Identification of a Nonlinear System with Integral Characteristic by DNN and Performance Verification

Abstract

In this paper, we consider nonlinear system identification when a plant contains an integral characteristic. When the plant includes an integral characteristic, applying the DNN directly to nonlinear system identification may result in large differences between the response of the actual equipment and the response of the DNN model. In order to reduce response errors, we propose a method that adds a structure constraint represnting the integral characteristics to the DNN model. The effectiveness of the proposed method is verified using input-output data from a nonlinear simulation of a pneumatic vibration isolator that includes integral characteristics.