Journal of Physics: Conference Series

PAPER • OPEN ACCESS

Recent citations
Optimization of Control Points Number at - Inverse Kinematic Problem of Movement
Coordinate Measurements based on the Monte- of Six Degrees of Freedom Robotic Arm
(Solved by a Numerical Method)
A V Kochetkov et al
Carlo Method - Problems and ways to reduce
measurement uncertainties in evaluating
To cite this article: A A Korolev et al 2018 J. Phys.: Conf. Ser. 944 012061 the geometric assembly parameters of gas
turbine engine assemblies
I A Grachev et al

View the article online for updates and enhancements.

This content was downloaded from IP address 109.175.99.249 on 03/12/2020 at 14:48

AMSD IOP Publishing
IOP Conf. Series: Journal of Physics: Conf. Series 944 (2018)
1234567890 ‘’“”012061 doi:10.1088/1742-6596/944/1/012061

Optimization of Control Points Number at Coordinate Meas-

urements based on the Monte-Carlo Method

A A Korolev1, A V Kochetkov2, and O V Zakharov1,3,4

¹Yuri Gagarin State Technical University, Saratov, Russia
2
Perm National Research Polytechnic University, Perm, Russia
3
Samara National Research University, Samara, Russia
4
Institute of Precision Mechanics and Control of RAS, Saratov, Russia

E-mail: [email protected]

Abstract. Improving the quality of products causes an increase in the requirements for the ac-
curacy of the dimensions and shape of the surfaces of the workpieces. This, in turn, raises the
requirements for accuracy and productivity of measuring of the workpieces. The use of coordi-
nate measuring machines is currently the most effective measuring tool for solving similar
problems. The article proposes a method for optimizing the number of control points using
Monte Carlo simulation. Based on the measurement of a small sample from batches of work-
pieces, statistical modeling is performed, which allows one to obtain interval estimates of the
measurement error. This approach is demonstrated by examples of applications for flatness, cy-
lindricity and sphericity. Four options of uniform and uneven arrangement of control points are
considered and their comparison is given. It is revealed that when the number of control points
decreases, the arithmetic mean decreases, the standard deviation of the measurement error in-
creases and the probability of the measurement -error increases. In general, it has been estab-
lished that it is possible to repeatedly reduce the number of control points while maintaining
the required measurement accuracy.

1. Introduction
Coordinate measurement traditionally involves measuring individual points on the workpiece surfaces,
replacing the cloud of control points with the nominal geometry of the workpiece, and comparing with
the tolerance for the corresponding dimensions. As a result, a decision is made to accept or reject the
defective workpiece. Evaluation of the error in the form of the surface consists of several components:
the form error after machining, the systematic and random measurement errors. The systematic meas-
urement error is related to the error of basing and the sampling error (the number of control points).
The random measurement error primarily depends on the errors of the touch sensor of the coordinate
measuring machine.
Obviously, increasing the accuracy requires the use of a large number of control points on the sur-
faces. However, the measurement productivity is significantly reduced. Therefore, the question arises
of the optimal choice of the number of control points. The idea of the need to reduce sample sizes in
the measurement and the possibility of a priori estimating the possible error or uncertainty of meas-
urement was repeatedly mentioned in a number of papers [1–12]. The information given in the studies
is of unquestionable scientific and practical value, but only concern certain surfaces or aspects of the
problem. Therefore, in practice this issue is solved directly by the CMM operator and the result largely
depends on its qualification. In this paper, we propose a new optimization method for determining the

Content from this work may be used under the terms of the Creative Commons Attribution 3.0 licence. Any further distribution
of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI.
Published under licence by IOP Publishing Ltd 1
AMSD IOP Publishing
IOP Conf. Series: Journal of Physics: Conf. Series 944 (2018)
1234567890 ‘’“”012061 doi:10.1088/1742-6596/944/1/012061

number and location of control points on surfaces, which is based on statistical simulations Monte
Carlo.

2. The formulation of the problem

The initial problem is formulated as follows – to measure the given surface with a certain error and
maximum productivity. The maximum productivity is achieved, firstly, by minimizing the number of
control points, and secondly, the location of control points, which ensures their best sequencing. As an
estimate of measurement error can be used absolute or relative error values, the rejection value for a
given tolerance, the probabilities of measuring I and II errors.
The new strategy is based on computer modeling of measurement errors for batches of workpieces
using the Monte Carlo method. The fundamentals of the application of the Monte Carlo method for
various applied problems are presented in [13]. The advantage of using the Monte Carlo method is to
obtain non-point, but interval error estimates for certain batches of workpieces.
Simulation is carried out in the following sequence. At the first stage, sequences of pseudo-random
numbers with given correlation and probability distribution laws modeling the random values of error
components at each test are generated. With a stable technological process, the error distribution law
does not change, and its parameters change insignificantly. Therefore, it becomes possible to deter-
mine the law and the error distribution parameters of various surfaces on the basis of a small sample
from a batch of workpieces. At the second stage, multiple calculations are performed for surfaces with
simulated deviations of the form, determining the values of the size, form or location errors. The num-
ber of iterations in modeling can be related to the number of workpieces in a batch or taken large for
statistical stability. At the third stage, statistical processing of the simulation results is performed - es-
tablish the law and determine the distribution parameters or calculate the statistical moments. Thus, an
interval estimation of the measurement error is obtained to compare the variants with different num-
bers and locations of control points on the surfaces.
The statistical modeling technique was previously described in detail in our paper [14].

3. Mathematical Modeling
As objects of research three types of surfaces are chosen: flatness, cylinder, sphere. It is these surfaces
that make up the overwhelming majority on various workpieces of machines and mechanisms. The
surfaces are described in a parametric form, where the simulated errors 1, 2, 3 are additionally intro-
duced into the equations.
The equation of the flatness with errors is:
X 1  xi ; 
 (1)
Y1  yi ; 
Z1  c  1;
where xi = [x1; x2], yi = [y1; y2] is the linear parameters that determine the dimensions of the workpiece;
c is the position of the plane.
The equation of the cylinder with errors is:
X 2  ( r2   2 ) cos ;

Y2  ( r2   2 ) sin  ;  (2)
Z 2  zi ; 

where zi = [z1; z2] is length of the cylinder;  = [0; 2] is angular parameter; r2 is radius of the cylinder.
The equation of the sphere with errors is:
X 3  ( r3   3 ) cos  cos  ;
 (3)
Y3  ( r3   3 ) cos  sin  ; 
Z 3  ( r3   3 ) sin ; 

where  = [0; 2],  = [0; ] are angular parameters; r3 is radius of the sphere.
The initial data for the simulation was the results of measuring the workpieces on the mobile CMM
Faro Arm Edge (FARO Swiss Holding GmbH, Switzerland). The measurements were carried out with
the following parameters: flatness of size 100100 mm, a cylinder 50 mm in diameter and 100 mm in

2
AMSD IOP Publishing
IOP Conf. Series: Journal of Physics: Conf. Series 944 (2018)
1234567890 ‘’“”012061 doi:10.1088/1742-6596/944/1/012061

length, a hemisphere of radius 60 mm. Was used uniform grid for linear and angular coordinates for
control points: 121 for the plane, 110 for the cylinder, and 101 for the sphere. For this, three-
dimensional models of workpieces with measurable surfaces with marking of control points were cre-
ated in the program Power Inspect.
The measurement was made at given points 30 times. The re-basing procedure was carried out 3
times. For the flatness, the array of measured points of all the experiments is shown in Fig. 1. The de-
viation of the shape at the point of the flatness is the distance between the nominal coordinates of the
points and the measured coordinates taken with the sign taken into account. The deviation of the shape
for the flatness as a whole is the difference between the maximum and minimum deviation of the
shape at the points of the flatness. In the case of ideal basing, the deviations of the shape are the coor-
dinates along the Z-axis (this axis is the normal to the flatness). In Fig. 2 shows the calculated devia-
tions in 110 points of 30 experiments after elimination of the error of basing.

Figure 1. An array of measured points for the flatness.

Figure 2. The errors of the flatness after elimination of

the error of basing.

On the basis of the data obtained, Monte Carlo simulations were carried out for batches of 100
workpieces with uniform and normal distribution of error components. Then, according to the standard
method, the error values are calculated: flatness, cylindricity, sphericity. Modeling and processing of
results was carried out according to the developed program in the Matlab environment.
In the simulation, four options of the location of the control points are considered:
 option 1 – uniform grid of linear and angular parameters of the surface, the number of control
points 121 for the plane, 110 for the cylinder and 101 for the sphere;
 option 2 – uneven mesh, sparse for the plane along the X axis, for the cylinder along the
length, for the sphere along the zenith angle, the number of control points 66 for the plane, 60
for the cylinder, 61 for the sphere;

3
AMSD IOP Publishing
IOP Conf. Series: Journal of Physics: Conf. Series 944 (2018)
1234567890 ‘’“”012061 doi:10.1088/1742-6596/944/1/012061

 option 3 – non-uniform grid, sparse for the plane along the Y axis, for the cylinder in the angle
of the circle, for the sphere in the azimuth angle, the number of control points 66 for the plane,
55 for the cylinder, 51 for the sphere;
 option 4 – uniform grid, sparse in angular and linear coordinates, the number of control points
36 for the plane, 30 for the cylinder, 31 for the sphere.
Option 1 was considered to be a basic one for modeling. All other variants were obtained from the
basic by eliminating a number of control points according to the above algorithm. The results of one of
the simulation implementations are shown in Fig. 3–5. Here, for a cylinder and a hemisphere, an error
scale of 500 times is used.

a b

c d

Figure 3. Simulation of the flatness error: a – option 1, b – option 2, c – option 3,

d – option 4.

a b c d

Figure 4. Simulation of the cylinder error: a – option 1, b – option 2, c – option

3, d – option 4.

4
AMSD IOP Publishing
IOP Conf. Series: Journal of Physics: Conf. Series 944 (2018)
1234567890 ‘’“”012061 doi:10.1088/1742-6596/944/1/012061

a b

c d

Figure 5. Simulation of the sphere error: a – option 1, b – option 2, c – option 3,

d – option 4.

4. Analysis of measurement simulation results

Statistical processing of the data resulting from the simulation showed that the measurement error is
well described by the normal distribution law. The evaluation was carried out by Pearson's agreement
criterion with a confidence probability of 95%. As an example, Fig. 6 gives empirical distributions of
the measurement error for flatness (a) and cylindricity (b).

a b

Figure 6. Histogram of the measurement error: a – flatness, b – cylindricity.

Therefore, the comparison of options was carried out on the arithmetic mean value and the standard
deviation of the measurement error. Also, the tolerances for the corresponding deviations of the form
were conventionally given and calculated the values of the defective workpieces in percent and meas-
uring I and II errors. The results of the calculations are given in Fig. 7 for the flatness, Fig. 8 – for the
cylinder, Fig. 9 for the sphere. Shown: I – the arithmetic mean, μm, II – the standard deviation, μm.

5
AMSD IOP Publishing
IOP Conf. Series: Journal of Physics: Conf. Series 944 (2018)
1234567890 ‘’“”012061 doi:10.1088/1742-6596/944/1/012061

28 12

24 10
8
20
6
16
4
12
2
8
0
4 I II
0
I II
Figure 7. Measuring error for the flat- Figure 8. Measuring error for the
ness for option 1–4. cylindricity for option 1–4.

It is established that for the considered variants only -error is characteristic. The results of calcu-
lating the -error in percent are shown in Fig. 10: I – flatness, II – cylinder, III – sphere.

16 25

20
12
15
8
10

4 5

0
0
I II III
I II
Figure 9. Measuring error for the sphe- Figure 10. Measuring -error error for
ricity for option 1–4. option 2–4: I – flatness, II – cylindricity,
III – sphericity.

Analysis of the obtained modeling data allows us to draw the following conclusions:
1) when using a smaller number of control points, the measured value of the surface error always
decreases, the largest variation with the base option 1 is given by option 4;
2) for all options only the measurement -error is characteristic, that is, the probability increases
that the defective workpieces is fit for measurement;
3) when measuring the plane, a satisfactory result gives a decrease in control points from 121 to 66,
while their location is not of fundamental importance (options 2, 3);
4) when measuring the cylinder, a satisfactory result corresponds to a non-uniform arrangement of
control points along the angular step, with a decrease in their number from 110 to 55 (option 3);
5) when measuring the sphere, a satisfactory result was obtained for all variants; therefore, a pref-
erence should be given to increase the productivity of variant 4 with the number of points 31, and to
ensure accuracy – to the variant with a decrease in the number of control points along the zenith angle
to 61 (option 2).

5. Conclusion
The method of optimization of the number of control points with the use of Monte Carlo simulation is
considered in the article. Based on the measurement of a small sample from batches of workpieces,
statistical modeling is performed, which allows one to obtain interval estimates of the measurement
error. The conducted researches established that in the measurement and analysis of flatness, cylin-
dricity and sphericity, the number and location of the control points significantly affects the magnitude
of the error. It is revealed that when the number of control points decreases, the arithmetic mean de-
creases, the standard deviation of the measurement error increases and the probability of the measure-

6
AMSD IOP Publishing
IOP Conf. Series: Journal of Physics: Conf. Series 944 (2018)
1234567890 ‘’“”012061 doi:10.1088/1742-6596/944/1/012061

ment -error increases. In general, it has been established that it is possible to repeatedly reduce the
number of control points while maintaining the required measurement accuracy.

Acknowledgments
The study was performed by a grant from the Russian Science Foundation (project № 16-19-
10204).

References
[1] Badar M A , Raman S and Pulat P S 2005 Experimental verification of manufacturing error pat-
tern and its utilization in form tolerance sampling International Journal of Machine Tools and
Manufacture 45 pp 63–73
[2] Obeidat Suleiman M, Abu Jadayil Wisam M and Mandahawi Nabeel F 2013 Intelligent Sam-
pling for Inspecting Milled Flat Plates International Review of Mechanical Engineering 7 pp
767–773
[3] Capello E and Semeraro Q 2001 The harmonic fitting method for the assessment of the substi-
tute geometry estimate error. Part II: statistical approach, machining process analysis and in-
spection plan optimisation International Journal of Machine Tools and Manufacture 41 pp
1103–1129
[4] Jones S D and Ulsoy A G 1995 Optimization strategy for maximizing coordinate measuring
machine productivity, part 2: Problem formulation, solution, and experimental results Journal of
engineering for industry 117 pp 610–618
[5] Pechenin V A and Bolotov M A 2016 Basing error in coordinate measurements of cylindrical
gears Russian Engineering Research 36 pp 630–634
[6] K. Carr and P. Ferreira, Verification of Form Tolerances Part I: Basic Issues, Flatness, and
Straightness Precision Engineering vol 17 1995 pp 131–143
[7] Poniatowska Małgorzata and Werner Andrzej 2012 Simulation tests of the method for determin-
ing a cad model of free-form surface deterministic deviations Metrol. Meas. Syst. XIX pp 151–
158
[8] Lee G, Mou J and Shen Y 1997 Sampling strategy design for dimensional measurement of ge-
ometric features using coordinate measuring machine International Journal of Machine Tools
and Manufacture 37 pp 917–934
[9] Raghunandan R and Rao P V 2007 Selection of an optimum sample size for flatness error esti-
mation while using coordinate measuring machine International Journal of Machine Tools and
Manufacture 47 pp 477–482
[10] Jalid A, Hariri S and Laghzale N E 2015 Influence of sample size on flatness estimation and un-
certainty in three-dimensional measurement International Journal of Metrology and Quality
Engineering 6 pp 102
[11] Ascione R, Moroni G, Petrò S and Romano D 2013 Adaptive inspection in coordinate metrolo-
gy based on kriging models Precision Engineering 37 pp 44–60
[12] Colosimo B M, Moroni G and Petrò S 2010 A tolerance interval based criterion for optimizing
discrete point sampling strategies Precision Engineering 34 pp 745–754
[13] Liu J S 2004 Monte Carlo Strategies in Scientific Computing (New York: Springer)
[14] Zakharov O V and Kochetkov A V 2016 Minimization of the systematic error in centerless
measurement of the roundness of parts Measurement Techniques 58 pp 1317–1321