SYNTHESIS OF OPTIMAL CLOSED-LOOP CONTROL OF THE SYSTEM „CRANE-LOAD” IN FORM OF FEEDBACK

Abstract

The article develops an approach to the synthesis of optimal motion control of the dynamic system „crane-load” for the problem of eliminating load oscillations. The approach is based on an artificial neural network. Its training was conducted by using the metaheuristic method ME-D-PSO via the reinforcement learning paradigm. All calculations are given for the mode of acceleration of the crane with a load on a flexible suspension. The optimization criterion was a complex indicator that takes into account the duration of the system motion and the RMS value of the dynamic component of drive power. In addition, there are kinematic and dynamic constraints in the problem statement, which are caused by the limited features of the frequency-controlled drive of the crane movement mechanism. The essence of the approach developed in the article is connected with finding the optimization criterion minimum in the space of weights and biases of an artificial neural network, which, in addition, satisfy the boundary conditions of the system and the imposed constraints. The weight tensor and the bias matrix of the neural network have been obtained during the calculations. They satisfy all of the conditions of the problem. The results are illustrated by graphical dependences of kinematic, energetic and dynamic characteristics of the dynamic system motion. In addition, the calculation of estimated indicators are given: maximum and RMS values of power, driving force, and load oscillations. The final part of the article presents the prospects for further research in the area