Optimal Adaptive and Predictive Control System of Buildings by Neuro-Fuzzy Systems

为了开发有效的建筑物控制系统，必须考虑到未来荷载的复杂性、不确定性等特点。本研究基于模糊理论、神经网络和遗传算法，建立了建筑物抗震的最优自适应和预测控制系统。这些系统具有以下特点：(1)用隶属度函数描述目标响应和控制变量；(2)进行地震输入的实时预测和结构识别；(3)采用模糊最大化决策进行优化。在地震输入预测方面，采用了条件模糊集规则和神经网络相结合的方法。结构识别采用分段线性响应方程和神经网络相结合的方法。提出了用遗传算法对结构识别神经网络训练数据进行系统选择的方法。目标结构假定为1、2和5个自由度系统，其顶部具有主动质量驱动器。采用等效变质量系统作为主动控制方法。建立了这些结构的总体主动控制系统，并进行了数字仿真。结果表明，所提出的控制系统可以根据假定的隶属度函数对结构响应进行控制。结果表明，所提出的主动控制方法对建筑物的控制是有效的。通过这项研究，我们可以得到许多关于建筑模糊最优控制系统的基础数据和系统。

To develop the effective control system of buildings, it should be necessary to take account of special features such as complexity, uncertainty of future loading. In this research, optimal adaptive and predictive control systems of buildings against earthquake loading are developed based on fuzzy theory, neural network and genetic algorithm. These system have following special features ; (1) Target responses and control variables are described with membership function, (2) Real time prediction of earthquake input and structural identification are performed, and (3) Optimization is performed by fuzzy maximizing decision. As for the prediction of earthquake input, conditioned fuzzy set rules and neural network are employed. As for the structural identification, piece-wise linear response equations and neural network are employed. The systematic selection method of the training data of neural network for the structural identification are also developed by genetic algorithm. Objective structures are assumed 1,2 and 5-degree-of-freedom systems with active mass driver at the top of it. An equivalent variable mass system is employed as an active control method. Total active control systems for these structures are developed and digital simulations are carried out. The results show that proposed control systems can control the structural responses in accordance with the assumed membership functions. Consequently, it is proved that the proposed active control methods are effective for the control of buildings. Thorough this research, we can get al lot of basic data and systems on the fuzzy optimal control system of buildings.