2000 年 31 巻 4 号 p. 99-106
A hydraulic servo system, which has a quick-response characteristic, high positioning accuracy and high power by itself is often used in airplanes, a construction machinery and the center of vibratory testing machines. Several papers for application of a Neural Network to servo systems have been reported. Most of them were for systems, which have slow speed response characteristics such as pneumatic servo systems. Recently application to improve the characteristics of a hydraulic servo system, a vibratory waveform control system and so forth are studied. In this study, we directly applied a Neural Network to the hydraulic pressure-waveform control system and discussed the basic characteristics such as the convergent problem of the output waveform for reference input signal and the stability problem of the system. The results were as follow :
(1) When the Neural Network is applied to the pressure-waveform control system and learns several times in one sampling period, the theoretical convergent condition equation could be obtained and confirmed by simulation and experiment.
(2) In this control system, there are convergent and divergent domains. We found that stable and unstable areas existed in the convergent domain.