PID CONTROL : TUNING AND ROBUSTNESS

Abstract

The Proportional-Integral-Derivative (PID) controllers are widely used in the process industry and this will continue to be the case in the future owing to their combination of simplicity and familiarity. However, many PID control loops in industries are found to be performing poorly mainly because of bad tuning practice and equipment problems. In view of this fact, this thesis has developed new insights and results on tuning and robustness of PID control. This work, we hope, will be a small but significant step in the attempt to improve PID control in industries. Firstly, this thesis presents a robustness analysis of well-known PID design methods. An important requirement of a feedback control system is its robustness or ability to remain stable despite model inaccuracies and process parameter variations. A control loop that lacks it cannot be left closed and therefore cannot benefit from the performance expected of the controller. Many well-known PID tuning formulas have been derived based on a first order plus deadtime model to satisfy certain time domain performance criteria. However, the effect of these design methods on the robustness of PID controller is not well understood. This thesis gives new insights on this work. Secondly, this thesis develops novel auto-tuning methods based on relay feedback. Automatic tuning is useful in reducing system start-up time and in tightening process control through regular retunings. This thesis improves on the conventional relay feedback auto-tuning method through the use of a relay with DC bias to estimate two points on the Nyquist curve at no extra tuning time and without any separate experiments. This thesis also develops a novel relay feedback auto-tuning method for processes with integration. This technique has the advantage that it can shorten the tuning time by as much as 100%. A novel relay feedback auto-tuning method is also developed for cascade PID controllers. This technique has the advantage that two cascaded PID controllers can be tuned automatically without any separate experiments. Thirdly, this thesis develops a simple and easily implemented analytical design method for multiloop PID controllers. Multiloop PID control is one of the most common control schemes for multi variable processes in the process industries. The controllers are also difficult to tune because of interactions between control loops. The proposed design method is based on a novel idea of shaping the Gershgorin bands to pass through two user specified points for desired robustness and performance. The design method can be easily combined with existing on-line process identification techniques to implement self-timing control. Finally, this thesis presents a proposal for an on-line expert system for supervision of process control systems. In practice, a large process may require a large number of control loops for which the performance requirements are such that a standard PID controller is sufficient. In this case, it is reasonable to introduce an expert system on a. high supervisory level that will continuously monitor the performance of several controllers in real-time and provide on-line fault diagnosis as well as controller design. Expert system techniques are suitable approaches to solving these problems involving knowledge and reasoning.