Результати аналізу залежності коефіцієнтів взаємокореляції вібраційних процесів гідроагрегату від його навантаження

Abstract

There has been investigated the dependence of cross-correlation coefficients, which are used as weight coefficients of artificial neurosimilar networks, on hydrounit power.

There has been carried out the analysis of results of a correlation method of definition of weight factors artificial neurosimilar networks intended for diagnosing of defects hydrounits.

The necessity of application of such network is substantiated by exclusive complexity of the hydrounit as dynamical hydroelectromechanical system and practical impossibility of the mathematical description of dependence vibrating signals on all factors which cause vibration. Therefore it makes sense to consider the hydrounit as «a black box», that is to model not its structure, and external functioning. The entrance information for artificial neurosimilar networks are peak-time-and-frequency spectrum (PTFS) of vibrating signals.

Weight factors are for each strip of frequencies PTFS and define importance of the account of wavelet-coefficients of strips of frequencies vibrating signals in level of reliability of every neuron.

There has been shown the efficiency of definition of weight factors as cross-correlation coefficients of strips of frequencies PTFS of vibrating signals, simultaneously received from different sensor controls. For this purpose it is possible to present the hydrounit simplified as rather stationary distributed quasilinear indissoluble elastic system with variables in rigidity. Another feature of hydroelectric generator will be operating on it spatially set in space factors of uncompensated mechanical forces of different nature, amplitude and vectorial direction that arbitrarily change in time. In connection with stochastic character of excitant uncompensated forces a hydroelectric generator can be considered the stochastic system; therefore there are intercorrelation connections between the reactions of vibrosignal in the different points of the prospected electric machine. It is possible fully adequate to count a hypothesis that application of coefficients of intercorrelation of vibrosignals between set in space knots of hydroelectric generator together with their instantaneous amplitudes will allow to get the array of data, containing information not only about amplitude of excitant influences, but also their spatial localization, that can be with large probability related to reasons of their origin.