def compute_predominant_period(gmr_name, gmrs, damping):
igmr = gmrs['name'].index(gmr_name)
T = np.arange(0.01, 2, 0.01)
Sv = utils.calculate_iml(igmr, gmrs, 'Sv', T, damping)
Tg = T[Sv == max(Sv)]
return Tg[0]
def compute_predominant_period(gmr_name, gmrs, damping):
igmr = gmrs['name'].index(gmr_name)
T = np.arange(0.01,2,0.01)
Sv = utils.calculate_iml(igmr, gmrs, 'Sv', T, damping)
Tg = T[Sv==max(Sv)]
return Tg[0]
def MSA_post_processing(PDM, gmrs, gmr_info, IMT, msa, damage_model, T,
damping, method, initial_damage):
Nrecords = msa['records per bin']
Nstripes = msa['n. bins']
no_gmrs = Nrecords * Nstripes
imls = []
for index in np.arange(0, no_gmrs, Nrecords):
igmr = gmr_info[index, 0]
scaling = gmr_info[index, 1]
imls.append(
utils.calculate_iml(int(igmr), gmrs, IMT, T, damping) * scaling)
# plt.plot(imls[0],Sds['iDS1'],marker='o', linestyle='None')
# plt.ylim((0,0.5));
no_assets = np.sum(PDM, axis=1, dtype=float)
cumPDM = np.fliplr(np.cumsum(np.fliplr(PDM), axis=1))
cumPDM = np.transpose(cumPDM)
cumPDM = cumPDM / no_assets
logmeans, logstddev = [], []
I = np.argsort(imls, axis=0)
imls = np.array(imls)
imls = imls[I]
if method == 'least squares':
for iDS in range(initial_damage, len(damage_model['damage_states'])):
cumPDM[iDS + 1] = cumPDM[iDS + 1][I]
solution, _ = optimize.leastsq(utils.residual_lognormal_dist,
[0.1, 0.6],
args=(imls, cumPDM[iDS + 1]))
logmeans.append(solution[0])
logstddev.append(solution[1])
elif method == 'max likelihood':
# the imls and the PDM are sorted to be used in the maximum likelihood approach
# the PDM should be a list of integers in this approach
# I = np.argsort(imls, axis=0)
# imls = np.array(imls)
# imls = imls[I]
PDM_mle, PoE = [], []
for iDS in range(initial_damage, len(damage_model['damage_states'])):
cumPDM[iDS + 1] = cumPDM[iDS + 1][I]
temp = cumPDM[iDS + 1] * no_assets
temp2 = temp.astype(int)
PDM_mle = temp2.tolist()
solution = utils.mle(PDM_mle, imls, no_assets, 1)
logmeans.append(solution[0])
logstddev.append(solution[1])
Rsquare = utils.calculate_correlation(logmeans, logstddev,
cumPDM[initial_damage:], imls)
fragility_model = {
'damage_states': None,
'logmean': None,
'logstddev': None,
'IMT': None
}
fragility_model['damage_states'] = damage_model['damage_states'][
initial_damage:]
fragility_model['logmean'] = logmeans
fragility_model['logstddev'] = logstddev
fragility_model['IMT'] = IMT
fragility_model['Rsquare'] = Rsquare
fragility_model['PoE points'] = cumPDM[initial_damage + 1:]
fragility_model['imls'] = imls
return fragility_model
def MSA_post_processing(PDM,gmrs,gmr_info,IMT,msa,damage_model,T,damping, method,initial_damage):
Nrecords = msa['records per bin']
Nstripes = msa['n. bins']
no_gmrs = Nrecords*Nstripes
imls = []
for index in np.arange(0,no_gmrs,Nrecords):
igmr = gmr_info[index,0]
scaling = gmr_info[index,1]
imls.append(utils.calculate_iml(int(igmr), gmrs, IMT, T, damping)*scaling)
# plt.plot(imls[0],Sds['iDS1'],marker='o', linestyle='None')
# plt.ylim((0,0.5));
no_assets = np.sum(PDM, axis=1, dtype=float)
cumPDM = np.fliplr(np.cumsum(np.fliplr(PDM), axis=1))
cumPDM = np.transpose(cumPDM)
cumPDM = cumPDM/no_assets
logmeans, logstddev = [], []
I = np.argsort(imls, axis=0)
imls = np.array(imls)
imls = imls[I]
if method == 'least squares':
for iDS in range(initial_damage,len(damage_model['damage_states'])):
cumPDM[iDS+1] = cumPDM[iDS+1][I]
solution, _ = optimize.leastsq(utils.residual_lognormal_dist,
[0.1, 0.6],
args=(imls, cumPDM[iDS+1]))
logmeans.append(solution[0])
logstddev.append(solution[1])
elif method == 'max likelihood':
# the imls and the PDM are sorted to be used in the maximum likelihood approach
# the PDM should be a list of integers in this approach
# I = np.argsort(imls, axis=0)
# imls = np.array(imls)
# imls = imls[I]
PDM_mle, PoE = [], []
for iDS in range(initial_damage,len(damage_model['damage_states'])):
cumPDM[iDS+1] = cumPDM[iDS+1][I]
temp = cumPDM[iDS+1]*no_assets
temp2 = temp.astype(int)
PDM_mle = temp2.tolist()
solution = utils.mle(PDM_mle,imls,no_assets,1)
logmeans.append(solution[0])
logstddev.append(solution[1])
Rsquare = utils.calculate_correlation(logmeans,logstddev,cumPDM[initial_damage:],imls)
fragility_model = {'damage_states': None, 'logmean': None,'logstddev': None, 'IMT': None}
fragility_model['damage_states'] = damage_model['damage_states'][initial_damage:]
fragility_model['logmean'] = logmeans
fragility_model['logstddev'] = logstddev
fragility_model['IMT'] = IMT
fragility_model['Rsquare'] = Rsquare
fragility_model['PoE points'] = cumPDM[initial_damage+1:]
fragility_model['imls'] = imls
return fragility_model
