基于RDRNN的变阻尼半主动结构控制振动试验研究

在一个1：4的五层模型刚架结构上进行了变阻尼半主动结构控制的振动台试验，应用多输入多输出分支动态递归神经网络模型RDRNN建立了神经网络控制器，利用已有的试验数据对神经网络控制器进行训练，然后应用该神经网络控制器对结构进行变阻尼半主动结构振动控制，输入了几种不同的地震波，对比分析了该神经网络控制器的控制效果。振动参试验结果表明，应用神经网络控制器对结构进行变阻尼半主动结构控制，可以达到较好的控制效果,所需控制信息较少，并且对不同的地震激励有较强的荷载适应性。

Shaking Table Test Investigation of a Building Structure with a Semi-Active Fluid Damper Based on RDRNN

This paper describes shaking table tests of a multi storey scale model building structure subjected to seismic excitation and controlled by a semi active fluid damper control system. A neural network controller is proposed based on ramose input and multi output dynamic recurrent neural network(RDRNN). The neural network controller was trained with measured responses of the structure. The semi actively controlled shaking table test was conducted using the trained controller. Several earthquake records were used to validate the performance of the control system. The experimental results demonstrate that the proposed neural controller is effective for semi active structural control. It is robust to different seismic excitation and needs less control information.