You are on page 1of 4

METROLOGIA-2003 – Metrologia para a Vida

Sociedade Brasileira de Metrologia (SBM)


Setembro 01−05, 2003, Recife, Pernambuco - BRASIL

UNCERTAINTY ANALYSIS OF POSITIONAL DEVIATIONS OF CNC


MACHINE TOOLS

Ignacio Lira, George Cargill

Pontificia Universidad Católica de Chile, Santiago, Chile

Abstract: The evaluation of the measurement uncertainty is The outline of the paper is as follows: section 2 summarizes
an indispensable task in all calibration procedures. By the method in ISO 230-2, section 3 presents our proposal,
international accord, this is to be done in accordance with section 4 discusses the proposal and section 5 provides the
the ISO Guide to the Expression of Uncertainty in conclusions.
Measurement (GUM). To measure the positional deviations
of computer numerically controlled (CNC) machine tools,
2. EVALUATION OF THE UNCERTAINTY IN THE
however, calibration laboratories will usually follow the
CALIBRATION OF CNC MACHINES ACCORDING
guidelines in ISO Standard 230-2, which does not comply
with the GUM. The main purpose of this paper is to present TO ISO 230-2
an uncertainty evaluation scheme that is firmly grounded in In the ISO 230-2 method the machine is programmed to
the GUM, and can therefore be of use as a guide to develop locate its moving part at a series of predefined target
appropriate uncertainty calculations in this and similar types positions along the axis under test. Fig. 1 shows a typical
of calibrations. test cycle; it consists of n unidirectional approaches in both
the positive (increasing) and negative (decreasing) directions
Keywords : uncertainty analysis, calibration, machine tools,
dimensional metrology. to m target positions. Symbols ↑ and ↓ are customarily used
to represent each of these directions of movement.
For point i in approach j, the positional deviation d ij is
1. INTRODUCTION
defined as the difference between the measured position lij
Various international standards provide guidance to and the target position p i , i.e.,
calibrate and verify the performance of computer
d ij = lij – pi (1)
numerically controlled (CNC) tool machines, such as
machining centers. In particular, ISO Standard 230-2 [1] is The unidirectional averages d i ↑ and d i ↓ are then computed.
accepted and used worldwide to determine the accuracy and Next, the experimental variances
positional repeatability of numerically controlled axes in
machine tools; it may be used for type testing, acceptance si 2 ↑ = ∑ j (d ij ↑ – d i ↑) 2 / ( n – 1) (2a)
tests, comparison testing, periodic verification and machine and
compensation.
si 2 ↓ = ∑ j (d ij ↓ – d i ↓) 2 / ( n – 1) (2b)
The method recommended in ISO 230-2 involves repeated
measurements at predefined positions of the axis under test, are calculated, these are taken to be the squares of the
which results are used to determine the positional undirectional standard uncertainties of the positional
deviations. The standard deviations of these measurements deviations. Finally, ISO 230-2 recommends to double the
are defined as the associated standard uncertainties, a result standard deviations to obtain the expanded uncertainties,
which does not agree with the rules and recommendations which are to be stated as the final uncertainty result in the
given in the ISO Guide to the Expression of Uncertainty in calibration certificate, possibly in graphic form.
Measurement (GUM) [2]. ISO 230-2 defines also a number of other parameters, such
In this paper we propose an uncertainty evaluation as repeatability (R) and accuracy (A). These other
procedure that differs from the method of ISO 230-2 on parameters will not be addressed here, since from the
several accounts: a) the measurement model is clearly discussion above it should be clear that the method in ISO
stated; b) from this model all relevant input quantities are 230-2 is at odds with the GUM: first, the variances (2) apply
identified; c) whenever possible, the standard uncertainties to each individually measured deviation, not to the reported
of these input quantities are derived following the averages; and second, a number of uncertainty sources are
procedures in the GUM; and d) for some input quantities not considered, namely: resolution, misalignment,
that are modeled nonlinearly in terms of other quantities temperature differences and calibration of the measuring
special methods are used. equipment.
3. A PROPOSAL TO EVALUATE THE This is the well-known type A evaluation of standard
UNCERTAINTY IN THE CALIBRATION OF CNC uncertainty [2]; it applies to a quantity that is measured
MACHINES repeatedly. The other quantities should instead be modeled
in terms of other quantities, whenever possible, or their
In order to evaluate any measurement uncertainty, the
uncertainties should be obtained as the standard deviations
quantities of interest must be defined precisely through one
of the applicable probability density functions (pdfs), the so-
or more measurement models. When the models are linear
called type B procedure.
or weekly nonlinear, they allow to express the standard
uncertainties of the output quantities in terms of the standard
3.2. Correction quantities
uncertainties of the input quantities by using the so-called
law of propagation of uncertainties, LPU [2,3]. Strictly, The thermal correction factor ft is usually modeled as
when applying this law all correlations between input
quantities should be taken into account. In the present case, ft = (1 + α∆T )−1 (8)
however, the common influences that may introduce where α is the linear coefficient of thermal expansion for the
interdependencies between measurements, such as scale’s material and ∆T is the difference between its
barometric pressure, temperature and humidity, are assumed temperature and the reference temperature (normally 20 °C).
to have negligible effects. Application of the LPU then gives

3.1. Basic model u 2 [ f t ] = ( ∆T 2 u 2 [α] + α2 u 2 [∆T] ) f t 4 (9)

The positional deviations will be modeled as Depending on the calibration procedure, the temperature
difference should be measured at several places along the
d i = li − p i (3) scale and at various times, for example, at each reversal of
where the li ´s are the corrected indications of the the measurement cycle. Doing this allows to establish a
measurement system, the p i ´s are the corrected target range of values wT for ∆T over which a rectangular pdf is
positions and the symbol  stands for either ↑ or ↓. assumed. Thus we set

In turn, the models for the quantities li  and p i should u 2 [∆T ] = wT2 /12 (10)
include all conceivable random and systematic effects. The coefficient α is normally not measured; its value is
While this depends on the particulars of the measurement imported from handbooks for the material one believes the
system, it is reasonable to assume that measurements are in scale to be made off. The standard uncertainty u[α ] should
general affected or limited by temperature, alignment,
then be evaluated from a rectangular pdf whose width wα
resolution and calibration of the test equipment. Therefore has to be assigned depending on how well one knows the
we write
material and on how well that material’s coefficient is
li = ft fa l mi + cc + cr m (4) known. In other words, wα corresponds to twice the
maximum possible error in the assignment of the value of α.
p i = p ni + crs (5)
With this assumption we get
where lmi  represents the arithmetic mean of the positions
u 2 [α] = wα2 /12 (11)
indicated by the measurement system in the n unidirectional
approaches to point p i , ft is a correction factor to compensate It should be noted, however, that in the vicinity of ∆T = 0
for thermal expansion of the positioning scale, fa is a equation (9) yields a value for u[ft ] that is independent of
correction factor to compensate for possible misalignment of the uncertainty in knowing α. This obviously unreasonable
the measurement system, cc is a correction associated with result is due to the nonlinearity of the model for the thermal
the calibration of the measurement system, crm is a expansion correction factor. Fortunately, unless wα is very
correction associated with the resolution of the measurement large, the non linearity of model (8) is weak and equation
system, p ni is the nominal (programmed) target position and (9) is adequate.
crs is a correction associated with the resolution of the axis
scale. Replacing (4) and (5) in (3), using the LPU and In turn, the alignment correction fa is modeled as
recognizing that no uncertainty should be attached to the fa = cos θ (12)
nominal positions, we obtain
where θ is the angle between the actual measurement
u 2 [d i ] = (ft fa u[lmi ]) 2 + (fa lmi u[ft ]) 2 + (ft lmi u[fa]) 2 + u 2 [cc] + direction and the axis direction. Because of the form of (12)
u 2 [crm] + u 2 [crs] (6) this correction factor is usually referred-to as the cosine
where the symbol  has been suppressed for clarity error. Application of the LPU gives

According to the GUM, the squares of the standard u[fa] = sin θ (13)
uncertainties of the averages lmi are evaluated as the However, normally the best estimate for the misalignment
variances (2) divided by the number n of unidirectional angle will be θ = 0, yielding u[fa] = 0 irrespective of the
measurements. Thus
uncertainty associated with θ. This is again an unreasonable
u 2 [lmi ] = si 2 / n (7) result, and is due to the strong non-linearity of cos θ in the
vicinity of θ = 0. Therefore, we recommend to estimate fa The measured deviations for a series of n = 5 positive
and its uncertainty from unidirectional approaches are given in Table 1, together with
the deviations computed from equation (3). Note that the
fa = sin θmax / θmax (14) latter are quite different from the average of the indicated
u 2 [fa] = 1/2 (fa cos θmax + 1) − fa 2 (15) deviations, this is due principally to the effect of the
temperature correction. Table 2 shows the numerical values
where θmax is an upper bound for the maximum deviation of each uncertainty term in the right-hand side of equation
angle, whose value has to be established based on the (6). It is seen that, except for the target positions closest to
particulars of the measurement system. Since normally θmax the axis´ origin, the contribution to the uncertainty of the
should be a small angle, fa should be nearly equal to 1 and calibration and resolution corrections are very small, while
u[fa] nearly equal to zero. Equations (14) and (15) are the contributions of the machine´s repeatability lmi , of the
derived in [3]. temperature correction factor ft and of the alignment
The rest of the correction quantities are easier to evaluate. correction factor fa, are of the same order of magnitude
The values of the resolution corrections crm and crs are both
taken as zero, but their standard uncertainties are evaluated 5. CONCLUSIONS
from rectangular pdfs of widths equal to each instrument’s
resolution, rm for the measurement system and rs for the axis In this paper we have presented a detailed procedure to
scale. Thus evaluate the uncertainty in the calibration of positional
deviations of the linear axes of CNC machine tools. In
u 2 [crm] = r m2 / 12 (16) contrast with the evaluation method proposed in currently
2 2
u [crs] = rs / 12 (17) used standard ISO 230-2, our method complies strictly with
the GUM´s recommendations: it is derived from a model for
Finally, the calibration correction cc depends on the the measurand, whose uncertainty is obtained through
instrument that is used to perform the measurements; its application of the law of propagation of uncertainties
value and associated standard uncertainty should normally together with type A and B evaluation methods.
be obtained from the instrument’s calibration certificate,
thus providing the traceability for the calibration of the CNC
machine REFERENCES
[1] ISO 230-2: Test Code for Machine Tools. Determination of
Accuracy and Repeatability of Positioning Numerically
4. DISCUSSION Controlled Axes. Geneva: International Organization for
From section 2 it is clear that ISO 230-2 considers the Standardization, 1997.
repeatability of the measuring system as the only source of [2] ISO Guide to the Expression of Uncertainty in Measurement.
uncertainty in the calibration of linear axes, and also that its Geneva: International Organization for Standardization,
proposed evaluation method does not comply with the rules 1995.
in the GUM. Our proposal corrects these deficiencies: it [3] Lira I. Evaluating the Uncertainty of Measurement:
properly uses the type A procedure and includes other Fundamentals and Practical Guidance. Bristol: Institute of
relevant sources of uncertainty. Some uncertainty Physics Publishing, 2002.
contributions may of course be negligible in comparison
with the others, but this cannot be known a priori.
In this section we show some typical results obtained in the Author: Dr Ignacio Lira, Departamento de Ingeniería Mecánica y
Metalúrgica, Pontificia Universidad Católica de Chile, Vicuña
calibration of a CNC machine using a laser interferometry
Mackenna 4860, Santiago, Chile. Tel.: +56 2 686 4629, fax: +56 2
system. The measurement system resolution is rm = 0.01 µm 686 5828, email: ilira@ing.puc.cl.
and the calibration certificate states that its relative error is er
= 0.1×10−7 with an expanded uncertainty U(er) = 2×10−7 and
a coverage factor k = 2. Therefore we take cci = − er p ni and
u[cci ] = U(er) p ni /2.
We used the system to calibrate the X axis of a CNC
mechanizing center with a scale resolution rs = 1 µm. The
programmed positions are shown in Table 1, these were
chosen with uneven spacing to avoid periodic sampling
errors. The temperature during calibration was (18 ± 0.5) o C,
such that ∆T = −2 o C (for a 20 o C reference temperature) and
wT = 1 o C. For the thermal expansion coefficient we took α=
11.7 × 10−6 oC−1 (for steel) with a maximum spread of 10 %
of this value to each side, giving wα = 2.34 × 10−6 oC−1 .
Finally, for θmax we chose 0.2o .
Table 1: Measured and calculated deviations for a series of n =5 positive unidirectional approaches to m = 11 nominal positions.

i p ni d i1 ↑ d i2 ↑ d i3 ↑ d i4 ↑ d i5 ↑ di↑

mm µm

0 0.000 0.00 0.00 0.00 0.00 0.00 0.00

1 55.327 7.35 5.05 5.98 6.22 5.78 7.26

2 113.260 8.38 6.06 6.25 7.15 6.94 9.38

3 169.810 6.88 4.80 5.20 5.46 5.56 9.21

4 222.723 7.94 5.82 6.22 6.25 6.59 11.32

5 280.000 16.35 14.63 14.47 14.56 15.32 21.05

6 336.145 20.71 18.13 18.66 18.30 18.78 26.10

7 393.418 24.50 21.62 21.80 22.27 22.17 30.88

8 446.111 29.07 25.65 26.32 26.38 26.62 36.34

9 504.917 23.01 19.57 19.88 20.37 20.80 31.51

10 555.234 24.90 21.49 22.22 22.37 22.40 34.54

Table 2: Components of the standard uncertainty, in microns, associated with the positional deviations in Table 1.

i fT fa u[lmi ] fa lmi u[fT] fT lmi u[fa] u[cc] u[crm] u[crs] u[d i ]

0 0.000 0.000 0.000 0.000 0.003 0.289 0.29

1 0.374 0.201 0.101 0.006 0.003 0.289 0.52

2 0.410 0.412 0.206 0.011 0.003 0.289 0.68

3 0.351 0.618 0.308 0.017 0.003 0.289 0.83

4 0.365 0.810 0.405 0.022 0.003 0.289 1.02

5 0.355 1.019 0.509 0.028 0.003 0.289 1.23

6 0.464 1.223 0.611 0.034 0.003 0.289 1.47

7 0.521 1.431 0.715 0.039 0.003 0.289 1.71

8 0.588 1.623 0.810 0.045 0.003 0.289 1.93

9 0.608 1.837 0.917 0.050 0.003 0.289 2.16

10 0.580 2.020 1.009 0.056 0.003 0.289 2.35

You might also like