading :math:`K^+` gets decreased at every time step. If the cohesive traction during reloading exceeds the upper limit given by equation :eq:`eq-smm-coh-tractions`, it is recomputed following the behavior of the linear decreasing cohesive law for crack opening. .. _ssect:smm:cl:coh-exponential: Exponential Cohesive Law ''''''''''''''''''''''''' Ortiz and Pandolfi proposed this cohesive law in 1999 :cite:`ortiz1999`. The traction-opening equation for this law is as follows: .. math:: T = e \sigma_c \frac{\delta}{\delta_c}e^{-\delta/ \delta_c} :label: eq:exponential_law -This equation is plotted in Figure :numref:`fig:smm:cl:ecl`. The term +This equation is plotted in :numref:`fig:smm:cl:ecl`. The term :math:`\partial{\vec{T}}/ \partial{\delta}` after the necessary derivation can expressed as .. math:: \frac{\partial{\vec{T}}} {\partial{\delta}} = \hat{\vec{T}} \otimes \frac {\partial{(T/\delta)}}{\partial{\delta}} \frac{\hat{\vec{T}}}{\delta}+ \frac{T}{\delta} \left[ \beta^2 \mat{I} + \left(1-\beta^2\right) \left(\vec{n} \otimes \vec{n}\right)\right] :label: eq:tangent_cohesive where .. math:: \frac{\partial{(T/ \delta)}}{\partial{\delta}} = \left\{\begin{array} {l l} -e \frac{\sigma_c}{\delta_c^2 }e^{-\delta / \delta_c} & \quad \text{if} \delta \geq \delta_{max}\\ 0 & \quad \text{if} \delta < \delta_{max}, \delta_n > 0 \end{array} \right. As regards the behavior in compression, two options are available: a contact penalty approach with stiffness following the formulation of the exponential law and a contact penalty approach with constant stiffness. In the second case, the stiffness is defined as a function of the tangent of the exponential law at the origin. .. figure:: figures/cl/cohesive_exponential.png :alt: Exponential cohesive law :name: fig:smm:cl:ecl :align: center Exponential cohesive law diff --git a/doc/dev-doc/manual/figures/cl/isotropic_hardening_plasticity.svg b/doc/dev-doc/manual/figures/cl/isotropic_hardening_plasticity.svg index 631e1c525..54dd40d38 100644 --- a/doc/dev-doc/manual/figures/cl/isotropic_hardening_plasticity.svg +++ b/doc/dev-doc/manual/figures/cl/isotropic_hardening_plasticity.svg @@ -1,1186 +1,181 @@ image/svg+xmlσεEhσyLinear Isotropic HardeningPerfect PlasticityEεpεeσ + width="545.57861" + height="444.22092" + viewBox="0 0 545.5786 444.22092" + sodipodi:docname="isotropic_hardening_plasticity.svg" + inkscape:version="0.92.4 (5da689c313, 2019-01-14)">image/svg+xml + + + + + +ε + +σ + +σ + + \ No newline at end of file diff --git a/doc/dev-doc/manual/figures/cl/stress_strain_el.svg b/doc/dev-doc/manual/figures/cl/stress_strain_el.svg index e7973b480..340d82871 100644 --- a/doc/dev-doc/manual/figures/cl/stress_strain_el.svg +++ b/doc/dev-doc/manual/figures/cl/stress_strain_el.svg @@ -1,166 +1,144 @@ image/svg+xmlεσEE + id="clipPath22" + clipPathUnits="userSpaceOnUse">ε + +σ + +E + +E + + \ No newline at end of file diff --git a/doc/dev-doc/manual/figures/cl/stress_strain_neo.svg b/doc/dev-doc/manual/figures/cl/stress_strain_neo.svg index a183cf930..2b8c92ac6 100644 --- a/doc/dev-doc/manual/figures/cl/stress_strain_neo.svg +++ b/doc/dev-doc/manual/figures/cl/stress_strain_neo.svg @@ -1,135 +1,209 @@ image/svg+xml + xmlns:xlink="http://www.w3.org/1999/xlink" + id="svg78" + version="1.2" + viewBox="0 0 408.828308 404.326477" + height="404.326477pt" + width="408.828308pt"> + + + + image/svg+xml + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + diff --git a/doc/dev-doc/manual/new-constitutive-laws.rst b/doc/dev-doc/manual/new-constitutive-laws.rst index 04c94884e..f8925e0fe 100644 --- a/doc/dev-doc/manual/new-constitutive-laws.rst +++ b/doc/dev-doc/manual/new-constitutive-laws.rst @@ -1,375 +1,373 @@ Adding a New Constitutive Law ----------------------------- There are several constitutive laws in ``Akantu`` as described in the previous Section :ref:`sect-smm-cl`. It is also possible to use a user-defined material for the simulation. These materials are referred to as local materials since they are local to the example of the user and not part of the ``Akantu`` library. To define a new local material, two files (``material_XXX.hh`` and ``material_XXX.cc``) have to be provided where ``XXX`` is the name of the new material. The header file ``material_XXX.hh`` defines the interface of your custom material. Its implementation is provided in the ``material_XXX.cc``. The new law must inherit from the :cpp:class:`Material ` class or any other existing material class. It is therefore necessary to include the interface of the parent material in the header file of your local material and -indicate the inheritance in the declaration of the class: - -.. code-block:: c++ +indicate the inheritance in the declaration of the class:: auto & solver = model.getNonLinearSolver(); solver.set("max_iterations", 1); solver.set("threshold", 1e-4); solver.set("convergence_type", SolveConvergenceCriteria::_residual); model.solveStep(); /* ---------------------------------------------------------------------- */ #include "material.hh" /* ---------------------------------------------------------------------- */ #ifndef __AKANTU_MATERIAL_XXX_HH__ #define __AKANTU_MATERIAL_XXX_HH__ namespace akantu { class MaterialXXX : public Material { /// declare here the interface of your material }; In the header file the user also needs to declare all the members of the new material. These include the parameters that a read from the material input file, as well as any other material parameters that will be computed during the simulation and internal variables. In the following the example of adding a new damage material will be presented. In this case the parameters in the material will consist of the Young's modulus, the Poisson coefficient, the resistance to damage and the damage threshold. The material will then from these values compute its Lamé coefficients and its bulk modulus. Furthermore, the user has to add a new internal variable ``damage`` in order to store the amount of damage at each quadrature point in each step of the simulation. For this specific material the member declaration inside the class will look as follows:: class LocalMaterialDamage : public Material { /// declare constructors/destructors here /// declare methods and accessors here /* -------------------------------------------------------------------- */ /* Class Members */ /* -------------------------------------------------------------------- */ AKANTU_GET_MACRO_BY_ELEMENT_TYPE_CONST(Damage, damage, Real); private: /// the young modulus Real E; /// Poisson coefficient Real nu; /// First Lame coefficient Real lambda; /// Second Lame coefficient (shear modulus) Real mu; /// resistance to damage Real Yd; /// damage threshold Real Sd; /// Bulk modulus Real kpa; /// damage internal variable InternalField damage; }; In order to enable to print the material parameters at any point in the user's example file using the standard output stream by typing:: for (UInt m = 0; m < model.getNbMaterials(); ++m) std::cout << model.getMaterial(m) << std::endl; the standard output stream operator has to be redefined. This should be done at the end of the header file:: class LocalMaterialDamage : public Material { /// declare here the interace of your material }: /* ---------------------------------------------------------------------- */ /* inline functions */ /* ---------------------------------------------------------------------- */ /// standard output stream operator inline std::ostream & operator <<(std::ostream & stream, const LocalMaterialDamage & _this) { _this.printself(stream); return stream; } However, the user still needs to register the material parameters that should be printed out. The registration is done during the call of the constructor. Like all definitions the implementation of the constructor has to be written in the ``material_XXX.cc`` file. However, the declaration has to be provided in the ``material_XXX.hh`` file:: class LocalMaterialDamage : public Material { /* -------------------------------------------------------------------- */ /* Constructors/Destructors */ /* -------------------------------------------------------------------- */ public: LocalMaterialDamage(SolidMechanicsModel & model, const ID & id = ""); }; The user can now define the implementation of the constructor in the ``material_XXX.cc`` file:: /* ---------------------------------------------------------------------- */ #include "local_material_damage.hh" #include "solid_mechanics_model.hh" namespace akantu { /* ---------------------------------------------------------------------- */ LocalMaterialDamage::LocalMaterialDamage(SolidMechanicsModel & model, const ID & id) : Material(model, id), damage("damage", *this) { AKANTU_DEBUG_IN(); this->registerParam("E", E, 0., _pat_parsable, "Young's modulus"); this->registerParam("nu", nu, 0.5, _pat_parsable, "Poisson's ratio"); this->registerParam("lambda", lambda, _pat_readable, "First Lame coefficient"); this->registerParam("mu", mu, _pat_readable, "Second Lame coefficient"); this->registerParam("kapa", kpa, _pat_readable, "Bulk coefficient"); this->registerParam("Yd", Yd, 50., _pat_parsmod); this->registerParam("Sd", Sd, 5000., _pat_parsmod); damage.initialize(1); AKANTU_DEBUG_OUT(); } During the intializer list the reference to the model and the material id are assigned and the constructor of the internal field is called. Inside the scope of the constructor the internal values have to be initialized and the parameters, that should be printed out, are registered with the function: ``registerParam``:: void registerParam(name of the parameter (key in the material file), member variable, default value (optional parameter), access permissions, description); The available access permissions are as follows: - ``_pat_internal``: Parameter can only be output when the material is printed. - ``_pat_writable``: User can write into the parameter. The parameter is output when the material is printed. - ``_pat_readable``: User can read the parameter. The parameter is output when the material is printed. - ``_pat_modifiable``: Parameter is writable and readable. - ``_pat_parsable``: Parameter can be parsed, *i.e.* read from the input file. - ``_pat_parsmod``: Parameter is modifiable and parsable. In order to implement the new constitutive law the user needs to specify how the additional material parameters, that are not defined in the input material file, should be calculated. Furthermore, it has to be defined how stresses and the stable time step should be computed for the new local material. In the case of implicit simulations, in addition, the computation of the tangent stiffness needs to be defined. Therefore, the user needs to redefine the following functions of the parent material:: void initMaterial(); // for explicit and implicit simulations void computeStress(ElementType el_type, GhostType ghost_type = _not_ghost); // for implicit simulations void computeTangentStiffness(const ElementType & el_type, Array & tangent_matrix, GhostType ghost_type = _not_ghost); // for explicit and implicit simulations Real getStableTimeStep(Real h, const Element & element); In the following a detailed description of these functions is provided: - ``initMaterial``: This method is called after the material file is fully read and the elements corresponding to each material are assigned. Some of the frequently used constant parameters are calculated in this method. For example, the Lam\'{e} constants of elastic materials can be considered as such parameters. - ``computeStress``: In this method, the stresses are computed based on the constitutive law as a function of the strains of the quadrature points. For example, the stresses for the elastic material are calculated based on the following formula: .. math:: \mat{\sigma } =\lambda\mathrm{tr}(\mat{\varepsilon})\mat{I}+2 \mu \mat{\varepsilon} Therefore, this method contains a loop on all quadrature points assigned to the material using the two macros: ``MATERIAL_STRESS_QUADRATURE_POINT_LOOP_BEGIN`` and ``MATERIAL_STRESS_QUADRATURE_POINT_LOOP_END`` .. code:: MATERIAL_STRESS_QUADRATURE_POINT_LOOP_BEGIN(element_type); // sigma <- f(grad_u) MATERIAL_STRESS_QUADRATURE_POINT_LOOP_END; The strain vector in Akantu contains the values of :math:`\nabla \vec{u}`, i.e. it is really the *displacement gradient*, - ``computeTangentStiffness``: This method is called when the tangent to the stress-strain curve is desired (see Fig \ref {fig:smm:AL:K}). For example, it is called in the implicit solver when the stiffness matrix for the regular elements is assembled based on the following formula: .. math:: \label{eqn:smm:constitutive_elasc} \mat{K } =\int{\mat{B^T}\mat{D(\varepsilon)}\mat{B}} Therefore, in this method, the ``tangent`` matrix (\mat{D}) is computed for a given strain. The ``tangent`` matrix is a :math:`4^{th}` order tensor which is stored as a matrix in Voigt notation. .. _fig:smm:AL:K: .. figure:: figures/tangent.svg :align: center :width: 60% Tangent to the stress-strain curve. .. \begin{figure}[!htb] \begin{center} \includegraphics[width=0.4\textwidth,keepaspectratio=true]{figures/tangent.pdf} \caption{Tangent to the stress-strain curve.} \label{fig:smm:AL:K} \end{center} \end{figure} - ``getCelerity``: The stability criterion of the explicit integration scheme depend on the fastest wave celerity~\eqref{eqn:smm:explicit:stabletime}. This celerity depend on the material, and therefore the value of this velocity should be defined in this method for each new material. By default, the fastest wave speed is the compressive wave whose celerity can be defined in ``getPushWaveSpeed``. Once the declaration and implementation of the new material has been completed, this material can be used in the user's example by including the header file:: #include "material_XXX.hh" For existing materials, as mentioned in Section~\ref{sect:smm:CL}, by default, the materials are initialized inside the method ``initFull``. If a local material should be used instead, the initialization of the material has to be postponed until the local material is registered in the model. Therefore, the model is initialized with the boolean for skipping the material initialization equal to true:: /// model initialization model.initFull(_analysis_method = _explicit_lumped_mass); Once the model has been initialized, the local material needs to be registered in the model:: model.registerNewCustomMaterials("name_of_local_material"); Only at this point the material can be initialized:: model.initMaterials(); A full example for adding a new damage law can be found in ``examples/new_material``. Adding a New Non-Local Constitutive Law ``````````````````````````````````````` In order to add a new non-local material we first have to add the local constitutive law in Akantu (see above). We can then add the non-local version of the constitutive law by adding the two files (``material_XXX_non_local.hh`` and ``material_XXX_non_local.cc``) where ``XXX`` is the name of the corresponding local material. The new law must inherit from the two classes, non-local parent class, such as the ``MaterialNonLocal`` class, and from the local version of the constitutive law, *i.e.* ``MaterialXXX``. It is therefore necessary to include the interface of those classes in the header file of your custom material and indicate the inheritance in the declaration of the class:: /* ---------------------------------------------------------------------- */ #include "material_non_local.hh" // the non-local parent #include "material_XXX.hh" /* ---------------------------------------------------------------------- */ #ifndef __AKANTU_MATERIAL_XXX_HH__ #define __AKANTU_MATERIAL_XXX_HH__ namespace akantu { class MaterialXXXNonLocal : public MaterialXXX, public MaterialNonLocal { /// declare here the interface of your material }; As members of the class we only need to add the internal fields to store the non-local quantities, which are obtained from the averaging process:: /* -------------------------------------------------------------------------- */ /* Class members */ /* -------------------------------------------------------------------------- */ protected: InternalField grad_u_nl; The following four functions need to be implemented in the non-local material:: /// initialization of the material void initMaterial(); /// loop over all element and invoke stress computation virtual void computeNonLocalStresses(GhostType ghost_type); /// compute stresses after local quantities have been averaged virtual void computeNonLocalStress(ElementType el_type, GhostType ghost_type) /// compute all local quantities void computeStress(ElementType el_type, GhostType ghost_type); In the intialization of the non-local material we need to register the local quantity for the averaging process. In our example the internal field *grad_u_nl* is the non-local counterpart of the gradient of the displacement field (*grad_u_nl*):: void MaterialXXXNonLocal::initMaterial() { MaterialXXX::initMaterial(); MaterialNonLocal::initMaterial(); /// register the non-local variable in the manager this->model->getNonLocalManager().registerNonLocalVariable( this->grad_u.getName(), this->grad_u_nl.getName(), spatial_dimension * spatial_dimension); } The function to register the non-local variable takes as parameters the name of the local internal field, the name of the non-local counterpart and the number of components of the field we want to average. In the *computeStress* we now need to compute all the quantities we want to average. We can then write a loop for the stress computation in the function *computeNonLocalStresses* and then provide the constitutive law on each integration point in the function *computeNonLocalStress*. diff --git a/doc/dev-doc/manual/solidmechanicsmodel.rst b/doc/dev-doc/manual/solidmechanicsmodel.rst index 12ae4a720..374e930f3 100644 --- a/doc/dev-doc/manual/solidmechanicsmodel.rst +++ b/doc/dev-doc/manual/solidmechanicsmodel.rst @@ -1,880 +1,880 @@ .. _sect-smm: Solid Mechanics Model ===================== The solid mechanics model is a specific implementation of the :cpp:class:`Model ` interface dedicated to handle the equations of motion or equations of equilibrium. The model is created for a given mesh. It will create its own :cpp:class:`FEEngine ` object to compute the interpolation, gradient, integration and assembly operations. A :cpp:class:`SolidMechanicsModel ` object can simply be created like this:: SolidMechanicsModel model(mesh); where ``mesh`` is the mesh for which the equations are to be solved. A second parameter called ``spatial_dimension`` can be added after ``mesh`` if the spatial dimension of the problem is different than that of the mesh. This model contains at least the following six ``Arrays``: :cpp:func:`blocked_dofs `: contains a Boolean value for each degree of freedom specifying whether that degree is blocked or not. A Dirichlet boundary condition can be prescribed by setting the **blocked_dofs** value of a degree of freedom to ``true``. A Neumann boundary condition can be applied by setting the **blocked_dofs** value of a degree of freedom to ``false``. The **displacement**, **velocity** and **acceleration** are computed for all degrees of freedom for which the **blocked_dofs** value is set to ``false``. For the remaining degrees of freedom, the imposed values (zero by default after initialization) are kept. :cpp:func:`displacement `: contains the displacements of all degrees of freedom. It can be either a computed displacement for free degrees of freedom or an imposed displacement in case of blocked ones (:math:`\vec{u}` in the following). :cpp:func:`velocity `: contains the velocities of all degrees of freedom. As **displacement**, it contains computed or imposed velocities depending on the nature of the degrees of freedom (:math:`\dot{\vec{u}}` in the following). :cpp:func:`acceleration `: contains the accelerations of all degrees of freedom. As **displacement**, it contains computed or imposed accelerations depending on the nature of the degrees of freedom (:math:`\ddot{\vec{u}}` in the following). :cpp:func:`external_force `: contains the external forces applied on the nodes (:math:`\vec{f}_{\st{ext}}` in the following). :cpp:func:`internal_force `: contains the internal forces on the nodes (:math:`\vec{f}_{\mathrm{int}}` in the following). Some examples to help to understand how to use this model will be presented in the next sections. Model Setup ----------- Setting Initial Conditions `````````````````````````` For a unique solution of the equations of motion, initial displacements and velocities for all degrees of freedom must be specified: .. math:: \vec{u}(t=0) & = \vec{u}_0\\ \dot{\vec u}(t=0) & = \vec{v}_0 The solid mechanics model can be initialized as follows:: model.initFull() This function initializes the internal arrays and sets them to zero. Initial displacements and velocities that are not equal to zero can be prescribed by running a loop over the total number of nodes. Here, the initial displacement in :math:`x`-direction and the initial velocity in :math:`y`-direction for all nodes is set to :math:`0.1` and :math:`1`, respectively:: auto & disp = model.getDisplacement(); auto & velo = model.getVelocity(); for (UInt node = 0; node < mesh.getNbNodes(); ++node) { disp(node, 0) = 0.1; velo(node, 1) = 1.; } .. _sect-smm-boundary: Setting Boundary Conditions ``````````````````````````` This section explains how to impose Dirichlet or Neumann boundary conditions. A Dirichlet boundary condition specifies the values that the displacement needs to take for every point :math:`x` at the boundary (:math:`\Gamma_u`) of the problem domain (:numref:`fig-smm-boundaries`): .. math:: \vec{u} = \bar{\vec u} \quad \forall \vec{x}\in \Gamma_{u} A Neumann boundary condition imposes the value of the gradient of the solution at the boundary :math:`\Gamma_t` of the problem domain (:numref:`fig-smm-boundaries`): .. math:: \vec{t} = \mat{\sigma} \vec{n} = \bar{\vec t} \quad \forall \vec{x}\in \Gamma_{t} .. _fig-smm-boundaries: .. figure:: figures/problem_domain.svg :align: center Problem domain :math:`\Omega` with boundary in three dimensions. The Dirchelet and the Neumann regions of the boundary are denoted with :math:`\Gamma_u` and :math:`\Gamma_t`, respecitvely. Different ways of imposing these boundary conditions exist. A basic way is to loop over nodes or elements at the boundary and apply local values. A more advanced method consists of using the notion of the boundary of the mesh. In the following both ways are presented. Starting with the basic approach, as mentioned, the Dirichlet boundary conditions can be applied by looping over the nodes and assigning the required values. :numref:`fig-smm-dirichlet_bc` shows a beam with a fixed support on the left side. On the right end of the beam, a load is applied. At the fixed support, the displacement has a given value. For this example, the displacements in both the :math:`x` and the :math:`y`-direction are set to zero. Implementing this displacement boundary condition is similar to the implementation of initial displacement conditions described above. However, in order to impose a displacement boundary condition for all time steps, the corresponding nodes need to be marked as boundary nodes using the function ``blocked``. While, in order to impose a load on the right side, the nodes are not marked. The detail codes are shown as follows .. code-block:: c++ auto & blocked = model.getBlockedDOFs(); const auto & pos = mesh.getNodes(); UInt nb_nodes = mesh.getNbNodes(); for (UInt node = 0; node < nb_nodes; ++node) { if(Math::are_float_equal(pos(node, _x), 0)) { blocked(node, _x) = true; // block dof in x-direction blocked(node, _y) = true; // block dof in y-direction disp(node, _x) = 0.; // fixed displacement in x-direction disp(node, _y) = 0.; // fixed displacement in y-direction } else if (Math::are_float_equal(pos(node, _y), 0)) { blocked(node, _x) = false; // unblock dof in x-direction forces(node, _x) = 10.; // force in x-direction } } .. _fig-smm-dirichlet_bc: .. figure:: figures/dirichlet.svg :align: center Beam with fixed support and load. For the more advanced approach, one needs the notion of a boundary in the mesh. Therefore, the boundary should be created before boundary condition functors can be applied. Generally the boundary can be specified from the mesh file or the geometry. For the first case, the function ``createGroupsFromMeshData`` is called. This function can read any types of mesh data which are provided in the mesh file. If the mesh file is created with Gmsh, the function takes one input strings which is either ``tag_0``, ``tag_1`` or ``physical_names``. The first two tags are assigned by Gmsh to each element which shows the physical group that they belong to. In Gmsh, it is also possible to consider strings for different groups of elements. These elements can be separated by giving a string ``physical_names`` to the function ``createGroupsFromMeshData`` .. code-block:: c++ mesh.createGroupsFromMeshData("physical_names"). Boundary conditions support can also be created from the geometry by calling ``createBoundaryGroupFromGeometry``. This function gathers all the elements on the boundary of the geometry. To apply the required boundary conditions, the function :cpp:func:`applyBC ` needs to be called on a :cpp:class:`SolidMechanicsModel `. This function gets a Dirichlet or Neumann functor and a string which specifies the desired boundary on which the boundary conditions is to be applied. The functors specify the type of conditions to apply. Three built-in functors for Dirichlet exist: :cpp:class:`FlagOnly `, :cpp:class:`FixedValue ` and :cpp:class:`IncrementValue `. The functor ``FlagOnly`` is used if a point is fixed in a given direction. Therefore, the input parameter to this functor is only the fixed direction. The ``FixedValue`` functor is used when a displacement value is applied in a fixed direction. The ``IncrementValue`` applies an increment to the displacement in a given direction. The following code shows the utilization of three functors for the top, bottom and side surface of the mesh which were already defined in the Gmsh .. code-block:: c++ model.applyBC(BC::Dirichlet::FixedValue(13.0, _y), "Top"); model.applyBC(BC::Dirichlet::FlagOnly(_x), "Bottom"); model.applyBC(BC::Dirichlet::IncrementValue(13.0, _x), "Side"); To apply a Neumann boundary condition, the applied traction or stress should be specified before. In case of specifying the traction on the surface, the functor :cpp:class:`FromTraction ` of Neumann boundary conditions is called. Otherwise, the functor :cpp:class:`FromStress ` should be called which gets the stress tensor as an input parameter .. code-block:: c++ Vector surface_traction{0., 0., 1.}; auto surface_stress(3, 3) = Matrix::eye(3); model.applyBC(BC::Neumann::FromTraction(surface_traction), "Bottom"); model.applyBC(BC::Neumann::FromStress(surface_stress), "Top"); If the boundary conditions need to be removed during the simulation, a functor is called from the Neumann boundary condition to free those boundary conditions from the desired boundary .. code-block:: c++ model.applyBC(BC::Neumann::FreeBoundary(), "Side"); User specified functors can also be implemented. A full example for setting both initial and boundary conditions can be found in ``examples/boundary_conditions.cc``. The problem solved in this example is shown in :numref:`fig-smm-bc_and_ic`. It consists of a plate that is fixed with movable supports on the left and bottom side. On the right side, a traction, which increases linearly with the number of time steps, is applied. The initial displacement and velocity in :math:`x`-direction at all free nodes is zero and two respectively. .. _fig-smm-bc_and_ic: .. figure:: figures/bc_and_ic_example.svg :align: center :width: 75% Plate on movable supports. .. \begin{figure}[!htb] \centering \includegraphics[scale=0.8]{figures/bc_and_ic_example} \caption{Plate on movable supports.\label{fig-smm-bc_and_ic}} \end{figure} As it is mentioned in Section \ref{sect:common:groups}, node and element groups can be used to assign the boundary conditions. A generic example is given below with a Dirichlet boundary condition:: // create a node group NodeGroup & node_group = mesh.createNodeGroup("nodes_fix"); /* fill the node group with the nodes you want */ // create an element group using the existing node group mesh.createElementGroupFromNodeGroup("el_fix", "nodes_fix", spatial_dimension-1); // boundary condition can be applied using the element group name model.applyBC(BC::Dirichlet::FixedValue(0.0, _x), "el_fix"); Material Selector ````````````````` If the user wants to assign different materials to different finite elements groups in ``Akantu``, a material selector has to be used. By default, ``Akantu`` assigns the first valid material in the material file to all elements present in the model (regular continuum materials are assigned to the regular elements and cohesive materials are assigned to cohesive elements or element facets). To assign different materials to specific elements, mesh data information such as tag information or specified physical names can be used. :cpp:class:`MeshDataMaterialSelector ` class uses this information to assign different materials. With the proper physical name or tag name and index, different materials can be assigned as demonstrated in the examples below:: auto mat_selector = std::make_shared>("physical_names", model); model.setMaterialSelector(mat_selector); In this example the physical names specified in a GMSH geometry file will by used to match the material names in the input file. Another example would be to use the first (``tag_0``) or the second (``tag_1``) tag associated to each elements in the mesh:: auto mat_selector = std::make_shared>( "tag_1", model, first_index); model.setMaterialSelector(*mat_selector); where ``first_index`` (default is 1) is the value of ``tag_1`` that will be associated to the first material in the material input file. The following values of the tag will be associated with the following materials. There are four different material selectors pre-defined in ``Akantu``. :cpp:class:`MaterialSelector ` and :cpp:class:`DefaultMaterialSelector ` is used to assign a material to regular elements by default. For the regular elements, as in the example above, :cpp:class:`MeshDataMaterialSelector ` can be used to assign different materials to different elements. Apart from the ``Akantu``'s default material selectors, users can always develop their own classes in the main code to tackle various multi-material assignment situations. For cohesive material, ``Akantu`` has a pre-defined material selector to assign the first cohesive material by default to the cohesive elements which is called :cpp:class:`DefaultMaterialCohesiveSelector ` and it inherits its properties from :cpp:class:`DefaultMaterialSelector `. Multiple cohesive materials can be assigned using mesh data information (for more details, see :ref:`sect-smm-intrinsic-insertion`). Insertion of Cohesive Elements `````````````````````````````` Cohesive elements are currently compatible only with static simulation and dynamic simulation with an explicit time integration scheme (see section :ref:`ssect-smm-expl-time-integration`). They do not have to be inserted when the mesh is generated (intrinsic) but can be added during the simulation (extrinsic). At any time during the simulation, it is possible to access the following energies with the relative function: .. code-block:: c++ Real Ed = model.getEnergy("dissipated"); Real Er = model.getEnergy("reversible"); Real Ec = model.getEnergy("contact"); A new model have to be call in a very similar way that the solid mechanics model: .. code-block:: c++ SolidMechanicsModelCohesive model(mesh); model.initFull(_analysis_method = _explicit_lumped_mass, _is_extrinsic = true); Cohesive element insertion can be either realized at the beginning of the simulation or it can be carried out dynamically during the simulation. The first approach is called ``intrinsic``, the second one ``extrinsic``. When an element is present from the beginning, a bi-linear or exponential cohesive law should be used instead of a linear one. A bi-linear law works exactly like a linear one except for an additional parameter :math:`\delta_0` separating an initial linear elastic part from the linear irreversible one. For additional details concerning cohesive laws see Section~\ref{sec:cohesive-laws}. .. _fig-smm-coh-insertion: .. figure:: figures/insertion.svg :align: center Insertion of a cohesive element. Extrinsic cohesive elements are dynamically inserted between two standard elements when .. math:: \sigma_\mathrm{eff} > \sigma_\mathrm{c} \quad\text {with} \quad \sigma_\mathrm{eff} = \sqrt{\sigma_\mathrm{n} ^ 2 + \frac{\tau ^ 2} {\beta ^ 2 }} in which :math:`\sigma_\mathrm { n } ` is the tensile normal traction and $\tau$ the resulting tangential one( :numref:`fig-smm-coh-insertion`). Extrinsic approach '''''''''''''''''' During the simulation, stress has to be checked along each facet in order to insert cohesive elements where the stress criterion is reached. This check is performed by calling the method :cpp:func:`checkCohesiveStress `, as example before each step resolution: .. code-block:: c++ model.checkCohesiveStress(); model.solveStep(); The area where stresses are checked and cohesive elements inserted can be limited using the method :cpp:func:`setLimit ` on the :cpp:class:`CohesiveInserter ` during initialization. As example, to limit insertion in the range :math:`[-1.5, 1.5]` in the $x$ direction: .. code-block:: c++ auto & inserter = model.getElementInserter(); inserter.setLimit(_x, -1.5, 1.5); Additional restrictions with respect to :math:`_y` and :math:`_z` directions can be added as well. .. _sect-smm-intrinsic-insertion: Intrinsic approach '''''''''''''''''' Intrinsic cohesive elements are inserted in the mesh with the method :cpp:func:`initFull `. Similarly, the range of insertion can be limited with :cpp:func:`setLimit ` before the :cpp:func:`initFull ` call. In both cases extrinsic and intrinsic the insertion can be restricted to surfaces or element groups. To do so the list of groups should be specified in the input file. .. code-block:: model solid_mechanics_model_cohesive [ cohesive_inserter [ cohesive_surfaces = [surface1, surface2, ...] cohesive_zones = [group1, group2, ...] ] material cohesive_linear [ name = insertion beta = 1 G_c = 10 sigma_c = 1e6 ] ] Static Analysis --------------- The :cpp:class:`SolidMechanicsModel ` class can handle different analysis methods, the first one being presented is the static case. In this case, the equation to solve is .. math:: \mat{K} \vec{u} = \vec{f}_{\mathrm{ext}} :label: eqn-smm-static where :math:`\mat{K}` is the global stiffness matrix, :math:`\vec{u}` the displacement vector and :math:`\vec{f}_{\st{ext}}` the vector of external forces applied to the system. To solve such a problem, the static solver of the :cpp:class:`SolidMechanicsModel ` object is used. First, a model has to be created and initialized. To create the model, a mesh (which can be read from a file) is needed, as explained in Section~\ref{sect:common:mesh}. Once an instance of a :cpp:class:`SolidMechanicsModel ` is obtained, the easiest way to initialize it is to use the :cpp:func:`initFull ` method by giving the :cpp:class:`SolidMechanicsModelOptions `. These options specify the type of analysis to be performed and whether the materials should be initialized with :cpp:func:`initMaterials ` or not .. code-block:: c++ SolidMechanicsModel model(mesh); model.initFull(_analysis_method = _static); Here, a static analysis is chosen by passing the argument :cpp:enumerator:`_static ` to the method. By default, the Boolean for no initialization of the materials is set to false, so that they are initialized during the ``initFull``. The method ``initFull`` also initializes all appropriate vectors to zero. Once the model is created and initialized, the boundary conditions can be set as explained in Section :ref:`sect-smm-boundary`. Boundary conditions will prescribe the external forces for some free degrees of freedom :math:`\vec{f}_{\st{ext}}` and displacements for some others. At this point of the analysis, the function :cpp:func:`solveStep ` can be called .. code-block:: c++ auto & solver = model.getNonLinearSolver(); solver.set("max_iterations", 1); solver.set("threshold", 1e-4); solver.set("convergence_type", SolveConvergenceCriteria::_residual); model.solveStep(); This function is templated by the solving method and the convergence criterion and takes two arguments: the tolerance and the maximum number of iterations (100 by default), which are :math:`10^{-4}` and :math:`1` for this example. The modified Newton-Raphson method is chosen to solve the system. In this method, the equilibrium equation (:eq:`eqn-smm-static`) is modified in order to apply a Newton-Raphson convergence algorithm: .. math:: \mat{K}^{i+1}\delta\vec{u}^{i+1} &= \vec{r} \\ &= \vec{f}_{\st{ext}} -\vec{f}_{\st{int}}\\ &= \vec{f}_{\st{ext}} - \mat{K}^{i} \vec{u}^{i}\\ \vec{u}^{i+1} &= \vec{u}^{i} + \delta\vec{u}^{i+1}~, where :math:`\delta\vec{u}` is the increment of displacement to be added from one iteration to the other, and :math:`i` is the Newton-Raphson iteration counter. By invoking the ``solveStep`` method in the first step, the global stiffness matrix :math:`\mat{K}` from (:eq:`eqn-smm-static`) is automatically assembled. A Newton-Raphson iteration is subsequently started, :math:`\mat{K}` is updated according to the displacement computed at the previous iteration and one loops until the forces are balanced (:cpp:enumerator:`SolveConvergenceCriteria::_residual `), i.e. :math:`||\vec{r}|| <` :cpp:member:`threshold `. One can also iterate until the increment of displacement is zero (:cpp:enumerator:`SolveConvergenceCriteria::_solution `) which also means that the equilibrium is found. For a linear elastic problem, the solution is obtained in one iteration and therefore the maximum number of iterations can be set to one. But for a non-linear case, one needs to iterate as long as the norm of the residual exceeds the tolerance threshold and therefore the maximum number of iterations has to be higher, e.g. :math:`100` .. code-block:: c++ solver.set("max_iterations", 100); model.solveStep(); At the end of the analysis, the final solution is stored in the **displacement** vector. A full example of how to solve a static problem is presented in the code ``examples/static/static.cc``. This example is composed of a 2D plate of steel, blocked with rollers on the left and bottom sides as shown in :numref:`fig-smm-static`. The nodes from the right side of the sample are displaced by :math:`0.01\%` of the length of the plate. .. _fig-smm-static: .. figure:: figures/static.svg :align: center :width: 75% Numerical setup. The results of this analysis is depicted in :numref:`fig-smm-implicit:static_solution`. .. _fig-smm-implicit:static_solution: .. figure:: figures/static_analysis.png :align: center :width: 75% Solution of the static analysis. Left: the initial condition, right: the solution (deformation magnified 50 times). Dynamic Methods --------------- Different ways to solve the equations of motion are implemented in the solid mechanics model. The complete equations that should be solved are: .. math:: \mat{M}\ddot{\vec{u}} + \mat{C}\dot{\vec{u}} + \mat{K}\vec{u} = \vec{f}_{\mathrm{ext}} :label: eqn-equation-motion where :math:`\mat{M}`, :math:`\mat{C}` and :math:`\mat{K}` are the mass, damping and stiffness matrices, respectively. In the previous section, it has already been discussed how to solve this equation in the static case, where :math:`\ddot{\vec{u}} = \dot{\vec{u}} = 0`. Here the method to solve this equation in the general case will be presented. For this purpose, a time discretization has to be specified. The most common discretization method in solid mechanics is the Newmark-:math:`\beta` method, which is also the default in ``Akantu``. For the Newmark-:math:`\beta` method, (:eq:`eqn-equation-motion`) becomes a system of three equations (see :cite:`curnier92a,hughes-83a` for more details): .. math:: \begin{eqnarray} \mat{M} \ddot{\vec{u}}_{n+1} + \mat{C}\dot{\vec{u}}_{n+1} + \mat{K} \vec{u}_{n+1} &={\vec{f}_{\st{ext}}}_{\, n+1}\\ \vec{u}_{n+1} &= \vec{u}_{n} + \left(1 - \alpha\right) \Delta t \dot{\vec{u}}_{n} + \alpha \Delta t \dot{\vec{u}}_{n+1} + \left(\frac{1}{2} - \alpha\right) \Delta t^2 \ddot{\vec{u}}_{n}\\ \dot{\vec{u}}_{n+1} &= \dot{\vec{u}}_{n} + \left(1 - \beta\right) \Delta t \ddot{\vec{u}}_{n} + \beta \Delta t \ddot{\vec{u}}_{n+1} \end{eqnarray} :label: eqn-equation-motion-discret In these new equations, :math:`\ddot{\vec{u}}_{n}`, :math:`\dot{\vec{u}}_{n}` and :math:`\vec{u}_{n}` are the approximations of :math:`\ddot{\vec{u}}(t_n)`, :math:`\dot{\vec{u}}(t_n)` and :math:`\vec{u}(t_n)`. Equation~(:eq:`eqn-equation-motion-discret`) is the equation of motion discretized in space (finite-element discretization), and the equations above are discretized in both space and time (Newmark discretization). The :math:`\alpha` and :math:`\beta` parameters determine the stability and the accuracy of the algorithm. Classical values for :math:`\alpha` and :math:`\beta` are usually :math:`\beta = 1/2` for no numerical damping and :math:`0 < \alpha < 1/2`. .. csv-table:: :header: ":math:`\\alpha`", "Method (:math:`\\beta = 1/2`)", "Type" ":math:`0`", "central difference", "explicit" ":math:`\frac{1}{6}`", "Fox-Goodwin(royal road)", "implicit" ":math:`\frac{1}{3}`", "Linear acceleration", "implicit" ":math:`\frac{1}{2}`", "Average acceleration (trapeziodal rule)", "implicit" The solution of this system of equations, (:eq:`eqn-equation-motion-discret`) is split into a predictor and a corrector system of equations. Moreover, in the case of a non-linear equations, an iterative algorithm such as the Newton-Raphson method is applied. The system of equations can be written as: - *Predictor*: .. math:: \vec{u}_{n+1}^{0} &= \vec{u}_{n} + \Delta t \dot{\vec{u}}_{n} + \frac{\Delta t^2}{2} \ddot{\vec{u}}_{n} \\ \dot{\vec{u}}_{n+1}^{0} &= \dot{\vec{u}}_{n} + \Delta t \ddot{\vec{u}}_{n} \\ \ddot{\vec{u}}_{n+1}^{0} &= \ddot{\vec{u}}_{n} - *Solve*: .. math:: \left(c \mat{M} + d \mat{C} + e \mat{K}_{n+1}^i\right) \vec{w} &= {\vec{f}_{\st{ext}}}_{\,n+1} - {\vec{f}_{\st{int}}}_{\,n+1}^i - \mat{C} \dot{\vec{u}}_{n+1}^i - \mat{M} \ddot{\vec{u}}_{n+1}^i\\ &= \vec{r}_{n+1}^i - *Corrector*: .. math:: \ddot{\vec{u}}_{n+1}^{i+1} &= \ddot{\vec{u}}_{n+1}^{i} +c \vec{w} \\ \dot{\vec{u}}_{n+1}^{i+1} &= \dot{\vec{u}}_{n+1}^{i} + d\vec{w} \\ \vec{u}_{n+1}^{i+1} &= \vec{u}_{n+1}^{i} + e \vec{w} where :math:`i` is the Newton-Raphson iteration counter and :math:`c`, :math:`d` and :math:`e` are parameters depending on the method used to solve the equations .. csv-table:: :header: "", ":math:`\\vec{w}`", ":math:`e`", ":math:`d`", ":math:`c`" "in acceleration", ":math:`\delta\ddot{\vec{u}}`", ":math:`\alpha\beta\Delta t^2`", ":math:`\beta\Delta t`", ":math:`1`" "in velocity", ":math:`\delta\dot{\vec{u}}`", ":math:`\alpha\Delta t`", ":math:`1`", ":math:`\frac{1}{\beta\Delta t}`" "in displacement", ":math:`\delta\vec{u}`", ":math:`1`", ":math:`\frac{1}{\alpha\Delta t}`", ":math:`\frac{1}{\alpha\beta \Delta t^2}`" .. note:: If you want to use the implicit solver ``Akantu`` should be compiled at least with one sparse matrix solver such as `Mumps `_ :cite:`mumps`. Implicit Time Integration ````````````````````````` To solve a problem with an implicit time integration scheme, first a :cpp:class:`SolidMechanicsModel ` object has to be created and initialized. Then the initial and boundary conditions have to be set. Everything is similar to the example in the static case (Section~\ref{sect:smm:static}), however, in this case the implicit dynamic scheme is selected at the initialization of the model:: SolidMechanicsModel model(mesh); model.initFull(_analysis_method = _implicit_dynamic); Because a dynamic simulation is conducted, an integration time step :math:`\Delta t` has to be specified. In the case of implicit simulations, ``Akantu`` implements a trapezoidal rule by default. That is to say :math:`\alpha = 1/2` and :math:`\beta = 1/2` which is unconditionally stable. Therefore the value of the time step can be chosen arbitrarily within reason:: model.setTimeStep(time_step); Since the system has to be solved for a given amount of time steps, the method ``solveStep()``, (which has already been used in the static example in Section~\ref{sect:smm:static}), is called inside a time loop:: /// time loop Real time = 0.; auto & solver = model.getNonLinearSolver(); solver.set("max_iterations", 100); solver.set("threshold", 1e-12); solver.set("convergence_type", SolveConvergenceCriteria::_solution); for (UInt s = 1; time ` class. In the initialization, the explicit scheme is selected using the ``_explicit_lumped_mass`` constant:: SolidMechanicsModel model(mesh); model.initFull(_analysis_method = _explicit_lumped_mass); .. note:: Writing ``model.initFull()`` or ``model.initFull();`` is equivalent to use the ``_explicit_lumped_mass`` keyword, as this is the default case. The explicit time integration scheme implemented in ``Akantu`` uses a lumped mass matrix :math:`\mat{M}` (reducing the computational cost). This matrix is assembled by distributing the mass of each element onto its nodes. The resulting :math:`\mat{M}` is therefore a diagonal matrix stored in the **mass** vector of the model. The explicit integration scheme is conditionally stable. The time step has to be smaller than the stable time step which is obtained in Akantu as follows:: critical_time_step = model.getStableTimeStep(); The stable time step corresponds to the time the fastest wave (the compressive wave) needs to travel the characteristic length of the mesh: .. math:: \Delta t_{\st{crit}} = \frac{\Delta x}{c} where :math:`\Delta x` is a characteristic length (\eg the inradius in the case of linear triangle element) and :math:`c` is the celerity of the fastest wave in the material. It is generally the compressive wave of celerity :math:`c = \sqrt{\frac{2 \mu + \lambda}{\rho}}`, :math:`\mu` and :math:`\lambda` are the first and second Lame's coefficients and :math:`\rho` is the density. However, it is recommended to impose a time step that is smaller than the stable time step, for instance, by multiplying the stable time step by a safety factor smaller than one:: const Real safety_time_factor = 0.8; Real applied_time_step = critical_time_step * safety_time_factor; model.setTimeStep(applied_time_step); The initial displacement and velocity fields are, by default, equal to zero if not given specifically by the user (see \ref{sect:smm:initial_condition}). Like in implicit dynamics, a time loop is used in which the displacement, velocity and acceleration fields are updated at each time step. The values of these fields are obtained from the Newmark:math:`-\beta` equations with :math:`\beta=1/2` and :math:`\alpha=0`. In ``Akantu`` these computations at each time step are invoked by calling the function ``solveStep``:: for (UInt s = 1; (s-1)*applied_time_step < total_time; ++s) { model.solveStep(); } The method ``solveStep`` wraps the four following functions: - ``model.explicitPred()`` allows to compute the displacement field at :math:`t+1` and a part of the velocity field at :math:`t+1`, denoted by :math:`\vec{\dot{u}^{\st{p}}}_{n+1}`, which will be used later in the method ``model.explicitCorr()``. The equations are: .. math:: \vec{u}_{n+1} &= \vec{u}_{n} + \Delta t \vec{\dot{u}}_{n} + \frac{\Delta t^2}{2} \vec{\ddot{u}}_{n}\\ \vec{\dot{u}^{\st{p}}}_{n+1} &= \vec{\dot{u}}_{n} + \Delta t \vec{\ddot{u}}_{n} - ``model.updateResidual()`` and ``model.updateAcceleration()`` compute the acceleration increment :math:`\delta \vec{\ddot{u}}`: .. math:: \left(\mat{M} + \frac{1}{2} \Delta t \mat{C}\right) \delta \vec{\ddot{u}} = \vec{f_{\st{ext}}} - \vec{f}_{\st{int}\, n+1} - \mat{C} \vec{\dot{u}^{\st{p}}}_{n+1} - \mat{M} \vec{\ddot{u}}_{n} The internal force :math:`\vec{f}_{\st{int}\, n+1}` is computed from the displacement :math:`\vec{u}_{n+1}` based on the constitutive law. - ``model.explicitCorr()`` computes the velocity and acceleration fields at :math:`t+1`: .. math:: \vec{\dot{u}}_{n+1} &= \vec{\dot{u}^{\st{p}}}_{n+1} + \frac{\Delta t}{2} \delta \vec{\ddot{u}} \\ \vec{\ddot{u}}_{n+1} &= \vec{\ddot{u}}_{n} + \delta \vec{\ddot{u}} The use of an explicit time integration scheme is illustrated by the example: ``examples/explicit/explicit_dynamic.cc``. This example models the propagation of a wave in a steel beam. The beam and the applied displacement in the :math:`x` direction are shown in :numref:`fig-smm-explicit`. .. _fig-smm-explicit: .. figure:: figures/explicit.svg :align: center :width: 90% Numerical setup. The length and height of the beam are :math:`L={10}\textrm{m}` and :math:`h = {1}\textrm{m}`, respectively. The material is linear elastic, homogeneous and isotropic (density: :math:`7800\mathrm{kg/m}^3`, Young's modulus: :math:`210\mathrm{GPa}` and Poisson's ratio: :math:`0.3`). The imposed displacement follow a Gaussian function with a maximum amplitude of :math:`A = {0.01}\textrm{m}`. The potential, kinetic and total energies are computed. The safety factor is equal to :math:`0.8`. .. include:: ./constitutive-laws.rst .. include:: ./new-constitutive-laws.rst diff --git a/test/test_model/test_structural_mechanics_model/CMakeLists.txt b/test/test_model/test_structural_mechanics_model/CMakeLists.txt index a7076ad1e..81dff85c7 100644 --- a/test/test_model/test_structural_mechanics_model/CMakeLists.txt +++ b/test/test_model/test_structural_mechanics_model/CMakeLists.txt @@ -1,64 +1,72 @@ #=============================================================================== # @file CMakeLists.txt # # @author Fabian Barras # # @date creation: Fri Sep 03 2010 # @date last modification: Fri Feb 09 2018 # # @brief Structural Mechanics Model Tests # # @section LICENSE # -# Copyright (©) 2010-2018 EPFL (Ecole Polytechnique Fédérale de Lausanne) Laboratory (LSMS - Laboratoire de Simulation en Mécanique des Solides) +# Copyright (©) 2010-2018 EPFL (Ecole Polytechnique Fédérale de Lausanne) +# Laboratory (LSMS - Laboratoire de Simulation en Mécanique des Solides) # -# Akantu is free software: you can redistribute it and/or modify it under the terms of the GNU Lesser General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version. +# Akantu is free software: you can redistribute it and/or modify it under the +# terms of the GNU Lesser General Public License as published by the Free +# Software Foundation, either version 3 of the License, or (at your option) any +# later version. # -# Akantu is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU Lesser General Public License for more details. +# Akantu is distributed in the hope that it will be useful, but WITHOUT ANY +# WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR +# A PARTICULAR PURPOSE. See the GNU Lesser General Public License for more +# details. # -# You should have received a copy of the GNU Lesser General Public License along with Akantu. If not, see . +# You should have received a copy of the GNU Lesser General Public License along +# with Akantu. If not, see . # #=============================================================================== # Adding sources register_gtest_sources( SOURCES test_structural_mechanics_model_bernoulli_beam_2.cc PACKAGE implicit structural_mechanics ) register_gtest_sources( SOURCES test_structural_mechanics_model_bernoulli_beam_3.cc PACKAGE implicit structural_mechanics ) register_gtest_sources( SOURCES test_structural_mechanics_model_discrete_kirchhoff_triangle_18.cc PACKAGE implicit structural_mechanics ) register_gtest_sources( SOURCES test_structural_mechanics_model_bernoulli_beam_dynamics.cc PACKAGE implicit structural_mechanics ) #=============================================================================== # Adding meshes for element types package_get_element_types(structural_mechanics _types) foreach(_type ${_types}) if(EXISTS ${CMAKE_CURRENT_SOURCE_DIR}/${_type}.msh) list(APPEND _meshes ${_type}.msh) #register_fem_test(fe_engine_precomputation ${_type }) else() if(NOT ${_type} STREQUAL _point_1) message("The mesh ${_type}.msh is missing, the fe_engine test cannot be activated without it") endif() endif() endforeach() #=============================================================================== # Registering google test register_gtest_test(test_structural_mechanics FILES_TO_COPY ${_meshes} ) diff --git a/test/test_model/test_structural_mechanics_model/test_structural_mechanics_model_bernoulli_beam_dynamics.cc b/test/test_model/test_structural_mechanics_model/test_structural_mechanics_model_bernoulli_beam_dynamics.cc index c18ecc36a..13ed7b1f3 100644 --- a/test/test_model/test_structural_mechanics_model/test_structural_mechanics_model_bernoulli_beam_dynamics.cc +++ b/test/test_model/test_structural_mechanics_model/test_structural_mechanics_model_bernoulli_beam_dynamics.cc @@ -1,367 +1,367 @@ /** * @file test_structural_mechanics_model_bernoulli_beam_dynamics.cc * * @author Sébastien Hartmann * * @date creation: Mon Jul 07 2014 * @date last modification: Wed Feb 03 2016 * * @brief Test for _bernouilli_beam in dynamic * * * Copyright (©) 2014-2018 EPFL (Ecole Polytechnique Fédérale de Lausanne) * Laboratory (LSMS - Laboratoire de Simulation en Mécanique des Solides) * * Akantu is free software: you can redistribute it and/or modify it under the * terms of the GNU Lesser General Public License as published by the Free * Software Foundation, either version 3 of the License, or (at your option) any * later version. * * Akantu is distributed in the hope that it will be useful, but WITHOUT ANY * WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR * A PARTICULAR PURPOSE. See the GNU Lesser General Public License for more * details. * * You should have received a copy of the GNU Lesser General Public License * along with Akantu. If not, see . * */ /* -------------------------------------------------------------------------- */ #include "test_structural_mechanics_model_fixture.hh" /* -------------------------------------------------------------------------- */ #include "dof_manager.hh" #include "mesh_accessor.hh" #include "non_linear_solver_newton_raphson.hh" #include "structural_mechanics_model.hh" /* -------------------------------------------------------------------------- */ #include #include #include /* -------------------------------------------------------------------------- */ using namespace akantu; /* -------------------------------------------------------------------------- */ static Real analytical_solution(Real time, Real L, Real rho, Real E, __attribute__((unused)) Real A, Real I, Real F) { Real omega = M_PI * M_PI / L / L * sqrt(E * I / rho); Real sum = 0.; UInt i = 5; for (UInt n = 1; n <= i; n += 2) { sum += (1. - cos(n * n * omega * time)) / pow(n, 4); } return 2. * F * pow(L, 3) / pow(M_PI, 4) / E / I * sum; } template class TestStructBernoulliDynamic : public TestStructuralFixture { using parent = TestStructuralFixture; public: const UInt nb_element{40}; const Real L{2}; const Real le{L / nb_element}; const UInt nb_nodes{nb_element + 1}; const Real F{1e4}; StructuralMaterial mat; void readMesh(std::string /*filename*/) override { MeshAccessor mesh_accessor(*this->mesh); auto & nodes = mesh_accessor.getNodes(); nodes.resize(nb_nodes); nodes.set(0.); for (auto && data : enumerate(make_view(nodes, this->spatial_dimension))) { auto & node = std::get<1>(data); UInt i = std::get<0>(data); node[_x] = i * le; } this->mesh->addConnectivityType(parent::type); auto & connectivities = mesh_accessor.getConnectivity(parent::type); connectivities.resize(nb_element); for (auto && data : enumerate(make_view(connectivities, 2))) { UInt i = std::get<0>(data); auto & connectivity = std::get<1>(data); connectivity = {i, i + 1}; } mesh_accessor.makeReady(); } void setNormals() override { if (this->spatial_dimension != 3) { return; } auto & normals = this->mesh->template getData("extra_normal", parent::type); normals.resize(nb_element); for (auto && normal : make_view(normals, this->spatial_dimension)) { normal = {0., 0., 1.}; } } AnalysisMethod getAnalysisMethod() const override { return _implicit_dynamic; } void addMaterials() override { this->mat.E = 1e9; this->mat.rho = 10; this->mat.I = 1; this->mat.Iz = 1; this->mat.Iy = 1; this->mat.A = 1; this->mat.GJ = 1; this->model->addMaterial(mat); } void setDirichletBCs() override { auto & boundary = this->model->getBlockedDOFs(); boundary.set(false); boundary(0, _x) = true; boundary(0, _y) = true; boundary(nb_nodes - 1, _y) = true; if (this->spatial_dimension == 3) { boundary(0, _z) = true; boundary(nb_nodes - 1, _z) = true; } } void setNeumannBCs() override { auto node_to_print = nb_nodes / 2; // Forces auto & forces = this->model->getExternalForce(); forces.zero(); forces(node_to_print, _y) = F; } void assignMaterials() override { this->model->getElementMaterial(parent::type).set(0); } }; using beam_types = gtest_list_t, element_type_t<_bernoulli_beam_3>>>; -TYPED_TEST_SUITE(TestStructBernoulliDynamic, beam_types); +TYPED_TEST_SUITE(TestStructBernoulliDynamic, beam_types, ); template void getElementMassMatrix(const StructuralMaterial & /*material*/, Real /*l*/, Matrix & /*M*/) {} template void getElementStifnessMatrix(const StructuralMaterial & /*material*/, Real /*l*/, Matrix & /*M*/) {} template <> void getElementMassMatrix>( const StructuralMaterial & material, Real l, Matrix & M) { auto A = material.A; auto rho = material.rho; // clang-format off M = rho * A * l / 420. * Matrix({ {140, 0, 0, 70, 0, 0}, { 0, 156, 22*l, 0, 54, -13*l}, { 0, 22*l, 4*l*l, 0, 13*l, -3*l*l}, { 70, 0, 0, 140, 0, 0}, { 0, 54, 13*l, 0, 156, -22*l}, { 0,-13*l, -3*l*l, 0, -22*l, 4*l*l}}); // clang-format on } template <> void getElementStifnessMatrix>( const StructuralMaterial & material, Real l, Matrix & K) { auto E = material.E; auto A = material.A; auto I = material.I; auto l_2 = l * l; auto l_3 = l * l * l; // clang-format off K = Matrix({ { E*A/l, 0, 0, -E*A/l, 0, 0}, { 0, 12*E*I/l_3, 6*E*I/l_2, 0, -12*E*I/l_3, 6*E*I/l_2}, { 0, 6*E*I/l_2, 4*E*I/l, 0, -6*E*I/l_2, 2*E*I/l}, {-E*A/l, 0, 0, E*A/l, 0, 0}, { 0, -12*E*I/l_3, -6*E*I/l_2, 0, 12*E*I/l_3, -6*E*I/l_2}, { 0, 6*E*I/l_2, 2*E*I/l, 0, -6*E*I/l_2, 4*E*I/l}}); // clang-format on } template <> void getElementMassMatrix>( const StructuralMaterial & material, Real l, Matrix & M) { auto A = material.A; auto rho = material.rho; // clang-format off M = rho * A * l / 420. * Matrix({ {140, 0, 0, 0, 0, 0, 70, 0, 0, 0, 0, 0}, { 0, 156, 0, 0, 0, 22*l, 0, 54, 0, 0, 0, -13*l}, { 0, 0, 156, 0, -22*l, 0, 0, 0, 54, 0, 13*l, 0}, { 0, 0, 0, 140, 0, 0, 0, 0, 0, 70, 0, 0}, { 0, 0, -22*l, 0, 4*l*l, 0, 0, 0, -13*l, 0, -3*l*l, 0}, { 0, 22*l, 0, 0, 0, 4*l*l, 0, 13*l, 0, 0, 0, -3*l*l}, { 70, 0, 0, 0, 0, 0, 140, 0, 0, 0, 0, 0}, { 0, 54, 0, 0, 0, 13*l, 0, 156, 0, 0, 0, -22*l}, { 0, 0, 54, 0, -13*l, 0, 0, 0, 156, 0, 22*l, 0}, { 0, 0, 0, 70, 0, 0, 0, 0, 0, 140, 0, 0}, { 0, 0, 13*l, 0, -3*l*l, 0, 0, 0, 22*l, 0, 4*l*l, 0}, { 0, -13*l, 0, 0, 0, -3*l*l, 0, -22*l, 0, 0, 0, 4*l*l}}); // clang-format on } template <> void getElementStifnessMatrix>( const StructuralMaterial & material, Real l, Matrix & K) { auto E = material.E; auto A = material.A; auto Iy = material.Iy; auto Iz = material.Iz; auto GJ = material.GJ; auto a1 = E * A / l; auto b1 = 12 * E * Iz / l / l / l; auto b2 = 6 * E * Iz / l / l; auto b3 = 4 * E * Iz / l; auto b4 = 2 * E * Iz / l; auto c1 = 12 * E * Iy / l / l / l; auto c2 = 6 * E * Iy / l / l; auto c3 = 4 * E * Iy / l; auto c4 = 2 * E * Iy / l; auto d1 = GJ / l; // clang-format off K = Matrix({ { a1, 0, 0, 0, 0, 0, -a1, 0, 0, 0, 0, 0}, { 0, b1, 0, 0, 0, b2, 0, -b1, 0, 0, 0, b2}, { 0, 0, c1, 0, -c2, 0, 0, 0, -c1, 0, -c2, 0}, { 0, 0, 0, d1, 0, 0, 0, 0, 0, -d1, 0, 0}, { 0, 0, -c2, 0, c3, 0, 0, 0, c2, 0, c4, 0}, { 0, b2, 0, 0, 0, b3, 0, -b2, 0, 0, 0, b4}, { -a1, 0, 0, 0, 0, 0, a1, 0, 0, 0, 0, 0}, { 0, -b1, 0, 0, 0, -b2, 0, b1, 0, 0, 0, -b2}, { 0, 0, -c1, 0, c2, 0, 0, 0, c1, 0, c2, 0}, { 0, 0, 0, -d1, 0, 0, 0, 0, 0, d1, 0, 0}, { 0, 0, -c2, 0, c4, 0, 0, 0, c2, 0, c3, 0}, { 0, b2, 0, 0, 0, b4, 0, -b2, 0, 0, 0, b3}}); // clang-format on } TYPED_TEST(TestStructBernoulliDynamic, TestBeamMatrices) { this->model->assembleMatrix("M"); this->model->assembleMatrix("K"); const auto & K = this->model->getDOFManager().getMatrix("K"); const auto & M = this->model->getDOFManager().getMatrix("M"); Matrix Ka(this->nb_nodes * this->ndof, this->nb_nodes * this->ndof, 0.); Matrix Ma(this->nb_nodes * this->ndof, this->nb_nodes * this->ndof, 0.); Matrix Ke(this->ndof * 2, this->ndof * 2); Matrix Me(this->ndof * 2, this->ndof * 2); getElementMassMatrix(this->mat, this->le, Me); getElementStifnessMatrix(this->mat, this->le, Ke); auto assemble = [&](auto && nodes, auto && M, auto && Me) { auto n1 = nodes[0]; auto n2 = nodes[1]; for (auto i : arange(this->ndof)) { for (auto j : arange(this->ndof)) { M(n1 * this->ndof + i, n1 * this->ndof + j) += Me(i, j); M(n2 * this->ndof + i, n2 * this->ndof + j) += Me(this->ndof + i, this->ndof + j); M(n1 * this->ndof + i, n2 * this->ndof + j) += Me(i, this->ndof + j); M(n2 * this->ndof + i, n1 * this->ndof + j) += Me(this->ndof + i, j); } } }; auto && connectivities = this->mesh->getConnectivity(this->type); for (auto && connectivity : make_view(connectivities, 2)) { assemble(connectivity, Ka, Ke); assemble(connectivity, Ma, Me); } auto tol = 1e-13; auto Ka_max = Ka.template norm(); auto Ma_max = Ma.template norm(); for (auto i : arange(Ka.rows())) { for (auto j : arange(Ka.cols())) { EXPECT_NEAR(Ka(i, j), K(i, j), tol * Ka_max); EXPECT_NEAR(Ma(i, j), M(i, j), tol * Ma_max); } } } TYPED_TEST(TestStructBernoulliDynamic, TestBeamOscilation) { Real time_step = 1e-6; this->model->setTimeStep(time_step); auto & solver = this->model->getNonLinearSolver(); solver.set("max_iterations", 100); solver.set("threshold", 1e-8); solver.set("convergence_type", SolveConvergenceCriteria::_solution); auto node_to_print = this->nb_nodes / 2; auto & d = this->model->getDisplacement()(node_to_print, _y); std::ofstream pos; std::string filename = "position" + std::to_string(this->type) + ".csv"; pos.open(filename); if (not pos.good()) { AKANTU_ERROR("Cannot open file \"position.csv\""); } pos << "id,time,position,solution" << std::endl; //#define debug #ifdef debug this->model->addDumpFieldVector("displacement"); this->model->addDumpField("blocked_dofs"); this->model->addDumpFieldVector("external_force"); this->model->addDumpFieldVector("internal_force"); this->model->addDumpFieldVector("acceleration"); this->model->addDumpFieldVector("velocity"); this->model->dump(); #endif this->model->getDisplacement().set(0.); Real tol = 1e-6; Real time = 0.; for (UInt s = 1; s < 300; ++s) { EXPECT_NO_THROW(this->model->solveStep()); time = s * time_step; auto da = analytical_solution(time, this->L, this->mat.rho, this->mat.E, this->mat.A, this->mat.Iy, this->F); pos << s << "," << time << "," << d << "," << da << std::endl; #ifdef debug this->model->dump(); #endif EXPECT_NEAR(d, da, tol); } }