One of the remaining unsolved problems in control design is the development of simple and practical controllers for real-time control, based on a sound theoretical basis, for systems with severe nonlinearities and constraints. Although all dynamic systems are nonlinear, classical approaches to analysis and design are almost universally based on linear time-invariant approximations to the dynamic characteristics. However, classical approaches are no longer adequate because of the increasing performance needs of modern industry, which require plants to be operated in regions where there are constraints and strong non-linear behaviour. Combined with the fact that most existing non-linear control techniques used by industry are empirically based and, as a result, difficult to tune and analyse, there is a real need for a scientifically more rigorous framework for practical non-linear multivariable control. This is a cooperative project in which the research work is divided between the Industrial Control Centre, University of Strathclyde and the Department of Aerospace Engineering, University of Glasgow. It builds upon the new nonlinear generalized minimum variance (NGMV) control design ideas developed at University of Strathclyde. The project will attack the problem of synthesising nonlinear controller design using two complementary philosophies: the traditional approach of a purely theoretical foundation tested through realistic case studies and the alternative, practical view of tailoring advanced control research to address specific engineering problems. Motivation for this second approach comes from the fact that bespoke nonlinear controllers already exist in many engineering systems and this practical experience should be harnessed in the search for complete theory. The project involves five very significant and difficult scientific challenges that should make the proposed method suitable for the most challenging real-time applications. These theoretical challenges include:
1. Development of a Nonlinear Predictive Control facility which is much faster than existing solutions for machinery controls and faster processes.
2. Include new Constraint Handling features in both the predictive and non-predictive versions.
3. Introduce measures to guarantee some minimum Robustness Margins.
4. Develop real time Labview based implementation with pre-specified low order Restricted Structure implementation.
5. Introduce a Learning and adaptation feature for applications where plant models change slowly.
The last feature may seem particularly ambitious but the NGMV family of control structures has a useful property that its structure is very suitable for learning or adaptive system. That is, any nonlinear plant subsystems affect the controller structure and solution in a very simple way.