强震作用下高层结构地震响应时程的多步预测

提出了一种结合经验模态分解(EMD)与极限学习机(ELM)的时间序列多步预测方法，对近断层强震作用下弹塑性高层框架结构的顶层加速度和位移响应时程进行了多步预测。首先，利用经验模态分解技术将高层结构非线性、非平稳地震响应分解为一系列具有不同特征尺度的固有模态函数序列(IMFs)。然后，利用极限学习机分别对固有模态函数子序列进行多步预测，再将各子序列的预测值叠加得到最终的预测值。预测结果表明，EMD-ELM预测方法能够高精度地实现强震作用下高层建筑动力响应的多步预测。建筑结构地震响应时程的短期预测可为主动、半主动控制系统预先提供准确的动力响应，从而有利于实现工程结构的在线实时减震控制。

Multi-step prediction for time histories of seismic responses of tall building structure subjected to strong earthquake ground motions

This paper proposes a new multi-step prediction method of EMD-ELM (empirical mode decomposition-extreme learning machine), and achieves the multi-step prediction for time histories of top story’s acceleration and displacement responses of elasto-plastic tall frame structure subjected to near-fault strong earthquake ground motions. Firstly, the seismic responses of tall structure with nonlinear and nonstationary property were decomposed into several components of intrinsic mode functions (IMFs) with different characteristic scales by the technique of EMD. Subsequently, the ELM method was utilized to conduct the multi-prediction of the IMF components. Moreover, the predicted values of each IMF component were superimposed to obtain the final predicted result of structural response. It is demonstrated that the EMD-ELM method can realize the multi-step prediction of seismic responses of tall building under strong ground motions with relatively high accuracy. Short-term prediction for time histories of seismic responses of building structures can provide in advance accurate dynamic responses for active and semi-active control system, thus facilitate the real-time online vibration-reduction control of engineering structures.