def func_radiation_patter_show(singlerun_data_folder, ff_data_sub_folder_name):
np_theta = 360
np_phi = 5
ff_export_folder = os.path.join(singlerun_data_folder,
ff_data_sub_folder_name)
# Get all the farfield export data file
farfield_data_file_list = pycst_ctrl.func_file_list_get(ff_export_folder,
ext='.txt')
for export_file in farfield_data_file_list:
theta_vec, dir_co_arr, dir_cx_arr, dir_abs_arr, \
dir_co_norm_arr, dir_cx_norm_arr, dir_abs_norm_arr, \
peak_co_cvec, peak_cx_cvec, peak_abs_cvec, \
ar_arr, ar_boresight_cvec = \
PyCstDataAnalyser.func_exp_ff_data_proc(ff_export_folder, export_file, np_theta, np_phi)
plt.figure()
plt.plot(theta_vec, dir_co_norm_arr[0, :], 'r-')
plt.plot(theta_vec, dir_co_norm_arr[1, :], 'b-')
plt.plot(theta_vec, dir_co_norm_arr[2, :], 'k-')
plt.plot(theta_vec, dir_cx_norm_arr[1, :], 'b--')
plt.xlim(-180, 180)
plt.ylim(-40, 0)
plt.minorticks_on()
plt.grid(True, which='both', linestyle=':')
plt.show()
def __init__(self, proj_cfg_dic, data_analyser_cfg_dic, optimizer_cfg_dic):
# Project configuration
# Settings from users
self.env = proj_cfg_dic.get('env', 'hpc')
self.solver = proj_cfg_dic.get('solver', 'T')
self.gpu_num = proj_cfg_dic.get('gpu_num', 1)
# Indicate if this is an optimization resumed from previous one
self.resume_flag = proj_cfg_dic.get('resume_flag', False)
self.project_folder = proj_cfg_dic.get('project_folder',
"/data/home/eex181/SmoothWallHornProfileRelative_Python")
self.sim_name = proj_cfg_dic.get('sim_name', "SmoothWallHornProfileRelative_Python")
self.base_para_file_name = proj_cfg_dic.get('base_para_file_name', "ProfiledSmoothWallHornRelativePara.txt")
# CST project controlled by matlab will be Farfields
self.ff_export_subfolder_name = proj_cfg_dic.get('ff_export_subfolder_name', "Farfield")
# Record csv files of parameters and objective values during optimization
self.rec_para_file_name = proj_cfg_dic.get('rec_para_file_name', 'opt_para_list.csv')
self.rec_para_name_lst = proj_cfg_dic.get('rec_para_name_lst', ['EMPTY'])
self.rec_objval_file_name = proj_cfg_dic.get('rec_objval_file_name', 'opt_obj_val.csv')
self.rec_obj_name_lst = proj_cfg_dic.get('rec_obj_name_lst', ['EMPTY'])
# Define cma object saved file name for resuming
self.cma_object_save_file_name = proj_cfg_dic.get('cma_object_save_file_name', '_saved-cma-object.pkl')
self.full_cma_object_save_file = os.path.join(self.project_folder, self.cma_object_save_file_name)
# Initialize run_id
self.run_id = proj_cfg_dic.get('init_run_id', 1) # Default run_id starts from 1
# Settings to generate
self.cst_file_name = self.sim_name + '.cst'
self.export_path = os.path.join(self.project_folder, self.sim_name, 'Export')
self.export_py_path = os.path.join(self.project_folder, self.sim_name, 'ExportPy')
self.parafile_subfolder_name = 'ParaFiles'
self.full_sim_para_file = os.path.join(self.project_folder, 'CurrentPara.txt')
self.full_cst_file = os.path.join(self.project_folder, self.cst_file_name)
# Create data analyser instance
self.data_analyser_ins = PyCstDataAnalyser(data_analyser_cfg_dic)
# Create optimizer instance
self.optimizer_ins = self.func_opt_cma_create(optimizer_cfg_dic)
def func_norm_radiation_pattern_freq_sim_show(ff_data_dir, ff_file_name,
np_theta, np_phi, axes_cfg_dic,
freq_str, pol_ind):
"""Display simulated radiation pattern in 0/45/90 deg at one frequency"""
# Parse the exported farfield data file
theta_vec, dir_co_arr, dir_cx_arr, dir_abs_arr, \
dir_co_norm_arr, dir_cx_norm_arr, dir_abs_norm_arr, \
peak_co_cvec, peak_cx_cvec, peak_abs_cvec, \
ar_arr, ar_boresight_cvec = \
PyCstDataAnalyser.func_exp_ff_data_proc(ff_data_dir, ff_file_name, np_theta, np_phi, pol_ind)
# Create plot configuration
data_plot_cfg_sim_radpat_dic = {
'x_data_vec': theta_vec,
'y_data_group_dic': {
r'LHCP, $\phi=0\degree$': dir_co_norm_arr[0, :],
r'LHCP, $\phi=45\degree$': dir_co_norm_arr[1, :],
r'LHCP, $\phi=90\degree$': dir_co_norm_arr[2, :],
r'RHCP, $\phi=45\degree$': dir_cx_norm_arr[1, :]
},
'color_list': ['r', 'k', 'b', 'k'],
'line_style': ['-', '-', '-', '--'],
# 'marker': [None]*4,
# 'marker_edge_color': None,
'line_width': 2
}
data_plot_cfg_dic_lst = [data_plot_cfg_sim_radpat_dic]
x_label = 'Theta (deg)'
y_label = 'Directivity (dB)'
fig_title = pol_ind + ' (%s)' % freq_str
func_group_data_show(data_plot_cfg_dic_lst, axes_cfg_dic, x_label, y_label,
fig_title)
def func_cst_radpat_cocx_subplot_show(np_theta, np_phi, pol_ind, ff_data_dir,
ff_data_file_dic, nrow, ncol,
axes_cfg_dic):
"""
This function is used to plot radiation patterns with co-pol and cx-pol
:param np_theta: number of sample points for theta, normally 360
:param np_phi: number of sample points for phi, if step is 45deg, np_phi is 5
:param pol_ind: 'LHCP', 'RHCP' or 'lin_dir'
:param ff_data_dir: directory path which contains farfield data
:param ff_data_file_dic: a dictionary list of radiation patterns to be shown, should be like:
ff_data_file_dic = {'75 GHz': "farfield_(f=75)_[1].txt",
'85 GHz': "farfield_(f=85)_[1].txt",
'90 GHz': "farfield_(f=90)_[1].txt",
'95 GHz': "farfield_(f=95)_[1].txt",
'100 GHz': "farfield_(f=100)_[1].txt",
'110 GHz': "farfield_(f=110)_[1].txt"
}
:param nrow: number of rows
:param ncol: number of columns
:param axes_cfg_dic: see func_group_data_plot() for details
:return:
"""
# Create label (legend) list according to the polarization
if pol_ind == 'LHCP':
lable_lst = [
'LHCP, ' + r'$\phi=0\degree$', 'LHCP, ' + r'$\phi=45\degree$',
'LHCP, ' + r'$\phi=90\degree$', 'RHCP, ' + r'$\phi=45\degree$'
]
unit_str = 'dBic'
elif pol_ind == 'RHCP':
lable_lst = [
'RHCP, ' + r'$\phi=0\degree$', 'RHCP, ' + r'$\phi=45\degree$',
'RHCP, ' + r'$\phi=90\degree$', 'LHCP, ' + r'$\phi=45\degree$'
]
unit_str = 'dBic'
else:
lable_lst = [
'Co-pol, ' + r'$\phi=0\degree$', 'Co-pol, ' + r'$\phi=45\degree$',
'Co-pol, ' + r'$\phi=90\degree$', 'Cx-pol, ' + r'$\phi=45\degree$'
]
unit_str = 'dBi'
data_plot_cfg_dic_lst = []
anno_text_lst = []
for freq_str, ff_file_name in ff_data_file_dic.items():
# Parse the exported farfield data file
theta_vec, dir_co_arr, dir_cx_arr, dir_abs_arr, \
dir_co_norm_arr, dir_cx_norm_arr, dir_abs_norm_arr, \
peak_co_cvec, peak_cx_cvec, peak_abs_cvec, \
ar_arr, ar_boresight_cvec = \
PyCstDataAnalyser.func_exp_ff_data_proc(ff_data_dir, ff_file_name, np_theta, np_phi, pol_ind)
# Create plot configuration
data_plot_cfg_sim_radpat_dic = {
'x_data_vec': theta_vec,
'y_data_group_dic': {
lable_lst[0]: dir_co_norm_arr[0, :],
lable_lst[1]: dir_co_norm_arr[1, :],
lable_lst[2]: dir_co_norm_arr[2, :],
lable_lst[3]: dir_cx_norm_arr[1, :]
},
'color_list': ['r', 'k', 'b', 'g'],
'line_style': ['-', '--', '-.', ':'],
# 'marker': [None]*4,
# 'marker_edge_color': None,
'line_width': 1.5
}
# Create data list, one data_dic for each subplot
data_plot_cfg_dic_lst.append(data_plot_cfg_sim_radpat_dic)
# Create annotation text
anno_text_lst.append('D={0:.1f} {1}'.format(np.amax(peak_co_cvec),
unit_str))
x_label = 'Theta (deg)'
y_label = 'Normalized Magnitude (dB)'
sub_titles = [sub_title for sub_title in ff_data_file_dic.keys()]
func_group_data_subplot_show(data_plot_cfg_dic_lst, axes_cfg_dic, nrow,
ncol, x_label, y_label, sub_titles,
anno_text_lst)
class PyCstProjSimulator:
""" Define the simulation project"""
def __init__(self, proj_cfg_dic, data_analyser_cfg_dic, optimizer_cfg_dic):
# Project configuration
# Settings from users
self.env = proj_cfg_dic.get('env', 'hpc')
self.solver = proj_cfg_dic.get('solver', 'T')
self.gpu_num = proj_cfg_dic.get('gpu_num', 1)
# Indicate if this is an optimization resumed from previous one
self.resume_flag = proj_cfg_dic.get('resume_flag', False)
self.project_folder = proj_cfg_dic.get('project_folder',
"/data/home/eex181/SmoothWallHornProfileRelative_Python")
self.sim_name = proj_cfg_dic.get('sim_name', "SmoothWallHornProfileRelative_Python")
self.base_para_file_name = proj_cfg_dic.get('base_para_file_name', "ProfiledSmoothWallHornRelativePara.txt")
# CST project controlled by matlab will be Farfields
self.ff_export_subfolder_name = proj_cfg_dic.get('ff_export_subfolder_name', "Farfield")
# Record csv files of parameters and objective values during optimization
self.rec_para_file_name = proj_cfg_dic.get('rec_para_file_name', 'opt_para_list.csv')
self.rec_para_name_lst = proj_cfg_dic.get('rec_para_name_lst', ['EMPTY'])
self.rec_objval_file_name = proj_cfg_dic.get('rec_objval_file_name', 'opt_obj_val.csv')
self.rec_obj_name_lst = proj_cfg_dic.get('rec_obj_name_lst', ['EMPTY'])
# Define cma object saved file name for resuming
self.cma_object_save_file_name = proj_cfg_dic.get('cma_object_save_file_name', '_saved-cma-object.pkl')
self.full_cma_object_save_file = os.path.join(self.project_folder, self.cma_object_save_file_name)
# Initialize run_id
self.run_id = proj_cfg_dic.get('init_run_id', 1) # Default run_id starts from 1
# Settings to generate
self.cst_file_name = self.sim_name + '.cst'
self.export_path = os.path.join(self.project_folder, self.sim_name, 'Export')
self.export_py_path = os.path.join(self.project_folder, self.sim_name, 'ExportPy')
self.parafile_subfolder_name = 'ParaFiles'
self.full_sim_para_file = os.path.join(self.project_folder, 'CurrentPara.txt')
self.full_cst_file = os.path.join(self.project_folder, self.cst_file_name)
# Create data analyser instance
self.data_analyser_ins = PyCstDataAnalyser(data_analyser_cfg_dic)
# Create optimizer instance
self.optimizer_ins = self.func_opt_cma_create(optimizer_cfg_dic)
def func_opt_cma_create(self, optimizer_cfg_dic):
if self.resume_flag is False:
# Create optimizer object
cma_init_para_list = optimizer_cfg_dic.get('cma_init_para_list', None)
cma_sigma = optimizer_cfg_dic.get('cma_sigma', 0.3)
cma_opts = optimizer_cfg_dic.get('cma_opts', None)
es = cma.CMAEvolutionStrategy(cma_init_para_list, cma_sigma, cma_opts)
else:
# resuming from saved file
es = pycst_ctrl.func_load_ins_from_file(self.full_cma_object_save_file)
# overwrite the log directory for outcmaes to current project directory
# because cmaes recorded the absolute path in es.logger.name_prefix
es.logger.name_prefix = os.path.join(self.project_folder, 'outcmaes')
es.logger.name_prefix = es.logger.name_prefix + os.sep
# Set max iteration number when we know the approximate iteration number for one run (10-day)
cma_opts = optimizer_cfg_dic.get('cma_opts', None)
if cma_opts is not None:
max_iter_num = cma_opts.get('maxiter', None)
if max_iter_num is not None:
es.opts['maxiter'] = max_iter_num
return es
@staticmethod
def func_cst_sim_run(env, solver, gpu_num, full_para_file, full_cst_file):
if solver == 'F':
solver_cmd = '-f'
elif solver == 'T_GPU':
solver_cmd = '-r -withgpu=%d' % gpu_num
elif solver == 'I_GPU':
solver_cmd = '-q -withgpu=%d' % gpu_num
elif solver == 'T':
solver_cmd = '-r'
else:
solver_cmd = '-r'
if env == "linux_pc": # Linux_PC:
cmd_cst = '"/opt/cst/CST_STUDIO_SUITE_2019/cst_design_environment" -m -par "%s" -r "%s"' % (
full_para_file, full_cst_file)
elif env == "hpc":
cmd_cst = "singularity run --nv /data/containers/cst/cst cst_design_environment -numthreads=${NSLOTS} " \
"-m -par \"%s\" %s \"%s\"" % (full_para_file, solver_cmd, full_cst_file)
elif env == "win":
cmd_cst = '"C:\\Program Files (x86)\\CST STUDIO SUITE 2019\\CST DESIGN ENVIRONMENT.exe" ' \
'-m -par \"%s\" %s \"%s\"' % (full_para_file, solver_cmd, full_cst_file)
elif env == "test":
cmd_cst = ""
else:
cmd_cst = ""
if cmd_cst != "":
if env == "win":
os.system('"' + cmd_cst + '"')
else:
os.system(cmd_cst)
@staticmethod
def func_cst_parafile_gen(project_folder, parafile_subfolder_name, base_para_file_name, run_id, cst_para_val_vec):
# Get full path to the base para file
full_para_file = os.path.join(project_folder, parafile_subfolder_name, base_para_file_name)
# Update parameters and generate new para file
with open(full_para_file, 'r') as fread:
content_list = fread.readlines()
for i in range(len(cst_para_val_vec)):
para_val = cst_para_val_vec[i]
para_set_str = content_list[i]
para_set_str = para_set_str % para_val
content_list[i] = para_set_str
para_sweep_file_name = os.path.splitext(base_para_file_name)[0] + '_' + str(run_id) + '.txt'
full_para_sweep_file = os.path.join(project_folder, parafile_subfolder_name, para_sweep_file_name)
with open(full_para_sweep_file, 'w') as fwrite:
fwrite.writelines(content_list)
# Prepare the para file for current simulation
full_cur_para_file = os.path.join(project_folder, 'CurrentPara.txt')
shutil.copy(full_para_sweep_file, full_cur_para_file)
@staticmethod
def func_cst_opt_res_save(project_folder, para_list_file_name, obj_val_file_name, para_val_vec, para_name_list,
obj_val_vec, obj_name_list, run_id):
# Convert ndarray to pandas series
para_df = pd.DataFrame(data=para_val_vec.reshape(1, -1), index=[run_id], columns=para_name_list)
obj_df = pd.DataFrame(data=obj_val_vec.reshape(1, -1), index=[run_id], columns=obj_name_list)
# prepare the file to write
full_para_list_file = os.path.join(project_folder, para_list_file_name)
full_obj_val_file = os.path.join(project_folder, obj_val_file_name)
if 1 != run_id:
write_mode = 'a' # append if already exists
header_mode = False
else:
write_mode = 'w' # make a new file if not
header_mode = True
para_df.to_csv(full_para_list_file, mode=write_mode, header=header_mode)
obj_df.to_csv(full_obj_val_file, mode=write_mode, header=header_mode)
def func_cst_proj_sim(self, opt_para_val_vec):
# Read project configuration
project_folder = self.project_folder
export_path = self.export_path
export_py_path = self.export_py_path
ff_export_subfolder_name = self.ff_export_subfolder_name
parafile_subfolder_name = self.parafile_subfolder_name
base_para_file_name = self.base_para_file_name
full_sim_para_file = self.full_sim_para_file
full_cst_file = self.full_cst_file
# Get current run_id
run_id = self.run_id
# Create parameter list
cst_para_val_vec = opt_para_val_vec
# Generate the parameter file for simulation
self.func_cst_parafile_gen(project_folder, parafile_subfolder_name, base_para_file_name, run_id,
cst_para_val_vec)
# Run the simulation
self.func_cst_sim_run(self.env, self.solver, self.gpu_num, full_sim_para_file, full_cst_file)
if self.env == "test":
# objval_vec = np.random.random_sample((5,)) * 5
# fitness = objval_vec.sum()
export_sim_subfolder_name = 'run' + str(run_id)
bak_sim_subfolder_path = os.path.join(export_py_path, export_sim_subfolder_name)
# Analyse the export data
fitness, objval_vec = self.data_analyser_ins.func_cst_data_analyse(bak_sim_subfolder_path, run_id,
ff_export_subfolder_name)
else:
# Backup the export data of the current simulation
export_sim_subfolder_name = 'run' + str(run_id)
bak_sim_subfolder_path = os.path.join(export_py_path, export_sim_subfolder_name)
shutil.copytree(export_path, bak_sim_subfolder_path)
# Analyse the export data
fitness, objval_vec = self.data_analyser_ins.func_cst_data_analyse(bak_sim_subfolder_path, run_id,
ff_export_subfolder_name)
# Save parameter values and objective values of this simulation to .csv record files
rec_objval_vec = np.append(objval_vec, fitness)
self.func_cst_opt_res_save(project_folder, self.rec_para_file_name, self.rec_objval_file_name,
opt_para_val_vec, self.rec_para_name_lst,
rec_objval_vec, self.rec_obj_name_lst, run_id)
# Save cma object to file for resume later
pycst_ctrl.func_save_ins_to_file(project_folder, self.cma_object_save_file_name, self.optimizer_ins)
# Update run_id
self.run_id += 1
return fitness