模糊神经网络在高层建筑横风向振动控制中的应用研究

提出了模糊神经网络方法控制高层建筑横风向风振反应。通过观测部分楼层加速度和控制力输出，建立了模糊神经网络控制器，解决了传统控制中有限的传感器数目对系统振动状态估计的困难．利用模糊神经网络控制器预测结构的控制行为，消除了闭环控制系统中存在的时滞。利用模糊神经网络控制器的自学习能力来确定模糊规则和语言变量隶属函数，解决了土木工程复杂结构模糊控制中，难于依据专家的主观经验来确定模糊控制规则和语言变量隶属函数等困难。模糊神经网络方法的优势在于算法自身的鲁棒性，处理结构非线性、参数不确定性及时变等问题的能力。通过对基准建筑的刚度不确定性分析，讨论了模糊神经网络控制器的鲁棒性。仿真分析表明，模糊神经网络控制策略能有效地抑制高层建筑的横风向风振反应，控制效果略优于LQG控制，而拥有LQG控制不具备的诸多优点。

Study on Application of Fuzzy Nenral Network to Active Control of Across wind Response of tall Buildings

A robust fuzzy neural network methodology is proposed for vibration mitigation of tall building under across wind excitation.Based on measurement of floor accelerations and control forces,a fuzzy neural network controller(FNNC) is designed,in which few sensors and no observer are needed.The FNNC by predicting control action eliminates effects of time delay in the control loop.Utilizing the characteristics of adaptation and learning of the FNNC,it's easy to obtain an appropriate set of rules and membership functions,which are difficult to obtain by experts'experience in a fuzzy logic control strategy.The main advantage of the proposed controller is its inherent robustness and ability to handle(nonlinearity),uncertainty and time-varying properties of structures.The robustness of the controller is demonstrated through the uncertainty in stiffness(+15% and-15% varying from initial stiffness) of a benchmark building.The results of simulation show a good performance in reducing across wind-induced response by the proposed controller for all cases tested.Also the results show that the proposed controller performs slightly better than the LQG controller,while possessing several advantages over the LQG controller.