A Python 3 implementation of the algorithm for calculating the system and survival signature from a binary decision diagram representation of a system's reliability structure function that was described in the paper "An efficient algorithm for exact computation of system and survival signatures using binary decision diagrams".
This software depends on the following libraries and their dependencies: NumPy (for multidimensional arrays), dd (for Binary Decision Diagrams) and python-igraph (for networks).
You can install them using pip:
pip install requirements.txt
Note: Windows users should follow the instructions here for installing the python-igraph library.
The compute_signatures function implements the algorithm that computes a multidimensional array representing the survival signature (and also the system signature if there is only a single component type) for a BDD where each variable is mapped to a type.
from dd import BDD
from signatures import sig_funs
# Create a simple BDD.
bdd = BDD()
bdd.declare('w','x','y','z')
w = bdd.var('w')
x = bdd.var('x')
y = bdd.var('y')
z = bdd.var('z')
t = bdd.apply('and', w, x)
u = bdd.apply('and', y, z)
v = bdd.apply('or', t, u)
# Create a mapping between the variables and their component types.
var_types = {'w':0, 'x':1, 'y':1, 'z':0}
# Compute signatures for the root node, v, of the BDD.
sig_dim_types, survival_sig, system_sig = sig_funs.compute_signatures(bdd, v, var_types)
# Print the survival signature as a formatted table.
sig_funs.print_survival_sig_table(survival_sig, sig_dim_types)
'''
Type 0 | Type 1 | Probability
0 | 0 | 0.000000
0 | 1 | 0.000000
0 | 2 | 0.000000
1 | 0 | 0.000000
1 | 1 | 0.500000
1 | 2 | 1.000000
2 | 0 | 0.000000
2 | 1 | 1.000000
2 | 2 | 1.000000
'''
# Repeat calculation but with all variables of same component type.
sig_dim_types, survival_sig, system_sig = sig_funs.compute_signatures(bdd, v)
# Print the system signature as a formatted table.
sig_funs.print_system_sig_table(system_sig)
'''
Number of Failures | Probability
1 | 0.000000
2 | 0.666667
3 | 0.333333
4 | 0.000000
'''
# Get the system survival probability from the survival signature.
sig_dim_probabilities = (0.85, 0.98)
sys_survival_prob = sig_funs.eval_prob_from_surv_sig(survival_sig, sig_dim_probabilities)
print("Probability: {:.6f}".format(sys_survival_prob))
'''
Probability: 0.0.972111
'''Some helper functions are included for computing signatures for the connectivity between two terminal nodes in a network with unreliable vertices. These are implemented to be familiar to users of the ReliabilityTheory R package and uses the same library (igraph) and notation.
Note that the helper function for constructing a BDD of a network relies on derivation of cut-sets and is computationally intensive for anything except small networks.
The following example computes and prints the survival signature for the network from Figure 1 of the paper "Generalizing the signature to systems with multiple types of components" (Coolen and Coolen-Maturi, 2012):
from igraph import Graph
from signatures.networks import get_component_types_dict, get_st_bdd_from_cutsets
from signatures.sig_funs import compute_signatures, print_survival_sig_table
# Create the network.
network = Graph.Formula("s -- A, A -- B:C, B -- D:E, C -- D:F, D -- E:F, E -- t, F -- t")
# Set the component types of the vertices.
for v in network.vs:
if v["name"] in ["A", "B", "E"]:
v["compType"] = 1
elif v["name"] in ["C", "D", "F"]:
v["compType"] = 2
# Get the binary decision diagram representing the Boolean structure function for connectivity of the 's' and 't' nodes in
# terms of the vertex states.
bdd, root = get_st_bdd_from_cutsets(network)
# Compute the signature of the binary decision diagram.
component_types = get_component_types_dict(network)
sig_dim_types, survival_sig, system_sig = compute_signatures(bdd, root, component_types)
# Print the survival signature as a formatted table.
print_survival_sig_table(survival_sig, sig_dim_types)
'''
Type 1 | Type 2 | Probability
0 | 0 | 0.000000
0 | 1 | 0.000000
0 | 2 | 0.000000
0 | 3 | 0.000000
1 | 0 | 0.000000
1 | 1 | 0.000000
1 | 2 | 0.111111
1 | 3 | 0.333333
2 | 0 | 0.000000
2 | 1 | 0.000000
2 | 2 | 0.444444
2 | 3 | 0.666667
3 | 0 | 1.000000
3 | 1 | 1.000000
3 | 2 | 1.000000
3 | 3 | 1.000000
'''The trees module contains some functions for computing the BDD for the structure function represented by a fault tree or success tree (complement of a fault tree). AND, OR and n-out-of-m gates are supported.
from signatures.trees import fault_tree_to_bdd
from signatures.sig_funs import compute_signatures, print_survival_sig_table, print_system_sig_table
# Each basic event is specified as a unique name (amongst all basic events and gates).
basic_events = ["BE1", "BE2", "BE3", "BE4"]
'''
Each gate is specified as a (name, logical_op, inputs) tuple, where:
name - unique identifier string amongst all gates and basic events.
logical_op - "*" to signify an AND gate, "+" to signify an OR gate, "(n/m)"
to signify an n-out-of-m gate (where n and m are integers).
inputs - tuple of the names of the child gates and basic events.
'''
top_gate_name = "System Event" # System failure (survives) for fault (success) tree
top_gate = (top_gate_name, "*", ("G1", "G2"))
G1 = ("G1", "+", ("BE1", "BE2"))
G2 = ("G2", "+", ("BE3", "BE4"))
gates = [top_gate, G1, G2]
component_types = {"BE1": 1, "BE2": 2, "BE3": 2, "BE4": 1}
# Convert tree to BDD.
bdd, root = fault_tree_to_bdd(top_gate_name, gates, basic_events)
# bdd, root = success_tree_to_bdd(top_gate_name, gates, basic_events)
# Compute signatures and print as table.
sig_dim_types, survival_sig, system_sig = compute_signatures(bdd, root, component_types)
print_survival_sig_table(survival_sig, sig_dim_types)
'''
Type 1 | Type 2 | Probability
0 | 0 | 0.000000
0 | 1 | 0.000000
0 | 2 | 0.000000
1 | 0 | 0.000000
1 | 1 | 0.500000
1 | 2 | 1.000000
2 | 0 | 0.000000
2 | 1 | 1.000000
2 | 2 | 1.000000
'''To run the unit tests:
python -m unittest
Sean Reed.
This project is licensed under the MIT License - see the LICENSE.md file for details.
Please use the following citation for this algorithm:
Reed, S. "An efficient algorithm for exact computation of system and survival signatures using binary decision diagrams", Reliability Engineering & System Safety, Volume 165, 2017, https://doi.org/10.1016/j.ress.2017.03.036.
Thank you.