NEW DEVELOPMENTS IN PID CONTROL AND AUTO-TUNING

Abstract

Manufacture of PID controllers has undergone many changes over the last decades. PID controllers have evolved from a pneumatic setup in the 1940's to electrical ones and now modern microprocessors-based devices. While the implementation has changed tremendously, the basic functions of controllers, the PID algorithm, has not and probably will never changed. For the user, the job of tuning the controller has also been essentially the same throughout the decades, although new PID designs have been continuously developed to cater to different sets of control requirements. However, new possibilities and functionalities have become possible with a micro­ processor driven PID controller. Process controllers today often contain much more than just the basic PID algorithm. Fault diagnosis, alarm handling, signals scaling, choice of type of output signal, filtering, simple logical and arithmetic operations are becoming common functions to be expected in modern PID controllers. Physical size of the controller has shrunk compared to the analog predecessors and yet the functions and performance have greatly increased. Furthermore, riding on the advances in adaptive control and techniques, the modern PID controllers are becoming intelligent. Many high-end controllers are appearing in the market equipped with auto-tuning and self-tuning features. No longer is tedious manual tuning an in­ evitable part of process control. The role of operators in PID tuning has been very much reduced to simple specifications and decision. In this thesis, some latest developments on PID controller tuning are presented. First, we consider for the first time the application of relay feedback to identification of the PID configuration and parameters, thus seeking to expand the application domain of current autotuning approaches to general and non-standard PID configurations. The method is focused on the two main classes of PID configuration, the parallel and the series type, even though the approaches developed will be equally applicable to other sub-classes if the user chooses to include them in the identification experiment. Then a robust on-line relay automatic tuning method is developed for general control applications. The proposed method is effective against many of the common constraints of the basic relay autotuning method, such as load disturbance, time-varying systems, run-away systems and unstable systems. It also retains the simplicity of the original method. Finally, for the first time, the detailed generalised predictive control (GPC) solution is developed for a linear system with time delay. The useful properties of GPC control are combined with the simplicity of PI structure by applying the GPC solution to tune PI control gains for first-order time-delay systems or those which may be adequately modelled as such. It is proven that the closed-loop system with input delay will be stable under the proposed control law. It provides significant improvement over the existing PID tuning methods. Simulation examples and a real-time experiment will illustrate the practical appeal of the proposed method.