Traceability and uncertainty analysis in volume measurements

S. Lorece
*
Istituto Nazionale di Ricerca Metrologica (INRIM), Strada delle Cacce, 73-10135 Torino, Italy
a r t i c l e i n f o
Article history:
Received 27 October 2008
Received in revised form 16 July 2009
Accepted 23 July 2009
Available online 22 August 2009
Keywords:
Glassware
Proving tank
Traceability
Calibration
a b s t r a c t
Laboratories must ensure that obtained results are reliable, for that it is necessary to cali-
brate glassware and proving tanks using the correct method and estimate of uncertainties.
The aim of this work is to present the gravimetric method as a suitable and traceable
method to calibrate glassware according to laboratory requirements. Moreover, the model
for the calculation of the measurement uncertainty is described and it is proved to be com-
pletely suitable for the calibration by gravimetric method through a numerical example
drawn from the INRIM uncertainty budget in the CCM.FF-K4 key comparison.
2009 Elsevier Ltd. All rights reserved.
1. Introduction
Volume measurements of liquids are important routine
operations required in many elds like health, chemistry,
biology, pharmacy, . . . and genetics, in the range from a
few lL to several m
3
.
When generally speaking about uids, the term volume
is not distinguished from capacity; both terms are gener-
ally used to refer to the amount of liquid or gas that may
be held in a container. Therefore, uids can be measured
by either weight or by volume.
It is often more accurate and convenient to measure the
volume of a liquid and then convert it to mass, if the den-
sity is known at the relevant temperature.
Many suppliers offer volumetric instruments and appa-
ratuses in varying shapes, qualities, capacity and made of
different material such as plastics, glass, . . ., and stainless
steel in order to meet the different demands, intending
both to contain and to deliver accurate quantities of liq-
uids. This is necessary in activities such as dispensing,
pipetting, . . ., titrating or where the legal metrology is in-
volved, which enters into all commercial transactions from
the trading of goods such as petroleum, natural gas or me-
tal ores in bulk to the retail sale of goods to the public in
the market place.
Many quality systems require that volumetric instru-
ments be calibrated or certied (the terms are often used
interchangeably), especially where precise measurements
are of critical importance. There are many documentary
standards and other publications, for example [19] that
provide the methods and procedures to be used with refer-
ence to different types of instruments or ow measuring
equipment. There are three general approaches to the cal-
ibration of volumetric instruments that are in wide use: (i)
geometric, (ii) volumetric, (iii) gravimetric. The choice of
the method or of the procedure is imposed by the nominal
capacity of the instrument, the shape, the conditions of
use, . . ., and on the required accuracy. However, in any
method there are some pitfalls and subtle sources of error
that must be taken into account in order to make full use of
the available accuracy.
The Italian Institute of metrology, INRIM, as well as sev-
eral National Metrology Institutes (NMIs), accepts high-
grade standards of volume for calibration. INRIM classies
such instruments as glassware, according to whether it is
specied for single volumes or multiple volumes (one-
mark or multi-mark), typically with capacity below 1 L,
and as proving tanks, typically for industrial use, with
capacity above 1 L.
0263-2241/$ - see front matter 2009 Elsevier Ltd. All rights reserved.
doi:10.1016/j.measurement.2009.07.016
* Tel.: +39 011 3919929; fax: +39 011 3919937.
E-mail address: [email protected]
Measurement 42 (2009) 15101515
Contents lists available at ScienceDirect
Measurement
j our nal homepage: www. el sevi er. com/ l ocat e/ measur ement
The INRIMprocedure for testing both volumetric instru-
ments is usually based on the gravimetric method, that is,
to weigh the amount of distilled water contained and/or
delivered with reference to the graduations marked on
the instrument, the volume being computed from the den-
sity of the water.
In recent years comparisons of primary standard mea-
surement facilities have been an essential activity of na-
tional metrology institutes as this is the best way to
detect aws in instruments and/or procedures. Several
projects of volume comparison have been carried out
among NMIs in the range 1 mL50 L [1014]. The INRIM
laboratories took part in some of these [13,14] in order to
conrm their own capabilities.
On the basis of the experience gained, although specic
applications are not covered in detail, the main purpose of
this paper is to introduce the gravimetric method to cali-
brate glassware according to laboratory requirements and
based on existing standards. Moreover the model for the
calculation of the measurement uncertainty is described
and it is proved to be completely suitable for the calibra-
tion by gravimetric method through a numerical example,
drawn from the INRIM uncertainty budget in the CCM.FF-
K4 key comparison.
2. Experimental
Volumetric glassware: burettes, pipettes, asks and cyl-
inders are measuring instruments for laboratory with dif-
ferent properties to contain (In) or to deliver (Ex). It is
common to have glassware calibrated to two different lev-
els of accuracy; Class A and Class B. Very often, though not
always, the tolerances for Class A glassware are one half of
those for Class B.
Usually the gravimetric method is preferred for glass-
ware calibration, although it is very time consuming. This
method is also used with very little difference in the proce-
dure both for determining the contained volume of a volu-
metric instrument or for measuring the volume delivered
from it.
In Fig. 1 a basic owchart related to the traceability of
volume calibration by gravimetric method is shown.
The reference value of volume/capacity is achieved at
INRIM applying a weighing procedure consisting of at least
10 weighing cycles, where in each cycle, the mass corre-
sponding to volume is obtained by the difference between
the lled instrument with distilled water at the reference
level m
2
, and the empty instrument m
1
.
The mathematical model that converts the mass of
water to volume V contained in or delivered for each mea-
surement at the temperature t is the following:
Vt
m
2
m
1

q
w
t q
a

1
q
a
q
c
_ _
1
where q
w
t is the density of water at the mean tempera-
ture value between the two temperatures, before and after
the water is weighed, q
a
the density of air during the
weighing and q
c
is the conventional value of the density
of the weights used to calibrate the balance, usually 8 g/
cm
3
.
The value of density of water at the calibration temper-
ature is computed according to the water density equation
recommended by BIPM, for temperatures between 0 C and
40 C and at a pressure of 101.325 kPa [15], the density of
air is calculated in agreement with [16], taking into ac-
count the mean value of the ambient temperature, humid-
ity and pressure, and nally the temperature t is the mean
value.
The volume at the reference temperature, usually at
20 C is:
V
20
Vt 1 c20 t 2
where c is the thermal coefcient of cubic expansion of the
material of the vessel, for example c is 10 10
6

C
1
for
borosilicate glass.
3. Main identied sources of uncertainty
The measurement uncertainty theory states that all
measurements are estimations; the best we can hope
would require that all sources of variation were identied
and controlled so that we reduce the amount of bias (or
distance from the true value) as much as possible [17].
There are several sources of inuence to consider in
measuring volume V(t) by weighing water. They are espe-
cially identied in the mathematical model given in Eqs.
(1) and (2). In principle there is also correlation between
some of these variables, e.g. due to the use of the same
instruments during the weighing or in environmental mea-
surements devoted to determine the air density, but the ef-
fect of correlation can be assumed to be negligible.
As an example, Fig. 2 shows the averaged values of the
main uncertainty components over all participating labo-
ratories as they resulted in the EUROMET project 692 In-
ter-laboratory calibration comparison of the volume of
100 mL Gay-Lussac Pycnometer [13].
Reliability of calibration and the level of uncertainty is
strictly connected with the equipment and the procedure
used in the measurement; the choice should be based on
obtaining the highest accuracy in the calibration. The main Fig. 1. Traceability chain for volume measurements at INRIM.
S. Lorece / Measurement 42 (2009) 15101515 1511
identied sources of uncertainty can be described as
follows:
3.1. Weighing of the glassware
The weighings of the empty and lled glassware are
usually performed by electronic balances; commercial bal-
ances with a resolution of 0.1 mg are suitable to obtain a
good accuracy. The standard uncertainty associated with
the weighing is obtained from contributions due to the bal-
ance calibration and the reference standards calibration,
but also from the repeatability, linearity, hysteresis, . . .,
and drifting of the measurements. Moreover, in agreement
with Eq. (1) the standard uncertainty associated with the
density of the standard weights should be considered.
3.2. Water density
The water should have the quality suitable for the pur-
pose of calibration; for most laboratory applications the
medium-quality reagent water with a resistivity of
> 1:0 MX (conductivity < 1 lS) is acceptable [18]. There
are several treatment processes used to remove all con-
taminants. The most common commercial systems use
activated carbon, microltration, ultraltration and re-
verse osmosis, followed by deionization or distillation.
Most laboratories are able to prepare the water, but very
few can afford to determine its density by direct measure-
ment. Instead, the laboratories generally use the measured
temperature and calculate the density according to one of
the density formulas based on previous experimental
works [19].
The evaluation of the uncertainty of the density of
water is the major source of uncertainty on the total cali-
bration uncertainty, Fig. 2, and probably it is also the com-
ponent more problematic. Such contribution mainly
depends on the judgment about how close the prepared
water is to the water on which the density formula is
based; other minor contributions include the evaporation
effect, the compressibility, the inuence of residual con-
taminants and, possibly, the air content.
3.3. Temperature of water
Thermal conditions play an important role during the
whole procedure; water density changes and density gra-
dients due to non-uniform temperature might have a sig-
nicant effect on volume measurement results even in a
temperature-controlled set-up. Unfortunately, signicant
temperature gradients cannot always be avoided for prac-
tical reasons, e.g. insufcient uniformity and stability of
the reference liquid and differences between the liquid
and ambient temperature. The temperature effect should
be expected in changing the apparent weight of the lled
glassware and the density of the water.
3.4. Air density
The measurement of air density is necessary to allow
buoyancy corrections to be made when weighing bodies
of different materials exhibiting different volumes, or
when making mass measurements to the highest accuracy.
For most volume measurements, the air density can be as-
sumed to be 1.2 kg/m
3
. A measurement of the actual air
density is usually only needed when it is required to know
a volume to a high accuracy, as in the case for some pyc-
nometers, whose relative expanded uncertainty can be less
than 0.0001 (0.01%). In such cases, most NMIs adopt the
Comit International des Poids et Mesures recommenda-
tion known as the CIPM81 air density equation [16]. Usu-
ally only the ambient pressure P
a
, temperature T
a
and
humidity H
a
are measured in laboratory, while it is sup-
posed that the mole fraction of CO
2
, v
CO
2
in atmosphere
be equal to 0.0004. From this equation the largest contri-
bution to uncertainty in air density is that due to pressure
measurements, followed by the temperature ones.
3.5. Others
The effect of other factors, such as the bubble formation,
the instrumental cleaning, the instrumental temperature,
the removal of the remaining water on the stopper, the sta-
tic electricity, weight differences, etc. can usually be kept
to a not signicant level by following a suitable measure-
ment protocol, or they can be taken into account all to-
gether under the entry for other inuence sources.
Specically, the volume contained in, or delivered by glass-
ware depends on the cleanliness of the internal glass sur-
face of the instrument. To attain greatest accuracy, all
glassware should be efciently cleaned, oil or grease may
be removed by suitable solvents, then the vessels must
be repeatedly rinsed with distilled water, until all traces
of detergent are removed.
3.6. Repeatability (standard deviation)
The volume measurement is performed n times (usually
at least 10 times); one of the most important uncertainty
components is the uncertainty of the operator, that is
usually included in the standard deviation of the mean.
Human errors can include inaccurate readings or misalign-
ments of the meniscus, depending on the operator skill, on
his fatigue or concentration level, as well as on the glass-
0.00000
0.00020
0.00040
0.00060
0.00080
0.00100
0.00120
0.00140
0.00160
M
a
s
s

p
i
e
c
e
s

d
e
n
s
i
t
y
W
a
t
e
r

d
e
n
s
i
t
y
A
i
r

d
e
n
s
i
t
y
E
x
p
a
n
s
i
o
n

c
o
e
f
i
c
i
e
n
t
T
e
m
p
e
r
a
t
u
r
e
M
a
s
s
f
i
l
l
e
d
O
t
h
e
r
S
t
a
n
d
a
r
d
d
e
v
i
a
t
i
o
n
Standard uncertainties
V
o
l
u
m
e

/

m
L
Fig. 2. Schematic representation of sources and main components of
measurement uncertainty, for the calibration procedure, performed
against national standards (EUROMET project 692) [13].
1512 S. Lorece / Measurement 42 (2009) 15101515
ware instruments, sharpness of their scale, and nally on
the used procedure.
4. Example of evaluation of the measurement
uncertainty at INRIM
In recent years the comparison of primary standard
measurement facilities has been an essential activity of na-
tional metrology institutes as this is the best way to detect
aws in instruments and/or procedures. These compari-
sons are usually conducted at the best possible level of
uncertainty, and involve a small number of NMIs drawn
from the membership of the relevant CC. Additional com-
parisons by RMOs allow to link the measurement capabil-
ities of other NMIs to these results.
INRIM has been involved as one of the linking laborato-
ries in the two different projects, carrying out rstly the re-
gional project EUROMET 692 and subsequently the CIPM
key comparison CCM.FF-K4, Fig. 3. The main purpose of
both projects was to compare the experimental method
and the uncertainty calculation in the volume determina-
tion for 100 mL Gay-Lussac pycnometers, being such
instruments expected to be representative of all types of
laboratory glassware.
INRIM used the same procedure and instrumentations
for both comparisons; specically the INRIM laboratory
operated at the ambient temperature of 20 C involving
the equipment whose specications are listed in Table 1.
Based upon the gravimetric measurement procedure
and the value of measurands, briey summarized above,
the model for measurement uncertainty analysis can be
established. The expression for the combined standard
uncertainty uV
20
was developed from Eq. (1) and (2),
respectively, using the error propagation formula.
uV
20

@V
20
@m
_ _
2
u
2
m
@V
20
@q
w
_ _
2
u
2
q
w

@V
20
@q
a
_ _
2
u
2
q
a

@V
20
@q
c
_ _
2
u
2
q
c

@V
20
@c
_ _
2
u
2
c
@V
20
@t
_ _
2
u
2
t u
2
V
_

_
_

_
3
where u(m) is the uncertainty of the weighing; uq
w
, the
uncertainty of water density; uq
a
, the uncertainty of air
density; uq
c
, the uncertainty of the mass pieces; uc,
the uncertainty of the coefcient of cubic thermal expan-
sion of the material; u(t), the uncertainty of the tempera-
ture and uV the repeatability of the 10 repeated volume
determinations.. The partial derivatives of V
20
with respect
to the input variables are the known sensitivity
coefcients.
As a numerical example, Table 2 lists the main uncer-
tainty components presented by INRIM in the key compar-
ison CCM.FF-K4 related to the pycnometer with Serial
number 03.01 13. The expanded uncertainty U at the 95%
level of condence was determined from the individual
standard uncertainty u and by evaluating the effective de-
grees of freedom m
eff
[12]. Moreover, the table shows for
each entry the type A or type B evaluation. The uncer-
tainty of the mass is a combination of the uncertainty of
the calibration of the balance, the repeatability of weighing
Fig. 3. Degrees of equivalence of each participating laboratory of EUROMET comparison with respect to the reference value determined in CCM.FF-K4 for
the Pycnometer no. TS 03.04.03 [14]. INRIM ( the former IMGC) took part in both comparisons, being one of the linking laboratories.
Table 1
Technical specication and traceability of used equipment by INRIM Laboratory in the EUROMET project 692 and CCM.FF-K4.
Upper range value Resolution Standard uncertainty Traceability
Balance 400 g 0.00001 g 0.17 mg IMGC (INRIM)
Weights 100 g 1 lg to 12 lg IMGC (INRIM)
Water thermometer 80 C 1 mK 0.015 C IMGC (INRIM)
Ambient thermometer 0.25 mK 0.01 C IMGC (INRIM)
Pressure 1 Pa 10 Pa IMGC (INRIM)
Relative humidity 0.1% 1% IMGC (INRIM)
S. Lorece / Measurement 42 (2009) 15101515 1513
and the resolution of reading. The uncertainty of the tem-
perature is essentially evaluated by the calibration of the
thermometer and its repeatability. The evaluation of the
uncertainty of the density of water is derived from the
uncertainty of the tabled values and from the inuence
of temperature. For air density, the standard uncertainty
is estimated by the individual contribution of ambient
quantities. The standard uncertainty of the expansion coef-
cient was reported by the technical protocol of the pro-
ject. The estimated contribution due to the position of
the stopper and the adjustment of meniscus are shown to-
gether as additional inuence quantities. Finally it is
shown the standard deviation or repeatability obtained in
10 repeated volume determinations.
5. Conclusion
Measurement traceability is extremely important if
reproducible and reliable results have to be obtained.
Although specic applications have not been dealt in
detail in this paper, the described work has the main pur-
pose of showing the gravimetric method to calibrate glass-
ware according to laboratory requirements and based on
existing standards.
A basic owchart related to the traceability of volume
calibration at INRIM and the mathematical model that con-
verts the mass of water to volume V contained in or deliv-
ered is also introduced.
Reliability of calibration and the level of uncertainty are
strictly connected with the equipment and the procedure
used in the measurement. Moreover the water density
and the operator have been identied, through a numerical
example based on the INRIM uncertainty budget in the
CCM.FF-K4 key comparison, as the major source of uncer-
tainty on the total calibration uncertainty.
References
[1] ISO 384:1987, Laboratory glassware Principles of design and
construction of volumetric glassware, International Organisation for
Standardization, Geneva.
[2] ISO 4787:2005, Laboratory glassware Volumetric glassware
Methods for use and testing of capacity, International Organisation
for Standardization, Geneva.
[3] ISO 1042:2000, Laboratory glassware One-mark volumetric asks,
International Organisation for Standardization, Geneva.
[4] ISO 385:2005, Laboratory glassware Burettes, International
Organisation for Standardization, Geneva.
[5] ISO 648:1977, Laboratory glassware One-mark pipettes,
International Organisation for Standardization, Geneva.
[6] ISO 3507:1999, Laboratory glassware Pyknometers, International
Organisation for Standardization, Geneva.
[7] ASTM E542:2000, Standard practice for calibration of laboratory
volumetric apparatus, ASTM International, USA.
[8] OIML R71:1985 (E), Fixed storage tanks general requirements,
International Organization of Legal Metrology OIML, Paris.
[9] OIML R120:1995 (E), Standard capacity measures for testing
measuring systems for liquids other than water, International
Organization of Legal Metrology OIML, Paris.
[10] J.M. Maldonado, R. Arias, H.-H. Oelze,V.E. Bean, J.F. Houser, C.
Lachance, C. Jacques, International comparison of volume
measurement standards at 50 L at the CENAM (Mexico), PTB
Table 2
Budget of uncertainty presented by INRIM in the key comparison CCM.FF-K4 related to the pycnometer with Serial number 03.01 13 of 100 cm
3
. For each entry
type A or type B evaluation is identied.
Entries Units Best
estimate, x
i
Source of
information
Standard
uncertainty, ux
i

Sensitivity
coefcient, c
i
Contribution,
u
i
V
20

Degrees of
freedom, m
Balance
Eccentricity g Negligible
Resolution g 0.00001 Manufact. 0.000003 1.0E+00 3.0E06 50
Linearity g Experim. 0.00015 1.0E+00 1.5E04 50
Weights (initial)
Calibration g 47.64389 Experim. 0.00007 1.0E+00 7.0E05 50
Density g/cm
3
7.900 Estimated 0.050 1.9E03 9.3E05 50
Weights (nal)
Calibration g 145.99335 Experim. 0.00030 1.0E+00 3.0E04 50
Density g/cm
3
7.900 Estimated 0.050 Unied with weights initial
Water density g/cm
3
0.998183 Equation 0.000002 9.9E+01 2.0E04 50
Water temperature C 20.086 Estimated 0.025 2.0E02 5.1E04 50
Calibration C Experim. 0.015 2.0E02 3.1E04 50
Temperature gradients within the TS C Estimated 0.01 2.0E02 2.0E04 50
Ambient conditions
Temperature C 20.24 Experim. 0.3 3.7E04 1.1E04 50
Relative humidity % 50.62 Experim. 4 9.2E06 3.7E05 50
Pressure kPa 99.073 Experim. 70 1.0E06 7.2E05 50
Artifact
Expansion coefcient C
1
9.9E06 Protocol 0.000001 8.6E+00 9.0E06 50
Temperature C 20.086 Estimated 0.115 9.8E04 1.1E04 50
Others
Stopper position + meniscus adjust. cm
3
Estimated 0.0003 1.0E+00 3.0E04 50
Repeatability cm
3
Experim. 0.000129 1.0E+00 1.3E04 9
Sum of squares 7.01E07
Standard uncertainty 8.4E04
Effective m
eff
248
t(95) 1.97
Volume at 20 C cm
3
98.6295 Expanded uncertainty, U 0.0017
1514 S. Lorece / Measurement 42 (2009) 15101515
(Germany), Measurement Canada and NIST (USA), Metrologia 39 (1)
(2002) 9195.
[11] C. Jacques, S. Trujillo Juarez, J.M. Maldonado, V. Bean, NORAMET
intercomparison of volume standards at 50 mL and 100 mL
(SIM.M.FF-S1), Metrologia 40 (Tech. Suppl. 07001) (2003).
[12] E. Batista, E. Filipe, B. Mickan, Volume calibration of 1000 lL
micropipettes. Inter-laboratory comparison, Accred. Qual. Assur. 13
(2008) 261266.
[13] E. Batista, Final report on EUROMET key comparison
EUROMET.M.FF-K4 for volume intercomparison of 100 mL Gay-
Lussac pycnometer, Metrologia 43 (Tech. Suppl. 07009) (2006).
[14] R. Arias, J. Wright, C. Jacques, C. Lachance, P. Lau, H. Tbben, G.
Cignolo, S. Lorece, J. Man, V.Y. Aibe, CCM.FF-K4: volume at 20 L and
100 mL, Metrologia 43 (Tech. Suppl. 07005) (2006).
[15] M. Tanaka, G. Girard, R. Davis, A. Peuto, N. Bignell,
Recommended table for the density of water between 0 C and
40 C based on recent experimental reports, Metrologia 38 (4)
(2001) 301309.
[16] A. Picard, R.S. Davis, M. Glser, K. Fujii, Revised formula for the
density of moist air (CIPM-2007), Metrologia 45 (2) (2008) 149155.
[17] ISO, Guide to the Expression of Uncertainty in Measurement,
(corrected and reprinted 1995), International Organisation for
Standardization, Geneva.
[18] ISO 3696:1987, Water for analytical laboratory use Specication
and test methods, International Organisation for Standardization,
Geneva.
[19] E. Batista, R. Paton, The selection of water property formulae for
volume and ow calibration, Metrologia 44 (6) (2007) 453463.
S. Lorece / Measurement 42 (2009) 15101515 1515