TABLE OF CONTENTS One of the earliest stages in the design of a flight control system (FCS) for a civil aircraft is based on the use of linear aircraft models to produce an initial controller structure. This is because linear models, usually based on small perturbations of the full nonlinear aircraft model at a point in the aircraft flight envelope, are easier and quicker to analyse mathematically. Both traditional single-input single-output (SISO) methods and most modern methods depend on this linear-synthesis stage to produce the basis for a controller. For this reason, much effort has been put into developing linearcontrol system design techniques which serve to produce a better controller for these linear aircraft models.
It is often found, however, that the resulting controller does not perform well when tested in a more realistic aircraft environment, where considerations such as a much wider flight envelope, varying aircraft configuration, system nonlinearities and model uncertainties come into being. The traditional industrial approach to solving this problem is to adjust the controller iteratively until the system is satisfactory. This process results in an enormous amount of time and money being spent on tuning the controller using nonanalytical procedures. Because of the time involved in adjusting the controller in this way, linear synthesis may constitute very little of the total design effort. In the finished software, this is further diluted by the fact that the original linear-model-based control algorithms may constitute as little as five to ten per cent...
