Excitation signals to state-space realization and linear-regression identification methods applied to a MIMO system

This work presents a new approach in the comparative analysis of three excitation signals applied in a system identification process. The excitation signals employed are Pseudo-Random Binary Sequence (PRBS), Sine Sweep (CHIRP), and Gaussian White Noise (GWN). Moreover, with comparative purposes, two algorithms are used to identify a Multiple-Input-Multiple-Output (MIMO) system of 100 degrees of freedom: a State-space identification method based on the traditional Eigensystem Realization Algorithm (ERA) and the linear-regression method ARMAX. The MIMO system is submitted to five levels of additive measurement noise: 1%, 1.5%, 2%, 2.5%, and 3%. The results are statistically analysed composed by 100 simulations for each set of algorithm parameters. The simulations show that, apart from the GWN, well-known as an appropriated excitation signal for application in systems with the characteristics portrayed in the present work, the PRBS can be another feasible signal to identify the MIMO system simulated accordingly with the identification method chosen.