PyJive workshop: FE^2

PyJive workshop: FE\(^2\)#

import sys
import numpy as np
sys.path.append('../../../pyjive')

import main
from utils import proputils as pu
from names import GlobNames as gn

%matplotlib widget

import contextlib
import os
from urllib.request import urlretrieve

def findfile(fname):
    url = "https://gitlab.tudelft.nl/cm/public/drive/-/raw/main/fe2workshop/" + fname + "?inline=false"
    if not os.path.isfile(fname):
        print(f"Downloading {fname}...")
        urlretrieve(url, fname)

findfile("macro.pro")
findfile("micro.pro")
findfile("dogbone.msh")
findfile("unitCell.msh")

Run multiscale analysis#

Task 1: Run predefined problem

Run the two-scale problem by executing the cell below. The case is that of a dog-bone specimen in uniaxial tension on the macroscale, linked to a unit-cell micromodel with FE\(^2\). Coarse meshes are provided for both scales, and only a small number of time steps is executed to limit the runtime.

globdat = main.jive(pu.parse_file('macro.pro'))

Task 2: Inspect the code

Answer the following questions

  • Which part of the main input file macro.pro defines that a two-scale problem is solved?

  • The macroscale input file is read on the notebook, how is the microscale input file processed?

  • Can you find how the macrostrain is imposed on the micromodel and how the homogenized microstress is returned to the macroscale?

Visualize results#

The resulting displacement field and its evolution can be visualized on the macroscale

viewmodule = globdat[gn.MODULEFACTORY].get_module('View','view')

viewprops = { 'view': {}}
viewprops['view']['plot'] = 'solution[dx]'
viewprops['view']['interactive'] = 'True'
viewprops['view']['constantLevels'] = 'False'
viewprops['view']['constantTicks'] = 'True'
viewprops['view']['step0'] = 3
viewprops['view']['ncolors'] = 10

viewmodule.init(viewprops, globdat)
viewmodule.plot(globdat)

And also on the microscale

microglobdats = globdat[gn.MODEL]._models[0]._mat._globdats
viewmodule.init(viewprops, microglobdats[0])
viewmodule.plot(microglobdats[0])

Task 3: Inspect the results

The cells above generate visualizations from both scales.

  • The microscale visualization is plotting results for one selected micromodel. Can you also plot results for a different one?

  • For the macroscale, the number of time steps agrees with what is specified in the input file. For the microscale, that is not the case. What defines the number of time steps that is plotted for the microscale? (tip: Results are stored in globdat every time the NonlinModule reaches the end of its run function, this is somewhat hidden, but done by the ControlModule which is the base class of the NonlinModule)

Task 4: A missing feature in the present implementation

Although the simulation is defined as a nonlinear one, it remains in the linear regime. In fact, the provided code is not entirely suitable for performing nonlinear simulations. There is an issue with how material history is stored. In the regular nonlinear finite element workflow in pyjive, the action COMMIT takes care of storing history.

  • At what point in the nonlinear solution algorithm is the COMMIT action taken for a single scale simulation?

  • What would need to be done to get consistent behavior for a nonlinear two-scale simulation?

Contents