MEASUREMENT SCIENCE REVIEW, 17, (2017), No.

6, 269-272

Journal homepage: http://www.degruyter.com/view/j/msr

Verification of the Indicating Measuring Instruments Taking

into Account their Instrumental Measurement Uncertainty
Igor Zakharov1, Pavel Neyezhmakov2, Olesia Botsiura3
1
Kharkiv National University of Radio Electronics, Department of Metrology and Technical Expertise, Faculty of
Infocommunications, University, Nauky Ave., 14, 61166, Kharkiv, Ukraine, [email protected]
2
National Scientific Centre “Institute of Metrology”, Myronosytska Str., 42, 61002, Kharkiv, Ukraine,
[email protected]
3
Kharkiv National University of Radio Electronics, Department of Higher Mathematics, Faculty of Information-Analytical
Technologies and Management, University, Nauky Ave., 14, 61166, Kharkiv, Ukraine

The specific features of the measuring instruments verification based on the results of their calibration are considered. It is noted that, in
contrast to the verification procedure used in the legal metrology, the verification procedure for calibrated measuring instruments has to
take into account the uncertainty of measurements into account. In this regard, a large number of measuring instruments, considered as
those that are in compliance after verification in the legal metrology, turns out to be not in compliance after calibration. In this case, it is
necessary to evaluate the probability of compliance of indicating measuring instruments. The procedure of compliance probability
determination on the basis of the Monte Carlo method is considered. An example of calibration of a Vernier caliper is given.

Keywords: Probability of compliance; uncertainty of measurement; calibration; verification; uncertainty budget; maximum permissible error.

1. INTRODUCTION limits of the tolerance region, respectively; y and U – the

The items 5.6.2.1.1 and 5.10.4.1.b of ISO 17025:2005 [1] estimate of measurand and expanded uncertainty,
prescribe that calibration certificates for measuring respectively.
instruments (МIs) shall contain “the measurement results, These expressions are combined into one, in which the
including the measurement uncertainty and/or a statement of measurement result is within the conformity region [5]:
compliance with an identified metrological specification”.
From this requirement, it follows that the presence of an LSL + U ≤ y ≤ USL − U . (1)
indication of conformity of the calibrated MI to the
established metrological requirements or separate A measuring instrument may be an indicating measuring
metrological characteristics is necessary in the calibration instrument (IMI) or a material measure. The measurand of
certificate. The conformity assessment of the gauge IMI calibration is the systematic error E X . In verification of
according to the specified requirements is considered in a an IMI, the modulus of the specification limits of tolerance
number of documents [2]-[8]. The main requirement of region is equal to its maximum permissible error (MPE):
these documents is the need to take into account the
uncertainty of measurement when performing conformity
=
LSLIMI =
USLIMI MPE .
assessment.

2. SUBJECT & METHODS It should be noted that the main sources of uncertainty of
IMI calibration are: instrumental uncertainty of the
2.1. Conformity region and probability of compliance
measurement standard, its instability, changes in its
Conformity with a specification is proved when the complete operating conditions, mutual influence of the measurement
measurement result falls within the tolerance region [5]: standard and the IMI to be calibrated; the observed variation
in the readings of the calibrated IMI; resolution of a
LSL ≤ y − U or y + U ≤ USL , displaying device of IMI. With all uncertainty components
taken into account, the extended measurement uncertainty
where LSL and USL – the lower and upper specification during calibration may be greater than MPE.
_________________
DOI: 10.1515/msr-2017-0033

269
 MEASUREMENT SCIENCE REVIEW, 17, (2017), No. 6, 269-272

In the example S10 of calibration of a Vernier caliper with Table 2. Uncertainty budget of calibration of Vernier caliper
a resolution of 0.05 mm, considered in [9], the measurement with measuring force control.
model is:
quantity estimate standard probability sensitivity uncertainty
E X= liX − lS + LS ⋅ α ⋅ ∆t + δliX + δlM , uncertainty distribution coefficient contribution
Xi xi u ( xi ) ci ui ( y )
where liX – indication of the caliper; lS – length of the 150.10
liX mm
– – – –
actual gauge block; LS – nominal length of the actual 150.00
lS 0.46 μm rectangular - 0.46 μm
gauge block; α – average thermal expansion coefficient of
-1.0
mm
the caliper and the gauge block; ∆t – difference in ∆t 0 1.15 K rectangular 1.7 μmK-1 2.0 μm
temperature between the caliper and the gauge block; δliX – δliX 0 14.4 μm rectangular 1.0 14.4 μm
correction for the finite resolution of the caliper; δlM – δlM 0 3.3 μm triangular 1.0 3.3 μm
correction for the mechanical effects.
The uncertainty budget is given in Table 1. There are 2 EX 0.10 mm – – – 15 μm
dominating rectangular contributions in this budget.
Therefore, the expanded measurement uncertainty (for
trapezoidal distribution law) was: It should be noted that the expression (1) is true for
probability compliance of no more than 0.95. In general, in
U=
k ⋅ u ( E X ) =⋅
1.83 0.0325 mm ≈ 0.06 mm. the documents [7]-[8] it is proposed to evaluate the
probability of compliance of IMIs in the following way:
The coverage factor 1.83 for trapezoidal distribution is 
substituted in clause S10.10 [9].  USL − y   MPE − E X 
pc =
ΦN  =ΦN  =Φ N ( z ) , (2)
Table 1. Uncertainty budget of Vernier caliper calibration [9].  u   u 
 
quantity estimate standard probability sensitivity uncertainty
uncertainty distribution coefficient contribution where Φ N ( z ) – standard normal distribution function with

Xi xi u ( xi ) ci ui ( y ) variable z , E X , u – estimation of IMI’s indication error
liX 150.10
– – – –
and its standard uncertainty, respectively.
mm To find Φ N ( z ) , it is proposed [8] to use the standard
150.00
lS 0.46 μm rectangular -1.0 - 0.46 μm normal distribution table (p. 53). However, when calibrating
mm
a large number of IMIs, such as Vernier calipers, the
∆t 0 1.15 K rectangular 1.7 μmK -1
2.0 μm
distribution function attributed measurand is often
δliX 0 14.4 μm rectangular 1.0 14.4 μm trapezoidal or even rectangular. This is due to the fact that
the dominant sources of uncertainty of the calibrated IMI are
δlM 0 29 μm rectangular 1.0 29 μm
often rectangular distributed corrections, such as the
EX 0.10 mm – – – 32.44 μm correction of the Vernier caliper resolution [9].
The cumulative distribution function (CDF) of the
trapezoidal distribution, which is a convolution of two
So, such a Vernier caliper will be unusable even if its uniform distributions with the ratio of standard
readings do not deviate from the value of the end length =
uncertainties γ u2 u1 ≤ 1 , has the form:
gauge, since it is usually equal to its resolution for MPE of
the Vernier caliper. If we neglect the unjustifiably high
uncertainty associated with the influence of the measuring 0, z < −A;
force (this is quite true for the Vernier calipers with a 
[ z 1 + γ + 3(1 + γ )]
2 2
measuring force control), this will lead to a reduced
uncertainty (Table 2.). There is only one dominating  , − A ≤ z < −B;
24γ
rectangular contribution in this budget. Therefore, expanded 
uncertainty in this case will be:  z 1 + γ 2 + 3
= FT ( z )  , − B ≤ z < B; (3)
U= k ⋅ u( EX ) =0.95 3 ⋅ 0.15 = 0.0247 mm.  2 3
 [ 3(1 + γ ) − z 1 + γ 2 +]2
The coverage factor 0.95 3 for rectangular distribution
1 − , B ≤ z < A;
was taken from the formula (S9.8) [9].
 24 γ

It should be noted that even in this case the condition 1, z ≥ A.
U<MPE/3, given in [3], is not observed. 

270
 MEASUREMENT SCIENCE REVIEW, 17, (2017), No. 6, 269-272

where= A 3(1 + γ ) 1 + γ 2 ;=
B 3(1 − γ ) 1 + γ 2 . 0, z < − 6;
For trapezoidal distribution with γ =0.5 (Table 1.) and 
( z + 6) 12, − 6 ≤ z < 0;
2

=
considering z ( MPE − E X ) u , we have: FT ( z ) =  (5)
1 − ( 6 − z ) 12, 0 ≤ z < 6;
2

EX
at =
0.05
=0, z = 1.538 , that is= pc 0.936 < 0.95 ; 
0.0325 1, z ≥ 6.

EX 0.025
at =
=0.025, z = 0.769 , that is pc = 0.75 ; The formulas (2) - (4) are obtained for rectangular,
0.0325 triangular and trapezoidal distributions of variable with zero

at E X =0.05, z = 0 , therefore pc = 0.5 . expectations and unit standard deviations. However, these
models are only an approximation of the real law of
The rectangular CDF has the form: distribution obtained as a result of calibrations. In those
cases, we recommend to evaluate the probability of
0, z < − 3; compliance with help of the Monte Carlo method [10].


FR ( z=
) ( z + 3) 2 3 , z ∈ [− 3; 3]; (4) 2.2. Monte Carlo procedure

1,
 z > 3. Monte Carlo procedure for construction of distribution
function includes the following operations, registered in
uncertainty budget (Table 3.):
For the rectangular distribution and data of Table 2., we
have: Table 3. Uncertainty budget.
 0.05
=
at E X =0, z = 3.33> 3 , that is pc = 1 ,
0.015 input standard probability sensitivity uncertainty
estimate
 0.025
quantity uncertainty distribution coefficient contribution

at E X =0.025,= z = 1.67 < 3 , that is pc = 0.98 , X1 x1 u ( x1 ) PDF 1 с1 u1 ( E X )
0.015 

at E X =0.05, z = 0 , therefore pc = 0.5 . X2 x2 u ( x2 ) PDF 2 с2 u2 ( E X )
Thus, a caliper will be usable with a probability of more : : : : : :

than 0.95 if its readings do not deviate from the value of the XN xN u ( xN ) PDF N сN uN ( EX )
end length gauge or equal to 0.5 MPE. Practice shows that
the number of such calipers is about 60 % of those arrived at combined
output coverage coverage expanded
the test. Thus, 40 % of the verified Vernier calipers are estimate standard
quantity probability factor uncertainty
unusable. uncertainty
 
Fig.1. shows the CDF for the uniform, triangular, EX EX uс ( E X ) 0.95 k U
trapezoidal, and normal distribution laws.
1. Recording the model equation:

E X = f ( X 1 , X 2 ,..., X N ) , (6)

where X 1 , X 2 ,..., X N - input quantities (first column of the

Table 3.).
2. Evaluation of the input quantities as x1 , x2 ,..., xN
(second column of the Table 3.).
3. Evaluation of standard uncertainties of the input
quantities as u ( x1 ), u ( x2 ),..., u ( xN ) (third column of the
Table 3.).
4. Assigning the probability density functions (PDFs) for
Fig.1. CDF F ( z ) for uniform (∙∙∙), trapezoidal (---) with γ = 0.5, input quantities (fourth column of the Table 3.).
triangular (−∙− ), and normal (─) distribution laws. 5. Selecting the number M of Monte Carlo trials to be
made ( M ≥ 104 ).
From Fig.1. it is seen that the CDF for the triangular and 6. Generating M trials of measurand for vector, by
normal laws practically coincide (with an error of no more
sampling from the assigned PDFs as realizations of the (set
than 2 % in probability), therefore, instead of a table with
of N) of the input quantities X i .
values of the normalized normal distribution given in [8],
one can use the dependence for the triangular distribution 7. For each such vector, forming the corresponding model
law: value of E X , yielding M model values E Xi .

271
 MEASUREMENT SCIENCE REVIEW, 17, (2017), No. 6, 269-272

8. Calculation of an estimate E X of E X by the formula: When carrying out verification of the IMIs applied in the
legal metrology, it would also be necessary to take into
M account the uncertainty of measurements.
1
EX =
M
∑E
i =1
Xi . (7)
REFERENCES
[1] International Organization for Standardization. (2005).
9. Calculation of unbiased estimate *
E Xi using the General requirements for the competence of testing
formula: and calibration laboratories. ISO/IEC 17025:2005.
[2] EURACHEM/CITAC Working Group. (2012).
Quantifying Uncertainty in Analytical Measurement.
E=
*
Xi E Xi − E X . (8) EURACHEM / CITAC Guide CG4, Third Edition.
ISBN 978-0-948926-30-3.
10. Sorting these M model values E Xi *
into strictly [3] International Organization for Standardization. (2003).
Statistical methods – Guidelines for the evaluation of
increasing order, using the sorted model values to provide an
conformity with specified requirements – Part 1:
implementation of the propagation of distributions G [10].
General principles. ISO 10576-1:2003.
11. Calculating the values of probability
[4] EURACHEM/CITAC Working Group. (2007). Use of
uncertainty information in compliance assessment.
p (=
i ) 100 ⋅ i M , i =1…M, EURACHEM / CITAC Guide, First Edition.
[5] International Organization for Standardization. (2013).
* Geometrical product specifications (GPS) – Inspection
which correspond to the values of E Xi .
by measurement of workpieces and measuring
12. Construction of dependence p (i ) = E Xi .
*
equipment – Part 1: Decision rules for proving
13. Finding the probability of compliance pc for the conformity or nonconformity with specifications. ISO
value of 14253-1:2013.
[6] Eurolab. (2008). Determination of conformance with
specifications or limit values with particular reference
= *
E Xi MPE − E X . to measurement uncertainties – possible strategies.
EUROLAB ”Cook Book” – Doc No. 8.0.
Realization of the steps 6-8 of the above described Monte www.eurolab.org/documents/Cookbook_No_8.pdf.
Carlo procedure gives the dependence pc (MPE − E X ) , [7] Joint Committee for Guides in Metrology. (2012).
Evaluation of measurement data – The role of
represented in Fig.2. measurement uncertainty in conformity assessment.
JCGM 106:2012.
[8] International Organization of Legal Metrology. (2017).
The role of measurement uncertainty in conformity
assessment decisions in legal metrology. OIML G 19,
Edition 2017 (E).
[9] EA Laboratory Committee. (2013). Evaluation of the
Uncertainty of Measurement In Calibration. EA-4/02
M: 2013.
[10] Joint Committee for Guides in Metrology. (2008).
Evaluation of measurement data – Supplement 1 to the
“Guide to the expression of uncertainty in
measurement” – Propagation of distributions using a
Monte Carlo method. JCGM 101:2008.
[11] International Organization for Standardization. (2003).
Statistical methods – Guidelines for the evaluation of
Fig.2. Dependence pc of MPE − E X .
conformity with specified requirements – Part 1:
General principles. ISO 10576-1:2003.
3. CONCLUSIONS
The approaches for compliance probability determination
of the IMIs with the specification requirements taking into Received September 9, 2017.
account the uncertainty of the measurements for the Accepted November 13, 2017.
abnormal laws of distribution of their error are presented.
The examples adduced in the article show that for all
distribution laws the condition U <MPE/3 given in [3] is not
observed.

272