AUTOMATIC TUNING OF PROCESS CONTROLLERS : PRACTICAL APPROACHES

Abstract

This thesis focuses on practical tuning approaches for PID controllers - the most widely used controller in the process industry. Through analysis and approximation, analytical tuning formulas based on gain margin and phase margin are derived. These analytical formulas are particularly useful in the context of adaptive control and auto-tuning where the controller parameters have to be computed on-line. Unlike other analytical formulas, the desired system performance as well as robustness can be set by users through gain and phase margin specifications. A simple relay auto-tuning procedure is presented to model the over-damped processes approximately by a reduced order process model in terms of a first or second order transfer function plus dead time. With the auto-tuning method, the formulas give reasonably accurate gain margin and phase margin with respect to the specifications for a wide range of process dynamics. The same method is extended to processes with integration. A comparison study with the Internal Model Control formula has been made. The two formulas have the same advantage that a user can tune the PID controllers through specifying only one or two parameters linked to performance and robustness. The study also reveals the limitation that the Internal Model Control design for the PID controller could only achieve certain gain and phase margin combination because of pole-zero cancellation. The simple relay auto-tuning technique for PID controllers is also extended to the Smith predictor. Parameters of the first order or second order plus dead time model are used directly in the predictor part as well as in computing the optimal settings of the primary controller of the Smith predictor. This auto-tuning technique is effective even for high order or non-minimum phase processes that exhibit apparent dead time-like characteristics in their dynamics. With an appropriate use of the FFT algorithm, a simple adaptive technique has been developed to adapt the Smith predictor to process dynamics changes. The signal from a load disturbance responses can also be used for adaptation. The variable set point weighting technique is also introduced in this thesis. This new technique can reduce overshoot of set point response without sacrificing the rise time. Thus the controller settings can be independently tuned for optimal load disturbance response. A self-tuning formula of set point weighting giving a systematic method of determining the switching instants and the set point weighting is derived. The good performance of the tuning formula has been substantiated by extensive simulation.