An adaptive and reconfigurable flight control system is developed for a general aviation aircraft. The flight control system consisting of two neural networks is developed using a two phase procedure called the pre-training phase and the online training phase. The adaptive critic method used in this thesis was developed by Ferrari and Stengel. In the pre-training phase the architecture and initial weights of the neural network are determined based on linear control. A set of local gains for the linearized model of the plant is obtained at different design points on the velocity v/s altitude envelope using an LQR method. The pre-training phase guarantees that the neural network controller meets the performance specifications of the linear controllers at the design points. Online training uses a dual heuristic adaptive critic architecture that trains the two networks to meet performance specifications in the presence of nonlinearities and control failures. The control system developed is implemented for a three-degree-of-freedom longitudinal aircraft simulation. As observed from the results the adaptive control system meets performance requirements, specified in terms of the damping ratio of the phugoid and short period modes, in the presence of nonlinearities. The neural network controller also compensates for partial elevator and thrust failures. It is also observed that the neural network controller meets the performance specification for large variations in the parameters of the assumed and actual models.