RESEARCH OF DECENTRALIZED VIBRATION CONTROL OF BUILDING STRUCTURES BASED ON A NEURAL NETWORK ALGORITHM

The article is devoted to the urgent problem of reducing the vibration response of building structures under the influence of dynamic loads, such as seismic impacts and strong wind. The aim of the study is to develop and analyze the effectiveness of a decentralized active vibration control system using a neural network algorithm to coordinate the operation of local actuators. The tasks include: analyzing the advantages of a decentralized approach compared to a centralized one, developing the architecture of a neural network controller for local control nodes, creating a mathematical model of a multi-storey building and simulating the operation of the proposed system under various scenarios of dynamic loading. The results obtained demonstrate that the proposed decentralized neural network algorithm is able to effectively reduce the oscillation amplitudes of the structure (up to 40–60% at peak accelerations, depending on the type of impact) and interstory distortions, while having increased fault tolerance and scalability. The scientific novelty lies in the development of an adaptive decentralized algorithm based on neural networks, which allows local controllers to make decisions based on limited data exchange, optimizing control actions for global vibration reduction.