You are on page 1of 33

Representation of

Risk Values and Risk


Acceptance Criteria
Unit 34
QRA in Safety Engineering
Spring 2016
1

References
n

Modarres, M., Risk Analysis in Engineering,


Taylor&Francis, 2006, Chap 7 on website
(Modarres, RAE)

Modarres, M., M. Kaminskiy, and Vasiliy Krivtsov, Reliability


Engineering and Risk Analysis, Taylor & Francis, 2010
(Modarres, RERA)

Jordaan, Ian, Decisions Under Uncertainty Probabilistic


Analysis for Engineering Decisions, Cambridge University
Press, 2005 (Jordaan, 2005)

Ayyub, B.M., Risk Analysis in Engineering and Economics,


Chapman & Hall/CRC, 2003 (Ayyub, RAEE)
2

Summary
n

Usefulness of Risk Acceptance Criteria:

Compare risk assessment methods and criteria


ranges, such as traditional ALARP, with other
criteria ranges for risk decision making as, e.g., we
saw were needed for each identified ALARP region

Develop a consistent and flexible expression of risk


standards and acceptance criteria values and
ranges to be discussed with educated stakeholders
within Risk Governance, and communicated to the
community (as shown on the Risk Map).

Build confidence in acceptable risk standards as a


foundation for management of risk at the plant.
3

Risk Measure for an Individual or Facility,


or a Group of People or Facilities
n

Individual risks probability of exposure of a person,


system, or plant to a hazard or a particular level of the
hazard. Ex.: risk of injury or fatality for an individual
performing work in a plant. Fatality 10-4 yr-1

Societal risks cumulative frequency or probability of


outcome for a number of people (or facilities) affected,
by a specified consequence level from exposure to
specified hazards, represented in, e.g., F-N profiles.
Fatality 10-6 yr-1

Context involving mixing Individual and Societal risk,


Fatality 10-4 10-6 yr-1

Risk Acceptance Criteria


n

Criteria for acceptance have been based on relative


comparisons to be included in informing and educational
messages of Risk Communication:
Compare

calculated risk, frequency, or exposure to a


hazard to historical levels of similar hazards

Compare

risk levels of observed events with


previously estimated risk levels of exposure to the
same hazard

Compare

a risk with more familiar risks in society and


daily life, e.g., individual fatality risk 10-4 /person-yr for
volunteer sports (see RCF* list)
*Risk Conversion Factor

Individual Risk Criteria Review


n

Tolerable (or acceptable) risk criteria have been


proposed for an individual, a facility, and for society.

A tolerable level of individual risk is historically


founded on the familiar risk of natural hazards or
disease.

Individual fatality risk is tolerably low if it is than the


fatality risk of natural hazards ~10-6 yr-1.

Individual risk is unacceptably high if it is near or


greater than the fatality risk of disease ~ 10-3 yr-1.
6

Acceptable Everyday Risk Levels


n

Higher risk levels are generally more acceptable for


familiar events than for unfamiliar events.

Accidental risk of 1x10-4 fatality/yr for 10-14 year old


(young, healthy) age group is a risk criterion for
voluntary activities, such as working in a plant

Maximum voluntary risk level for an death rate^ of


1x10-4/yr has been advocated by the NRC (Nuclear
Regulatory Commission)

Targeting risk reductions based on a fraction of basic


risk,~10-4/yr, to increase safety index, SI = -log10
^Reactor

Safety Study, NRC, WASH-1400, 1979

Individual Risk Acceptance Limits

Fatality rate

Historical

(illnesses)

voluntary
air travel
involuntary

(natural hazards)

(Modarres, RAE)

Presentation of Risk Results: Risk Profiles


Risk profiles or distributions of values
n
n

Generalized Risk Profile


Not acceptable

Co-ordinates can be:


Risk

vs. Consequence

Probability

vs. Consequence

Risk

vs. Frequency

Risk

vs. Time (of risk exposure)

Risk

vs. Space, or Distance

Magnitude

of exposure vs. Time

Performance
Reliability

F 7.1, Limits for risk

vs. Time

vs. Time

acceptability threshold

Frequency vs. Consequence:


(Modarres, RAE)

Acceptable

f ic = k

Utility/disutility function

e.g., Frequency

> 1, risk averse


= 1, risk neutral
< 1, risk seeking

, risk averse parameter

Generalized Risk Curve, Risk Profile


n Form:
Continuous
Discrete

distribution

distribution

Combination
n

of both discrete and continuous

Expression for acceptability limits as a power law


of frequency (or probability) and consequence
f i c = k, constant
> 1, risk averse
= 1, risk neutral
< 1, risk seeking

k = risk acceptability threshold


= risk averse parameter

k, levels set by policy and risk


acceptability of organization or
decision maker

10

Representation of Risk Results


n Population

mortality rate

Measure

of societal or system risk of an activity


in terms of its impact on life expectancy
conditional failure rate = (i), shown here as h(i)

Number that fail (or die) in the i + 1 year of life


h(i) =
Number that function (live) at age i
Discrete version of the conditional failure rate
11

Life Expectancy (LE) Example


n

Consider death rate from all causes and from


circulatory diseases^. Calculate loss of life
expectancy (LLE) due to diseases of the
blood circulatory system.

Analyze: Use the given data and equations


for h(i), conditional failure rate, all causes,
and hj(i) (for circulatory system causes) to
calculate the mortality rates due to circulatory
diseases for each age group.
^in

New Zealand, 2000

12

Example Data, Solution

h j (i) =

# at age i +1 who die of j


# who live in age i

13

(Modarres, RAE)

All causes

Circulatory

LE (all causes) ~ 79.2 yr


LEj (w/o circulatory) ~ 86.3 yr
LLEj (circulatory) ~ 7.1 yr
Example of targeting incremental risk reduction to increase SI

(Modarres, RAE)

(Modarres, RAE)

Societal

LLEj

14

Risk Criteria Related to Effect on


Fatality Risk
n

Most criteria for health and injuries are related to


fatality risk and to the influence of injuries on the
probability of fatality.

Reasons: There are wide ranges of health and


injury severity outcomes, but an individual death
outcome has a fixed severity, and fatality rates
from illnesses (10-3) and events of nature (10-6)
are known and accepted.

15

Health and Safety Risk Acceptance


Criteria Point Values, Ranges
n

Categories of Acceptance Criteria


Individual risk

Groups, community, societal risk

Absolute or relative criteria

Continuous or incidental hazard exposures


n Continuous: Living near a contaminated area:
10-5 -10-6 depending on effective distance from

Occasional voluntary working in a


contaminated area, ~10-4-10-5

plant

n Incidental:

16

Individual Criteria: Voluntary, Involuntary


n

Acceptability of risk levels for hazard exposure


during voluntary activities is usually much higher
(~ 102 to 103 or greater) than for involuntary
activities.

Voluntary activities include working at a plant,


mountain climbing, motorcycling, and smoking.

Involuntary include living near a plant, diseases,


and natural events

See Risk Conversion Factor list

17

Societal Risk Acceptance Criteria


n

Societal risk acceptance criterion: risk to human life^


k
P(Nd > n) n < A absolute risk criterion, acceptability threshold

P (Nd > n) < A/nk

~ f c = k , Slide 9
, risk averse parameter

Nd = # fatalities in 1 year following a single accident


P (Nd > n) = probability or frequency of ND > n
n = number of fatalities
A = 0.001 /year to 1 /year (less tolerant more tolerant)

k = 1 to 2 risk averse parameter: higher values of k


(or ): increased aversion or intolerance to severe
consequences
^International Organization for Standardization (ISO 2394)

18

Societal

P (Nd > n) = A/nk

FN curves

less stringent criterion


Old
Technology
Not acceptable

Cumulative

Not acceptable
Acceptable

more stringent criterion


and higher risk aversion, k
New
Technology
Similar to the Farmer F-N curve but with the risk averse parameter k
Acceptable

Exceeding the upper line: unacceptable

Below the lower limit: acceptable/negligible risk level

(Modarres, RAE)

19

Risk Criteria for High-Dose Levels


n Societal

radiation exposures at high-dose


levels: risk acceptance level < 10-7 /yr

n F-N

curves with 3 areas

Area
Acceptable
ALARP
Unacceptable

Risk Value
A2 < 10-7
10-5 10-7
A1 > 10-5

Nfatalities
10
10
10

ALARP: as low as reasonably practical

<10-6 for N = 1

>10-4 for N = 1

20

Societal Risk Curve


A1

P (Nd > n) < A/nk

A 2< A 1

same k
same
slope

(Modarres, RAE)

21

Economic Risk and Performance


Acceptance Criteria
n

Economic Risk limits

Cumulative Cost of Consequences vs. Cost of


Control profiles (as in risk of fatality exposure),
mitigation, or protection from specific hazard
exposures

Similar to F-N curves of fatality risk limits

Example of ALARP analysis, which shows that


Risk criteria and uncertainty analysis are
22
necessary in applying ALARP.

Economic Performance Criteria

ALARP region (acceptable only if it


also falls within the acceptable risk
region within the uncertainty)
Protection and Mitigation Costs
(Modarres, RAE)

23

lower risk by targeting higher SI values


ALARP Criterion Target
through risk effectiveness analysis (RE).

SI = -log10 pf
RE = Cost/R

SI = 4

10-5
SI = 5

Target values

10-4,

Safety Index (SI) target risk levels

Work/all causes, < 10-4 , SI > 4


Specific causes, 10-5 , SI 5

10-6, SI = 6

(Jordaan, 2005)

24

Economic Risk and Performance


Acceptance Criteria for Risk
n

Risk acceptance criteria using, for example, Figures


of Merit: risk-related measures used in nuclear
engineering to support decisions such as test and
maintenance requirements to limit unavailability of
critical components, such as reducing SAD to ~0 and
thereby increasing SI, the Safety Index.

Metrics used to determine acceptability of changes to


system design and operation with changes in risk
that affect the system as represented by a risk model
25

Instantaneous Value Limits


n

Judged by effect on the overall change in risk level, R

Absolute acceptability limits are applied to event


frequencies such as CDF (reactor core damage
frequency) or toxic releases

Categorize events:
n

Upper Limits to identify and control risk spikes, such


as occur during maintenance

Lower Limits screen out events that are not risk


significant and do not require further analysis

26

Probability of Failure on Demand, PFD


n

Recall probability of failure on demand, pfd, for


random failure, not including mean down time, M ,
during the time regime t0 << 1 and the distribution
leading terms from exponential, pfd ~ t0 and from
Weibull, pfd ~ (t0)/(1+) as discussed previously.

For development of cost effect test and maintenance


schedules and time between tests, t0, use minimum
reliability levels between tests for critical components.

For Weibull with = 1.5, = 0.5/yr, given the reliability


criterion 0.95 based on acceptable risk,
(t0)/(1+) = (t0)1.5/(2.5) 0.05; t0 (0.125)0.67/0.5 yr
0.5 yr, which is the determined time between tests to
meet the minimum reliability and the corresponding risk
standard.

27

Unavailability on Demand, QOD


n

Unavailability on Demand is more realistic than PFD in


capturing the true failure probability due to all of:
Component

random failure

Component

test and maintenance, MTTR

Organizational

delays: supply & administration, SAD

Supply and administration delays, SAD should be quite


low ~ 0 and are low in a High Resilience Organization.

QOD = t0 + MTTR/Tm + SAD/Tm, for exponential

QOD = (t0)/(1+) + MTTR/Tm + SAD/Tm, for Weibull

Tm is total time: Tm = t0 + MTTR + SAD,


where expected down time is M = MTTR + SAD
28

Risk Acceptance Based on Value of a


Statistical Human Life: highly subjective
n

Based on lost gross output of employee

Livelihood Approach assigns a value in direct


proportion to income earned

Insurance Method: based on amount of insurance


purchased (value placed on own life)

Compensation Method: whether risk could have


been foreseen (less compensation than if risk level
could not have been foreseen)

WTP (willingness-to-pay) spent by organization to


29
increase safety or to reduce mortality risk

WTP Estimate Examples


n

Assuming an average WTP of US$ 6 per person


for risk reduction of 1 life in 1 million individuals,
the value of the statistical life would be US$ 6
million.

Assuming that workers receive premiums of US$


600 to compensate for an increased annual work
fatality risk of R = 10-4, the resulting WTP of
~ US$ 6 million.
Risk Effectiveness, RE:

Cost $600
RE =
=
=6M
4
Risk 10

Other WTP values have been estimated based on


wage premiums and mortality probabilities.
30

Risk Conversion Factors


n

Risk categories with risk factors, such as familiarity,


old/new, to estimate perceived risks for a variety of
activities

Compare perceived risks in one category (that


invoke similar perceptions and yield similar risk
acceptance), such as familiarity = old, with perceived
risks in another category, such as familiarity = new.

Use a risk conversion factor, such as risk of


new = 10risk of old, for an activity category to
estimate acceptable risk levels in another activity
category.
31

Laws of Acceptable Risk Level


n

Example: Public will accept voluntary risks


about 1001000 times (or more) greater than for
involuntary imposed risks.

The statistical death rate (historically, 10-3,


illnesses; 10-6, natural events) is a
psychological metric for the level of
acceptability of other risk levels.

The acceptability of risk is ~ proportional to the


3rd power of the perceived benefits

Risk (Benefit)

3
32

Psychometric Risk Factor Model


n

Risk conversion factors (RCF) represent bias


against risks that are unfamiliar (by a factor of
10), catastrophic (30), involuntary (100 to 1000),
uncontrollable (5 10), or have immediate
consequences (30).

RCF are estimates for direction of bias


(up or down in level) and degree of bias in
perception of risk.
33

You might also like