GENERIC FIRE RISK ASSESSMENT FOR

(1)

G

ENERIC

F

IRE

R

ISK

A

SSESSMENT

F

OR

RESIDENTIAL AND INDUSTRIAL

B

UILDINGS

Malta,

13.

April

2012

Gianluca

De

Sanctis

(2)

Design Format of the Codes

Prescriptive Performance-Based

Safety Level

Building

Design Format of the Codes

Safety Level

Human Losses

Financial Losses Safety

Investments Expected Human Portfolio Losses Expected Financial Portfolio Losses Portfolio Safety Investments

Design Format of the Codes

Prescriptive Performance-Based Safety Level Expected Human Portfolio Losses Expected Financial Portfolio Losses Portfolio Safety Investments

(3)

G

ENERIC

R

ISK

A

SSESSMENT

Requirements

for

a

generic

risk

model



Applicable

for

most

buildings

in

a

portfolio



Quantitative

assessment

of

the

risk



Physical

effects

of

the

required

fire

safety

measures



Unbiased

assessment

of

the

portfolio

risk

Institute of Structural Engineering – ETH Zurich

Generic

System

Representation

(4)

R

ISK

A

NALYSIS

 Engineering models are able to quantify the losses based on the risk

indicators and physical or empirical understanding of the problem

 Probabilistic engineering models allow to consider the uncertainties

of the risk indicators consistently

 Due to simplification and approximations, engineering models will

never predict the real behaviour of a system exactly. The resulting

loss estimation will be biased.

 This lack of fit could be considered by model parameters

M

odel

Pa

rame

(5)

C

OMBINING AN ENGINEERING APPROACH WITH

D

ATA

e = evidence

e

 A calibration of the model with observed data leads to an unbiased estimation of

the expected losses on portfolio level

 The aim is to quantify the model parameters through the available data

 Methods:

– Maximum Likelihood Method  select model parameters such that the

observed values obtain the greatest probability

– Bayesian Updating  for involving prior information

 Model parameter and its associated uncertainties can be assessed for the hole

(6)

P

ORTFOLIO

R

ISK

A

SSESSMENT

BuildingBuilding BuildingBuilding

Design Format of the Codes

Safety Level

Generic Risk Assessment

Generic Design of the Building Risk Indicators

Probabilistic Engineering Models

Human Losses

Financial Losses Safety

Investments Expected Human Portfolio Losses Expected Financial Portfolio Losses Portfolio Safety Investments 1 2 n… Model Parameters

(7)

R is k to lif e Safety investments Acceptable region LQI-threshold

E

FFICIENT

F

IRE

S

AFETY

M

EASURES

Life

Safety



A

rational

acceptance

criterion

for

decisions

regarding

life

safety

can

be

derived

based

on

the

Life

Quality

Index

(LQI)



The

LQI

define

a

threshold

for

the

societal

willingness

to

pay

for

a

marginal

increase

in

life

safety

Financial

Losses



Minimizing

the

total

expected

costs

while

the

fire

safety

measures

fulfil

the

Life

Safety

conditions

derived

from

the

LQI

principle

F in a n c ia ll o ss Optimum Investment costs Expected consequences Total expected costs

(8)

S

INGLE

F

AMILY

H

OUSES

An

economical

examination:

should

smoke

alarms

be

required

for

single

family

houses?

Model Parameters 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Discovery Time CDF No Smoke Alarm With Smoke Alarm

Smoke Alarm:

 Increasing probability of fire discovery

(derived from statistical data[1][2]₎

 Increasing probability that the fire

brigade will arrive sooner

[1]_Holborn,_Nolan_&_{Golt (2004)}

(9)

S

INGLE

F

AMILY

H

OUSES

105 106 10-4 10-3 10-2 10-1 Financial Loss [CHF] E x ceedance Pr obab ility 1 -F (x ) No Smoke Alarm With Smoke Alarm

102 103 104 105 106 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Financial Loss [CHF] C u m u la tiv e d is tr ib u tio n fu n c tio n F (x ) Data Model Model Data e e e

Estimation

of

the

model

parameters

using

a

dataset

of

1996

fires

incidents

(10)

S

INGLE

F

AMILY

H

OUSES

105 106 10-4 10-3 10-2 10-1 Financial Loss [CHF] E xceedance P robab ility 1 -F (x ) No Smoke Alarm With Smoke Alarm

Risk

reduction

by

demanding

Smoke

Alarms

for

single

family

houses:



 







no Smoke Alarm Smoke Alarm no Smoke Alarm 4.8 7.4% 64.5 CHF E L E L _{a SFH} r CHF E L a SFH  _    

(11)

N

EXT

S

TEPS

Ongoing research

at

ETH



Risk

based

review

of

the

Swiss

Fire

Safety

Regulations



Quantitative

influence

of

fire

protection

measures

on

the

risk



Efficiency

of

fire

safety

measures

Future

research

possibilities



Application

of

the

concept

to

performance

based

codes



Calibration

of

the

Fire

Safety

Code

for

a

portfolio

target

(12)

P

ROBLEMS

Generic System Representation and Available Data:

 Available data does not correspond to the needed input / output for

the engineering models

 Error in the estimation of input parameters based on the

available data

 Data (Switzerland) does not classify the damage according the

fire spread in the building Probabilistic Engineering Models

 Missing “simple” engineering models for natural fire conditions, i.e.

charring rate for timber, fire resistance of fire barriers (EI), fire spread

in multi‐compartment buildings etc. Investments cost

 Generic assessment of safety investment costs dependent on the

(13)

T

HANK YOU FOR YOUR

A

TTENTION

Y

OUR

Q

UESTIONS PLEASE

!

Gianluca

De

Sanctis

[email protected]

(14)

G

ENERIC

R

ISK

M

ODEL

Building Volume Constr. Year Living Area Number of Rooms Number of cells Initial Area A₀ Room arrangement I_RA Building Area A_tot Available Data Model‐ Input Model‐ Output Damage Ratio A/A_tot

Damaged Area A Loss Ratio S/V Pr obabilis tic Engineering Model Loss S Insurance Value Model Parameters Buildings 1:N_G

=

Room Height Available Data Fire Brigade Intervention