def __init__(self,
template_filename, target_filename,
common_dir, uncertain_params, cpo_name, link_xmlfiles=False):
# Check that user has specified the objests to use as template
if template_filename is None:
msg = ("CPOEncoder must be given 'template_filename': a CPO file.")
logging.error(msg)
raise RuntimeError(msg)
self.template_filename = template_filename
self.target_filename = target_filename
self.common_dir = common_dir
self.uncertain_params = uncertain_params
self.cpo_name = cpo_name
self.link_xmlfiles = link_xmlfiles
self.fixture_support = True
# The CPO object
cpo_filename = os.path.join(common_dir, template_filename)
self.cpo_core = read(cpo_filename, cpo_name)
# Mapping with cpo file
# TODO move to switcher tools routine
self.mapper = {
"Te_boundary" : self.cpo_core.te.boundary.value[0],
"Ti_boundary" : self.cpo_core.ti.boundary.value[0][0]
}
def build_cpo(filelist, codename, cponame, cpoarray):
#from IPython.parallel import Client
#c = Client()
#csize = len(c.ids)
#print 'exec on '+str(csize)+' processes'
l = len(filelist)
for i in range(l):
cpoarray[i] = ascii_cpo.read(filelist[i], cponame, cpoarray[i])
def get_flux_index(corep_file, coret_file):
corep = read(corep_file, 'coreprof')
coret = read(coret_file, 'coretransp')
# rho_tor_norm_transp_flux
rt = coret.values[0].rho_tor_norm
n_flux = len(rt)
print('rho_cores = ', rt)
# rho_tor_norm vector in coreprof
r = corep.rho_tor_norm
# the closest rho_tor in coreprof
rp = [r.flat[np.abs(r - rt[i]).argmin()] for i in range(n_flux)]
print('rho_corep = ', rp)
# the correponding index
idx = [list(r).index(rp[i]) for i in range(n_flux)]
return idx
def parse_sim_output(self, run_info={}):
out_path = self._get_output_path(run_info, self.target_filename)
# The CPO object
cpo_core = read(out_path, self.cpo_name)
# Get Quantity of Intersets
qoi_values = self._get_qoi_values(cpo_core)
quoi_dict = {}
for qoi in self.output_columns:
quoi_dict.update({qoi: qoi_values[qoi]})
# Output data frame
data = pd.DataFrame(quoi_dict)
return data
def get_cpo_from_dir(path, codename, cponame, time_pattern):
"""
Read CPO files corresponding to pattern in order to build and return a time serie.
\nParameters
----------
path : str
Path to the directory containing CPO files
codename : str
Name of the code which produced the CPO
cponame : str
Name of the CPO
time_pattern : str
Pattern to match the time slice part of the cpo file
e.g: 'codename_cponame_timeslice\.cpo' where timeslice='0...' for slices between 0 and 999
\nReturns
-------
cpo_list :
CPO array
"""
db = ual.itm()
prevdir = os.getcwd()
os.chdir(path)
ll = os.listdir('.')
ll.sort()
pattern = re.compile(codename+'_'+cponame+'_'+time_pattern+'\.cpo')
lcpo = pattern.findall(ll.__str__())
lcpolen = len(lcpo)
cpoobj = getattr(db,cponame+'Array')
cpoobj.resize(lcpolen)
cpoarr = getattr(cpoobj,'array')
for i in range(lcpolen):
cpoarr[i] = ascii_cpo.read(lcpo[i],cponame)
os.chdir(prevdir)
return cpoobj
default_campaign_dir_prefix='ETS_Campaign_')
campaign_dir = ets_campaign.campaign_dir
# Copy xml files needed in the ETS code
os.system("mkdir " + campaign_dir + "/workflows")
os.system("cp ../../workflows/ets.xml " + campaign_dir + "/workflows")
os.system("cp ../../workflows/ets.xsd " + campaign_dir + "/workflows")
# Copy CPO files in common directory
common_dir = campaign_dir + "/common/"
os.system("cp " + cpo_dir + "/*.cpo " + common_dir)
# Get uncertain parameters distrubutions
coret_file = common_dir + "ets_coretransp_in.cpo"
coret = read(coret_file, "coretransp")
diff_eff = coret.values[0].te_transp.diff_eff
list_dist = [cp.Normal(diff_eff[i], 0.2 * diff_eff[i]) for i in range(4)]
# Define the parameters dictionary
for i in range(4):
ets_campaign.vary_param(uncert_params[i], dist=list_dist[i])
# Create the sampler
ets_sampler = uq.elements.sampling.PCESampler(ets_campaign)
# Generate runs
ets_campaign.add_runs(ets_sampler)
ets_campaign.populate_runs_dir()
# Execute runs
# Add the BC app (automatically set as current app)
my_campaign.add_app(name="uq_ets_chease",
params=params,
encoder=encoder,
decoder=decoder)
# Create a collation element for this campaign
collater = uq.collate.AggregateSamples(average=False)
my_campaign.set_collater(collater)
# Get uncertain parameters values
#coret_file = common_dir + "ets_coretransp_in.cpo"
#coret = read(coret_file, "coretransp")
#diff_eff = coret.values[0].te_transp.diff_eff
corep_file = common_dir + "ets_coreprof_in.cpo"
corep = read(corep_file, "coreprof")
Te_boundary = corep.te.boundary.value[0]
Ti_boundary = corep.ti.boundary.value[0][0]
# Create the sampler
#vary = { uparams[k]: cp.Normal(diff_eff[k], 0.2*diff_eff[k]) for k in range(4)}
vary = {
uparams[0]: cp.Normal(Te_boundary, 0.2 * Te_boundary),
uparams[1]: cp.Normal(Ti_boundary, 0.2 * Ti_boundary)
}
my_sampler = uq.sampling.PCESampler(vary=vary, polynomial_order=4)
# Associate the sampler with the campaign
my_campaign.set_sampler(my_sampler)
pctl_te = results['percentiles']['Te']
stot_te = results['sobols_total']['Te']
stats_ti = results['statistical_moments']['Ti']
pctl_ti = results['percentiles']['Ti']
stot_ti = results['sobols_total']['Ti']
print('>>> Ellapsed time: ', time.time() - time0)
# Graphics for Descriptive satatistics
__PLOTS = False # If True create plots subfolder under outputs folder
if __PLOTS:
from utils import plots
uparams_names = list(params.keys())
corep = read(os.path.join(cpo_dir, "ets_coreprof_in.cpo"), "coreprof")
rho = corep.rho_tor
uparams_names = list(params.keys())
test_case = cpo_dir.split('/')[-1]
plots.plot_stats_pctl(rho, stats_te, pctl_te,
xlabel=r'$\rho_{tor} ~ [m]$', ylabel=r'$Te$',
ftitle='Te profile ('+test_case+')',
fname='plots/Te_STAT_'+test_case)
plots.plot_sobols(rho, stot_te, uparams_names,
ftitle=' Total-Order Sobol indices - QoI: Te',
fname='plots/Te_SA_'+test_case)
plots.plot_stats_pctl(rho, stats_ti, pctl_ti,
#st_te = results['sobol_total_order']['te']
stats_ti = results['statistical_moments']['ti']
pctl_ti = results['percentiles']['ti']
#s1_ti = results['sobol_first_order']['ti']
#st_ti = results['sobol_total_order']['ti']
# Elapsed time
end_time = time.time()
print('======= Elapsed times')
print('- EXEC : ', exec_time / 60.)
print('- TOTAL : ', (end_time - start_time) / 60.)
# Graphics for descriptive satatistics
corep_file = common_dir + '/ets_coreprof_in.cpo'
corep = read(corep_file, 'coreprof')
rho = corep.rho_tor
plots.plot_stats_pctl(rho,
stats_te,
pctl_te,
xlabel=r'$\rho_{tor} ~ [m]$',
ylabel=r'$T_e ~ [eV]$',
ftitle='UQ: Te profile (Uncertenties in Ion sources)',
fname='figs/te_sr_ions-stats.png')
plots.plot_stats_pctl(rho,
stats_ti,
pctl_ti,
xlabel=r'$\rho_{tor} ~ [m]$',
ylabel=r'$T_i ~ [eV]$',
header=0)
# Add the ETS app (automatically set as current app)
my_campaign.add_app(name="uq_ets",
params=params,
encoder=encoder,
decoder=decoder
)
# Create a collation element for this campaign
collater = uq.collate.AggregateSamples(average=False)
my_campaign.set_collater(collater)
# Get uncertain parameters values
coret_file = common_dir + "ets_coretransp_in.cpo"
coret = read(coret_file, "coretransp")
diff_eff = coret.values[0].te_transp.diff_eff
corep_file = common_dir + "ets_coreprof_in.cpo"
corep = read(corep_file, "coreprof")
Te_boundary = corep.te.boundary.value[0]
Ti_boundary = corep.ti.boundary.value[0][0]
# Create the sampler
print('Create the sampler')
vary = {uncertain_params[k]: cp.Normal(diff_eff[k], 0.2*diff_eff[k]) for k in range(4)}
vary.update({
uncertain_params[4]: cp.Normal(Te_boundary, 0.2*Te_boundary),
uncertain_params[5]: cp.Normal(Ti_boundary, 0.2*Ti_boundary)
})
def ets_test(uncert_params):
# The ets_run executable (to run the ets model)
bin_file = "../bin/" + SYS + "/ets_run "
# Input/Output template
input_json = "inputs/ets_in.json"
output_json = os.path.join(TMP_DIR, "out_ets.json")
# Initialize Campaign object
ets_campaign = uq.Campaign(name='ETS_Campaign',
state_filename=input_json,
workdir=TMP_DIR,
default_campaign_dir_prefix='UQP2_ETS_')
campaign_dir = ets_campaign.campaign_dir
# Copy xml files needed in the ETS code
os.system("mkdir " + campaign_dir + "/workflows")
os.system("cp ../../workflows/ets.xml " + campaign_dir + "/workflows")
os.system("cp ../../workflows/ets.xsd " + campaign_dir + "/workflows")
# Copy CPO files in common directory
common_dir = campaign_dir + "/common/"
os.system("cp " + CPO_DIR + "/*.cpo " + common_dir)
# Get uncertain parameters distrubutions
coret_file = common_dir + "ets_coretransp_in.cpo"
coret = read(coret_file, "coretransp")
diff_eff = coret.values[0].te_transp.diff_eff
list_dist = [cp.Normal(diff_eff[i], 0.2 * diff_eff[i]) for i in range(4)]
# Define the parameters dictionary
for i in range(4):
ets_campaign.vary_param(uncert_params[i], dist=list_dist[i])
# Create the sampler
ets_sampler = uq.elements.sampling.PCESampler(ets_campaign)
# Generate runs
ets_campaign.add_runs(ets_sampler)
ets_campaign.populate_runs_dir()
# Execute runs
cmd = bin_file + common_dir + " ets_input.nml"
ets_campaign.apply_for_each_run_dir(uq.actions.ExecuteLocal(cmd))
# Aggregate the results from all runs.
output_filename = ets_campaign.params_info['out_file']['default']
output_columns = ['c']
aggregate = uq.elements.collate.AggregateSamples(
ets_campaign,
output_filename=output_filename,
output_columns=output_columns,
header=0,
)
aggregate.apply()
# Analysis
analysis = uq.elements.analysis.PCEAnalysis(ets_campaign,
value_cols=output_columns)
dist, cov = analysis.apply()
# Results
cov_mnat = cov["c"]
return dist["c"]
return stats, sobols
# Main
if __name__ == "__main__":
# Uncertain parameters
uncert_params_in = ["D1", "D2", "D3", "D4"]
# The 1st box: ETS
dist = ets_test(uncert_params_in)
# The 2nd box: CHEASE
stats, sobols = chease_test(dist)
# PLOTS
eq_file = CPO_DIR + "/ets_equilibrium_in.cpo"
eq = read(eq_file, "equilibrium")
rho = eq.profiles_1d.rho_tor
plots.plot_stats(rho,
stats,
xlabel=r'$\rho_{tor} ~ [m]$',
ylabel=r'<B>',
ftitle='B_av profile',
fname='figs/b_prof.png')
plot.plot_sobols_4(rho, sobols, uncert_params_in, 'b_av')
if range_pattern == None:
range_pattern = '....'
pattern = re.compile(codename + '_' + cponame + '_' +
range_pattern + '\.cpo')
lcpo = pattern.findall(ll.__str__())
lcpolen = len(lcpo)
cpoobj = getattr(db, cponame + 'Array')
cpoobj.resize(lcpolen)
cpoarr = getattr(cpoobj, 'array')
logger.info("Reading CPO %s from %s", cponame, codename)
for i in range(lcpolen):
logger.info("slice #%i", i)
cpoarr[i] = read(lcpo[i], cponame)
logger.info("Writing CPO into database")
getattr(cpoobj, "setExpIdx")(db.expIdx)
cpoobj.put()
########################## STORE PROVENANCE DATA ############################
logger.critical("Saving provenance data into topinfo CPO")
# todo: get real name instead of username
db.topinfo.dataprovider = usr
db.topinfo.description = entry.get("DESCRIPTION")
db.topinfo.firstputdate = datetime.now().strftime(
'%Y-%m-%d (%Hh%Mm%Ss)')
db.topinfo.dataversion = ver
workflow = simulation.get('WORKFLOW')
import pylab
from ascii_cpo import read
time = []
te_1 = []
te_2 = []
te_3 = []
te_4 = []
te_5 = []
te_6 = []
print("reading coreprof and coretransp CPOs")
for i in range(50):
corep = read('cpos/ets_coreprof_'+str(i+1).zfill(4)+'.cpo', 'coreprof')
time.append(corep.time)
te = corep.te.value
te_1.append(te[0])
te_2.append(te[19])
te_3.append(te[39])
te_4.append(te[59])
te_5.append(te[79])
te_6.append(te[99])
r = corep.rho_tor_norm
fig = pylab.figure(figsize=(12,9))
ax1 = pylab.subplot(111)
ax1.set_xlabel('time')
ax1.set_ylabel(r'$t_e \quad [eV]$')
def plot_dv(args, step, ext='pdf', show=True):
if args.title == None:
title = "Static=0-20, Lim=0.2/0.09, Tau=0.01, Step=99"
else:
title = args.title
sstep = str(step).zfill(4)
corep = ascii_cpo.read(args.path + "/ets_coreprof_" + sstep + ".cpo",
"coreprof")
coretw = ascii_cpo.read(
args.path + "/ets_coretransp-work_in_" + sstep + ".cpo", "coretransp")
coretdv = ascii_cpo.read(
args.path + "/imp4dv_coretransp_" + sstep + ".cpo", "coretransp")
coretgem = ascii_cpo.read(args.path + "/gem_coretransp_" + sstep + ".cpo",
"coretransp")
fulltitle = title + ", Step=" + sstep
if args.Te:
f, (ax1, ax2) = pylab.subplots(2,
figsize=(8, 12),
sharex=True,
sharey=False)
ax1.set_title(fulltitle)
ax1.plot(corep.rho_tor_norm,
coretw.values[0].te_transp.diff_eff,
'b',
label="used in ETS")
ax1.plot(coretdv.values[0].rho_tor_norm,
coretdv.values[0].te_transp.diff_eff,
'bo',
label="IMP4dv coefs")
ax1.plot(coretgem.values[0].rho_tor_norm,
coretgem.values[0].te_transp.diff_eff,
'g^',
label="GEM coefs")
ax1.set_ylabel("Te/Diff_eff")
ax1.legend()
ax2.plot(corep.rho_tor_norm,
coretw.values[0].te_transp.vconv_eff,
'b',
label="used in ETS")
ax2.plot(coretdv.values[0].rho_tor_norm,
coretdv.values[0].te_transp.vconv_eff,
'bo',
label="IMP4dv coefs")
ax2.plot(coretgem.values[0].rho_tor_norm,
coretgem.values[0].te_transp.vconv_eff,
'g^',
label="GEM coefs")
ax2.set_xlabel("rho_tor_norm")
ax2.set_ylabel("Te/Vconv_eff")
f.subplots_adjust(hspace=0)
#pylab.tight_layout()
if show:
pylab.show()
if args.out != None:
f.savefig("Te_" + sstep + "_" + args.out + "." + ext)
if args.Ti:
f, (ax1, ax2) = pylab.subplots(2,
figsize=(8, 12),
sharex=True,
sharey=False)
ax1.set_title(fulltitle)
ax1.plot(corep.rho_tor_norm,
coretw.values[0].ti_transp.diff_eff,
'b',
label="used in ETS")
ax1.plot(coretdv.values[0].rho_tor_norm,
coretdv.values[0].ti_transp.diff_eff,
'bo',
label="IMP4dv coefs")
ax1.plot(coretgem.values[0].rho_tor_norm,
coretgem.values[0].ti_transp.diff_eff,
'g^',
label="GEM coefs")
ax1.set_ylabel("Ti/Diff_eff")
ax1.legend()
ax2.plot(corep.rho_tor_norm,
coretw.values[0].ti_transp.vconv_eff,
'b',
label="used in ETS")
ax2.plot(coretdv.values[0].rho_tor_norm,
coretdv.values[0].ti_transp.vconv_eff,
'bo',
label="IMP4dv coefs")
ax2.plot(coretgem.values[0].rho_tor_norm,
coretgem.values[0].ti_transp.vconv_eff,
'g^',
label="GEM coefs")
ax2.set_xlabel("rho_tor_norm")
ax2.set_ylabel("Ti/Vconv_eff")
f.subplots_adjust(hspace=0)
#pylab.tight_layout()
if show:
pylab.show()
if args.out != None:
f.savefig("Ti_" + sstep + "_" + args.out + "." + ext)
import pylab
import ascii_cpo
corep = ascii_cpo.read("ets_coreprof_in.cpo", "coreprof")
pylab.figure(1)
pylab.suptitle("Initial profiles for JET#92436/270")
pylab.subplot(321)
pylab.plot(corep.rho_tor_norm, corep.profiles1d.q.value)
#pylab.xlabel("rho_tor_norm")
pylab.ylabel("q")
pylab.subplot(322)
pylab.plot(corep.rho_tor_norm, corep.profiles1d.pr_th.value)
#pylab.xlabel("rho_tor_norm")
pylab.ylabel("pr_th")
pylab.subplot(323)
pylab.plot(corep.rho_tor_norm, corep.te.value)
#pylab.xlabel("rho_tor_norm")
pylab.ylabel("te")
pylab.subplot(324)
pylab.plot(corep.rho_tor_norm, corep.ti.value)
#pylab.xlabel("rho_tor_norm")
pylab.ylabel("ti")
pylab.subplot(325)
pylab.plot(corep.rho_tor_norm, corep.ne.value)
pylab.xlabel("rho_tor_norm")
