Professional Documents
Culture Documents
Abstract
The fundamental quantities in the depth sensing hardness test are force and displacement. The penetration depth used for
the calculation of a certain hardness parameter (for instance HU, Eq. (1)) makes it necessary to determine the contact point.
As shown in the paper, fitting of the force–displacement curve, F(h), by a second order polynomial and extrapolating to
F 5 0 is not always the right way to achieve a sufficiently small uncertainty of depth. The feature of that procedure is
analysed by a simple mathematical model. In dependence of the testing force, the scatter of the experimental data, and the
required limit of depth uncertainty, a profound selection of the fitting range, the fitting function, and the extrapolation
function should be done. 2000 Elsevier Science Ltd. All rights reserved.
Fig. 4. Extrapolated contact points according to Eq. (4) after Fig. 5. Indentation data record of float glass measured by a
fitting the experimental data of float glass (tin side, R a 5 0.02 mm) hardness machine with an enhanced scatter of force before
in the range 0.05 N,F , F2 . The range of uncertainty according contact. If the fitting range, F1 , F ,1 N, is extended to the four
to Eq. (5) is shown by the two dashed lines. Note the missing real indicated data points (open squares) the shape of the curve for
solution in the four grey regions. The specifications of the extrapolating to F 5 0 becomes very different while differences of
hardness machine are given in the caption of Fig. 2. the curves within the fitting range are not visible.
C. Ullner / Measurement 27 (2000) 43 – 51 47
of extrapolation is very different whereas the func- and further materials has shown that a more detailed
tions cover within the fitting range above 1 mm well. analysis of the applied extrapolation is necessary.
The different fitting ranges result not only in differ-
ent contact points but also in cases of missing real
solutions (curves 1 and 2 in Fig. 5). 4. Analysis of extrapolation
Certainly, the test up to 10 N selected for the
demonstration of the extrapolation problems in Figs. The determination of the contact point by ex-
4 and 5 is extreme. However, fitting ranges which do trapolating the function F(h) fitted in F1 # F # F2 is
not result in real solutions of contact point h 0 after demonstrated in Fig. 7. The indentation force is
extrapolation can also exist in certain indentation normalized to the lower limit of the fitting range, F1
tests up to 1 N. Two indentation curves of the test (equals the standard deviation of the force, u F , for
plotted in Fig. 1 are selected to demonstrate the instance). Random fluctuations of the data points
indentation tests up to 1 N. Two indentation curves within the fitting range are not taken into account
of the test plotted in Fig. 1 are selected to demon- because the extrapolation problem arises from the
strate the results of h 0 after extrapolation (Fig. 6). certain shape of the extrapolation function. Instead of
The test 06 leads to a slightly variable contact point fitting within F1 # F # F2 the parameters a 0 , a 1 , a 2
if the upper limit of the fitting range is increased up of the second order polynomial, Eq. (3), are de-
to 0.1Fmax according to the standard [3]. In contrast, termined from the three data points (h 0 , 0), (h 1 , F1 ),
the test 02 does not give real solutions for the fitting (h 2 , F2 ) which give a linear inhomogenous equation
ranges among 0.00004 N,F ,0.009 N and 0.00004 system
N,F ,0.028 N. However, there are additional real
solutions of the contact point for the fitting ranges Fi 5 a 0 1 a 1 h i 1 a 2 h 2i ; i 5 0, 1, 2; F0 5 0. (7)
among 0.003 N,F ,0.028 N and 0.003 N,F ,
The zero point (0, 0) is defined by the condition
0.028 N if only the fitting range near the contact
point is reduced. F F
The practice involving further hardness machines ]21 5 ]22 . (8)
h1 h2
the rule of similarity. The real indentation curve in plotted in Fig. 7. Comparing with experimental data
F1 # F # F2 can diverge from Eq. (6). In this sets demonstrated in Figs. 1–3 the shapes of the
analysis, it is defined by the given contact point (h 0 , curves in Fig. 7 make clear that there is not a
0). The contact point, h 0 , must range in 0 # h 0 # h lin favoured extrapolating function. The uncertainty due
with to the artificial extrapolating function can be esti-
mated by the difference of the calculated contact
h 1 F2 2 h 2 F1
h lin 5 ]]]]. (9) points after Eq. (12). Taking into account the
F2 2 F 1
uncertainty of the force origin, u F , the maximal
Although such a deterministic analysis cannot give contact point, h 0max , is h 01 with F1 ⇒ F1 2 u F and
the variances and covariance of the parameters a 0 , the minimal contact point, h 0min is h 02 with
a 1 , a 2 the uncertainty of the extrapolated contact F1 ⇒ F1 1 u F . The difference reads after Eq. (12)
point can be roughly estimated according to Eq. (5) F1 1 3u F
because the first term is dominant Dh 0 5 h 0max 2 h 0min 5 ]]]. (13)
a1 1 a2h1
uF
ush 0d . ]]] 5 U. (10) Finally this difference is used for normalization of
2a 2 R 0.5
the main part of the uncertainty according to Grau et
The uncertainty in force measurement is u F . We are al. [10], Eq. (10),
going to calculate the portion U of the uncertainty. U
U a1 1 a2h1
is the uncertainty for the approach of F(h) to the ]] 5 ]]]]]]]]]]]]
ideal case, Eq. (6), that means h 0 → 0 or R → 0. For
this, U is normalized to the additional uncertainty of
Dh 0 F1
]13
uF S DœS D
]
a1 2 a0
2a 2
2]
a2
the contact-point determination due to the use of the
fitting function for the extrapolation to F 5 0 instead S D F1 F2
5 f h 0, ], ] .
uF uF
(14)
of the use of a real physical function F(h). It can be
assumed that the force–depth dependence closed to Using Eqs. (7), (9), and (14) the normalized
contact, Fex (h), equals the indentation response in the uncertainty, U /Dh 0 , is plotted in Fig. 8 as a function
fitting range F1 # F # F2 . To estimate this type of of h 0 /h 0lin for three different ‘fitting ranges’ normal-
extrapolation uncertainty it is assumed that the ized to the uncertainty of the force origin, F1 /u F #
unknown force dependence Fex (h) in 0 # F # F1 is
accomplished between the two functions as follows:
on the one hand the simplest function is the linear
dependence with the same slope, A, like the fitting
function at (h 1 , F1 ). On the other hand the function is
based on the rule of similarity (ideal case). The two
functions can be written:
Fexshd 5 A msh 2 h 0md m ; m 5 1, 2 (11)
with the condition of the same slope A m 5 A of the
fitting function, Eq. (3), at F 5 F1 . The extrapolated
contact points h 01 and h 02 are
mF1
h 0m 5 h 1 2 ]]]2 ; m 5 1, 2 (12)
a 1 1 2a 2 h 1
Fig. 8. Normalized uncertainty of the contact point as a function
with h 1 as the displacement at F 5 F1 and a 1 , a 2 as
of the portion of the linear dependence of F(h) within the fitting
the limits of Eq. (3) calculated by Eq. (7). range. U /Dh 0 . 1 means the uncertainty due to the use of the
The three extrapolating functions, F(h) after Eq. second order polynomial is greater than the uncertainty of the
(3), Fex (h) after Eq. (11) with m 5 1 and m 5 2 are unknown extrapolating function.
C. Ullner / Measurement 27 (2000) 43 – 51 49
F /u F # F2 /u F . In this way the curves in Fig. 8 are • The tests with problems in extrapolation could be
independent of the uncertainty in force measurement, removed. A procedure for the selection of accept-
u F . The portion of linearity concerning the ‘fitting able tests must be found. However, the result is
function’, F(h), is given by the value h 0 /h 0lin rang- artificial regarding the stochastic materials re-
ing from h 0 /h 0lin 5 0 (ideal case according to the sponse.
rule of geometric similarity), to h 0 /h 0lin 5 1 (linear • The fitting range is varied unless the extrapolation
dependence). gives a real solution with minimal uncertainty.
However, the indefinite limits of the fitting range
can result in reduced reproducibility between
5. Discussion different machines.
• The need for determination of the contact point is
As is demonstrated by the deterministic model on avoided with the depth sensing hardness accord-
the property of contact-point extrapolation there are ing to Eq. (2). However, the current depth
three critical points of the obvious way to use a dependence of the hardness cannot be detected.
second order polynomial for both fitting and ex- • A more robust determination of contact point can
trapolating. It can be concluded from Fig. 8: be performed by the use of the inverse indentation
dependence,
1. the normalized uncertainty becomes unlimited for
the ideal case according to the rule of geometric h 5 b0 1 b1F 1 b2F 2 (15)
similarity (the procedure needs a fitting function
with a portion of linearity);
2. the normalized uncertainty drops substantially or
below one for increasing the portion of linearity
]
(the uncertainty caused by the unknown ex- h 5 b 0 1 b 1ŒF 1 b 2 F (16)
trapolating function in the contact range, Fex (h) ,
F1 , becomes dominant); with the contact point h 0 5 b 0 . The results accord-
3. the normalized uncertainty is higher if the fitting ing to the two functions are plotted in Fig. 9 using
range is closer to F 5 0 (the disadvantage of point the extreme example of Fig. 4. As shown in Fig. 9
1 cannot be weakened by a closer fitting range). the contact points for small fitting ranges below
F2 51 N are very realistic but the model error
In practice, a laboratory using one type of a increases at higher F2 related to that of the second
hardness machine can obtain results of HU with very order polynomial, F(h).
small uncertainty caused by the contact-point ex-
trapolation, if the fitting range has been optimized
regarding the certain character of data sets (sampling In every case the estimation of the uncertainty is
rate, scatter of force and displacement, portion of fundamental for the examination of the hardness
linearity). However, the uncertainty caused by the machine including the procedure of extrapolation.
unknown function, Fex (h) , F1 , near the contact Taking into account the results of the analysis it is
point can be much larger than the uncertainty caused proposed to improve the comparability as follows.
by the statistical scatter (U /Dh 0 < 1 for h 0 /h 0lin . Twice the uncertainty of the contact point is de-
0.3 in Fig. 8). For that reason, the scatter between termined from the sum of the uncertainty, Dh 0 ,
results of HU which were measured by several caused by the unknown extrapolating function,
hardness machines of different types (in different Fex (h) , F1 , and the standard deviations of the two
laboratories with different procedures for determin- types of contact points according to Eq. (12), u h 01 or
ing the contact point) can become much higher than m 5 1, and u h 02 for m 5 2,
the scatter of results within one laboratory.
There are several ways to avoid the critical points Dh 0 1 u h 01 1 u h 02
which have been indicated by the analysis. Uh 0 5 ]]]]] (17)
2
50 C. Ullner / Measurement 27 (2000) 43 – 51
6. Conclusion
The two determinations of the contact point accord- [1] A. Martens, in: Handbuch der Materialkunde fur ¨ den Mas-
ing to Eqs. (4), (5), (17), and (18) are compared in chinenbau (Handbook of Materials Science for Structural
Fig. 9. It can be seen that the upper limit of Engineering), Springer, Berlin, 1898, p. 234.
[2] W. Weiler, Hardness testing — a new method for economical
uncertainties is nearly equal and the difference
and physically meaningful microhardness testing, Br. J. Non-
between the contact points h 0 and h 02 is smaller destructive Test. 31 (1989) 253–258.
related to the size of ranges. However, the method [3] DIN 50359 Universal Hardness Testing, Part 1: Test Method,
according to Eq. (17), (18) yields real solutions for October, 1997.
C. Ullner / Measurement 27 (2000) 43 – 51 51
[4] H.-J. Weiss, On deriving Vickers hardness from penetration [7] P.M. Sargent, Use of the indentation size effect on mi-
depth, Phys. Stat. Sol. 99 (1987) 491–501. crohardness of material characterisation. In: P.J. Blau, B.R.
[5] W.C. Oliver, G.M. Pharr, An improved technique for de- Lawn (Eds.), ASTM STP No. 889, 1984, pp. 160–174.
termining hardness and elastic modulus using load and [8] P. Grau, Ch. Ullner, H.-H. Behncke, Uncertainty of depth
displacement sensing indentation experiments, J. Mater. Res. ¨
sensing hardness, Materialprufung 39 (9) (1997) 362–367.
7 (1992) 1564–1583. [9] Ch. Ullner, J. Beckmann, Effect of microstructure of
[6] Ch. Ullner, Different hardness parameter from continuous ceramics on the instrumented indentation test, in preparation.
force–depth indentation tests — modelling of their un- ¨
[10] P. Grau, G. Berg, W. Franzel, H. Meinhard, Recording
certainties, in: B. Michel, T. Winkler (Eds.), Proceedings of hardness testing. Problems of measurement at small indenta-
the Micro Mat ’97, Berlin, 1997, pp. 1166–1168. tion depth, Phys. Stat. Sol. 146 (1994) 537–548.