ニューラルネットワークを用いた時系列の学習および線形近似モデルの導出法

概要

論文の詳細を見る
Generally, real objective systems for controlling are nonlinear dynamic one. On the other hand, usual control theories are linear. Therefore, they are not so suitable for such nonlinear dynamic systems. Recently, neural network (NN) which has an ability to describe nonlinear dynamic characteristics are studied and applied in many technical fields. Proposed here is a control method which is suitable for nonlinear dynamic systems. This method uses real time identification of the transient system by NN and applying an usual linear control theory. These identification and applying are carried on at an interval of each control cycle time. Therefore, it can be applied for strong nonliner dynamic systems on the whole. There are 3 simulation examples applied the method, in the paper, and all show relatively good results.
一般社団法人日本機械学会の論文
1994-07-25