The Linear Fractional Transformation (LFT) is a general, flexible and powerful framework to represent uncertain systems. Linear Fractional Representations (LFRs) are the basis for the application of many modern robust control techniques (e.g., robust H-infinity control design, mu-synthesis/analysis). For several classes of uncertain systems, it is in principle straightforward to generate equivalent LFRs. However, the resulting LFRs are generally not unique, a theory for the generation of LFRs with minimal complexity does not exist and the pure application of existing ad-hoc realization methods generally yields LFRs of high complexity. LFR-based modern robust control methods are numerically highly demanding and of high computational complexity, e.g., many methods require to solve a large system of Linear Matrix Inequalities (LMIs). Therefore the application of these methods is restricted to LFRs of reasonable complexity, otherwise the computation time will be unacceptable or the methods may even fail. To realize LFRs of low complexity, a three step procedure is employed in this thesis consisting of (i) symbolic preprocessing of uncertain system models using improved and newly developed decomposition techniques, (ii) object-oriented LFR realization based on a newly developed generalized/descriptor LFT, (iii) numerical multidimensional order reduction based on newly implemented numerical reliable and efficient routines. All the techniques are implemented in version 2 of the LFR-toolbox for Matlab and allow to realize an LFR of almost minimal complexity for one of the most complex parametric aircraft models available in the literature. Using this LFR, a reliable LFR-based robust stability analysis covering the whole flight envelope has been performed, which was not possible with earlier generated LFRs of high complexity. An LFR of minimal complexity is generated for a parametric vehicle model. Based on this LFR, a mu-synthesis controller and a gain-scheduled Linear Parameter Varying (LPV) controller are synthesized. Both controllers show better performance and are robust with respect to a considerably larger parametric uncertainty domain than recently developed controllers using the Parameter Space (PS) method.
«
The Linear Fractional Transformation (LFT) is a general, flexible and powerful framework to represent uncertain systems. Linear Fractional Representations (LFRs) are the basis for the application of many modern robust control techniques (e.g., robust H-infinity control design, mu-synthesis/analysis). For several classes of uncertain systems, it is in principle straightforward to generate equivalent LFRs. However, the resulting LFRs are generally not unique, a theory for the generation of LFRs wi...
»