def resp(ns, strt, name, set_id, bc, newdir):
start = time.time()
C = const()
f = h5py.File("responses.hdf5", 'a')
# nwd = os.getcwd() + '\\' + newdir
nwd = os.getcwd() + '/' + newdir # for unix
os.chdir(nwd)
fname = "%s_%s.txt" % (name, bc)
tmp = np.loadtxt(fname)
dset_name = "yield_%s_%s" % (bc, set_id)
f.create_dataset(dset_name, data=tmp[strt:strt+ns, 2])
dset_name = "stiffness_%s_%s" % (bc, set_id)
f.create_dataset(dset_name, data=tmp[strt:strt+ns, 5])
# return to the original directory
os.chdir('..')
f.close()
end = time.time()
timeE = np.round((end - start), 3)
msg = 'responses read from .txt file for %s_%s: %s seconds' \
% (name, bc, timeE)
rr.WP(msg, C['wrt_file'])
def combine():
C = constants.const()
filename = 'log_combine_coef.txt'
""" Combine the results of the coefficient determination"""
# coeff is the combined vector of coefficients as calculated by the
# orthogonal regression
coef = np.zeros((C['cmax'], 10), dtype='complex128')
c = 0
# for tnum in xrange(596):
for tnum in xrange(C['integrate_njobs']):
fn.WP(str(tnum), filename)
# load partially filled coefficient arrays from each file
f = h5py.File(C['integrate_output'] % str(tnum).zfill(5), 'r')
coef_prt = f.get('coef_prt')[...]
f.close()
clen = coef_prt.shape[0]
coef[c:c + clen, :] = coef_prt
c += clen
# save the coefficients file
f = h5py.File(C['combinecoef_coef'], 'w')
f.create_dataset('coef', data=coef)
f.close()
def read_euler(strt, ns, set_id, newdir, funit):
start = time.time()
C = const()
euler = np.zeros([ns, 3, C['el']**3])
# nwd = os.getcwd() + '\\' + newdir
nwd = os.getcwd() + '/' + newdir # for unix
os.chdir(nwd)
for ii in xrange(ns):
sn = strt + ii + 1
filename = "Ti64_Dream3D_v01_Output_%s.vtk" % sn
euler[ii, :, :] = rr.read_vtk_vector(filename=filename)
if funit == 1:
euler = euler * (np.pi / 180.)
# return to the original directory
os.chdir('..')
f = h5py.File("spatial.hdf5", 'a')
dset_name = 'euler_%s' % set_id
f.create_dataset(dset_name, data=euler)
f.close()
end = time.time()
timeE = np.round((end - start), 3)
msg = 'euler angles read from .vtk file for %s: %s seconds' \
% (set_id, timeE)
rr.WP(msg, C['wrt_file'])
def fegrab(ns, strt, set_id, dir):
# FINITE ELEMENT RESPONSES
st = time.time()
C = const()
r_fem = np.zeros([ns, C['el']**3])
nwd = os.getcwd() + '/' + dir # for unix
# nwd = os.getcwd() + '\\' + direc
os.chdir(nwd)
for ii in xrange(ns):
sn = strt + ii + 1
filename = 'sq21_50test_%s.dat' % sn
r_fem[ii, ...] = res_red(filename)
r_fem = r_fem.reshape(ns, C['el'], C['el'], C['el'])
os.chdir('..')
f = h5py.File('responses.hdf5', 'a')
f.create_dataset('y_sim_%s' % set_id, data=r_fem)
f.close()
msg = 'Load FE results from .dat files for set %s%s: %s seconds' \
% (ns, set_id, np.round((time.time() - st), 3))
rr.WP(msg, C['wrt_file'])
def combine():
C = constants.const()
filename = 'log_Xcalc_combine.txt'
f_master = h5py.File(C['combineXcalc_output'], 'w')
"""load the cosine basis evaluations"""
f_cos = h5py.File(C['Xcalccos_output'], 'r')
for name in f_cos.keys():
fn.WP(name, filename)
tmp = f_cos.get(name)[...]
f_master.create_dataset(name, data=tmp)
del tmp
f_cos.close()
"""load the GSH basis evaluations"""
for jobnum in xrange(C['XcalcGSH_njobs']):
f_gsh = h5py.File(C['XcalcGSH_output'] % str(jobnum).zfill(5), 'r')
for name in f_gsh.keys():
fn.WP(name, filename)
tmp = f_gsh.get(name)[...]
f_master.create_dataset(name, data=tmp)
del tmp
f_gsh.close()
f_master.close()
def regress(ns, set_id):
st = time.time()
C = const()
"""load the feature data"""
f = h5py.File("pre_regress_%s.hdf5" % set_id, 'r')
X = f.get('X')[...]
f.close()
"""load the dependent variable data"""
f = h5py.File("responses.hdf5", 'r')
y = f.get('fip_%s' % set_id)[...]
y = y.reshape((C['n_samp']))
f.close()
# clf = svm.SVR()
# clf = neighbors.KNeighborsRegressor(n_neighbors=1, weights='uniform')
clf = tree.DecisionTreeRegressor(max_depth=10)
clf.fit(X, y)
joblib.dump(clf, 'modelfit.pkl')
timeE = np.round(time.time() - st, 1)
msg = "fit completed: %s s" % timeE
rr.WP(msg, C['wrt_file'])
def calculate():
C = constants.const()
filename = 'Xcalc_log_cos.txt'
""" Load info from collected simulation info file """
f = h5py.File(C['combineread_output'], 'r')
var_set = f.get('var_set')
theta = np.sort(np.unique(var_set[:, 0]))
et_norm = np.sort(np.unique(var_set[:, 4]))
f.close
f = h5py.File(C['basis_eval_cos'], 'a')
"""Evalute the cosine basis functions for theta"""
for q in xrange(C['N_q']):
vec = np.cos(q * np.pi * theta / C['L_th'])
set_id = 'q_%s' % str(q).zfill(5)
f.create_dataset(set_id, data=vec)
fn.WP(set_id, filename)
"""Evalute the cosine basis functions for en"""
for r in xrange(C['N_r']):
vec = np.cos(r * np.pi * (et_norm - C['a']) / C['L_en'])
set_id = 'r_%s' % str(r).zfill(5)
f.create_dataset(set_id, data=vec)
fn.WP(set_id, filename)
f.close()
def read_fip(ns, set_id, newdir):
start = time.time()
C = const()
fip = np.zeros([ns, C['el']**3])
# nwd = os.getcwd() + '\\' + newdir
nwd = os.getcwd() + '/' + newdir # for unix
os.chdir(nwd)
sn = 0
for filename in os.listdir(nwd):
if filename.endswith('%s.vtk' % C['step']):
fip[sn, :] = rr.read_vtk_scalar(filename=filename)
sn += 1
"""return to the original directory"""
os.chdir('..')
f = h5py.File("responses.hdf5", 'a')
f.create_dataset('fip_%s' % set_id, data=(1e9) * fip)
f.close()
end = time.time()
timeE = np.round((end - start), 3)
msg = 'fip values read from .vtk file for %s: %s seconds' % (set_id, timeE)
rr.WP(msg, C['wrt_file'])
def read_euler(ns, set_id, newdir, funit):
start = time.time()
C = const()
euler = np.zeros([ns, 3, C['el']**3])
# nwd = os.getcwd() + '\\' + newdir
nwd = os.getcwd() + '/' + newdir # for unix
os.chdir(nwd)
sn = 0
for filename in os.listdir(nwd):
if filename.endswith('%s.vtk' % C['step']):
euler[sn, :, :] = rr.read_vtk_vector(filename=filename)
sn += 1
if funit == 1:
euler = euler * (np.pi / 180.)
"""return to the original directory"""
os.chdir('..')
f = h5py.File("spatial.hdf5", 'a')
f.create_dataset('euler_%s' % set_id, data=euler)
f.close()
end = time.time()
timeE = np.round((end - start), 3)
msg = 'euler angles read from .vtk file for %s: %s seconds' % (set_id,
timeE)
rr.WP(msg, C['wrt_file'])
def pltcorr(ns, set_id, step, sn, iA, iB):
C = const()
f = h5py.File("spatial.hdf5", 'r')
dset_name = 'euler_%s' % set_id
euler = f.get(dset_name)[sn, 0, :].reshape(C['el'], C['el'], C['el'])
corr = f.get('ff_%s' % set_id)[sn, iA, iB, ...]
f.close()
corr_centered = np.fft.fftshift(corr)
"""Plot slices of the response"""
plt.figure(num=1, figsize=[8, 2.7])
plt.subplot(121)
ax = plt.imshow(euler[0, :, :],
origin='lower',
interpolation='none',
cmap='magma')
plt.colorbar(ax)
plt.title('phi1 field')
plt.subplot(122)
ax = plt.imshow(corr_centered[10, :, :],
origin='lower',
interpolation='none',
cmap='viridis')
plt.colorbar(ax)
plt.title('ff: %s, %s' % (iA, iB))
plt.show()
def calculate():
C = constants.const()
filename = 'Xcalc_log_cos.txt'
""" Load info from collected simulation info file """
f = h5py.File(C['combineread_output'], 'r')
var_set = f.get('var_set')
theta = np.sort(np.unique(var_set[:, 0]))
msg = "theta vec: %s" % str(theta * (180 / np.pi))
fn.WP(msg, filename)
f.close
f = h5py.File(C['basiscos_output'], 'a')
"""Evalute the cosine basis functions for theta"""
for q in xrange(C['N_q']):
vec = np.cos(q * np.pi * theta / C['L_th'])
set_id = 'q_%s' % str(q).zfill(5)
f.create_dataset(set_id, data=vec)
fn.WP(set_id, filename)
f.close()
def combine():
C = constants.const()
f1 = h5py.File(C['combineread_output'], 'w')
alldata = f1.create_dataset("var_set", (C['n_eul'] * C['n_th'], 14))
c = 0
for tt in xrange(C['n_th']):
print "Deformation Mode: %s deg" % str((tt + 0.5) * C['inc'])
# create file for pre-database outputs
f2 = h5py.File(C['read_output'] % str(tt).zfill(5), 'r')
ep_tmp = f2.get("var_set")
stt = (c) * C['n_eul']
print "start index: %s" % stt
end = (c + 1) * C['n_eul']
print "end index: %s" % end
alldata[stt:end, :] = ep_tmp
f2.close()
c += 1
print alldata.shape
f1.close()
def correlate(ns, set_id):
st = time.time()
C = const()
f = h5py.File("spatial.hdf5", 'a')
M = f.get('M_%s' % set_id)[...]
ff = f.create_dataset("ff_%s" % set_id,
(ns, C['H'], C['H'], C['el'], C['el'], C['el']),
dtype='float64')
S = C['el']**3
cmax = C['H'] * C['H']
cmat = np.unravel_index(np.arange(cmax), [C['H'], C['H']])
cmat = np.array(cmat).T
for c in xrange(cmax):
ii, jj = cmat[c, :]
if np.mod(c, 20) == 0:
print str([ii, jj])
M1 = M[:, ii, ...]
mag1 = np.abs(M1)
ang1 = np.arctan2(M1.imag, M1.real)
exp1 = np.exp(-1j * ang1)
term1 = mag1 * exp1
del M1, mag1, ang1, exp1
M2 = M[:, jj, ...]
mag2 = np.abs(M2)
ang2 = np.arctan2(M2.imag, M2.real)
exp2 = np.exp(1j * ang2)
term2 = mag2 * exp2
del M2, mag2, ang2, exp2
FFtmp = term1 * term2 / S
del term1, term2
tmp = np.fft.ifftn(FFtmp, [C['el'], C['el'], C['el']], [1, 2, 3])
ff[:, ii, jj, ...] = tmp.real
if c == 0:
szgb = np.round(C['H'] * C['H'] * FFtmp.nbytes / (1e9), 3)
msg = "ff = %s gb" % szgb
rr.WP(msg, C['wrt_file'])
f.close()
timeE = np.round(time.time() - st, 5)
msg = "correlations computed for %s: %ss" % (set_id, timeE)
rr.WP(msg, C['wrt_file'])
def new_space(ns_set, set_id_set):
st = time.time()
C = const()
print "H: %s" % C['H']
n_corr = C['H']**2
ns_tot = np.sum(ns_set)
allcorr = np.zeros((ns_tot, n_corr * C['el']**3), dtype='float64')
f_stats = h5py.File("spatial.hdf5", 'a')
"""here we will treat the real and imaginary parts of ff as
separate dimensions prior to applying PCA"""
c = 0
for ii in xrange(len(set_id_set)):
tmp = f_stats.get('ff_%s' % set_id_set[ii])[...]
ff = tmp.reshape(ns_set[ii], n_corr * C['el']**3)
allcorr[c:c + ns_set[ii], ...] = ff
c += ns_set[ii]
f_stats.close()
msg = "correlations combined"
rr.WP(msg, C['wrt_file'])
f_master = h5py.File("pca_data.hdf5", 'w')
"""Note that when whiten=True the information about the
relative variances of the pc vectors. This may be desirable
when using regression to find a linkage to reduce some
numerical issues"""
pca = PCA(n_components=C['n_pc_tot'], whiten=True)
pca.fit(allcorr)
ratios = 100 * pca.explained_variance_ratio_
f_master.create_dataset('ratios', data=ratios)
ratios = np.round(ratios, 1)
msg = "pca explained variance: %s%%" % str(ratios)
rr.WP(msg, C['wrt_file'])
f_master.close()
msg = "PCA completed: %ss" % np.round(time.time() - st, 5)
rr.WP(msg, C['wrt_file'])
return pca
def calculate():
C = constants.const()
filename = 'Xcalc_log_cos.txt'
""" Load info from collected simulation info file """
f = h5py.File(C['combineread_output'], 'r')
var_set = f.get('var_set')
theta = var_set[:, 0]
et_norm = var_set[:, 4]
f.close
f = h5py.File(C['Xcalccos_output'], 'a')
"""Evalute the cosine basis functions for theta"""
st = time.time()
for q in xrange(C['N_q']):
vec = np.cos(q*np.pi*theta/C['L_th'])
set_id = 'q_%s' % str(q).zfill(5)
f.create_dataset(set_id, data=vec)
fn.WP(set_id, filename)
msg = "Cosine basis evaluation for theta complete: %ss" \
% np.round(time.time()-st, 3)
fn.WP(msg, filename)
"""Evalute the cosine basis functions for en"""
st = time.time()
for r in xrange(C['N_r']):
vec = np.cos(r*np.pi*(et_norm-C['a'])/C['L_en'])
set_id = 'r_%s' % str(r).zfill(5)
f.create_dataset(set_id, data=vec)
fn.WP(set_id, filename)
msg = "Cosine basis evaluation for en complete: %ss" \
% np.round(time.time()-st, 3)
fn.WP(msg, filename)
f.close()
def read_meas(ns, set_id, comp, tensor_id, newdir):
start = time.time()
C = const()
typ = ['sigma', 'epsilon_t', 'epsilon_p']
# nwd = os.getcwd() + '\\' + newdir
nwd = os.getcwd() + '/' + newdir # for unix
os.chdir(nwd)
compd = {'11': 0, '22': 4, '33': 8, '12': 1, '13': 6, '23': 5}
compp = compd[comp]
r_fem = np.zeros([ns, C['el'], C['el'], C['el']])
sn = 0
for filename in os.listdir(nwd):
if filename.endswith('%s.vtk' % C['step']):
r_temp = rr.read_vtk_tensor(filename=filename,
tensor_id=tensor_id,
comp=compp)
r_fem[sn, ...] = r_temp.reshape([C['el'], C['el'], C['el']])
sn += 1
"""return to the original directory"""
os.chdir('..')
f = h5py.File("responses.hdf5", 'a')
f.create_dataset('%s_%s' % (typ[tensor_id], set_id), data=r_fem)
f.close()
# """FFT OF RESPONSE FIELD"""
# f = h5py.File("D_%s%s_s%s.hdf5" % (ns, set_id, step), 'a')
# tmp = np.fft.fftn(r_fem, axes=[1, 2, 3])
# print tmp.shape
# f.create_dataset('rfft%s_%s' % (comp, typ[tensor_id]), data=tmp)
# f.close()
end = time.time()
timeE = np.round((end - start), 3)
msg = 'The measure of interest has been read from .vtk file' \
' for %s, component %s, type %s: %s seconds' % (set_id, comp,
typ[tensor_id],
timeE)
rr.WP(msg, C['wrt_file'])
def features(ns, set_id):
st = time.time()
C = const()
"""gather the independent variable data"""
f = h5py.File("spatial.hdf5", 'r')
neig = f.get('neig_%s' % set_id)[...]
neig = neig.reshape((C['n_samp'], C['H'], C['cmax']))
f.close()
"""calculate the X matrix"""
X = np.zeros((C['n_samp'], C['xmax']), dtype='float64')
c = 0 # keep track of position in X
"""for 0th order polynomial"""
X[:, 0] = 1
c += 1
"""for 1st order polynomial"""
Imax = C['H']*C['cmax']
Imat = np.unravel_index(np.arange(Imax), (C['H'], C['cmax']))
Imat = np.array(Imat).T
for I in xrange(Imax):
h, pos = Imat[I, :]
X[:, c] = neig[:, h, pos]
c += 1
"""for 2nd order polynomial"""
Imax = C['H']*C['cmax']**2
Imat = np.unravel_index(np.arange(Imax), (C['H'], C['cmax'], C['cmax']))
Imat = np.array(Imat).T
for I in xrange(Imax):
h, pos1, pos2 = Imat[I, :]
X[:, c] = neig[:, h, pos1]*neig[:, h, pos2]
c += 1
f = h5py.File("pre_regress_%s.hdf5" % set_id, 'w')
f.create_dataset('X', data=X)
f.close()
timeE = np.round(time.time()-st, 1)
msg = "features extracted for %s: %s s" % (set_id, timeE)
rr.WP(msg, C['wrt_file'])
def res_red(filename):
"""
Summary:
This function reads the E11 values from a .dat file and reorganizes
the data into a el x el x el array with the correct organization
It will also plot a certain x-slice in the dataset if called within
this script.
Inputs:
filename (string): the name of the '.dat' file containing the
FEM response
el (int): the number of elements per side of the microstructure cube
Outputs:
r_mat ([el,el,el],float): the FEM response of the '.dat' file of
interest
"""
C = const()
f = open(filename, "r")
linelist = f.readlines()
# finds a location several lines above the start of the data
# linelist[n] reads the entire line at location n
for ln in xrange(1000):
if 'THE FOLLOWING TABLE' in linelist[ln]:
break
# line0 is the index of first line of the data
line0 = ln + 5
r_mat = np.zeros([C['el']**3, 8])
c = -1
# this series of loops generates a 9261x8 dataset of E11s
# (element x integration point)
for k in xrange(C['el']**3):
for jj in xrange(8):
c += 1
r_mat[k, jj] = linelist[line0 + c].split()[2]
f.close()
# here we average all 8 integration points in each element cell
r_mat = np.mean(r_mat, axis=1)
return r_mat
def get_M(ns, set_id):
start = time.time()
C = const()
"""get the euler angle files"""
f = h5py.File("spatial.hdf5", 'a')
euler = f.get('euler_%s' % set_id)[...]
mf = np.zeros([ns, C['H'], C['el']**3], dtype='float64')
c = 0
for h in xrange(C['H']):
tmp = gsh.gsh_eval(euler.swapaxes(1, 2), [h])
tmp = np.squeeze(tmp)
mf[:, c, :] = tmp.real
c += 1
# mf[:, h, :] = (2*indxvec[h, 0]+1)*tmp # 2*l+1 included in maple generator
end = time.time()
timeE = np.round((end - start), 3)
msg = "Conversion from Euler angles to GSH coefficients completed:" + \
" %s seconds" % timeE
rr.WP(msg, C['wrt_file'])
mf = mf.reshape([ns, C['H'], C['el'], C['el'], C['el']])
# MICROSTRUCTURE FUNCTIONS IN FREQUENCY SPACE
start = time.time()
M = np.fft.fftn(mf, axes=[2, 3, 4])
del mf
f.create_dataset('M_%s' % set_id, data=M)
f.close()
end = time.time()
timeE = np.round((end - start), 3)
msg = "FFT3 conversion of mf to M for %s: %s seconds" % \
(set_id, timeE)
rr.WP(msg, C['wrt_file'])
msg = 'Size of M: %s gb' % str(M.nbytes/(1e9))
rr.WP(msg, C['wrt_file'])
def pltcheck(ns, set_id):
C = const()
"""load the simulated and predicted responses"""
f = h5py.File("responses.hdf5", 'r')
r_sim = f.get('fip_%s' % set_id)[...]
r_sim = r_sim.reshape((C['n_samp']))
f.close()
f = h5py.File('validation_%s.hdf5' % set_id, 'r')
r_fit = f.get('r_fit')[...]
f.close()
r_sim = np.exp(r_sim)
r_fit = np.exp(r_fit)
"""plot the prediction equal to simulation line"""
plt.figure(num=6, figsize=[9, 8.5])
minval = np.min([r_sim])
maxval = np.max([r_sim])
valrange = maxval - minval
minval_ = minval - 0.5 * valrange
maxval_ = maxval + 0.5 * valrange
line = np.array([minval_, maxval_])
plt.plot(line, line, 'k-')
plt.plot(r_sim,
r_fit,
marker='o',
markersize=7,
color=[.7, .1, .1],
linestyle='')
minval_ = minval - 0.1 * valrange
maxval_ = maxval + 0.1 * valrange
plt.axis([minval_, maxval_, minval_, maxval_])
plt.title("predicted versus simulated ln(FIP)")
plt.xlabel("simulation")
plt.ylabel("prediction")
plt.xticks(rotation=20)
plt.yticks(rotation=20)
plt.show()
def validate(ns, set_id):
st = time.time()
C = const()
"""load the feature data"""
f = h5py.File("pre_regress_%s.hdf5" % set_id, 'r')
X = f.get('X')[...]
f.close()
"""gather the dependent variable data"""
f = h5py.File("responses.hdf5", 'r')
r_sim = f.get('fip_%s' % set_id)[...]
r_sim = r_sim.reshape((C['n_samp']))
f.close()
"""retrieve the coefficient set from the regression"""
f = h5py.File("regress_results.hdf5", 'r')
coef = f.get('coef')[...]
f.close()
"""evalute the fit response"""
r_fit = np.dot(coef, X.T)
f = h5py.File('validation_%s.hdf5' % set_id, 'w')
f.create_dataset('r_fit', data=r_fit)
f.close()
"""evalute error metrics"""
err = np.abs(r_sim-r_fit)
err_mean = err.mean()
err_max = err.max()
r_sim_mmm = np.array([r_sim.min(), r_sim.mean(), r_sim.max()])
r_fit_mmm = np.array([r_fit.min(), r_fit.mean(), r_fit.max()])
print "r_sim min, mean and max: %s" % str(r_sim_mmm)
print "r_fit min, mean and max: %s" % str(r_fit_mmm)
print "mean error: %s" % err_mean
print "max error: %s" % err_max
timeE = np.round(time.time()-st, 1)
msg = "validation completed for %s: %s s" % (set_id, timeE)
rr.WP(msg, C['wrt_file'])
def pltevd(sn, set_id):
C = const()
"""get the x, y data for plotting the evd"""
f = h5py.File("raw_responses.hdf5", 'r')
rawfip = f.get('fip_%s' % set_id)
x = np.sort(rawfip[sn, :])
x = x[np.int64(C['pcnt']*x.size):, None]
y = (np.arange(x.size)+1)/np.float32(x.size)
f.close
"""retrieve the coefficients"""
f = h5py.File("responses.hdf5", 'a')
c0 = f.get('c0_%s' % set_id)[sn]
c1 = f.get('c1_%s' % set_id)[sn]
c2 = f.get('c2_%s' % set_id)[sn]
f.close()
"""plot the original data and the fits"""
plt.figure()
plt.plot(np.log(x), y, 'b.', markersize=3)
tmp = np.linspace(np.log(x).min(), np.log(x).max(), 100)
x_ = np.exp(tmp)
plt.plot(np.log(x_), ss.gamma.cdf(x_, c0, loc=c1, scale=c2),
'r-', lw=2, label='gamma cdf')
ymin = y.min()
ymax = y.max()
rng = ymax - ymin
ymin += -0.1*rng
ymax += 0.1*rng
plt.ylim((ymin, ymax))
plt.show()
def read(ns, strt, set_id, newdir):
start = time.time()
C = const()
tmp = np.loadtxt("micr.txt", dtype='int16', delimiter=',').T
micr = tmp[strt:strt+ns, :]
f = h5py.File("spatial.hdf5", 'a')
f.create_dataset('micr_%s' % set_id, data=micr)
f.close()
end = time.time()
timeE = np.round((end - start), 3)
msg = 'microstructures read from .txt file for %s: %s seconds' % (set_id,
timeE)
rr.WP(msg, C['wrt_file'])
def get_mf(ns, set_id):
st = time.time()
C = const()
"""get the microstructure files"""
f = h5py.File("spatial.hdf5", 'a')
micr = f.get('micr_%s' % set_id)[...]
mf = np.zeros([ns, C['H'], C['el']**3], dtype='int16')
for h in xrange(C['H']):
mf[:, h, :] = micr == h
f.create_dataset('mf_%s' % set_id, data=mf)
f.close()
end = time.time()
timeE = np.round((end - st), 3)
msg = "mf calculated: %s seconds" % timeE
rr.WP(msg, C['wrt_file'])
def neighbors(ns, set_id):
st = time.time()
C = const()
f = h5py.File("spatial.hdf5", 'a')
mf = f.get('mf_%s' % set_id)[...]
mf = mf.swapaxes(1, 2)
mf = mf.reshape((ns, C['el'], C['el'], C['el'], C['H']))
exth = np.floor(0.5 * C['ext'])
cvec = np.arange(C['cmax'])
cmat = np.unravel_index(cvec, (C['ext'], C['ext'], C['ext']))
cmat = np.array(cmat).T
neig = np.zeros((ns, C['el'], C['el'], C['el'], C['H'], C['cmax']),
dtype='int16')
for cc in cvec:
ii, jj, kk = cmat[cc, :]
inx = np.int16(ii - exth)
iny = np.int16(jj - exth)
inz = np.int16(kk - exth)
tmp = np.roll(mf, inx, 1)
tmp = np.roll(tmp, iny, 2)
tmp = np.roll(tmp, inz, 3)
neig[..., cc] = tmp
neig = neig.reshape((ns, C['el']**3, C['H'], C['cmax']))
neig = f.create_dataset('neig_%s' % set_id, data=neig)
f.close()
timeE = np.round(time.time() - st, 5)
msg = "neighbors found for %s: %ss" % (set_id, timeE)
rr.WP(msg, C['wrt_file'])
def get_mf(ns, set_id):
st = time.time()
C = const()
"""get the euler angle files"""
f = h5py.File("spatial.hdf5", 'a')
euler = f.get('euler_%s' % set_id)[...]
mf = np.zeros([ns, C['H'], C['el']**3], dtype='float64')
for h in xrange(C['H']):
tmp = gsh.gsh_eval(euler.swapaxes(1, 2), [h])
tmp = np.squeeze(tmp)
mf[:, h, :] = tmp # 2*l+1 included in maple generator
f.create_dataset('mf_%s' % set_id, data=mf)
f.close()
end = time.time()
timeE = np.round((end - st), 3)
msg = "mf calculated: %s seconds" % timeE
rr.WP(msg, C['wrt_file'])
def transform(ns, set_id, pca):
st = time.time()
C = const()
n_corr = C['H']**2
f_red = h5py.File("spatial_reduced.hdf5", 'a')
f_stats = h5py.File("spatial.hdf5", 'r')
ff = f_stats.get('ff_%s' % set_id)[...]
ff = ff.reshape(ns, n_corr * C['el']**3)
tmp = pca.transform(ff)
f_red.create_dataset('reduced_%s' % set_id, data=tmp, dtype='float64')
f_red.close()
f_stats.close()
timeE = np.round(time.time() - st, 2)
msg = "transform to low dimensional space, %s: %s s" % (set_id, timeE)
rr.WP(msg, C['wrt_file'])
import plot_correlation as pltcorr import plot_pc_map as pltmap import plot_response as pr import plot_linkage_check as plc import explained_variance as ev import plot_pc_vs_poly as pltpcpoly import numpy as np from constants import const C = const() names_cal = C['names_cal'] set_id_cal = C['set_id_cal'] strt_cal = C['strt_cal'] ns_cal = C['ns_cal'] dir_cal = C['dir_cal'] names_val = C['names_val'] set_id_val = C['set_id_val'] strt_val = C['strt_val'] ns_val = C['ns_val'] dir_val = C['dir_val'] par = 'c4' # """Plot an autocorrelation""" # sn = 0 # iA = 1 # iB = 1 # pltcorr.pltcorr(ns_cal[0], set_id_cal[0], sn, iA, iB) """Plot the percentage explained variance"""
def evalf(theta, euler, var_id, thr, LL_p):
"""variable assignments
var_id 0: sigma'11
var_id 1: sigma'22
var_id 2: sigma'33
var_id 3: sigma'12
var_id 4: sigma'12
var_id 5: sigma'23
var_id 6: total shear rate
var_id 7: w12
var_id 8: w13
var_id 9: w23
"""
filename = "log_eval.txt"
C = constants.const()
f = h5py.File('coef.hdf5', 'r')
coef = f.get('coef')[:, var_id]
f.close()
basis_info = gsh.gsh_basis_info()
N_p_tmp = np.sum(
basis_info[:, 0] <= LL_p) # number of GSH bases to evaluate
N_pts = theta.size
"""Select the desired set of coefficients"""
msg = "cmax: %s" % C['cmax']
fn.WP(msg, filename)
cmat = np.unravel_index(np.arange(C['cmax']), C['N_tuple'])
cmat = np.array(cmat).T
cuttoff = thr * np.abs(coef).max()
print "cutoff: %s" % cuttoff
cuttoffvec = (np.abs(coef) > cuttoff) * \
(np.arange(C['cmax']) < N_p_tmp*C['N_q'])
print "cuttoffvec.shape: %s" % str(cuttoffvec.shape)
indxvec = np.arange(C['cmax'])[cuttoffvec]
N_coef = indxvec.size
pct_coef = 100. * N_coef / (N_p_tmp * C['N_q'])
fn.WP("number of coefficients retained: %s" % N_coef, filename)
fn.WP("percentage of coefficients retained %s%%" % np.round(pct_coef, 4),
filename)
"""Evaluate the parts of the basis function individually"""
st = time.time()
p_U = np.unique(cmat[indxvec, 0])
q_U = np.unique(cmat[indxvec, 1])
all_basis_p = np.zeros([N_pts, C['N_p']], dtype='complex128')
for p in p_U:
all_basis_p[:, p] = np.squeeze(gsh.gsh_eval(euler, [p]))
fn.WP("number of p basis functions used: %s" % p_U.size, filename)
all_basis_q = np.zeros([N_pts, C['N_q']], dtype='complex128')
for q in q_U:
all_basis_q[:, q] = np.cos(q * np.pi * theta / C['L_th'])
fn.WP("number of q basis functions used: %s" % q_U.size, filename)
"""Perform the prediction"""
Y_ = np.zeros(theta.size, dtype='complex128')
for ii in indxvec:
p, q = cmat[ii, :]
basis_p = all_basis_p[:, p]
basis_q = all_basis_q[:, q]
ep_set = basis_p * basis_q
Y_ += coef[ii] * ep_set
if np.mod(ii, 1000) == 0:
msg = "evaluation complete for coefficient" +\
" %s out of %s" % (ii, N_coef)
fn.WP(msg, filename)
Ttime = np.round(time.time() - st, 3)
msg = "total interpolation time: %ss" % Ttime
fn.WP(msg, filename)
msg = "interpolation time per point: %s" % (Ttime / theta.size)
fn.WP(msg, filename)
return Y_
import h5py import sys """ in this version of the code the id of the tensor is an argument to the script. trying to reduce the amount of data to analyse by half sampling in the angular variable """ # initialize important variables tnum = np.int64(sys.argv[1]) C = constants.const() # these indices are defined for the sampled db inputs sub2rad_eul = C['inc_eul']*np.pi/180. sub2rad_th = C['inc_th']*np.pi/180. # here we determine the sampling for en a_std = 0.0050 b_std = 0.0085 en_inc = 0.0001 # en increment et_norm = np.linspace(.0001, .0100, 100) ai = np.int64(np.round(a_std/en_inc))-1 # index for start of en range bi = np.int64(np.round(b_std/en_inc))-1 # index for end of en range sample_indx = np.arange(ai, bi+5, 3) n_en = sample_indx.size
