diff --git a/kadmos/graph/graph_data.py b/kadmos/graph/graph_data.py
index f3e3121acf95941e4bd97a7dc7e720610bda5904..7b2d2fad0f0a4b21571d93311c7df85e3a4cfd2a 100644
--- a/kadmos/graph/graph_data.py
+++ b/kadmos/graph/graph_data.py
@@ -3044,8 +3044,7 @@ class FundamentalProblemGraph(DataGraph, KeChainMixin):
if node_selection:
nodes_to_partition = list(node_selection)
# For IDF, all functions in the optimizer loop are partitioned except for the objective and constraint functions
- elif 'system_architecture' in self.graph['problem_formulation'] and \
- self.graph['problem_formulation']['system_architecture'] == self.OPTIONS_ARCHITECTURES[2]:
+ elif self.graph['problem_formulation']['mdao_architecture'] == self.OPTIONS_ARCHITECTURES[2]:
# Get post-design-variables, coupled and post-coupling functions
mg_function_ordering = self.get_mg_function_ordering()
@@ -3217,8 +3216,7 @@ class FundamentalProblemGraph(DataGraph, KeChainMixin):
# Update the function order
if node_selection:
function_order = [node for partition in partitions for node in partition]
- elif 'system_architecture' in self.graph['problem_formulation'] and \
- self.graph['problem_formulation']['system_architecture'] == self.OPTIONS_ARCHITECTURES[2]:
+ elif self.graph['problem_formulation']['mdao_architecture'] == self.OPTIONS_ARCHITECTURES[2]:
# noinspection PyUnboundLocalVariable
pre_desvars_order = self.sort_nodes_for_process(mg_function_ordering[self.FUNCTION_ROLES[3]])
partitioned_nodes_order = [node for partition in partitions for node in partition]
@@ -3237,7 +3235,7 @@ class FundamentalProblemGraph(DataGraph, KeChainMixin):
self.graph['problem_formulation']['coupled_functions_groups'] = best_partitions
self.graph['problem_formulation']['local_convergers'] = convergers
- self.graph['problem_formulation']['partition_convergence'] = []
+ self.graph['problem_formulation']['jacobi_convergence'] = []
self.graph['problem_formulation']['sequence_partitions'] = []
if not node_selection:
self.graph['problem_formulation']['function_order'] = function_order
@@ -3411,7 +3409,7 @@ class FundamentalProblemGraph(DataGraph, KeChainMixin):
'partitions']
self.graph['problem_formulation']['local_convergers'] = partition_results[partition_range[idx]][
'local_convergers']
- self.graph['problem_formulation']['partition_convergence'] = []
+ self.graph['problem_formulation']['jacobi_convergence'] = []
self.graph['problem_formulation']['sequence_partitions'] = []
self.graph['problem_formulation']['function_order'] = partition_results[partition_range[idx]]['function_order']
@@ -3435,16 +3433,6 @@ class FundamentalProblemGraph(DataGraph, KeChainMixin):
partitions = self.graph['problem_formulation']['coupled_functions_groups']
coupling_dict = self.get_coupling_dictionary()
- # Select system architecture # todo: remove? System acrhictecture should be given before graph is partitioned
- system_architectures = ['unconverged-MDA', 'converged-MDA', 'MDF', 'IDF', 'unconverged-OPT']
- options = []
- for idx, arch in enumerate(system_architectures):
- options.append([idx, arch])
- printing.print_in_table(options, headers=['Option', 'System architecture'])
- msg = 'Please select system architecture:'
- system_architecture = str(prompting.user_prompt_select_options(*system_architectures, allow_empty=False,
- allow_multi=False, message=msg)[0])
-
# Select which partitions need a local converger
msg = 'Please select which partitions need a local converger:'
while True:
@@ -3492,13 +3480,11 @@ class FundamentalProblemGraph(DataGraph, KeChainMixin):
# todo: remove or improve
print 'local converger:', local_convergers
- print 'system architecture:', system_architecture
print 'jacobi convergence:', jacobi_convergence
print 'sequence partitions:', sequence_partitions
- self.graph['problem_formulation']['system_architecture'] = system_architecture
self.graph['problem_formulation']['local_convergers'] = local_convergers
- self.graph['problem_formulation']['partition_convergence'] = jacobi_convergence
+ self.graph['problem_formulation']['jacobi_convergence'] = jacobi_convergence
self.graph['problem_formulation']['sequence_partitions'] = sequence_partitions
return
@@ -3518,8 +3504,8 @@ class FundamentalProblemGraph(DataGraph, KeChainMixin):
pre_functions = self.graph['problem_formulation']['function_ordering'][self.FUNCTION_ROLES[0]]
mg_function_ordering = dict(self.graph['problem_formulation']['function_ordering'])
- # OPTIONS_ARCHITECTURE: IDF, MDF, unc-OPT, unc-DOE, con-DOE, distr-conv, CO, BLISS-2000
- if mdao_arch in self.OPTIONS_ARCHITECTURES[2:10]:
+ # OPTIONS_ARCHITECTURE: IDF, MDF, unc-OPT, unc-DOE, con-DOE, CO, BLISS-2000
+ if mdao_arch in self.OPTIONS_ARCHITECTURES[2:9]:
del mg_function_ordering[self.FUNCTION_ROLES[0]]
if pre_functions:
target_set = set()
@@ -3541,8 +3527,6 @@ class FundamentalProblemGraph(DataGraph, KeChainMixin):
post_desvars_funcs = []
mg_function_ordering[self.FUNCTION_ROLES[3]] = pre_desvars_funcs
mg_function_ordering[self.FUNCTION_ROLES[4]] = post_desvars_funcs
- if mdao_arch == self.OPTIONS_ARCHITECTURES[2]: # IDF
- mg_function_ordering[self.FUNCTION_ROLES[2]].append(self.CONSCONS_STRING)
return mg_function_ordering
@@ -3879,7 +3863,8 @@ class FundamentalProblemGraph(DataGraph, KeChainMixin):
# TODO: Imco check docstring
# Initiate dictionary
- global_functions = [global_objective_function] + global_cnstrnt_functions
+ global_functions = [global_objective_function] if global_objective_function else []
+ global_functions += global_cnstrnt_functions
system_level_function_dict = dict()
system_level_function_dict[self.FUNCTION_ROLES[1]] = []
system_level_function_dict[self.FUNCTION_ROLES[3]] = []
@@ -3889,8 +3874,12 @@ class FundamentalProblemGraph(DataGraph, KeChainMixin):
# Get and check function groups
if mg_function_ordering is None:
mg_function_ordering = self.get_mg_function_ordering()
- pre_desvars = mg_function_ordering[self.FUNCTION_ROLES[3]]
- post_desvars = mg_function_ordering[self.FUNCTION_ROLES[4]]
+ if self.FUNCTION_ROLES[0] in mg_function_ordering: # TODO: This should be fixed when pre-coupling category is eliminated
+ pre_desvars = mg_function_ordering[self.FUNCTION_ROLES[0]]
+ post_desvars = []
+ else:
+ pre_desvars = mg_function_ordering[self.FUNCTION_ROLES[3]]
+ post_desvars = mg_function_ordering[self.FUNCTION_ROLES[4]]
post_couplings = mg_function_ordering[self.FUNCTION_ROLES[2]]
# Add pre-design variables functions to the dictionary
@@ -3974,7 +3963,10 @@ class FundamentalProblemGraph(DataGraph, KeChainMixin):
# Get and check function groups
if mg_function_ordering is None:
mg_function_ordering = self.get_mg_function_ordering()
- post_desvars = mg_function_ordering[self.FUNCTION_ROLES[4]]
+ if self.FUNCTION_ROLES[0] in mg_function_ordering: # TODO: This should be fixed when pre-coupling category is eliminated
+ post_desvars = []
+ else:
+ post_desvars = mg_function_ordering[self.FUNCTION_ROLES[4]]
post_couplings = mg_function_ordering[self.FUNCTION_ROLES[2]]
additional_post_couplings = []
additional_post_desvars = []
@@ -4133,17 +4125,28 @@ class FundamentalProblemGraph(DataGraph, KeChainMixin):
basic_group_couplings = list(group_couplings)
# Determine objective function based on objective value
- global_objective_function = self.get_sources(objective_node)[0]
+ if objective_node:
+ global_objective_function = self.get_sources(objective_node)[0]
+ else:
+ global_objective_function = None
# TODO: Assert that objective function only has one output
# Determine local and global design variables
- global_des_vars, local_des_vars, des_vars_group_idxs = \
- self.determine_scope_design_variables(des_vars=des_var_nodes)
+ if des_var_nodes:
+ global_des_vars, local_des_vars, des_vars_group_idxs = \
+ self.determine_scope_design_variables(des_vars=des_var_nodes)
+ else:
+ global_des_vars, local_des_vars, des_vars_group_idxs = [], [], dict()
# TODO: assess that each discipline group is dependent on at least one design variable (?)
# Get global and local constraints and their functions
- global_cnstrnt_vars, global_cnstrnt_funcs, local_cnstrnt_vars, local_cnstrnt_funcs, cnstrnt_vars_group_idxs, \
- cnstrnt_funcs_group_idxs = self.determine_scope_constraint_functions(cnstrnt_vars=constraint_nodes)
+ if constraint_nodes:
+ global_cnstrnt_vars, global_cnstrnt_funcs, local_cnstrnt_vars, local_cnstrnt_funcs, \
+ cnstrnt_vars_group_idxs, cnstrnt_funcs_group_idxs = \
+ self.determine_scope_constraint_functions(cnstrnt_vars=constraint_nodes)
+ else:
+ global_cnstrnt_vars, global_cnstrnt_funcs, local_cnstrnt_vars, local_cnstrnt_funcs, \
+ cnstrnt_vars_group_idxs, cnstrnt_funcs_group_idxs = [], [], [], [], [], dict()
# Create dictionary of pre-desvar, post-desvar, and post-coupling functions for the system optimizer
sys_functions_dict = self.get_system_level_functions(global_objective_function, global_cnstrnt_funcs,
@@ -4244,6 +4247,19 @@ class FundamentalProblemGraph(DataGraph, KeChainMixin):
# Set up MDAO data graph
mdg = graph.create_mdg(mg_function_ordering, name=name)
+ # Check for partitions in monolithic architectures
+ if mdao_arch in graph.OPTIONS_ARCHITECTURES[1:4] and 'coupled_functions_groups' in \
+ graph.graph['problem_formulation']:
+ partitions = graph.graph['problem_formulation']['coupled_functions_groups']
+ system_analysis = self._analyze_distributed_system(des_var_nodes, objective_node, constraint_nodes,
+ mg_function_ordering,
+ add_system_functions_to_subsystems=False)
+ distr_function_ordering = system_analysis['functions_dicts']
+ mdg.graph['distr_function_ordering'] = distr_function_ordering
+ sub_func_orderings = distr_function_ordering[1]
+ else:
+ partitions = None
+
# Manipulate data graph
if mdao_arch == graph.OPTIONS_ARCHITECTURES[0]: # unconverged-MDA
if allow_unconverged_couplings:
@@ -4259,32 +4275,52 @@ class FundamentalProblemGraph(DataGraph, KeChainMixin):
# Connect system inputs and outputs to the coordinator
mdg.connect_coordinator()
elif mdao_arch == graph.OPTIONS_ARCHITECTURES[1]: # converged-MDA
- conv = graph.CONVERGER_STRING
- # Connect converger
- mdg.connect_converger(conv, conv_type, coup_functions, True)
+ if partitions:
+ # Connect partitions
+ # noinspection PyUnboundLocalVariable
+ mdg.connect_partitions(mdao_arch, sub_func_orderings, coup_functions)
+ else:
+ sys_conv, sys_conv_label = graph.CONVERGER_STRING, graph.CONVERGER_LABEL
+ # Connect system converger
+ mdg.connect_converger(sys_conv, conv_type, coup_functions, True, label=sys_conv_label)
# Connect QOIs to the coordinator
mdg.connect_qoi_nodes_as_input(qoi_nodes, graph.COORDINATOR_STRING, True)
# Connect remaining system inputs and outputs to the coordinator
mdg.connect_coordinator()
elif mdao_arch == graph.OPTIONS_ARCHITECTURES[2]: # IDF
- opt = graph.OPTIMIZER_STRING
- # Connect optimizer as a converger using the consistency constraint function
- mdg.connect_converger(opt, graph.OPTIONS_ARCHITECTURES[2], coup_functions, True)
- # Connect optimizer w.r.t. design variables, objective, contraints
+ if partitions:
+ sys_opt = graph.SYS_PREFIX + graph.OPTIMIZER_STRING
+ sys_opt_label = graph.SYS_PREFIX + graph.OPTIMIZER_LABEL
+ # Connect partitions
+ # noinspection PyUnboundLocalVariable
+ mdg.connect_partitions(mdao_arch, sub_func_orderings, coup_functions)
+ else:
+ sys_opt, sys_opt_label = graph.OPTIMIZER_STRING, graph.OPTIMIZER_LABEL
+ # Connect optimizer as a converger using the consistency constraint function
+ mdg.connect_converger(sys_opt, graph.OPTIONS_ARCHITECTURES[2], coup_functions, True,
+ label=sys_opt_label, converger_is_optimizer=True)
+ # Connect optimizer w.r.t. design variables, objective, constraints
# noinspection PyUnboundLocalVariable
- mdg.connect_optimizer(opt, des_var_nodes, objective_node, constraint_nodes)
+ mdg.connect_optimizer(sys_opt, des_var_nodes, objective_node, constraint_nodes, label=sys_opt_label)
# Connect QOIs to the coordinator
mdg.connect_qoi_nodes_as_input(qoi_nodes, graph.COORDINATOR_STRING, True)
# Connect remaining system inputs and outputs to the coordinator
mdg.connect_coordinator()
elif mdao_arch == graph.OPTIONS_ARCHITECTURES[3]: # MDF
- opt = graph.OPTIMIZER_STRING
- conv = graph.CONVERGER_STRING
- # Connect converger
- mdg.connect_converger(conv, conv_type, coup_functions, True)
+ if partitions:
+ sys_opt = graph.SYS_PREFIX + graph.OPTIMIZER_STRING
+ sys_opt_label = graph.SYS_PREFIX + graph.OPTIMIZER_LABEL
+ # Connect partitions
+ # noinspection PyUnboundLocalVariable
+ mdg.connect_partitions(mdao_arch, sub_func_orderings, coup_functions)
+ else:
+ sys_opt, sys_opt_label = graph.OPTIMIZER_STRING, graph.OPTIMIZER_LABEL
+ sys_conv, sys_conv_label = graph.CONVERGER_STRING, graph.CONVERGER_LABEL
+ # Connect system converger
+ mdg.connect_converger(sys_conv, conv_type, coup_functions, True, label=sys_conv_label)
# Connect optimizer
# noinspection PyUnboundLocalVariable
- mdg.connect_optimizer(opt, des_var_nodes, objective_node, constraint_nodes)
+ mdg.connect_optimizer(sys_opt, des_var_nodes, objective_node, constraint_nodes, label=sys_opt_label)
# Connect QOIs to the coordinator
mdg.connect_qoi_nodes_as_input(qoi_nodes, graph.COORDINATOR_STRING, True)
# Connect remaining system inputs and outputs to the coordinator
@@ -4331,144 +4367,7 @@ class FundamentalProblemGraph(DataGraph, KeChainMixin):
mdg.connect_doe_block(doe, des_var_nodes, qoi_nodes)
# Connect remaining system inputs and outputs to the coordinator
mdg.connect_coordinator()
- elif mdao_arch == self.OPTIONS_ARCHITECTURES[7]: # distributed-convergence
-
- # Input checks
- assert 'coupled_functions_groups' in graph.graph['problem_formulation'], 'Graph is not partitioned.'
- assert 'system_architecture' in graph.graph['problem_formulation'], 'No architecture selected for ' \
- 'distributed convergence.'
-
- # Load extra variables from fpg
- partitions = graph.graph['problem_formulation']['coupled_functions_groups']
- system_arch = graph.graph['problem_formulation']['system_architecture']
- local_convergers = graph.graph['problem_formulation']['local_convergers']
- parallel_conv = graph.graph['problem_formulation']['partition_convergence']
-
- sys_opt, sys_conv, sub_convs = self.get_architecture_node_ids(self.OPTIONS_ARCHITECTURES[7],
- number_of_groups=len(partitions))
- sys_opt_label, sys_conv_label, sub_convs_labels = self.get_architecture_node_labels(
- self.OPTIONS_ARCHITECTURES[7], number_of_groups=len(partitions))
- system_analysis = self._analyze_distributed_system(des_var_nodes, objective_node, constraint_nodes,
- mg_function_ordering,
- add_system_functions_to_subsystems=False)
- distr_function_ordering = system_analysis['functions_dicts']
- mdg.graph['distr_function_ordering'] = distr_function_ordering
- sub_func_orderings = distr_function_ordering[1]
-
- if system_arch == 'converged-MDA' or system_arch == 'MDF':
- # Connect partitions
- for partition in range(len(partitions)):
- # Get convergence type of partitions (default is Gauss-Seidel, unless indicated otherwise)
- conv_type_partition = graph.OPTIONS_CONVERGERS[0] if partition in parallel_conv else \
- graph.OPTIONS_CONVERGERS[1]
- # Connect disciplines to local converger if present, otherwise system converger
- converger = sub_convs[partition] if partition in local_convergers else sys_conv
- converger_label = sub_convs_labels[partition] if partition in local_convergers else sys_conv_label
- nodes = [node for type in sub_func_orderings[partition] for node in
- sub_func_orderings[partition][type]] if partition not in local_convergers and \
- conv_type_partition == graph.OPTIONS_CONVERGERS[0] else sub_func_orderings[partition]['coupled']
- mdg.connect_converger(converger, conv_type_partition, nodes, True, label=converger_label)
- # Connect remaining couplings to system converger
- mdg.connect_converger(sys_conv, graph.OPTIONS_CONVERGERS[0], coup_functions, True,
- system_converger=True, label=sys_conv_label)
- # Connect optimizer if present
- if system_arch == 'MDF':
- mdg.connect_optimizer(sys_opt, des_var_nodes, objective_node, constraint_nodes, label=sys_opt_label)
- # Connect qoi
- mdg.connect_qoi_nodes_as_input(qoi_nodes, graph.COORDINATOR_STRING, True)
- # Connect coordinator
- mdg.connect_coordinator()
- elif system_arch == 'IDF':
- # Connect partitions
- # TODO: quick-fix to get IDF working
- # Add concisitency constraint block
- assert not self.has_node(self.CONSCONS_STRING), 'Consistency constraint name already in use in FPG.'
- mdg.add_node(self.CONSCONS_STRING,
- label=self.CONSCONS_LABEL,
- category='function',
- architecture_role=self.ARCHITECTURE_ROLES_FUNS[8],
- function_type='consistency')
- distr_function_ordering[0][self.FUNCTION_ROLES[2]].append('Gc')
- mdg.graph['distr_function_ordering'] = distr_function_ordering
- for partition in range(len(partitions)):
- converger = sub_convs[partition] if partition in local_convergers else sys_opt
- converger_label = sub_convs_labels[partition] if partition in local_convergers else sys_opt_label
- # Get convergence type for the partition
- if partition in parallel_conv:
- if converger == sys_opt:
- conv_type_partition = graph.OPTIONS_ARCHITECTURES[2]
- else:
- conv_type_partition = graph.OPTIONS_CONVERGERS[0]
- else:
- conv_type_partition = graph.OPTIONS_CONVERGERS[1]
- converger_is_optimizer = True if converger_label == sys_opt_label else False
- nodes = [node for type in sub_func_orderings[partition] for node in
- sub_func_orderings[partition][type]] if partition not in local_convergers and \
- conv_type_partition in [graph.OPTIONS_CONVERGERS[0], graph.OPTIONS_ARCHITECTURES[2]] else \
- sub_func_orderings[partition]['coupled']
- mdg.connect_converger(converger, conv_type_partition, nodes, True, label=converger_label,
- converger_is_optimizer=converger_is_optimizer)
- # Connect remaining couplings to optimizer as system converger
- mdg.connect_converger(sys_opt, graph.OPTIONS_ARCHITECTURES[2], coup_functions, True,
- system_converger=True, label=sys_opt_label, converger_is_optimizer=True)
- # Connect optimizer
- mdg.connect_optimizer(sys_opt, des_var_nodes, objective_node, constraint_nodes, label=sys_opt_label)
- # Connect qoi
- mdg.connect_qoi_nodes_as_input(qoi_nodes, graph.COORDINATOR_STRING, True)
- # Connect coordinator
- mdg.connect_coordinator()
- elif system_arch == 'unconverged-MDA' or system_arch == 'unconverged-OPT':
- # Connect partitions
- for partition in range(len(partitions)):
- # Get convergence type of partitions (default is Gauss-Seidel, unless indicated otherwise)
- conv_type_partition = graph.OPTIONS_CONVERGERS[0] if partition in parallel_conv else \
- graph.OPTIONS_CONVERGERS[1]
- # Connect disciplines to local converger if present
- if partition in local_convergers:
- mdg.connect_converger(sub_convs[partition], conv_type_partition,
- sub_func_orderings[partition]['coupled'], True,
- label=sub_convs_labels[partition])
- else:
- if conv_type_partition == graph.OPTIONS_CONVERGERS[1]:
- mdg.manipulate_coupling_nodes(sub_func_orderings[partition]['coupled'],
- remove_feedback=True, remove_feedforward=False,
- converger=None, include_couplings_as_final_output=False)
- else:
- nodes = [node for type in sub_func_orderings[partition] for node in
- sub_func_orderings[partition][type]]
- mdg.manipulate_coupling_nodes(nodes, remove_feedback=True, remove_feedforward=True,
- converger=None, include_couplings_as_final_output=False)
- # Connect remaining couplings
- mdg.manipulate_coupling_nodes(coup_functions, remove_feedback=True, remove_feedforward=True,
- converger=None, include_couplings_as_final_output=False,
- system_converger=True)
- # Connect optimizer if present
- if system_arch == 'unconverged-OPT':
- mdg.connect_optimizer(sys_opt, des_var_nodes, objective_node, constraint_nodes, label=sys_opt_label)
- # Connect qoi
- mdg.connect_qoi_nodes_as_input(qoi_nodes, graph.COORDINATOR_STRING, True)
- # Connect coordinator
- mdg.connect_coordinator()
-
- # Resolve problematic variables # todo
- prob_var = mdg.find_all_nodes(subcategory='all splittable variables')
- for var in prob_var:
- sources = mdg.get_sources(var)
- targets = mdg.get_targets(var)
- function_order = []
- for converger in local_convergers:
- if sub_convs[converger] in sources:
- function_order.extend([sub_convs[converger]])
- function_order.extend([target for target in targets if target in partitions[converger]])
- if sys_conv in sources:
- function_order.extend([sys_conv])
- function_order.extend([target for target in targets if target not in function_order])
- if sys_opt in sources:
- function_order.extend([sys_opt])
- function_order.extend([target for target in targets if target not in function_order])
- mdg.split_variables(var, function_order=function_order)
-
- elif mdao_arch == graph.OPTIONS_ARCHITECTURES[8]: # CO
+ elif mdao_arch == graph.OPTIONS_ARCHITECTURES[7]: # CO
coupled_functions_groups = graph.graph['problem_formulation']['coupled_functions_groups']
n_groups = len(coupled_functions_groups)
sys_opt, sub_opts = self.get_architecture_node_ids('CO', number_of_groups=n_groups)
@@ -4543,7 +4442,7 @@ class FundamentalProblemGraph(DataGraph, KeChainMixin):
# Write function_ordering to the graph
mdg.graph['distr_function_ordering'] = sa['functions_dicts']
- elif mdao_arch == graph.OPTIONS_ARCHITECTURES[9]: # BLISS-2000
+ elif mdao_arch == graph.OPTIONS_ARCHITECTURES[8]: # BLISS-2000
coupled_functions_groups = graph.graph['problem_formulation']['coupled_functions_groups']
n_groups = len(coupled_functions_groups)
sys_opt, sys_conv, sys_sms, sub_smbds, sub_does, sub_opts, sub_smbs = \
@@ -5057,7 +4956,7 @@ class MdaoDataGraph(DataGraph, MdaoMixin):
for func in functions:
self.nodes[func]['architecture_role'] = 'pre-iterator analysis'
# Add optimizer / DOE
- if mdao_arch in self.OPTIONS_ARCHITECTURES[2:5]: # IDF, MDF, unc-OPT
+ if mdao_arch == self.OPTIONS_ARCHITECTURES[4]: # unc-OPT
assert not self.has_node(self.OPTIMIZER_STRING), 'Optimizer name already in use in FPG.'
self.add_node(self.OPTIMIZER_STRING,
category='function',
@@ -5084,8 +4983,7 @@ class MdaoDataGraph(DataGraph, MdaoMixin):
self.nodes[func]['architecture_role'] = self.ARCHITECTURE_ROLES_FUNS[6]
# Converger required
- if mdao_arch in [self.OPTIONS_ARCHITECTURES[1]] + [self.OPTIONS_ARCHITECTURES[3]] + \
- [self.OPTIONS_ARCHITECTURES[6]]: # con-MDA, MDF, con-DOE
+ if mdao_arch == self.OPTIONS_ARCHITECTURES[6]: # con-DOE
# Add converger
assert not self.has_node(self.CONVERGER_STRING), 'Converger name already in use in FPG.'
self.add_node(self.CONVERGER_STRING,
@@ -5102,15 +5000,7 @@ class MdaoDataGraph(DataGraph, MdaoMixin):
# Add post-coupling functions
for func in mg_function_ordering[self.FUNCTION_ROLES[2]]:
- if func != self.CONSCONS_STRING:
- self.nodes[func]['architecture_role'] = self.ARCHITECTURE_ROLES_FUNS[8]
- else:
- assert not self.has_node(self.CONSCONS_STRING), 'Consistency constraint name already in use in FPG.'
- self.add_node(self.CONSCONS_STRING,
- label=self.CONSCONS_LABEL,
- category='function',
- architecture_role=self.ARCHITECTURE_ROLES_FUNS[8],
- function_type='consistency')
+ self.nodes[func]['architecture_role'] = self.ARCHITECTURE_ROLES_FUNS[8]
return
@@ -5997,6 +5887,74 @@ class MdaoDataGraph(DataGraph, MdaoMixin):
return
+ def connect_partitions(self, mdao_arch, sub_func_orderings, coup_functions):
+ """Method to connect partitions in a data graph of a monolithic architecture
+ """
+
+ partitions = self.graph['problem_formulation']['coupled_functions_groups']
+ local_convergers = self.graph['problem_formulation']['local_convergers']
+ jacobi_conv = self.graph['problem_formulation']['jacobi_convergence']
+
+ sys_opt, sys_conv, sub_convs = self.get_architecture_node_ids(mdao_arch, number_of_groups=len(partitions))
+ sys_opt_label, sys_conv_label, sub_convs_labels = self.get_architecture_node_labels(
+ mdao_arch, number_of_groups=len(partitions))
+
+ # Connect partitions
+ for partition in range(len(partitions)):
+ # Connect nodes to local converger if present, otherwise system converger or optimizer
+ if partition in local_convergers:
+ conv, conv_label = sub_convs[partition], sub_convs_labels[partition]
+ elif mdao_arch == self.OPTIONS_ARCHITECTURES[2]:
+ conv, conv_label = sys_opt, sys_opt_label
+ else:
+ conv, conv_label = sys_conv, sys_conv_label
+ # Get convergence type
+ if partition in jacobi_conv:
+ conv_type = self.OPTIONS_ARCHITECTURES[2] if conv == sys_opt else self.OPTIONS_CONVERGERS[0]
+ else:
+ conv_type = self.OPTIONS_CONVERGERS[1]
+ # Check whether optimizer needs to be connected as converger
+ converger_is_optimizer = True if conv_label == sys_opt_label else False
+ # Get nodes in partition
+ if partition not in local_convergers and conv_type in [self.OPTIONS_CONVERGERS[0],
+ self.OPTIONS_ARCHITECTURES[2]]:
+ nodes = [node for func_role in sub_func_orderings[partition] for node in
+ sub_func_orderings[partition][func_role]]
+ else:
+ nodes = sub_func_orderings[partition]['coupled']
+ # Connect nodes to (sub)system converger or system optimizer
+ self.connect_converger(conv, conv_type, nodes, True, label=conv_label,
+ converger_is_optimizer=converger_is_optimizer)
+
+ # Connect remaining couplings to the system converger or optimizer
+ if mdao_arch == self.OPTIONS_ARCHITECTURES[2]:
+ conv, conv_label, conv_type = sys_opt, sys_opt_label, self.OPTIONS_ARCHITECTURES[2]
+ else:
+ conv, conv_label, conv_type = sys_conv, sys_conv_label, self.OPTIONS_CONVERGERS[0]
+ converger_is_optimizer = True if conv == sys_opt else False
+ self.connect_converger(conv, conv_type, coup_functions, True, system_converger=True, label=conv_label,
+ converger_is_optimizer=converger_is_optimizer)
+
+ # Resolve problematic variables
+ prob_var = self.find_all_nodes(subcategory='all splittable variables')
+ for var in prob_var:
+ sources = self.get_sources(var)
+ targets = self.get_targets(var)
+ function_order = []
+ for converger in local_convergers:
+ if sub_convs[converger] in sources:
+ function_order.extend([sub_convs[converger]])
+ function_order.extend([target for target in targets if target in partitions[converger]])
+ if sys_conv in sources:
+ function_order.extend([sys_conv])
+ function_order.extend([target for target in targets if target not in function_order])
+ if sys_opt in sources:
+ function_order.extend([sys_opt])
+ function_order.extend([target for target in targets if target not in function_order])
+ self.split_variables(var, function_order=function_order)
+
+ return
+
def manipulate_coupling_nodes(self, func_order, remove_feedback, remove_feedforward, converger=None,
include_couplings_as_final_output=False, system_converger=False):
"""Method to manipulate the coupling nodes in a data graph in order to remove unwanted feedback/feedforward.
@@ -6076,6 +6034,17 @@ class MdaoDataGraph(DataGraph, MdaoMixin):
self.copy_edge((coupling[2], coupling[1]), (coupling_copy_node, coupling[1]))
# Create equation label
equation_label = coupling[2].split('/')[-1]
+ # Add consistency constraint function to the graph (if it's not already there)
+ if not self.has_node(self.CONSCONS_STRING):
+ self.add_node(self.CONSCONS_STRING,
+ label=self.CONSCONS_LABEL,
+ category='function',
+ architecture_role=self.ARCHITECTURE_ROLES_FUNS[8],
+ function_type='consistency')
+ if 'distr_function_ordering' in self.graph:
+ self.graph['distr_function_ordering'][0][self.FUNCTION_ROLES[2]].append(self.CONSCONS_STRING)
+ if 'mg_function_ordering' in self.graph:
+ self.graph['mg_function_ordering'][self.FUNCTION_ROLES[2]].append(self.CONSCONS_STRING)
# Connect original and copied coupling node to the consistency constraint function
self.add_edge(coupling[2], self.CONSCONS_STRING, equation_label=equation_label)
self.add_edge(coupling_copy_node, self.CONSCONS_STRING, equation_label=equation_label+'c')
@@ -6160,6 +6129,17 @@ class MdaoDataGraph(DataGraph, MdaoMixin):
coup_functions = mdg.sort_nodes_for_process(mg_function_ordering[mdg.FUNCTION_ROLES[1]])
post_functions = mg_function_ordering[mdg.FUNCTION_ROLES[2]]
+ # Check for partitions in monolithic architectures
+ if mdao_arch in mdg.OPTIONS_ARCHITECTURES[1:4] and 'coupled_functions_groups' in \
+ mdg.graph['problem_formulation']:
+ partitions = mdg.graph['problem_formulation']['coupled_functions_groups']
+ distr_function_ordering = mdg.graph['distr_function_ordering']
+ pre_desvars_funcs = distr_function_ordering[0][mdg.FUNCTION_ROLES[3]]
+ post_desvars_funcs = distr_function_ordering[0][mdg.FUNCTION_ROLES[4]]
+ post_functions = distr_function_ordering[0][mdg.FUNCTION_ROLES[2]]
+ else:
+ partitions = None
+
# Set up MDAO process graph
mpg = mdg.create_mpg(name=name)
@@ -6171,35 +6151,53 @@ class MdaoDataGraph(DataGraph, MdaoMixin):
sequence = [coor] + pre_functions + coup_functions + post_functions
mpg.add_process(sequence, 0, mdg, end_in_iterative_node=coor)
elif mdao_arch == mdg.OPTIONS_ARCHITECTURES[1]: # converged-MDA
- conv = mdg.CONVERGER_STRING
- sequence = [coor] + pre_functions + [conv]
- mpg.add_process(sequence, 0, mdg)
- sequence2 = [conv] + coup_functions
- mpg.add_process(sequence2, mpg.nodes[sequence[-1]]['process_step'], mdg, end_in_iterative_node=conv)
+ _, sys_conv, _ = self.get_architecture_node_ids(mdao_arch, number_of_groups=len(partitions)) if \
+ partitions else ([], mdg.CONVERGER_STRING, [])
+ sequence1 = [coor] + pre_functions + [sys_conv]
+ mpg.add_process(sequence1, 0, mdg)
+ if partitions:
+ mpg.add_process_partitions([sys_conv], partitions, [], mpg.nodes[sequence1[-1]]['process_step'], mdg,
+ end_in_iterative_node=sys_conv)
+ else:
+ sequence2 = [sys_conv] + coup_functions
+ mpg.add_process(sequence2, mpg.nodes[sequence1[-1]]['process_step'], mdg,
+ end_in_iterative_node=sys_conv)
if post_functions:
- sequence3 = [conv] + post_functions
- mpg.add_process(sequence3, mpg.nodes[conv]['converger_step'], mdg, end_in_iterative_node=coor)
+ sequence3 = [sys_conv] + post_functions
+ mpg.add_process(sequence3, mpg.nodes[sys_conv]['converger_step'], mdg, end_in_iterative_node=coor)
else:
- mpg.connect_nested_iterators(coor, conv)
+ mpg.connect_nested_iterators(coor, sys_conv)
elif mdao_arch == mdg.OPTIONS_ARCHITECTURES[2]: # IDF
- opt = mdg.OPTIMIZER_STRING
- sequence1 = [coor] + pre_desvars_funcs + [opt]
+ sys_opt, _, _ = self.get_architecture_node_ids(mdao_arch, number_of_groups=len(partitions)) if \
+ partitions else (mdg.OPTIMIZER_STRING, [], [])
+ sequence1 = [coor] + pre_desvars_funcs + [sys_opt]
mpg.add_process(sequence1, 0, mdg)
- sequence2 = [opt] + post_desvars_funcs + coup_functions + post_functions
- mpg.add_process(sequence2, mpg.nodes[sequence1[-1]]['process_step'], mdg, end_in_iterative_node=opt)
- mpg.connect_nested_iterators(coor, opt)
+ if partitions:
+ sequence2 = [sys_opt] + post_desvars_funcs
+ sequence3 = post_functions
+ mpg.add_process_partitions(sequence2, partitions, sequence3, mpg.nodes[sequence1[-1]]['process_step'],
+ mdg, end_in_iterative_node=sys_opt)
+ else:
+ sequence2 = [sys_opt] + post_desvars_funcs + coup_functions + post_functions
+ mpg.add_process(sequence2, mpg.nodes[sequence1[-1]]['process_step'], mdg, end_in_iterative_node=sys_opt)
+ mpg.connect_nested_iterators(coor, sys_opt)
elif mdao_arch == mdg.OPTIONS_ARCHITECTURES[3]: # MDF
- opt = mdg.OPTIMIZER_STRING
- conv = mdg.CONVERGER_STRING
- sequence1 = [coor] + pre_desvars_funcs + [opt]
+ sys_opt, sys_conv, _ = self.get_architecture_node_ids(mdao_arch, number_of_groups=len(partitions)) if \
+ partitions else (mdg.OPTIMIZER_STRING, mdg.CONVERGER_STRING, [])
+ sequence1 = [coor] + pre_desvars_funcs + [sys_opt]
mpg.add_process(sequence1, 0, mdg)
- sequence2 = [opt] + post_desvars_funcs + [conv]
+ sequence2 = [sys_opt] + post_desvars_funcs + [sys_conv]
mpg.add_process(sequence2, mpg.nodes[sequence1[-1]]['process_step'], mdg)
- sequence3 = [conv] + coup_functions
- mpg.add_process(sequence3, mpg.nodes[sequence2[-1]]['process_step'], mdg, end_in_iterative_node=conv)
- sequence4 = [conv] + post_functions
- mpg.add_process(sequence4, mpg.nodes[conv]['converger_step'], mdg, end_in_iterative_node=opt)
- mpg.connect_nested_iterators(coor, opt)
+ if partitions:
+ mpg.add_process_partitions([sys_conv], partitions, [], mpg.nodes[sequence2[-1]]['process_step'], mdg,
+ end_in_iterative_node=sys_conv)
+ else:
+ sequence3 = [sys_conv] + coup_functions
+ mpg.add_process(sequence3, mpg.nodes[sequence2[-1]]['process_step'], mdg,
+ end_in_iterative_node=sys_conv)
+ sequence4 = [sys_conv] + post_functions
+ mpg.add_process(sequence4, mpg.nodes[sys_conv]['converger_step'], mdg, end_in_iterative_node=sys_opt)
+ mpg.connect_nested_iterators(coor, sys_opt)
elif mdao_arch == mdg.OPTIONS_ARCHITECTURES[4]: # unconverged-OPT
opt = mdg.OPTIMIZER_STRING
sequence1 = [coor] + pre_desvars_funcs + [opt]
@@ -6229,62 +6227,7 @@ class MdaoDataGraph(DataGraph, MdaoMixin):
else:
mpg.connect_nested_iterators(doe, conv)
mpg.connect_nested_iterators(coor, doe)
- elif mdao_arch == mdg.OPTIONS_ARCHITECTURES[7]: # distributed-convergence
-
- # Input checks
- assert 'coupled_functions_groups' in mdg.graph['problem_formulation'], 'Graph is not partitioned.'
- assert 'system_architecture' in mdg.graph['problem_formulation'], 'No architecture selected for ' \
- 'distributed convergence.'
-
- # Load extra variables from fpg
- partitions = mdg.graph['problem_formulation']['coupled_functions_groups']
- system_arch = mdg.graph['problem_formulation']['system_architecture']
- distr_function_ordering = mdg.graph['distr_function_ordering']
- system_pre_desvars_funcs = distr_function_ordering[0][mdg.FUNCTION_ROLES[3]]
- system_post_desvars_funcs = distr_function_ordering[0][mdg.FUNCTION_ROLES[4]]
- system_post_functions = distr_function_ordering[0][mdg.FUNCTION_ROLES[2]]
-
- sys_opt, sys_conv, _ = self.get_architecture_node_ids(self.OPTIONS_ARCHITECTURES[7],
- number_of_groups=len(partitions))
-
- if system_arch == 'unconverged-MDA':
- sequence1 = [coor] + system_pre_desvars_funcs + system_post_desvars_funcs
- sequence2 = system_post_functions
- mpg.add_process_partitions(sequence1, partitions, sequence2, 0, mdg, end_in_iterative_node=coor)
- elif system_arch == 'converged-MDA':
- sequence1 = [coor] + system_pre_desvars_funcs + system_post_desvars_funcs + [sys_conv]
- mpg.add_process(sequence1, 0, mdg)
- mpg.add_process_partitions([sys_conv], partitions, [], mpg.nodes[sys_conv]['process_step'], mdg,
- end_in_iterative_node=sys_conv)
- sequence2 = [sys_conv] + system_post_functions
- mpg.add_process(sequence2, mpg.nodes[sys_conv]['converger_step'], mdg, end_in_iterative_node=coor)
- elif system_arch == 'IDF':
- sequence1 = [coor] + system_pre_desvars_funcs + [sys_opt]
- mpg.add_process(sequence1, 0, mdg)
- sequence2 = [sys_opt] + system_post_desvars_funcs
- sequence3 = system_post_functions
- mpg.add_process_partitions(sequence2, partitions, sequence3, mpg.nodes[sys_opt]['process_step'], mdg,
- end_in_iterative_node=sys_opt)
- mpg.connect_nested_iterators(coor, sys_opt)
- elif system_arch == 'MDF':
- sequence1 = [coor] + system_pre_desvars_funcs + [sys_opt]
- mpg.add_process(sequence1, 0, mdg)
- sequence2 = [sys_opt] + system_post_desvars_funcs + [sys_conv]
- mpg.add_process(sequence2, mpg.nodes[sys_opt]['process_step'], mdg)
- mpg.add_process_partitions([sys_conv], partitions, [], mpg.nodes[sys_conv]['process_step'], mdg,
- end_in_iterative_node=sys_conv)
- sequence3 = [sys_conv] + system_post_functions
- mpg.add_process(sequence3, mpg.nodes[sys_conv]['converger_step'], mdg, end_in_iterative_node=sys_opt)
- mpg.connect_nested_iterators(coor, sys_opt)
- elif system_arch == 'unconverged-OPT':
- sequence1 = [coor] + system_pre_desvars_funcs + [sys_opt]
- mpg.add_process(sequence1, 0, mdg)
- sequence2 = [sys_opt] + system_post_desvars_funcs
- sequence3 = system_post_functions
- mpg.add_process_partitions(sequence2, partitions, sequence3, mpg.nodes[sys_opt]['process_step'], mdg,
- end_in_iterative_node=sys_opt)
- mpg.connect_nested_iterators(coor, sys_opt)
- elif mdao_arch == mdg.OPTIONS_ARCHITECTURES[8]: # CO
+ elif mdao_arch == mdg.OPTIONS_ARCHITECTURES[7]: # CO
distr_function_ordering = mdg.graph['distr_function_ordering']
n_groups = len(distr_function_ordering[1])
sys_opt, sub_opts = self.get_architecture_node_ids(mdao_arch, number_of_groups=n_groups)
@@ -6301,7 +6244,7 @@ class MdaoDataGraph(DataGraph, MdaoMixin):
end_in_iterative_node=sub_opts[idx])
mpg.connect_nested_iterators(sys_opt, sub_opts[idx])
mpg.connect_nested_iterators(coor, sys_opt)
- elif mdao_arch == mdg.OPTIONS_ARCHITECTURES[9]: # BLISS-2000
+ elif mdao_arch == mdg.OPTIONS_ARCHITECTURES[8]: # BLISS-2000
distr_function_ordering = mdg.graph['distr_function_ordering']
n_groups = len(distr_function_ordering[1])
sys_opt, sys_conv, _, sub_smbds, sub_does, sub_opts, sub_smbs = \
diff --git a/kadmos/graph/graph_kadmos.py b/kadmos/graph/graph_kadmos.py
index 7461448f24712a5c54e2d40a7078f1d76e627e50..1b786868d2a885aaa8af9fd8c77062762eca5b5b 100644
--- a/kadmos/graph/graph_kadmos.py
+++ b/kadmos/graph/graph_kadmos.py
@@ -58,9 +58,8 @@ class KadmosGraph(nx.DiGraph, EquationMixin, VistomsMixin):
'unconverged-OPT', # 4
'unconverged-DOE', # 5
'converged-DOE', # 6
- 'distributed-convergence', # 7
- 'CO', # 8
- 'BLISS-2000'] # 9
+ 'CO', # 7
+ 'BLISS-2000'] # 8
OPTIONS_DOE_METHODS = ['Full factorial design', # 0
'Latin hypercube design', # 1
'Monte Carlo design', # 2
@@ -811,8 +810,8 @@ class KadmosGraph(nx.DiGraph, EquationMixin, VistomsMixin):
# Add analysis blocks
for idx in range(0, graph_mpg.number_of_nodes()):
node_list = graph_mpg.find_all_nodes(attr_cond=['diagonal_position', '==', idx])
- assert len(node_list) == 1, 'Somehow, a unique diagonal position %d could not be found in the PSG' \
- ' data graph.' % idx
+ assert len(node_list) == 1, 'Somehow, a unique diagonal position {} could not be found in the MDG.'\
+ .format(idx)
node = node_list[0]
try:
role_index = self.ARCHITECTURE_ROLES_FUNS.index(graph_mpg.nodes[node]['architecture_role'])
@@ -3103,7 +3102,7 @@ class KadmosGraph(nx.DiGraph, EquationMixin, VistomsMixin):
:rtype: tuple
"""
- if mdao_architecture == self.OPTIONS_ARCHITECTURES[7]: # Distributed-convergence
+ if mdao_architecture in self.OPTIONS_ARCHITECTURES[1:4]: # Distributed convergence of monolithic architectures
assert number_of_groups is not None, \
'number_of_groups should be specified for this architecture ({}).'.format(mdao_architecture)
assert number_of_groups > 0, 'number_of_groups should be larger than 0.'
@@ -3112,7 +3111,7 @@ class KadmosGraph(nx.DiGraph, EquationMixin, VistomsMixin):
sub_convs = ['{}{}{}{}'.format(self.SUBSYS_PREFIX, self.CONVERGER_STRING, self.SUBSYS_SUFFIX, item) for
item in range(number_of_groups)]
return sys_opt, sys_conv, sub_convs
- elif mdao_architecture == self.OPTIONS_ARCHITECTURES[8]: # CO
+ elif mdao_architecture == self.OPTIONS_ARCHITECTURES[7]: # CO
assert number_of_groups is not None, \
'number_of_groups should be specified for this architecture ({}).'.format(mdao_architecture)
assert number_of_groups > 0, 'number_of_groups should be larger than 0.'
@@ -3120,7 +3119,7 @@ class KadmosGraph(nx.DiGraph, EquationMixin, VistomsMixin):
sub_opts = ['{}{}{}{}'.format(self.SUBSYS_PREFIX, self.OPTIMIZER_STRING, self.SUBSYS_SUFFIX, item) for
item in range(number_of_groups)]
return sys_opt, sub_opts
- elif mdao_architecture == self.OPTIONS_ARCHITECTURES[9]: # BLISS-2000
+ elif mdao_architecture == self.OPTIONS_ARCHITECTURES[8]: # BLISS-2000
assert number_of_groups is not None, \
'number_of_groups should be specified for this architecture ({}).'.format(mdao_architecture)
assert number_of_groups > 0, 'number_of_groups should be larger than 0.'
@@ -3149,7 +3148,7 @@ class KadmosGraph(nx.DiGraph, EquationMixin, VistomsMixin):
:rtype: tuple
"""
- if mdao_architecture == self.OPTIONS_ARCHITECTURES[7]: # Distributed-convergence
+ if mdao_architecture in self.OPTIONS_ARCHITECTURES[1:4]: # Distributed convergence of monolithic architectures
assert number_of_groups is not None, \
'number_of_groups should be specified for this architecture ({}).'.format(mdao_architecture)
assert number_of_groups > 0, 'number_of_groups should be larger than 0.'
@@ -3158,7 +3157,7 @@ class KadmosGraph(nx.DiGraph, EquationMixin, VistomsMixin):
sub_convs_labels = ['{}{}{}{}'.format(self.SUBSYS_PREFIX, self.CONVERGER_LABEL, self.SUBSYS_SUFFIX, item)
for item in range(number_of_groups)]
return sys_opt_label, sys_conv_label, sub_convs_labels
- elif mdao_architecture == self.OPTIONS_ARCHITECTURES[8]: # CO
+ elif mdao_architecture == self.OPTIONS_ARCHITECTURES[7]: # CO
assert number_of_groups is not None, \
'number_of_groups should be specified for this architecture ({}).'.format(mdao_architecture)
assert number_of_groups > 0, 'number_of_groups should be larger than 0.'
@@ -3166,7 +3165,7 @@ class KadmosGraph(nx.DiGraph, EquationMixin, VistomsMixin):
sub_opts_labels = ['{}{}{}{}'.format(self.SUBSYS_PREFIX, self.OPTIMIZER_LABEL, self.SUBSYS_SUFFIX, item)
for item in range(number_of_groups)]
return sys_opt_label, sub_opts_labels
- elif mdao_architecture == self.OPTIONS_ARCHITECTURES[9]: # BLISS-2000
+ elif mdao_architecture == self.OPTIONS_ARCHITECTURES[8]: # BLISS-2000
assert number_of_groups is not None, \
'number_of_groups should be specified for this architecture ({}).'.format(mdao_architecture)
assert number_of_groups > 0, 'number_of_groups should be larger than 0.'
diff --git a/kadmos/graph/graph_process.py b/kadmos/graph/graph_process.py
index 31d36a7237def1fd83d868912f7a70cb6cc7f33e..e892d984a29a81f7b9a8e3848d0c7792cfe46a65 100644
--- a/kadmos/graph/graph_process.py
+++ b/kadmos/graph/graph_process.py
@@ -275,12 +275,10 @@ class MdaoProcessGraph(ProcessGraph):
bliss2000 = False
co = True
distr_conv = False
- elif mdao_architecture == 'distributed-convergence':
+ else:
bliss2000 = False
co = False
distr_conv = True
- else:
- raise NotImplementedError('Invalid MDAO architecture {} found.'.format(mdao_architecture))
mg_function_ordering = self.graph['distr_function_ordering']
syslevel_ordering = mg_function_ordering[0]
subsyslevel_orderings = mg_function_ordering[1]
@@ -329,8 +327,8 @@ class MdaoProcessGraph(ProcessGraph):
attr_cond=['architecture_role', '==', self.ARCHITECTURE_ROLES_FUNS[2]]) # converger
if len(convs) >= 1:
sys_conv = [item for item in convs if self.SYS_PREFIX in item]
- if distr_conv and self.graph['problem_formulation']['system_architecture'] not in \
- self.OPTIONS_ARCHITECTURES[1] + self.OPTIONS_ARCHITECTURES[3]:
+ if distr_conv and mdao_architecture not in self.OPTIONS_ARCHITECTURES[1] + \
+ self.OPTIONS_ARCHITECTURES[3]:
assert len(sys_conv) == 0, '{} system convergers found, none expected'.format(len(sys_conv))
else:
assert len(sys_conv) == 1, '{} system convergers found, one expected.'.format(len(sys_conv))
@@ -572,8 +570,8 @@ class MdaoProcessGraph(ProcessGraph):
local_convergers = self.graph['problem_formulation']['local_convergers']
sequences_partitions = self.graph['problem_formulation']['sequence_partitions']
sublevel_function_orderings = mdg.graph['distr_function_ordering'][1]
- _, _, sub_convs = self.get_architecture_node_ids(self.OPTIONS_ARCHITECTURES[7],
- number_of_groups=len(partitions))
+ mdao_arch = mdg.graph['problem_formulation']['mdao_architecture']
+ _, _, sub_convs = self.get_architecture_node_ids(mdao_arch, number_of_groups=len(partitions))
# Add process first sequence
self.add_process(previous_sequence, process_step, mdg)
@@ -803,10 +801,11 @@ class MdaoProcessGraph(ProcessGraph):
process_list = []
for step in range(max_step + 1):
process_step_nodes = self.find_all_nodes(attr_cond=['process_step', '==', step])
+ updated_process_step_nodes = copy.deepcopy(process_step_nodes)
for process_step_node in process_step_nodes:
if self.nodes[process_step_node]['architecture_role'] in self.ARCHITECTURE_ROLES_FUNS[:4]:
- process_step_nodes.remove(process_step_node)
- process_list.extend(process_step_nodes)
+ updated_process_step_nodes.remove(process_step_node)
+ process_list.extend(updated_process_step_nodes)
return process_list
def get_process_hierarchy(self):
diff --git a/kadmos/utilities/general.py b/kadmos/utilities/general.py
index 2475509188e8ad6a3a927cfefd9755467ad57342..7c98f710d6caaabc21df8c35536dbaf447155332 100644
--- a/kadmos/utilities/general.py
+++ b/kadmos/utilities/general.py
@@ -157,9 +157,8 @@ def get_mdao_setup(mdao_setup):
'unconverged-DOE-J', # 13
'converged-DOE-GS', # 14
'converged-DOE-J', # 15
- 'distributed-convergence', # 16
- 'CO', # 17
- 'BLISS-2000'] # 18
+ 'CO', # 16
+ 'BLISS-2000'] # 17
if mdao_setup.endswith(('-FF', '-MC', '-LH', '-CT')):
mdao_setup = mdao_setup[:-3]
@@ -228,14 +227,10 @@ def get_mdao_setup(mdao_setup):
mda_type = 'Jacobi'
allow_unconverged_couplings = False
elif mdao_setup == mdao_defintions[16]:
- mdo_architecture = 'distributed-convergence'
- mda_type = None
- allow_unconverged_couplings = False
- elif mdao_setup == mdao_defintions[17]:
mdo_architecture = 'CO'
mda_type = None
allow_unconverged_couplings = False
- elif mdao_setup == mdao_defintions[18]:
+ elif mdao_setup == mdao_defintions[17]:
mdo_architecture = 'BLISS-2000'
mda_type = None
allow_unconverged_couplings = False