УПРАВЛЕНИЕ ТЕМПЕРАТУРНО-ВЯЗКОСТНЫМ КЛИНОМ В ПОДШИПНИКЕ ЖИДКОСТНОГО ТРЕНИЯ НА ОСНОВЕ АЛГОРИТМА ГЛУБОКОГО ОБУЧЕНИЯ С ПОДКРЕПЛЕНИЕМ

Abstract

. This paper discusses the possibility of using intelligent systems based on reinforcement learning and deep neural networks to control the lubrication system of a mechatronic fluid friction bearing. The training of the agent of the control system was carried out on a simulation model of a rotary machine, the purpose of training was to reduce energy losses for friction and reduce the level of vibrations under conditions of a random external force action of a periodic nature with a random frequency. The ability to switch modes has reduced the negative effect of the resonance phenomenon. The results of the training demonstrated the effectiveness of using such intelligent systems for the developed mechatronic bearing arrangement.