<![CDATA[In the modelling of a system based on a system identification approach, data acquisition is the first procedure that must be carried out. The data acquisition process from a real system typically yields large amounts of data. This may lead to unacceptable computational time during the identification process. Raw data may also suffer severe noise disturbance, especially in the high frequency region. In addition, bias estimation will occur if one only considers direct identification from a closed-loop system. To overcome this problem, in this paper a multivariable frequency sampling filter approach is introduced. Multi-input-multi-output (MIMO) raw data are analyzed in order to obtain only relevant and meaningful parameters that describe the empirical model of the analyzed data. By performing this procedure, compressed, cleaned and unbiased data are produced. The efficacy of the MIMO frequency sampling filters was demonstrated by compressing two sets of data: pH neutralization process data and steam generator plant data. The results show that the amount of raw data was successfully compressed and that the output was ready for identification purposes with less computational time, i.e. they could be further used to develop a model of the system, to conduct time and frequency response analysis, and also for developing a new control system design.]]>
Copyrights © 2013
