Current jet fighters and modern airliners are hugely complex pieces of machinery. The drawback of this complexity lies in the number of systems and subsystems that may fail for one reason or another. Given the systems complexity of aircraft it is no longer easily possible for the crew to establish what exactly has happened when these fail. This motivates the need to provide means for the diagnosis of failures and automated recovery, i.e. fault tolerant flight control. In general the procedure to make a system fault-tolerant consists of two steps: 1) Fault diagnosis: the existence of faults has to be detected and the faults have to be identified, 2) Control re-design: the controller has to be adapted to the faulty situation so that the overall system continues to satisfy its goal. This thesis focuses on the latter through the application of modern control methods towards reconfigurable flight control. This thesis investigates fault tolerant flight control in the event of actuator or plant faults. A literature survey suggests that model predictive control (MPC) is very suitable for use as fault tolerant flight control method due to its ability to incorporate various constraints. Application of MPC in this setting is the central topic in this thesis. MPC is applied in two different ways in the text. First, a method is presented for finding both a state-observer and the cost function associated with a model predictive controller, based on an already existing output feedback controller. The goal of this exercise is to retain the properties of the existing controller, while adding the constraint handling capabilities of MPC. The second way features the combination of model-based predictive control and the inversion of the dynamics of the system under control into a constrained and globally valid control method for fault-tolerant flight-control purposes. The fact that the approach allows the incorporation of constraints creates the possibility to incorporate additional constraints in case of a failure. Such failures range from relatively straightforward actuator failures to more complicated structural breakdowns where, through the addition of constraints, the aircraft can be kept within its remaining flight envelope. Furthermore, the method is model-based, which allows adaptation of the system model in case of a failure. Both of these properties lead to the fault-tolerant qualities of the method presented. Projection of a polytope onto a lower dimensional polytope is an important element in the combination of MPC and dynamic inversion. A method is presented that avoids the computation of the polytope’s vertices and the application of linear programming methods. The theory presented in this thesis is applied to a benchmark model which constitutes a detailed simulation model of a Boeing 747-200 aircraft, like the aircraft that crashed in the Amsterdam Bijlmer area in 1992.
|Award date||10 May 2017|
|Publication status||Published - 2017|