DAMASK_EICMD/processing/pre/geom_fromVoronoiTessellatio...

273 lines
10 KiB
Python
Executable File

#!/usr/bin/env python3
import os
import sys
import multiprocessing
from io import StringIO
from optparse import OptionParser,OptionGroup
import numpy as np
from scipy import spatial
import damask
scriptName = os.path.splitext(os.path.basename(__file__))[0]
scriptID = ' '.join([scriptName,damask.version])
def Laguerre_tessellation(grid, seeds, grains, size, periodic, weights, cpus):
def findClosestSeed(fargs):
point, seeds, myWeights = fargs
tmp = np.repeat(point.reshape(3,1), len(seeds), axis=1).T
dist = np.sum((tmp - seeds)**2,axis=1) -myWeights
return np.argmin(dist) # seed point closest to point
copies = \
np.array([
[ -1,-1,-1 ],
[ 0,-1,-1 ],
[ 1,-1,-1 ],
[ -1, 0,-1 ],
[ 0, 0,-1 ],
[ 1, 0,-1 ],
[ -1, 1,-1 ],
[ 0, 1,-1 ],
[ 1, 1,-1 ],
[ -1,-1, 0 ],
[ 0,-1, 0 ],
[ 1,-1, 0 ],
[ -1, 0, 0 ],
[ 0, 0, 0 ],
[ 1, 0, 0 ],
[ -1, 1, 0 ],
[ 0, 1, 0 ],
[ 1, 1, 0 ],
[ -1,-1, 1 ],
[ 0,-1, 1 ],
[ 1,-1, 1 ],
[ -1, 0, 1 ],
[ 0, 0, 1 ],
[ 1, 0, 1 ],
[ -1, 1, 1 ],
[ 0, 1, 1 ],
[ 1, 1, 1 ],
],dtype=np.float)*size if periodic else \
np.array([
[ 0, 0, 0 ],
],dtype=np.float)
repeatweights = np.tile(weights,len(copies)).flatten(order='F') # Laguerre weights (1,2,3,1,2,3,...,1,2,3)
for vec in copies: # periodic copies of seed points ...
try: seeds = np.append(seeds, seeds+vec, axis=0) # ... (1+a,2+a,3+a,...,1+z,2+z,3+z)
except NameError: seeds = seeds+vec
damask.util.croak('...using {} cpu{}'.format(options.cpus, 's' if options.cpus > 1 else ''))
arguments = [[arg,seeds,repeatweights] for arg in list(grid)]
if cpus > 1: # use multithreading
pool = multiprocessing.Pool(processes = cpus) # initialize workers
result = pool.map_async(findClosestSeed, arguments) # evaluate function in parallel
pool.close()
pool.join()
closestSeeds = np.array(result.get()).flatten()
else:
closestSeeds = np.zeros(len(arguments),dtype='i')
for i,arg in enumerate(arguments):
closestSeeds[i] = findClosestSeed(arg)
# closestSeed is modulo number of original seed points (i.e. excluding periodic copies)
return grains[closestSeeds%seeds.shape[0]]
def Voronoi_tessellation(grid, seeds, grains, size, periodic = True):
KDTree = spatial.cKDTree(seeds,boxsize=size) if periodic else spatial.cKDTree(seeds)
devNull,closestSeeds = KDTree.query(grid)
return grains[closestSeeds]
# --------------------------------------------------------------------
# MAIN
# --------------------------------------------------------------------
parser = OptionParser(option_class=damask.extendableOption, usage='%prog option(s) [seedfile(s)]', description = """
Generate geometry description and material configuration by tessellation of given seeds file.
""", version = scriptID)
group = OptionGroup(parser, "Tessellation","")
group.add_option('-l',
'--laguerre',
dest = 'laguerre',
action = 'store_true',
help = 'use Laguerre (weighted Voronoi) tessellation')
group.add_option('--cpus',
dest = 'cpus',
type = 'int', metavar = 'int',
help = 'number of parallel processes to use for Laguerre tessellation [%default]')
group.add_option('--nonperiodic',
dest = 'periodic',
action = 'store_false',
help = 'nonperiodic tessellation')
parser.add_option_group(group)
group = OptionGroup(parser, "Geometry","")
group.add_option('-g',
'--grid',
dest = 'grid',
type = 'int', nargs = 3, metavar = ' '.join(['int']*3),
help = 'a,b,c grid of hexahedral box')
group.add_option('-s',
'--size',
dest = 'size',
type = 'float', nargs = 3, metavar=' '.join(['float']*3),
help = 'x,y,z size of hexahedral box')
group.add_option('-o',
'--origin',
dest = 'origin',
type = 'float', nargs = 3, metavar=' '.join(['float']*3),
help = 'origin of grid')
group.add_option('--nonnormalized',
dest = 'normalized',
action = 'store_false',
help = 'seed coordinates are not normalized to a unit cube')
parser.add_option_group(group)
group = OptionGroup(parser, "Seeds","")
group.add_option('-p',
'--pos', '--seedposition',
dest = 'pos',
type = 'string', metavar = 'string',
help = 'label of coordinates [%default]')
group.add_option('-w',
'--weight',
dest = 'weight',
type = 'string', metavar = 'string',
help = 'label of weights [%default]')
group.add_option('-m',
'--microstructure',
dest = 'microstructure',
type = 'string', metavar = 'string',
help = 'label of microstructures [%default]')
group.add_option('-e',
'--eulers',
dest = 'eulers',
type = 'string', metavar = 'string',
help = 'label of Euler angles [%default]')
group.add_option('--axes',
dest = 'axes',
type = 'string', nargs = 3, metavar = ' '.join(['string']*3),
help = 'orientation coordinate frame in terms of position coordinate frame')
parser.add_option_group(group)
group = OptionGroup(parser, "Configuration","")
group.add_option('--without-config',
dest = 'config',
action = 'store_false',
help = 'omit material configuration header')
group.add_option('--homogenization',
dest = 'homogenization',
type = 'int', metavar = 'int',
help = 'homogenization index to be used [%default]')
group.add_option('--phase',
dest = 'phase',
type = 'int', metavar = 'int',
help = 'phase index to be used [%default]')
parser.add_option_group(group)
parser.set_defaults(pos = 'pos',
weight = 'weight',
microstructure = 'microstructure',
eulers = 'euler',
homogenization = 1,
phase = 1,
cpus = 2,
laguerre = False,
periodic = True,
normalized = True,
config = True,
)
(options,filenames) = parser.parse_args()
if filenames == []: filenames = [None]
for name in filenames:
damask.util.report(scriptName,name)
table = damask.Table.from_ASCII(StringIO(''.join(sys.stdin.read())) if name is None else name)
size = np.zeros(3)
origin = np.zeros(3)
for line in table.comments:
items = line.lower().strip().split()
key = items[0] if items else ''
if key == 'grid':
grid = np.array([ int(dict(zip(items[1::2],items[2::2]))[i]) for i in ['a','b','c']])
elif key == 'size':
size = np.array([float(dict(zip(items[1::2],items[2::2]))[i]) for i in ['x','y','z']])
elif key == 'origin':
origin = np.array([float(dict(zip(items[1::2],items[2::2]))[i]) for i in ['x','y','z']])
if options.grid: grid = np.array(options.grid)
if options.size: size = np.array(options.size)
if options.origin: origin = np.array(options.origin)
size = np.where(size <= 0.0,grid/grid.max(),size)
seeds = table.get(options.pos) * size if options.normalized else table.get(options.pos) - origin
if options.eulers in table.labels:
eulers = table.get(options.eulers)
grains = table.get(options.microstructure) if options.microstructure in table.labels else np.arange(len(seeds))+1
grainIDs = np.unique(grains).astype('i')
NgrainIDs = len(grainIDs)
coords = damask.grid_filters.cell_coord0(grid,size).reshape(-1,3)
damask.util.croak('tessellating...')
if options.laguerre:
indices = Laguerre_tessellation(coords,seeds,grains,size,options.periodic,
table.get(options.weight),options.cpus)
else:
indices = Voronoi_tessellation (coords,seeds,grains,size,options.periodic)
config_header = []
if options.config:
if options.eulers in table.labels:
config_header += ['<texture>']
for ID in grainIDs:
eulerID = np.nonzero(grains == ID)[0][0] # find first occurrence of this grain id
config_header += ['[Grain{}]'.format(ID),
'(gauss)\tphi1 {:.2f}\tPhi {:.2f}\tphi2 {:.2f}'.format(*eulers[eulerID])
]
if options.axes: config_header += ['axes\t{} {} {}'.format(*options.axes)]
config_header += ['<microstructure>']
for ID in grainIDs:
config_header += ['[Grain{}]'.format(ID),
'(constituent)\tphase {}\ttexture {}\tfraction 1.0'.format(options.phase,ID)
]
config_header += ['<!skip>']
header = [scriptID + ' ' + ' '.join(sys.argv[1:])]\
+ config_header
geom = damask.Geom(indices.reshape(grid,order='F'),size,origin,
homogenization=options.homogenization,comments=header)
damask.util.croak(geom)
geom.to_file(sys.stdout if name is None else os.path.splitext(name)[0]+'.geom',pack=False)