Date of Award

2002

Document Type

Master Thesis

Degree Name

Master of Engineering (Research)

Department

Electronic Engineering

First Advisor

Dr. Conor Downing

Abstract

One of the more popular control strategies, of late, is the Generalised Predictive Controller (GPC). It has received widespread dissemination in academia and numerous successful applications in industry testify to its usefulness. Its popularity partly stems from its superiority to its predecessors, e.g. the generalised minimum variance and pole-placement controllers, but also from the GPC ‘tuning knobs’, which have simple time-domain interpretations and extremely robust default settings. However, despite the widespread interest, there would appear to be some issues relating to the specification of the closed-loop serv^o response that have not been covered in any great detail in the GPC literature. This thesis attempts to redress this deficiency.

Existing tuning strategies are presented and analysed and modifications proposed where appropriate. However the issue of programmed control (relating to setpoint sequences that are known a priori) as well as the use of a setpoint prefilter to generate the future reference trajectory is not covered in any detail m the GPC literature. These two techniques are implemented and critically examined in this thesis. It is shown that an improved response can be achieved when controlling non-minimum phase (NMP) inverse response processes using programmed control. An analytical expression for the servo transfer function when using a setpoint prefilter is derived and its superior model following abilities are demonstrated.

In order to achieve high performance tracking of the desired trajectory an accurate model of the process is required. Modelling procedures, based on the closed-loop step response, are proposed with the intention of reducing test times for plants with vary large time constants and for the identification of second order plants with little or no overshoot.

Plant model identification in closed-loop is necessary in some cases, can prove to be advantageous in other cases and is necessary in indirect adaptive control. The recently developed closed-loop output error recursive parameter estimation algorithm is critically analysed and some unexpected behaviour is reported on.

For plants with time varying and non-linear behaviour a fixed robust controller may be unable to produce the desired servo performance. In such cases an adaptive solution is required. In this thesis three such adaptive techniques are compared: adaptive control, adaptive tracking with (fixed) robust regulation and multiple model control. A novel feature in the multiple model case is the use of a GPC controller kernel.

Access Level

info:eu-repo/semantics/openAccess

Share

COinS