Evaluation and extension of the potential for application of the behavioural framework to practical engineering problems

  • Raphael Pfaff

    Student thesis: Doctoral ThesisDoctor of Philosophy

    Abstract

    Motivated by the theoretical advantages of the Behavioural Framework for modelling and control, which is based on model formulations without preconception of the input/output structure of the system, this thesis sets out to investigate the applicability of this framework to practical control engineering problems. While the Behavioural Framework is consistent and all-encompassing, certain areas are identified where more practically oriented approaches are likely to improve the applicability and thus increase the popularity of the Behavioural Framework. In these areas, novel techniques are developed and tested, extending the Behavioural Framework with immediately applicable techniques.

    The development, documentation and exchange of models based on physical principles forms the basis for many practical problems in modelling and control. This topic is addressed with a view on model validity and graphical representations, both for documentation and exchange as well as as input for simulation software.

    The validity of a model can be increased further by taking into account system nonlinearities. Commonly used nonlinear model classes are analysed with application in the Behavioural Framework in mind and the best suited class, the class of bilinear systems, is represented, analysed and tested on practical problems.

    The identification of systems from recorded data, termed approximate modelling, is important in modelling and control. Here a novel practically viable approach, the combined mist/complexity approach, is presented. This approach weighs the model mist versus the model complexity and in this way resembles a heuristic approach.

    A novel technique for adaptive control in the Behavioural Framework is developed based on the properties of practical control systems, among these I are the mixed time axis and the limited availability of controller representations. The scheme, based on a recursive Errors-in-Variables estimator and an interconnected controller, is applied to linear time varying as well as bilinear plants.

    The application of existing and novel techniques to practical control engineering problems forms a leitmotif of this thesis.
    Date of Award2013
    Original languageEnglish
    Awarding Institution
    • Coventry University
    SupervisorKeith Burnham (Supervisor)

    Cite this

    '