Input PRBS design for identification of multivariable systems.
(January 2018)Abstract
This paper presents a systematic procedure for designing the input signals to identify multivariable processes. The procedure is based on time domain specifications and can be applied to multivariable processes with m-outputs and ninputs, which can be operating in closed-loop. The design of the input signals, which are pseudo random binary sequences, are based on the old information about the process model and the
controller, together with the measures of the input and output variances of the process. The method proposes excitation in the frequency interval where the model needs to be accurate for robust feedback control. The method is illustrated using the Wood & Berry distillation column model, which is a 2-inputs-2- outputs benchmark in process control.