Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
net/publication/231151106
CITATION READS
1 209
2 authors:
All content following this page was uploaded by Gorazd Bombek on 22 September 2016.
This article has been downloaded from IOPscience. Please scroll down to see the full text article.
(http://iopscience.iop.org/0957-0233/21/8/085101)
View the table of contents for this issue, or go to the journal homepage for more
Download details:
IP Address: 164.8.20.42
The article was downloaded on 22/06/2010 at 13:36
Abstract
This paper presents a method for measuring the velocity of a flow of particles accelerated on a
shot-blasting wheel and then expanding into space at a wide angle of approximately 45◦ . The
method uses the pulsed nature of the flow characteristic for turbo machinery with a finite
number of wheel blades and calculates the velocity from the time shift between the particles
hitting two targets at a known distance. This method does not depend on the material
properties of the particles; however, a large number of particles is required for a successful
measurement. The impacts are detected with a microphone covered by a steel membrane,
making the exposed parts cheap and easily replaceable. This makes the method suitable for
industrial test-and-development purposes, including the efficiencies of acceleration
measurements. A cross-correlation of the signals was used, but the characteristic cycle (one
rotation of the wheel) had to be determined beforehand by overlapping and averaging several
cycles in order to compensate for the fact that different particles would be hitting the targets.
Keywords: shot blasting, wheel, velocity, particle, measurement, efficiency, microphone,
overlapping
(Some figures in this article are in colour only in the electronic version)
0957-0233/10/085101+10$30.00 1 © 2010 IOP Publishing Ltd Printed in the UK & the USA
Meas. Sci. Technol. 21 (2010) 085101 G Bombek and A Hribernik
Figure 1. The expansion of the waves, the important dimensions (in mm), and the idea behind the measurement method.
2.1. Development of the sensor refers to every collision between a particle and the head of
the sensor. A side impact, on the other hand, refers to
The requirements for the sensor are challenging, and every collision between a particle and any other part of the
sometimes one requirement excludes another. The sensor must measurement system, including girders and other sensors on
be able to withstand a harsh environment, detect the impact of the same girder, generally resulting in noise. The aim was
a single particle, have the minimum vibration transfer between to obtain the largest-possible response from a frontal impact
the sensors and must enable a separation of impacts that occur and the smallest-possible response from a side impact. The
in 1 ms or less. Similar issues were presented in [11], which development process and the test results are presented in
dealt with much smaller particles and a much lower mass [12–14], but the focus of this paper is on organizing the data,
flow of particles. As a result, a special hybrid sensor was extracting the results and then validating them (so table 1 only
developed. This sensor uses a microphone (costing approx. presents the essential parameters). The ringing time is the
€5) for sensing the sound when a particle collides with the time needed for the signal to return to the base value or to stop
front plate of the sensor. The principle of the sensor is oscillating, as shown in figure 3.
presented in figure 2. The current, optimized version (type D) is presented in
The development of the sensor was a long process. figure 2; it enables a good separation between the frontal and
Different types of sensors were tested using an impact ball side impacts resulting from the impact ball test, the results of
to replicate the momentum of a 1 mm diameter spherical which are presented in figure 3.
particle travelling at 80 m s−1 . The responses were studied The results of the impact-ball test (presented in figure 3)
in terms of frontal and side impacts. Here, a frontal impact were acquired while testing the sensors on the same girder
2
Meas. Sci. Technol. 21 (2010) 085101 G Bombek and A Hribernik
2.5
2.0
0.5
0.0
0 0.5 1 1.5 2
time (ms)
Figure 3. Signal of the type-D sensor (transient response of the optimized sensor to single front and side impacts) [13] in the case of an
impact-ball test.
Figure 4. Three sensors on a common girder (all the dimensions are in mm).
Table 1. Sensor development and the essential parameters. was used, and digital filtering and the threshold method were
Amplitude Ringing Amplitude Sensor necessary in order to obtain the results. However, it was hard
frontal time ratio side/ diameter to estimate the time shift caused by the filtering and to specify
Type impact (V) (ms) frontal (mm) the optimum threshold. These parameters had a great influence
on the results, so a decision was made to avoid filtering (if at
A 0.015 3 0.67 24 all possible) and to make our measurements as representative
B 0.1 3.5 0.33 21
C 0.04 1.5 0.08 22 as they could be, which suggested that some kind of averaging
D 1.7 0.6 0.06 11 would be the most likely solution. The signals detected in the
shot-blasting chamber are presented in figure 5.
3
Meas. Sci. Technol. 21 (2010) 085101 G Bombek and A Hribernik
1.4
1.2
0.8
sensor1
signal (V)
0.6 sensor2
sensor3
0.4
0.2
-0.2
0 5 10 15 20 25
time (ms)
1.4
1.2
0.8
sensor1
Signal (V)
0.6 sensor2
sensor3
0.4
0.2
-0.2
1.5 2 2.5 3 3.5
time (ms)
industrial conditions, so contamination with other (smaller) determination of the time shift difficult, and so a special
particles was likely; however, most of the particles smaller than algorithm had to be developed.
0.3 mm were removed by a filter. As presented in figures 5
and 14, the mass flow depends on the position of the sensors, 2.2. Development of the algorithm for the time-shift
and it varies from wave to wave. The signals from the shot- determination
blasting chamber show that the particles actually travel in
waves. The problem is that more than one particle hits each 2.2.1. Data-acquisition parameters. The need for a large
number of cycles (averaging) and high rates (short distances
sensor with almost every wave. There is also the possibility
between the targets) resulted in the following parameters for
that no impact occurs or is sensed during a wave. The first
the data acquisition:
wave from figure 5 is shown in detail in figure 6.
As can be seen in figure 6, there are several impacts during • sampling rate 100 kHz (per channel)
each wave. Any difference in the signal amplitude is mostly • number of samples 1 M (10 s)
caused by the different gaps between the microphones and • number of channels 3 (4).
the sensor heads of the individual sensors. The number of The frequency of the rotor is 50 Hz, so every measurement
impacts also varies between the sensors, which makes any covers about 500 rotations of the wheel. Since the wheel
4
Meas. Sci. Technol. 21 (2010) 085101 G Bombek and A Hribernik
663000
662000
661000
660000
659000
658000 c-correl1-2
c-correl value
657000 c-correl1-3
656000 c-correl2-3
655000
654000
653000
-500 -400 -300 -200 -100 0 100 200 300 400 500
measuring point (10-5 s)
0.5
0
signal (V)
-0.5
-1
-1.5
1 201 401 601 801 1001 1201 1401 1601 1801 2001
-2
measuring point (10-5s)
has six blades, this results in 3000 waves impacting on the of shockwaves can be expected [16]. This method is based
sensor. on a time-of-flight determination, but a non-stationary cross-
correlation function is used to compensate for the noise and
2.2.2. Conventional methods for time-shift determination. the disturbances. Our main problem with existing methods
The usual method for determining the time shift is cross- lay in the poor repeatability of the waves, and so we had to
correlation. The results of the cross-correlation are presented find a way of improving this. This is why we developed a new
in figure 7. approach to determining the time shift.
The peak of the cross-correlation between signals 2 and 3
was, as expected, located in the middle of the signal. However, 2.2.3. Overlapping and averaging of the signals. In the
the exact locations of these cross-correlation peaks between case of a constant rotational wheel speed, the impacts occur
signals 1 and 2, and between 1 and 3, were almost impossible at constant time intervals. However, the repeatability of the
to establish, even though a frequency-domain cross-correlation signal can be substantially improved by clustering six waves
was used (a direct method gives even rougher results). This together (i.e. one rotation of the wheel). It is possible to
was due to the poor repeatability of the waves. Generally, calculate the average intensity of a signal by overlapping a
those particles that hit the first sensor do not hit the other signal interval representing one rotation of the wheel, and then
sensors. This means that the number of particles hitting the averaging the amplitudes. The result of this signal overlapping
second (or third) sensor reduces with the distance between is presented in figure 8. Accurate overlapping is only possible
the sensors, and so some similarities with the propagation if the reference points at which the signals have to be cut for
5
Meas. Sci. Technol. 21 (2010) 085101 G Bombek and A Hribernik
0.6
0.5
0.4
0.2 sensor 1
sensor 2
0.1 sensor 3
0.0
-0.1
1 201 401 601 801 1001 1201 1401 1601 1801 2001
measuring point (10-5s)
Figure 9. Averaged impact signal derived from the data presented in figure 5 (6 blades ∼1 wheel rotation).
10
6
c-correl value
4 c-correl1-2
c-correl1-3
2 c-correl2-3
-2
-500 -400 -300 -200 -100 0 100 200 300 400 500
-4
measuring point (10-5s)
overlapping can be correctly obtained. The first overlapping It was decided to use cross-correlation (a direct method), and
procedure used frequency analysis, but the frequency of the the results are presented in figure 10.
wheel’s rotation was not sufficiently constant and a problem From the data presented in figure 10, the maximum values
occurred when the frequencies of the impacts were transferred and their locations can be easily found and used for the velocity
to the number of samples. In other words, if it is not a whole calculation. As already presented in figure 4, the distance
number (and generally it is not) it has to be rounded to the between the sensors is set at 60 mm in order to minimize the
nearest whole. As a result, the error of the interval position size of the sensor system and to prevent sensors 2 and 3 being in
increased with each interval, and so the peaks shifted. The the shadow of sensor 1. The sensor is approximately 800 mm
solution to this problem was to introduce a fourth sensor for away from the wheel. We assumed that approximately 20%
sensing the position of the wheel, and to use this signal as a fewer particles hit sensors 2 and 3 compared with sensor 1,
reference point. A very good repeatability and stability was but the velocity of the particles was assumed to be constant.
achieved when using this procedure, as can be seen in figure 8, Our method is based on determining the time shift between
with only 22 cycles out of 450 being presented (for reasons of the peaks of the average impact signal, so the amplitude of
clarity). the signal and the number of hits is unimportant if the average
In figure 8, the waves are clearly visible. The average impact signal is representative. This is why 3000 particle
intensity of all three signals is presented in figure 9. waves were used to determine the average impact signal. The
The signals (averaged from 450 cycles) presented in locations of the peaks correspond well, but the amplitudes
figure 9 are much smoother than the original raw signals depend on the signal amplitude, which is higher in the case of
(figure 5), and the time shift can be determined manually. sensor 3, as presented in figures 6 and 9. However, as already
It is, however, faster and more objective to use an algorithm. mentioned, this does not influence the results.
6
Meas. Sci. Technol. 21 (2010) 085101 G Bombek and A Hribernik
7
Meas. Sci. Technol. 21 (2010) 085101 G Bombek and A Hribernik
8
Meas. Sci. Technol. 21 (2010) 085101 G Bombek and A Hribernik
80 1.5
78 1.35
76 1.2
74 1.05
veelocity (m/s) (5.5 kg/s)
velocity (m/s)
72 0.9
rel. mass flow
Figure 16. Vertical velocity profile in accordance with the relative mass flow.
9
Meas. Sci. Technol. 21 (2010) 085101 G Bombek and A Hribernik
10