基于RDRNN的变阻尼半主动结构控制遗传算法

Semi - active structural control GA by variable dampers based on RDRNN

提出了基于ＲＤＲＮＮ的变阻尼半主动结构控制遗传算法，应用多输入多输出动态递归神经网络模型ＲＤＲＮＮ预测结构的响应，并利用遗传控制算法进行变阻尼控制力寻优，实现了结构振动的变阻尼半主动最优控制。ＲＤＲＮＮ模型针对结构控制中结构状态变量、控制变量和外界荷载对结构的响应有不同的影响，采用分支输入递归处理，不但结构响应预测精度好，而且神经网络的训练效率也高；基于ＲＤＲＮＮ给出的预测结果，遗传控制算法直接在控制力的解空间进行全局寻优，逐步搜索到满足变阻尼控制力约束条件并使性能指标趋于最小的控制力。算例仿真表明，应用所提算法进行变阻尼半主动结构控制，控制效果明显，是一种很有发展前景的控制策略。

The semi - active structural control method based on RDRNN is proposed in the paper. The semi - active structural control by variable dampers is performed, in which the structural response is predicted by RDRNN and the control force is calculated by GA. The ramose - input and multi -output Dynamic Recurrent Neural Network RDRNN is efficient on training and is accurate on predicting the structural response. The inputs that have different affections on the structural response are treated respectively through ramose input. The genetic algorithm GA uses the results from RDRNN to search the optimal control force to minimize the performance function while satisfying the specified constraints. A simulation result demonstrates that the proposed method is a promising structural control strategy for semi - active structural control by variable dampers.