Nonlinear modeling of laboratory model AMIRA DR300 by feedforward neural network

Abstract

Purpose of this paper is to design a nonlinear feedforward neural network model of a laboratory model AMIRA DR300 (made by AMIRA, Duisburg, Germany). This device consists of two direct-current engines, whose shafts are connected together by a fixed shaft coupling. An output value of the laboratory model is a rotation speed of this shaft coupling and it can be measured either by a tachometer generator, or an incremental position sensor. The first engine was modeling by neural network. The second one is used to apply as a generator of a faulty measured value.