The application of artificial neural networks for aircraft motion control, in particular, for creation of nonlinear algorithms of the aircraft remote control system (RCS) is considered. Aircraft as a control object is represented as a multidimensional nonlinear dynamic system and nonlinear control methods are used to operate this system. The control loop is constructed using the method of inverse dynamics based on the feedback linearization principle. The nonlinear control law is represented as a neural network being learned (adjusted) by recorded or incoming measurements of motion parameters. Synthesis and testing of neural network control algorithms is performed with the fully nonlinear mathematical model of a maneuverable aircraft for three control channels. Simulation results of the closed system are presented.In this paper, we solve the problem of obtaining desired aircraft controllability characteristics by means of automation. The problem being solved by algorithmic methods consists of the high-precision control throughout the flight envelope, including the predominance of nonlinear object characteristics. A classic way of eliminating nonlinearity is an increase of gain factors in the control loop, which linearizes the control object to some extent. However, the possibilities of this approach are limited by stability of the closed system.Another way is to use nonlinear control methods. Linearization of output feedback (the basis of the inverse dynamics method) has become a generally accepted technique for constructing the control loop for nonlinear dynamic systems. Furthermore, the aim is to use the measurements of flight parameters, that is to "automate" acquisition of the nonlinear control law.The artificial neural networks are used to solve this problem that is due to their two properties, namely, first, neural networks have a property of universal approximation and, therefore, they can approximately implement a wide class of nonlinear functions; second, they are nonlinear statistical models and they are adjusted (trained) by incoming or recorded measurement signals. Thus, from the practical point of view, neural networks are of interest as universal models of nonlinear regression. In this paper, the controlling neural network implements the nonlinear control law being used in the main loop.Synthesis and testing of neural network control algorithms is performed on the fully nonlinear mathematical model of a maneuverable aircraft through three control channels for the current flight conditions. Control actuators are presented by oscillating elements with a frequency of 20 rad/s and restrictions on speed and position. Due to specific character of the configuration methods of the neural network control law, an internal loop of angular velocity control is distinguished in the structure of the system. This loop is built by the inverse dynamics method, and an external angle of attack control loop includes the PI controller.

No abstract

No abstract