DETERMINING THE REPEATABILITY OF ACOUSTIC EMISSION ...mqa.ogpta.polsl.pl/files-articles/25/pdf/177-192_Boczar_Lorenz.pdf · DETERMINING THE REPEATABILITY OF ACOUSTIC EMISSION

Molecular and Quantum Acoustics vol. 25, (2004) 177

DETERMINING THE REPEATABILITY OF ACOUSTIC EMISSION GENERATED BY THE HSU-NIELSEN CALIBRATING SOURCE

Tomasz BOCZAR, Marcin LORENC

Technical University of Opole, Department of Electrical Engineering and Automatic Control31 K. SosnkowskiegoStr., 45-272 , Opole, POLAND

e-mail: [email protected], [email protected]

This paper characterizes the idea of calibrating measurement paths, which make the measurement of the acoustic emission (AE ) pulses generated by partial discharges (PDs) in insulation of power appliances possible, by using the Hsu-Nielsen method. It also presents the results of the analyses done in the time and frequency domains of the measured AE pulses generated by the Hsu-Nielsen calibrating source. The measurement and analysis of the AE pulses were performed on the lid of the distributive transformer tub, for various distances between the source of calibrating signals and a measurement transducer in the range from 2.5 to 10 cm. The analysis of the results obtained was carried out based on a series of selected descriptors characterizing the AE signals measured, separately for the time and frequency domains.

Moreover, the paper presents the results of parametric and non-parametric tests of goodness of fit, which were carried out to determine the probability distribution of the AE pulses measured and to examine the repeatability of the frequency analysis results obtained for a given distance between the transducer and the source of calibrating signals. Statistical analyses were carried out based on the values of three descriptors: shape coefficient, peak coefficient and median frequency, which make it possible to identify basic PD forms occurring in insulation oil.

Keywords: Hsu-Nielsen calibration methods, partial discharges, time and frequency analyses of AE pulses, repeatability of the measurement results

1. INTRODUCTION

Recent years have witnessed a dynamic development of the AE method. It finds a

wider and wider application in many industrial branches, e.g.: raw materials, materials,

machines, electronics, and power engineering [1]. Very important applications of the AE

method in power appliance diagnostics are detection, measurement and location of PDs in

power transformer insulation, high-voltage transformers, bushing insulators, SF6 switching

stations. [2]. The main factor influencing progress in the AE measurements is the

improvement of methods and measuring apparatus as well as deepening the knowledge of

physical processes accompanying generation and propagation of the AE signals.

The results of the AE measurements provide electrical quantities (e.g. voltage) the

values of which do not make it possible to determine the absolute value of a pulse registered.

mailto:[email protected]

178 Boczar T., Lorenc M.

It often happens that it is not possible to measure the size of an AE signal in the place of its

generation. It is usually so because the source of the AE is as a rule inside the dielectric under

study, which causes that when a signal reaches a transducer it is suppressed and reflected

many times. Moreover, the medium coupling a transducer with an insulation system under

study can cause the suppression of the primary signal. Also the measuring system introduces

some indeterminacy of the signal registered in relation to the signal in the place of generation.

All these are incentives for taking quantitative measurements of the AE based on relative

measurements. Quantitative measurements consist in comparing the pulses under study with

model signals generated by artificial sources. A theoretical study of this issue was given by

Breckenridge [3].

Since the AE can be either continuous or discrete, the attempts to produce model

signals have been made, which for discrete signals correspond with delta Dirac function and

in the case of continuous emission with white noise. As an impulse model source breaking a

glass capillary was suggested [4]. McBride and Huchison proposed a method consisting in

the outflow of gas from a nozzle for calibration of apparatus for registration of continuous AE

[5]. The calibration signal should be characteristic of the parameters determined in the time

and frequency domains and be as close to the AE signal as possible. The most proper

calibration methods are the methods that make it possible to produce a pulse of strictly

determined and repeatable AE parameters, which are easy to produce in laboratory conditions.

They provide the possibility to compare the results obtained in various laboratories dealing

with measurements and analyses of the AE signals. Calibration of the apparatus also becomes

vital when the AE measurements are taken in a costly or unrepeatable experiment.

The scope of the research work carried out, the results of which are presented in this

paper, comprised the measurements of the AE signals generated by calibrating heads made

according to the guidelines defined by Hsu-Nielsen and the analysis of the AE signals

measured in the time and frequency domains. Moreover, the influence of the distance between

the calibrating head position and the measuring transducer on the obtained descriptor values

characterizing time and frequency runs of the AE signals measured was examined. Using

statistical tests the repeatability of the calibrating signals generated was tested. The evaluation

of the repeatability was carried out for selected descriptors characterizing frequency spectra

measured for a given distance between the source and the transducer of the AE signals

generated by the Hsu-Nielsen calibrating source.


2. THE IDEA OF CALIBRATION USING THE HSU-NIELSEN METHOD

The Hsu-Nielsen method was elaborated at the beginning of the 1980s [6]. It is a

simple and easy method to apply and it does not require any complex auxiliary apparatus or

training people taking the measurements. The calibrating device comes down to making a

proper head, automatic pencil and marking points of 2H hardness and the gauge of 0.5 mm.

The calibration signal is generated in the Hsu-Nielsen system at the moment of breaking a

sensitive graphite pin cased by a properly adapted tip put on the automatic pencil. The head

and the marking point pulled out to 3 mm ensure the same breaking angle at each test, which

also means generation of a repeatable acoustic surface wave. This method is classified as one

of the pulse calibration methods. The process of breaking gives the AE pulse congenial with

the signal obtained in the process of breaking the glass capillary. There is, however, some

difference between them, which consists in the occurrence of an inconsiderable minimum in

the initial phase of the run. It results from the fact that the breaking of the marking point is

done by a manual pressing of the pencil which is of a dynamic character. The point of the

pencil presses the surface directly in one point. This method found its application in

calibration of the AE detectors in the process of their production. Owing to its simplicity it is

used for calibration of measuring systems installed in technical conditions. It was also

standardized [7].

3. TECHNICAL PRODUCTION OF THE CALIBRATION SOURCE AND SELECTION OF THE MEASURING SYSTEM

The calibration head was made based on the data from the literature [7]. The draft of

the head is shown in Fig. 1. Following the draft the head was made from methyl

polymethacrylate. This material guaranties durability and constancy of the required shape and

measurements of the head. The head was turned from a rod of the diameter of 7 mm. After

giving the head the required shape and measurements, a concentric opening of the diameter of

0.9 mm was bored. Such calibration heads comply with the requirements described by the

French standard [7].

Since the main aim of the measurement taking was the selection of the way of

calibrating the systems for measuring the AE generated by PDs, the fundamental tests were

done on a transformer of the following parameters:

- type TP 60/110,

- upper voltage of 110 kV,

- total mass of 345 kg.


4

3

0 , 92

0,5 7

Fig. 1. Draft of the Hsu-Nielsen calibration head

The transducers calibrated were placed on the upper lid of a transformer tub assuming

four different distances between the model AE source and the transducer. To ensure an

effective mechano-electrical coupling the transducer was fixed to the transformer through a

layer of cup grease. Fig. 2 shows an overall view of the transducer calibrated and installed on

the transformer.

Fig 2. Overall view of the transducer calibrated by the Hsu-Nielsen method, installed on a transformer


4. COLLECTIVE LISTING OF THE RESULTS

The AE signals generated during calibration were registered by the apparatus

described in the papers [8 ]. For each distance assumed the calibration test was repeated five

times. Based on the results obtained criterion calibration descriptors were determined and

defined [8]. For the particular measurement series the following statistical parameters were

also determined: average value, standard deviation and variance. The measurement results are

listed in Tables 1 through 4. Table 1. shows parameters characterizing time runs of the signals

registered such as: the length of a signal, the sum of the AE counts, and the sum of transitions

through a selected level of the discrimination threshold. During the whole experiment the PD

discrimination threshold was adopted at the level of PD = 0.2 V.

Table 1Listing of the descriptor values calculated for the time runs of the AE signals generated by the Hsu-Nielsen method

Distance between the

transducer and the generation

place [mm]

Statistical quantitiesSignal length[ms]

Sum of the AE counts [n.imp.]

Sum of transitions through the

adopted discrimination level [n.imp.]

25Average value

Standard deviationVariance

2,68 1169 209

0,30 172 31

0,09 29563 959

37Average value


2,40 932 191

0,31 66 18

0,09 4318 340

62Average value


2,84 829 148

0,47 170 36

0,22 28766 1281

100Average value


2,45 713 155

0,40 145 24

0,16 21060 559


Table 2Listing of the descriptor values calculated for the time runs of the energy of the AE signals generated by the Hsu-Nielsen method

Distance between the

transducer and the generation

place [mm]

Statistical quantitiesAverage

amplitude[V]

ARMS value[V]

Surfaceabove the

average value[V2]

Peak coefficient[ - ]

25Average value 0,48 0,30 132,60 5,38

Standard deviation 0,03 0,06 30 0,69Variance 1,14E-03 3,78E-03 921 0,47

37Average value 0,47 0,26 85,75 5,31

Standard deviation 3,85E-02 0,02 14 0,58Variance 1,48E-03 5,59E-04 188 0,34

62Average value 0,42 0,26 74,40 4,93

Standard deviation 0,03 0,03 20 0,46Variance 1,20E-03 1,04E-03 392 0,21

Average value 0,397 0,203 56,800 4,78Standard deviation 0,032 0,033 16 0,745

Variance 0,001 1,10E-03 271 0,555

Table 3Listing of the descriptor values calculated for the spectral amplitude density of the AE signals generated by the Hsu-Nielsen method

Distance between the transducer

and the generationplace [mm]

Statistical quantities

Max. value of

the spectral

line in the spectrum

[V]

Average value of

the spectral


[V]

Frequency for the max.

amplitude density[kHz]

ARMS value

Peak coefficient

[ - ]

Median frequency

[kHz]

25

Average value 5,10 0,08 5,46 0,33 15,36 15,90

Standard deviation 0,39 0,01 0,80 0,04 2,28 2,44

Variance 0,156 1,41E-04 0,64 1,80E-03 5,20 5,94

37

Average value 4,24 0,07 4,88 0,27 15,52 17,78


Variance 1,22 7,00E-06 0,00 6,18E-04 12,41 4,44

62

Average value 4,29 0,06 5,37 0,27 15,58 15,90


Variance 0,96 6,07E-05 0,00 1,03E-03 5,65 1,48

100

Average value 2,79 0,046 5,71 0,215 12,38 15,8


Variance 0,34 4,25E-05 0,10 1,40E-03 0,62 1,80


Table 4Listing of the descriptor values calculated for the spectral energy density of the signal energy generated by the Hsu-Nielsen method

Distance between the transducer

and the generationplace [mm]

Statistical quantities

Max. value of

the spectral


[V]

Average value of

the spectral


[V]

Frequency for the max.

amplitude density[kHz]

ARMS value

Peak coefficient

[ - ]

Median frequency

[kHz]

25

Average value 24,30 0,09 5,76 1,50 41,70 6,93

Standard deviation 3,78 2,80E-02 0,80 2,03 6,77 1,04

Variance 14,30 7,84E-04 0,64 4,12 45,9 1,09

37

Average value 23,78 0,066 4,88 0,52 36,01 6,77


Variance 2,40 1,50E-04 0,00 5,52E-04 226,90 0,77

62

Average value 19,30 0,07 5,37 0,52 36,15 6,10


Variance 97,08 3,21E-04 0,00 1,36E-03 271,39 0,15

100

Average value 7,18 0,04 5,71 0,46 15,27 7,03


Variance 8,10 1,61E-04 0,11 1,46E-03 24,75 0,01

5. THE ANALYSIS OF THE RESULTS OBTAINED

Table 1. shows the values descriptors determined for the time runs that were calculated

for four distances between the model source and the measuring transducer. The increase of the

distance did not influence significantly the time of the signals registered. The values

determined were in the range from 2.45 ms to 2.84 ms. In the case of the other two descriptors

the influence of the distance change on the values obtained was observed. With the increase

of the length of the signal the values of the sum of the counts and the sum of transitions

through a selected discrimination threshold decrease. This is caused by the influence of the

object configuration in which the AE signals generated propagate. The upper part of the

transformer is of a non-homogenous structure. The increase of the distance between the

source and the transducer causes the shortening of a free path. The elastic wave when

propagating in a non-homogeneous medium can be subject to reflection or refraction. The

wave motion consists in propagation of a spherical disturbing wave which causes particle

0

20

40

60

80

100

I

II

III

IV

- 25 mm

- 37 mm

- 62 mm

- 100 mm

Average amplitude

Surface above the average value

Peak coefficient

ARMS

value


deflection from the state of equilibrium. At some point the primary wave and the wave

formed through transformation reach the same point of the medium. Each of them causes

deflection characteristic for its kind, i.e. such deflection as it would cause if the other wave

did not exist. The resultant deflection from the state of equilibrium of the medium point under

study is equal to the geometrical sum of constituent deflections. This law is binding in the

range of the Hookes Law applicability i.e. in the range in which the whole experiment was

carried out. In consequence, it causes the increase of the maximum amplitude, the sum of

counts and the sum of transitions through a selected discrimination threshold. Fig. 2 shows an

exemplary run of the model signals registered, generated from the Hsu- Nielsen source.

2000 4000 6000 80002.5

1.25

0

1.25

2.5

pip

ip

Fig. 2. An exemplary time run of the model AE signal generated by the Hsu-Nielsen source

The effect of elastic waves transformation is also noticeable for time runs it has an

effect on the decrease of the average amplitude value and the size of the surface determined

above the average value. With the increase of the pulse duration time the value of the energy

derivatives such as the root-mean-square value ARMS and the peak coefficient also decrease

with the increase of the distance.

Fig. 3. Relative changes of the descriptors of derivative time runs determined for the AE signal generated in the Hsu-Nielsen system


Analyzing the descriptor values obtained in the frequency domain calculated for the

spectral amplitude density and the spectral energy density the influence on the medium values

obtained in which the propagation of waves took place is also observable. An exemplary

amplitude spectrum of a model AE signal is shown in Fig. 4.

0 50 100 150 200 250 300 350 400 450 500100

80

60

40

20

0

ff1k

frk

Fig. 4. Amplitude spectrum of the model AE signal shown in Fig. 2

The shortening of the free path, on which the wave propagates, caused the decrease of

the value of the maximum spectral line and the ARMS value. The geometry of the medium did

not influence the frequency of the maximum spectral line, the value of which was from 4.88

kHz to 5.71 kHz. The values of the median frequency showed similar trends, which means

that the division of the power transferred by the particular harmonics is unchanging. However,

the values of the peak coefficient were fluctuating significantly due to considerable changes

of the energy values for those runs.

Fig. 5. Relative descriptor values calculated for the spectral amplitude density of the AE signals generated in the Hsu-Nielsen system

Maximum value in a spectrum

Average value in a spectrum

Frequency for the maximum

spectral density

ARMS

value Peak coefficient

Median frequency

0

20

40

60

80

100

I

II

III

IV

- 25 mm

- 37 mm

- 62 mm

- 100 mm


Fig. 6 shows an exemplary energy spectrum of a signal. The main portion of the

energy is transferred by the waves of the frequency in the range from 0 to 10 kHz. In the

group of descriptors determined for the spectral energy density there could be observed

analogous changes as for the descriptors determined based on the amplitude spectrum.

Relative values of descriptors for the spectral energy density are shown in Fig. 7.

0 10 20 30 400

10

20

30

40

fmk

frkFig. 6. Spectral energy density of the AE signal shown in Fig. 2

Fig. 7. Relative descriptor values calculated for the energy density spectrum of the calibrating signals generated in the Hsu-Nielsen system

0

20

40

60

80

100

I

II

III

IV

- 25 mm

- 37 mm

- 62 mm

- 100 mm

Maximum value in a spectrum

Average value in a spectrum

Frequency for the maximum

spectral densityA

RMS value Peak

coefficientMedian

frequency


6. CHARACTERISTICS OF THE MEASUREMENT RESULTS OF CALIBRATING SIGNALS GENERATED IN THE HSU-NIELSEN SYSTEM AS RANDOM VARIABLES

In order to determine the repeatability of the calibrating signals generated by the Hsu-

Nielsen source the methods of statistical inference were used. To carry out a statistical

analysis for the population of the results obtained there were created samples consisting of

randomly selected elements that represented them. Statistical tests were done for

measurements obtained at the distance of 10 cm between the transducer and calibrating head.

That is the distance which can be used during the calibration of measurement paths that

measure AE signals generated by PDs in paper-oil insulation of power transformers.

For the designated random samples, function runs and density probability cumulative

distribution functions were determined, where using histograms the empirical distribution of

the values measured was presented, and using a continuous line their theoretical distribution

was presented. In order to verify the statistical hypothesis of equivalence of the probability

density of the AE pulses analyzed with the assumed theoretical probability density a non-

parametric 2 text of goodness of fit was used. Within the analyses carried out it was proved

that with a 5% tolerance of error making there are no bases for rejecting the zero Ho hypothesis which assumes a normal type of distribution of the AE pulses measured generated

in the Hsu-Nielsen system. The detailed results of the test verifications carried out have been

presented, among others, in the works [9, 11, 12]. Moreover, in order to verify graphically the

normal distribution probability of the random variables tested probability diagrams of the

normal and semi-normal probability were used.

Figs 8-11 show in order the probability density and cumulative density function, the

normal and semi-normal probability diagram for random samples representing the AE pulses

generated in the Hsu-Nielsen system.

U [ V ]

frequ

ency

0

500

1000

1500

2000

2500

3000

3500

4000

-2,5 -2,0 -1,5 -1,0 -0,5 0,0 0,5 1,0 1,5 2,0 2,5

Fig. 8 Probability density functions (PDF) determined for the AE pulses generated the

U [ V ]

frequ

ency

0100020003000400050006000700080009000

100001100012000130001400015000160001700018000

-2,5 -2,0 -1,5 -1,0 -0,5 0,0 0,5 1,0 1,5 2,0 2,5

Fig. 9 Cumulative density function determined for the AE pulses generated the Hsu-Nielsen


Hsu-Nielsen system system

observed value

norm

al v

alue

-5

-3

-1

1

3

5

-3 -2 -1 0 1 2 3

Fig. 10 Normal probability plot determined for the AE pulses generated the Hsu-Nielsen

system

observed value

norm

al v

alue

-0,5

0,5

1,5

2,5

3,5

4,5

-3 -2 -1 0 1 2 3

Fig. 11 Semi-normal probability plot determined for the AE pulses generated the

Hsu-Nielsen system

7. TEST RESULTS OF THE MEASUREMENT REPEATIBILITY

The repeatability tests were done for three descriptors, i.e. median frequency, peak

coefficient and shape coefficient, which can be used for identification of basic PD forms

[10-13]. The measurement cycles took place in similar environment conditions, e.g.

temperature, humidity and air pressure.

The conclusions from the analyses carried out were drawn based on the test

of significance that is based on the variance analysis for many averages of a single

classification [14-20]. This test is based on F Snedecor distribution and the assumption that

there are given k populations of a normal distribution N(mi, i), where i =1, 2, , k, or close

to a normal distribution, and the variances of all k populations are equal, i.e. 12 = 22 = ... =

k2 = 2, but they do not have to be known. Hence, in order to begin the test it was necessary

to check first whether the data of measurements were within a normal distribution. Following

the earlier research, already published [9, 12, 13], it was decided, based on the assumption of

the concord of the data with the distribution of a normal type, to find out whether there also

occurs the equality of the variance of the data. A homogeneity test of many variances [14-20]

is based on the distribution 2 and the assumption that there are k normal populations N(mi,

i), where i = 1, 2, ..., k, of ni number. The selected measurement data on which the

calculations were performed are presented in Table 5.


Table 5Comparative listing of coefficient values calculated for the results obtained and read from the tables for 2 and F tests for the peak, shape and median frequency coefficients calculated for the amplitude spectrum and energy density of the AE pulses generated by the Hsu-Nielsen calibration source.

DescriptorAverage

value Standard deviation 2 2

F F

Spec

tral a

mpl

itude

den

sity Peak

coefficient[ - ]

12,38 0,069 3,88

Shape coefficient [ - ] 4,15 0,045 2,66

Median frequency

[ kHz ] 15,80 0,987 7,13

Spec

tral e

nerg

y de

nsity Peak

coefficient[ - ]

15,27 0,191 3,92

Shape coefficient [ - ] 5,74 0,052 4,42

Median frequency

[ kHz ] 7,03 0,627 8,01

11,07

0,78

1,00

1,88

0,79

1,12

1,98

2,53

The hypotheses of the test assume, respectively:H0: 12 = 22 = 32 = 42 = 52 = 62; where is a variance of a given population,

H1 : not all variances of results are equal.

The calculated value of 2 distribution and the value of the significance level = 0.05

readout from the table are presented in Table 5.

Since in each case there is the inequality 2 < , there are no bases for rejecting the

Ho hypothesis which says that all variances of the descriptors determined are equal to one

another. It should be added that the limitations of the paper length do not make it possible to

present all the calculation results tested, i.e. for the remaining three distances between the

transducer and the calibrating head. However, homogeneity tests of many variances were

performed for all the data collected, and as a result of which it was possible to state that there

are no bases for rejecting the H0 hypothesis.

Because the calculations presented in Table 5 can be described by a normal

distribution and their variances are equal, it was possible to carry out a test of variance

analysis for many averages in order to determine the repeatability of the test results.


The test hypotheses assume, respectively:

H0 : m1 = m2 = m3 = m4 = m5; where m is the average from population.

H1 : not all averages are equal to one another

The calculated value of distribution F and the value readout from the table for the

significance level = 0.05 is presented in Table 5, respectively. Since in each case the

inequality F < F takes place, there are no bases for rejecting the Ho hypothesis which says

that all the selected descriptors are equal to one another. This statement proves, at the same

time, that in the setup under study the repeatability of the experiment results takes place with

a 5% error tolerance.

8. SUMMARY

Based on the statistical tests carried out it can be observed that the AE signals

generated by breaking the marking point placed in the Hsu-Nielsen calibrating head are

characteristic of the repeatability for a given distance between the generation place and the

measuring transducer. This ascertainment is true with a 5% error margin. Therefore this

method can be used for calibration of measuring paths applied in the acoustic method of

evaluating the condition of power appliance insulation.

Acoustic emission is one of the non-destructive research methods finding its wider and

wider application in practice. This paper offers another insight into important issues dealing

with generation and propagation of the repeatable model AE pulses. The results of the

analysis of the AE pulses generated in the Hsu-Nielsen system refer to an actual transformer.

This fact is important as the AE method is more and more often used in professional power

engineering for monitoring and diagnosing the condition of insulation in high-power

transformers, high-power measuring transformer, flow insulators, power capacitors and

switching stations with SF6. The AE method provides unique information on the location and

intensity of PDs occurring in these appliances. In the conditions of regular work the PD

measurement by electrical methods is not possible due to a strong interaction of power

disturbances. The spectra of the pulses registered are not always measured during tests

performed in technical conditions. The measurements are often limited to determining basic

descriptors of time change derivatives. The authors of this paper wanted to draw attention to

the fact that a signal occurring in the generation place differs significantly from the pulse

received at the end of the measuring system.


The size of these differences depends on the transition function, which characterizes

each element of the path that a signal has to cover, and on the medium in which it propagates.

The results presented in this paper confirm the thesis that the Hsu-Nielsen source complies

with the conditions of the repeatable AE source, is easy to make and be used in technical

conditions.

ACKOWLEDGEMENTS

The research was carried out within the grant KBN no. 3 T10A 031 27

REFERENCES

[1] Kaiser J.: Messung von Geruschen bei Zugbeanspruchung von metallischen Werkstoffen, Arch. f. Eisenhttenwesen, 25, 43, 1953.

[2] Skubis J.: Emisja akustyczna w badaniach izolacji urzdze elektroenergetycznych, IPPT PAN Warszawa, 1993

[3] Breckenridge F., Tschiegg C., Greenspan M.: Acoustic emission: Some Application of Lambs Problem, J. Acoustic Soc. Vol.57, No. 3, March 1975.

[4] Hsu N. N., Breckenridge F. R..: Characterization and Calibration of Acoustic Emission Sensors, Mat. Evaluation, 39, 60, 1981.

[5] Mc Bribge S.L., Hutchison T.: Absolute Calibration of the Helium Gas Jet Noise Source, Canadian J. of Phis., 56, 504, 1978.

[6] Brel & Kjael Technical Review, No 2, 38-40, 1981.

[7] Norma francuska Badania nieniszczce, sownictwo wykorzystywane w EA, NF - A 09-350.

[8] Skubis J., Lorenc M.: Measurements and analysis of acoustic emission standard impulses generated in Hsu-Nielsen source. Archives of Electrical Engineering Vol. XLVII. No 1 pp. 13-24, Warszawa 1998.

[9] Lorenc M.: Ocena powtarzalnoci sygnaw emisji akustycznej generowanej przez rda wzorcowe testem Chi-kwadrat, KNT Transformatory w eksploatacji; Sieniawa, 2002, pp.143-148

[10] T. Boczar: Identification of a Specific Type of Partial Discharges form Acoustic Emission Frequency Spectra, IEEE Transactions on Dielectrics and Electrical Insulation, Vol. 8, No 4 August 2001, pp. 598-606.

[11] T. Boczar., D. Zmarzy: Application of Wavelet Analysis to Acoustic Emission Pulses Generated by Partial Discharges, IEEE Transactions on Dielectrics and Electrical Insulation, Vol. 11, No 3, June 2004, pp. 433-449.

[12] T. Boczar, S. Wolny: Application of calculus of probability for the randomness evaluation of acoustic emission signals generated by partial discharges, 31st International


Conference Defektoskopie 2001, Czech Society for Nondestructive Testing, Prague, Czech Republic, 2001, pp. 55-62.

[13] T. Boczar, S. Wolny: Application of statistical methods in the analysis of the acoustic emission signals generated by partial discharges, Proceedings Scientific Colloquium on High Voltage Engineering, Slovakia, Kosice, 2002.

[14] R. Gnanadesikan: Methods for statistical data analysis of multivariate observations, John Wiley and Sons, New York, 1997.

[15] M. Hollander, D. A. Wolfe: Nonparametric statistical methods, John Wiley and Sons, New York, 1997.

[16] D. C. Montgomery, G. C. Runger: Applied statistics and probability for engineers, John Wiley and Sons, New York, 1999.

[17] J. Neter, M. H Kutner, C. J. Nachtsheim, W. Wasserman: Applied linear statistical models, Richard D. Irwin, Inc. and Times Mirror Higher Education Group, Inc, Chicago, 1996.

[18] J. Stevens: Applied multivariate statistics for the social sciences, Hillsdale Erlbau, new York, 1986.

[19] S. B. Vardeman: Statistics for engineering problem solving, PWS Publishing Company, Boston, 1994.

[20] R. J. Van Brut: Stochastic properties of partial discharges phenomena, IEEE Transactions on EL, vol. 26, no. 5, 1990, pp. 642-665.

REFERENCES

Documents

DETERMINING THE REPEATABILITY OF ACOUSTIC EMISSION ...mqa.ogpta.polsl.pl/files-articles/25/pdf/177-192_Boczar_Lorenz.pdf · DETERMINING THE REPEATABILITY OF ACOUSTIC EMISSION