Polynomial Models Identification Using Real Data Acquisition Applied to Didactic System

Models of real systems are of fundamental importance for its analysis, making it possible to simulate or predict its behavior. Additionally, advanced techniques for controller design, optimization, monitoring, fault detection and diagnosis components are also based on process models. One of the most used techniques to model a system is by identification. System identification or process identification is the field of mathematical modeling of systems, in which the parameters are obtained from test or experimental data. Given the importance of obtaining a model able to represent the dynamics of real processes, we developed a software that aggregates identification algorithms using Least squares (LS), Least squares Extended (ELS), Generalized Least Squares (GLS) Recursive least squares with Compensator Polarization (BCRLS). This identification package is used in this paper to identify an educational level plant. Its actual data was inserted in the package and thus, results from different identification techniques implemented in the algorithm were compared. All steps necessary to carry out the identification and analysis of the autocorrelation of the output data for the definition of the sampling period, the design of excitation signals and data collection were taken in consideration. The conclusions reached are that the software provides consistency and the implemented algorithms return a model capable of representing the linear part of the system´s dynamics.