issn: 0975-766x coden: ijptfi available online through ......are riveted connections of compressor...

Emil Rinatovich Saifullin*et al. /International Journal of Pharmacy & Technology

IJPT|Dec-2016 | Vol. 8 | Issue No.4 | 24343-24356 Page 24343

ISSN: 0975-766X

CODEN: IJPTFI

Available Online through Research Article

www.ijptonline.com

CONTROL FOR RIVETED CONNECTIONS BY FREE OSCILLATION TECHNIQUE.

HARDWARE AND SOFTWARE SYSTEM. EVALUATION OF MEASUREMENT ERROR Emil Rinatovich Saifullin

1, Yury Vitalevich Vankov

2, Andrey Vladimirovich Busarov

3

1Kazan Federal University, Kazan, Russian Federation.

2 Kazan State Power Engineering University, Kazan, Russian Federation.

3 Limited Liability Company Center of Industrial Safety Expertise, Kazan, Russian Federation.

Email: [email protected]

Received on 14-08-2016 Accepted on 20-09-2016

Abstract

The paper presents the results of studies on the development of techniques and diagnostic suite for instrumentation

control for quality of riveted joints by the free oscillation technique in automatic mode. The suite consists of a recording

system and a signal processing system. The registration system includes a device for positioning the inspected articles.

Excitation of oscillations in an inspected article is carried out with a calibrated strike. Signal processing system is

composed of a hexadecimal analog-to-digit converter and a digital computer. Acoustic signals of vane vibrations are

recorded on the hard disk, and their processing is performed with a software package. The description of DetectFault

software created in LabVIEW environment. The software controls the movement of a guide wheel and a striker bar,

ensures registration and transformation of signals from analog to digital form with the set number of samples and

sampling interval value. The control algorithm is described. Acceptance or rejection of a riveted connection is made

depending on the results of comparison of a reference and the current spectrum for the Spearman correlation coefficient

and the rejectable level (higher - passed, below - rejected). To classify a riveted connection as "fit" or "defective", an

approach should be used that is specific to procedures on rejection of anomalies by a confidence interval. A robust

weighting algorithm is used when forming the reference spectrum. The error estimation of the measurement results was

carried out. The results of reproducibility of the data during the monitoring of riveted joints.

Keywords: Riveted connections, non-destructive testing, free oscillation technique, spectra comparison, decision rule

Introduction: Experience in operation of gas compressor units showed that defects of riveted connections in guide cases

significantly reduce their reliability. Development of a method for instrumentation control of riveted connections will


IJPT|Dec-2016 | Vol. 8 | Issue No.4 | 24343-24356 4344

make it possible to solve the problem of sound prediction of no-failure service life for equipment and its resource

assignment in accordance with actual condition of guide vanes.

1. Description of the automated diagnostic suite

The basis of the automated diagnostic suite developed for control of riveted connections is free oscillation technique

which relies on registration of free oscillations parameters excited in a unit under test [1-6]. Units tested with this suite

are riveted connections of compressor vane rings guide wheels for 4 ÷ 8 stages of a gas turbine engines NK-16-18ST and

NK-16ST. The block schematic diagram of the suite is shown in Figure 1. The suite consists of a signal recording system

and a signal processing system.

Fig. 1 Block schematic diagram of the suite.

The recording system includes a device for positioning the inspected articles. Excitation of oscillations in an inspected

article is performed by a calibrated strike. Strike force depends on the weight and configuration of an inspected article

and can be adjusted during the testing process. The signal processing system consists of hexadecimal analog-digital

converter (ADC) and a computer. Acoustic signals caused by vane vibrations are recorded on the hard drive, and then

processed by the software package. All devices and mechanisms of the suite which appearance is shown in Fig. 2, are

located on the vibration-isolated table.

Fig. 2 Automated diagnostic suite.



The suite consists of:

The positioning device consisting of a guide case mounting assembly with a fixed value of coupling force along the

perimeter of the guide wheel and a centering device, and also a mechanism for rotation (stepping movement) of the guide

case with reversible electric motor connected through a gearbox to the hub (Fig. 3).

Shock vibration excitation device which consists of an electromagnet and a striker bar which is located on a stand

with a counterweight (Fig.4).

Receiving acoustic device consisting of a microphone placed in an acoustic pipe disposed on a separate stand (Fig.5).

Control unit - a controller that forms the motor and striker bar control signals (Fig.6).

Computer which performs the processing of electrical signals and their analysis, recording the inspection results,

setting the control parameters for the control unit (Fig.7).

Fig. 3 - Positioning device

1 - motor, 2 - hub, 3 - receiver

Fig. 4 - Shock vibration excitation device.

1 - Counterweight, 2 - stand, 3 - electromagnet, 4 - striker bar, 5 - guide case.



Fig. 5 - Receiving acoustic device

1 - acoustic pipe 2 - stand.

Fig. 6 - Control unit.

Fig. 7 - Schematic diagram of the control unit.



Fig. 8 - Soundproofed room.

The suite is located in a soundproofed room (Fig. 8).

The suite operates as follows. After installation of the guide wheel on the measuring position a signal is supplied to the

controller which activates the striker bar. A series of strikes with normalized force is applied to a vane shroud of the

riveted connection of the tested guide case. Microphone detects vibrations of the article, converts them into electrical

signals which are transmitted to the audio input of the computer. Analog electric signals using the 16-bit ADC is

converted in the computer to digital form, recorded on the hard disk and subsequently processed by the specialized

program DetectFault [7].

Upon completion of a series of strikes a signal should be applied to the controller and the guide wheel rotates to the next

riveting. At this time, the program analyzes the recorded signals and outputs the result on the condition of the riveted

connection (defective or defect-free). Then the cycle is repeated until all riveted connections of the guide wheel will be

subjected to test. After that results should be logged

2. Description of the software

DetectFault program developed in LabVIEW environment [8, 9] controls operation of the suite and consists of the

following modules for:

Acoustic signals recording;

Formation of reference spectra;



Comparison with the reference spectra.

The folder "Diagnostics" is created on computer hard disk upon running the program for first time. Then in the folder

"Diagnostics" there should be created the folders for the following: acoustic signals recording program, reference spectra

formation program, program for comparison with the reference spectra; the program creates corresponding folders

"Reference", "Processing" and "Comparison". The reference spectra is written in the folder "Reference". Folder

"Processing" is the storage of the recorded signals. Reports with the results of comparing recorded signals with the

reference signals should be written in the folder "Comparison".

Program of recording acoustic signals controls the movement of the striker bar and the guide wheel, provides signals

recording, their conversion from analog to digital form with a predetermined number of samples and sampling interval

value.

On the front panel of the program interface (Fig. 9), the user sets the number and the cipher code of the tested guide

wheel, the rivet number with which testing starts, and the number of rivets on the guide wheel.

In the tab, the reference spectrum file is additionally selected, the last used file is read from the registry by default, the

time for the guide wheel movement from a rivet to the next rivet is set, the function to print the report is selected, and the

siren signal is set. If in the process of testing a defective rivet is detected, a siren sounds.

Figure 9. The front panel of the program interface for recording acoustic signals.

The program works as follows:

a report file is generated (the file name comprises the test starting time);



a programmable relay starts, and the striker bar performs a series of strikes on the tested riveting;

the microphone records acoustic feedbacks of strikes (the file name comprises the guide wheel cipher

code and the vane number, file format is wav.);

a signal is issued to rotate the guide wheel to the next rivet;

the cycle is repeated as many times as there are rivets on the guide wheel.

While the guide wheel is rotating to the next rivet, the program on comparison of spectra with the reference works:

FFT algorithm generates a spectrum for each signal;

Spearman correlation coefficient is calculated for the current and the reference spectra;

coefficients obtained are compared with the confidence interval boundary, and a conclusion is made on

fitness of the part;

the information about conditions of the tested rivet is appended to the report;

upon completion of the work, the end of the testing time should be registered in the report file, and the

report should be printed out.

The program on generation of the reference spectra is intended for generation of the reference and the confidence

interval. The task is that the generated spectrum should be the most "typical" for those set of spectra for which it is

generated, i.e. it has to include the most common characteristics of the spectra, and exclude the individual characteristics

of each spectrum. In the front panel of the program interface (Figure 10) the user sets the number of count from which a

signal will be processed (zero count by default), the sample length of the signal to obtain the spectrum (2048 counts by

default), and the frequency intervals within which spectra processing will be carried out .Additionally there is the

possibility of spectrum normalization. Normalization refers to the division of each spectral component to the maximum

amplitude of the spectrum. As a result of this operation, affect of the strike force instability on the result of the spectra

comparison will be eliminated.

The program of reference spectra generation works as follows: the user selects signals for which the reference spectrum

will be generated; a reference spectrum is formed from the selected signals; signal spectra are compared with the

reference spectrum with the use of the Spearman correlation coefficient for the values of the coefficients; confidence

intervals with a predetermined significance level will be formed. The reference spectrum, calculated coefficients, the



boundaries of the confidence intervals are recorded in a .txt file. The file name is the date and time of creation of the

reference spectrum.

Fig. 10.The appearance of the front panel of the program of formation of reference spectra.

When forming the reference spectrum, the robust (error-correcting) weighing algorithm is used that allows spectral

components into three categories to separate: feasible data, area in doubt, certainly obvious outliers (distant area) [10-

14]. Observations from the area of feasible data are considered fit, observations from the area in doubt are to be assigned

to having reduced weight which is the smaller, the more distant are the data from the area of feasible data; observations

from the distant area are considered as overshoots, and zero weight is given to them, thus they are excluded from the

assessment. The program for spectra comparison with the reference (Fig. 11) is designed to assess the differences

between each of the original spectrum of signals recorded from the reference one [15].

Fig. 11 Front panel view of the program interface for spectra comparison with the reference.



The front panel of the program interface allows the user to select the reference spectrum (set to the last used reference by

default); appearance of displaying the coefficients (one or group for each article), an additional flag is set to print a

report. The program works as follows: read the reference file, read signals, a spectrum is generated for each signal; for

each spectrum Spearman correlation coefficients are calculated; the obtained coefficients related to one riveted

connection are averaged; averaged coefficients are compared with the boundary of the confidence interval, an opinion on

fitness of the riveted joint is issued, and a report is generated.

Acceptance or rejection of a riveted connection is made depending on the results of comparison of the reference and the

current spectra by the Spearman correlation coefficient value and the rejecting level (higher - "passed", lower -

"rejected"). Spearman's correlation coefficient r is calculated by the formula

𝑟 = 1 −6

𝑛(𝑛2 − 1) (𝑟𝑎𝑛𝑘 𝑎𝑖 − 𝑟𝑎𝑛𝑘 𝑎𝑠𝑖)

2,

𝑛

𝑖=1

where n - number of comparable frequencies in the spectrum; rank ai - the amplitude a i rank in the variational series of

amplitudes of the checked spectrum (place number which this amplitude takes between all the amplitudes of the

spectrum in ascending order); rank asi - asi amplitude rank for reference spectrum.

For the classification of riveted connections as "fit" or "defective" (Fig. 12) the approach is used that is specific to

procedures on rejection of anomalies.

Fig. 12. The principle of building the decision rule "fit-defective".

r - Spearman correlation coefficient, N - number of riveted connections.



Analysis algorithm interprets the set of calculated Spearman correlation coefficient values as a plurality of measured

values of an abstract parameter and applies to this set of values the following procedure: position estimate p is

calculated; scatter estimate S is calculated as the median of absolute deviations with respect to position estimate; for a

given level of significance a confidence interval is built as follows

𝑝 ± 𝑆𝑡 1 −𝛼

2,𝑚 − 2 ,

Where 𝑡 𝛼,𝑚 - 𝛼 - quantile of the t-distribution with m degrees of freedom.

Error estimation of the diagnostic suite was carried out by the probabilistic and statistical method providing for error

calculation with the use of characteristics of errors distribution laws of measuring instruments which compose the

system.

3. Measurement result error estimation: Estimation of measurement results is carried out experimentally on a riveted

connection of the guide case under study. The arithmetic mean value of the fifteen values of Spearman correlation

coefficient (r) were obtained:

𝑟𝑐 = 𝑟𝑖

𝑛𝑖=1

𝑛

As a result of observations, the average Spearman correlation coefficient was

r a = 0.6974

The standard deviation of the measurement results:

𝑆 𝑟𝑐 = (𝑟𝑖 − 𝑟𝑐)2𝑛

𝑖=1

𝑛(𝑛 − 1)

Given rc , we obtain, respectively, S (r c) = 0.004652.

Confidence limits (excluding the sign) of the random measurement result error:

ε = tS (rc), where t - Student's ratio which is equal to 2.086 for P = 0,95.

Accordingly, εc = 0.009705.

Confidence boundaries of the residual systematic measurement result error:

Θ = 𝑘 Θ𝑖2

𝑚

𝑖=1



where k = 1,1 at P = 0,95, Θi - boundary of i-th residual systematic error (excluding the sign).

The following formula holds for the experimental installation used for measurement of free oscillation parameters:

Θс = 𝑘 Θ𝑚𝑖𝑐2 + Θ𝑎𝑚𝑝

2 + Θ𝐴𝐷𝐶2

Where Θmic, Θamp, and ΘADC are boundaries of i-th residual systematic error of a measurement microphone, amplifier, and

ADC, respectively, which are as follows:

Θmic = ± 0.028, Θamp = ± 0.012, Θ ADC = ± 0.0075 V.

From there, we have: Θ c = 0.03451.

Check inequality: 0,8 ≤ Θ

𝑆 (𝑟𝑐)≤ 8. For Θс the ratio

Θ

𝑆 (𝑟𝑐)= 7.417532, therefore, the absolute error in both cases should be

calculated as follows:

Δ = KS Σ,

where K - coefficient which depends on the ratio of random to residual systematic errors,

SΣ - estimation of total mean square deviation of the measurement result which is determined by the formulas:

𝐾 =𝜀 + Θ

𝑆2 𝑟𝑐 + Θ𝑖

2

3𝑚𝑖=1

; 𝑆 =

Θ𝑖2

3+ 𝑆2(𝑟𝑖)

𝑚

𝑖=1

Taking into account the values obtained for the automated diagnostic suite in order to test a riveted connection of GTU

vanes Δ = ± 0.045596 with confidence probability P = 0.95 while the average measurement results of Spearman

correlation coefficient rc = 0.6974.

4. Estimation of experimental data reproducibility

High requirements to the reliability of the experimental studies and requirement to increasing the sensitivity of the

considered acoustic control methods impose strict limitations to the experimental conditions such as:

the need for uniform placement of the tested guide case on the positioning device,

an impact to excite natural oscillations should be applied with a constant force strictly in a certain point of

the vane shroud

The measurement microphone must be placed at the same distance from the tested riveted composition

that should not be changed during the experimental measurements.



Compliance with these requirements is implemented in the design features of the experimental facilities and the use of

special equipment to control the conditions of the experiments. In the experiments, control for parameter constancy is

performed by multiple repetition of acoustic characteristics measurements for a same vane with the same measurement

conditions. When studying acoustic characteristics of free oscillations in the vane No. 5, 15-fold measurement of its

natural oscillations parameters was held. The vane was exposed to impacts, and amplitude-frequency response

characteristics have been measured by the microphone located in the acoustic tube. After each measurement the guide

case was removed from its place and fit again. The remaining experimental conditions were unchanged.

Results of comparison of the Spearman correlation coefficient values are presented in Table 1.

Table 1.

Result

No. 1 2 3 4 5 6 7 8

r 0.721 0.715 0.701 0.655 0.679 0.688 0.699 0.681

Result

No. 9 10 11 12 13 14 15

r 0.721 0.718 0.688 0.707 0.696 0.697 0.695

The Spearman correlation coefficient used for comparison has showed that its values are within the range of the

confidence probability P = 0.95. Thus, the system of measuring and processing results provides reproducibility of the

measured values.

Results: The method of riveted connections instrumental control is developed. The software DetectFault is developed.

The software controls the movement of a striker bar and guide ring, ensures the recording, transformation of signals from

analog to digital with desired number of samples and sampling interval value.

Conclusions: Acceptance or rejection of riveted connections are made depending on the results of the comparison of the

reference and a current spectra by the values of Spearman correlation coefficient and the rejectable level. Taking into

account the values obtained for the automated diagnostic suite to test riveted connections of GTU vanes Δc = ± 0.045596

with confidence probability P = 0.95 while the average measurement results for Spearman correlation coefficient rc =

0.6974.

Conclusion

As a result of the works executed, we have developed the automated diagnostic suite to test riveted connections of vanes

in guide cases of gas turbine engines, based on the free oscillation method. Detect Fault program was created to control



suite operation, signal recording and analysis, implementation of an algorithm for defects detection in riveted joints by

parameters of their vibro-acoustic signal in LabVIEW environment.

Acknowledgements

The work is performed according to the Russian Government Program of Competitive Growth of Kazan Federal

University.

References

1. Yu.V. Vankov, Influence of defects on natural oscillation frequencies of gas turbine engine blades, Russian

Aeronautics. 3 (2003) 69-71 (in Russ.)

2. E.V. Belov, Yu.V. Vankov, I.V. Ivshin, A.V. Kochergin, D.N. Pervukhin, A.P. Tunakov, Diagnostics of Gas-

Turbine Engines by Acoustic Sensing of Their Components, Acoustical Physics.42(1)(1996) 14-17.

3. Yu.V. Vankov, R.B. Kazakov, D.N. Pervukhin, Diagnostic complex for detection of defects of axes railway cars in

parameters of fluctuations, Industrial Laboratory. 70(3) (2004) 34-38 (in Russ.).

4. Yu.V. Vankov, The acoustic spectral defectoscope for detection of defects of composite materials (Part 1), Devices

and Control Systems.1 (2004) 45-53 (in Russ.).

5. Yu.V. Vankov, The acoustic spectral defectoscope for detection of defects of composite materials (Part 1), Devices

and Control Systems. 2 (2004) 53-58 (in Russ.).

6. Yu.V. Vankov, Sh.G. Ziganshin, E.V. Izmailova, A.R. Zagretdinov, The effect of the size and position of transverse

cracks in the natural frequencies of the rods, IOP Conference Series: Materials Science and

Engineering.86(1)(2015)012036.

7. Certificate of state registration of the computer program № 2008610104 DetectFault / Authors and copyright

holders: Akutin M.V., Busarov A.V. Vankov Yu.V., Ziganshin Sh.G. Registered in the Register of the computer

programs, January 9, 2008.

8. Peter Blum. LabVIEW. Programming Style. LabVIEW: Style Book, Series: Complete guide on LabVIEW: DMK

Press, 2008, 400 p.

9. V.C. Batovrin, A.S. Bessonov, V.V. Moshkin. LabVIEW: Workshop on electronics and microprocessor technology.

M: DMK Press, 2005



10. Daniel K. The use of statistics in the industrial experiment. M .: Mir, 1979, 299 p.

11. N. Johnson, Lyon Ph. Statistics and experimental design in engineering and science. -M .: World, Vol.1, 1980, - 610

p., Vol.2. 1981 - 520 p.

12. F. Hampel, Ronchetti E., P. Routh, V. Shtael. Robustness in statistics. Approach based on the influence function. -

M .: Mir, 1989.- 512 p.

13. P. Huber. Robustness in statistics. - M .: Mir, 1984.-304 p.

14. Holender M. Wolfe D. Non-parametric statistical methods. - M .: Finance and Statistics, 1983 - 518 p.

15. Bendat J., Peirsol A. The use of correlation and spectral analysis. - M .: Mir, 1983, 312 p.

issn: 0975-766x coden: ijptfi available online through ......are riveted connections of compressor...

Documents