Publication

Reference
PURPOSE
This document provides guidance for calibration and testing laboratories involved in
hardness measurements, as well as their assessors. It has been produced to improve the
harmonization within EA in determination of uncertainties in hardness measurements with
special emphasis on Rockwell hardness. It is based on international standards with
respect to traceability reuirements and calibration procedures.
!ctobre "# rev."" $age # of %&
EA-10'#(
Hardness Measurements
of Uncertainty in
on the Estimation
EA Guidelines
EA-10/16 EA Guidelines on the Estimation of Uncertainty in Hardness Measurements
Authorship
This document has been prepared by the EA )aboratory *ommittee based on a draft of
the Task +orce ,-ardness .easurements/ of the E0pert 1roup ,.echanical
.easurements/.
Official language
The te0t may be translated into other languages as reuired. The English language
version remains the definitive version.
Copyright
The copyright of this te0t is held by EA. The te0t may not be copied for resale.
Guiance publications
This document represents a consensus of EA member opinion and preferred practice on
how the relevant clauses of the accreditation standards might be applied in the conte0t of
the sub2ect matter of this document. The approaches taken are not mandatory and are for
the guidance of accreditation bodies and their client laboratories. 3evertheless, the
document has been produced as a means of promoting a consistent approach to
laboratory accreditation amongst EA member bodies, particularly those participating in the
EA .ultilateral Agreement.
!urther infor"ation
+or further information about this publication, contact your national member of EA or the
*hairman of the EA )aboratory *ommittee, e4mail address5 klaus.brinkmann6ptb.de.
$lease check our website for up4to4date information http5''www.european4
accreditation.org
Date of endorsement : May 2001
Date of implementation : May 2002
Transitional period :
!ctobre "# rev."" $age % of %&
EA-10/16 EA Guidelines on the Estimation of Uncertainty in Hardness Measurements
CO#$E#$S
1 INTRODUCTION 4
2 PARAMETERS THAT AFFECT THE UNCERTAINTY OF INDENTATION HARDNESS
MEASUREMENT 7
2.1 Reference/test material 7
2.2 Hardness machine 9
2.3 Enironment 9
2.! "#erator 1$
3 GENERAL PROCEDURE FOR CALCULATING THE UNCERTAINTY OF
HARDNESS MEASUREMENT 10
4 APPLICATION TO THE ROCKWELL C SCALE: EVALUATION AND PROPAGATION
OF UNCERTAINTY 13
!.1 %ali&ration uncertainty of hardness testin' machines (direct cali&ration method) 13
!.2 %ali&ration uncertainty of the indirect cali&ration method 1*
5 REFERENCES 24
!ctobre "# rev."" $age 7 of %&
EA-10/16 EA Guidelines on the Estimation of Uncertainty in Hardness Measurements
1 %#$RO&UC$%O#
#.# In the field of hardness measurement a wide variety of methods and euipment is
applied which may differ according to the material. A hardness measurement is
useful when the results obtained at different sites are compatible to within a
determined interval of measurement uncertainty. The guide aims to demonstrate
the concepts of measurement uncertainty applied to this special field. !nly
uncertainties of the commonly used indentation hardness measuring methods for
metals 89rinell, Rockwell, :ickers; are discussed, for the ranges generally
employed in engineering practice where universal metrological methods have
already been implemented in industrial countries.
#.% A hardness value is the result of a measurement performed on a test piece under
standard conditions, and it is based on an agreed convention. The hardness
determination is essentially performed in two steps5
#. An indentation is made under prescribed conditions,
%. The determination of a characteristic dimension of the indentation 8mean
diameter, mean diagonal or indentation depth;.
#.7 The dissemination of hardness scales is based on three main elements5
a) !" !#$%&"'' '(#)" %"*+&++,&: description of the measurement method, the
relevant tolerances of the uantities involved and the limiting ambient
conditions.
b) !" !#$%&"'' $"*"$"&(" -#(!+&": metrological devices that materialise the
hardness scale definitions. <istinction should be made between primary
standard machines! which constitute the best possible realisation of the
hardness scale definitions, and cali"ration machines! used for the industrial
production of hardness reference blocks.
c) !" !#$%&"'' $"*"$"&(" .),(/: !ne may distinguish between primary
hardness reference "loc#s, calibrated by primary hardness standard
machines and used when the highest accuracy is reuired, e.g. for verification
and calibration of hardness calibration machines, and hardness reference
"loc#s intended mainly for the verification and calibration of industrial
hardness testing machines.
#.= +igure #.# shows the four4level structure of the metrological chain necessary to
define and disseminate hardness scales. 3ote that at each level both direct
calibration and indirect calibration are reuired. <irect calibration gives any
possible reference to mass, length and time national standards, and checks the
conformity to tolerances reuired by the scale definition. Indirect calibration is
reuired because a number of factors, not yet completely defined 8e. g.
displacement4time pattern during the indentation, shape irregularities and
mechanical performances of the indenter; cannot be evaluated by direct
measurement. *omparisons like international comparisons for the $rimary
-ardness >tandard .achines, comparisons with $rimary -ardness >tandard
9locks for the -ardness *alibration .achines and finally comparisons with
-ardness Reference 9locks for -ardness Testing .achines are considered,
therefore, as indirect measurements. <irect calibration and indirect calibration
cover, as shown before, different contributions to the uncertainty, so that different
e0pressions of the uncertainty, with different meaning, can be obtained5
!ctobre "# rev."" $age = of %&
EA-10/16 EA Guidelines on the Estimation of Uncertainty in Hardness Measurements
a; uncertainty of the scale definition, produced by the tolerances adopted and by
the lack of definition of some influence factors?
b; uncertainty of the nominal materialisation of the scale definition, produced by
the uncertainty of the factors defined by the scale definitions 8taken into
account by the direct calibration;?
c; uncertainty of the effective materialisation of the scale definition, produced by
the factors not defined by the scale definitions 8taken into account by the
indirect calibration;.
3otice that contribution a; is inherent to the definition itself and therefore shall
always be combined with contributions b; and c; that are, at least partially,
overlapping, so that one can take the ma0imum value of the two separate
evaluations.
#.& The metrological chain starts at the international level using international
definitions of the various hardness scales to carry out international
intercomparisons.
#.( A number of primary hardness standard machines at the &#+,&#) )"0") @produce@
primary hardness reference "loc#s for the calibration laboratory level. 3aturally,
direct calibration and the verification of these machines should be at the highest
possible accuracy.
#.A 3o international standards are available for this first step in the materialisation of
hardness scales. <ue to the small number of laboratories at the national level,
their work is regulated by internal operation procedures for the primary machines
only and, of course, by the regulations for international intercomparisons.
#.B At the (#)+.$#+,& )#.,$#,$1 )"0")2 the primary hardness reference "loc#s are
used to ualify the hardness calibration machines, which also have to be
calibrated directly and indirectly. These machines are then used to calibrate the
hardness reference "loc#s for the user level.
#.C At the 3'"$ )"0")2 hardness reference blocks are used to calibrate the industrial
hardness testing machines in an indirect way, after they have been directly
calibrated.
#.#" The stability of hardness scales is essentially underpinned by this two4step
calibration procedure for hardness machines5
I) Direct cali"ration ensures that the machine is functioning correctly in
accordance with the hardness definitions and regarding the appropriate
parameters?
II) $ndirect cali"ration with hardness reference blocks covers the performance of
the machine as a whole.
#.## The main reuirements for the hardness of reference blocks are stability with time
and uniformity over the block surface.
#.#% In some cases hardness blocks calibrated by primary standard machines are used
directly for the verification and calibration of industrial hardness testing machines.
This is not in line with the four4level structure of figure #.#, but there are good
reasons for it. In hardness metrology the classical rule of thumb 4 namely that the
reference instrument should be an order of magnitude or at least a factor of three
better than the controlled device 4 in many cases cannot be applied.
!ctobre "# rev."" $age & of %&
EA-10/16 EA Guidelines on the Estimation of Uncertainty in Hardness Measurements
The uncertainty gap between the national level and the user level is fairly small and
each step from one level to the ne0t adds an additional contribution to the total
uncertainty? so the four4level hierarchy may lead to uncertainties too large for
reliable hardness values at the user level. .ost metrological problems of hardness
comparison, of error propagation and traceability to standards have their origins in
this fact. The calculations in section = illustrate this problem.
I&"$&#+,&#) )"0") International International
comparisons definitions
N#+,&#) )"0") $rimary hardness <irect
standard machines calibration
C#)+.$#+,& $rimary hardness -ardness calibration <irect
)#.,$#,$1 )"0")
reference blocks machines calibration
U'"$ )"0") -ardness reference -ardness <irect
blocks testing machines calibration
Reliable
hardness values
F+45 151: T!" '$3(3$" ,* !" -"$,),4+(#) (!#+& *,$ !" %"*+&++,& #&%
%+''"-+&#+,& ,* !#$%&"'' '(#)"'
!ctobre "# rev."" $age ( of %&
EA-10/16 EA Guidelines on the Estimation of Uncertainty in Hardness Measurements
' PARA(E$ERS $)A$ A!!EC$ $)E U#CER$A%#$* O! %#&E#$A$%O#
)AR&#ESS (EASURE(E#$
%.".# Indentation hardness measurement can often be rightly considered non4
destructive since the tested part is still usable afterwards. -owever, destruction at
the actual point of test makes it impossible to verify the uncertainty of the process
by a repeated measurement at that same point. It is therefore important that every
single measurement be performed to a high degree of accuracy 8see section %.=;.
%.".% There are several influencing parameters that affect the uncertainty of hardness
measurements more or less seriously? they are listed in table %.# and divided into
groups according to their origins5
#. Test piece
%. -ardness testing machine
7. Environment
=. !perator
%.".7 The table lists more than %" sources of uncertainty which may all contribute
significantly to the total uncertainty of a hardness measurement. These sources of
uncertainty may not always contribute to every measurement at every level of the
metrological chain illustrated in figure #.#.
251 R"*"$"&("6"' -#"$+#)
%.#.# Table %.# shows that the test piece material introduces a significant number of
uncertainties. +or e0ample, the test piece thic#ness may affect the measured
hardness if the wrong method is selected. The deeper the indent, the thicker the
test piece needs to be. .aterial which is too thin will yield harder results because
of the anviling effect. In addition, if the material is too thin to support the test force
during measurement, the indenter itself could be damaged and this will undermine
the reliability of any further measurement performed with that indenter.
%.#.% The surface %uality of the test piece may also considerably influence the results of
hardness measurements. A rougher surface would reuire a greater force and'or a
larger indenter to produce a larger indentation. The 9rinell method may be the
most appropriate since it is less affected by a rough surface than Rockwell or
:ickers. Although 9rinell measurements are more tolerant of varying finish, there
are limits to the permissible surface roughness for this method too. In general,
uniformity of surface finish is important for accurate and reproducible results.
2&1&' (urface cleanliness is also critical for precise and reproducible hardness
measurement. >urface soiling with grease, o0ides or dust may cause
considerable deviations in the results? moreover, the test material or reference
block may even be irreversibly damaged.
!ctobre "# rev."" $age A of %&
EA-10/16 EA Guidelines on the Estimation of Uncertainty in Hardness Measurements
T#.)" 251: P#$#-""$' !# #**"( !" 3&("$#+&1 ,* +&%"&#+,& !#$%&"'' -"#'3$"-"&
I&*)3"&(+&4 *#(,$ S,3$(" ,* 3&("$#+&1 R"-#$/' P#$#-""$' (,&'+%"$"% *,$
(#)(3)#+,&
15 T"' 7+"(" Test piece thickness too low
>tiffness of the support
1rain structure too coarse !nly relevant,
>urface roughness if the chosen test method
Inhomogeneous distribution of hardness is not appropriate.
>urface cleanliness
25 H#$%&"'' "'+&4 -#(!+&"
a; .achine frame +riction loss
Elastic deflection
.isalignment of the indenter holder
b#; <epth measuring system Indicating error !nly relevant for Rockwell indentation depth h
$oor resolution
3onlinearity
-ysteresis
b%; )ateral measuring system Indicating error !nly relevant for 9rinell, :ickers, Dnoop
$oor resolution
3umerical aperture of lens or illuminator
Inhomogeneous illumination of the indentation
c; +orce application system <eviation from nominal forces preliminary'total test force )", )
<eviation from time intervals of the testing
cycle
preliminary'total test force
dwell time
t", t
+orce introduction
!verrun of test forces indentation velocity *
d; Indenters <eviation from the ideal shape indenter radius and angle r!
<amage
<eformation under force !nly relevant for metal indenters
35 E&0+$,&-"& Temperature deviation or drift
:ibration and shocks
45 O7"$#,$ Erong selection of test method
-andling, reading, evaluation errors
August #= rev."" $age B of %&
EA-10/16 EA Guidelines on the Estimation of Uncertainty in Hardness Measurements
252 H#$%&"'' -#(!+&"
%.%.# The design, assembly and condition of the hardness testing machine are all
critical to accurate results. E0cessive friction can cause bias and non4
repeatability. Even instruments that are operated properly can give poor
results due to e0cessive friction in the force applying system. >imilar
uncertainty contributions due to small amounts of friction can be e0pected
from the depth measuring system.
%.%.% E0cessive deflections of the supporting frame of the testing machine and the
test piece support system can cause problems too. <eviations of # to 7
hardness units are not uncommon due to improper support of the test piece
and e0cessive deflection of the instrumentFs frame.
%.%.7 <ue to the very small dimensions that are measured, the measuring system
is critical. +or e0ample, one regular Rockwell scale unit is euivalent to only
% Gm indentation depth and the superficial scale is half of that, so measuring
system uncertainty is very important.
%.%.= The force application system must constantly apply accurate forces. -igh4
uality measuring euipment should be able to apply forces well within the
limits of #," H for the user level, and even within ",# H of the nominal
force for calibration machines.
%.%.& Application of the forces reuires that both the velocity and the dwell time of
the forces be defined. :ariations of testing cycle parameters that may occur
with some manually controlled machines can produce variations in the result
of up to # -R* at (" -R*. >ofter materials and materials sub2ect to work
hardening could give significantly higher uncertainties. In these cases
contributions of dwell time uncertainty and indentation velocity shall be
evaluated specifically for the material tested.
%.%.( The properties of the indenter also influence the uncertainty of hardness
measurements. It is relatively easy to manufacture a ball to the reuired
shape. -owever, the ball holder is the main source of uncertainty.
%.%.A <iamond indenters are more difficult to manufacture to the reuired shape.
The potential sources of uncertainty are significant, but in this conte0t it is not
necessary to categorise the effect of each in detail. It is important to note here
that the best Rockwell diamond indenters manufactured today will e0hibit
variations up to ",& -R* when compared on the same testing machine.
)ower uality indenters will give significantly larger variations.
253 E&0+$,&-"&
%.7.# Ambient temperature may have considerable influence on the results of
hardness measurements, especially if small lengths have to be determined.
The lower limit for :ickers indentations is %" m, and the minimum depth for
Rockwell scales 3 and T is only ( m to A m. According to the relevant
standards, the temperature ranges are #"I* to 7&I* for the test methods
and 8%7J&;I* for the calibration of reference blocks. These ranges are too
wide for some hardness scales, but operation outside these ranges should in
any case be cause of concern. If this is unavoidable, comparative
measurements should be performed to assess the influence of temperature.
August #= rev."" $age C of %&
EA-10/16 EA Guidelines on the Estimation of Uncertainty in Hardness Measurements
%.7.% :i"rations! electrical interference and lac# of cleanliness! can cause
significant problems that are difficult to uantify. Kltra4low force
microhardness measurements of course reuire an absolutely vibration4free
environment, whereas vibration reuirements for test forces above %"" m3
are not so critical.
254 O7"$#,$
.easurement positions on the surface of the sample become important in
many cases. .easurements, for instance, near the edge of a piece or at
points close to each other must be properly located to ensure accurate
results. Kncertainties of up to % -R* are not uncommon here. !verall
monitoring of the operation is very important. >ome modern testing
machines have features that minimise operator influence? nevertheless, the
latter is still essential for a successful hardness measurement.
+ GE#ERA, PROCE&URE !OR CA,CU,A$%#G $)E U#CER$A%#$* O!
)AR&#ESS (EASURE(E#$
The following procedure is based on EA'=4"% L#M 8cf. worked e0amples in
section =;.
a) E0press the relationship between the measured hardness H 8output
uantity; and the input uantities +
i
8model function; in mathematical
terms5
H N f8+
#
,+
%
,...,+
,
; 8#;
3otice that in the case of -ardness a mathematical relationship
connecting input uantities +
i
with the output uantity H is not #no-n at
the state of the art& The connection is .i*en "y the scale definitions that
are empirical procedures. The model function, therefore, does not give
much more than a list of factors affecting the measurement results. In
practice this is sufficient for establishing a procedure based on EA'=4"%,
providing that special care is adopted for evaluating standard
uncertainties of the input uantities and sensitivity coefficients, as shown
here after.
b; Identify and apply all significant corrections.
c; )ist all sources of uncertainty in the form of an uncertainty analysis in
accordance with the following table5
August #= rev."" $age #" of %&
EA-10/16 EA Guidelines on the Estimation of Uncertainty in Hardness Measurements
T#.)" 351: S(!"-#+( ,* #& ,$%"$"% #$$#&4"-"& ,* !" 83#&++"'2
"'+-#"'2 '#&%#$% 3&("$#+&+"'2 '"&'++0+1 (,"**+(+"&'
#&% 3&("$#+&1 (,&$+.3+,&' 3'"% +& !" 3&("$#+&1
#&#)1'+' ,* # !#$%&"'' -"#'3$"-"&
uantity
+
i
estimate
/
i
standard
uncertainty
u0/
i
1
sensitivity
coefficient
c
i
contribution to
the standard
uncertainty u
i
0H1
+
1
/
1
u0/
1
1 c
1
u
1
0H1
... ... ... &&& ...
+
n
/
n
u0/
n
1 c
n
u0H1
n
-ardness H u0H1
The uantities in table 7.# are defined as follows5
+
i
uantities, reported in table %.#, affecting the measurement result
H. As said in #.= the uncertainty can be evaluated in two separate
ways5 the first way involving the physical uantities used for the
scale definitions 8forces, lengths, times, velocities etc.;, refers to
the direct calibration? the second way, involving all the factors of
influence present in practice, refers to the indirect calibration.
3otice that one could suppose that this second way contains all the
uncertainty contributions, therefore can give alone the uncertainty
value reuired, but this is no always true. +or instance it is possible
to perform a very careful indirect calibration that produces an
uncertainty lower than the uncertainty produced by the tolerances
accepted for direct calibration L%M. +or this reason both ways shall
be followed and the larger of the two uncertainty values obtained
taken as the result.
/
i
estimate values of the uantities +
i
u0/
i
1 standard uncertainties of the estimates /
i
. >ome ways can be
followed for determining u0/
i
1. +or the part connected with the
uncertainty of hardness scale definitions one shall take the
tolerance fields of the definition L7M as variability fields, and
evaluate the uncertainty contributions of type 9. Type 9
uncertainties shall be used in any case when only a declaration of
conformity is available. +or the part connected with direct
calibration it is possible to determine u0/
i
1 by the uncertainty
declared in calibration certificates of the measurement instruments
used for direct measurements. +or the part connected with indirect
calibration, that is comparisons performed using hardness blocks,
the relevant uncertainty of type A shall be evaluated.
c
i
is the sensiti*ity coefficient associated with the input estimate /
i
.
The sensitivity coefficient c
i
describes the e0tent to which the
hardness H is influenced by variations of the input estimate /
i
& As
said before at the state of the art the mathematical connection
between /
i
and H is unknown, therefore the sensitivity coefficients
shall be evaluated e0perimentally by the change H in the
hardness H due to a change /
i
in the input estimate /
i
as follows5
August #= rev."" $age ## of %&
EA-10/16 EA Guidelines on the Estimation of Uncertainty in Hardness Measurements
n n
x X x X
i
i
x
H
c
= =

,...,
1 1
8%;
The e0perimental evaluation of the sensitivity coefficients is usually
time consuming, therefore usually it is advantageous to use the
e0perimental results given in literature L=, &M and adopted for the
e0amples attached, but one shall be careful when the relevant
factors depend on the characteristics of the material tested 8dwell
time and indentation velocity;. In this case some e0periments with
the specific material are necessary.
u
i
0H1 is the contribution to the standard uncertainty associated with the
hardness H resulting from the standard uncertainty u0/
i
1 associated
with the input estimate /
i
5
u
i
0H1 2 c
i
u0/
i
1 87;
d) +or uncorrelated input uantities the suare of the standard
uncertainty u0H1 associated with the measured hardness H is given
by5

=
=
n
i
i
H u H u
1
2 2
) ( ) (
8=;
e) *alculate for each input uantity +
i
the contribution u
i
0H1 to the
uncertainty associated with the hardness H resulting from the input
estimate /
i
according to Es. 8%; and 87; and sum their suares as
described in E. 8=; to obtain the suare of the standard
uncertainty u0H1 of the hardness H.
f) *alculate the e0panded uncertainty U by multiplying the standard
uncertainty u0H1 associated with the hardness H by a coverage
factor #2%5
U 2 #u0H1 8&;
>hould the effective degrees of freedom
eff
in e0ceptional cases be
less than #&, then calculate the coverage factor # according to
EA'=4"%, Anne0 E L#M.
g; Report the result of the measurement as follows5 in calibration
certificates, the complete result of the measurement comprising the
estimate H of the measurand and the associated e0panded
uncertainty U shall be given in the form 0HU1& To this an
e0planatory note must be added which in the general case should
have the following content5
The reported e0panded uncertainty of measurement has been
obtained by multiplying the combined standard uncertainty by the
coverage factor #N% that, for a normal distribution, corresponds to a
confidence level p of appro0imately C&H. The combined standard
uncertainty of measurement has been determined in accordance
with EA'=4"% L#M.
August #= rev."" $age #% of %&
EA-10/16 EA Guidelines on the Estimation of Uncertainty in Hardness Measurements
- APP,%CA$%O# $O $)E ROC./E,, C SCA,E0 E1A,UA$%O# A#&
PROPAGA$%O# O! U#CER$A%#$*
The relevant standard documents L%M reuire that both direct and indirect
calibration methods be used, at least with new, revised or reinstalled
hardness testing machines. It is always good practice to use both calibration
methods together.
451 C#)+.$#+,& 3&("$#+&1 ,* !#$%&"'' "'+&4 -#(!+&"' 9%+$"( (#)+.$#+,&
-"!,%:
=.#.# The direct calibration method is based on the direct measurement of the
hardness scale parameters prescribed by I>! (&"B4% L%M. Even though it is
not possible to establish an analytical function to describe the connection
between the defining parameters and the hardness result L=M, some
e0periments L&M do allow, as described in section 7, to evaluate uncertainty
propagation. Oet one should be careful in the application because some of
the parameters are primarily connected with the measuring system
8preliminary test force, total test force, indentation depth, indenter geometry,
frame stiffness;, whereas others refer to the measurand 8creep effect, strain4
hardening effect;.
=.#.% The measurand related parameters can be described as an indication based
on results obtained with hardness reference blocks, but should be evaluated
directly for the specific measurand. The creep effect depends on both the
measuring system and the material characteristics? the amount of creep is a
function of the creep characteristic of the material, also depending on the
time reuired by the measuring system to register the force. +or a manual
zeroing machine, creep has generally stopped when zero is finally reached.
Even automatic machines are more or less prompt. A machine that takes & s
to apply the preliminary test force produces a different creep rela0ation than
a machine taking only # s, and the strict observance of a = s force dwell time
will not help to obtain compatible results.
=.#.7 There is call for caution in interpreting numerical values because the results
obtained with old manual machines cannot represent those of a modern
automatic hardness testing machine, designed to produce indentations in
the shortest possible time.
=.#.= The evaluation of uncertainty is described in the relevant EA'=4"% document
L#M. The uncertainty calculation must be done in different ways, depending on
the types of data available. The first step is the evaluation of the appropriate
variances corresponding to the measurement parameters involved
8independent variables;.
=.#.& The measurement results given in a calibration certificate, with the uncertainty
usually uoted for #2% coverage factor, permit the calculation of the standard
uncertainty. It is sufficient to divide the given uncertainty by the stated
coverage factor. *onformity declaration can also be used to evaluate the
standard uncertainty, taking the tolerance interval a into account. A
rectangular distribution function should be used, with euivalent variance u
2
2
a
2
37.
August #= rev."" $age #7 of %&
EA-10/16 EA Guidelines on the Estimation of Uncertainty in Hardness Measurements
=.#.( The second step is the calculation of the combined standard uncertainty.
Theoretically, if the hardness H is the measurand 8dependent variable;, it
can be represented as a function of the measurement independent
variables. The symbols used are indicated in table =.#5
H 2 f 0)
0
4 )4 r4 4 t
0
4 t4 *4 h4 ,4 (1 8(;
.ore e0plicitly, the euation is5
i
i
/
/
H
(
h
, H

+ =

8A;
where 0
i
are the independent variables in e. 8C;.
=.#.A Ksing the appropriate sensitivity coefficients, namely the partial derivatives
of the dependent variable H against the independent variables /
i
, one
obtains the formula for evaluating the uncertainty propagation in the
appro0imation of uncorrelated independent variables5

= =
=
n
i
i i
n
i
i
x u c H u H u
1
2 2
1
2 2
) ( ) ( ) (
8B;
In practice, the partial derivatives can be appro0imated by the incremental
ratios5
; 8 ; 8 ; 8 ; 8
; 8 ; 8 ; 8 ; 8 ; 8
%
%
%
%
%
%
"
%
%
"
%
%
%
%
%
%
"
%
%
"
%
h u
h
H
* u
*
H
t u
t
H
t u
t
H
u
H
r u
r
H
) u
)
H
) u
)
H
H u

+
+

8C;
=.#.B The standard uncertainty can be evaluated for different conditions. As an
e0ample, Table =.% shows the evaluation of the standard uncertainty u0H1,
and the e0panded uncertainty with coverage factor #N%, for a conformity
assessment of hardness testing machines and indenters to the relevant
standard L%M. This was done using the appropriate tolerances to calculate
type 9 standard uncertainties.
August #= rev."" $age #= of %&
EA-10/16 EA Guidelines on the Estimation of Uncertainty in Hardness Measurements
T#.)" 451: '1-.,)' 3'"%
H .easured hardness t total test force dwell
time
ud uncertainty of
hardness scale
definition
)0 $reliminary test
force
* indentation velocity um uncertainty of primary
hardness standard
machine
) total test force h indentation depth us stability uncertainty of
calibration machine
r Indenter radius , constant number
dependent by the scale
uf fitting uncertainty
indenter angle ( constant number
dependent by the scale
i degrees of freedom
t0 $reliminary test
force dwell time
H" mean hardness measurement result of
primary hardness reference block
sc >tandard deviation of the
measurements Hc
H"i single hardness measurement result of
primary hardness reference block
(ci >tandard deviation of the
measurements Hci
u"d *alibration uncertainty of primary
hardness reference blocks considering
the scale definition
Hc .ean hardness values of the scale of
the calibration machine
u"m *alibration uncertainty of primary
hardness reference blocks considering
the uncertainty of the primary hardness
standard machine
Hci >ingle hardness values of the scale
of the calibration machine
s" >tandard deviation of the measurement
H"
ucdf *alibration machine uncertainty
considering the scale definition
uncertainty and the fitting uncertainty
s"i >tandard deviation of the
measurements H"i
ucmf *alibration machine uncertainty
considering the primary standard
machine uncertainty and the fitting
uncertainty
ucd *alibration uncertainty of the
calibration machine considering the
scale definition
ucdu *alibration machine uncertainty
considering the scale definition
uncertainty and the calibration results
uncorrected
ucm *alibration uncertainty of the
calibration machine considering the
uncertainty of the primary hardness
standard machine
ucmu *alibration machine uncertainty
considering the primary standard
machine uncertainty and the
calibration results uncorrected
H *orrection value
August #= rev."" $age #& of %&
EA-10/16 EA Guidelines on the Estimation of Uncertainty in Hardness Measurements
T#.)" 452: E0#)3#+,& ,* !" 3&("$#+&1 7$,7#4#+,& *,$ (,&*,$-+1
#''"''-"& ,* !" !#$%&"'' "'+&4 -#(!+&" #&% +&%"&"$
/
i
a
i
( )
7
%
% i
a
i
/ u =
>ensitivity coefficients at
different hardness levels

i
i
/
H
c

=
*ontributions to u50H;'-R*P
at different hardness levels

= =
=
n
i
i i
n
i
i
x u c H u H u
1
2 2
1
2 2
) ( ) ( ) (
%" to %& =" to =& (" to (& %" to %& =" to =& (" to (&
)
0
'3 % #,7Q#"
"
#,%Q#"
4#
A,"Q#"
4%
&,"Q#"
4%
#,CQ#"
4%
(,=Q#"
47
7,7Q#"
47
)'3 #& A,&Q#"
R#
4=,"Q#"
4%
47,"Q#"
4%
4%,"Q#"
4%
#,%Q#"
4#
(,BQ#"
4%
7,"Q#"
4%
'I ",7& =,#Q#"
4%
#,7Q#"
R"
B,"Q#"
4#
=,"Q#"
4#
(,CQ#"
4%
%,(Q#"
4%
(,(Q#"
47
r'mm ","# 7,7Q#"
4&
#,&Q#"
R#
7,"Q#"
R#
&,"Q#"
R#
A,=Q#"
47
7,"Q#"
4%
B,7Q#"
4%
h'Gm # 7,7Q#"
4#
4&,"Q#"
4#
4&,"Q#"
4#
4&,"Q#"
4#
B,7Q#"
4%
B,7Q#"
4%
B,7Q#"
4%
v'
8Gm's;
%& %,#Q#"
R%
4%,"Q#"
4%
","
.
#"
"
7,"Q#"
4%
B,=Q#"
4%
","
..
#"
"
#,CQ#"
4#
t
0
's #,& A,&Q#"
4#
#,"Q#"
4%
&,"Q#"
47
=,"Q#"
47
A,&Q#"
4&
#,CQ#"
4&
#,%Q#"
4&
t's % #,7Q#"
"
4A,"Q#"
4%
4=,"Q#"
4%
47,"Q#"
4%
(,=Q#"
47
%,#Q#"
47
#,%Q#"
47
T!TA)
% % % %
-R* -R*

=
i
u u
",7C ",%% ",="
>tandard uncertainty
-R* u
",(% ",=( ",(7
E0panded uncertainty
-R* -R* #u U =
#,%& ",C7 #,%(
=.#.C Table =.7 shows the evaluation of standard and e0panded uncertainty for
calibration certificates for the hardness testing machine and indenter. -ere
the e0ample is for the hardness level %" -R* to %& -R*. 3ote that the
differences between the parameter and nominal values are known, together
with their uncertainties, and it is therefore possible to estimate both a
correction H
i
and its uncertainty u0 H
i
1 using the same sensitivity
coefficients as before.
=.#.#" Ehilst in the case of type 9 uncertainty contributions the degrees of freedom

i
of the various parameters can be considered large enough to apply the
1aussian distribution, in this case
i
depends on the adopted measurement
procedure. Table =.7 uotes typical values of
i
.
August #= rev."" $age #( of %&
EA-10/16 EA Guidelines on the Estimation of Uncertainty in Hardness Measurements
T#.)" 453: E0#)3#+,& ,* !" 3&("$#+&1 7$,7#4#+,& +& (#)+.$#+,&
("$+*+(#"' *,$ !" !#$%&"'' "'+&4 -#(!+&" #&% *,$ !"
+&%"&"$ *,$ 20 HRC , 25 HRC !#$%&"'' )"0")
*ertificate data .easured hardness
2
i
i
/ U
i
021
i
i
i
/
H
c

=
i
H u
i
'
3)4 u
i
-
3)4/
i
-R* -R*
%
-R*
=
)
0
33 ",B ",% B #,%Q#"
4#
",#" #,=Q#"
4=
%,(Q#"
4C
)'3 4=,7 #,& B 4=,"Q#"
4%
",#A C,"Q#"
4=
#,"Q#"
4A
'I ",% ",# B #,7Q#"
"
",%( =,%Q#"
47
%,%Q#"
4(
r'mm ",""A ",""% B #,&Q#"
R#
",## %,7Q#"
4=
(,7Q#"
4C
h'Gm 4",& ",% 7 4&,"Q#"
4#
",%& %,&Q#"
47
%,#Q#"
4(
*'8Gm's; %" & % 4%,"Q#"
4%
4",=" %,&Q#"
47
7,#Q#"
4(
t
0
's # ",& 7 #,"Q#"
4%
","# (,7Q#"
4(
#,7Q#"
4##
t's # ",& 7 4A,"Q#"
4%
4","A 7,#Q#"
4=
7,#Q#"
4B
Total ",=% ","## A,(Q#"
4(
>tandard uncertainty
-R* u
",#"
<egrees of freedom #&
*overage factor # for confidence level p N C&H %,#7
E0panded uncertainty
-R* -R* #u U =
",%%
Ehere
i i i
/ c H =
and ; 8 ; 8
% % %
i i i
/ u c H u
=.#.## This method can only be used correctly if nominal values are defined for the
various parameters. If, as is the case with current standards, there are
parameters which are not defined as nominal values with a given tolerance
but as uniform probability intervals, then the reference to a @nominal value@ is
not possible. In conseuence, the uncertainty calculated in this way can only
be accepted where there is a preliminary agreement on the @nominal values@
of the measurement parameters.
452 C#)+.$#+,& 3&("$#+&1 ,* !" +&%+$"( (#)+.$#+,& -"!,%
=.%.".# The indirect calibration method is based on a metrological chain. A typical
seuence is 8cf. +igure #.#;5
a; definition of the hardness scale?
August #= rev."" $age #A of %&
EA-10/16 EA Guidelines on the Estimation of Uncertainty in Hardness Measurements
b; materialisation of the hardness scale definition by a primary hardness
standard machine?
c; calibration of primary hardness reference blocks for the dissemination of
the hardness scale?
d; calibration of a hardness calibration machine for the industrial
production of hardness reference blocks?
e; calibration of hardness reference blocks?
f; calibration of industrial hardness testing machines using hardness
reference blocks.
g; hardness measurement performed with industrial hardness testing
machines.
=.%.".% It is also possible to go directly from step c; to step f;, or, after step e; to add
the calibration of a frontline hardness testing machine from the industrial
uality system and, within the uality system, to calibrate the hardness
reference blocks necessary for the calibration of other hardness testing
machines used within the uality system itself. 3ote that after step d; the
subseuent steps are repetitions of the previous ones. In conseuence, the
description of the uncertainty evaluation can be restricted to the first four
steps.
!.2.1 U&("$#+&1 u

,* !" R,(/;")) !#$%&"'' '(#)" %"*+&++,&

=.%.#.# The evaluation of the uncertainty u
d
of the hardness scale definition and its
materialisation is similar to the evaluation of the uncertainty due to the direct
calibration method, taking the tolerances prescribed by I>! (&"B47 L7M into
account. Table =.= presents an e0ample of uncertainty evaluation. 3ote that
uncertainty contributions are of type 9, therefore a coverage factor #N% is
used.
T#.)" 454 : E0#)3#+,& ,* !" 3&("$#+&1 u

%3" , !" %"*+&++,& ,* !"

R,(/;")) C S(#)" #&% +' -#"$+#)+'#+,&
August #= rev."" $age #B of %&
EA-10/16 EA Guidelines on the Estimation of Uncertainty in Hardness Measurements
2
i
a
i
( )
7
%
% i
i
a
/ u =
>ensitivity coefficients at
different hardness levels
i
i
/
H
c

=
*ontributions to u
2
0H1'-R*P
at different hardness levels

= =
=
n
i
i i
n
i
i
x u c H u H u
1
2 2
1
2 2
) ( ) ( ) (
%" to
%&
=" to =& (" to (& %" to %& =" to =& (" to (&
August #= rev."" $age #C of %&
EA-10/16 EA Guidelines on the Estimation of Uncertainty in Hardness Measurements
)
0
'3 ",% #,7Q#"
4%
#,%Q#"
4#
A,"Q#"
4%
&,"Q#"
4%
#,CQ#"
4=
(,=Q#"
4&
7,7Q#"
4&
)'3 #,& A,&Q#"4# 4=,"Q#"
4
%
47,"Q#"
4%
4%,"Q#"
4%
#,%Q#"
47
(,BQ#"
4=
7,"Q#"
4=
'I ",# 7,7Q#"
47
#,7Q#"
"
B,"Q#"
4#
=,"Q#"
4#
&,(Q#"
47
%,#Q#"
47
&,7Q#"
4=
r'mm ",""& B,7Q#"
4(
#,&Q#"
R#
7,"Q#"
R#
&,"Q#"
R#
#,CQ#"
47
A,&Q#"
47
%,#Q#"
4%
h'Gm ",% #,7Q#"
4%
4&,"Q#"
4
#
4&,"Q#"
4#
4&,"Q#"
4#
7,7Q#"
47
7,7Q#"
47
7,7Q#"
47
*'
8Gm's;
#" 7,7Q#"
R#
4%,"Q#"
4
%
","
.
#"
"
7,"Q#"
4%
#,7Q#"
4%
","
.
#"
"
7,"Q#"
4%
t
0
's #,& A,&Q#"
4#
#,"Q#"
4%
&,"Q#"
47
=,"Q#"
47
A,&Q#"
4&
#,CQ#"
4&
#,%Q#"
4&
t's % #,7Q#"
"
4A,"Q#"
4
%
4=,"Q#"
4%
47,"Q#"
4%
(,=Q#"
47
%,#Q#"
47
#,%Q#"
47
T!TA)
% % % %
-R* -R*

=
i d
u u ","7 ","% ","(
>tandard uncertainty
-R*
d
u ",#B ",#7 ",%=
E0panded uncertainty
-R* -R*
d
#u U = ",7( ",%( ",=A
=.%.#.% The evaluated values are confirmed by results obtained during international
comparisons, in particular that involving the largest number of participants,
which shows a spread of results of about J ",& -R*.
45252 U&("$#+&1 ,* !" -#"$+#)+'#+,& ,* !" R,(/;")) !#$%&"'' '(#)"
%"*+&++,&
=.%.%.# To demonstrate an uncertainty evaluation for state of the art characteristics
of primary hardness standard machines, one may do a calculation similar to
that in table =.7, taking relevant uncertainties as shown in table =.& into
account. The results are optimistic because significant parameters, such as
the performance of the indenter, are not accounted for, yet these must be
considered as inherent in the uncertainty due to the definition. It can be seen
that the uncertainty of the machine is almost negligible compared to the
effect of the tolerances given by the definition, with the uncertainty
contributions from influencing uantities missing in the definition itself.
T#.)" 455: E0#)3#+,& ,* !" 3&("$#+&1 3- .#'"% ,& !" '#" ,* !"
#$ ,* 7$+-#$1 !#$%&"'' '#&%#$% -#(!+&"' *,$ !" 20 HRC
, 25 HRC !#$%&"'' )"0")5
*ertificate data .easured hardness
2
i x
i
U
i
3'4
$
i
i
/
H
c

=
i
H u
i
2
0H1 u
i
6
0H13
i
-R* -R*
%
-R*
=
August #= rev."" $age %" of %&
EA-10/16 EA Guidelines on the Estimation of Uncertainty in Hardness Measurements
)
0
'3 ","# ","# %" #,%Q#"
4#
#,%Q#"
47
7,(Q#"
4A
(,&Q#"
4#&
)33 ",#& ","& %" 4=,"Q#"
4%
4(,"Q#"
47
#,"Q#"
4(
&,"Q#"
4#=
'I ","& ","% %" #,7Q#"
"
(,&Q#"
4%
#,AQ#"
4=
#,=Q#"
4C
r'mm ",""7 ",""# %" #,&Q#"
R#
=,&Q#"
4%
&,(Q#"
4&
#,(Q#"
4#"
h'8Gm; ",# ","& %" 4&,"Q#"
4#
4&,"Q#"
4%
#,(Q#"
4=
#,%Q#"
4C
*38Gm's; & % #" 4%,"Q#"
4%
4#,"Q#"
4#
=,"Q#"
4=
#,(Q#"
4B
t
0
's ",& ",% #" #,"Q#"
4%
&,"Q#"
47
#,"Q#"
4(
#,"Q#"
4#7
t's ",& ",% #" 4A,"Q#"
4%
47,&Q#"
4%
=,CQ#"
4&
%,=Q#"
4#"
Total 4","A ",""# #,CQ#"
4B
>tandard uncertainty
-R*
m
u ","7
<egrees of freedom 7(
*overage factor # for confidence level p N C&H %,"7
E0panded uncertainty
-R* -R* #u U =
","(
Ehere
i i i
/ c H =
and ; 8 ; 8
% % %
i i i
/ u c H u
=.%.%.% The value of the uncertainty is therefore primarily the result of tolerances of
the measuring parameters prescribed by relevant standards. Although table
=.= does not take the contribution due to the primary hardness standard
machine into account for the materialisation of the definition itself, it can still
be considered a comprehensive evaluation.
45253 U&("$#+&1 ,* !" (#)+.$#+,& ,* R,(/;")) 7$+-#$1 !#$%&"'' $"*"$"&("
.),(/'
=.%.7.# The primary hardness reference block is calibrated by a primary hardness
standard machine making five hardness measurements H
"i
. The mean value
H
"
is taken as the hardness value of the block.
=.%.7.% Repeating the measurement reveals the effects of non4uniformity of the
reference block surface and the repeatability of the primary hardness
standard machine, including its resolution. !ther effects, such as the
hardness stability of reference blocks, must be estimated from e0perience
with the reference blocks and their maintenance conditions.
=.%.7.7 E0cept for a possible drift that must be evaluated separately, the uncertainty
u
"d
or u
"m
of H
"
can be evaluated from the uncertainty due to the scale
definition u
d
! given in Table =.=, combined with the standard deviation s
"
of
H
"
evaluated using the standard deviation s
"i
of the measurements H
"i
&
=.%.7.= The uncertainties u
"d
or u
"m
are given by5
=
; 8
&
#
%

=
i
" "i
"i
H H
s
8#";
August #= rev."" $age %# of %&
EA-10/16 EA Guidelines on the Estimation of Uncertainty in Hardness Measurements
&
"i
"
s
s =
8##;
% %
" d "d
s u u + = or
% %
" m "m
s u u + = 8#%;
=.%.7.& The calibration certificate shall at least state the value of the standard
uncertainty u
"d
. Also reuired is the value u
"m
. E0plicit values for the
uncertainty contributions L&, (M can be included for information.
45254 U&("$#+&1 ,* !" (#)+.$#+,& ,* R,(/;")) (#)+.$#+,& -#(!+&"'
=.%.=.# The hardness reference block is calibrated by a hardness calibration making
five hardness measurements H
ci
. The mean value H
c
is compared with the
block hardness H
"
to calibrate the machine for that scale and that hardness
8H2H
c
7H
"
;.
=.%.=.% Repeating the measurement reveals the effects of non4uniformity of the
reference block surface and the repeatability of the hardness calibration
machine, including its resolution. Therefore, e0cept for the stability of the
calibration machine u
s
that must be evaluated separately because it depends
on the working conditions, the uncertainty u
cd
or u
cm
of H
c
can be evaluated
by combining the relevant uncertainty due to the hardness reference block
u
"d
or u
"m
with the standard deviation s
c
of H
c
calculated using the standard
deviation s
ci
of the measurements H
ci
.
=.%.=.7 To minimise the uncertainty, the correction H should be applied by the
measured hardness. To derive the uncertainty u
cdf
or u
cmf
at any point of the
machine scale one should interpolate the results H. The uncertainty due to
fitting u
f
depends on the structure and the working characteristics of the
calibration machine, and should therefore be determined to characterise the
machine itself by a calibration on five hardness levels, comparing the least
suares parabola with the parabola passing through the three points at the
hardness level chosen for the subseuent periodic checks.
=.%.=.= +or the uncertainties u
cdf
or u
cmf
we have5
=
; 8
&
#
%

=
i
c ci
ci
H H
s
8#7;
&
ci
c
s
s =
8#=;
% %
c "d cd
s u u + = or
% %
c "m cm
s u u + = 8#&;
% %
f cd cdf
u u u + = or
% %
f cm cmf
u u u + = 8#(;
if the correction H is not applied, the uncertainty u
cdu
and u
cmu
are calculated
using5
% %
H u u
cd cdu
+ =

or % %
H u u
cm cmu
+ =
8#A;
=.%.=.& The calibration certificate shall at least state the value of the standard
uncertainty u
cdf
. Also reuired is the value of u
cmf
& E0plicit values of the
uncertainty contribution L&, (M can be included for information.
August #= rev."" $age %% of %&
EA-10/16 EA Guidelines on the Estimation of Uncertainty in Hardness Measurements
45255 U&("$#+&1 ,* !" (#)+.$#+,& ,* !#$%&"'' $"*"$"&(" .),(/' #&% "'+&4
-#(!+&"'
+or the calibration of hardness reference blocks and hardness testing
machines the same procedures are used as those described above for
calibration of primary hardness reference blocks and hardness calibration
machines. The formulae given for those cases shall be used.
4525< N3-"$+(#) "=#-7)"
The uncertainty evaluation can be set out as in the following e0ample in
Table =.(.
August #= rev."" $age %7 of %&
EA-10/16 EA Guidelines on the Estimation of Uncertainty in Hardness Measurements
T#.)" 45< I&%+$"( (#)+.$#+,& (!#+& - U&("$#+&1 "0#)3#+,&
H#$%&"'' )"0") %" to %& =" to =& (" to (&
<efinition and standard machine uncertainty 8u
d
; 8see Table
=.=;
",#B ",#7 ",%=
P$+-#$1 !#$%&"'' $"*"$"&(" .),(/ (#)+.$#+,&
3umber of indentations & & &
3on4uniformity of primary hardness reference block and
machine reproducibility. Relevant standard deviation 8s
"i
;
8E.#";
",%7 ",#A ",#%
>tandard deviation of the mean of indentations 8s
"
; 8E.##; ",#" ","B ","&
Kncertainty of the hardness value of reference blocks 8u
"d
or
u
"m
; 8E.#%;
",%# ",#& ",%&
C#)+.$#+,& ,* !#$%&"'' (#)+.$#+,& -#(!+&"
3umber of indentations & & &
3on4uniformity of primary hardness reference block and
machine reproducibility. Relevant standard deviation 8s
ci
;
8E.#7;
",%C ",%7 ",#A
>tandard deviation of the mean of indentations 8s
c
; 8E.#=; ",#7 ",#" ","B
+itting uncertainty u
f
","C ","= ","(
Kncertainty of the hardness scale of the calibration machine
8u
cdf
or

u
cmf
; 8E.#& and E.#(;
",%( ",#B ",%(
H#$%&"'' $"*"$"&(" .),(/ (#)+.$#+,&
3umber of indentations & & &
3on4uniformity of hardness reference block and machine
reproducibility. Relevant standard deviation 8s
"i
; 8E. #";
",%C ",%7 ",#A
>tandard deviation of the mean of indentations 8s
"
; 8E.##; ",#7 ",#" ","B
Kncertainty of the hardness value of hardness reference
blocks 8u
"d
or u
"m
; 8E.#%;
",%C ",%% ",%A
Effective degrees of freedom
i
& 7" %( =%
*overage factor %,"= %,"( %,"%
E0panded uncertainty U ",&C ",== ",&&
August #= rev."" $age %= of %&
EA-10/16 EA Guidelines on the Estimation of Uncertainty in Hardness Measurements
5 RE!ERE#CES
L#M EA'=4"%5 E/pression of the Uncertainty of Measurement in 8ali"ration,
<ecember #CCC
L%M I>! (&"B4%5#CCC5 Metallic materials 9 :oc#-ell hardness test 9 ;art 2:
<erification and cali"ration of the testin. machine
L7M I>! (&"B475#CCC5 Metallic materials 9 :oc#-ell hardness test 9 ;art ':
8ali"ration of reference "loc#s
L=M 9arbato, 1.? <esogus, >.5 The meanin. of the .eometry of :oc#-ell
indenters $MG8 Technical :eport! ,o& :12=! 1>?=! 6
L&M $etik, +.5 The Unification of Hardness Measurement, 9I.), $aris, #CC#,
p.((4(C
L(M @$MA (; 1>3(: 6: 8ompte7rendu de la comparaison internationale des
Bchelles de duretB 9I.), #CB=
August #= rev."" $age %& of %&