Синтез електромеханічних систем з нейронною мережею та фрикційним навантаженням

Abstract

Thesis for the degree of candidate of technical sciences in the specialty 05.09.03 "Electrotechnical complexes and systems" – National Technical University "Kharkov Polytechnic Institute", Ministry of Education and Science of Ukraine, Kharkov, 2019. The thesis is devoted to the synthesis of electromechanical systems with a per-ceptron neural network and frictional load, which is nonlinear in real electric drives and materialize in electromechanical systems in the form of friction self-oscillations. Friction self-oscillations is a special electric drives operation mode, adversely affecting its functioning. There are various ways to eliminate the friction self-oscillations in the electric drive. A new technique for eliminating of frictional self-oscillations and electromechanical systems dynamic properties improving by using neuro-regulators, which are based on multi-layer headon neural perceptron network, synthesized by the genetic algorithm method, is proposed in the thesis. The choice of structure, neural network teaching method, and the choice of the automated control system feedbacks is motivated. Mathematical models of one-mass and two-mass electromechanical systems with a neural network controller in dimensionless generalized parameters, and also models of electric drives taking into account the load mechanical characteristic nonlinearity are obtained. The control object parameters influence analysis on the dynamic modes of the synthesized one- and two-mass electromechanical systems with a neural network is carried out. A new criterion for neural networks teaching has been proposed and a technique for the electromechanical systems with nonlinear friction neural network regulators synthesis has been developed. The developed technique was tested by synthesis of various DC and AC electric drives of specific machines and mechanisms using the corresponding real data. The examples cited in the thesis are related to the researches according to the plans of the Ministry of Education and Science of Ukraine and the initiative themes requested by enterprises. The efficiency of neural networks using for a wide variety of electric drives of machines and mechanisms is shown.