Varfolomiyev O. Synthesis of the neuro-controller for the armament aiming and stabilizing system

Object of research: dynamical processes in the aiming and stabilizing systems subject to resilient elements, nonlin-earities of external friction and gaps in kinematic units. Research objective: development of synthesis methods for aiming and stabilizing systems to improve precision based on neural controllers. Methods of testing: fundamentals of automatic control theory, mathematical modeling methods, artificial neural networks and neural control theories. Theoretical and practical results: grounded and formalized aiming and stabilizing systems synthesis area based of neural network structure and predictive control. Developed synthesis methods of neural network structures with pre-diction for aiming and stabilizing systems on basis of two-mass electromechanical systems subject to nonlinear exter-nal friction. Expanded mathematical models of aiming and stabilizing systems control objects with complex kinematic constraints and provided a mathematical model of aiming and stabilizing system as a multi-layered feed forward net-work. Proposed and synthesized neural network aiming and stabilizing systems with P- and PD-regulators in the posi-tion loop, as well as with velocity pre-control, that guarantee the high control quality subject to nonlinearities of ex-ternal friction in the control object and in the operation unit under the different input actions and thirty-percent fluc-tuation of control object parameters, and in the presence of external disturbances and measurement noises. Novelty of the work is in the theoretical development and generalization of the aiming and stabilizing system synthesis on basis of neural networks with prediction, that provides high precision and meets system requirements. Implementation ex-tent: scientific production enterprise "Khartron-ARCOS", UIPA. Sphere of appliance: production of state-of-the-art new and modernization of the existing high-precision aiming and stabilizing systems.