'iLPM+IVVF': uma abordagem eficaz para identificar módulos em redes dinâmicas

Abstract

Classical methods for identifying systems embedded in dynamic networks require the estimation of parametric models for all modules that comprise the system, commonly characterized by having multiple inputs and only one output. This process requires the preselection of the order of a parametric model for each system that composes the experiment, which increases computational complexity as the size of the network grows. This article discusses a two-stage semi-parametric approach for identifying local modules in complex dynamic networks, which demonstrate to efficiently estimate module dynamics with reduced computational complexity. The first stage utilizes a non-parametric indirect method, called iLPM, to estimate the frequency response function of the target module. The curve thus estimated is then smoothed through the IVVF parametric estimator. A case study is presented to validate the effectiveness of the proposed approach. The results show that the ’iLPM + IVVF’ method is effective and can be used to identify local modules in complex dynamic networks.