Design and synthesis of time maximum disturbance for stable control systems

Validation of control system performance is a common task in certification of aircraft and other vehicles. The closed loop control specifications often require a tolerance to various levels of white noise disturbance without disruption of the normal operation of the aircraft. Recent specifications require a tolerance to bounded wind gusts without loss of stable performance. Testing of control systems in the worst case bounded disturbance situation is the objective of the thesis research summarized here.; It is shown in this thesis that the most stressful bounded disturbance for a linear stable second order system with numerator dynamics is of bang-bang type and should be realized in feedback form. To test the control system robustness and dynamics, a bounded disturbance is selected to maximize a severity index, usually time distance. A switch curve can be used to realize the disturbance in feedback form. The switch curve has a general analytic form for second order systems. This closed analytic form requires less computational effort and gives an intuitive feeling for how the switch curve will change with the other parameters.; The bang-bang disturbance can also be implemented with a switch curve that is constructed by using isochrones as the level sets for the disturbance index in a state space setting. Some new results on how to maximize the severity index using both bang-bang disturbances and impulse actions with a recharge constraint on the impulse firings are described below. Bounded impulse disturbance action is added to usual bang-bang disturbance action to further test stabilized control system robustness.; However there are some obvious weaknesses; sensitivity to model uncertainties and the time consuming work to construct the switch curves (disturbance switch curve and impulse firing curve). We can suggest an efficient way to approximate the switch curves. It overcomes the shortcomings of the original switch curves which are constructed through time backwards computation and implements sufficiently to test the control stable system. The approximation also relieves some of the burden of computational construction of the switch curves.