GRANT
journal
ISSN 1805-062X, 1805-0638 (online), ETTN 072-11-00002-09-4
EUROPEAN GRANT PROJECTS | RESULTS | RESEARCH & DEVELOPMENT | SCIENCE
Figure 1 Sensor
head
Figure 2 Sensor
contact
The integral ring gauges with 3 nominal dimensions (diameters of
16 mm 50 mm and 90 mm) were used to simulate the real part. They
are marked in the test as small - medium - large diameter, see Fig. 3.
Figure 3 Ring gauges
The following parameters were evaluated on each gauge
Diameter of the gauge at two depths
Circularity of the gauge at two depths
Cylinder diameter
Cylindricality
When setting the test conditions, the mathematical definition of
individual measured elements was used. The number of points was
set as follows, see Table 1.
Number of points to construct a ring
50, 100, 500, 1000, 3000, 5000 points
Number of points to construct a
cylinder
100, 200, 1000, 2000, 6000, 10000 points
Table 1 Number of points for a circle and a cylinder
The rings were measured in two cross sections. The measurement
was performed using contact scanning. All measurements were
repeated five times. All measurements were made in a clockwise
direction.
4.
DATA ANALYSIS
When analyzing the data, it was divided into appropriate categories
(diameter at two measuring points, roundness, cylindricality,
cylinder diameter). The data were tested for normality in the
Minitab program. Normality was confirmed for all measurements.
Subsequently, in the procedure for determining the optimal number
of points, the standard deviation of individual categories was
calculated. The standard deviations are used for the initial quick
creation to give an idea of how the data behaves.
The number
of points in
the element
50
100
500
1000
3000
5000
Diameter 1
0.000519
0.000587
0.000524
0.000542
0.000542
0.000572
Diameter 2
0.000560
0.000528
0.000575
0.000573
0.000599
0.000547
Circularity 1
0.001016
0.000615
0.000347
0.001414
0.000484
0.010426
Circularity 2
0.000531
0.000719
0.000488
0.000347
0.001532
0.005635
Table 2 Table with reset of standard deviation
The number of
points in the
element
100
200
1000
2000
6000
10000
Diameter of
cylinder
0.000571
0.000545
0.000591
0.000526
0.000541
0.000564
Cylindricity
0.000706
0.000852
0.011972
0.004158
0.011402
0.016926
Table 3 Table with reset of standard deviation
Tables 2 and 3 show the changes in standard deviations for the
individual measured elements. Looking at tables 2 and 3, it is
possible to clearly identify trends in the behaviour of the evaluated
data. There are two trends, depending on which element is
evaluated. The first trend confirms the findings in article 2, which
led to the conclusion that 30 points are sufficient to evaluate the
average. The second trend indicates that when scanning and
evaluating circularity or cylindricity, the points used for the test are
probably not sufficient for the correct evaluation of the given
elements, due to the fact that their standard deviation is still
increasing. This suggests that we will need to test even higher
numbers of points in order to evaluate circularity and cylindricity.
4.1
Analysis of circle and cylinder diameters
A multi-factor ANOVA method was used to evaluate the effect of
points. The evaluation of each diameter (gauge) was performed
separately [5,6]. However, bearing in mind that both the evaluation
of the individual evaluated parameters (diameter, roundness,
cylindricality,…) and the evaluation as a whole is important for us.
4.1.1 Large diameter–evaluation of element diameter
Here we provide an example of the evaluation of a gauge with a
large diameter. The processing and evaluation of the test took place
according to the same conditions and procedures as in article 2. The
method of evaluation was left the same due to the relevance and
possible interconnection of the achieved outputs.
Table 4 shows the results for a one-factor ANOVA method. It can
be seen that for the result based on the p-value in the input test, the
number of points will not have such a significant effect on the
measured values. However, at the same time, this table indicates that
our evaluation lacks some other influences that have a significant
impact on the achieved results. [6,7,8] This statement is also based
on the value of the accuracy of the prediction model, which is very
low; the value of R-sq is only 6%.
Source
DF
Adj SS
Adj MS
F-Value
P-Value
Numberofpoints 5
0.000000 0.000000
0.32
0.896
Error
24 0.000003 0.000000
Total
29 0.000003
S
R-sq
R-sq(adj)
R-sq(pred)
0.0003256
6.25%
0.00%
0.00%
Table 4 Table with ANOVA result
Based on this finding, the measured values were displayed in a
graph (Fig. 4) where the effect which was neglected during the
initial evaluation, is very clearly visible. This is the effect of the
measuring loop (stepped arrangement of points).
Vol. 9, Issue 1
101