Effective optimal control of a fighter aircraft engine
Typical modem fighter aircraft use two-spool, low by-pass ratio, turbojet engines to provide the thrust needed to carry out the combat manoeuvres required by present-day air warfare tactics. The dynamic characteristics of such aircraft engines are complex and non-linear. The need for fast, accurate control of the engine throughout the flight envelope is of paramount importance and this research was concerned with the study of such problems and subsequent design of an optimal linear control which would improve the engine's dynamic response and provide the required correspondence between the output from the engine and the values commanded by a pilot. A detailed mathematical model was derived which, in accuracy and complexity of representation, was a large improvement upon existing analytical models, which assume linear operation over a very small region of the state space, and which was simpler than the large non-analytic representations, which are based on matching operational data. The non-linear model used in this work was based upon information obtained from DYNGEN, a computer program which is used to calculate the steady-state and transient responses of turbojet and turbofan engines. It is a model of fifth order which, it is shown, correctly models the qualitative behaviour of a representative jet engine. A number of operating points were selected to define the boundaries used for the flight envelope. For each point a performance investigation was carried out and a related linear model was established. By posing the problem of engine control as a linear quadratic problem, in which the constraint was the state equation of the linear model, control laws appropriate for each operating point were obtained. A single control was effective with the linear model at every point. The same control laws were then applied to the non-linear mathematical model adjusted for each operating point, and the resulting responses were carefully studied to determine if one single control law could be used with all operating points. Such a law was established. This led, naturally, to the determination of an optimal linear tracking control law, and a further investigation to determine whether there existed an optimal non-linear control law for the non-linear model. In the work presented in this dissertation these points are fully discussed and the reasons for choosing to find an optimal linear control law for the non-linear model by solving the related two-point, boundary value problem using the method of quasilinearisation are presented. A comparison of the effectiveness of the respective optimal control laws, based upon digital simulation, is made before suggestions and recommendations for further work are presented.