notes98_105

  • Upload
    peach5

  • View
    217

  • Download
    0

Embed Size (px)

Citation preview

  • 7/30/2019 notes98_105

    1/9

    98

    RISK ANALYSIS

    Risk is defined as

    Risk =

    N

    n=1(likelihood of hazardous event i)

    (Consequence of hazardous event i )

    Risk analysis (assessment) is a technique for identifying, characterizing, quantifying and

    evaluating hazards. It generally consists of three components:

    1. Identifying and characterizing hazardous events.

    2. Estimation of the likelihood of these events.

    3. Estimation of the consequences of events.

    To illustrate the first two components using the event/fault-tree approach, consider the

    pressure tank example:

  • 7/30/2019 notes98_105

    2/9

    99

    If the likelihood of the basic events are known, then the likelihood of a rupture occurringcan be quantified. Consequences of the rupture can be various, including damage to the

    tank system, environment and possible loss of life.

    Since the consequence of an event can contain many components, risk is a vector quan-

    tity, in general. However, for simplification and comparative purposes, often a single

    measure is used.

  • 7/30/2019 notes98_105

    3/9

    100

  • 7/30/2019 notes98_105

    4/9

    101

    Fatal Accident Frequency Rate

    From Various Daily Activities (per 108 hour)

    Some Example Risk Profiles

  • 7/30/2019 notes98_105

    5/9

    102

    Risk Perception and Acceptability

    While a quantitative measure has been defined for risk, risk perception is often based

    upon subjective judgement, beliefs and societal bias rather that objective measure. Gen-

    erally, risk associated with unfamiliar and incorrectly publicized activities is perceived as

    much higher than its actual value.

    Risk of motor and aviation accidents is perceived to be 100 times lower.

    Risk of nuclear power and food coloring is over estimated by a factor of 10,000.

    Risk conversion and compensating factors must be applied to account for public bias

    against events that are

    unfamiliar (X 10),

    catastrophic (X 30)

    inv oluntary (X 100)

    uncontrollable (X 5-10)

    with immediate consequences (X 30).

    For example, the risk from nuclear power production in the U.S. is less than 103 early

    fatalities/year and the risk of flying is about 10 deaths/year. Howev er, nuclear power pro-

    duction is regarded to have higher risk compared to flying because nuclear power is unfa-

    miliar (103 10 = 102), nuclear accidents may have catastrophic consequences

    (102 30 = 0. 3) and is involuntary (0. 3 100 = 30). Similarly, the figure above shows

    that while the risk associated with working in a chemical plant and domestic activities

    such as eating, washing, dressing are about the same (2. 5 108/hour vs.

    3. 5 108/hour) the latter risk is usually regarded as much less since it is familiar, volun-

    tary and controllable.

    It is interesting to note that the public bias is consistent with results from Bayesian

    statistics. Suppose there is a debate about the safety of a new facility. The facility is

    designed to withstand accidents. It is estimated that an accident yields 1 fatality with

    probability 0.01 and 1 fatality with probability 0.99. However, if a defect exists in the

    design or construction an accident yields 100 fatalities with probability 0.99 and 1 fatality

    with probability 0.01 (i.e. catastrophic consequence). The public believes the and non-

    existence of the defect are equally probable. Now consider what happens following an

    accident with 100 fatalities. Define the events

    A: Defect existsB: 100 fatalities in the accident

    From the Bayes theorem

  • 7/30/2019 notes98_105

    6/9

    103

    P(A|B)=P(B|A)P(A)

    P(B|A)P(A) + P(B|A)P(A)=

    (0. 99)(0. 5)

    (0. 99)(0. 5) + (0. 01)(0. 5)= 0. 99.

    Then once the accident with 100 fatalities occurs the public may be justified in thinking

    that the probability of a defect in the design/construction is high and hence future acci-

    dents will also yield 100 fatalities. The Chernobyl incident is a good example to this bias

    (but not the number of fatalities). The next two tables illustrate the difference between

    the technical expert and lay public to risks.

  • 7/30/2019 notes98_105

    7/9

    104

  • 7/30/2019 notes98_105

    8/9

    105

    Steps in Conducting a Probabilistic Risk Assessment (PRA)

    1. Methodology Definition: Includes required computer codes, facility experts and

    analytical experts and provides a road map for the analysis.

    2. Familiarization and Information Assembly: Acquiring a general knowledge of thephysical system layout, administrative controls, maintenance and test procedures

    and safety systems. Physical interactions among all major systems should be identi-

    fied. Past major failures and abnormal events should be noted and studied.

    3. Identification of Initiating Events: Delineation and grouping of external and inter-

    nal off-normal conditions. Combine into different groups the initiating events that

    directly break all hazard barriers,

    break the same hazard barriers (not necessarily all the barriers),

    require the same group of mitigating personnel or automatic actions,

    simultaneously disable the normal process as well as some of the mitigating

    human or automatic actions.

    4. Sequence or Scenario Development: Description of the probabilistic consequence

    ev olution such as by using the event/fault-tree approach with computer codes mod-

    eling the relevant processes.

    5. Dependent Failure Considerations: Identify items that are

    similar such as similar pumps, valves, diesel generators.

    susceptible to common cause failure (e.g. devices powered by the same

    source),

    functional dependencies (e.g. generator is driven by the turbine).

    6. Failure Data Analysis: Determine

    generic failure data for each component in the fault-trees,

    test, repair, outage data (from experience if available),

    frequency of initiating events from experience, expert judgement or generic

    sources,

    common cause probability for similar items.

    6. Quantification: Quantification of the event/fault trees using Boolean algebra as dis-

    cussed earlier.

    7. Damage Assessment: Quantification of consequences using, for example, atmo-

    spheric dispersion codes and medical data for pollutant leakage into the atmosphere.

  • 7/30/2019 notes98_105

    9/9

    106