COMPARISON OF FREQUENCY RESPONSE AND NEURALNETWORK TECHNIQUES FOR SYSTEM IDENTIFICATION OF ANACTIVELY CONTROLLED STRUCTURE

摘要

System identification methodsare generally used to obtain the dynamic properties of structural systems. The dynamic propertiesare used for various purposes, such as model updating, structural health monitoring, and control synthesis. This paper presents the identificationof an actively controlled structure with an active mass damper based on input-outputrelationships.The input signals include accelerations inthe base of the structure and control force inputs while the output signals are the accelerations of the structure due to the inputs. In this paper,the system identification using frequency response functions iscompared with non-linear relationships obtained by using artificial neuralnetworks (ANN) for bothasingle-input, single-output, and multiple-inputsingle-output (MISO) system. The results indicate that for the MISOstructural system,the ANN technique providesa more accurate identification than identifications obtained with frequency responsemethods.