Enhancing Performance of Level System Modeling with Pseudo-Random Signals

Obtaining mathematical models for typical-level sys-tems still presents significant challenges due to the nonlinearities inherent to these systems. The choice of appropriate identification signals plays a fundamental role in this context since non-stimulated features cannot be adequately represented in the model. This article aims to investigate the modeling of a MonoTank Level System (MTLS) using the Pseudo Random Binary Signal (PRBS) to obtain robust linear models. Additionally, the models will be compared with the classical identification approach based on the reaction curve. The results demonstrated the superiority of the models obtained using PRBS. When compared to classical methods, there was a notable average reduction of 18.18% in Root Mean Squared Error (RMSE). Depending on the classic approach, this reduction could reach up to 35.54%. This improvement in model accuracy suggests that PRBS are more efficient in capturing the nonlinearities inherent to the studied level system and deserve more attention in the process control literature.