Sie sind auf Seite 1von 4

Representing Societal Risk as an FN Curve and Calculating the Expectation Value

Risk management in occupational safety often requires the numerical quantification of risks. For situations such as the failure of a guard at a power press, the appropriate formulation of quantified risk is individual risk. For situations having the potential to harm many people, so-called major hazards, the appropriate formulation of quantified risk is societal risk. Individual risk is the simply the predicted frequency of the undesired outcome. ocietal risk takes account of !oth the severity of the range of possi!le outcomes and the frequency at which they each are predicted to occur. It is usually presented as a two dimensional relationship !etween frequency and cumulative severity of outcome, called an F" curve. #his is !est understood from a simple e$ample.

A Simple FN Curve
%onsider a major hazard site storing a compressed liquefied to$ic gas, such as chlorine, in a !ulk tank. In a comprehensive assessment, loss of containment of the chlorine from a !ulk tank might !e modelled as possi!ly &' idealised scenarios. (dditional scenarios would !e used to model loss of containment of the chlorine from associated pipework and vessels. For this e$ample a single scenario will !e used. )ense gas dispersion models are used to model each release for a range of com!inations of wind speed and atmospheric sta!ility. For to$ic gases the use of * com!inations is common. For this e$ample two com!inations representing daytime and night-time conditions will !e used. Risk calculations may assume that the likelihood of the wind !lowing in a particular direction is equally distri!uted around the compass. #his is using a uniform wind rose. #his approach is normally regarded as inadequate where suita!le meteorological data is availa!le, such as in +reat ,ritain. Risk assessment for to$ic gases would typically aggregate the meteorological data into &sectors. For this e$ample meteorological data aggregated into * sectors will !e used. #he num!er of fatalities are predicted for each scenario, for each direction, for each speed.sta!ility com!ination. For this e$ample the results can !e represented !y the following diagram.

0.15 30 20 5 10 0.2 10 17 18 15



#his indicates that a wind !lowing towards the "/ in daytime conditions is predicted to kill &' persons, and in night-time conditions to kill -0 persons. #he wind !lows in that direction for *'1 of the time day and night. If the loss of containment that leads to these outcomes is * chances in a million per year 2* $ &'-34 then the following ta!le shows the calculation of the frequencies of the individual outcomes and the e$pectation value for each outcome. Event Base Direction Day/Night utco!e utco!e E"#ectation $a%ue Frequency Frequency &#er year' &(' &(' &#er year' &)ata%ities' &)ata%ities #er year' 4.0E(* 0.4 0.8 1.3E(* 10 1.3E(5 4.0E(* 0.4 0.2 3.2E(7 25 8.0E(* 4.0E(* 0.25 0.8 8.0E(7 17 1.4E(5 4.0E(* 0.25 0.2 2.0E(7 35 7.0E(* 4.0E(* 0.2 0.8 *.4E(7 5 3.2E(* 4.0E(* 0.2 0.2 1.*E(7 15 2.4E(* 4.0E(* 0.15 0.8 4.8E(7 20 -.*E(* 4.0E(* 0.15 0.2 1.2E(7 50 *.0E(* 4.0E-6 6. E-!

NE(Day NE(Night +E(Day +E(Night +,(Day +,(Night N,(Day N,(Night

#he e$pectation value of 3.5 $ &'-0 is often used for valuing safety improvements in monetary terms. If a chemical plant is considering fitting an additional safety measure to reduce the likelihood of this loss of containment then an initial screening assumption might !e that the event will eliminated !y the measure. If the plant lifetime is 5' years, and the value placed on an averted fatality is 6&.million, then the monetary value of implementing the measure is7 5' $ 23.5 $ &'-04 $ 2&.- $ &'34 8 6--39 #his value would then !e compared with the costs of implementing the measure and it must !e implemented unless the costs are grossly disproportionate to the monetary value of the !enefit. In practice, there would !e more harms averted than just the fatalities used here. ome initial information in occupational safety cost !enefit analysis can !e found on the : / we! site2&4. #he author;s views on gross disproportion can !e found in a technical paper2-4. In order to produce a F" curve the ta!le of outcomes is then sorted into decreasing num!er of fatalities and an additional column added for the cumulative frequency. Event Base Freq Direction Day/Night &(' 0.15 0.25 0.4 0.15 0.25 0.2 0.4 0.2 &(' 0.2 0.2 0.2 0.8 0.8 0.2 0.8 0.8 utco!e utco!e .u!u%ative Frequency Frequency &#er year' &)ata%ities' &#er year' 1.2E(7 50 1.2E(7 2.0E(7 35 3.2E(7 3.2E(7 25 *.4E(7 4.8E(7 20 1.1E(* 8.0E(7 17 1.-E(* 1.*E(7 15 2.1E(* 1.3E(* 10 3.4E(* *.4E(7 5 4.0E-6

&#er year' N,(Night 4.0E(* +E(Night 4.0E(* NE(Night 4.0E(* N,(Day 4.0E(* +E(Day 4.0E(* +,(Night 4.0E(* NE(Day 4.0E(* +,(Day 4.0E(*

& - <+ross )isproportion, tep !y tep - ( =ossi!le (pproach to /valuating (dditional >easures at %?>(: ites<, !y >artin : +oose, -''3 2availa!le from the author4

#he resulting values of fatalities 2"4 are then plotted against cumulative frequency 2F4 conventionally on a log.log plot.

FN .urve
1.0E( 5

4.0E( * 3.4E( *

.u!u%ative Frequency &#er year'

2.1E( * 1.-E( *

1.1E( * 1.0E( *

*.4E( 7

3.2E( 7

1.2E( 7 1.0E( 7 1 10 Nu!/er o) Fata%ities 100

In a real assessment to produce an F" curve the num!er of outcomes analysed and plotted is likely to !e hundreds and possi!ly thousands. #he F" plot generated would then !e compared with appropriate societal risk criteria. #he criteria usually take the format of a pair of straight lines similar to those shown in the ne$t page.

>artin +oose 0th >arch -'&'

Societal Risk Criterion


Intolerable/ALARP boundary Broadly acceptable/ALARP boundary



Frequency of (n) or more Fatalities (yr -1)

R2P2 Point

Canvey Point



ACMH 1 Point

Intolerable Region




Tolerable if ALARP Region



Broadly Acceptable Region










Number of Fatalities (n)

Note: These criterion lines are now adopted HID criteria (HID Circular SPC/Permissioning/09 - Annex)

Verwandte Interessen