Transient at Finite Switching Time: Monitoring by Neural Net.

Electromechanical switching equipment has a finite response time of up to several milliseconds. The switching process is intermittent, caused by the bouncing of mechanical contacts and the formation of an electrical arc. The transient process is accompanied by high-frequency current and voltage oscillations, the phase, amplitude, and spectral composition of which largely have a random character dependent on initial conditions. The article presents the results of a laboratory analysis of signal oscillations during the switching on and off of an asynchronous single-phase motor. It is shown that the structure of high-frequency oscillations is determined by both parasitic parameters of the switching apparatus and electrical circuits in the external circuit. A qualitative analysis of the signal structure based on analytical expressions is provided, demonstrating the possibility of controlling the time intervals between current and voltage surges using neural-network algorithms. An algorithm for training a neural network based on simple equations and the use of a basic perceptron for this purpose is described. The ability to identify time intervals corresponding to contact bouncing in an automated mode with a maximum delay of tens of microseconds is demonstrated. The results can be used in testing electromechanical switching equipment. [ABSTRACT FROM AUTHOR]