Sie sind auf Seite 1von 7

E L S E VI E R Measurement 16 (1995) 51 57

Measurement
On the application of the Guide to the Expression of Uncertainty
in Measurement to measuring instruments
E r n e s t o A r r i a , , , F r a n c o C a b i a t i b, S a v e r i o D ' E m i l i o b, L u i g i G o n e l l a a
a Politecnico di Torino, Corso Duca Degli Abruzzi, 24, 1-10129 Torino, Italy
b I EN Galileo Ferraris, Torino, Italy
Abstract
The problem situation related to the calibration and use of measuring instruments is addressed in general terms
with reference to the ISO Guide to the Expression of Uncertainty in Measurement. Some relevant concepts and
definitions, not explicitly addressed in the Guide, are considered, such as the role of the compatibility of different
measurements of the same measurand and the calibration diagram of the measuring instruments. These additional
concepts are expressed in terms coherent with the philosophy of the Guide. Situations commonly met in measurement
practice, such as calibration at different levels of accuracy, evaluation of traceability levels throughout the measurement
hierarchy (from the primary standards to the calibration laboratories and shop floor tests) and interlaboratory
comparisons, are analyzed and discussed.
Keywords: Measurement uncertainty; Measurement compatibility; Measuring instrument; Calibration diagram;
Traceability level; Interlaboratory comparison
1. Introduction
The Gui de t o the Expr essi on of Uncer t ai nt y in
Meas ur ement [ 1] represent s the mos t compr e-
hensi ve document agreed on at i nt er nat i onal level
and consi st ent wi t h t he r ecommendat i ons of the
Wor ki ng Gr o u p convened by BI PM in 1980 and
of CI P M in 1981 and 1986. The Gui de, as st at ed
in its scope, is meant t o be appl i cabl e t o a l arge
spect r um of scientific and t echni cal appl i cat i ons,
"f r om the shop fl oor t o f undament al research",
i ncl udi ng bot h the cal i br at i on of meas ur i ng i nst ru-
ment s and tests t hr oughout a nat i onal measur e-
ment system, and t he devel opment , mai nt enance
and compar i s on of reference st andar ds. However ,
t he act ual di scussi on and t he exampl es are limited
to the eval uat i on of t he uncer t ai nt y in i ndi rect
*Corresponding author.
0263-2241/95/$09.50 1995 Elsevier Science B.V. All rights reserved
SSDI 0263-2241(95)00017-8
meas ur ement s where the meas ur and is funct i onal l y
tied to di rect l y measur ed quant i t i es. I n the det ermi -
nat i on of t he uncer t ai nt y component s of these
var i ous di rect meas ur ement s a f undament al rol e is
pl ayed by the uncer t ai nt y due to t he di r ect - r eadi ng
i nst r ument itself, a rol e t hat overri des all ot hers in
the case, t he mos t usual in i ndust ri al pract i ce,
where t he meas ur ement is carri ed out by a single
meas ur ement action.
The i nst r ument al uncer t ai nt y is referred to by
t he Gui de as a dat um suppl i ed by t he manuf ac-
t urer, mat hemat i cal l y t reat ed as a squar e di st ri bu-
tion. Of course the Gui de' s aut hor s assume t hat
the eval uat i on of such an i mpor t ant i nst r ument al
charact eri st i c is t o be per f or med t hr ough a cal i bra-
t i on process. However , in deal i ng wi t h measur i ng
i nst rument s, mor e at t ent i on shoul d be pai d to this
cal i br at i on process and t o the rol e of t raceabi l i t y.
Obvi ousl y, some concept ual and t er mi nol ogi cal
52 E. Arri et al./Measurement 16 (1995) 51-57
integration is required in additional documents to
be issued to address these topics. We are here
endeavouring to suggest a path for these docu-
ments, which are needed as a frame of reference
for the standards related to specific types of
instruments.
To this purpose, we must add to the terms and
definitions given in the Guide and the last edition
of VIM [2] a few terms and concepts concerning
the calibration and use of measuring instruments,
the measurement compatibility and its role in
evaluating measurements of the same measurand
carried out by different instruments, the traceability
through the successive steps of any hierarchical
chain of measurement. For such an integration we
shall follow the operative approach adopted in the
Italian frame standard on terms and definitions in
measurement [3] and the proposal presented to
IEC TC 66 [4]. For some terms and concepts
concerning the national standards and the dissemi-
nation activity more specific references are con-
sidered [5-7]. To some extent, these documents
are in substantial agreement with the lines of the
Guide, even if some mismatch is present and needs
to be overcome in future specific standards.
As stated in the Guide, the measurement uncer-
tainty is a parameter characterizing "the quality of
a result of a measurement" (clause 0.1) or "the
dispersion of the values that could be reasonably
attributed to the measurand" (clause 2.2.3 and
VIM entry 3.9). This parameter is expressed,
according to the recommendation of the CIPM
Working Group, by a value related to the estimated
standard deviation. In the standard [3] the uncer-
tainty is considered an integral part of the informa-
tion given by the measurement result, rather than
an additional or supplementary part of this infor-
mation, to the point that the measurement result
is not expressed just by a single value, but by a
whole interval of values. This interval is identified
by a measurement value (usually its mid element)
and its measurement uncertainty (usually the half
width of the interval). Both approaches renounce
a more detailed description of the uncertainty (e.g.
statistical distribution, fuzzy set), which would
depend on the specific features of the measurement
conditions, and limit the information on the uncer-
tainty to its amplitude only, for the sake of simplic-
ity and uniformity.
2. Additional concepts and definitions
The statement in clause 1.2 "This Guide is
primarily concerned with the expression of uncer-
tainty in the measurement of a well defined physical
quantity, the measurand, that can be characterized
by an essentially unique value" is in line with the
concept of true value, that, however, the Guide
overcomes for the reasons expounded in Annex D.
In fact, every measurand has an intrinsic uncer-
tainty, given by the minimum uncertainty that can
be assigned in its description (clause 2.3.1 in [3]).
If one has to measure a given quantity with uncer-
tainty lower than its own intrinsic uncertainty, one
is compelled to redefine the quantity using a more
sophisticated model. In any case, the intrinsic
uncertainty has to be included as a component of
the total uncertainty, as acknowledged in clause
D.I.1 of the Guide. From this point of view, the
measurement uncertainty reflects both the level of
definition and the level of knowledge of the measur-
and rather than just "the lack of exact knowledge
of the value of the measurand", as stated in the
Guide (clause 3.3.1). The result of a measurement
carried out with an uncertainty equal to the intrin-
sic uncertainty of the measurand may be called the
best measurement of the measurand (clause 2.3.1 in
[3]). The fact that a quantity can be defined only
at a finite level of detail makes the true value
concept questionable even before the measurement
uncertainty is accounted for. As well known, even
in dealing with primary or national standards, the
metrologists evaluate and declare the uncertainty
associated with the value(s) of each standard [5].
Another basic concept strictly linked to the
uncertainty, mentioned but not expounded in the
Guide (clause 1.4), is the measurement compatibility.
This is an important concept because from the
operative point of view it may be taken as the
basis of the definition of uncertainty. The concept
of measurement compatibility is in complete
agreement with the CIPM Recommendation,
which eliminates the traditional concept of true
value of the measurand, substituted in the operative
E. Ar r i et a l . / Me a s u r e me n t 16 ( 1 9 9 5 ) 5 1 - 5 7 53
appr oach by t hat of best measur ement of the
measurand. Specifically, t he measur ement compat i -
bility plays a cent ral role in evaluating:
- measur ement s of the same measur and in the
same measur ement condi t i ons obt ai ned using
different i nst rument s or by different l aborat ori es
(e.g. i nt er compar i sons of nat i onal st andards,
i nt er l abor at or y compar i sons in t he f r amewor k
of cal i brat i on services);
- t r a c e a b i l i t y checks in t he accredi t at i on of
cal i brat i on and testing l aborat ori es, whose final
aim is t o assure t he i nt erchangeabi l i t y and integ-
rat i on of product s.
In the operat i ve appr oach pr omot ed in [ 3] and
confi rmed in [ 4] , a measuri ng i nst r ument can be
represent ed as a funct i onal block, which ei t her
(a) supplies an out put readi ng dependent on the
quant i t y appl i ed to its input, or
(b) makes available an out put quant i t y identified
by a nomi nal value.
In the VI M t he i nst rument s of t ype (a) are called
"measuri ng t ransducers or i ndi cat i ng and record-
ing i nst r ument s" and the i nst rument s of t ype (b)
are called "mat eri al measures". The l at t er ones
include an increasing number of i nst rument s, e.g.
the mul t i funct i on and mul t i range calibrators,
where the nomi nal value may be single or selected
within a set of discrete or cont i nuous values. The
t r eat ment of uncer t ai nt y and t he cal i brat i on pro-
cess are essentially the same for bot h types of
i nst rument s, because the readi ng of type-(a) instru-
ment s plays the same role as the nomi nal val ue of
t ype-(b) instruments.
A clear di st i nct i on has t o be made bet ween the
r e adi ng val ue r of an i nst rument (or the nomi nal
v al ue of a mat eri al measure) and the me a s u r e me n t
r e s ul t M = I r a - u, m + u] assigned t o represent t he
measurand, where m is the me a s u r e me n t val ue and u
t he me a s u r e me n t u n c e r t a i n t y . The rel at i on between
the readi ng val ue r and t he measur ement result M
is det ermi ned t hr ough a cal i brat i on pr ocedur e that
involves a cal i brat ed i nst rument , havi ng in general
an uncer t ai nt y l ower or negligible in compar i son
with t hat of the i nst r ument under cal i brat i on. Thi s
is a general rule, except in t he br anch of met r ol ogy
t hat concerns the devel opment of t he pr i mar y
standards. In fact, these i ncor por at e a real i zat i on
of t he relevant SI units, for which t he knowl edge
of t he realized physical quant i t y cannot rely
directly on the cal i brat i on of anot her reference
st andard, but can be val i dat ed t hr ough inter-
nat i onal compari sons.
The result of a cal i brat i on pr ocedur e is generally
issued in t he form of tables. However, t hi nki ng in
t erms of a di agram of M versus r can be useful
from t he concept ual poi nt of view: accordi ng t o
[ 3, 4] , this di agr am is named c al i br at i on di agr am.
Due t o t he looseness i nt r oduced by uncer t ai nt y u,
the di agram must be conveni ent l y t hought of as a
strip di agram (see Fig. 1 ).
For the maj ori t y of moder n i nst rument s, the mid
line of the cal i brat i on di agram ( cal i br at i on curve)
is close to a straight line and can be expressed by
a par amet er dependi ng on the readi ng value r.
This par amet er may be either the c al i br at i on f a c t o r
k(r) = m/ r or the fractional devi at i on d(r) from the
straight-line value knr, where kn is the nomi nal
val ue of k. In the l at t er case it is
d(r) = ( m - k , r ) / k , r , m = k , r [ 1 + d(r)]. ( 1 t
In the cal i brat i on process, the same measur and
is put in common to the i nst rument under test and
to t he measuri ng system assumed as reference,
whose cal i brat i on di agr am is known. As a result,
the cal i brat i on di agram of the i nst r ument under
test can be generat ed from the readings of the t wo
Measurement
values
/ ~ calibration curve
m +
m - u
i
I D -
r Reading wdues
Fig. 1. Conceptual representation of the calibration diagram,
obt ai ned by combi ni ng t he reading values (r) with the
correspondi ng measurement values (m) and the associated
uncertainty (u). The calibration di agram is intended as a strip
diagram whose mid line is the calibration curve.
54 E. Arri et al./Measurement 16 (1995) 51- 57
i nst rument s and using t he known cal i brat i on dia-
gram. The i nst rument s of t ype (a) and (b) can be
combi ned in different ways, as bot h t he reference
and t he i nst r ument under test can bel ong to
ei t her type.
It is i nt erest i ng t o not e t hat onl y readi ng values
r and measur ement values m are i nvol ved in the
cal i brat i on process, while no assumpt i on is needed
beside the val i di t y of t he cal i brat i on di agram of
the i nst r ument used as reference and t he measur-
and stability duri ng t he calibration. Measur ement
i nt eract i ons are our onl y source of knowl edge on
t he physical quantities, and t he cal i brat i on of an
i nst r ument against known measur ands is t he onl y
source of i nf or mat i on on the rel at i on bet ween t he
readings r and t he measur ement values m of the
measurands. Therefore, at the ul t i mat e levels of t he
cal i brat i on chai n the stability can onl y be ascer-
t ai ned by t he met rol ogi cal experi ence on t he over-
all consi st ency of the readi ngs of t he reference
systems i nt eract i ng with given measurands. At t he
next levels of cal i brat i on, one can assume t hat t he
reference systems and t he measur ands are stable,
i.e. t hat t hei r vari at i ons are negligible with respect
t o the uncer t ai nt y of t he i nst r ument under test, or
at least one knows enough about such vari at i ons
(with time or ot her condi t i ons) as t o be able t o
account for t hem when eval uat i ng t he uncer t ai nt y
u in t he cal i brat i on diagram.
It is wor t h poi nt i ng out t hat t he knowl edge of
t he measur and is t he nor mal ai m of scientific and
technical measurement , whereas in the field of
appl i ed met r ol ogy one most l y deals with "known"
measur ands and "unknown" instruments: t he aim
of t he measur ement is not t o assign a val ue t o a
physical quant i t y but t o assign a cal i brat i on dia-
gram to an i nst rument . Thi s makes evi dent the
i mpor t ance of the cal i brat i on di agram, what ever
t he f or m adopt ed for the cert i fi cat i on of cal i brat i on
results is.
The me a s u r e me n t c o mp a t i b i l i t y is defined in [ 3,4]
as t he pr oper t y of measur ement results of the
same measur and consisting in t hei r being expres-
sed by non-di sj oi nt intervals. Accordingly, t wo
measur ement results M l =[ ml - ux, m l + u l ] and
M z = [m E -- u2, m 2 -n t- U2] are compat i bl e if
Ira1 - m z l ~ < u ~ --~-u 2 . (2)
A different cri t eri on is suggested by WECC in
connect i on with t he eval uat i on of an i nt erl abora-
t or y compar i son [ 7] :
I m l - m 2 1 < ux/~l +Uez. ( 3 )
In bot h cases the difference of t he measur ement
values m~ and m2 is compar ed with a combi nat i on
of the uncert ai nt i es Ul and u2 of M~ and M2. A
mor e general definition of compat i bi l i t y, valid not
onl y in the case of compl et el y i ndependent meas-
urement s but also when t hey are part i al l y corre-
lated, coul d be
2 2
I ml - m21 < u~2 = x / u~ + U 2- 2 R l z u l u 2 ,
( 4 )
where u12 is the uncer t ai nt y of the difference
ml - m2 eval uat ed accordi ng t o the Gui de, and R12
is t he cor r el at i on coefficient bet ween M~ and M 2.
Fr om this rel at i on one can derive the ones valid
in any specific situation, e.g. for measurement s
compl et el y i ndependent (R12=0) or compl et el y
correl at ed (R12 = + 1). In addi t i on, a compat i bi l i t y
i ndex bet ween t wo measur ement results M~ and
M a mi ght be defined as follows:
I c = Ira1 - - mz [ / U1 2 , (5)
and M1 and M 2 are compat i bl e if I~ is not larger
t han 1. The definition (4) of compat i bi l i t y is not
far from t he classical conf or mi t y test criteria or
from t hose discussed in [-8]: in any case the
difference bet ween t wo measur ement values ml and
m2 of the same measur and is compar ed with the
est i mat ed st andar d devi at i on of this difference
multiplied by a suitable factor. The above consider-
ations show how much the concept of uncert ai nt y
is tied in with compat i bi l i t y.
In a met rol ogi cal system, t he compat i bi l i t y of
measurement s is at t ai ned t hr ough t he t raceabi l i t y
of t he measuri ng i nst rument s t o pr i mar y standards.
A commonl y used definition of traceability, in
subst ant i al agreement with the one given in [ 2] ,
is: t raceabi l i t y is the pr oper t y t hat a measur ement
i nst r ument acquires when it is cal i brat ed using
measur ands whose values and uncert ai nt i es have
been assigned with reference t o recogni zed nat i onal
or i nt ernat i onal st andards (clause 7.1.1 in [ 3] ) .
E. Arri et al./Measurement 16 (1995) 51- 57 55
3. Appl i cat i on t o s ome pract i cal s i t uat i ons
The concept of t raceabi l i t y implies a ranki ng of
the cal i brat i on l aborat ori es accordi ng to t hei r met-
rol ogi cal capabilities, from the nat i onal l abor at o-
ries, devot ed t o mai nt ai n and i mpr ove the pr i mar y
st andards, down t o t he cal i brat i on st at i ons at the
shop floor. It is wort hwhi l e to define quant i t at i vel y
a t r a c e a b i l i t y l e v e l able to assess the posi t i on of a
l abor at or y in the t raceabi l i t y chain, a level t hat is
of course different for different quant i t i es and
ranges. Thi s level may be defined by t he uncert ai nt y
resulting from the combi nat i on of the uncert ai nt i es
of the reference i nst rument s used in the cal i brat i on
pr ocedur e and the uncert ai nt i es i nt r oduced in the
cont r ol of the influence quantities, of t he coupl i ng
of the measur ands to the i nst rument under test,
etc., i.e. all t hose component s of uncer t ai nt y t hat
are i ndependent of the tested i nst rument itself. In
nor mal oper at i ons of cal i brat i on checking the
uncer t ai nt y expressing the t raceabi l i t y level ought
to be negligible with respect to the uncert ai nt y of
the i nst r ument under test. If this is not so, the
t raceabi l i t y level has to be t aken i nt o account in
the uncert ai nt y budget with procedures t hat ought
to be st andardi zed.
Within a met rol ogi cal system based on traceabil-
ity, the definition of uncer t ai nt y given in the Gui de
in general t erms (clause 2.2.3) assumes a mor e
definite meaning. In such a system, indeed, the
measur ement values "t hat can be reasonabl y at t ri b-
ut ed to the measur and", whose di spersi on is char-
act eri zed by the uncert ai nt y, are the measur ement
values t hat can be obt ai ned with a t raceabl e meas-
uri ng i nst rument .
The t raceabi l i t y level of a given l abor at or y for
a given ki nd of i nst r ument can be assessed by a
pr ocedur e usually called "audi t " or "i nt erl abora-
t or y compar i son". The pr ocedur e consists in circu-
lating a travelling st andar d of a suitable and known
cal i brat i on di agr am among a gr oup of N l abor at o-
ries, each of t hem member of an established met r o-
logical system. As the lines of t raceabi l i t y for such
a system are clearly defined, following explicitly
stated cal i brat i on chains, t he t raceabi l i t y levels are
det ermi ned. Thus, t he measur ement results Mi,
i = l , 2 , . . . , N , obt ai ned by each l abor at or y are
expect ed to be compat i bl e, so t hat the exercise
coul d be seen as a compat i bi l i t y test and t he results
eval uat ed on t he basis of a compat i bi l i t y criterion.
Mor e precisely, if M1 and M 2 a r e the results
obt ai ned by the l aborat ori es 1 and 2, the compat i -
bility index is given by (5) and (4), where now ml,
m 2 are t he measur ement values, and u~, u2 their
uncert ai nt i es eval uat ed t aki ng i nt o account the
t raceabi l i t y levels.
Fr om the N measur ement results a global meas-
ur ement result Mm can be derived, whose measure-
ment value m and uncert ai nt y Um are given by the
weighted mean of t he measur ement values m~ and
t hei r uncer t ai nt y ui. An eval uat i on of every single
result M~ coul d be based upon a compat i bi l i t y
i ndex obt ai ned using (5). If the N measur ement
results can be consi dered i ndependent , it is (see
Appendix):
I i = I m i - m [ / x / u 2 - UZm . (61
Also, a general i ndi cat i on of the efficiency of the
met rol ogi cal system comes from the mean of the
indexes Ici of all l aborat ori es.
Since the result of a cal i brat i on may be expressed
in t erms of the cal i brat i on factor, one is t herefore
easily t empt ed to t hi nk of the cal i brat i on fact or as
embodyi ng the whol e i nf or mat i on on the calibra-
tion, and to t reat the uncert ai nt y of the i nst rument
as it were an uncert ai nt y of its cal i brat i on factor,
a posi t i on from which simple algebraic t r eat ment
mi ght be derived. One must keep in mind, however,
t hat by itself t he cal i brat i on fact or does not yield
any i nf or mat i on on the uncert ai nt y of the instru-
ment, being onl y an algebraic device poi nt i ng out
the posi t i on of the mid poi nt s of the cal i brat i on
di agram of which the uncert ai nt y is the hal f wi dt h
(see Fig. 1). The nor mat i ve t er mi nol ogy ought to
make a clear di st i nct i on bet ween the operat i ons of
cal i brat i on (or re-calibrationt, meant to det ermi ne
the cal i brat i on di agram of an i nst rument , hence its
i nst rument al uncert ai nt y, and the operat i ons of
cal i brat i on check, meant onl y to verify whet her
t he act ual readings of an i nst rument fall within
the range prescri bed by its cal i brat i on diagram:
the former i mpl y the statistical el abor at i on of
readi ngs under different condi t i ons sweeping over
the whol e range of quant i t i es of influence and
ot her oper at i ng condi t i ons, while t he l at t er may
56 E. Ar r i et al . / Measurement 16 ( 1995) 51 57
be carri ed out with few readings and be simply
expressed in t erms of the cal i brat i on factor.
When a quant i t y Y depends on N ot her
quant i t i es X 1 , . . . , X N t hr ough a funct i onal rel at i on
Y = f ( X ~ , . . . , X N ) and the measur ement result of Y is
obt ai ned f r om the measur ement results of X I , . . . , X N ,
t he same funct i onal rel at i on t hat describes the
physical system, t o which the measur ed quant i t i es
bel ong as paramet ers, is assumed t o hol d also for
t he measur ement values and t he usual rules for
the eval uat i on of uncer t ai nt y in indirect measure-
ment s are applied. Specifically, t he measur ement
value y of Y is det ermi ned using the following
rel at i on bet ween the measur ement values: y =
f ( k l r l , . . . , k N r N ) , where k i r l is t he measur ement value
of X~, k~ the cal i brat i on f act or and r i the readi ng
val ue of the i nst r ument used for measuri ng or
generat i ng Xi. As suggested in clause A.1 of [ 6] ,
t he funct i on f shoul d express not simply a physical
law but the measur ement process used, and in
part i cul ar, it shoul d cont ai n all quant i t i es X~ t hat
can cont r i but e a significant uncer t ai nt y component
t o t he measur ement result (e.g. cor r ect i on factors,
quant i t i es t hat t ake i nt o account ot her sources
of vari abi l i t y such as time and measur ement
conditions).
given by the rel at i on
~{m~ m ~ _ rnN~
rh " = u" / - - ~- r "~- "'" + ~-N)
\ u~ u2
( A . 1 )
where u~ is t he uncer t ai nt y associated with m i , and
u,, is the uncert ai nt y associated with m, for which
it is
1
2 (A.2)
U m - -
1 1 1
. ~ + .75 +... + .-i-
U 1 U2 UN
Usi ng (A.1) and (A.2) the uncert ai nt y of m ~ - m
results:
U i m = N / N 2 __ Um2. (A.3)
Fr om (A.3) the expressi on (6) of t he compat i bi l -
ity i ndex bet ween Mi and M is obt ai ned. Mor eover ,
from (A.3) and (4) it follows t hat the correl at i on
coefficient bet ween M~ and M is given by
U m
Ri,, - (A.4)
U i "
4 . C o n c l u s i o n s
Fur t her concept s and definitions not explicitly
consi dered in t he I SO Gui de (such as cal i brat i on
di agr am and factor, measur ement compat i bi l i t y
and t raceabi l i t y level) seem necessary to fulfil t he
scope of t he Gui de, part i cul arl y in it bei ng suitable
to appl i cat i ons at all levels of the measur ement
hi erarchy. The utility of i nt r oduci ng these concept s
in addi t i onal , mor e specific guidelines or st andar ds
is shown by t hei r appl i cat i on t o some significant
cases of cur r ent measur ement practice.
A p p e n d i x
Let mi, with i =l , 2, . . , N, be t he measur ement
values obt ai ned by different l abor at or i es consid-
ered i ndependent . Then the weighted mean m is
It can also be not ed t hat in the part i cul ar case
N= 2 from (6) it follows:
I~1 = I~2 = I ml - m21/ V / ~ l + uZ2 ,
(A.5)
i.e. t he compat i bi l i t y indexes with the weighted
mean M of ei t her M1 or M2 are equal and, in
addi t i on, are equal to t he compat i bi l i t y index
bet ween M1 and M2.
R e f e r e n c e s
[1] BIPM, IEC, IFCC, ISO, IUPAC, IUPAP and OIML,
Guide t o t he Expressi on o f Uncert ai nt y in Measurement .
Fi r s t Edition, 1993, ISBN 92-67-10188-9.
[2] BIPM, IEC, IFCC, ISO, IUPAC, IUPAP and OIML,
Int ernat i onal Vocabul ary o f Basi c and General Ter ms in
Met rol ogy, Second Edition, 1993, ISBN 92-67-01075-1.
1-3] Norma UNI 4546, Mi sure e misurazioni. Termi ni e
def i ni zi oni f ondament al i , Novembre 1984.
[4] Document 66(Italy)ll, New Edition of IEC 359 2nd
E. Arri et al./Measurement 16 (1995) 51-57 57
Edition 1987 Expression of the Performance of Electrical
and Electronic Measuring Instruments, June 1990.
I-5] S. D' Emilio, M.L. Rastello, P. Soardo, A. Peut o and S.
Sartori, I campioni de| sistema nazionale di t arat ura, AEI
80/ 5) (1993) 36-43.
I-6] B.N. Taylor and C.E. Kuyatt, Guidelines for evaluating
and expressing the uncertainty of NIST measurement
results, NI S T Technical Note 1297, 1994 edition.
[ 7] WECC Doc. 15, Gui de for the organization of interlabora-
tory comparisons, 1992 edition.
[ 8] K. Weise and W. W6ger, A Bayesian theory of measure-
ment uncertainty, Meas. Sci. Technol. 411993) 1 11.

Das könnte Ihnen auch gefallen