Stability analysis of nonlinear systems and upset recovery strategies for agile maneuvering aircraft

This thesis investigates the upset recovery problem of aircrafts as well as stability analysis of general nonlinear systems. In order to deal with upset recovery problem, the strategies which contains nonlinear control techniques are adopted. In order to validate through comparison, linear control technique which make use of linear quadratic regulator is also implemented. As a nonlinear control techniques nonlinear dynamic inversion (NDI) based angular rate regulator and double loop NDI controller for angle of attack, sideslip and bank angle are derived for High fidelity F-16 aircraft model. The choice of aircraft model enabled us to observe conditions with high angular rates which can be encountered in agile maneuver executions. Implemented controllers are evaluated through a novel framework where the recovery envelopes and recovery times are used as the performance criteria. Three dimensional recovery envelopes are constructed using three body angular rates which are used as initial points for simulations and included or excluded to recovery envelopes based on the success in recovery attempt. Successful recovery conditions are restricted so that the aircraft would not count as recovered if it exceed structural or altitude limits along with the constraints employed in actuator models. Based on the evaluations which are made by the proposed framework, another strategy is proposed which adopts the strong sides of single loop and double loop NDI controllers such that it utilizes the single loop NDI controller until the high angular rates are regulated below some predetermined threshold, then uses the double loop NDI controller to recover aerodynamic angles. This new strategy is called as switching NDI controller and its performance superiority is validated using the proposed framework. Thereafter conducting analysis based on simulation study, another approach is implemented to mitigate the computational burden which comes along with the simulations. Neural network model is constructed to reproduce the recoverability envelopes obtained via simulations. By this way, recoverability envelopes are created with high fidelity using only small portion of the whole simulations executed before.