Multi-level periodic perturbation signals for system identification in the
frequency domain are defined. The design and properties of these signals, a
nd the output signal processing methods used with them to reduce or elimina
te the effects of noise and nonlinearity, are developed. Both pseudo-random
and multi-frequency perturbation signals are considered, and an applicatio
n in which these signals were used to identify the dynamics of steel strip
in an industrial process is described. It is concluded that the use of thes
e signals leads to considerable benefits, and that they are to be preferred
for identification schemes of this kind. (C) 1999 Elsevier Science Ltd. Al
l rights reserved.