基于神经网络算法的建筑结构振动分散控制研究

针对地震作用下建筑结构振动分散控制问题,引入神经网络算法,研究结构振动分散神经网络控制策略,来解决分散控制中各子系统的耦合问题和神经网络算法的训练成本问题.利用径向基函数RBF(Radical Basis Function)神经网络模型并基于newrb函数构建了RBF神经网络控制器,对某20层Benchmark结构模型分别进行集中控制和多工况子系统划分分散控制的数值模拟分析,结果表明,提出的各子系统耦合的分散RBF神经网络振动控制策略考虑了子系统间的信息共享,可有效控制结构的振动响应,且子系统达到理想训练结果所需的训练次数与BP网络相比显著降低.

Decentralized control of building structure based on neural network algorithm

Aiming at decentralized vibration control of structures under an earthquake, a neural network algorithm is introduced to study the decentralized neural network control strategy of structural vibration, so as to solve the coupling problem of individual subsystems in the decentralized control and reduce the training cost of the neural network algorithm.Employing the Radial Basis Function (RBF) neural network model, an RBF neural network controller is formed on the basis of the newrb function.And a 20-layer Benchmark structure model is respectively tested by centralized control and multi-condition subsystems-division decentralized control, the data of which is later processed by numerical simulation analysis.The simulation analysis shows that the decentralized RBF neural network vibration control strategy for the coupling of individual subsystem herein takes into account the information sharing between the subsystems, which can effectively control the vibration response of the structure and rationalize the training frequency required for the subsystems to achieve the ideal training result.Compared with that in BP network, the required frequency is significantly reduced.