Scribd Logo
Upload

Welcome to Scribd!

  • 19 views

    PFD (In) Sensitivity To PTC

    Uploaded by

    Raja Ganguli
    This document discusses the sensitivity of Probability of Failure on Demand (PFD) calculations to errors in estimates of proof test coverage. It shows that PFD is relatively insensitive to errors in proof test coverage compared to other uncertainties like failure rates. A 10% error in estimated coverage would result in only around a 10-20% change in calculated PFD. While more accurate coverage estimates are preferable, the uncertainties involved mean efforts to increase rigor may not be worthwhile given broader analysis uncertainties. Safety integrity levels account for an order of magnitude of uncertainty.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd

    PFD (In) Sensitivity To PTC

    Uploaded by

    Raja Ganguli
    19 views3 pages
    This document discusses the sensitivity of Probability of Failure on Demand (PFD) calculations to errors in estimates of proof test coverage. It shows that PFD is relatively insensitive to errors in proof test coverage compared to other uncertainties like failure rates. A 10% error in estimated coverage would result in only around a 10-20% change in calculated PFD. While more accurate coverage estimates are preferable, the uncertainties involved mean efforts to increase rigor may not be worthwhile given broader analysis uncertainties. Safety integrity levels account for an order of magnitude of uncertainty.

    Original Description:

    Functional safety

    Original Title

    PFD (In)Sensitivity to PTC

    Copyright

    © © All Rights Reserved

    Available Formats

    PDF, TXT or read online from Scribd

    Did you find this document useful?

    Is this content inappropriate?

    This document discusses the sensitivity of Probability of Failure on Demand (PFD) calculations to errors in estimates of proof test coverage. It shows that PFD is relatively insensitive to errors in proof test coverage compared to other uncertainties like failure rates. A 10% error in estimated coverage would result in only around a 10-20% change in calculated PFD. While more accurate coverage estimates are preferable, the uncertainties involved mean efforts to increase rigor may not be worthwhile given broader analysis uncertainties. Safety integrity levels account for an order of magnitude of uncertainty.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    19 views3 pages

    PFD (In) Sensitivity To PTC

    Uploaded by

    Raja Ganguli
    This document discusses the sensitivity of Probability of Failure on Demand (PFD) calculations to errors in estimates of proof test coverage. It shows that PFD is relatively insensitive to errors in proof test coverage compared to other uncertainties like failure rates. A 10% error in estimated coverage would result in only around a 10-20% change in calculated PFD. While more accurate coverage estimates are preferable, the uncertainties involved mean efforts to increase rigor may not be worthwhile given broader analysis uncertainties. Safety integrity levels account for an order of magnitude of uncertainty.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    of 3

    Functional Safety in Practice 4th Edition

    Addendum: PFD (In)Sensitivity to Proof Test Coverage Error


    Harvey T. Dearden

    In calculating average Probability of Failure on Demand (PFD), IEC 61511 requires that rather than
    using the average failure rate, a higher rate corresponding with an upper bound of 70% confidence
    should be used, to introduce a degree of conservatism to the evaluation. This specification does not
    place any constraints on the rate distribution, which could be relatively broad or narrow. There is a
    30% chance that the actual failure rate is greater than the 70% confidence figure, and without any
    specification as to the distribution spread, there is no constraint on how much higher the actual failure
    rate might be. The variation in available data from vendors and industry points to the possibility of
    actual failure rate being a half-order magnitude (a factor 3) greater than the 70% figure. [Ref 1]

    Proof test coverage is concerned with the fraction of dangerous unrevealed failures that will be
    discovered by a proof test; the actual rates are not a concern.

    Although Failure Mode Effects Diagnostic Analysis (FMEDA) may be employed to estimate this
    fraction, PFD is relatively insensitive to errors in proof test coverage (as shown below), and it is very
    much less burdensome to nominate this fraction simply from categorisation of equipment type and
    the proof test technique. [Ref 2, 3] (FMEDA would require the identification of a failure rate for each
    individual component of an element, together with an evaluation of the possible modes of failure,
    their impact on the element function, and whether or not they would be discoverable through a proof
    test.) Note that any FMEDA study might have to be revisited if the test approach changed. I suggest
    FMEDA may be much more useful to manufacturers in identifying the automatic diagnostic coverage
    (DC) of their products – this, unlike proof test coverage, is not dependent upon the end user’s testing
    approach – it is dictated by the manufacturers’ design provisions.

    We may take the basic 1oo1 formula for PFD average, using dangerous unrevealed failure rate (λ), the
    relationship between the ratio (n) of full test interval (TFull) and partial test interval (T), and the
    coverage factor claimed (C), to identify the sensitivity of PFD to errors in the estimate of proof test
    coverage.
    1 1
    𝑃𝐹𝐷 = 𝐶. 𝜆𝛵 + (1 − 𝐶). 𝜆𝛵𝐹𝑢𝑙𝑙
    2 2

    𝜕𝑃𝐹𝐷 1 1
    = 𝜆𝛵 − 𝜆𝛵𝐹𝑢𝑙𝑙
    𝜕𝐶 2 2

    1 1
    𝛿𝑃𝐹𝐷 = [ 𝜆𝛵 − 𝜆𝛵𝐹𝑢𝑙𝑙 ] × 𝛿𝐶
    2 2

    1 1
    𝛿𝑃𝐹𝐷 [2 𝜆𝛵 − 2 𝜆𝛵𝐹𝑢𝑙𝑙 ] × 𝛿𝐶
    =
    𝑃𝐹𝐷 1 1
    𝐶. 2 𝜆𝛵 + (1 − 𝐶). 2 𝜆𝛵𝐹𝑢𝑙𝑙

    v1 Page 1 of 3
    SISSuite: An advanced set of software tools for the functional
    safety lifecycle. Check out www.sissuite.com for details.

    If 𝑇𝐹𝑢𝑙𝑙 = 𝑛𝑇
    1 1
    𝛿𝑃𝐹𝐷 [2 𝜆𝛵 − 2 𝜆. 𝑛𝑇] × 𝛿𝐶
    =
    𝑃𝐹𝐷 1 1
    𝐶. 2 𝜆𝛵 + (1 − 𝐶). 2 𝜆. 𝑛𝑇

    𝛿𝑃𝐹𝐷 [1 − 𝑛] × 𝛿𝐶
    =
    𝑃𝐹𝐷 𝐶 + (1 − 𝐶). 𝑛
    Note the sensitivity relationship is independent of the failure rate and the absolute values of the test
    intervals. We may identify an error sensitivity factor (S) as:
    𝛿𝑃𝐹𝐷 [1 − 𝑛]
    𝑆 =( )⁄𝛿𝐶 =
    𝑃𝐹𝐷 𝐶 + (1 − 𝐶). 𝑛
    Taking a typical 1-year partial test interval and 5-year full test interval, and a claimed coverage of 75%,
    a shift in actual coverage of 10% (so, to 65% or 85%), would give a 20% shift in PFD (to 0.8x PFD or
    1.2x PFD), since S=2. When compared with the factor 3 associated with the failure rate itself (where
    the actual PFD could be 300% of the identified value), it can be seen that the uncertainty in proof test
    coverage is not a dominant concern.

    The table below identifies a range of values:

    n C ∆PFD %
    0.6 11
    0.7 13
    3
    0.8 14
    0.9 17
    0.6 15
    0.7 18
    5
    0.8 22
    0.9 29
    0.6 20
    0.7 24
    10
    0.8 32
    0.9 47

    Table: Shift in PFD corresponding with a 10% shift in Coverage, as a function of n and C.

    It should be remembered that broad uncertainties pertain throughout the analyses, calculations, and
    estimates employed in the functional safety lifecycle. In the face of these, attempts to introduce more
    rigour may well be misplaced, particularly when considered alongside the systematic concerns that
    are not susceptible to calculation. It is for these very reasons that safety integrity levels are notionally
    separated by an order magnitude.

    References:

    v1 Page 2 of 3
    SISSuite: An advanced set of software tools for the functional
    safety lifecycle. Check out www.sissuite.com for details.

    1. The Myth About Proof Test Coverage and Mission Time, Rev 0 2020, Mirek Generowicz,
    I+E Systems Pty. Ltd.
    2. Functional Safety in Practice 4th Ed., Ch. 37, Harvey T. Dearden 2022, SISSuite Ltd.
    3. Failure Modes & Test Coverage, Harvey T. Dearden, Briefing Note from InstMC, 2021.

    v1 Page 3 of 3

    You might also like