The methodology of multi objective design of neural network two-degree-of-freedom nonlinear electromechanical stabilization systems based on metaheuristic optimization algorithm

Problem. Most existing nonlinear electromechanical stabilization systems (NESS) of equipment spatial position of moving object used typical proportional-differential controllers, which in most cases fail to meet the modern ever-increasing stringent requirements that are presented to the dynamic characteristics of such systems in various modes of their operation. Goal. To develop the methodology of multi objective design of NESS for wide class moving object based on deep learning artificial neural network (ANN) of two-degree-of-freedom control (TDFC), in with both closed-loop control with error feedback and open-loop feedforward control with reference and disturbance used simultaneously to improve system accuracy. Methodology. The problem of multi objective design of NESS based on choosing of weight matrices with the help of which the quality criterion of deep learning ANN is calculated. The calculation of these weight matrices performed based on the zero-sum vector game solution. The vector game solution calculated using a metaheuristic optimization algorithm based on the behavior of blood-sucking leeches. Results. The results of multi objective design of NESS in vertical and horizontal planes based on deep learning ANN are given. Based on both modeling and experimental studies results established, that with the help of designed controllers, it is possible to reduce the variance of the stabilization error under random disturbances by more than 5 times in comparison with the system with typical regulators. Scientific novelty. For the first time the methodology of multi objective design of NESS based on deep learning ANN is developed. Practical value. The possibility of solving the problem of multi objective design of NESS based on deep learning ANN is shown. References 48, figures 8.