Correlation Method For Calculation Of Weight Coefficients Of Artificial Neural-Like Networking Hydraulic Units' Diagnostic Systems

The paper proposes a new method for calculating the weight coefficients of an artificial neural network in the systems of technical diagnostics of hydro aggregates, in which it is proposed to use the coefficients of correlation between vibration signals in spatially distributed points of a hydro aggregate. A mathematical model and algorithm for calculation of weight coefficients of an artificial neural network are developed. The expediency of use of wavelet transformation of time realizations of a vibration signal is shown, as a result of which the received vibration signal is divided into amplitude-frequency-time spectrum, which leads to increase its informativeness. Experimentally confirmed the presence of strong inter-correlation links between spatially distributed points of the hydro aggregate and their dependence on the nature and place of application of disturbing forces. The dependence of the correlation coefficients on the load of the hydro aggregate and the water pressure in the reservoir is established. The obtained results can be considered as an experimental confirmation of the expediency of using the proposed method for calculating the weight coefficients of an artificial neural network.