TAVARES NETO, J. I. H.; http://lattes.cnpq.br/4287246487119988; TAVARES NETO, Júlio Inácio Holanda.
Résumé:
This work intends to be a contribution to the field of system identification, which is one of the
methods available for the construction of models and has been gradually becoming the
method of choice among the different system modeling alternatives. The system identification
method addressed in this study was developed using a signal known as a Rseudorandom
Binary Sequence (PRBS). This class of excitation signal has been increasingly used due to its
interesting excitation characteristics and fewer generation and application difficulties. One
characteristic that has led to the acceptance of PRBS in different industries is the small degree
of disturbance to plant operations, which is why it is commonly called a "friendly signal".
However, the currently proposed method for PRBS signals applied to industrial processes
does not consider the dynamics of the input variable. The present work offers an alternative
methodology for the projection of this type of signal that includes the influence of the input
variable in the modeling process using the system identification method. Two case studies
were carried out in order to assess and compare the proposed methodology with the currently
available methodology for the projection of this type of excitation signal. In the first, the
influence of the final control element in the identified model was assessed, revealing that the
signal applied using the proposed methodology was able to generate data that lead to a more
representative model of the system under study. In the second case study, the influence of the
input variable was simulated, revealing that the proposed methodology is more adequate for
exciting the system in the conditions under which the system was simulated. The results
demonstrate that the novel methodology proposed for the projection of PRBS signals, which
considers the dynamics of the input variable, has a predictive capacity for generating a better
process model that the methodology currently adopted in the literature