Author(s) :

Yusuf, A.S., Mohammad, A., Hamisu, A.A., Ahmad, K..

Volume/Issue :

Abstract :

This paper presents a study of the effect of different identification signals on the richness of identification data for illconditioned processes. Previous works have suggested that the use of correlated test signals result in identification that is not persistently exciting. However, there have been suggestions that for ill-conditioned systems, the use of correlated signals gives better excitation. This paper aims to investigate the effect of the use of correlated signals for the excitation of a nonlinear ill-conditioned process. To achieve this, three different test signal pairs were used to excite a non-linear model of the high purity binary distillation column. The three tests were carried out using an uncorrelated signal pair, a partly correlated signal pair and a signal pair formed by the summation of an uncorrelated signal with a correlated signal. The effects of these tests were compared using plots of output directions. The use of uncorrelated signals resulted in a maximum range of 0.001 in the low gain direction. The use of two specially designed partly correlated signals resulted in maximum ranges of 0.005 and 0.01 giving improvements of 400% and 900% respectively. Therefore, significant benefits in terms of excitation were achieved by using a partly correlated and a summation of correlated and uncorrelated signals. This means, models developed from the specially designed partly correlated signals would capture more relevant information in the low gain direction

No. of Downloads :

6