A influência do ruído no controle e na modelagem de sistemas dinâmicos lineares discretos

Abstract

This study investigates how different types of noise influence the modeling of discrete linear dynamic systems. The research comprises three simulated experiments for controlling discrete linear systems under different noise conditions: no noise, white noise, pink noise, and brown noise. Additionally, real-world experiments are proposed through data acquisition from a digital accelerometer, the LIS3DH model, using an Arduino Nano. ARIMA models are employed to characterize the accelerometer’s behavior, with sampling intervals of 100 ms, 500 ms, and 1000 ms. MATLAB software is utilized for simulation, analysis, and model derivation. The results from the simulated experiments indicate that the Mean Squared Error (MSE) is lower for white noise compared to the other noise types. The simulations also demonstrate that adding noise to the system output increases the error and induces undesirable oscillations. The experimental results with the accelerometer reveal that the models were able to represent up to 30% of the measured signal. The accuracy of the models was constrained by uncertainties associated with data acquisition and variations in system behavior, leading to a satisfactory and functional approximation of the system’s modeling.