195 lines
10 KiB
Python
Executable File
195 lines
10 KiB
Python
Executable File
#!/usr/bin/env python
|
|
# -*- coding: UTF-8 no BOM -*-
|
|
|
|
import os,sys,math,string
|
|
import numpy as np
|
|
from optparse import OptionParser
|
|
import damask
|
|
|
|
scriptID = string.replace('$Id$','\n','\\n')
|
|
scriptName = os.path.splitext(scriptID.split()[1])[0]
|
|
|
|
#--------------------------------------------------------------------------------------------------
|
|
# MAIN
|
|
#--------------------------------------------------------------------------------------------------
|
|
parser = OptionParser(option_class=damask.extendableOption, usage='%prog options [file[s]]', description = """
|
|
|
|
Generate geometry description and material configuration from EBSD data in given square-gridded 'ang' file.
|
|
Two phases can be discriminated based on threshold value in a given data column.
|
|
|
|
""", version = scriptID)
|
|
|
|
parser.add_option('--column', dest='column', type='int', metavar = 'int',
|
|
help='data column to discriminate between both phases [%default]')
|
|
parser.add_option('-t','--threshold', dest='threshold', type='float', metavar = 'float',
|
|
help='threshold value for phase discrimination [%default]')
|
|
parser.add_option('--homogenization', dest='homogenization', type='int', metavar = 'int',
|
|
help='homogenization index for <microstructure> configuration [%default]')
|
|
parser.add_option('--phase', dest='phase', type='int', nargs = 2, metavar = 'int int',
|
|
help='phase indices for <microstructure> configuration %default')
|
|
parser.add_option('--crystallite', dest='crystallite', type='int', metavar = 'int',
|
|
help='crystallite index for <microstructure> configuration [%default]')
|
|
parser.add_option('-c', '--configuration', dest='config', action='store_true',
|
|
help='output material configuration [%default]')
|
|
parser.add_option('--compress', dest='compress', action='store_true',
|
|
help='lump identical microstructure and texture information [%default]')
|
|
parser.add_option('-a', '--axes', dest='axes', nargs = 3, metavar = 'string string string',
|
|
help='Euler angle coordinate system for <texture> configuration x,y,z = %default')
|
|
parser.add_option('-p', '--precision', dest='precision', choices=['0','1','2','3'], metavar = 'int',
|
|
help = 'euler angles decimal places for output format and compressing (0,1,2,3) [2]')
|
|
|
|
parser.set_defaults(column = 11)
|
|
parser.set_defaults(threshold = 0.5)
|
|
parser.set_defaults(homogenization = 1)
|
|
parser.set_defaults(phase = [1,2])
|
|
parser.set_defaults(crystallite = 1)
|
|
parser.set_defaults(config = False)
|
|
parser.set_defaults(compress = False)
|
|
parser.set_defaults(axes = ['y','x','-z'])
|
|
parser.set_defaults(precision = '2')
|
|
(options,filenames) = parser.parse_args()
|
|
|
|
|
|
for i in options.axes:
|
|
if i.lower() not in ['x','+x','-x','y','+y','-y','z','+z','-z']:
|
|
parser.error('invalid axes %s %s %s' %(options.axes[0],options.axes[1],options.axes[2]))
|
|
|
|
# --- loop over input files -------------------------------------------------------------------------
|
|
|
|
if filenames == []: filenames = [None]
|
|
|
|
for name in filenames:
|
|
try:
|
|
table = damask.ASCIItable(name = name, outname = os.path.splitext(name)[0] +'%s'%('_material.config' if options.config else '.geom'),
|
|
buffered = False, labeled = False, readonly=True)
|
|
except: continue
|
|
table.croak(damask.util.emph(scriptName)+(': '+name if name else ''))
|
|
|
|
info = {
|
|
'grid': np.ones(3,'i'),
|
|
'size': np.zeros(3,'d'),
|
|
'origin': np.zeros(3,'d'),
|
|
'microstructures': 0,
|
|
'homogenization': options.homogenization
|
|
}
|
|
|
|
coords = [{},{},{1:True}]
|
|
pos = {'min':[ float("inf"), float("inf")],
|
|
'max':[-float("inf"),-float("inf")]}
|
|
|
|
phase = []
|
|
eulerangles = []
|
|
|
|
# --------------- read data -----------------------------------------------------------------------
|
|
while table.data_read():
|
|
words = table.data
|
|
if words[0] == '#': # process initial comments/header block
|
|
if len(words) > 2:
|
|
if words[2].lower() == 'hexgrid':
|
|
table.croak('The file has HexGrid format. Please first convert to SquareGrid...\n')
|
|
break
|
|
else:
|
|
currPos = words[3:5]
|
|
for i in xrange(2):
|
|
coords[i][currPos[i]] = True
|
|
currPos = map(float,currPos)
|
|
for i in xrange(2):
|
|
pos['min'][i] = min(pos['min'][i],currPos[i])
|
|
pos['max'][i] = max(pos['max'][i],currPos[i])
|
|
eulerangles.append(map(math.degrees,map(float,words[:3])))
|
|
phase.append(options.phase[int(float(words[options.column-1]) > options.threshold)])
|
|
|
|
# --------------- determine size and grid ---------------------------------------------------------
|
|
info['grid'] = np.array(map(len,coords),'i')
|
|
info['size'][0:2] = info['grid'][0:2]/(info['grid'][0:2]-1.0)* \
|
|
np.array([pos['max'][0]-pos['min'][0],
|
|
pos['max'][1]-pos['min'][1]],'d')
|
|
info['size'][2]=info['size'][0]/info['grid'][0]
|
|
eulerangles = np.array(eulerangles,dtype='f').reshape(info['grid'].prod(),3)
|
|
phase = np.array(phase,dtype='i').reshape(info['grid'].prod())
|
|
|
|
limits = [360,180,360]
|
|
if any([np.any(eulerangles[:,i]>=limits[i]) for i in [0,1,2]]):
|
|
table.croak('Error: euler angles out of bound. Ang file might contain unidexed poins.\n')
|
|
for i,angle in enumerate(['phi1','PHI','phi2']):
|
|
for n in np.nditer(np.where(eulerangles[:,i]>=limits[i]),['zerosize_ok']):
|
|
table.croak('%s in line %i (%4.2f %4.2f %4.2f)\n'
|
|
%(angle,n,eulerangles[n,0],eulerangles[n,1],eulerangles[n,2]))
|
|
continue
|
|
eulerangles=np.around(eulerangles,int(options.precision)) # round to desired precision
|
|
for i,angle in enumerate(['phi1','PHI','phi2']):
|
|
eulerangles[:,i]%=limits[i] # ensure, that rounded euler angles are not out of bounds (modulo by limits)
|
|
|
|
if options.compress:
|
|
formatString='{0:0>'+str(int(options.precision)+3)+'}'
|
|
euleranglesRadInt = (eulerangles*10**int(options.precision)).astype('int') # scale by desired precision and convert to int
|
|
eulerKeys = np.array([int(''.join(map(formatString.format,euleranglesRadInt[i,:]))) \
|
|
for i in xrange(info['grid'].prod())]) # create unique integer key from three euler angles by concatenating the string representation with leading zeros and store as integer
|
|
devNull, texture, eulerKeys_idx = np.unique(eulerKeys, return_index = True, return_inverse=True)# search unique euler angle keys. Texture IDs are the indices of the first occurrence, the inverse is used to construct the microstructure
|
|
msFull = np.array([[eulerKeys_idx[i],phase[i]] for i in xrange(info['grid'].prod())],'i8') # create a microstructure (texture/phase pair) for each point using unique texture IDs. Use longInt (64bit, i8) because the keys might be long
|
|
devNull,msUnique,matPoints = np.unique(msFull.view('c16'),True,True)
|
|
matPoints+=1
|
|
microstructure = np.array([msFull[i] for i in msUnique]) # pick only unique microstructures
|
|
else:
|
|
texture = np.arange(info['grid'].prod())
|
|
microstructure = np.hstack( zip(texture,phase) ).reshape(info['grid'].prod(),2) # create texture/phase pairs
|
|
formatOut = 1+int(math.log10(len(texture)))
|
|
textureOut =['\n\n<texture>']
|
|
|
|
eulerFormatOut='%%%i.%if'%(int(options.precision)+4,int(options.precision))
|
|
outStringAngles='(gauss) phi1 '+eulerFormatOut+' Phi '+eulerFormatOut+' phi2 '+eulerFormatOut+' scatter 0.0 fraction 1.0\n'
|
|
for i in xrange(len(texture)):
|
|
textureOut += ['[Texture%s]\n'%str(i+1).zfill(formatOut) +
|
|
'axes %s %s %s\n'%(options.axes[0],options.axes[1],options.axes[2]) +
|
|
outStringAngles%tuple(eulerangles[texture[i],...])
|
|
]
|
|
formatOut = 1+int(math.log10(len(microstructure)))
|
|
microstructureOut =['<microstructure>']
|
|
for i in xrange(len(microstructure)):
|
|
microstructureOut += ['[Grain%s]\n'%str(i+1).zfill(formatOut) +
|
|
'crystallite\t%i\n'%options.crystallite +
|
|
'(constituent)\tphase %i\ttexture %i\tfraction 1.0\n'%(microstructure[i,1],microstructure[i,0]+1)
|
|
]
|
|
|
|
info['microstructures'] = len(microstructure)
|
|
|
|
#--- report ---------------------------------------------------------------------------------------
|
|
table.croak('grid a b c: %s\n'%(' x '.join(map(str,info['grid']))) +
|
|
'size x y z: %s\n'%(' x '.join(map(str,info['size']))) +
|
|
'origin x y z: %s\n'%(' : '.join(map(str,info['origin']))) +
|
|
'homogenization: %i\n'%info['homogenization'] +
|
|
'microstructures: %i\n\n'%info['microstructures'])
|
|
|
|
if np.any(info['grid'] < 1):
|
|
table.croak('invalid grid a b c.\n')
|
|
continue
|
|
if np.any(info['size'] <= 0.0):
|
|
table.croak('invalid size x y z.\n')
|
|
continue
|
|
|
|
|
|
#--- write data/header --------------------------------------------------------------------------------
|
|
table.info_clear()
|
|
if options.config:
|
|
table.output_write(line for line in microstructureOut+ textureOut)
|
|
else:
|
|
table.info_append([' '.join([scriptID] + sys.argv[1:]),
|
|
"grid\ta %i\tb %i\tc %i"%(info['grid'][0],info['grid'][1],info['grid'][2],),
|
|
"size\tx %f\ty %f\tz %f"%(info['size'][0],info['size'][1],info['size'][2],),
|
|
"origin\tx %f\ty %f\tz %f"%(info['origin'][0],info['origin'][1],info['origin'][2],),
|
|
"microstructures\t%i"%info['microstructures'],
|
|
"homogenization\t%i"%info['homogenization'],
|
|
])
|
|
table.head_write()
|
|
if options.compress:
|
|
matPoints = matPoints.reshape((info['grid'][1],info['grid'][0]))
|
|
table.data = matPoints
|
|
table.data_writeArray('%%%ii'%(1+int(math.log10(np.amax(matPoints)))),delimiter=' ')
|
|
else:
|
|
table.output_write("1 to %i\n"%(info['microstructures']))
|
|
table.output_flush()
|
|
|
|
# --- output finalization --------------------------------------------------------------------------
|
|
|
|
table.close()
|