(1996) Sergeev, S. A.; Klepikov, V. B.; Mahotilo, K. V.
The paper interprets characteristics of a neural-control-based DC servodrive in terms of the classical theory of automatic control. It also touches on the problem of choosing training patterns to synthesize a nonlinear PID-controller with a desired amplitude-frequency characteristic and analyses the efficiency of using for this purpose input signals in form of a step function and a harmonic one. Synthesis of the neurocontroller has been performed within the framework of a three-layer perceptron. To train it, a genetic algorithm has been developed.
