Identification of continuous-time delayed systems using an input pulse of arbitrary shape

This article resolves difficulties of continuous identification often encountered in field testing, such as unsteady initial states, unknown disturbances, and noise-corrupted measurements. The unique feature of the proposed method is that it divides the identification problem into two simpler estimation problems based on pulse testing of arbitrary shape. Sequential algorithms are developed to deal with the first estimation problem subjected to the zero portion of the input and then to solve the remaining problem subjected to the arbitrary portion of the input. The method provides a convenient way to fit model predictions to output measurements in the face of unknown initial states and static disturbances. Simulation results demonstrate that the use of the integral filter renders the method rather robust with respect to noise and model structure mismatch.