Measurement noise attenuation in modified Smith predictor and automatic offset controllers for integrator plus dead-time system.

The paper compares the recently modified Åström-Smith predictor (ASP) developed for IPDT models with an automatic offset controller (AOC). While an excellent performance can be achieved with ASP in idealized conditions, unacceptable transients with extraordinary high excessive controller effort and a steady-state control error result in the presence of measurement noise. AOC combines a possibly higher-order (HO) stabilizing controller (SC) with compensation of disturbances using a full disturbance observer (DOB). Nominal full DOB includes both the model dead-time and inversion of the integral mode of the model. By increasing the number of output derivatives used in the SC of the AOC, together with increasing the order of the low-pass filter used both in the SC and the DOB, it is possible to significantly increase the speed and robustness of responses in time-delayed processes, together with decreasing the measurement noise impact. The AOC based on the ultralocal IPDT model can be used to replace the higher-order PID in an universal controller for a wide class of processes with dominant first-order dynamics. Significant improvements in measurement noise attenuation can be demonstrated also in application of equivalent low-pass filters to the ASP. But even after such a modification of the noise attenuation, the filtered ASP can exhibit permanent control error or even instability at higher noise amplitudes. Hence, although even the ASP can be used as a universal controller for processes with dominant first-order dynamics, the benefit of its use should always be verified depending on the amplitude of the measurement noise. From this point of view, the use of AOC is simpler and more reliable. Despite the need for appropriate selection of the degree of derivatives used in SC and the tuning of the low-pass filters used. The conclusions of the article are illustrated by simulation experiments of unstable process control and real-time thermal process control.