132 自己組織化ニューラルネットワークによる適応的振動制御

概要

論文の詳細を見る
Making use of learning and nonlinear mapping capability, neural network has been applied to engineering problems such as pattern recognition, optimization, identification, and control. Though neural network has a great potential in realization of intelligence, several defects have been pointed out; it takes long time for tuning network parameters and the accuracy of mapping is not always high enough because of local minimum solutions in optimizing procedure. In the present paper, neurons are arranged in two dimensional space and each connections of neurons were created automatically using Cellular Automata. The self-organized neural network was applied to the identification part of an adaptive neural network control system of a flexible structure controlled by a magnetostrictive actuators. It is shown numerically and experimentally that the vibration response at the first and second natural frequency of the structure can be reduced by controlling with the present method.
一般社団法人日本機械学会の論文
2000-09-01