THEORETICAL-EXPERIMENTAL METHOD OF CALCULATION OF WEIGHT COEFFICIENTS OF THE ARTIFICIAL NEURAL-LIKE NETWORK IN DIAGNOSTIC SYSTEMS OF HYDROAGREGATES

2019;
: pp. 5-10
1
Vinnytsya National Technical University, Ukraine
2
Vinnytsya National Technical University, Ukraine
3
Vinnytsya National Technical University, Ukraine

In the article was proposed a new theoretical-experimental method for determining the weight coefficients of an artificial neural-like network. The authors propose using coefficients of correlation between the spatially distributed points of the hydro unit as these coefficients, The article contains theoretical substantiation of the dependence nature of the correction factors on the load of the hydro unit and the water pressure in the reservoir. These theoretical positions are confirmed by experimental research. This research was carried out at the industrial equipment of the Hydro power station. A serious problem in frequency analysis of vibration signals of hydro units consists in the non-periodical character of the signals. Therefore, is proposed an original method for determining the coefficients of correlation between the spatially distributed points of the hydro unit. It involves the prior use of a wavelet transformation. In the course of an experimental research it was found that that the dependence of the coefficients of correlation between the pressures of water in the reservoir at low frequencies is due to the dependence of the lapidary flow from the pressure of water in the reservoir. The reason for the change in the coefficients of inter-correlation in the thirteenth frequency band is the cavitations phenomenon. As a result of the research, it has been proved that in the indicated frequency bands, the coefficients of the correlation of vibration signals in the spatially distributed quasi-symmetric points of the hydro aggregate change significantly when the water pressure in the reservoir changes. This makes it possible to consider them as an additional sign of the presence of the hydrodynamic component of the vibration.

[1]  A.  Shirman,  A.  Solovyov.  Practical  vibration-based  diagnostics  and  monitoring  of  mechanical  equipment condition. Moscow, RF: Mashinostroyenie, 1996. 

[2]  O.  Kolesnitsky,  E.  Gordyshevskaya,  S.  Lukash, “Computer  modeling  of  the  recognition  method  of  multi-sensors signals of gases on the basis of pulsed neural network”, Meas.  and  Comp.  Devices  in  Techn.  Proc.,  Kyiv,  Ukraine:  No. 1, pp. 121–126, 2013.

[3]  P.  Broersen.  Automatic  autocorrelation  and  spectral analysis. London, GB: Springer-Verlag London Limited, 2006.

[4] S. Rao. Vibration of continuous systems. New York, USA: Jon Wiley & Sons, 2007.

[5]  V.  Hraniak,  V.  Kukharchuk,  V.  Kucheruk,  A. Khassenov, “Using instantaneous cross-correlation coefficients of  vibration  signals  for  technical  condition  monitoring  in rotating  electric  power  machines”,  Bull.  Karaganda  Un., Physics Series, No. 1(89), pp. 72–80, 2015.

[6]  A.  Kobzar.  Applied  mathematical  statistics. Moscow, RF: Fizmatlit, 2006.

[7] V. Vladislavlev. Vibration of hydro units of power plants. Moscow, USSR: Energy, 1972.

[8] V. Hraniak, S. Katsyv, V. Kukharchuk, “Results of the dependence analysis of  the mutual correlation coefficients of  hydro  aggregate  vibration  processes  on  its  load”,  Bull. Vinnytsia Polytechn. Inst., Ukraine: No. 4, pp. 7–14, 2018.