Publication Date: 6/1/74
    Pages: 2
    Date Entered: 2/22/84
    Title: STATISTICAL EVALUATION OF MATERIAL UNACCOUNTED FOR
    June 1974
    U.S. ATOMIC ENERGY COMMISSION
    REGULATORY GUIDE
    DIRECTORATE OF REGULATORY STANDARDS
    REGULATORY GUIDE 5.33
    STATISTICAL EVALUATION OF MATERIAL UNACCOUNTED FOR
A. INTRODUCTION
    Paragraph 70.51(e) of 10 CFR Part 70 requires certain AEC
    licensees authorized to possess special nuclear material to calculate
    material unaccounted for (MUF) quantities and their associated
    statistical limits of error (LEMUF) as part of their material control
    and accounting procedures for use in assuring that special nuclear
    material in their possession is accounted for. Paragraph 70.53(b)(1)
    requires that a report be made to the Commission if any single MUF
    exceeds its associated LEMUF and certain specified quantities and that
    the report include a statement of the probable reasons for the MUF and
    actions taken or planned. This guide identifies methods acceptable to
    the Regulatory staff for evaluating the statistical significance(1) of
    observed MUF values.
B. DISCUSSION
    While there may be mechanisms of process control involving
    identification of process anomalies that can provide an indication of
    possible missing material, the only positive means for assuring that the
    material is not missing is to measure all of the material and establish
    a measured material balance. Records are maintained of the measured
    quantities received into a plant and the measured quantities removed
    from a plant. The difference between these quantities should be on
    inventory in the plant. A measured physical inventory will either
    confirm that this quantity is present or indicate that some material is
    missing. Assuming that there is no inventory at the beginning of the
    time interval, the balance can be expressed by the equation:
    ----------
    (1) As defined in Regulatory Guide 5.3, "Statistical Terminology
    and Notation for Special Nuclear Materials Control and Accountability,"
    February 2, 1973.
    ----------
    Receipts - Removals = Inventory
    If there was material on hand at the beginning of the time
    interval for which a balance is being taken, a beginning inventory
    component would have to be added to the equation to give:
    Beginning Inventory + Receipts - Removals = Ending Inventory (1) Because of uncertainties(1) in measurements or unknown removals
    such as losses or thefts, this equation seldom balances in practice,
    indicating some observed material unaccounted for (MUF). Equation (1)
    can be recast as:
    (Beginning Inventory + Receipts) - (Removals + Ending Inventory) = MUF
    (2) If equation (1) balances, the MUF of equation (2) is zero. When
    equation (1) does not balance, the MUF of equation (2) represents some
    finite quantity of SNM. The significance of this quantity could
    represent only the uncertainties of the measurements, or it could
    include an unknown loss or theft. The first step in determining the
    significance of the MUF is to determine what value might be attributable
    to uncertainties of the measurement system.
    Each of the measured quantities in the material balance will have
    some uncertainty associated with it. Combination of these individual
    uncertainties by appropriate statistical methodology will result in
    limits in terms of SNM quantities by which equation (1) could be
    expected to be out of balance due only to the measurement system
    uncertainties, i.e., the MUF in equation (2) that could be expected.
    Since measurement uncertainties may be either positive or negative, a
    confidence interval(1) is established within which the MUF could be
    expected due only to measurement uncertainty. This confidence interval
    is defined by the "limits of error of the material unaccounted for"
    (LEMUF). It is the magnitude of this LEMUF value that determines, at a
    95% confidence level, how well the total material balance can confirm
    that all SNM is present or detect whether some is missing. When the MUF
    quantity is equal to or smaller than the LEMUF, the indication is that
    MUF could have occurred by chance due to measurement variation. If the
    MUF is greater than the LEMUF in either a positive or negative
    direction, the indication is that the MUF could not have occurred by
    chance due to measurement variation but that some other mechanism has
    had an effect. This could be a loss or theft of material (positive MUF)
    or an error in the system causing an unaccounted for gain (negative
    MUF).
    If MUF is due solely to random variations in the measurement
    system, MUF values taken over time from a series of balances around the
    same process or plant should tend toward zero. If this is not the case,
    and MUF values tend to show a consistent difference from zero, factors
    other than random measurement variations are indicated. Consistently
    positive MUF values would indicate that there may be some biased
    measurements, consistent measurement or recording mistakes, unknown or
    unrecorded inventory, or some continual small losses or thefts of
    material. Consistently negative MUF values would indicate that there
    may be some biased measurements or consistent measurement or recording
    mistakes.
    MUF values that are not consistent with measurement system
    variations, i.e., are in excess of such variations within stated
    statistical probabilities, are considered statistically significant.
    The purpose of this guide is to provide guidance in assessing the
    significance of MUF values but not in the investigation of MUF values
    that are found to be statistically significant.
C. REGULATORY POSITION
    The concepts, principles, and methods discussed and referenced
    below are generally acceptable to the Regulatory staff for evaluating
    the significance of MUF values resulting from measured material balances
    as specified in paragraph 70.51(e) of 10 CFR Part 70. Individual MUF
    values (short-term MUF) that are statistically significant (i.e., those
    that exceed the LEMUF values specified in paragraph 70.51(e)(5) and that
    exceed the minimum quantities specified in paragraph 70.51(e)(5)) are
    required to be investigated and reported to the AEC. In addition,
    combinations or sequences of MUF values (long-term MUF) which show
    trends significantly different from zero should be investigated to
    determine and correct the causes.
    ----------
    (1) As defined in Regulatory Guide 5.3, "Statistical Terminology
    and Notation for Special Nuclear Materials Control and Accountability,"
    February 2, 1973.
    ----------
    The evaluation of either short-term or long-term MUF should
    consist of testing whether the MUF quantity observed over a single
    material balance period or a cumulative period exceeds the LEMUF for
    that period. Specifically, the test criterion should be whether MUF @@
    LEMUF contains zero, i.e.,
    O epsilon(MUF Plus or Minus LEMUF) (3) Thus, an appropriate decision rule is that if observed MUF is
    greater than LEMUF, i.e.:
    MUF > LEMUF (4)then MUF should be declared significant and investigated. This 95%
    confidence interval test is equivalent to a two-sided hypothesis(1) test
    with a null hypothesis that MUF is zero at a 5% level of significance.
    Regulatory Guide 5.18, "Limit of Error Concepts and Principles of
    Calculation in Nuclear Materials Control," describes acceptable methods
    for calculation of limits of error.
    An example of how single, isolated observed MUF values should be
    evaluated is presented by John L. Jaech, Statistical Methods in Nuclear
    Material Control, TID-26298, December 1973, Section 7.1, specifically
    answers 7.A and 7.B of Section 7.1.3. However, the following
    specifications should be applied:
1. The value of M(o) should be set to zero.
2. The limit of error is c(alpha)sigma.
3. Absolute signs should be placed around x, the observed MUF,
    in order to indicate a two-sided test of significance.
4. A significance level of 5% should be assigned.
5. If the method of Section 7.1.3, question 7.B, is used, a
    confidence coefficient of 95% should be chosen.
    An example of how combinations and sequences of MUF values should
    be evaluated is given in Section 7.2 of the above reference,
    specifically answer 7.E of Section 7.2.2 with the following conditions
    specified:
1. Any related M(o) should be zero.
2. Any test of significance should be two-sided.
3. Control charts in answer 7.F would be more appropriate as
    the correlation between successive MUF values increases.
4. A 5% level of significance should be chosen.
5. The term denoted as a systematic-error variance in equation
    (7.9) and throughout TID-26298 is clearly defined on an applications
    basis in that reference. It should be noted in current practice that it
    may be determined by a mean square error or the variance of the
    estimated bias. Caution must be used in selecting which is appropriate
    for the process in question.
    29