Fault trees structure

Arturo Pérez Mulas arturo_perez at wanadoo.es
Tue Dec 12 01:17:35 CET 2000


An introctury fault-tree, including a very brief introduction:

http://www.riskspectrum.com/docs/methods_2.htm

I have to admit that, at this point I have not been able to find any more
"introductory" pieces. However, I include below a schematic description
of the underlying structure for fault-trees; if anyone is interested and teh
discussion proceeds, I will post a more detailed description of the
'quantification' and operations of fault-trees, along with a description
of what they are really used for, and how (fault tree analysis).

I will try to describe the Fault Tree Structure with my own words: a
fault-tree is a tree-like structure (similar to the binary tree), developed
from a unique top-event or top-node. The main characteristics of a
Fault Tree are the following:

- They have only ONE top node

- Each node can have unlimited inputs (opposed to two allowed for binary
trees)

- Nodes can be of any of the following different types:

a) 'Gates' : are nodes that admit further inputs; there are different types
of gates,
            AND, OR, XOR, n/m, NOT but this is not important at this point,
since
            from the tree structure they are all similar (except type n/m
gates which
            must have exactly 'm' inputs)

b) 'Basic Events': or final nodes, which cannot have any more inputs; for
operation
            purposes, a basic event has a numeric 'probability' between 0
and 1, but I
            leave this for latter.

c) 'House Events': special events, similar to the Basic Events, but they do
not have
            a numerical probability, but a "on/off" status flag.

d) 'Links', or 'link gates', are nodes which actually point to other
fault-tree or other
            node in a different fault tree. They are used to crop large
fault-trees in pieces,
            or in larger structures, when several fault-trees are used, they
may share smaller
            pieces, so 'links' are used not to repeat the same structure in
several places.
            Note that usually a link points to the top event of another
fault tree, but
            this is not mandatory, and in some cases links pointing to
intermediate nodes
            (gates) in other fault tree can be found.

- Considering the previous description, the following 'properties' should be
considered
for the different types of nodes (along with a 'name' and 'description' for
all of them):

a) 'Gates':     - type (AND, OR, XOR...)
                    - inputs

b) 'Basic Events':    - probability (float between 0 to 1)
                                - other probability data (not important at
this point), such as uncertainity distributions

c) 'House Events':    - logical state (boolean switch)

d) 'Links':    - none

I have developed a very simple objects hierarchy for this, but I am not very
satisfied with it (I considered other
issues when creating that, based on the sources from which I will obtain the
fault trees, and the formats to which
I have to send them later). If anyone is finally interested, I will post
some info on the structure and we can discuss it.


I would greatly appreciate some discussion here, since this is the first
time I use OOP to model something!

Arturo.















More information about the Python-list mailing list