Estudo e comparação de métodos de identificação de sistemas dinâmicos

Abstract

This work presents a comparative study between some identification methods for dynamical systems. For systems characterization, different signals are applied, such as step function, sine and square waves. Four graphical techniques are implemented using the step response: Ziegler and Nichols, Yuwana and Seborg, Sundaresan and Analysis of Transient Response; as result the system transfer function is obtained. By the application of the sine and square waves, the frequency response is obtained. For sine waves the procedure is done frequency-by-frequency. For square waves, in turn, different harmonics are considered at once, as result of the spectral analysis of the square wave and the respective output signal. From the frequency response, which is represented as a Bode diagram, the system transfer function is extracted. All methods are experimentally tested in electrical passive circuits, electronic circuits implemented using operational amplifiers and a DC servo-mechanism. All estimated models are compared to the theoretical ones by using Matlab/Simulink and PSpice software. The methods are also compared to each other. The performance of each technique is evaluated by the Sum of Squares Errors and Multiple Correlation Coefficient.