Sie sind auf Seite 1von 9

AdMet 2012 Paper No.

UM 001

Guard-banding Methods-An Overview

Swanand Rishi

ETDC, STQC Directorate, Department of IT, Govt. of India,


Agriculture College Campus, Shivajinagar, Pune 411005
Email: snrishi@stqc.nic.in

ABSTRACT : Guard-banding is a method of setting test limits lower than Specication


limits (SL) in order to optimise False-Accept & False-Reject risks. A guard-band is the
offset from the SL to the acceptance test limit (TL) that is used for the pass/fail decision.
It is used to reduce the False-Accept Risk, which increases the condence that a device
is really in-specication by factoring in the measurement errors. This involves the risk
of false-accept as well as false-reject, owing to various factors like specication limits,
TUR, measurement uncertainty & a-priory probability of the device being in-tolerance.
This paper presents various prevalent guard-banding methods in order to manage such
risks in calibration.

Keywords : Guard-band, TUR, Uncertainty, Specication Limit & Tolerance Limits

INTRODUCTION : Probability of False Acceptance (PFA). ISO/


IEC 17025 is silent on maximum level of false
In a near-ideal calibration scenario, accept risk. However, Clause 5.3 of ANSI/
one would expect reading of Device Under NCSL 540.3 [2] species PFA or Consumers
Calibration (DUC) closely matching the risk to be not more than 2%.
standards value, indicating high process
capability. The Specication Limits (SL) will There are two types of False-Accept
be much wider under such conditions. With risks- unconditional & conditional risks. An
consistent improvement in process capability, Unconditional False-Accept Risk is the average
manufacturer would tighten the specs & risk for a population of calibrated devices. It is
shorten the SL band. In such a real-life an appropriate statistic for managing a large
scenario, the scatter of readings is not always no. of instruments - typically a manufacturing
close to standard value; many values found scenario. On the other hand, conditional False-
lying close to the specication limits. This is Accept Risk is appropriate when dealing with
particularly the case after a few years of use & a specic instrument - typically a recalibration
when an item is subjected to recalibration. The scenario.
ISO IEC 17025:2005[1] calls for incorporation
of uncertainty while making compliance The False-Reject risk is the probability that a
statement. As such, the readings which are device which is actually in-tolerance is rejected
close to SL are likely to fall outside SL, leading due to measurement process error. The False-
to non-compliance. Reject risk is also called as Producers risk or
Probability of False-Rejection (PFR). This risk is
The False-Accept risk is the probability that evaluated in lieu of anticipated additional costs
a device which is actually out-of-tolerance is like re-calibration, adjustment or rectication.
accepted due to measurement process error One must note that increasing Guard-band for
or the uncertainty. A False-Accept occurs when reducing False-Accept risk (Consumers risk)
the UUT bias is out of tolerance, but deviation disproportionately increases the False-Reject
is not. It is also called as Consumers risk or risk (Producers risk) as seen from Fig. 1.

1
Table 1: False-Accept & False-Reject risk at
different TURs at SL = 2

Probability of False Acceptance (PFA)


could be altered by ne-tuning of calibration
system control tools like-
Measurement reliability
Calibration intervals
Fig. 1 (a) : Producer Risk
Calibration process uncertainty
Calibration adjustments
Guard-bands
This paper deliberates on the last option.
Before delving into various methods, some
important terms need attention.

Guard-band : It is nothing but range by


which the specication limit (SL) is reduced
so that there is more condence in the Pass/
Fail decisions, which are taken based on
tolerance limit (TL) rather than SL. Particularly
employed when TURs are lower than 4:1, this
is a safety margin for tightening an acceptance
(pass) limit. Thus, Guard-band GB = SL - TL.

It is also expressed as a multiple of SL, i.e.


GB = KSL; K being 1.
Fig. 1(b) : Consumer Risk
The Guard-band is usually set at a
This risk shows sharp rise at lower TURs. As point equal to the specication minus the
there are cost implications, balance between uncertainty but is often adjusted to get the
the two risks is necessary. The selection of same condence that results at a TUR of 4:1
a Guard-band method & its decision rule is & acceptance limit set at the SL. Guard-bands
more a business decision for manufacturers are employed by manufacturers as well as
whereas it is a matter of quality policy for the calibration/test labs. The concept is illustrated
test & calibration labs. in Fig. 2. In conditional False-Accept risk,
generally larger guard bands are used,
Table 1 shows some values from graph in while smaller guard bands are employed in
Fig. 1 for specication limit of 2 (i.e. approx. unconditional False-Accept risk. Note that
95% CL or In-tolerance probability & assumed Guard-band doesnt altogether avoid False-
normal distribution for UUT specications as Accept or False-Reject risk, but increases
well as uncertainty). condence in pass/fail decisions.

2
A-Priory Probability Distribution :
The evaluation of PFA assumes a-priory
probability that the device being calibrated is
in-tolerance prior to actual calibration.

Depending upon test-point reliability or


reliability of population of UUTs, the PFA
varies. Reliability gures of 80 to 95 % are
Fig. 2: Guard-band (shown at Uncertainty) common. The higher the reliability, the lower
the PFA.
Specication Limits (SL): These
are specied by manufacturer either as
VARIOUS GUARD-BANDING METHODS:
rectangular, usually two-sided symmetrical
specs, or at 95% condence level.
ISO17025:2005, cl. 5.10.4.2 states:
When statements of compliance are made,
It is stated in GUM[3] that, when a
the uncertainty of measurement shall be
specication is quoted for a given coverage
taken into account. However, it provides no
probability, then a Normal (Gaussian)
specic guidance for taking the measurement
distribution can be assumed. The methods
uncertainty into account when assigning
described in the paper are based on this
Pass/Fail status nor does it specify maximum
assumption.
level of false accept risk. This allows different
guard banding strategies to co-exist; some of
Test Uncertainty Ratio (TUR) : A TUR
the prevalent methods are presented here.
is the ratio of DUC specications to the
measurement process uncertainty, both
Note that tables below are based on
specied at same condence level. A higher
Specication Limit (SL) normalised to 100 for
TUR reduces probability of wrong Pass-Fail
ease of use & understanding. The Combined
decisions. A TUR of 10:1 is recommended in
Uncertainty Uc (although it appears only in
many standards & is in force for a couple of
method 2) assumes uncertainty evaluated at
decades.
approx 95% coverage probability i.e. at k = 2.
The old US military standard MIL-
1. Guard Banding at Measurement
STD-45662A[4] & old ANSI STD Z 540.1[5]
Uncertainty : ILAC-G8[6], ISO 14253-1[7]
recommend a minimum TUR of 4:1. Due to
recommend to take measurement uncertainty
technological advancements & declining
into account directly (without any factor),
costs of precision equipment, a TUR of 4:1
while taking Pass/Fail decision or making
is widely accepted nowadays. (Labs cannot
compliance statement. ILAC-G8 cl. 2.6 states
afford to keep standards to maintain TUR of
- In calibration, measurement uncertainty
>4:1 for all workload/test points) The False-
shall always be taken into account when
Accept risk & False-Reject risk become major
compliance with specication is made.
concerns when the TUR is worse than 4:1,
other factors remaining same. TUR has direct
Thus, the Tolerance limits (TL) are set
effect on the risk probability- the lower the
by subtracting the uncertainty from the
TUR the higher the False-Accept risk as well
specication limits (SL).
as False-Reject risk. Thus a lower TUR is not
desirable from both consumers & producers
The Guard-band in Fig. 2 is approximately
risk perspective.
set at uncertainty for the sake of illustration. In
ILAC-G8 this gure is shown without Guard-
Measurement Uncertainty : It is the
band & Pass/Fail legends.
probable measurement process error, usually
specied at approx 95% condence level at
The diamonds in the gure are various
a coverage factor k = 2 & is mandatory for
measurement results with uncertainty bands
accreditation per ISO/IEC 17025.
above & below. Clause 2.3 of ILAC-G8
recommends the following compliance / non-
3
compliance criteria (which is more or less Table 2 : Values of TL for Guard-banding at
same for other methods as well): uncertainty

(a) Compliance : If the specication limit


is not breached by the measurement result
plus the expanded uncertainty with a 95%
coverage probability, then compliance with
the specication can be stated. In calibration
this is often reported as Pass (See Case 1
of Fig. 2)

(b) Non-compliance : If the specication Table 3 : Application of uncertainty for status-


limit is exceeded by the measurement result decision
minus the expanded uncertainty with a 95%
coverage probability, then noncompliance with
the specication can be stated. In calibration
this is often reported as Fail (See Case 4 of
Fig. 2)

(c) If the measurement result plus/minus


the expanded uncertainty with a 95 %
coverage probability overlaps the limit (See
Case 2 and 3 of Fig. 2), it is not possible to
state compliance or non-compliance. The
measurement result and the expanded
uncertainty with a 95 % coverage probability
The method is highly favourable to the
should then be reported together with a
consumer & extremely taxing for the producer.
statement indicating that neither compliance
Further, it has a large discontinuity at TUR just
nor non-compliance was demonstrated. In
below 4:1, e.g. even at a TUR of say 3.99:1, TL
Case 2 of Fig. 2 it is possible to indicate, that
must be set at 75% of SL; whereas for a TUR
the measurement is below the limit, which
of 4.01:1, TL may be set at SL (as it is invoked
can be done using a similar statement- It
at a TUR of 4:1). Thus the producer runs
is not possible to state compliance using a
comparatively higher risk of False-Rejects for
95 % coverage probability for the expanded
TURs just below 4:1 & enjoys lower risk of
uncertainty although the measurement result
False-Rejects for TURs just above 4:1.
is below the limit
2. Guard-banding as per UKAS M3003-M2
The formula to calculate the Test (or
Tolerance) limit TL at Guard-band is:
As the uncertainty is normally stated at a
coverage probability of approximately 95%,
Eq. (1)
statements of compliance will generally be
given at that level. But this method (explained
Table 2 shows different values of TL & GB
in 1) sets quite a wide Guard-band & hence
with this method.
the suppliers run higher risk of False- Reject. A
method suggested in UKAS M3003[8] section
Table 3 shows application of uncertainty
M2 allows compliance or non-compliance
for deciding pass/fail status. Fig. 2 may be
at 95% condence level as well as at other
referred for decisions regarding the Pass,
condence levels. (It assumes, as per cl.
Fail, Pass1 & Fail1 status* (the last column
M2.18, that the uncertainty breaches any one
in Table 3). A Guard-band of 0.0002 V has
of the specication limits and is sufciently
been taken in this table. The decisions under
small so that an insignicant portion of the
Pass1 & Fail1 status are worth noticing & shall
distribution approaches the other limit). The
be clearly & explicitly reported as explained
higher risk of False-Reject could be reduced
under (c) above.
4
by following UKAS M3003-M2, which suggests
setting TL inside the SL by a factor, Ks, equal UKAS M3003 in section M3 suggests the
to 1.64 of Uc for 95% coverage probability. For following method in such cases, which is akin
a coverage probability of 99%., ks is equal to to RSS evaluation (The correct name rather
2.32 of Uc. Hence the formula to calculate the would be RDS-Root Difference Square).
test limit at 95% coverage probability is:
If both the uncertainty U and the
TL = SL - (Uc 1.64) Eq. (2)
specication SL are stated at the same
where, UC is Combined uncertainty. Table coverage probability, then the acceptance test
4 shows TL & GB values with this method. limits (TL) are established by the square root
of the difference of the squared uncertainty
Table 4: Values of TL for Guard-banding per
(U) from the squared specication (SL).
UKAS M3003-M2:

Hence the formula to calculate the test


limit is:
Eq. (3)

It gives False-Accept risk of less than 0.8%


for TURs from 4:1 to 1.5:1. However, the
False-Reject risk increases from 2% at TUR
of 4:1 to 8.3% at TUR of 1.5:1. Table 5 shows
different values of TL & GB with this method.

Table 5: Values of TL for Guard-banding per


This method gives small False-Accept
M3003-M3
risk and a lower False-Reject risk than earlier
method. It is appropriate when the specication
limits have been treated as absolute,
analogous to a rectangular probability
distribution, as suggested by M3003. This
may not always be the case. Further, in case
if it is not possible to decide compliance/non-
compliance at 95% CL, cl. M 2.12 states to
reduce the uncertainty, possibly by applying
corrections to the result, using more accurate
equipment or by taking the mean of a large
number of readings. Such an approach, it
says, should be taken by agreement with and 4. RSS Method :
understanding of the customer.
It appears in a paper by Deaver David[9] &
Clause M2.16 of M3003 gives values of Ks gives the following equation.
from 0.52 to 3.29 corresponding to coverage
probabilities from 70% to 99.9%. The formula
Eq. (4)
for ks relates to information about Uc, the result
of measurement & SL. Table 6 shows different values of TL &
GB using equation 4. This gives lower False-
3. Guard-banding per UKAS M3003-M3 : Accept risk; but higher False-Reject risk than
the previous method a). {Also compare this
Some manufacturers mention formula & table with that in method 1, Eqn.
specications in their manual at certain (1)}
condence level, typically at 95%. It is stated
in GUM that, when a specication is given at
a given coverage probability, then a normal
distribution can be assumed.

5
Table 6: Values of TL for Guard-banding by greater than that which would result from
RSS using a 4:1 ratio. To accomplish this, the TL
is reduced by a fraction M such that False-
Accept Risk is equivalent to that of a 4:1 TUR.
Thus, TL = SL M

The False-Accept Risk is 0.79 % for a


TUR of 4:1. With multiplier M as shown in
table 8, the False-Accept Risk is maintained
less than 0.79 % and the False-Reject Risk
is maintained under reasonable limit, provided
both uncertainty and specication are given at
95% condence level.
5. Guard-banding per NCSL
Recommended Practice 10 (RP-10) : Table 8 also provides some values of TL
& GB for different TURs. For all TURs, the
The method described in the NCSL- False-Accept risk is restricted to about 0.77%.
RP10[10] is given by the following formula:
Table 8: Values of M & TL for Guard-banding
per MIL-STD- 45662A & ANSI Z540.1

Eq. (5)
This equation can be applied for TURs
of 4:1 & worse. (For TURs better than 4:1,
formula gives TL > SL) As seen from the Table
7, for TUR of 4:1, TL = SL.

This method is slightly better than method


1 from producers risk perspective as it is not
discontinuous even just below 4:1 TUR. It
gives competitively low False-Accept Risk but The method is in the best interest of
still high False-Reject Risk. supplier as False-Accept risk is far less than
ANSI Z540.3 requirement. In contrast, the
Table 7 shows different values of TL & GB False-Reject risk being high, it is burdensome
using this equation. for the producer.

Table 7: Values of TL for Guard-banding per 7. Guard-banding by Managed Risk :


NCSL-RP 10
The new ANSI Z540.3: 2006, states that
the False-Accept risk should be capped to
2% when pass/fail decisions are to be taken.
Michael Dobbert of Agilent Technologies in his
paper[11] proposed a very innovative method to
achieve it, which caps the False-Accept Risk
to 2% with limited knowledge of the a priori
probability that a device is in-tolerance.
6. Guard-Banding Per MIL-STD-45662A &
ANSI Z540.1: With this method, the False-Reject risk is
also reduced to a great extent.
The old US military standard MIL-STD-
45662A and the old ANSI Z540.1 stipulate The test limits (TL) in this method are
a TUR of more than 4:1. But when it is not established by equation (6)
possible to follow this ratio, test limits can TLPFA2% = SL U95% [1.04 e (0.38ln(TUR)-0.54)]
beset such that the False-Accept risk is no Eq. (6)
6
Using this method Table 9 shows values
of TL & GB for TURs 4:1 & below. (At TURs
above 4:1, TLs become more than 100) The
above equation can also be written as-
TLPFA 2% = SL U 95% M, where the
multiplier M is

M = 1.04 e(0.38ln(TUR)-0.54) Eq. (7)

Table 9 : Values of TL for Guard-banding by


Managed Risk
Fig. 3 : Comparison of three methods

A summary of all methods is given in the


Table 11. The SL in all tables is taken as
100 for ease & convenience. The TL is the
proportion of SL & be suitably modied as per
actual specications used.

Table 11: Summary of values of TL for


various methods discussed.

The values for M are given in Table 10.

Table 10: Values of Multiplier M for Managed


risk method

# TL here is curtailed to 100 for TURs above 4:1

Fig. 4 is a graphical presentation of Table


11. The interpolation is done by drawing smooth
curve. More intermediate values of TUR would
give smoother curve & interpolation would be
more ne. Also note that, for TURs better that
4:1, graph for Managed risk is curtailed up-to
100 as their calculated values are more than
100. For NCSL RP-10, MIL-STD-45662A &
The graphs in Fig. 3, from Michael ANSI Z540.1, TURs better than 4:1 are not
Dobberts paper, show the comparison of No- applicable.
Guard-band, 95% uncertainty Guard-band &
Managed risk Guard-band. The advantages
of the Managed risk method are obvious
from the graphs for both the supplier & the
consumer. The method is very inventive as
it meets the ANSI Z 540.3 criteria of False-
Accept risk not more than 2% on one hand &
balances the false-reject risk on the other.

Fig. 4: Tolerance Limit (normalised to 100 %)


vs. TUR for various methods

7
CONCLUSION: below; while ILAC G-8, UKAS M3003 & RSS
methods can be applied up-to 10:1 TUR.
There are various methods being adopted
for deciding a Guard-band. Depending upon All methods are pertinent but shall be
the Guard-banding technique chosen, both used in right context. The method to be
False-Accept & False-Reject risks vary followed shall be an informed decision with
considerably. the customer in the loop. The signicance,
criticality and reliability of the application (e.g.
ILAC-G8, M3003-M2, and RP-10 methods space, military or industrial) as well as the
are based on the conditional False-Accept risk nancial implications should be factored in
while Managed Risk, & M3003-M3 methods while adopting any of the methods.
are based on the unconditional False-Accept
risk. The ASME B89.7.3 working group[12]
states that the selection of a decision rule is a
The 95% expanded uncertainty guard business decision, and the exibility of having
band as per ILAC-8, is the widest one & a continuum of rules ranging from stringent to
offers lowest False-Accept risk but highest relaxed acceptance or rejection is needed in
False-Reject Risk. ILAC-G8 allows other than order to satisfy a broad range of industries.
95% coverage probability for the expanded
uncertainty subject to agreement between I sincerely thank Shri. Gautam Pal,
the laboratory/supplier and the customer. It Director, ETDC, Pune for the encouragement
also states that coverage probabilities for & support provided to publish this paper.
the expanded uncertainty higher than 95 %
might be chosen while lower values should be REFERENCES :
avoided.
[1] ISO17025:2005, General requirements
M3003-M2 comes next to ILAC-G8 method for the competence of testing and
as regards Guard-band & closely follows it calibration laboratories.
throughout TURs from 1.5:1 to 10:1.
[2] ANSI/NCSL Z540.3-2006, Requirements
RSS and RP-10 methods provide closer for the Calibration of Measuring and
Guard-bands than M3003-M2 & ILAC-G8, Test Equipment, National Conference of
and follow closely till a TUR of 1.5:1 to 4:1 & Standard Laboratories, 2006.
provide a better False-Reject Risk than 95%
expanded uncertainty guard band but still are [3] Guide to the Expression of
on higher side compared to UKAS M3003-M3 Uncertainty in Measurement (GUM),
& Managed risk methods. BIPM, IEC, IFCC, ISO, IUPAC, IUPAP,
OIML - International Organization for
Managed Risk and UKAS M3003-M3 Standardization, 1995.
methods provide comparable results with
a controlled False-Accept Risk. The False- [4] Military Standard 45662A-Calibration
Reject risk is signicantly lower than ILAC-G8, System Requirements, United States of
M3003-M2, RSS and NCSL RP-10 methods. America, Department of Defence, 1988.
Both can be applied either with specication
quoted as absolute limit or when specication [5] ANSI/NCSL Z540.1-1994, Calibration
are stated with normal distribution. Laboratories and Measuring and Test
Equipment - General Requirements,
MIL-STD-45662A & ANSI Z 540.1 give National Conference of Standard
smallest Guard-band & hence pose highest Laboratories International, 1994.
False-Accept risk among Guard-banding
methods discussed. MIL-STD-45662A, ANSI [6] ILAC-G8:03/2009: Guidelines on
Z 540.1, NCSL RP-10 & Managed Risk the Reporting of Compliance with
methods are applicable for TURs from 4:1 & Specication.

8
[7] ISO14253-1:1998, Geometrical Product [11] Dobbert, Michael, A Guard-Band
Specications (GPS) -- Inspection Strategy for Managing False-Accept
by measurement of workpieces and Risk, 2008 NCSL International
measuring equipment -- Part 1: Decision Workshop & Symposium.
rules for proving conformance or non-
conformance with specications, [12] ASME B89.7.3.1-2001 Guidelines
International Organization for for Decision Rules: considering
Standardization, 1998. Measurement Uncertainty in Determining
Conformance with Specications.
[8] UKAS M3003:2007, The Expression
of Uncertainty and Condence in [13] Marcello Lucano, Differences in
Measurement, United Kingdom Guard-banding Strategies-A Beginners
Accreditation Service, 2007, pp 69-73. Guide Agilent Technologies Italia S.p.A.

[9] Deaver, David K., Guard-banding and [14] Deaver, David K., How to Maintain
the World of ISO Guide 25. Is There Only Your Condence (In a World of Declining
One Way? 1998 NCSL International Test Uncertainty Ratios), 1993 NCSL
Workshop & Symposium. International Workshop & Symposium

[10] NCSL RP-10: Establishment and


Operation of an Electrical Utility Metrology
Laboratory, National Conference of
Standard Laboratories, 1991.

Das könnte Ihnen auch gefallen