I N T E G R A T I O N F O R

hartebeestgrassAI and Robotics

Nov 7, 2013 (3 years and 5 months ago)

57 views


I n t e r n a t i o n a l

C o n f e r e n c e

o n

Environmental Knowledge for Disaster Risk
Management

KNOWLEDGE

AND

DATA

I N T E G R A T I O N

F O R

MODELLI NG

OF

RI SK

Prof
.

J
.

Durgaprasad

Civil

Engineering

Department

Gyan

Ganga

College

of

Technology

Jabalpur,

M
.
P
.

Plan of
Presentation

Knowledge and Data Integration for Modelling of Risk






Problem
Definition



Methodology



Case Study



In
spite of the
advances
in

science
and
technology

»
society
continues to face
new
hazards




New
problems of hazards


with increasing complexity

»
risk analyst
-

unfamiliarity

of the problem




Necessary

to develop
continually

Efficient
methodologies and techniques

»
for
moderating risks

to be within acceptable limits





R i s k

A n a l y s i s


Engineering
Paradox:

(
Waterman 1986)

»
More
competent

the domain
-
experts become,
less
able

they
are
to
describe
their knowledge


»
Domain
-
Experts
may not be able to express clearly
about his
conceptual
or

abstract

state of knowledge



How to
Resolve
: (
Ford
K.M,
Bradshaw 1993)

»
By not
restating a
coherent body

of knowledge that already exists in
the minds of Domain
-
experts


»
Rather, domain
-
experts are
engaged

in a constructive modelling
process, in the context of which formal representations are
newly
created

and

shaped


Domain
-
specific

model of human

expertise
-

Risk analysis



Lack
of

correspondence

between

»
the
basis for human skill
ed performance and

»
its
representation
in communicable models




What is required ?

»
Future
standard methods

and

»
Standard
forms


for
reporting data

that
are suitable for
electronic media
storage



which
facilitates the development of effective
Domain
-
Knowledge (DN) & DSS




Communicable

models



Fragments

of
domain
-
knowledge consists of


Conflicts



G
aps



R
edundancies




Unknown

interdependencies

among parameters




Number
of parameters

to be considered




No
clean
-
data available,
which is free from
inconsistencies

and
missing
information




Contain
inconsistent information




Further
compounded when
multiple experts

provide input to the KB




Complexity

of the Problem


Domain Knowledge




Knowledge
-
Base
(KB)

»

KB gives
Input

to Bayesian Network (BN)
systems


»

Existing BN
systems generally require the parameter
interdependency
information

to be coded as part of the
KB


»
Requiring the developer of Decision Support System (DSS) to specify
them
beforehand



»
Developed
incrementally







Domain
-
Knowledge
for modelling


BN

can be used at any stage of a
risk analysis
,
and


may
substitute both
fault trees

and
event
trees



Complexity




as stated by Haiqin,
2004

»
Building

of
BN considered
the main difficulty

&

»
when
applying to real
-
world problems




as stated by
Ann Devitt et. al, 2006

»
Extremely
difficult
to
build
BN

for complex problems

»
which
has
limited

their
application to real world problems



Bayesian networks (BN)
in

Risk Assessment





Processing

of Knowledge



Fragments
of knowledge
elicited
from the
domain
-
experts



I
nspection for errors of
consistency

and
completeness


»
Consistency
errors include
(i)
redundancy
,
(ii)
conflict

and
(iii)
circularity


»
Completeness
errors include
deadends

(unreachable destination), and
sufficiency

of Knowledge






Graph

theoretic techniques






Case Study
on Windstorm
-
induced Damage


During
the
past 40 years
, engineers have begun to make
increasingly closer examinations of windstorm
-
induced
damage


Researchers
and practitioners around the world have
documented

wind induced damage caused by extreme windstorms

(Chiu et al. 1983;
Dikkers et al. 1971; Eaton and Judge 1975; Mehta et al. 1975; Minor and Mehta 1979; Krishna and
Pande 1975; Walker 1975; Wolde
-
Tensae et al.
1985)


Each
fragment of knowledge
may
be in the
verbal
-
form
that is in the form
of a
relationship


For example
: Verbal
-
format


I
ntensity
of wind speed (p22) is a major
factor,
since an increase in wind speed
increases debris potential (p3) and results in higher intensity of debris hazard
(p2
)


Fragment
of knowledge,
Set { fk1 }
:


fk1 =
{ debris hazard (p2), debris potential (p3), wind speed grade (p22)
}


List
of
12

different
fragments (
fk1 to fk12
)
of
knowledge acquired

Representing
Fragments of Knowledge

fk1

fk1

fk1

Debris
hazard
(p2)

Wind speed
grade (p22)

Debris
potential
(p3)

fk2

fk2

fk2

Debris
exposure
(p1)

Terrain
exposure
(p20)

Debris
potential
(p3)

fk3

fk3

fk3

fk3

fk3

fk3

Internal
pressure due to
damage to glass
shutters caused
by debris (p5)

Internal pressure
due to damage to
overhead doors
caused by wind
(p6)

Internal pressure
due to damage to
sliding doors and
shutters caused
by wind (p7)

Net internal
pressure (p8)

fk4

fk4

fk4

Glass debris
damage
potential (p4)

Percentage
of glass
(p10)

Shutters
(p18)

Relating
and Building
non
-
Directed

Coherent body
of
Domain
-
Experts’ Fragmented
-
Knowledge

fk8

fk8

fk8

fk6

fk6

fk6

fk10

fk10

fk10

fk9

fk9

fk9

fk7

fk7

fk7

fk12

fk12

fk12

fk5

fk5

fk5

fk5

fk5

fk5

fk3

fk3

fk3

fk3

fk3

fk3

fk2

fk2

fk2

fk1

fk1

fk1

fk11

fk11

fk4

fk4

fk4

Terrain exposure
grade (p21)

Wind
speed zone
(p23)

Debris
exposure
(p1)

Debris
hazard
(p2)

Internal pressure
due to damage to
glass shutters
caused by debris
(p5)

Internal pressure
due to damage to
sliding doors and
shutters caused by
wind (p7)

Percentage
of glass
(p10)

Net internal
pressure (p8)

Potential
hazard
(p11)

Roof
covering
grade
(p14)

Roof
covering
(p13)

Internal pressure due
to damage to
overhead doors
caused by wind (p6)

Overhead
doors (p9)

Sliding
doors
(p19)

Roof
damage
grade
(p15)

Wind speed
grade (p22)

Debris
potential (p3)

Roof
geometry
(p16)

Roof
geometry
grade
(p17)

Shutters
(p18)

Glass debris
damage
potential (p4)

Terrain
exposure (p20)

Prescriptive
Code (p12)

fk11

Defining the
Problem
of
Windstorm
-
induced Risk















































Input

Output

Known

Unknown

Processed, and Directed Coherent body
of
Domain
-
Experts’
Knowledge
-

BN

fk11

fk8

fk8

fk6

fk6

fk10

fk10

fk9

fk9

fk7

fk7

fk12

fk12

fk5

fk5

fk5

fk3

fk3

fk3

fk2

fk2

fk1

fk1

fk11

fk4

fk4

Terrain exposure
grade (p21)

Wind
speed zone
(p23)

Debris
exposure
(p1)

Debris
hazard
(p2)

Internal pressure
due to damage to
glass shutters
caused by debris
(p5)

Internal pressure
due to damage to
sliding doors and
shutters caused
by wind (p7)

Percentage
of glass
(p10)

Net internal
pressure (p8)

Potential
hazard
(p11)

Roof
covering
grade
(p14)

Roof
covering
(p13)

Internal pressure
due to damage to
overhead doors
caused by wind (p6)

Overhead
doors (p9)

Sliding
doors
(p19)

Roof
damage
grade
(p15)

Wind speed
grade (p22)

Debris
potential (p3)

Roof
geometry
(p16)

Roof
geometry
grade
(p17)

Shutters
(p18)

Glass debris
damage
potential (p4)

Terrain
exposure (p20)

Prescriptive
Code (p12)

Bayesian Network

for Roof Damage
Risk


Software packages available
for supporting BNs for analysing
risk:

(i)
GeNIE
, (ii) Ergo, (iii) BNIF, (iv) Hugin, (v) Netica, (vi) KI and (vii) Norsys etc.



GeNIE

is selected for implementing the obtained BN as shown in
below:


Thank You