You are on page 1of 14

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/271195042

Simplified model to estimate uncertainty in CMM

Article in Journal of the Brazilian Society of Mechanical Sciences and Engineering · January 2015
DOI: 10.1007/s40430-014-0157-8

CITATIONS READS

14 2,735

4 authors:

Rosenda Arencibia Cláudio Costa Souza


Universidade Federal de Uberlândia (UFU) Universidade Federal de Uberlândia (UFU)
88 PUBLICATIONS 270 CITATIONS 4 PUBLICATIONS 34 CITATIONS

SEE PROFILE SEE PROFILE

Henara Costa Antonio Piratelli-Filho


Universidade Federal do Rio Grande (FURG) University of Brasília
110 PUBLICATIONS 3,722 CITATIONS 61 PUBLICATIONS 365 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Antonio Piratelli-Filho on 21 January 2015.

The user has requested enhancement of the downloaded file.


Simplified model to estimate uncertainty in
CMM

Rosenda Valdés Arencibia, Cláudio


Costa Souza, Henara Lilian Costa &
Antônio Piratelli-Filho

Journal of the Brazilian Society of


Mechanical Sciences and Engineering

ISSN 1678-5878
Volume 37
Number 1

J Braz. Soc. Mech. Sci. Eng. (2015)


37:411-421
DOI 10.1007/s40430-014-0157-8

1 23
Your article is protected by copyright
and all rights are held exclusively by The
Brazilian Society of Mechanical Sciences and
Engineering. This e-offprint is for personal
use only and shall not be self-archived in
electronic repositories. If you wish to self-
archive your article, please use the accepted
manuscript version for posting on your own
website. You may further deposit the accepted
manuscript version in any repository,
provided it is only made publicly available 12
months after official publication or later and
provided acknowledgement is given to the
original source of publication and a link is
inserted to the published article on Springer's
website. The link must be accompanied by
the following text: "The final publication is
available at link.springer.com”.

1 23
Author's personal copy
J Braz. Soc. Mech. Sci. Eng. (2015) 37:411–421
DOI 10.1007/s40430-014-0157-8

TECHNICAL PAPER

Simplified model to estimate uncertainty in CMM


Rosenda Valdés Arencibia • Cláudio Costa Souza •

Henara Lilian Costa • Antônio Piratelli-Filho

Received: 16 February 2012 / Accepted: 15 October 2013 / Published online: 12 March 2014
 The Brazilian Society of Mechanical Sciences and Engineering 2014

Abstract This paper aims to present and discuss simpli- Nomenclature


fied models to estimate the measurement uncertainty when a, b and c Variables
determining circularity and cylindricity deviations using DCIR Circularity deviation, lm
coordinate measuring machines (CMM). The work was DCYL Cylindricity deviation, lm
carried out according to the following steps: (i) analysis of Di Distance of each point Pi to the centre of
the known Standards and documents used for uncertainty the circle (Pc), mm
determination; (ii) analysis of the measurement system EA Errors of the CMM probing system, lm
(CMM), identifying the variables that influence the deter- f Function
mination of geometry deviations; (iii) development of a k Coverage factor
mathematical model for the measurement process; (iv) L CMM indication, mm
development of a routine to evaluate uncertainty; (v) vali- MQ Least squares
dation of the model through experimental tests, comparing n Number of measurement cycles
the obtained results with those using models existing in the Pmax Most distant point in relation to Pc
literature. It was concluded that the models presented are Pmin Least distant point in relation to Pc
adequate and easy to be applied. R CMM resolution, mm
r Radius of the circle, mm
Keywords Uncertainty  Form deviations  Cylindricity  s Standard deviation, mm
Circularity u Standard uncertainty, lm
uc Combined standard uncertainty, lm
Up Expanded uncertainty, lm
Up 2003 The values obtained with the method
proposed by Vieira Sato [42], lm
Technical Editor: Marcelo Trindade. Up 2011 The values obtained with the method
here proposed, lm
R. V. Arencibia (&)  C. C. Souza  H. L. Costa
Var(T) Temperature variation during
Faculty of Mechanical Engineering, Universidade Federal de
Uberlândia, Uberlândia, MG 38400-902, Brazil measurement, C
e-mail: arvaldes@mecanica.ufu.br W1, W2,…, WN Input variables
C. C. Souza X, Y and Z Coordinates of point, mm
e-mail: claudio_costasouza@hotmail.com xc and yc Coordinates of the centre of the circle,
H. L. Costa mm
e-mail: ltm-henara@ufu.br V Output variable

A. Piratelli-Filho
Greek symbols
Department of Mechanical Engineering, Faculty of Technology,
Universdade de Brası́lia, Brası́lia, DF 70910-900, Brazil aE Coefficient of thermal expansion of the scale,
e-mail: pirateli@unb.br C-1

123
Author's personal copy
412 J Braz. Soc. Mech. Sci. Eng. (2015) 37:411–421

ap Coefficient of thermal expansion of the Sousa and Wandek [36] have identified deficiencies in
workpiece, C-1 the understanding and proper application of GD&T in
Da Difference between coefficients of thermal design and manufacturing of parts in Brazil. The main
expansion of the scale and workpiece, C-1 deficiencies found were related to a superficial knowledge of
D(Pmax) Correction associated with the Pmax the existing standards for geometrical specification of parts
D(Pmin) Correction associated with the Pmin and to a little experience of designers and manufacturing
DI Uncertainty associated with the CMM workers in the extrapolation of such knowledge to help
indication system, lm defining good measurement practices, which have caused
Ds(DCIR) Correction associated with the circularity various problems for Brazilian industry. According to the
deviation, lm authors, in addition to the difficulties related to GD&T
Ds(DCYL) Correction associated with the cylindricity specifications, difficulties to evaluate uncertainty measure-
deviation, lm ments and to verify if they are adequate according to the
DEA Correction associated with the errors of the specified tolerance are another problem. It is important to
CMM probing system, lm invest time and money in a series of actions to minimize
DIT Correction due to the uncertainty associated such problems, in particular in the qualification of personnel
with the thermometer indication system, C involved in the whole design and manufacturing chain.
DR Correction associated with the CMM Also, dimensional and geometrical control goes beyond
resolution, lm the manufacturing chain and gains vital importance in
DRT Correction due to the thermometer resolution, manufacturing research which can enable many manufac-
C turing processes to be improved [5, 13, 34, 39].
DT Difference between the room temperature and In this context, coordinate measuring machines (CMM)
20 C, C are metrological instruments that have been proving to
DT20 Correction due to the difference between the meet modern manufacturing demands with respect to
room temperature and 20 C, C dimensional and geometric control. They enable the mea-
DdT Correction due to the temperature variation surement of complex structures and the simultaneous
during measurement, C control of various metrological characteristics, and there-
fore have been replacing a large proportion of instruments
Subscripts conventionally used in metrology laboratories.
I Number of points CMMs are fast, accurate, flexible and allow a reliable
N Number of input variables quality control. Nevertheless, their performance has been
limited by several factors, which can have complex inter-
actions, generating the so-called volumetric error. The
main error sources that can limit the performance of a
1 Introduction
CMM include:
The main goal of any machining operation is to produce • Geometric errors: they represent the largest fraction of
interchangeable parts with maximum functionality at reason- the volumetric error [8] and are considered to have the
able costs. Such need requires each part or assembly of parts of strongest effect on accuracy and precision for mea-
a final product to be manufactured according to predefined surements using CMMs [11, 27]. The systematic nature
specifications for dimensions, geometry and surface finish. of the geometric errors allows their correction or
The evolution of machining tools has enabled the compensation via computational programs [6, 23, 46].
manufacturing of parts to be faster and more practical. New • Measuring probe errors: these errors are associated with
technologies implemented in machining processes have electric, kinematic and mechanical characteristics of
also contributed to improve final products significantly. the probe [28]. Extensive research has been published
Despite all of that, the occurrence of dimensional and presenting methodologies to determine systematic
geometrical deviations is inevitable. An appropriate errors related to the measuring probe in CMMs [15,
understanding and use of the existing standards related to 16, 24, 26, 27, 37].
GD&T (Geometric Dimensioning and Tolerancing) and of • Errors due to computational programs: computational
the ISO-GPS (Geometric Product Specification) standards methods to adjust the part geometry (least squares
both in engineering departments and in metrology rooms method, minimum zone criterion, etc.) and simplica-
are fundamental to guarantee that the deviations occurred tions used to reduce computational time are examples
during manufacturing do not jeopardize the proper of possible error sources associated with the computa-
assembly and functioning of the manufactured parts. tion of the measurements using CMMs [28, 31].

123
Author's personal copy
J Braz. Soc. Mech. Sci. Eng. (2015) 37:411–421 413

• Thermally induced errors: CMMs are sensitive to expression of uncertainty in measurement) requirements
variations in the environment conditions, in particular using a coordinate measuring machine (CMM). We expect
to temperature variations and gradients, which have the this work to help personnel involved in design, manufac-
strongest effect on accuracy and precision [9, 25]. turing and quality control to estimate measurement
• Errors related to the measurement strategy: generally, uncertainty using a simple and easy model, to adequate
the CMM operator chooses the measurement strategy such measurements to ISO/IEC 17025 (2005) [22]. We also
based on intuition and experience. According to expect to contribute to the improvement of scientific rigour
Weckenmann et al. [43], it is necessary to develop in research and therefore in publications in the area.
intelligent systems able to analyse the geometry of part
to be measured and then decide the algorithm, the
number of measurement points and the most adequate 2 Theoretical background
distribution of measurement points.
• Errors due to properties of the part to be measured: A significant amount of manufactured parts has cylindrical
physical and mechanical properties of the part can shape or at least some portions with a cylindrical section.
influence the measurement result. For example, a part Form tolerances (circularity and cylindricity) are fre-
can deform elastically or plastically due to the contact quently applied to such parts, which are particularly
of the probe tip, which will alter the coordinates of the important in designs that require high accuracy. The
measurement points. The part surface finish (surface development of instruments and procedures to verify if
roughness, burr formation, etc.) can also lead to errors these tolerances are in accordance with the design
during the interpretation of the measurement results requirements is of utmost importance to ensure the inter-
[28]. changeability and functionality of the manufactured parts.
• Vibrations: external vibrations can induce relative
movement between the measuring probe, the axis and 2.1 Circularity and cylindricity deviations
the part being measured, which can affect measurement
accuracy and precision. Manufacturers of CMMs Circularity deviation is graphically equivalent to the min-
provide values for the maximum vibration to be imum radial distance between two concentric circumfer-
allowed and suggest installation procedures to mini- ences within which the real profile of the part must be
mize vibration effects during measurements [25]. contained (NBR 6409, 1997) [2]. As shown in Fig. 1a, the
International standards present tests to assess and largest circle inscribed in the cross section of the part and
analyse vibration [3, 10]. the smallest circumscribed circle are considered to mini-
mize the distance between the circles.
For many years, time and effort has been dedicated to
The determination of circularity deviation is carried out
the mathematical modelling of CMM to compensate for
in a circular section of a cylindrical part. It must be
the measurement errors. Various techniques have been
emphasized that some measurement systems take into
used, such as vector analysis [46]; homogeneous trans-
account an infinite number of points of the analysed cross
formations [12, 14, 18, 19, 41] and statistical analysis [17,
section, while others can only make a discrete evaluation
29, 30, 40].
by considering a determined number of points over a given
However, independently of the mathematical tool used
cross section.
in the modelling, the resulting mathematical models will be
In many projects, only the application of circularity
complex due to the large number of factors that can con-
tolerances is not sufficient to guarantee a good performance
tribute to the uncertainty in CMMs [45]. The machine
versatility also contributes to the complexity of the model,
since it allows the measurement of several metrological
features of a workpiece [4]. These factors can increase the
difficulty to evaluate measurement uncertainty in CMMs,
as shown in [1, 24, 41, 42].
Alternatively, several authors have used of Monte Carlo
methods to calculate CMM measurement uncertainties;
these methods make use of Virtual CMMs that should
simulate realistic CMM behaviour [4, 32, 35].
This paper aims to present procedures to estimate cir-
cularity and cylindricity measurements according to ISO Fig. 1 Graphic representation of circularity (a) and cylindricity
TAG 4/WG [21], popularly known as GUM (Guide to the (b) deviations

123
Author's personal copy
414 J Braz. Soc. Mech. Sci. Eng. (2015) 37:411–421

of the manufactured parts. In such cases, cylindricity tol- The application of ISO TAG 4/WG [21] requires a
erances must also be used to limit the maximum deviations. mathematical model of the measurement process, e.g., the
Cylindricity deviation is defined as the radial difference output variable (measurand) must be expressed as a func-
between two coaxial cylinders between which the real tion of the input variables, as shown Eq. (1).
surface of the part must be contained (Fig. 1b). This dif- V ¼ f ðW1 ; W2 ; . . .; WN Þ ð1Þ
ference must be equal or less the specified cylindricity
tolerance. It is important to point out that cylindricity where V represents the output variable and W1, W2,, WN
deviation is a composed form deviation used to control are the input variables.
both circularity and straightness of the generatrix. The law of uncertainty propagation can be applied to
A classic measurement system to verify the form tol- mathematical model to determine the combined standard
erance of parts is composed of a dial gauge attached to a uncertainty related to the measurement, so that Eq. (2) is
tailstock device, which has a simple working principle and obtained.
is relatively cheap. However, the recent literature in the XN  
of 2 2
area shows that other measurement systems are more u2c ðvÞ ¼ u ðwi Þ
i¼1
owi
widely used in research related to form control of machined
parts, in particular coordinate measuring machines [5, 13, X
N 1 XN
of of  
þ2 uðwi Þ  u wj  rðwi ; wj Þ ð2Þ
34, 39]. i¼1 j¼iþ1
owi owj
Despite the different possible configurations of CMMs,
their working principle is similar. They work by digitally where v is the estimate of the output variable V, wi is the
storing the coordinates of the measurement points (X, Y and estimate of variable Wi, u2(wi) is the variance associated
Z). Computational programs use these coordinates to cal- with wi, for i varying from 1 to N, N is the number of
culate the desired feature (circle diameter, sphere diameter, variables that affect the measurement result V, u(wi) is the
distance, angle, form deviations, etc.). uncertainty associated with the error source represented by
These programs are based on analytical geometry and the estimate wi and r(wi, wj) is the correlation coefficient
vector analysis principles. They generally use least squares between the estimates wi and wj.
methods to adjust the part geometry. The second term in Eq. (2) expresses the correlation
The verification of circularity and cylindricity deviations between two uncertainty sources wi and wj, i = j.
is carried out in production lines during the form control of The International Committee for Weights and Measures
machined parts, as well as during research in machining. (abbreviated CIPM from the French Comité International
To follow the standards related to GPS, research has been des Poids et Mesures) proposes the description of mea-
conducted in the area. Sami et al. [33] have described a surement uncertainty using the intervals that represent the
method to analyse uncertainty in cylindricity measure- expected values for the measurement error, with a known
ments using a CMM. Zhao et al. [47] have proposed a probability. The CIPM uses the term expanded uncertainty
method to estimate uncertainty propagation during the (Up) to describe such interval, as shown in Eq. (3).
verification of cylindricity deviations according to GPS. Up ¼ k  uc ð3Þ
Sun et al. [38] have estimated uncertainty in the mea-
surement of form deviations in CMMs following the where uc is the combined standard uncertainty and k [ 0 is
requirements of the new GPS system. the coverage factor.
The coverage factor k relates to the probability distri-
2.2 Measurement uncertainty bution of the values obtained during measurement, which
normally presents a normal distribution. For situations
Measurement uncertainty is defined as a non-negative where this probability is not normally distributed, The
parameter that characterizes the dispersion of the values Central Limit Theorem together with the t Student distri-
attributed to a measurand based on the information used bution can provide a coverage factor based on the effective
[7]. degree of freedom of the standard measurement uncertainty.
To estimate the measurement uncertainty, the concepts The effective degree of freedom meff is obtained using the
and recommendations presented in GUM must be properly Welch–Satterthwaite formulation, as shown in Eq. (4).
known. This document, published first in 1993, is the result u4c ðyÞ
of an international consensus about how to calculate veff ¼ : ð4Þ
P u4i ðwi Þ
N
measurement uncertainty. The latest version for GUM is vi
i¼1
from 2008. This guide presents three types of uncertainties:
standard uncertainty, combined standard uncertainty, and The calculation of the uncertainty measurement requires
expanded uncertainty. the operator to be perfectly knowledgeable about the

123
Author's personal copy
J Braz. Soc. Mech. Sci. Eng. (2015) 37:411–421 415

measurement system and the measurement procedure. Only X   2


MQ ¼ a  xi þ b  yi þ x2i þ y2i þ c : ð10Þ
a qualified operator is able to: define properly the measu-
rand; choose the most adequate measurement system and The least squares coefficients are determined by equat-
procedure; define a correct measurement strategy; and ing the partial derivatives of MQ in relation to a, b and c to
define environment factors that influence the measurement zero, Eqs. (11, 12, 13).
result. oMQ X   
Mathematical models for CMMs are not simple, due to ¼2 a  xi þ b  yi þ x2i þ y2i þ c  ðxi Þ ¼ 0
oa
the large number of variables that influence the measure-
ð11Þ
ment result, as shown in the literature [1, 24, 41, 42].
oMQ X   
An important contribution was given by Weckenmann ¼2 a  xi þ b  yi þ x2i þ y2i þ c  ðyi Þ ¼ 0
et al. [44], who evaluated measurement uncertainty for ob
automotive parts. They considered influences associated ð12Þ
with the measurement system (CMM), the part, the oMQ X   
environment, and the operator. Their work concluded that, ¼2 a  xi þ b  yi þ x2i þ y2i þ c  1 ¼ 0:
oc
after compensating for thermal effects, the variables ð13Þ
associated with the measurement system and with the
operator had the strongest contribution to the final In the matrix form, we obtain the linear system (14).
2 P P P 3 2 3
uncertainty. x2i xi  yi xi a
According to Vieira Sato [42], the uncertainty associ- 6P P 2 P 7 6 7
4 x i  y i y y 
i5 4b5
ated with the adjustment method must also be considered P P i
2 xi P 3 yi n3 c
to estimate the measurement uncertainty. For that, the law
of propagation of uncertainty must be applied in the  ðxi þ xi  y2i Þ
6 P 7
mathematical model until the calculation of the final ¼ 4  ðx2i :yi þ y3i Þ 5: ð14Þ
P 2 2
characteristic. Using this approach, the author proposed a  ðxi þ yi Þ
methodology to estimate the uncertainty associated to
In sequence, the distance (Di ) of each point Pi to the
circularity deviations using a moving bridge manual
centre of the circle (Pc ) is calculated, identifying the most
CMM.
distant point (Pmax ) and least distant point (Pmin ) in relation
This model starts with the definition of the number of
to Pc (Eq. 15).
measuring points on the part surface. During the measure- qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ment, a computer software stores the coordinates X, Y and DPi ¼ DðPc ; Pi Þ ¼ ðxi  xc Þ2 þðyi  yc Þ2 þðzi  zc Þ2
Z of each point, projecting them in the so-called projection
plane. In sequence, the radius (r) and the coordinates of the ð15Þ
centre (xc, yc) of the circle that adjusts best to the n points where (xc, yc, zc) and (xi, yi, zi) are the coordinates of the
projected onto the plane XY are determined (Eq. 5). points Pc and Pi respectively.
ðx  xc Þ2 þðy  yc Þ2 ¼ r 2 : ð5Þ The circularity deviation is given by the difference
between the maximum and minimum distances, as shown
To estimate the radius and coordinates of the centre of by Eq. (16).
this circle, the expression given by Eq. (6) must be
DCIR ¼ DPmax  DPmin : ð16Þ
minimized.
X X 2 This process must be repeated at least three times and
e2i ¼ ri2  r 2
X    2 the results must be averaged by calculating the arithmetic
¼ 2  xi  xc  2  yi  yc þ x2i þ y2i þ x2c þ y2c  r 2 : mean. Measurement repetition is necessary to obtain a
ð6Þ representative sample, to allow statistical treatment of the
data, detection of possible gross errors, and evaluation of
Rearranging variables as shown in Eq. (7, 8, 9), the the uncertainty measurement.
expression given by Eq. (6) can be rewritten in a linear The deviation measurement uncertainty (DCIR) is influ-
form [Eq. (10)]. enced by the uncertainty of the coordinates of the points
a ¼ 2  xc ð7Þ that generate the circle and by the uncertainty of the
coordinates of the points Pmax and Pmin . Therefore, the law
b ¼ 2  yc ð8Þ
of propagation of uncertainty must be applied to express
c ¼ x2c þ y2c  r 2 ð9Þ the combined standard uncertainty as:

123
Author's personal copy
416 J Braz. Soc. Mech. Sci. Eng. (2015) 37:411–421

N 
X 
  X N  
oDr 2 oDr 2 2   Substituting (20) in (8) and (9) and applying the law of
u2c ðDCIR Þ ¼ 2
 u xj þ  u yj
j¼1
oxj j¼1
oyj propagation of uncertainty, Eqs. (22) and (23) are obtained.
 2 "  #
    oDCIR oDCIR 2 2
oDPmax 2 2 oDPmax 2 2 u2C ðDCIR Þ ¼ u2 ðDsðDCIR ÞÞ þ 2  u ðDRÞ
þ u ðxPmax Þ þ u ðyPmax Þ oDsðDCIR Þ oDR
oxPmax oyPmax " 2 # "  #
    oDCIR oDCIR 2 2
oDPmax 2 2 oDPmin 2 2 þ2  u2 ðDEAÞ þ 2  u ðDI Þ
þ u ðzPmax Þ þ u ðxPmin Þ oDEA oDI
ozPmax oxPmin " 2 # "  #
oDCIR oDCIR 2 2
    þ2  2
u ðDT20 Þ þ 2  u ðDdT Þ
oDPmin 2 2 oDPmin 2 2 oDT20 oDdT
þ u ðyPmin Þ þ u ðzPmin Þ:
oyPmin ozPmin ð22Þ
ð17Þ  2 "  #
oDCYL oDCYL 2 2
u2C ðDCYL Þ ¼ u2 ðDsðDCYL ÞÞ þ 2  u ðDRÞ
In sequence, all the partial derivatives in Eq. (17) must oDsðDCYL Þ oDR
" 2 # "  #
be calculated. The whole development can be found in oDCYL oDCYL 2 2
þ2  u2 ðDEAÞ þ 2  u ðDI Þ
Vieira Sato [42]. The calculations are complex and require oDEA oDI
the use of applicative programming languages such as " 2 # "  #
oDCYL oDCYL 2 2
Mathematica or Matlab. Therefore, the complexity of this þ2  u2 ðDT20 Þ þ 2  u ðDdT Þ :
oDT20 oDdT
methodology makes it difficult to be adopted by the
majority of the community dedicated to the geometric ð23Þ
control of machined parts. The standard uncertainty related to s(DCIR) and s(DCYL)
The present paper proposes alternative mathematical can be calculated as shown in Eq. (24).
models that are easy to understand and simple to be s
sðDCIR Þ ¼ sðDCYL Þ ¼ pffiffiffi ð24Þ
implemented, to popularize the uncertainty calculation n
during circularity and cylindricity deviation measurements.
This new formulation is presented in Eqs. (18) and (19). where s is the standard deviation of the deviation readings
and n is the total number of measurement cycles.
DCIR ¼ DsðDCIR Þ þ DPmax þ DPmin ð18Þ The correction due to the CMM resolution is given by
DCYL ¼ DsðDCYL Þ þ DPmax þ DPmin : ð19Þ Eq. (25).
These equations take into account the variability of the R
uðDRÞ ¼ pffiffiffi : ð25Þ
circularity deviations (Ds(DCIR)) or cylindricity deviations 2 3
(Ds(DCYL)), considering the n measurement cycles and the The correction due to errors of the CMM probing system
corrections associated with the furthest point in relation to is given by Eq. (26).
the centre (Pmax ) and the point least distant from the centre
EA
(Pmin ). uðDEAÞ ¼ pffiffiffi : ð26Þ
On the other hand, the uncertainty of those two points 6
is a function of: corrections associated to the CMM The correction due to the uncertainty associated with the
resolution (DR); errors of the CMM probing system CMM indication system is given by Eq. (27).
(DEA); uncertainty associated with the CMM indication sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
    ffi
system (DI); corrections associated with the difference DIðxÞ 2 DIðyÞ 2 DIðzÞ 2
uðDIÞ ¼ þ þ ð27Þ
between coefficients of thermal expansion of the scale kðxÞ kðyÞ kðzÞ
and workpiece (Da); corrections due to the distancing of
The correction due to difference between coefficients of
the temperature in relation to 20 C (DT20); and cor-
thermal expansion of the scale and workpiece is given by
rections due to temperature variation during measurement
Eq. (28).
(DdT).
Where aE is the coefficient of thermal expansion of the
DPmax ¼ DPmin ¼ DR þ DEA þ DI þ LDaDT20 þ LDadT scale and ap is the coefficient of thermal expansion of the
ð20Þ workpiece.
L is the CMM indication and can be calculated by 0:01ðaP  aE Þ
uðDaÞ ¼ pffiffiffi : ð28Þ
Eq. (21), where Xi, Yi and Zi are the coordinates of the 3
point, Pmax and Pmin, respectively.
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi Both variables related to room temperature variation
L ¼ Xi2 þ Yi2 þ Zi2 ð21Þ were measured using the same measurement system.
Therefore, they were treated as correlated.

123
Author's personal copy
J Braz. Soc. Mech. Sci. Eng. (2015) 37:411–421 417

The correction due to the distancing of the temperature


in relation to 20 C (DT20) is determined using Eq. (29).
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 2    2
DT DRT 2 DIT
uðDT20 Þ ¼ pffiffiffi þ pffiffiffi þ ð29Þ
3 2 3 kT

where DT is the difference between the room temperature


and 20 C; DRT is the correction in relation to the ther-
mometer resolution and DIT is the uncertainty associated
with the thermometer indication system.
The uncertainty due to temperature variation during
measurement is given by Eq. (30).
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
     2ffi
VarðTÞ 2 DRT 2 DIT
uðDdTÞ ¼ pffiffiffi þ pffiffiffi þ : ð30Þ
3 2 3 kT

3 Methodology

Circularity and cylindricity deviations were measured for


four geometries, using a CMM, type moving bridge, man-
ufactured by Mitutoyo. The equipment, shown in Fig. 2, has Fig. 2 Coordinate measurement machine used in the experiments
a resolution of 0.001 mm and a work volume of 400 mm
(Axis X) 9 400 mm (Axis Y) 9 300 mm (Axis Z). Table 2 shows circularity and cylindricity values for all
The measurements were carried out at a controlled room the features measured in the tests.
temperature of (20 ± 1) C. A thermo-hygrometer with a The values for the combined standard uncertainty (uc)
digital increment of 0.1 C and measurement range of -20 and for the expanded standard uncertainty (Up) associated
to 60 C was used to monitor the temperature. with the measurements of the circularity and cylindricity
All the instruments and parts used in the measurement for the four measurands are shown in Figs. 3 and 4. These
tests were exposed to this temperature for approximately values were obtained using the new formulation proposed
12 h before the measurements. To remove dust or other dirt in this paper. It should be pointed out that two significant
particles that could interfere with the measurement results, digits were added during the calculations to minimize
all the instruments and parts were cleaned using isopropyl rounding errors.
alcohol, gloves, cotton buds and dry cloths. The temperature variation during the measurements was
The number of points used in the circularity measure- not considered, because they were carried out during a
ments varied according to the diameter of the feature short time interval. During the evaluation of the measure-
measured, or measurand. Number of points of 5, 7, 11 and ment uncertainty, thermal effects were not considered
13 were used for the different diameters, probed on a single because the measurement temperature was 20.3 C with a
cross section of the feature evaluated. For the cylindricity variation of 0.2 C.
measurements, the number of points was given by 2n ? 1, Table 3 shows circularity and cylindricity values for all
where n represents the number of points probed in the the features measured in the tests.
circularity measurements. The probing points were dis- According to Fig. 4, the combined uncertainty for the
tributed in two parallel planes along the length of the circularity deviation measurements varied between 2.3 and
feature. All measurements were repeated five times. 3.1 lm. The values for the expanded uncertainty varied
According to the calibration certificate, the CMM used between 4.5 and 6.3 lm, for k = 2.4 and 95.45 % of
in the measurements has linear expanded uncertainties of coverage probability.
(1.2 ? L/1,300) lm and k = 2.2 for the X axis; of Figure 5 shows that the combined uncertainty for the
(1.2 ? L/1,300) lm for the Y axis; and of (1.0 ? L/1,300) cylindricity deviation measurements varied between 2.3
for the Z axis. The probing error is 2.9 lm. and 3.0 lm. The values for the expanded uncertainty var-
ied between 4.6 and 6.6 lm, for a coverage factor k = 2.2
and 95.45 % of coverage probability.
4 Results and discussion In all cases, the largest contribution for the final
uncertainty was given by the CMM probing error and by
The mean values of the diameters measured in the exper- the variability of the values of circularity and cylindricity
iments are shown in Table 1. deviations for each measurand.

123
Author's personal copy
418 J Braz. Soc. Mech. Sci. Eng. (2015) 37:411–421

Table 1 Diameters of the features measured in the tests Circularity deviation uncertainty
Measurand Diameter (mm) 8 uc Up

Mean Standard deviation

Uncertainty (µm)
6
Feature 1 8.008 0.004
Feature 2 19.936 0.016 4
Feature 3 47.953 0.013
Feature 4 67.957 0.005 2

0
1 2 3 4
Table 2 Circularity and cylindricity deviations Feature
Measurand Circularity deviation (lm) Cylindricity deviation
Fig. 4 Uncertainties for the circularity deviation measurements
(lm)
Mean Standard deviation Mean Standard deviation

Feature 1 10 2 11 2 Table 3 Uncertainty of circularity and cylindricity deviations


Feature 2 57 4 61 5 Measurand Circularity deviation Cylindricity deviation
Feature 3 95 5 102 5
uc (lm) Up (lm) uc (lm) Up (lm)
Feature 4 92 5 105 3
Feature 1 2.3 4.6 2.3 4.5
Feature 2 2.9 6.1 3.0 6.6
Feature 3 3.1 6.8 2.7 6.4
Feature 4 3.0 6.5 2.6 5.4

methodology proposed Vieira Sato [42]. The input data in


both cases were the same, to identify differences between
uncertainty values provided by both models, i.e., the
measured geometries, the operator, the measurement
strategy and the environment conditions were identical. It
is relevant to point out that the methodology proposed by
Vieira Sato [42] is universal and can be applied to any
CMM. Therefore, it was implemented using Matlab.
According to Instone [20], any of the recognized
methods to combine uncertainty contributors gives results
that are approximately 10 % similar between themselves.
Their main weakness is the large number of measurements,
which results in elevated measurement costs to be trans-
ferred to the final customer. If the instrument is calibrated
Fig. 3 Workpiece and features measured
using a number of points significantly lower, the risk to
ascertain a measurement increases.
The simplified models here proposed, implemented Figures 6 and 7 show the expanded uncertainty values
using an Excel spreadsheet, proved to be viable and easy to using both methods, for circularity and cylindricity devia-
apply to evaluate the uncertainty measurements associated tions, respectively. The values obtained with the method
to both geometric deviations here investigated. here proposed were called Up 2011 and the values obtained
with the method proposed by Vieira Sato [42] were called
Up 2003.
5 Validation of the proposed models Table 4 shows circularity and cylindricity values for all
the features measured in the tests.
In sequence, the models here proposed were validated, by The values of expanded uncertainty shown in Figs. 5
comparing the results obtained with those given by the and 6 obtained with the formulation here proposed are

123
Author's personal copy
J Braz. Soc. Mech. Sci. Eng. (2015) 37:411–421 419

Cylindricity deviation uncertainty Circularity uncertainty - Comparison


8 uc Up 8 Up 2011 Up 2003
Uncertainty (µm)

Uncertainty (µm)
6

4
4
2
2
0
1 2 3 4
Feature 0
1 2 3 4
Fig. 5 Uncertainties for the cylindricity deviation Feature

Fig. 6 Comparison between the values obtained by both methods for


circularity deviation
similar to those obtained by the methodology proposed by
Vieira Sato [42]. The largest difference between the
expanded uncertainty values for the circularity deviation
was obtained for the measurement of the second measu- Cylindricity uncertainty - Comparison
rand. This value was 0.9 lm, which represents a difference 8 Up 2011 Up 2003
of approximately 14.4 %. In all the other cases the differ-
ences were lower than 5 %. 6
Incerteza (µm)

For the cylindricity deviation, the largest difference


between the values obtained by the two models was
4
0.2 lm, which represents approximately 4.5 %.
Although the new formulation does not take into
2
account the correction due to the geometry adjustment
method, the results obtained are considered adequate. The
0
differences observed are small and can be attributed to the
1 2 3 4
sensibility coefficient values. In the new formulation, these Feature
coefficients are unitary for the first four variables in Eqs.
(18) and (19), and only for two last variables they assume Fig. 7 Comparison between the values obtained by both methods for
cylindricity deviation
non-unitary values. However, in such cases, the standard
uncertainty values are very small and were not considered.
On the other hand, in Vieira Sato’s model, the sensibility
coefficients result from the solution of complex partial Table 4 Expanded uncertainty obtained by both methods for circu-
derivatives and are non-unitary for all variables in most larity and cylindricity deviations
cases. Therefore, the weights for the standard uncertainties Measurand Circularity deviation Cylindricity deviation
are different in the two models.
Up 2003 Up 2011 Up 2003 Up 2011
Figure 5 shows that the expanded uncertainty values (lm) (lm) (lm) (lm)
obtained are in fact larger when calculated using the new
formulation (features 1, 2 and 4) than those obtained using Feature 1 4.4 4.6 4.7 4.5
the model by Vieira Sato [42]. This indicates that the Feature 2 5.5 6.1 6.6 6.6
adjustment method is not very significant and therefore can Feature 3 6.8 6.8 6.3 6.4
be neglected. Feature 4 6.4 6.5 5.4 5.4

6 Conclusions In all cases, the CMM probing error and the reading
variability had the strongest influence on the final uncer-
The simplified models proposed in this paper have proved tainty values.
to be viable and of easy application to evaluate circularity The expanded uncertainty values were between 4.5 and
and cylindricity deviation uncertainty measurement. 6.3 lm for circularity deviation, whereas for cylindricity

123
Author's personal copy
420 J Braz. Soc. Mech. Sci. Eng. (2015) 37:411–421

deviation they varied between 4.6 and 6.6 lm. In both 13. Cavalcante FJN (2010) Análise de Utilização de Broca Escalo-
cases the coverage probability was 95.45 %. nada de Canal Reto no Processo de Furação em Liga de Alumı́nio
Utilizada na Indústria Automotiva. Universidade Tecnológica
Comparing the expanded uncertainty results obtained by Federal do Paraná, Curitiba
the two models, the largest difference was observed for the 14. Di Giacomo B, Orrego RMM, Viera Sato DP (1999) An
second measurand. The difference was 0.4 lm, which exploratory study about the second order errors in mathematical
represents 14.4 %. The difference was lower than 5 % in models of coordinate measuring machines. Mfg the Brown
Sharpe Publication of Precision Manufacturing, EUA, 6(1): 26
all other cases. Despite this, the model presented an 15. Estler WT, Phillips SD, Borchardt B, Hopp T, Levenson M,
uncertainty larger than the obtained with that proposed by Eberhardt K, McClain M, Shen Y, Zhang X (1997) Practical
Vieira Sato [42], showing that the uncertainty associated Aspects of touch-trigger probe error compensation. Precis. Eng.
with the adjustment method can be neglected. 21:1–17
16. Estler WT, Phillips SD, Borchardt B, Hopp T, Witzgall C, Le-
The uncertainty component related to the adjustment venson M, Eberhardt K, McClain M, Shen Y, Zhang X (1996)
method does not represent a significant contribution to Error compensation for CMM touch-trigger probes. Precis. Eng.
the total uncertainty measurement and therefore can be 19:85–97
neglected. 17. Guye JJ (1978) Metrological inspection of machining centers, jig
boring machines and measuring machines through a statistical
The comparison of the values obtained with mathe- approach. Sociéte Genovice d’Instrumets de Physique, Geneva,
matical models proposed by Vieira Sato proves the model pp 141–145
proposed in this paper to be applicable, in particular if the 18. Hermann G (2007) Geometric error correction in coordinate
calculation simplicity and the easy interpretation of the measurement. Acta Polytechnica Hungarica 4(1):47–62
19. Hocken RJ, Simpson JA, Borchardt B, Lazar J, Reeve C, Stein P
influencing variables are taken into account. (1977) Three dimensional metrology. Annals CIRP 26(2):403–408
20. Instone I (1996) Simplified method for assessing uncertainties in
Acknowledgments The authors are grateful to FAPEMIG/Brazil a commercial production environment. Teste and measurement.
for financial support. Metrology Forum. London, UK, 10 Oct. pp 1–7
21. ISO TAG 4/WG 3 (2008) Guide to the expression of uncertainty
in measurement. Geneva Switzerland
22. ISO/IEC 17025 (2005) General requirements for the competence
References of testing and calibration laboratories
23. Kunzmann H, Wäldele F, Ni J (1995) Accuracy enhancement. In:
1. Abackerli AJ, Pereira PH, Calônego N Jr (2010) A case study on Bosch JA (ed) Coordinate measuring machines and systems.
testing CMM uncertainty simulation software (VCMM). J Braz Marcel Dekker Inc, New York
Soc Mech Sci Eng 32(1):8–14 24. Miguel PA, Cauchick King T, Abackerli AJ (2003) CMM touch
2. ABNT (1997) NBR 6409-Tolerâncias Geométricas—Tolerâncias de trigger performance verification using a probe test apparatus.
Forma, Orientação, Posição e Batimento—Generalidades, Sı́mbo- J Braz Soc Mech Sci Eng 25(2):147–153
los, Definições e Indicações em Desenho. Rio de Janeiro, Brazil 25. Ni J (1995) Environmental control. In: Bosch JA (ed) Coordinate
3. ASME B89.4.1 (1997) Methods of performance evaluation of measuring machines and systems. Marcel Dekker, New York
coordinate measuring machines 26. Pahk H, Kim J, Lee K (1996) Integrated compensation system for
4. Balsamo A, Di Ciommo M, Mugno R, Rebaglia BI, Ricci E, errors introduced by measurement probe and machine geometry
Grella R (1999) Evaluation of CMM uncertainty through Monte in commercial CMMs. Int J Mach Tools Manuf 36(9):1045–1058
Carlo simulations. Annals CIPR 48(1):425–428 27. Peggs GN (1990) Traceability for coordinate measurement
5. Barbosa LM (2007) Monitoramento do Comportamento da Broca technology, 27th International MATADOR Conference, Man-
Chamdrill na Usinagem do Ferro Fundido Cinzento. In: Pro- chester, pp 463–468
ceedings of the 48 Congresso Brasileiro de Engenharia de Fa- 28. Phillips SD (1995) Performance evaluations. In: Bosch JA (ed)
bricação, Estância de S. Pedro, Brazil Coordinate measuring machines and systems. Marcel Dekker,
6. Belforte G, Bona B, Canuto E, Donati F, Ferraris F, Gorini IL, New York
Morei S, Peisino M, Sartori S (1987) Coordinate measuring 29. Piratelli-Filho A (1997) Método para avaliação do desempenho
machines and machine tools selfcalibration and error correction. de máquinas de medir a três coordenadas através do planejamento
Annals of the CIRP, Vol. 36 de experimentos. Tese (Doutorado)—Escola de Engenharia de
7. BIPM, IEC, IFCC, ILAC, ISO, IUPAC, IUPAP, OIML (2008) São Carlos, Universidade de São Paulo, São Carlos
International vocabulary of metrology—basic and general con- 30. Poole AB (1983) The calibration of coordinate measuring
cepts and associated terms (VIM) machines by statistical method. Qual Assur 9(2):71–97
8. Bosch JA (1995) Coordinate measuring machines and systems. 31. Porta C, Wäldele F (1986) Testing of three coordinate measuring
Marcel Dekker, Inc, New York, p 444 machine evaluation algorithms. Report EUR 10909 EN, Com-
9. Bryan JB (1995) Temperature fundamentals. In: Bosch JA (ed) mission of the European Communities
Coordinate measuring machines and systems. Marcel Dekker Inc, 32. Ramu P, Yague J, Hocken R, Miller J (2011) Development of a
New York parametric model and virtual machine to estimate task specific
10. BS 6808 (1989) Part 1–3: British Standard—Coordinate mea- measurement uncertainty for a five axis multi sensor coordinate
suring machines measuring machine. Precis Eng 35:431–439
11. Burdekin MS, Voutsadopoulos C (1981) Computer aided cali- 33. Sami A, Farooqui TD, Chittaranjan S (2009) Uncertainty analysis
bration of the geometric errors of multi-axis coordinate measur- of cylindricity measurements using bootstrap method. Measure-
ing machines. Proc Instn Mech Eng 195:231–239 ment 42(4):524–531
12. Cardoza JAS (1995) Máquinas virtuais de medir a três coorde- 34. Santos RG (2004) Avaliação do processo de alargamento de guias
nadas. Tese (Doutorado), EESC-USP de válvulas. Universidade Federal do Paraná, Curitiba

123
Author's personal copy
J Braz. Soc. Mech. Sci. Eng. (2015) 37:411–421 421

35. Schwenke H, Siebert B, Wäldele F, Kunzmann H (2000) Dissertação (Mestrado)Escola de Engenharia de São Carlos.
Assessment of uncertainties in dimensional metrology by Monte Universidade de São Paulo, São Carlos
Carlo simulation: proposal of a modular and visual software. 41. Valdés RA, Di Giacomo B, Paziani FT (2005) Synthesization of
CIRP Annals Manuf Technol 49(1):395–398 thermally induced errors in coordinate measuring machines.
36. Sousa AR, Wandek M (2009) Deficiências da metrologia indus- J Braz Soc Mech Sci Eng 27(2):170–177
trial no Brasil no correto entendimento do GD&T e na definição 42. Vieira Sato DP (2003) Determinação da Incerteza de Medição a
de estratégias de medição consistentes para o seu controle ge- Três Coordenadas. Universidade de São Paulo, S. Carlos, Brazil
ométrico, V Congresso Brasileiro de Metrologia. Salvador, Ba- 43. Weckenmann A, Eitzert H, Garmer M, Weber H (1995) Func-
hia—Brasil tionality-oriented evaluation and sampling strategy in coordinate
37. Stone J, Muralikrishnan B (2011) Geometric effects when mea- metrology. Precis Eng 17(4):244–252
suring small holes with micro contact probes. J Res Nat Inst 44. Weckenmann A, Knauer M, Killmaier T (2001) Uncertainty of
Stand Technol 16(2):573–587 coordinate measurements on sheet-metal parts in the automotive
38. Sun Y, Xu W, Zhu L, Du H, Zhang Y (2009) Estimation of industry. J Mater Process Technol Amsterdam 115:9–13
uncertainty in form error Cmm measurement according to new 45. Wilhelm R, Hocken R, Schwenke H (2001) Task specific
GPS standard system. In: Proceedings of the International Tech- uncertainty in coordinate measurement. CIRP Annals Manuf
nology and Innovation Conference 2009, Xian, China, pp 1–4 Technol 50(2):553–563
39. Tedesco ME, Zeilmann RP, Tentardini EK, Baumvol IJR (2006) 46. Zhang G, Veale R, Charlton T, Borchard B, Hocken R (1985)
Estudo sobre qualidade superficial e desvio dimensional do Aço Error compensation of coordinate measuring machines. Annals
AISI 4140 Submetido ao Processo de Torneamento. Proceedings CIRP 34(1):445–448
of the 178 Congresso Brasileiro de Engenharia e Ciência dos 47. Zhao F, Zhang L, Zheng P (2010) Estimation of the uncertainty
Materiais, Foz do Iguaçu, Brazil propagation in verification operator of cylindricity errors. Pro-
40. Valdés RA (1999) Equacionamento das Componentes do Erro ceedings of the 6th International Symposium on Precision Engi-
Volumétrico em Máquinas de Medir a Três Coordenadas. neering Measurements and Instrumentation, Hangzhou, China

123

View publication stats

You might also like