New approach for design and automatic tuning of the Smith predictor controller

Abstract

In this paper, we present a new Smith control design using a deliberate mismatch between the actual process and process model. With this particular representation, every Smith control system has an associated 'Smith compensated' (SC) process that is an interactive function of the actual process and process model. Under this alternative Smith control structure, it is both intuitive and demonstrative by simulation study that the 'best' Smith control design does not necessarily arise from the use of the perfect process model but rather the use of an adequately chosen model that results in the SC process possessing good dynamical properties from the viewpoint of the controller. Thus, the process modeling phase of the Smith control design may be viewed as a process precompensation design, and the final primary controller design will be carried out with respect to the SC process. For ease of practical applications, the entire procedure from modeling to control design may be automated and carried out online, using a new online relay tuning method. Robustness analysis of the new Smith control design is further carried out in the paper by drawing on existing results for single-input- single-output feedback systems. Simulations are provided to illustrate the applicability and effectiveness of both the online autotuning approach and the new Smith control design technique.