Download as pdf or txt
Download as pdf or txt
You are on page 1of 10

Measurement 158 (2020) 107692

Contents lists available at ScienceDirect

Measurement
journal homepage: www.elsevier.com/locate/measurement

Investigation of the effect of temperature and other significant factors on


systematic error and measurement uncertainty in CMM measurements
by applying design of experiments
Branko Štrbac a, Bojan Ačko b, Sara Havrlišan c, Ivan Matin a, Borislav Savković a, Miodrag Hadžistević a,⇑
a
University of Novi Sad, Faculty of Technical Sciences, Department of Production Engineering, Trg Dositeja Obradovica 6, 21000 Novi Sad, Serbia
b
University of Maribor, Faculty of Mechanical Engineering, 2000 Maribor, Slovenia
c
Josip Juraj Strossmayer University of Osijek, Mechanical Engineering Faculty, 35000 Slavonski Brod, Croatia

a r t i c l e i n f o a b s t r a c t

Article history: The design of experiments was used in this paper in order to determine the significance of particular fac-
Received 7 November 2019 tors for the parameters that indicate the quality of measurement results obtained by means of a coordi-
Received in revised form 7 February 2020 nate measuring machine (CMM). Special emphasis was placed on the interaction of temperature and
Accepted 3 March 2020
other factors. In accordance with the design of experiments, the diameter and roundness of a ring gauge
Available online 18 March 2020
were measured at the reference temperature and at the temperature of 26 °C. The ISO 15530-3:2011
standard for the assessment of measurement uncertainty of CMM measurements distinguishes system-
Keywords:
atic error from measurement uncertainty. Thus, separate analyses were done on how temperature, in the
Temperature
Systematic error
first place, but also workpiece position on the CMM table, coordinate system, stylus tip diameter, sam-
Measurement uncertainty pling size and interaction of temperature with other factors affect systematic error and measurement
DoE uncertainty. The experimental results showed an increase in systematic error at the higher temperature
CMM when the diameter was measured, whereas with roundness, systematic error was even smaller at the
higher temperature. As for measurement uncertainty, measurements made at the higher temperature
have a significantly greater uncertainty when the diameter is measured, unlike with roundness, where
the effect of temperature is almost negligible.
Ó 2020 Elsevier Ltd. All rights reserved.

1. Introduction reference when investigating the accuracy of measuring systems


whose traceability is difficult to verify [3]. These are the reasons
Modern production requires designing and constructing work- why evaluating quality of CMM measurements is extremely impor-
pieces within exact specification limits with the aim of satisfying tant and why this has been one of the main research topics in the
functional requirements. Therefore, verification processes have to field of dimensional metrology over the past three decades [4].
be almost ideal and a quantitative indicator of the quality of mea- So far numerous experimental, simulation and analytical
surement results has to be given. Since measurement accuracy, methods of evaluating measurement uncertainty of CMM mea-
expressed via measurement error, can be assessed only when mea- surements [5–9] have been developed. Some of these have been
suring reference workpieces, it is necessary to assess the accompa- ISO standardized [10,11]. It is generally believed that including
nying measurement uncertainty in order to evaluate the quality of all influential factors and their interactions in the models of evalu-
measurement results [1,2]. ating uncertainty is impossible because CMM measuring is very
Coordinate measuring machines (CMMs) are most appreciated complex and measurement uncertainty is not a characteristic of
measuring instruments in verification procedures of dimensional the measuring instrument but of the measuring task itself [12].
and geometric specifications of workpieces. CMMs are successfully The following factors of uncertainty have been seen as the most
used both in state-of-the-art metrological laboratories and indus- influential ones: geometric errors of the machine, probe errors,
try, providing high measurement accuracy with complex products. number and position of points in the measuring strategy, form
Equally, measurement results obtained by CMMs can be used as a error of the workpiece, and evaluation algorithm. However, despite
the fact that temperature affects measurement uncertainty to a
great extent, this factor has not been fully investigated in CMM
⇑ Corresponding author. measurements [13]. The reason for this lies in the fact that the
E-mail address: miodrags@uns.ac.rs (M. Hadžistević).

https://doi.org/10.1016/j.measurement.2020.107692
0263-2241/Ó 2020 Elsevier Ltd. All rights reserved.
2 B. Štrbac et al. / Measurement 158 (2020) 107692

developed models are primarily intended for laboratory conditions between the examined factors and for assessing the conditions in
where deviation from the reference temperature is considered neg- which measurement uncertainty is the smallest and the greatest.
ligible. This mainly refers to simulation models. Kruth et al. based Barini et al. used the design of experiments in order to determine
their simulation model on modeling hardware errors and work- the effect of four factors on the value of shape deviation of a com-
piece form errors, and pointed out that future research should plex surface and on measurement uncertainty of measurements
include integration of virtual surroundings with an emphasis on made on a CMM [26]. C.-X.J. Feng et al. used the design of experi-
temperature effects [14]. Gaska et al. recognized this problem. ments in order to determine the effect of factors and their interac-
They extended their model [15] so it could be used in industry, tions on measurement uncertainty when measuring the position of
including the effect of temperature via appropriate distribution the centre of a hole by means of a CMM [27]. Piratelli-Filho and Di
of residual errors for different temperatures [16]. Their model Giacomo used CMM performance tests by measuring the ball bar
examined the effect of temperature on the machine hardware [28]. They used the 23 design in order to examine the effect of ori-
errors, whereas the effect of temperature on the workpiece was entation and length of the ball bar on measurement uncertainty. A.
omitted. Experimental and analytical models investigate the effect M.A. Al-Ahmari and J. Aalam used Taguchi designs for parameter
of temperature on measurement uncertainty only through the optimization when the CMM was used for reverse engineering,
coefficient of thermal expansion of the workpiece and the CMM i.e., for reconstruction of complex surfaces [29] Research data were
[17,18]. They do not take into account the fact that temperature obtained by means of a general full factorial plan of the experiment
has a strong effect on geometric errors of the machine which have with five factors (Table 1). The input factors coordinate system,
to do with the kinematics of the machine. stylus tip diameter, sampling size and temperature were varied
The models developed so far have mostly presumed that mea- on two levels, whereas the factor workpiece position on the
surements are made in temperature-controlled laboratories where CMM table was varied on four levels. The repetition of every com-
temperature is held at the reference temperature of 20 °C. How- bination of factor levels was carried out five times with the aim of
ever, a question arises over what happens when measurements investigating variation within each level. The total number of
are performed in industrial conditions, where the economic aspect experiments was 320 (41  24  5). The output variables were
is extremely important. When the aim is to reduce production the diameter and roundness of the ring gauge. The experiments
costs by reducing the time needed for the transport of workpieces were conducted during two days because it was impossible to con-
to metrological laboratories and the time needed for thermal stabi- duct a completely randomized experiment in one day due to the
lization of a workpiece, the effect of temperature becomes very temperature factor. On the first day, the experiments were carried
important [21]. Peres et al. investigated making measurements in out according to the DoE at the temperature of 20 °C. Four days
a machine tool while a workpiece is still in the fixture [22]. In such later, the experiments were carried out at the temperature of
cases, there are significant temperature deviations from the refer- 26 °C. Thus, the presented DoE was in two blocks [30].
ence temperature which affect both the CMM and the workpiece, The factor ‘‘workpiece position on the CMM table” includes the
and the effect of temperature has to be assessed. Generally, if pos- effect of CMM hardware errors on measurement uncertainty. The
sible, temperature compensation is advised in order to achieve uncertainty of a sampled point is not uniform across the CMM
necessary measurement accuracy [23,24]. However, it would be table and it is the resultant of 21 CMM geometric errors. This
interesting to examine how significant deviation from the refer- way it is considered that the resultants of these errors differ
ence temperature affects the uncertainty of CMM measurements depending on the position of the sampling system. When sampling
without activating temperature compensation. each point in the measurement plan, the sampling system starts
Unlike previous research, this investigation has experimentally from the reference coordinate system (MCS). As increase in tem-
determined how an increase in temperature which affects both the perature affects the value of CMM geometric errors, it is assumed
CMM and the workpiece influences the quality of CMM measure- that this factor will strongly affect the examined outputs when
ments. The design of experiments was used which, apart from tem- interacting with temperature. This factor is present on four levels,
perature, included several other factors of uncertainty. The aim i.e. the workpiece was positioned in the corners of the table (Fig. 1).
was to determine the significance of individual factors of uncer- Although standards recommend taking measurements in the cen-
tainty and their interactions with temperature. The experiment tre of the CMM table, this position was not used in this research
was conducted on a ring gauge and the uncertainty was assessed because the positions in the corners of the CMM table were consid-
for the diameter and form error (roundness) of the ring gauge. ered to have covered the ranges of geometric errors. The position
and orientation of the workpiece were chosen randomly by the
CMM operator. The holes on the CMM table served for approximate
2. The experimental set-up positioning 1, 2, 3 and 4.
The ‘‘coordinate system” factor considers the assessment of
The aim of the investigation was to examine the effect of tem- workpiece features if the coordinate system for the sampled points
perature on measurement uncertainty of a specific measuring task, is positioned in a different way. This way evaluation strategy was
as well as the interaction of temperature and other factors of taken into account. Fig. 1 also shows that on the first level of this
uncertainty. For that purpose, we measured a ring gauge with factor, coordinates of points are expressed using the machine coor-
the nominal diameter D = 60 mm on a Carl Zeiss Contura g2 RDS dinate system (MCS), whereas on the second level the coordinates
(Germany), coordinate measuring machine whose maximum per- of points are expressed using the coordinate system of the work-
missible error amounts to MPEE= (1.9 + L/330) mm (L is the length piece (PCS).
of measurement expressed in mm). The passive VAST XXT sensor The factor ‘‘stylus tip diameter” introduces mechanical filtration
was used for sampling in the ‘‘point – by - point” mode with Rpt, for sampling deviations from the workpiece. This factor is particu-
MPE = 1 mm repeatability. The point-by-point sampling mode was larly important when measuring roundness because it filters out
used for two reasons: (1) the maximum permissible error in this the effect of roughness and waviness on the form error (Fig. 2).
mode is smaller than with scanning [25] and (2) scanning speed The factor ‘‘sampling size” refers to the number of points which
is a factor that can have a significant effect on experimental results. describe real geometry (Fig. 3). A larger number of points in the
For the purpose of determining the significance of uncertainty fac- measurement strategy approximate real geometry more closely,
tors, the design of experiments (DoE) was applied. DoE is often whereas a smaller number of points shorten the time needed for
used in research and it is suitable for the analysis of interaction inspection. In this experiment, a measurement strategy with uni-
B. Štrbac et al. / Measurement 158 (2020) 107692 3

Table 1
Analyzed factors and levels.

Factor Level Code


Workpiece position on the CMM table 1 2 3 4 A
Coordinate system CMM Workpiece B
Stylus tip diameter 5 mm 15 mm C
Sampling size 15 150 D
Temperature 20 °C 26 °C E

Fig. 1. Factors ‘‘workpiece position on the CMM table ‘‘and ‘‘coordinate system and
Fig. 3. Factor ‘‘sampling size‘‘.
their levels.

form profile

Fig. 2. Factor ‘‘stylus tip diameter”.

form distribution of 15 and 150 points was used to measure the Fig. 4. Experimental set-up and temperature control at 20 °C and 26 °C.
circle. Sampling was performed in the ‘‘point- by- point” mode.
The CMM and the workpiece were placed in a temperature-
controlled metrological laboratory where temperature was main- Table 2
Calibrated values of the ring gauge.
tained within an interval of ±0.5 °C at the 95% confidence level.
The experiments were carried out at the temperatures of 20 °C Characteristic Reference value in mm Standard uncertainty in mm
and 26 °C. Temperature compensation of the CMM was not acti- Diameter 60.0013 0.07
vated during the experiment, whereas the maintenance of air tem- Roundness 0.0006 0.05
perature in the laboratory, on the workpiece and the machine was
observed by means of a type K thermocouple (Fig. 4). The four-
channel device PCE-T395, Germany, was used.
Before the experiment, the workpiece was calibrated on a from the calibrated values, the values obtained on the CMM were
length measurement machine using a laser interferometer and assessed by means of the least squares method (LS) for diameter
mechanical probing. Calibrated values together with standard and minimum zone method (MZ) for roundness. The Machine soft-
uncertainties are presented in Table 2. In order to assess deviation ware Calypso was used for obtaining the association feature.
4 B. Štrbac et al. / Measurement 158 (2020) 107692

The standard uncertainty of calibration was obtained from the


expanded measurement uncertainty with k = 2 coverage factor.

3. Measurement results and statistical analysis

Minitab 17 statistical software was used for the statistical anal-


ysis of the experimental data and the order of performing the ran-
domized experiment. The significance of factors and their
interactions for the deviation from reference values and standard
deviation were assessed by means of ANOVA analysis of variance
with the significance threshold of a = 0.05. As the focus of this
paper is the effect of temperature, the used models examined
two-factor interactions.
According to the revised ISO 15330-3:2011 standard, which is
considered to be the reference standard in assessing measurement Fig. 6. Interaction between position of workpiece on CMM table and temperature
uncertainty of CMM measurements, the systematic error b is not a for systematic error of diameter.
part of measurement uncertainty. However, it is a part of the over-
all measurement result and it has to be stated. The systematic error
is defined as the difference between the value obtained by means workpiece position on the CMM table and the number of points is
of a CMM, and the calibrated value. Therefore, this is an opportu- more significant (F = 22.08). Statistical analysis showed that apart
nity to perform a separate analysis of the effect of the examined from the mentioned factors and interactions, other significant fac-
factors on the systematic error, and a separate analysis of the effect tors include stylus tip diameter and interactions between work-
of these factors on measurement uncertainty. piece position*Stylus tip diameter, Stylus tip diameter*Sampling
size, and Sampling size*Temperature. It can be seen that stylus
3.1. Analysis of a systematic error tip diameter and the number of points have a greater effect on
the systematic roundness error than on the systematic diameter
As assumed, statistical analysis of the experimental data error. In this case, the adequacy of the model was R-sq
showed a high significance of the effect of temperature and work- (pred) = 83.36%.
piece position on the CMM table (Fig. 5) and their interaction on In both analyses, the effect of coordinate system on the system-
the systematic error of diameter measurement (Fig. 6). Based on atic error was almost negligible. The effect of this factor would be
F–statistics, the most significant factors that affect the systematic more significant in measuring workpieces characterized by greater
error of diameter are temperature (F = 18605,47), workpiece posi- form deviation (production parts). This also applies to the effect of
tion on the CMM table (F = 308) and their interaction (F = 172). stylus tip diameter and number of points, especially on the form
Equally, based on the given significance threshold and the error.
assessed value p (p < 0.05), other statistically significant factors Considering the reference values of the measured characteris-
include stylus tip diameter, sampling size, and the following inter- tics, it can be seen that systematic measurement errors are greater
actions: workpiece position*Stylus tip diameter, workpiece posi- when the diameter is measured at higher temperature, whereas
tion*Sampling size and Sampling size*Temperature. The they remain almost unchanged when roundness is measured at
statistical conclusions were drawn based on the adequate model the higher temperature (Fig. 9). However, when the values of stan-
R-sq (pred) = 98.32%. dard deviation for both characteristics are analyzed, these values
The same analysis was performed in measuring roundness. As are lower when measurements are made at the higher tempera-
with diameter, the most significant factors affecting form error ture. When the diameter is measured, standard deviation amounts
are temperature (F = 37.44) and workpiece position on the CMM to 0.62 mm at the temperature of 20 °C, whereas it is 0.47 mm at the
table (F = 25.94) (Fig. 7). It can be seen that systematic errors in temperature of 26 °C. With roundness, these values amount to
roundness measurements are less significant at higher tempera- 0.95 mm and 0.6 mm respectively.
tures. The interaction of these two factors is significant as well Tables 3 and 4 present the minimum, maximum and mean val-
(F = 17.86) (Fig. 8). However, in this case, the interaction between ues of the systematic error in measuring diameter and roundness.

Fig. 5. Main effects plot for systematic error of diameter.


B. Štrbac et al. / Measurement 158 (2020) 107692 5

Fig. 7. Main effects plot for systematic error of roundness.

3.2. Assessment of measurement uncertainty

The research into the topic of assessing measurement uncer-


tainty in CMM measurements has mostly relied on the following
methods so far: Guide to the expression of uncertainty in measure-
ment (GUM), expert judgement, and standardized methods for
CMMs - ISO 15530-3 and ISO 15530-4 [12]. As the conditions
required by the standard ISO 15530-3:2011 have been satisfied,
measurement uncertainty in this paper was assessed according
to this standard. This standard demands repeated measurements
of the calibrated piece (at least 20) in a way that the workpiece
takes a different position and orientation in the CMM workspace
every time. In order to include as many uncertainty factors as pos-
sible, apart from position and orientation of the workpiece (factor
1), a few other influential factors were varied using the DoE. Since
Fig. 8. Interaction between position of workpiece on CMM table and temperature in the newer version of this standard the systematic error is sepa-
for systematic error of roundness. rated from measurement uncertainty, uncertainties assessed
according to this standard are much smaller than with, for exam-
ple, the simulation method [16]. According to this standard, the
expanded measurement uncertainty is calculated using formula
(1).
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
U ¼k u2cal þ u2p þ u2b þ u2w ; ð1Þ

The standard uncertainty of repeated measurement (up) can be


determined using the expression (2):
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 2
1 Xn 1 Xn
up ¼ x i  x i : ð2Þ
n1 i¼1 n i¼1

Tables 5 and 6 show the mean, minimum and maximum values


of the standard uncertainty at two different temperatures.
The tables show that the mean and the maximum values of
standard uncertainty are smaller at the higher temperature for
both measured characteristics. Standard uncertainty was derived
from the standard deviation calculated from the repeated mea-
surements. Thus, it can be said that this specific task has a better
repeatability at the higher temperature than at the reference tem-
perature. The reason for this might be in the fact that while the
temperature was being held at 26 °C, the temperature gradient
was lower. It can also be seen in the tables that the minimum val-
ues of standard uncertainty are 0 in both cases.
The values of standard uncertainties vary depending on the
level of the investigated factors. Tables 7 and 8 present levels of
factors that correspond to the minimum and maximum values of
the standard uncertainty of diameter and roundness, respectively.
Fig. 9. CMM systematic errors in diameter and roundness measurements at Figs. 10 and 11 show a considerable effect of hardware errors on
different temperatures. the standard uncertainty, which is expected. Unlike measurement
6 B. Štrbac et al. / Measurement 158 (2020) 107692

Table 3
Analysis of systematic diameter error at different temperatures.

Temp in °C Mean in mm Min in mm Max in mm


Systematic error 20 0.0005 0.0024 0.0004
Diameter 26 0.0026 0.0016 0.0036

Table 4
Analysis of systematic roundness error at different temperatures.

Temp in °C Mean in mm Min in mm Max in mm


Systematic error 20 0.0008 0.0001 0.0044
Roundness 26 0.0005 0.0003 0.0034

Table 5
Analysis of standard uncertainty of diameter error at different temperatures.

Temp in °C Mean in mm Min in mm Max in mm


St Unc Err Dia 20 0.00072 0 0.00137
26 0.00037 0 0.00061

Table 6
Analysis of standard uncertainty of roundness error at different temperatures.

Temp in °C Mean in mm Mini in mm Max in mm


St Unc Err Rou 20 0.00074 0 0.00169
26 0.00067 0 0.00162

Table 7
Levels of factors for minimum/maximum values of standard uncertainty for diameter error.

Factors A B C D E St Dev
Levels 2 CMM 5 150 26 min
3 Workpiece 5 15 20 max

Table 8
Levels of factors for minimum/maximum values of standard uncertainty for roundness error.

Factors A B C D E St Dev
Levels 4 CMM 5 150 26 min
2 Workpiece 15 15 20 max

errors where the factor of coordinate system was negligible, this of temperature according to the GUM (ISO JCGM 100:2008) Annex
factor has a statistically significant effect on measurement uncer- H.1. [32].
tainty. The standard uncertainty for both measured characteristics The first thing that can be determined is the uncertainty of the

is somewhat larger if coordinates of the sampled points are average thermal expansion uðaÞ. The interval of ±1  106 °C1 is
expressed via the workpiece coordinate system. The explanation defined from the literature on steel. Standard uncertainty at the
for this lies in the fact that when the workpiece coordinate system supposed rectangular distribution is, formula (4):
is used, a new coordinate system has to be defined every time by
measuring geometric primitives that constitute it. The strategy of 1  106  C1
sampling these primitives is always different and chosen by the ub2 ¼ uw2 ¼ uðaÞ ¼ pffiffiffi ¼ 0:58  106  C1 ð4Þ
3
operator, which is the source of uncertainty.
Mean values of standard uncertainty shown in Tables 5 And 6 The sensitivity coefficient in this case is 0.1 °C1  L.
were used to assess total uncertainty. Also, the uncertainty of temperature difference can be deter-
According to ISO 15530-3, ub and uw refer to the workpiece and mined. The ring is fixed on the same base plate and the tempera-
CMM because CMM do not use temperature compensation. These ture of the plate is measured. The temperature of the plate
values are calculated as, formula 3: during the measurement is limited to (20 ± 0.5) °C and (26 ± 0.5)
 °C. The difference interval of 0.1 °C is defined based on some exper-
ub1 ¼ uw1 ¼ ðT  20 CÞ  ua  L ð3Þ iments. Standard uncertainty at the assumed normal distribution
Since there was no temperature compensation, apart from the is:
uncertainties ub1 and uw1 (uncertainty of the coefficient of expan- 

sion the calibrated ring gauge ua = 0.12  106 L [31]) this paper 0:1 
uw3 ¼ ub3 ¼ uðdT Þ ¼ C ¼ 0:05 C ð5Þ
will include a few more standard uncertainties related to the effect 2
B. Štrbac et al. / Measurement 158 (2020) 107692 7

Fig. 10. Main effects plot for standard uncertainty of diameter error.

Fig. 11. Main effects plot for standard uncertainty of roundness error.

It is also necessary to determine the standard uncertainty of the Since the measurement is comparative, this difference is influ-
difference between the expansion coefficients of the CMM (pro- encing the measurement result only on the half measurement
ducer’s specification: a = 7.8  106 °C1) and the calibrated ring length in the worst case. Therefore, we use L/2 in the sensitivity
(steel; a = 11  106 C1) Ida = 4  106 C1 is supposed as the coefficient.
deviation interval. Standard uncertainty at the supposed rectangu- The uncertainty of average temperature deviation from 20 °C
lar distribution is, formula (6): and 26 °C u(h) can also be considered. The temperature expansion
6 
pffiffiffi is not corrected, but the allowed temperature deviation during the
uw4 ¼ ub4 ¼ uðdaÞ ¼ 4  10 C = 3 ¼ 2:3  106  C1
1
ð6Þ
8 B. Štrbac et al. / Measurement 158 (2020) 107692

calibration is limited to ±0.5 °C. Standard uncertainty at the sup- The overall measurement results for the two measured charac-
posed rectangular distribution is, formula (7): teristics at different temperatures are given in Fig. 12. The mea-
pffiffiffi surement result (D, R) and the systematic error b are presented
 
uw5 ¼ ub5 ¼ uðTmÞ ¼ 0:5 C= 3 ¼ 0:29 C ð7Þ as mean values.
In Fig. 12 the effect of temperature on the systematic error b
Since the measurement is comparative, we apply mean differ-
and on the total measurement uncertainty U can be seen when
ence in a in sensitivity coefficient calculation. Average difference
the diameter (length) is measured. Therefore, when dimensional
in a is assumed to be 2.5  106 °C1.
characteristics are measured at the higher stable temperature of
The uncertainty budget for standard uncertainties ub2 and uw2,
the CMM and the workpiece, less accurate measurement values
ub3 and uw3, ub4 and uw4, ub5 and uw5 that will be named uncertainty
and higher measurement uncertainty are expected. However, in
components ubT and uwT is presented in Table 10.
the experiment, repeatability (one of the uncertainty components)
What follows is that standard uncertainty uw, i.e. ub is, formula
was greater at the higher temperature than at the reference tem-
(8):
perature. As for measuring roundness (form), the situation with
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi the results is opposite. The systematic error and the total measure-
ub ¼ uw ¼ u2bT þ u2b1 ð8Þ

The last standard uncertainty is the uncertainty of the ring


gauge calibration ucal. Table 9 presents the measurement uncer-
tainty budget for measuring diameter at the reference tempera-
ture. Combined uncertainty uAB is calculated from all the
mentioned standard uncertainties, and based on the effective
degree of freedom, using t-distribution, k = 2 coverage factor is
chosen for 95% probability. Based on this, expanded measurement
uncertainty U95 is calculated. The table also presents other
expanded uncertainties for measuring diameter at the higher tem-
perature and measuring roundness both at the reference and the
higher temperature. When assessing uncertainty of roundness
measurements, standard uncertainties which contain temperature
effects have the value of 0 because L  0. Consequently, apart from
being present in the CMM and the workpiece, temperature effects
of measurement uncertainty cannot be contained in the standard
uncertainties ub and uw. The observations made by Jakubiec should
also be mentioned [19]. In order to obtain these values, it is neces-
sary to include, along with the coefficients of thermal expansion,
the value of length. However, when geometric characteristics such
as straightness, flatness, roundness or cylindricity are measured, a Fig. 12. Overall measurement results defined via mean values of measurement,
systematic error and expanded measurement uncertainty.
question arises about what length L is [20].

Table 9
Uncertainty budget for measuring diameter at 20 °C and expanded uncertainties for all cases.

Uncertainty component Source of uncertainty Distribution Standard Value of standard Effective degree of
Symbol for measurement uncertainty type measurement uncertainty in freedom
uncertainty component lm
up Repeated measurement according to none A 0.72
DoE
uw1 = ub1 From the manufacturing process and uniform B (20.5–20)ua  L
the systematic error
uwT = ubT Other temperature effects uniform B 1.37106  L
ucal Uncertainty of ring gauge calibration normal B 0.07
uAB Combined uncertainty 0.79 >30
U95 Expanded uncertainty of diameter 1.59
measurements at 20 °C
U95 Expanded uncertainty of diameter 3.98
measurements at 26 °C
U95 Expanded uncertainty of roundness 1.48
measurements at 20 °C
U95 Expanded uncertainty of roundness 1.34
measurements at 26 °C

Table 10
The uncertainty budget for other temperature effects.

Quantity Xi Standard uncertainty u(xi) Distribution Sensitivity coefficient ci Uncertainty contribution ui(y) = ci  u(xi)

a 0.58  106°C1 Rectangular 0.1 °C1  L 0.06  106  L
uðdT Þ 0.05 °C Normal 9.5  106°C1  L 0.47  106  L
da 2.3  106°C1 Rectangular 0.5 °C  L/2 0.57  106  L
h 0,29 °C Rectangular 2.5  106°C1 0.72  106  L
Total ubT and uwT 1.37  106  L
B. Štrbac et al. / Measurement 158 (2020) 107692 9

ment uncertainty have lower values when measurements are Declaration of Competing Interest
made at the higher temperature. This seems to be very interesting
since the assessment of CMM measurement uncertainty in mea- The authors declare that they have no known competing finan-
suring form error is much more interesting for investigation than cial interests or personal relationships that could have appeared
the dimensional measurements. Based on the results of the exper- to influence the work reported in this paper.
iment, it can be concluded that in measuring the form error, if the
CMM and the workpiece are thermally stable at the higher temper-
ature, the CMM is more accurate and more precise. The research References
[15] also presented reduced measurement uncertainty when mea-
[1] B. Acko, S. Brezovnik, L. Crepinsek Lipus, R. Klobucar, Verification of statistical
suring the form error (flatness and cylindricity) on the CMM if
calculations in interlaboratory comparisons by simulating input datasets, Int. J.
measurements were made at temperatures other than the refer- Simulat. Model. 14(2) (2015) A004, 227–237. https://doi:10.2507/IJSIMM 14
ence temperature. (2)4.288.
[2] R. D’Amato, J. Caja, P. Maresca, E. Gómez, Use of coordinate measuring machine
The results of this investigation can be applied in measuring
to measure angles by geometric characterization of perpendicular planes.
real workpieces because the most significant factors (CMM hard- Estimating uncertainty, Measurement 47 (2014) 598–606, https://doi.org/
ware and temperature) are the same for this case, whereby the 10.1016/j.measurement.2013.09.027.
amount of measurement uncertainty will probably be expanded [3] A. Horvatic Novak, B. Runje, J. Stepanic, Capabilities of industrial computed
tomography in the field of dimensional measurements, Adv. Product. Eng.
with uncertainty factors such as workpiece form deviation, sam- Manage. 12 (3) (2017) 245–253, https://doi.org/10.14743/apem2017.3.255.
pling strategy, their interaction, and so on. [4] W. Płowucha, Point-straight line distance as model for uncertainty evaluation
of coordinate measurement, Measurement 135 (2019) (2019) 83–95, https://
doi.org/10.1016/j.measurement.2018.11.008.
[5] C. Cui, S. Fu, F. Huang, Research on the uncertainties from different form error
evaluation methods by CMM sampling, Int. J. Adv. Manuf. Technol. 43 (1–2)
4. Conclusion (2009) 136–145, https://doi.org/10.1007/s00170-008-1681-4.
[6] H. Li, X. Chen, Y. Cheng, H. Liu, H. Wang, Z. Cheng, H. Wang, Uncertainty
Evaluation of CMM measurement uncertainty is a challenging modeling and evaluation of CMM task-oriented measurement based on
SVCMM, Measur. Sci. Rev. 17 (5) (2017) 226–231, https://doi.org/10.1515/
task. It is necessary to include all influential factors in the evalua- msr-2017-0027.
tion model and describe them adequately. Deviation from the ref- [7] J. Beaman, E. Morse, Experimental evaluation of software estimates of task
erence temperature can affect measurement uncertainty to a great specific measurement uncertainty for CMMs, Precis. Eng. 34 (1) (2010) 28–33,
https://doi.org/10.1016/j.precisioneng.2009.08.006.
extent. The design of experiments used in this research included
[8] R. Calvo, R. D’Amato, E. Gómez, R. Domingo, Integration of error compensation
some of the most influential factors affecting measurement uncer- of coordinate measuring machines into feature measurement: Part I—Model
tainty. It was examined how an increase in temperature interacting development, Sensors 16 (10) (2016) 1610, https://doi.org/10.3390/
s16101610.
with other factors of uncertainty affects quality parameters of a
[9] R. Calvo, R. D’Amato, E. Gómez, R. Domingo, Integration of error compensation
CMM. Temperature compensation of the CMM was deactivated of coordinate measuring machines into feature measurement: Part II—
during the experiment. Measuring the diameter and roundness of experimental implementation. Sensors 16(10) (2016) 1705. https://doi.org/
a ring gauge, it was found that an increase in temperature has a 10.3390/s16101705.
[10] ISO/TS 15530-3:2011. Geometrical product specifications (GPS) – coordinate
significant effect on the systematic diameter error, whereas its measuring machines (CMM): technique for determining the uncertainty of
influence on roundness is considerably weaker. Additionally, measurement – part 3: use of calibrated workpieces or standards.
strong interaction was observed between temperature and posi- [11] ISO/TS 15530-4:2008. Geometrical product specifications (GPS) – coordinate
measuring machines (CMM): technique for determining the uncertainty of
tion of workpiece on the CMM table which includes the effect of measurement – part 4: evaluating task-specific measurement uncertainty
geometric errors of the machine. As for measurement uncertainty, using simulation.
the effect of temperature is conspicuous when measuring the [12] R.G. Wilhelma, R. Hockena, H. Schwenkeb, Task specific uncertainty in
coordinate measurement, CIRP Ann. – Manufact. Technol. 50 (2001) 553–
diameter, whereas in measuring roundness, the value of uncer- 563, https://doi.org/10.1016/S0007-8506(07)62995-3.
tainty is smaller than the value of uncertainty at the reference tem- [13] R. Hocken, P . Pereira (editors), Coordinate measuring machines and systems,
perature. Experimental research is useful for measurements in 2nd ed, CRC Press, 2011.
[14] J.-P. Kruth, N. Van Gestel, P. Bleys, F. Welkenhuyzen, Uncertainty
industry because the majority of measurements are made at tem-
determination for CMMs by Monte Carlo simulation integrating feature form
peratures which significantly deviate from the reference tempera- deviations, CIRP Ann. – Manufact. Technol. 58 (1) (2009) 463–466, https://doi.
ture. Although this case study showed better results at the higher org/10.1016/j.cirp.2009.03.028.
[15] J. Sładek, A. Ga˛ska, Evaluation of coordinate measurement uncertainty with
temperatures, the general conclusion is that measuring roundness
use of virtual machine model based on Monte Carlo method, Measurement 45
at higher temperatures, which is often the case in industry, will not (6) (2012) 1564–1575. https://doi.org/10.1016/j.measurement.2012.02.020.
affect the measurement error or measurement uncertainty to a [16] A. Ga˛ska, W. Harmatys, P. Ga˛ska, M. Gruza, K. Gromczak, K. Ostrowska, Virtual
great extent. Future research would include the development of a CMM-based model for uncertainty estimation of coordinate measurements
performed in industrial conditions, Measurement 98 (2017) 361–371, https://
simulation model which would, apart from modeling hardware doi.org/10.1016/j.measurement.2016.12.027.
structure and measurement strategy, model the behavior of CMMs [17] I. Vrba, R. Palencar, M. Hadzistevic, B. Strbac, V. Spasic-Jokic, J. Hodolic,
at higher temperatures, but also the behavior of workpieces at Different approaches in uncertainty evaluation for measurement of complex
surfaces using coordinate measuring machine, Measurement Sci. Rev. 15 (3)
higher temperatures. (2015) 111–118, https://doi.org/10.1515/msr-2015-0017.
[18] T. Tasic, B. Acko, Integration of a laser interferometer and a CMM into a
measurement system for measuring internal dimensions, Measurement 44 (2)
(2011) 426–433, https://doi.org/10.1016/j.measurement.2010.11.002.
CRediT authorship contribution statement [19] W. Plowucha, W. Jakubiec, Proposal for changes in the ISO series of standards.
Quality – Access to Success 13(5) (2012) 237–240.
[20] B. Štrbac, V. Radlovački, V. Spasić-Jokić, M. Delić, M. Hadžistević, The difference
Branko Štrbac: Conceptualization, Methodology, Investigation, between GUM and ISO/TC 15530–3 method to evaluate the measurement
Writing - original draft. Bojan Ačko: Methodology, Investigation, uncertainty of flatness by a CMM, MAPAN 32 (2017) 251–257, https://doi.org/
10.1007/s12647-017-0227-3.
Writing - review & editing, Resources. Sara Havrlišan: Conceptual-
[21] R. Klobucar, B. Acko, Experimental evaluation of ball bar standard thermal
ization, Writing - review & editing. Ivan Matin: Conceptualization, properties by simulating real shop floor conditions, Int. J. Simulat. Modell. 15
Investigation, Writing - review & editing. Borislav Savković: (3) (2016) 511–521, https://doi.org/10.2507/IJSIMM15(3)10.356.
Writing - review & editing, Visualization. Miodrag Hadžistević: [22] P. Pérez, S. Aguado, J.A. Albajez, J. Santolaria, Influence of laser tracker noise on
the uncertainty of machine tool volumetric verification using the Monte Carlo
Conceptualization, Methodology, Writing - review & editing, method, Measurement 133 (2019) 81–90, https://doi.org/10.1016/j.
Supervision. measurement.2018.10.012.
10 B. Štrbac et al. / Measurement 158 (2020) 107692

[23] S. Naeem, S. Mian, A.P. Longstaff Fletcher, Reducing the latency between [28] A. Piratelli-Filho, B. Di Giacomo, CMM uncertainty analysis with factorial
machining and measurement using FEA to predict thermal transient effects on design, Precis. Eng. 27 (3) (2003) 283–288, https://doi.org/10.1016/S0141-
CMM measurement, Measurement 135 (2019) 260–277, https://doi.org/ 6359(03)00035-7.
10.1016/j.measurement.2018.11.034. [29] A.M.A. Al-Ahmari, Javed Aalam. Optimizing parameters of freeform surface
[24] J.-P. Kruth, P. Vanhercka, C. Van den Bergh, Compensation of static and reconstruction using CMM, Measurement 64 (2015) 17–28, http://dx.doi.org/
transient thermal errors on CMMs, CIRP Annals, 50(1) (2001) 377–380, 10.1016/j.measurement.2014.12.031.
https://doi.org/10.1016/S0007-8506(07)62144-1. [30] K. Simunovic, G. Simunovic, T. Saric, Single and multiple goal optimization of
[25] ISO 10360-2:2009: Geometrical product specifications (GPS) — Acceptance structural steel face milling process considering different methods of cooling/
and reverification tests for coordinate measuring machines (CMM), Geneva. lubricating, J. Cleaner Prod. 94 (2015) 321–329, https://doi.org/10.1016/j.
[26] E.M. Barini, G. Tosello, L. De Chiffre, Uncertainty analysis of point-by-point jclepro.2015.02.015.
sampling complex surfaces using touch probe CMMs. DOE for complex [31] T. Doiron, J. Stoup, Uncertainty and dimensional calibrations, J. Res. Nat. Inst.
surfaces verification with CMM, Precis. Eng. 34 (1) (2010) 16–21, https://doi. Stand. Technol. 102 (6) (1997) 647–676, https://doi.org/10.6028/jres.102.044.
org/10.1016/j.precisioneng.2009.06.009. [32] BIPM, IEC, IFCC, ILAC, ISO, IUPAC, IUPAP and OIML, Guide to the Expression of
[27] C.-X.J. Feng, A.L. Saal, J.G. Salsbury, A.R. Ness, G.C.S. Lin, Design and analysis of Uncertainty in Measurement, JCGM 100:2008 (GUM 1995 with minor
experiments in CMM measurement uncertainty study, Precis. Eng. 31 (2) corrections), 2008. http://www.bipm.org/utils/common/documents/jcgm/
(2007) 94–101, https://doi.org/10.1016/j.precisioneng.2006.03.003. JCGM_100_2008_E.pdf.

You might also like