|
Experimental
Data Analysis
The
aim of this chapter is to check the following conditions:
- Revealing
the type of distribution of the data population according
to the characteristic parameters and systematic error of
the GDV technique
- Influence
of liquid ionisation due to the process of GDV measurements
by comparison of results of subsequent GDV shots.
-
Sensitivity of the GDV technique to the concentration of
diluted salt.
-
Difference between different salts.
Study
of test-object GDV parameters
As a test-object we are using metal cylinder of titanium with
thoroughly polished sides. The test-object chosen is easily
produced; its parameters may be absolutely identical, irrespective
of the place and time of production. The GDV-image of this
object may be used to compare both the parameters of Visualisation
Systems and metrological and control problems, i. e. to check
working capacity of every concrete system. Special program
has been developed for metrological testing of the GDV devices
using test-object.
The
test-object was put on the working camera surface, and its
gas-discharge images were photographed. To avoid illumination
of the black-and-white GDV-image, the test-object was covered
by a special light-proof cloth. Previously, before each series
of experiments, the electrode surface was cleaned using alcohol
to move away the surface discharge, arising on the electrode
in the GDV-process. In each series not less than ten consecutive
measurements were carried out in succession. To test the time-dependence
of parameters, sixty series of experiments were carried out
during three month using different samples of GDV devices.
The following tests were carried out.
-
Two series of 50 consecutive measurements in different days
at the same environmental conditions (T = 22 C0, humidity
42 %).
-
For every device three series were made at a five-minute
and one-hour time interval, several tests were made with
the same device in different days. These intervals were
chosen to test the influence of day-time and environmental
conditions (atmospheric pressure, humidity, etc.) upon the
GDV-image of the object.
Before the measurements the GDV device was warmed up for
15 minutes.
Analysis of parameters received for the test-object.
Example of experimental data processing for test-object are
presented in Table 3.1 and fig.3.1 - fig.3.3 of Appendix 1
.These figures demonstrate the ratio errors of form factors,
fractal coefficients and luminescence areas. Fig. 3.1 demonstrates
data of measurements with five-minute time interval, fig.3.2-3.3
data taken at different days.
As
we see from these data the random errors of measurements for
form factors and fractal coefficients do not exceed 10%, and
the luminescence area inaccuracy - 8%. Different values of
inaccuracies are explained by the fact that fractal coefficient
and form factor are more variable than luminescence area,
i.e. more changeable from one experiment to the other.
The
results of data processing presented in Tables 3.1 indicate
that in all the cases the samples are most likely to be extracted
from the population of normal distribution. In all cases the
average value almost coincides with the median; the median
lies approximately in the middle between the 25th and 75th
percentile. Furthermore, the main values (from 95% to 98%)
are within the range of two standard abmodalities, and from
66% to 73% of the values - within the range of one standard
abmodality.
Results
obtained suggest as follows:
1. for the GDV technique standard instrumental error does
not exceed 8 % ;
2. data of the measurements follow the normal distribution
statistics ;
3. difference of the measurements data in different time intervals
during the day and in different days statistically does not
significant, that prove reliability of the GDV Technique.
Research
on revealing statistically significant differences between
distilled water, normal solution of an extra-pure alkalis
and their ionised analogues.
Measurements
for distilled water.
Based on the data of more than 200 measurements of distilled
water samples as a "null hypothesis" a supposition
may be taken that differences between the measurements have
randomicity. Significance test is taken equal to 5%. We compare
the data with the help of an ordinary t-criterion of Student.
The results of data processing indicate that in all the cases
the samples are most likely to be extracted from the population
of normal distribution. In all cases the average value almost
coincides with the median; the median lies approximately in
the middle between the 25th and 75th percentile. Furthermore,
the main values (from 95% to 98%) are within the range of
two standard abmodalities, and from 66% to 73% of the values
- within the range of one standard abmodality. Data, received
in different days may be accepted as extracts from the same
general aggregate.
Influence
of the GDV measuring process on the liquid samples.
Table 3.2 shows the data for distilled water. It was revealed
that when comparing distilled water with its ionised analogue
the possibility of the "null hypothesis" validity
is equal to 0,021, which is lower than the significance test
and, hence, the "null hypothesis" may be discarded.
This fact is also indicated by the confidence interval of
the difference, as it does not include the zero value. The
63,8 % sensitivity of test is quite enough to point to the
fact that the differences are statistically significant with
a high reliability.
A
similar comparison of NaCl and its ionised analogue, demonstrated
in Table 3.3, brings to the same conclusions concerning statistical
significance of the differences, with the only exception that
on account of a larger sampling and NaCl characteristics the
validity probability of the "null hypothesis" is
equal to 0,008, i.e. much closer to the zero. The sensitivity
of test is equal to 76%, at that.
Statistically
significant differences between water and NaCl are evident
even in small samples with the sensitivity equal to one, as
demonstrated in Table 3.4 and at fig.3.4.
Basing
on the data received, a conclusion may be drawn that for a
5% significance test according to the sensitivity table the
sensitivity at 60% level may be planned at the sampling quantity
of 30-40.
Revealed
data testify that in the process of measurements we can detect
influence of electromagnetic field to the liquid sample and
the difference is statistically significant.
Investigation
of Various Solution»
«
Back to Home
gdvtechnique.com
@ 2001. All rights Reserved
|
